economics

Money, Crises, and Transition

Money, Crises, and Transition
“This impressive collection of papers is a fitting tribute to one of the greatest
economists of the late twentieth century.” Essays in Honor of Guillermo A. Calvo
—Michael Bordo, Department of Economics, Rutgers University

Essays in Honor of
Carmen M. Reinhart is Professor of Economics at the “Guillermo Calvo’s work has always been rigorous in its theoretical treatment edited by Carmen M. Reinhart, Carlos A. Végh,

Guillermo A. Calvo
University of Maryland. Carlos A. Végh is Professor of the most important macroeconomic phenomena of our time. This fine and Andrés Velasco
of Economics at the University of Maryland. Andrés collection of papers is very much in the spirit of his work, tackling critical
Velasco, on leave as Sumitomo Professor of Interna- problems with new theories and careful empirical analysis. It includes impor-
tional Finance and Development at Harvard’s John tant new research by the leaders in the profession whose papers deal with
F. Kennedy School of Government, is currently serv- financial crises, monetary regimes, policy rules, and determinants of economic
ing as Chile’s Minister of Finance. All three are Re- growth. This is a masterful achievement, one that no serious scholar of inter-
search Associates at the National Bureau of Economic national finance can afford to be without.”
Research. —Andrew K. Rose, Rocca Professor, Haas School of Business, University
of California, Berkeley edited by
Carmen M. Reinhart Guillermo A. Calvo, one of the most influential macro-
Contributors Carlos A. Végh economists of the last thirty years, has made pathbreak-
ing contributions in such areas as time inconsistency,
Cover photograph by Pilar Bilecky. Leonardo Auernheimer Graciela L. Kaminsky Assaf Razin Andrés Velasco lack of credibility, stabilization, transition economies,
Fabrizio Coricelli Michael Kumhof Carmen M. Reinhart debt maturity, capital flows, and financial crises. His
Padma Desai Amartya Lahiri Francisco Rodriguez work on macroeconomic issues relevant for developing
Allan Drazen I. Igal Magendzo Efraim Sadka countries has set the tone for much of the research in
Sebastian Edwards Enrique G. Mendoza Ratna Sahay this area and greatly influenced practitioners’ thinking
Roque B. Fernández Frederic S. Mishkin Rajesh Singh in Latin America, Eastern Europe, Asia, and elsewhere.
Stanley Fischer Igor Masten Evan Tanner In Money, Crises, and Transition, leading specialists in
Ricardo Hausmann Pritha Mitra Rodrigo Wagner Calvo’s main areas of expertise explore the themes
Bostjan Jazbeč Alejandro Neut Carlos A. Végh behind this impressive body of work.
Peter Isard Maurice Obstfeld Andrés Velasco The essays take on the issues that have fascinated
Edmund S. Phelps Calvo most as an academic, a senior advisor at the In-
ternational Monetary Fund, and as the chief economist
at the Inter-American Development Bank: monetary and
exchange rate policy (both in theory and practice); finan-
cial crises; debt, taxation, and reform; and transition and
The MIT Press 978-0-262-18266-9 growth. A final section provides a behind-the-scenes
Massachusetts Institute of Technology
Cambridge, Massachusetts 02142 look at Calvo’s career and intellectual journey and in-
http://mitpress.mit.edu cludes an interview with Calvo himself.
Money, Crises, and Transition
Guillermo A. Calvo. Photograph by Pilar Bilecky.
Money, Crises, and Transition
Essays in Honor of Guillermo A.
Calvo

edited by Carmen M. Reinhart,

Carlos A. Végh, Andrés Velasco

The MIT Press

Cambridge, Massachusetts
London, England
( 2008 Massachusetts Institute of Technology

All rights reserved. No part of this book may be reproduced in any form by any electronic or mechani-
cal means (including photocopying, recording, or information storage and retrieval) without permis-
sion in writing from the publisher.

MIT Press books may be purchased at special quantity discounts for business or sales promotional use.
For information, please e-mail [email protected] or write to Special Sales Department,
The MIT Press, 55 Hayward Street, Cambridge, MA 02142.

This book was set in Palatino on 3B2 by Asco Typesetters, Hong Kong and was printed and bound in
the United States of America.

Library of Congress Cataloging-in-Publication Data

Money, crises, and transition : essays in honor of Guillermo A. Calvo / edited by Carmen M. Reinhart,
Carlos A. Végh, Andrés Velasco.
p. cm.
Includes bibliographical references and index.
ISBN 978-0-262-18266-9 (hbk. : alk. paper)
1. Monetary policy. 2. Foreign exchange rates. 3. Financial crises. 4. International finance. I. Calvo,
Guillermo A. II. Reinhart, Carmen M. III. Végh Gramont, Carlos A., 1958– IV. Velasco, Andrés.
HG230.3.M682 2008
332 0 .042—dc22 2007039868

10 9 8 7 6 5 4 3 2 1
 Contents

List of Contributors ix
Acknowledgments xi
Introduction xiii

I Monetary and Exchange Rate Policy in Theory 1

1 Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate 3

Maurice Obstfeld

2 Optimal Exchange Rate Regimes: Turning Mundell-Fleming’s Dictum on Its

Head 21
Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

3 Monetary Policy Rules, the Fiscal Theory of the Price Level, and (Almost) All
that Jazz: In Quest of Simplicity 41
Leonardo Auernheimer

II Monetary and Exchange Rate Policy in Practice 69

4 Can Inflation Targeting Work in Emerging Market Countries? 71

Frederic S. Mishkin

5 Why Should Emerging Economies Give up National Currencies? A Case for

‘‘Institutions Substitution’’ 95
Enrique G. Mendoza

6 Hard Currency Pegs and Economic Performance 121

Sebastian Edwards and I. Igal Magendzo
vi Contents

III Financial Crises 157

7 Asset Prices and Self-Fulfilling Macroeconomic Pessimism 159

Andrés Velasco and Alejandro Neut

8 The Center and the Periphery: The Globalization of Financial Turmoil 171
Graciela L. Kaminsky and Carmen Reinhart

9 Why Do Some Countries Recover More Readily from Financial Crises? 217
Padma Desai and Pritha Mitra

IV Debt, Taxation, and Reforms 247

10 Government Debt: A Key Role in Financial Intermediation 249

Michael Kumhof and Evan Tanner

11 Capital Income Taxation in the Globalized World 279

Assaf Razin and Efraim Sadka

12 Can Public Discussion Enhance Program ‘‘Ownership’’? 295

Allan Drazen and Peter Isard

V Transition and Growth 325

13 Sources and Obstacles for Growth in Transition Countries: The Role of

Credit 327
Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

14 Growth in Transition Economies: Domestic Policies, External Assistance, and

Institution Building 349
Stanley Fischer and Ratna Sahay

15 Growth Collapses 377

Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner
Contents vii

VI The Man Behind the Mind 429

16 The Columbia Years 431

Edmund S. Phelps

17 The Practitioner 437

Roque B. Fernández

18 In His Own Words: An Interview with Guillermo Calvo 445

Enrique G. Mendoza

Publications of Guillermo A. Calvo 463

Index 471
 Contributors

Leonardo Auernheimer Michael Kumhof

Texas A&M University International Monetary Fund
Fabrizio Coricelli Amartya Lahiri
University of Siena and University of University of British Columbia
Ljubljana I. Igal Magendzo
Padma Desai Central Bank of Chile
Columbia University Igor Masten
Allan Drazen University of Ljubljana
University of Maryland Enrique G. Mendoza
Sebastian Edwards University of Maryland
University of California, Los Angeles Frederic S. Mishkin
Roque B. Fernández Columbia University
Universidad del CEMA Pritha Mitra
Stanley Fischer International Monetary Fund
Bank of Israel Alejandro Neut
Ricardo Hausmann BBVA Economic Research
Harvard University Maurice Obstfeld
Peter Isard University of California, Berkeley
International Monetary Fund Edmund S. Phelps
Bostjan Jazbeč Columbia University
University of Ljubljana and Slovenian Assaf Razin
National Bank Tel Aviv University and Cornell
Graciela L. Kaminsky University
George Washington University
x Contributors

Carmen Reinhart
University of Maryland
Francisco Rodriguez
Wesleyan University
Efraim Sadka
Tel Aviv University
Ratna Sahay
International Monetary Fund
Rajesh Singh
Iowa State University
Evan Tanner
International Monetary Fund
Carlos A. Végh
University of Maryland
Andrés Velasco
Harvard University
Rodrigo Wagner
Harvard University
 Acknowledgments

This festschrift grew out of a conference in honor of Guillermo Calvo, held at the
International Monetary Fund’s headquarters April 15–16, 2004. We are grateful to
the Fund—and, in particular, to the Research Department—for its hospitality and
generosity in sponsoring this event. While putting together the conference pro-
gram, we were fortunate to have the support of the Research Department’s staff,
particularly Patricia Medina, who worked tirelessly for many months on confer-
ence logistics and participants’ travel arrangements.
Numerous conference participants—including many who endured a long jour-
ney from Guillermo’s native Argentina—contributed to the conference’s success
as presenters, lunch and dinner speakers, discussants, and session chairs. In addi-
tion to all of this volume’s contributors, we are thankful to Michael Bordo,
Eduardo Borensztein, Ricardo Caballero, Agustin Carstens, Gerardo Della Pao-
lera, Michael Dooley, Martin Eichenbaum, Andrew Feltenstein, Raquel Fernan-
dez, Ronald Findlay, Robert Flood, Pablo E. Guidotti, Alejandro Izquierdo,
Olivier Jeanne, Moshin Khan, Leonardo Leiderman, Ashoka Mody, Pablo Andres
Neumeyer, Juan Pablo Nicolini, Guilllermo Perry, Eswar Prasad, Raghu Rajan,
Anoop Singh, Federico Sturzenegger, Lars Svensson, Ernesto Talvi, John Taylor,
and Martin Uribe for making this such a special occasion.
Finally, we are grateful to Elizabeth Murry for her early and enthusiastic sup-
port for this project on behalf of The MIT Press and to John Covell, Lianna Kong,
and Sandra Minkkinen for helping us take it to completion.
 Introduction

Most of the chapters in this volume were prepared for a conference in honor of
Guillermo Calvo, organized by the International Monetary Fund’s Research De-
partment and held at Fund headquarters in Washington, DC, on April 15–16,
2004. At the editors’ request, a couple of chapters were specially prepared after
the conference for inclusion in this volume. The Fund was a natural and gracious
host since Guillermo had a distinguished affiliation with the Fund’s Research De-
partment from 1987 to 1994. Under his intellectual leadership, the Research
Department carried out path-breaking research on, among other issues, capital
flows, debt maturity, and inflation stabilization. Guillermo also made important
contributions to the internal discussion and formulation of Fund policies, particu-
larly in Eastern Europe, the former Soviet Union, and Latin America.
The conference brought together renowned academics and policy makers who
have had the privilege of being associated with Guillermo throughout his illustri-
ous career. Some of Guillermo’s former colleagues at Columbia University—
where he began his academic career in the early 1970s—also gave fascinating
behind-the-scenes accounts of Guillermo’s early professional blossoming. One
such account, by Edmund Phelps, recent winner of the Nobel Prize in economic
sciences, is included in this volume.
Delivering the opening remarks, Agustin Carstens (at the time IMF Deputy
Managing Director) struck a theme that would resonate throughout the confer-
ence. In Carstens’s words, ‘‘Guillermo’s ability to reduce complex problems to its
essential elements has taught us that complex models are for lesser minds—for
those who cannot grasp the essential elements out of a given reality. In Guiller-
mo’s hands, the chaos of reality has always yielded simple and illuminating mod-
els.’’ Indeed, in an era in which increasingly complicated paradigms have become
the rule, Guillermo’s simple—but never simplistic—models are a constant
reminder that intellectual elegance and policy relevance can indeed go hand in
hand.
xiv Introduction

A second theme that pervaded the conference was Guillermo’s crucial role in
bringing Latin American policy issues to the forefront of the academic discussion.
In fact, Guillermo is arguably the single most important person in bringing mod-
ern economics to bear on the problems of nations south of the Rio Bravo. He
brought rigor, discipline, and intellectual clarity to the difficult task of illuminat-
ing some of the region’s most pressing economic problems. As a token of grati-
tude, numerous academics and policymakers crossed the Rio Bravo to be present
at the conference. Among them was Roque Fernandez—former Argentine finance
minister—who spoke on Guillermo’s contributions to recent Argentine economic
policy and whose remarks are also part of this volume.
This volume is divided into six parts. The first five parts cover many of the
issues that have been dear to Guillermo’s heart over his thirty-year career: mone-
tary and exchange rate policy, financial crises, debt, taxation, reform, growth, and
transition. The last part sheds light on the mind behind the man. In one capacity
or another, all the chapters’ authors have been closely associated with Guillermo
and have had the privilege of learning and benefiting from Guillermo’s seemingly
infinite wisdom.
Since some of Guillermo’s most influential contributions have been in monetary
and exchange rate policy, it seems only appropriate for the volume to open with
theoretical advances in this area. A perennial and important question in open
economy macroeconomics is: should countries fix or float? In other words, should
an open economy peg its currency against a strong world currency or should it let
the currency float? The standard response, based on the Mundell-Fleming model,
is that the optimal exchange rate regime should depend on the type of shock hit-
ting the economy. If shocks are predominantly of real origin, flexible exchange
rates are optimal as they allow a quicker adjustment of relative prices through
changes in the nominal exchange rate. If shocks are predominantly monetary,
however, then fixed (or predetermined) exchange rates are preferred because
adjustments to the nominal money supply are automatically carried out by the
central bank. As Guillermo has put it, this is a result that ‘‘every well-trained
economist carries on the tip of her tongue.’’ The first two chapters in the book
deal with challenges to this basic tenet.
In chapter 1, Maurice Obstfeld takes issue with a recent finding by Michael
Devereux and Charles Engel, which holds that even if an economy is hit by real
shocks, fixed exchange rates may be preferable if the nominal prices of both
exports and imports are preset in the domestic currency. The key behind the
Devereux-Engel result is that, under such a pricing assumption, flexible rates
cease to be a useful tool to effect adjustments in relative prices. Obstfeld argues
that a minor modification of the Devereux-Engel set up—adding nontradable
goods to their model—restores the optimality of flexible exchange rates under
Introduction xv

real shocks, though for different reasons than those emphasized by the traditional
Mundell-Fleming model. In Obstfeld’s set up, flexible exchange rates become opti-
mal again because the nominal interest rate is freed as a monetary stabilization
instrument.
In chapter 2, Amartya Lahiri, Rajesh Singh, and Carlos Végh approach the same
problem from a very different angle by departing altogether from the traditional
Mundell-Fleming set up. Based on the idea that asset market imperfections are
likely to be as—if not more—important than goods market frictions in developing
countries, they present a model of flexible prices where some economic agents do
not have access to capital markets (in other words, there is asset market segmen-
tation). In such a world, the authors show that the Mundell-Fleming result is
turned on its head: flexible rates become optimal if monetary shocks dominate,
while fixed rates are preferable if real shocks dominate. The key to understanding
these results is that asset market segmentation critically affects the standard ad-
justment mechanism that operates under pegged exchange rates as it does not al-
low some agents to adjust their nominal money balances. In contrast, adjustments
in the nominal exchange rate hit all agents equally. The chapter’s punchline is
thus that the choice of the optimal exchange rate regime should depend not only
on the type of shock (real versus monetary), but also on the type of friction (goods
market versus asset market).
In chapter 3, Leonardo Auernheimer goes back to basics and revisits some of
the themes that Guillermo has focused on in many of his contributions: the rele-
vance of the government budget constraint, the concern with levels versus rates
of change, and the nominal interest rate as a policy instrument. The central ques-
tion addressed in this chapter is the well-known (but controversial) fiscal theory
of the price level. As Auernheimer points out, the literature on this topic is exten-
sive, occasionally obscure, and not lacking in sound and fury (witness Willem
Buiter’s claim that ‘‘the fiscal theory of the price level is fatally flawed’’). As
though guided by Guillermo’s own invisible hand, Auernheimer sets up an ex-
tremely simple model and takes the reader on an illuminating journey into the fis-
cal theory of the price level (and some related issues), stressing some of its truths
and debunking some of its myths.
The second part of the book turns to practical and empirical aspects of mone-
tary and exchange rate policy. In chapter 4, Frederic Mishkin notes that Guillermo
has been rightly skeptical of the adoption of inflation targeting schemes in emerg-
ing markets. Guillermo has convincingly argued in various pieces that, given
weak monetary and fiscal institutions, high degrees of dollarization, and financial
vulnerability, the consequences of giving policy makers too much discretion could
be disastrous. While taking Guillermo’s concerns seriously, Mishkin presents a
strong case for inflation targeting in emerging markets provided that countries
xvi Introduction

can find ways of dealing effectively with large exchange rate fluctuations. He
illustrates his arguments by looking closely at the cases of Chile and Brazil.
In chapter 5, Enrique Mendoza goes a step further and forcibly argues that, all
things considered, emerging market countries would be better off giving up their
national currency altogether. In his view, while giving up the national currency
would certainly not eliminate business cycle fluctuations or financial problems
associated with loose fiscal policies, it would considerably reduce an emerging
market’s vulnerability to abrupt changes in capital flows and enable it to free-
ride on some of the credibility already gained by the country issuing the hard
currency.
But how do countries without a national currency fare in practice? There is vir-
tually no hard evidence. In chapter 6, Sebastian Edwards and Igal Magendzo take
a novel approach to analyzing the inflation and growth performance of countries
with no currency of their own. To this end, they address an important method-
ological problem: joining a common-currency area (or dollarizing) may itself be
an endogenous decision that could be related to economic performance. Their in-
novative answer is to resort to a ‘‘treatment effects model’’ (a technique often used
in the labor literature) that jointly estimates the probability of being a common
currency country and macroeconomic performance equations. Their main finding
is that countries with no currency of their own have had significantly lower infla-
tion and enjoyed higher growth, but suffered from higher macroeconomic volatil-
ity. The jury is thus still out on the net benefits of giving up a national currency.
Ever since the now (in)famous Mexican crisis of December 1994 (aptly referred
to as ‘‘the first financial crisis of the twenty-first century’’), Guillermo’s research
on financial crises has set the agenda for much of the academic work in this area.
The three chapters in part III provide new insights into this critical topic.
In chapter 7, Andrés Velasco and Alejandro Neut show how, in a simple world
with capital market imperfections (that cause borrowing to be collateralized), pes-
simism can lead to self-fulfilling crises. To understand the basic idea, suppose
that, due to a sudden burst of pessimism, stock market prices are expected to fall.
If investment is subject to collateral constraints, the fall in stock market prices
reduces the value of firms and hence investment. But lower investment leads to
smaller returns in the future which, in turn, reduce stock market prices, thus vali-
dating the initial pessimism. Hence, bouts of pessimism can be self-fulfilling
(Keynes’ animal spirits?), a scary thought indeed. Velasco and Neut proceed to
analyze policies that might make this possibility less likely, such as expansionary
monetary and fiscal policy, debt restructuring, and financial and legal reform.
But in practice, once a crisis erupts, how does it spread out? And what deter-
mines the speed with which countries recover? In chapter 8, Graciela Kaminsky
and Carmen Reinhart tackle the first question, analyzing how financial turbulence
Introduction xvii

in emerging countries can spread across borders. They conclude that a financial
crisis will spread globally only if it affects some major financial center; otherwise,
it spreads, at most, regionally. For example, during the Asian crisis, Japanese
banks’ exposure to Thailand—and their subsequent curtailment of lending to
other Asian countries—played a critical role in spreading the crisis.
The second question is tackled by Padma Desai and Pritha Mitra in chapter 9.
They find that export performance before a crisis appears to be a critical variable
in predicting the speed of recovery. Through investor expectations, export perfor-
mance explains most of the difference in postcrisis interest rate and exchange rate
movements. In contrast, the fiscal position and national savings do not seem to
matter as much. This casts some serious doubts on the desirability of contraction-
ary fiscal policies, typically advocated by the International Monetary Fund.
Some of Guillermo’s most influential contributions have been in the area of
debt, taxation, and reform. In joint work with Pablo Guidotti, Guillermo showed
that, under certain conditions, unanticipated increases in government spending
are optimally financed with unanticipated inflation, which, in the model, imposes
no costs. But does unanticipated inflation really have few costs? In chapter 10—
the opening chapter of part IV—Michael Kumhoff and Evan Tanner argue that
the reason why policy makers avoid unanticipated levies on government debt
(through outright default or sudden burst of high inflation) is that government
debt plays a key role in financial intermediation. Hence, defaulting or inflating it
away would be akin to a negative technological shock. To make an empirical case
for this argument, they show that in developing countries (1) government domes-
tic debt is now larger than external debt, and growing relative to external debt; (2)
banks voluntarily hold a very large fraction of their assets in domestic govern-
ment debt; and (3) a deep and stable government bond market is critical for fur-
ther development of domestic financial markets. In this light, explicit or implicit
defaults on government debt would have real costs that would need to be traded
off against the benefits of lower distortionary taxation.
As countries develop, they increasingly rely on (personal and capital) income
taxes. But how feasible is it to rely on capital income taxation in an increasingly
globalized economy? As globalization makes it easier for firms to either locate in
low-tax countries and/or shift operations across countries to avoid taxation, will
there be a race to the bottom? In chapter 11, Assaf Razin and Efraim Sadka build
a political economy model to assess how feasible it is for governments to impose
capital income taxes. They conclude that a race to the bottom will indeed take
place and argue that available evidence for European countries bears this out.
It is easy for theory to suggest this or that policy. But in practice implementa-
tion of different policies is fraught with difficulties. In particular, the IMF’s track
record in helping developing countries achieve successful reforms is, at best,
xviii Introduction

open to question. In chapter 12, Allan Drazen and Peter Isard argue that an oft-
cited reason IMF programs get off track is a lack of so-called program ‘‘owner-
ship,’’ which loosely refers to a country’s commitment to a given set of reforms
and stabilization policies specified in an IMF program. Drazen and Isard provide
a theoretical framework for thinking about the role of public discussion in build-
ing and demonstrating ownership. In turn, this framework points to various
directions in which IMF programs can be strengthened.
During his seven years at the IMF’s Research Department (1987–1994), Guil-
lermo was witness to one of the most extraordinary economic events of the twen-
tieth century: the transformation of the former communist countries into market
economies. Guillermo’s work in this area was very influential both within and
outside the IMF. In particular, in joint work with Fabrizio Coricelli, Guillermo
forcefully argued that one of the main reasons behind the massive output collapse
at the outset of the transition was a severe credit crunch.
In chapter 13, Fabrizio Coricelli, along with Bostjan Jazbeč and Igor Masten,
takes this line of research one step further. The authors study the macroeconomic
performance of the central European countries in the process of entering the Euro-
pean Union. They conclude that the relative underdevelopment of credit markets
in these countries is partly responsible for low output growth and high volatility.
Joining the European Union should thus allow these economies to further build
their financial systems and to grow faster.
Like Guillermo, Stanley Fischer was a privileged witness to the economic trans-
formation in the former communist countries, first as the World Bank’s Chief
Economist (1988–1990) and later as the IMF’s Deputy Managing Director (1994–
2001). In fact, Fischer (with various collaborators) wrote some of the most influen-
tial pieces on the macroeconomic aspects of the transition. In chapter 14, Fischer
teams up once again with one of his main collaborators, Ratna Sahay, to ex-
amine the role of institutions and initial conditions in the transition process. They
find that, while adverse conditions constrained growth in the initial years, their
effect wore off with time. Fischer and Sahay also reject the charge that interna-
tional financial organizations—and mainstream economists—paid inadequate
attention to institution building. They argue instead that institution building—
typically with substantial assistance from international financial institutions—
complemented policy and structural reforms in promoting growth.
In recent years, Guillermo (with various collaborators) has focused much of his
energy on the consequences of sudden stops in capital inflows—particularly on
output—and how countries appear to recover quickly from such crises without
much internal or external credit. In chapter 15, Ricardo Haussman, Francisco
Rodriguez, and Rodrigo Wagner take this analysis a step further and tackle the
broader question: why do developing economies fall into deeper and longer reces-
Introduction xix

sions than developed ones? Clearly, as the authors note, the phenomenon of deep
and prolonged recessions presents a challenge to standard macroeconomic theory
by suggesting that the typical growth picture of economic fluctuations around an
increasing trend does not quite apply to the developing world. By focusing on
a large sample, the authors conclude that while countries fall into these growth
crises for multiple reasons—including wars, export collapses, sudden stops, and
political transitions—most of these variables do not help predict the episode’s du-
ration. They do find, however, that a measure of the density of a country’s export
base is significantly associated with lower crisis duration.
The volume closes with three short chapters that provide a rare and fascinating
behind-the-scenes look at Guillermo’s career and intellectual journey from his
early days in Buenos Aires through his doctorate at Yale, his years at Columbia,
and his tenure at the IMF’s Research Department, to his more recent and highly
influential role as the Inter-American Development Bank’s Chief Economist. In
chapter 16, Edmund Phelps (recent Nobel Prize winner in Economics and Guiller-
mo’s colleague at Columbia University for many years) first reminisces about the
years at Columbia (where a macroeconomic ‘‘dream team’’ had assembled). In
chapter 17, Roque Fernandez (former Central Bank President and Finance Min-
ister in Argentina) reveals Guillermo’s advisory role at a critical juncture in
Argentina’s recent economic history. Finally, in chapter 18, Enrique Mendoza con-
ducts a probing interview with Guillermo, which offers an incisive and penetrat-
ing look into Guillermo’s intellectual evolution and his views about the major
ideas that have shaped the profession, both academically and from a public policy
point of view.
I Monetary and Exchange Rate
Policy in Theory
1 Pricing-to-Market, the Interest-
Rate Rule, and the Exchange
Rate

Maurice Obstfeld

The beauty of economics as an intellectual pursuit is its position at the intersection

of formal theory, statistical analysis, and human events—coupled with its ulti-
mate potential to improve peoples’ every day lives. A master economist must as-
sume away the distracting inessential details of a situation in the interest of
mathematical clarity. At the same time, he or she must see how relevant subtleties
may affect the interpretation of data and the applicability of different models in
the real world. Because the policy decisions at stake are so complex, with such
vast potential to do harm or good, economics (and especially macroeconomics) is
perpetually unsettled, subject to constant questioning and reassessment. Guil-
lermo Calvo must be ranked as one of the great masters of economics, and one of
the most unsettling. Time and again, starting with his classic work on the dy-
namic inconsistency problem, he has uncovered the hidden crux of a major scien-
tific issue and forced the profession to rethink conventional beliefs.
One area that has undergone extensive rethinking of late is the classic Milton
Friedman case for exchange-rate flexibility, according to which floating exchange
rates are helpful in cushioning national economies from real idiosyncratic shocks.
Paradoxically, one influential assault on the Friedman case originates in the idea
that the pass-through of exchange-rate changes to domestic prices may be high,
another in the idea that pass-through may be very low.
According to Calvo and Reinhart (2002), one factor behind the reluctance of
emerging markets to allow large swings in nominally floating exchange rates is a
relatively rapid pass-through of those swings to consumer prices. Rapid pass-
through of this kind has two implications. Exchange-rate changes will have a
greater potential in the short term to affect domestic inflation, and thereby to im-
pede the pursuit of an inflation target. At the same time, rapid pass-through to all
the prices consumers face blunts the exchange rate’s impact on international rela-
tive prices, and thereby reduces its potential expenditure-switching effects. On
both counts, the costs of exchange-rate volatility are higher in emerging markets
compared with the benefits that Friedman claimed.
4 Maurice Obstfeld

Another way the expenditure-switching effects of the exchange rate can be

eliminated is if there is zero pass-through—both to domestic and import prices.
This is the polar opposite of the case Calvo and Reinhart emphasize, but it would
prevail if both domestic firms and foreign producers of a country’s imports were
to preset their prices in terms of the local currency. Devereux and Engel (2003)
analyze a formal model that includes this type of local-currency pricing. In their
model, they show that welfare-maximizing monetary policies may entail fixed
exchange rates. This theoretical analysis is viewed as a major challenge to the
Friedman case, and one that is applicable to industrial rather than emerging
economies.1 A foreign-based exporter presetting its price in an emerging-market
currency would implicitly be acquiring a contingent asset denominated in that
currency while issuing a contingent liability denominated in goods. This practice
would therefore contradict the observation of ‘‘original sin,’’ which restricts
emerging borrowers to issuing liabilities indexed to international currencies.2 As
a result, local-currency pricing of imports is not expected to characterize emerging
economies.
The modest goal of the present paper is to reinstate Friedman’s case in the
industrial-country setting while retaining the Devereux-Engel local-currency pric-
ing framework.3 A minor modification of the Devereux-Engel model—the in-
troduction of nontraded goods—is enough to restore the need for exchange
flexibility, even when all shocks are real. In my modified model, exchange-rate
changes still have absolutely no expenditure-switching effects in goods markets.
They are necessary, however, to allow countries to pursue independent interest-
rate policies in a world of international capital mobility. That is, the rationale for
exchange flexibility does not necessarily originate in goods markets, as in Fried-
man, but in asset markets. Divergent interest-rate movements may be needed, in
turn, to support the divergent consumption movements implied by idiosyncratic
national technology shocks in the presence of nontraded goods. Exchange-rate
movements can also enhance risk sharing when there are nontraded goods. A by-
product of my argument is an analysis of equilibria and optimal policies in terms
of interest-rate rules rather than money-supply rules—instead, the money supply
is endogenous below.

1.1 The Model

I adopt the basic setup outlined by Devereux and Engel (2003) but modify it in
two ways. First, I model monetary policy as a choice by the central bank of the
nominal interest rate (rather than a monetary aggregate). Second, and more
importantly, I introduce nontraded goods in order to illustrate the scope for an in-
dependent interest-rate policy in the Devereux-Engel local-currency pricing (LCP)
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate 5

framework. What is the intuition for this last effect? With nontraded goods and
flexible prices, a national productivity shock has a disproportionate effect on
home consumption, introducing an ex post asymmetry between countries. To
mimic this response under sticky prices—thereby achieving the best possible
(second-best) ex post allocation—authorities must apply a disproportionate
interest-rate stimulus in their domestic economies.
The basic setup of the model is as follows: there are two (ex ante) symmetrical
countries, Home and Foreign. Each country produces a continuum of tradable
goods (Home’s are indexed by ½0; 1Þ, Foreign’s by ½1; 2
) and a continuum of non-
tradable goods (indexed by ½0; 1Þ).
Each Home representative agent is an atomistic yeoman producer of one differ-
entiated tradable good i and one differentiated nontradable good i, and also sup-
plies labor.4 The producer of generic goods i maximizes
( " #)
Xy 1r
t ðCt ðiÞÞ
U0 ðiÞ ¼ E0 b  kLt ðiÞ ;
t¼0
1r

where C is a consumption index, L is labor supply, r > 0 and b A ð0; 1Þ. Because of
the assumption that the monetary instrument is the nominal interest rate, and that
the money supply adjusts endogenously, there is no need to model explicitly the
demand for money (see Woodford 2003), and I will assume that any money-
demand effects on welfare are negligibly small.
A critical assumption in the model is that of market segmentation between
Home and Foreign. A Home producer of tradables can practice third-degree price
discrimination, charging distinct same-currency prices in the Home and Foreign
markets. By assumption, Home and Foreign consumers (who are also producers
of other goods) face prohibitively high costs of arbitraging the resulting interna-
tional price differentials.
Let Qt be nominal marketable wealth at the start of period t, Pt the nominal
price of consumption during the period, Tt transfer payments from the govern-
ment, and Rtþ1 the nominal ex post return on the agent’s portfolio. Furthermore,
let Yj ðiÞ be the level of output that Home producer i supplies to the Home trad-
ables market ð j ¼ hÞ and to the Home nontradables market ð j ¼ nÞ; let Pj ðiÞ be
the corresponding domestic-currency price charged. To the Foreign market, Home
producer i supplies Yh ðiÞ at Foreign-currency price Ph ðiÞ. Then the flow budget
constraint for producer i is

Qtþ1 ðiÞ  ð1 þ Rtþ1 ðiÞÞQt ðiÞ

 
¼ Ph; t ðiÞYh; t ðiÞ þ Et Ph; t ðiÞYh; t ðiÞ þ Pn; t ðiÞYn; t ðiÞ þ Tt  Pt Ct ðiÞ; ð1:1Þ
6 Maurice Obstfeld

where E is the Home-currency price of Foreign currency (the nominal exchange

rate). There are isomorphic intertemporal maximization problems for the Foreign
agents (whose supplies are denoted by asterisks).
To maximize utility, each producer must grasp the nature of Home and Foreign
demand, which depend in turn on the form of the consumption index. As in
Obstfeld and Rogoff (2000), overall consumption depends on consumption of
tradables and nontradables,

Ctg Cn1g
C¼ ;
g g ð1  gÞ 1g
where the tradables subindex depends on consumption of Home- and Foreign-
produced tradables,

Ct ¼ 2Ch1=2 Cf1=2 :

In turn, Ch , Cf , and Cn are CES functions of the available varieties,

ð 1 y=ðy1Þ
ðy1Þ=y
Cj ¼ Cj ðiÞ di ;
0

with substitution elasticity y.5 Based on these assumptions, demands for the
goods produced by individual i take the forms
     
g Ph ðiÞ y Ph 1 Pt 1
Ch ðiÞ ¼ C;
2 Ph Pt P
     
g Ph ðiÞ y Ph 1 Pt 1 
Ch ðiÞ ¼ C ; and
2 Ph Pt P

 y  1
Pn ðiÞ Pn
Cn ðiÞ ¼ ð1  gÞ C:
Pn P

The exact price indexes entering the price indexes that Home consumers face
are defined as follows:

P ¼ Ptg Pn1g ;

Pt ¼ Ph1=2 Pf1=2 ; and

Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate 7

(Ð1
½ Pj ðiÞ 1y di
1=ð1yÞ ; j ¼ h; n;
Pj ¼ Ð 02
½ 1 Pj ðiÞ 1y di
1=ð1yÞ ; j ¼ f:

There are complete markets in claims on future money payments. As in Backus

and Smith (1993), and given the ex ante symmetry of the two countries, the result-
ing ex post allocation satisfies the condition

Cr
t ðC  Þr
¼ t  ð1:2Þ
Pt Et Pt

in all dates and states, where C  is Foreign consumption and P  is the Foreign
price level measured in Foreign currency. Since purchasing power parity need
not hold ex post in this model, the preceding condition does not generally equal-
ize marginal utilities of consumption internationally.
Production functions for every variety are given by the generic form
Yh ¼ ALh ; Yn ¼ ALn

in Home and

Yf ¼ A Lf ; Yn ¼ A Ln

in Foreign, so that A and A are economy-wide productivity shocks. Letting

lowercase letters (except for interest rates) denote natural logarithms, I write the
stochastic processes followed by the productivity shocks as

at ¼ lat1 þ ut ; at ¼ lat1


þ ut ; ð1:3Þ

where l A ½0; 1
and the shocks u and u  are normally distributed with means of
zero and a common variance of su2 .
Finally, the economy’s nominal anchor is provided by the nominal interest-rate
setting rule followed by the central bank,

logð1 þ it Þ ¼ i þ cpt  ah ut  ah ut ; ð1:4Þ

where it is the nominal interest earned between dates t and t þ 1. (See also
Benigno and Benigno 2008.) Foreign’s central bank has a corresponding rule,

logð1 þ it Þ ¼ i þ cpt  af ut  af ut : ð1:5Þ

It would be possible to add ‘‘noise’’ to these reaction functions by positing that

central banks observe with error some of the variables to which they respond, but
I will not pursue that generalization at this stage.
8 Maurice Obstfeld

1.2 The Flexible-Price Equilibrium

Consider next the model’s equilibrium when all prices are flexible and the central
banks do not respond to productivity innovations (that is, the a coefficients in the
interest-rate rules are all zero). Under flexible prices, producers set domestic-
money prices at a fixed gross markup, y=ðy  1Þ, over nominal marginal cost
(equal to W=A in Home and W  =A in Foreign), where W and W  are the Home
and Foreign nominal wage rates. Using the conditions for the optimal labor-
consumption tradeoff,

W r W
C ¼ k ¼  ðC  Þr ; ð1:6Þ
P P

along with the price-index definitions, one can derive the flex-price levels of over-
all consumption for the two countries:

  
  1=r
y  1 1g=2  g=2 1=r  y1  1g=2 g=2
C¼ A ðA Þ ; C ¼ ðA Þ A : ð1:7Þ
yk yk

Observe that C ¼ C  always in the flex-price equilibrium when all goods are
tradable (that is, when g ¼ 1). But the equality need not hold when g < 1. In the
latter case, a country’s flex-price consumption depends disproportionately on its
own productivity shock. The reason is simple: that shock affects the nontradable
as well as the tradable sector.
The formulas in equation 1.7 suggest already that a differential response of
national interest rates to global and national productivity shocks—and hence,
exchange-rate flexibility—will be necessary under sticky prices to mimic the
flexible-price consumption responses to productivity shocks.
Using the price-index definitions and equation 1.6, one can also establish that in
the flex-price equilibrium,

Ph A Ph
¼ ¼ :
Pf A Pf

Thus, despite discriminatory price setting, in this flexible-price setting Home

and Foreign consumers face the same international relative prices in equilibrium,
and the global allocation of resources is constrained efficient (subject to monopoly
and the nontradability of nontradable goods). Of course, Pn ¼ Ph , and the same is
true in Foreign.
Equilibrium real interest rates must be consistent with the path of expected con-
sumption growth described by equations 1.3 and 1.7. Nominal interest rates and
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate 9

the resulting path for the overall money price level must, in turn, be consistent
with the required path of equilibrium real interest rates.
Nominal interest rates have their relevant impact on the economy through the
intertemporal Euler equation for nominal bond holdings,
 r 
Cr C
t
¼ ð1 þ it ÞbEt tþ1 : ð1:8Þ
Pt Ptþ1

(There is a parallel equation for Foreign.) Taking logs of the preceding equality
and noting that consumption is lognormally distributed, I derive

r2 2 1 2
rct  pt ¼ logð1 þ it Þ þ log b  rEt ctþ1  Et ptþ1 þ s þ s þ rscp :
2 c 2 p

The variances above are endogenous, but because they will be constant over
time, it is simple to compute them once we have solutions for the equilibrium lev-
els of c and p in terms of current shocks and the means and variances of future
variables. After substituting the policy rule in equation 1.4 for logð1 þ it Þ above,
we obtain a difference equation with the unique stable price-level solution:

Xy    
1 sþ1t 1 r2 2 1 2
pt ¼ rðEt fcsþ1  cs gÞ  log b þ i þ sc þ sp þ rscp : ð1:9Þ
s¼t
1þc c 2 2

Above, consumption can be expressed in terms of the underlying technology

shocks using equation 1.7, allowing one to directly compute the equilibrium val-
ues of sc2 , sp2 , and scp . This computation is straight-forward and I therefore omit it
here. At this point I observe only that higher current and expected future con-
sumption growth rates are associated with a higher price level today. The reason
is that higher consumption growth requires higher real interest rates. A higher
price level raises the real interest rate through a policy channel—a higher nominal
interest rate—and through a lowering of inflation expectations. Once the nominal
price levels P and P  are determined, nominal wages and the nominal exchange
rate, which is given by E ¼ W=W  in this model, are likewise pinned down.

1.3 The Model with Preset Nominal Prices

In the sticky-price version of the model, producers of tradables set their domestic
and export prices one period in advance of sales, and must meet all demand that
materializes at that price. Prices can be reset fully after one period, but again must
be maintained for a period thereafter. Exporters set prices in the purchaser’s
currency—there is LCP as in Devereux and Engel (2003). Nontradables producers
10 Maurice Obstfeld

simply set prices in their respective domestic currencies. While these price dynam-
ics would be oversimplified for many purposes, they do allow us to consider the
qualitative stabilization roles of interest and exchange rates in a usefully transpar-
ent setting.
Let’s consider the price-setting problem of a generic Home producer i who sets
prices for date t on date t  1. Because the decision has no repercussions beyond

date t, we may imagine that the producer chooses prices Ph; t , Ph; t , and Pn; t so as
to maximize
( )
Ct ðiÞ 1r
Et1  kLt ðiÞ
1r

subject to equation 1.1,

Yh; t ðiÞ þ Yh  ; t ðiÞ þ Yn; t ðiÞ

Lt ðiÞ ¼ ;
At

and the demand equations

     
g Ph; t ðiÞ y Ph; t 1 Pt; t 1
Yh; t ðiÞ ¼ Ct ;
2 Ph; t Pt; t Pt
!y !1  1
 
 g Ph; t ðiÞ Ph; t Pt; t
Yh; t ðiÞ ¼  Ct ; and
2 Ph; t Pt; t Pt

   
Pn ðiÞ y Pn 1
Yn; t ðiÞ ¼ ð1  gÞ Ct :
Pn P

In performing this maximization the atomistic producer takes macro quantities

and price indexes (not indexed by i above) as given.
Because all domestic prices are preset and, I assume, known as of date t  1, the
first-order condition for Ph; t ðiÞ, for example, is

y kPt Et1 fCt =At g

Ph; t ðiÞ ¼  :
y  1 Et1 fCt ðiÞr Ct g

(I am making an assumption of common knowledge on the part of price setters.)

All producers are symmetric, so Ct ðiÞ ¼ Ct in equilibrium, for all i, and therefore
all set their Home tradable prices equal to

y kPt Et1 fCt =At g

Ph; t ¼  : ð1:10Þ
y1 Et1 fCt1r g
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate 11

Similar reasoning leads to the Home firms’ pricing formulas for exports and
nontradables, respectively:

 y kPt Et1 fCt =At g y kP  Et1 fCt =At g

Ph; t ¼  r  ¼  t ; ð1:11Þ
y  1 Et1 fEt Ct Ct g y  1 Et1 fðCt Þ 1r g

y kPt Et1 fCt =At g

Pn; t ¼  ¼ Ph; t : ð1:12Þ
y1 Et1 fCt1r g

The second equality in (1.11) above is derived using equation 1.2. These three
pricing equations have isomorphic counterparts for Foreign producers.
From equations 1.10–1.12, the relative tradables prices Home consumers face
are

Ph; t Et1 fCt =At g

¼ :
Pf; t Et1 fCt =At g

(Recall that Pf; t is preset, in domestic-currency terms, by Foreign exporters to

Home.) Correspondingly Foreigners face relative tradables prices

Ph; t Et1 fCt =At g
¼ :
Pf; t Et1 fCt =At g


The pricing formulas, equations 1.10–1.12, also yield useful information about
expected consumption levels. Because Pn ¼ Ph , the overall price level is P ¼
1ðg=2Þ g=2
Ph Pf . Thus, 1.10 can be written as
 
Ph; t g=2 y kEt1 fCt =At g
¼  : ð1:13Þ
Pf; t y  1 Et1 fCt1r g

Likewise, using the formula for the Foreign exporter’s price, Pf , one finds that
 1g=2
Pf; t y kEt1 fCt =At g
¼  : ð1:14Þ
Ph; t y  1 Et1 fCt1r g

Combining the last two expressions to eliminate relative prices yields

 1=gð2gÞ
yk ðEt1 fCt =At gÞ1=2g ðEt1 fCt =At gÞ1=2ð2gÞ
1¼ :
y1 ðEt1 fCt1r gÞ1=gð2gÞ

One similarly derives the corresponding expression involving expected Foreign

consumption,
12 Maurice Obstfeld

 1=gð2gÞ
yk ðEt1 fCt =At gÞ1=2g ðEt1 fCt =At gÞ1=2ð2gÞ
1¼ :
y1 ðEt1 fðCt Þ 1r gÞ1=gð2gÞ

Lognormality implies the following solution for the expected logarithm of

Home consumption:
      
1 y1 gð2  gÞ Et atþ1 þ scu Et atþ1 þ scu  su2 r 2
Et ctþ1 ¼ log þ þ   1  sc
r yk r 2g 2ð2  gÞ 2r 2
ð1:15Þ

(with a parallel solution for Et ctþ1 ). This expression is a critical ingredient in the
welfare analysis of monetary policies, because it contributes (via the consumption
Euler equation, 1.8) to the contemporaneous innovation in consumption, ct , and
hence to the variance of consumption and its covariance with technology shocks.
In the present setting the overall price level is known a period in advance, so the
log of the Euler equation 1.8 is


1 r2
ct ¼ Et ctþ1  log b þ logð1 þ it Þ  ð ptþ1  pt Þ þ sc2 : ð1:16Þ
r 2

To solve for the price level now, substitute the interest-rate rule, equation 1.4,
into 1.16, and take date t  1 expectations to derive a difference equation for
pt ¼ Et1 pt ,
  
1 r2
pt ¼ Et1 ptþ1 þ rðEt1 ctþ1  Et1 ct Þ  log b þ i þ sc2 :
1þc 2

I am now allowing the a coefficients in the interest-rate rule to differ from zero,
but because the price level for date t is determined a period earlier, the a values
do not enter its solution, which is

Xy    
1 sþ1t 1 r2
pt ¼ rðEt1 fcsþ1  cs gÞ  log b þ i þ sc2 :
s¼t
1þc c 2

This is the natural extension of equation 1.9 to the case in which pt is a function
only of information dated t  1 or earlier. Using 1.3 and 1.15 to substitute for the
expected consumption terms above, one finds that


  
glð2  gÞðl  1Þ at1 at1 1 r2
pt ¼ þ  log b þ i þ sc2 : ð1:17Þ
1þcl 2g 2ð2  gÞ c 2

A complete solution of the model requires an expression for realized consump-

tion, ct , in terms of the date-t shocks. That expression, in turn, allows computation
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate 13

of the equilibrium values of the key moments sc2 , scu , and scu  . Combination of
equation 1.16 with 1.3, 1.4, 1.15, and 1.17 yields
 
  
cl 2g g 1
ct ¼ at þ at þ ðah ut þ ah ut Þ þ W; ð1:18Þ
1þcl 2r 2r r

where W is a function only of date t  1 (or earlier) information.

Equations 1.7 and 1.18 disclose the key difference in consumption dynamics be-
tween the flexible and fixed-price cases. In the flexible-price case, assuming that
ah ¼ ah ¼ 0, the responses of consumption to technology shocks are given by

dct 2  g dct g
¼ ; ¼ : ð1:19Þ
dat 2r dat 2r

With sticky prices, however, 1.18 shows that the responses of consumption are
muted whenever l < 1. Why? For l ¼ 1, technology follows a random walk and
so does log consumption; according to equation 1.18, current consumption there-
fore can adjust fully with no change in the real rate of interest. When l < 1, how-
ever, consumption is mean-reverting, and current consumption can adjust to its
flex-price level only if the real interest rate falls. In the flexible-price case, pt in-
deed does fall, creating a lower real interest rate both through higher expected in-
flation and through the associated policy-induced fall in the nominal interest rate
it . In contrast, if pt is rigid in the short run, the required real interest rate response
is muted and so is the rise in ct . By appropriate choice of the policy response coef-
ficients ah and ah in equation 1.4, however, the central bank can induce the full
flex-price consumption responses, and I show later that it will wish to do so.
That result also holds in the Devereux-Engel (2003) model with no nontraded
goods, as the authors show. Because g ¼ 1 in their setting, however, flex-price
consumption responses to technology shocks are symmetrical, and so central
banks’ policy responses are absolutely symmetrical as well. That is not the case
when g < 1, for then a relatively more forceful Home interest rate intervention is
needed to mimic the flexible-price consumption response. Variable international
interest-rate differentials imply exchange-rate variation, however, even though
the exchange rate has no expenditure switching effects between Home and For-
eign goods in this model.
As a last step before a formal welfare analysis of policy rules, I derive the
endogenous covariances entering into the model. To simplify the algebra, let us
assume that the productivity shocks are independent, so that suu  ¼ 0. From
equation 1.18,
(
   
    )
2 cl 2g ah 2 cl g ah 2 2
sc ¼ þ þ þ su ; ð1:20Þ
1þcl 2r r 1 þ c  l 2r r
14 Maurice Obstfeld

   
cl 2g ah 2
scu ¼ þ s ; ð1:21Þ
1þcl 2r r u

and

   
cl g a
scu  ¼ þ h su2 : ð1:22Þ
1þcl 2r r

The expressions involving Foreign consumption c  are analogous but involve

the Foreign interest-rate policy coefficients, af and af . Those coefficients do not en-
ter the Home covariances because of the highly insulating role of exchange-rate
changes in this particular LCP setting. Exchange-rate changes merely facilitate in-
dependent monetary policies in a world of international capital mobility.

1.4 Welfare and Optimal Monetary Policy Rules

To assess welfare, observe that the Home labor supply must be consistent with
  1     1   
g Ph; tþ1 Ctþ1 g Ph; tþ1 C
Et Ltþ1 ¼ 1 Et þ  Et tþ1
2 Ptþ1 Atþ1 2 Ptþ1 Atþ1

  g=2   !1g=2 
 
g Pf; tþ1 Ctþ1 g Pf; tþ1 Ctþ1
¼ 1 Et þ  Et :
2 Ph; tþ1 Atþ1 2 Ph; tþ1 Atþ1

Using equation 1.13 and the Foreign analog of 1.14 to eliminate the relative
price terms above, one finds that
 
  
y1 g 1r g
Et Ltþ1 ¼ 1  Et fCtþ1 
g þ Et fðCtþ1 Þ 1r g :
yk 2 2

As a result, period Home expected utility can be written as

( 1r
) ( 1r  
  )
Ctþ1 Ctþ1 y1 g 1r g  1r
Et  kLtþ1 ¼ Et  1  Et fCtþ1 g þ Et fðCtþ1 Þ g
1r 1r y 2 2
   
y  1  2g ðy  1Þð1  rÞ 1r g y1
¼ Et fCtþ1 g  
Et fðCtþ1 Þ 1r g
ð1  rÞy 2 y
  ( )
y  1  2g ðy  1Þð1  rÞ ð1  rÞ 2 2
¼ exp ð1  rÞEt ctþ1 þ sc
yð1  rÞ 2
  ( )
g y1  ð1  rÞ 2 2
 exp ð1  rÞEt ctþ1 þ sc  :
2 y 2
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate 15

I have already noted that the distribution of Foreign consumption, C  , does not
depend on the Home interest-rate rule. Therefore, in considering Home’s optimal
interest-rate rule, I need only consider maximization of the first summand in the
last equation with respect to the feedback coefficients ah and ah . Moreover, using
equation 1.15, it is sufficient to maximize
    
ð1  rÞ 2 1 y1 gð2  gÞ Et atþ1 þ scu Et atþ1 þ scu 
Et ctþ1 þ sc ¼ log þ þ
2 r yk r 2g 2ð2  gÞ

su2 sc2
  ;
2r 2

or even more simply, to maximize

ð2  gÞscu gscu  sc2

V1 þ  :
2r 2r 2

Equations 1.20–1.22 imply that

   
   
cl 2g cl g
V z ð2  gÞ þ ah þ g þ ah
1þcl 2 1þcl 2
(
   2
   2 )
cl 2g cl g
 þ ah þ þ ah :
1þcl 2 1þcl 2
Maximization with respect to the policy parameters yields the procyclical (posi-
tive) responses
 
2g cl
ah ¼ 1 and
2 1þcl
 
 g cl
ah ¼ 1 :
2 1þcl

The Foreign response coefficients in 1.5 are analogous,6 with

 
g cl
af ¼ 1 and
2 1þcl
 
 2g cl
af ¼ 1 :
2 1þcl

A comparison of equations 1.18 and 1.19 reveals that these policy responses
yield a consumption response to innovations in technology that is identical to the
16 Maurice Obstfeld

flex-price response. They make the variance of consumption equal to its flex-price
variance, and induce the flex-price covariances with the shocks to technology. But
interest-rate intervention alone cannot bring the world economy to the flex-price
consumption levels. Policy optimization thus yields a strictly second-best alloca-
tion, with welfare below the flexible-price level. In particular, as Devereux and
Engel (2003) note, consumers will in general face the wrong relative prices in the
preset-price equilibrium, prices that do not reflect true levels of relative economic
scarcity.

1.5 The Need for Exchange-Rate Flexibility

A key point about the preceding second-best interest-rate rules is that they predict
asymmetric national responses to technology shocks—except in the special case of
no nontradables ðg ¼ 1Þ that Devereux and Engel analyze. A useful way to think
about this asymmetry is to define the mutually orthogonal global and idiosyn-
cratic shocks

u þ u
uw 1 and
2

u  u
ud 1 :
2

Then one can express the second-best interest-rate rules for Home and Foreign, re-
spectively, in the simple forms
   
cl cl
logð1 þ it Þ ¼ i þ cpt  1  uw; t  ð1  gÞ 1  ud; t and
1þcl 1þcl
   
cl cl
logð1 þ it Þ ¼ i þ cpt  1  uw; t þ ð1  gÞ 1  ud; t :
1þcl 1þcl

The countries respond identically to the global shock in all cases, but have
oppositely signed responses to the idiosyncratic shock when there are nontrad-
able goods and, consequently, g < 1. I noted the intuition for this result in the in-
troduction: when g ¼ 1, productivity shocks in either country have perfectly
symmetrical consumption effects in the flex-price equilibrium, so internationally
symmetrical interest-rate responses always suffice to induce the flex-price re-
sponse to any shock. That is, when all goods are tradable, it is optimal for central
banks to respond only to global shocks and to respond with equal interest-rate
changes. Nontradables change this. Because a domestic technology shock has a
stronger effect on domestic than on foreign consumption, a relatively more force-
ful domestic interest-rate response may be required.
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate 17

This asymmetry has implications for exchange rates because, given an interest-
rate parity condition, divergent interest-rate movements will call for exchange-
rate changes. To see this formally, observe that the Home and Foreign bond
Euler equations 1.8 of a Home investor may be combined to yield the exchange-
rate equation
  r
1 þ it Et fEtþ1 Ctþ1 g
Et ¼ :
1 þ it Et fCr
tþ1 g

After taking logs and substituting the optimal interest-rate rules, the interna-
tional risk-sharing condition, and the equations 1.18 for ex post consumption
levels, one concludes that, apart from additive predetermined terms, the log ex-
change rate under optimal monetary policies is given by
   
cl 2ð1  gÞc cl Et etþ1
et ¼ 2ð1  gÞ 1  ud; t þ ad; t þ :
1þcl 1þc 1þcl 1þc

This expression makes it clear that idiosyncratic technology shocks will induce
exchange-rate movements through the asymmetric response of consumption,
something that does not occur in this model when g ¼ 1 and all goods are
tradable.7
It is not an expenditure-switching function of exchange rates in commodity
markets that gives them a role in optimal second-best monetary policies. Instead,
the rationale for exchange-rate flexibility lies in the asset markets. Exchange-rate
adjustment makes room for expenditure-changing interest-rate policies, and it
does so by offsetting the incipient expected return differentials that divergent
interest-rate movements would otherwise cause. Exchange-rate movements can
also enhance risk sharing. To enjoy the benefits of both activist monetary policy
and open capital markets, governments must allow the exchange rate to move.

1.6 Conclusion

Even when the exchange rate plays no expenditure-switching role, as in the

Devereux-Engel model, countries may wish to have flexible exchange rates in or-
der to free the domestic interest rate as a stabilization tool. This can be true even
when all shocks are real. Why does no need for exchange flexibility arise in the
Devereux-Engel model with exclusively tradable goods? There, national consump-
tions move in a perfectly synchronized fashion when all shocks are real, whether
prices are flexible or preset. Optimal monetary policy simply raises the sticky-
price consumption response to its flexible-price level, a job that can be accom-
plished through globally symmetric monetary policies that maintain asset-market
equilibrium without the need for exchange-rate changes. In contrast, nontraded
18 Maurice Obstfeld

goods make national consumption responses to asymmetric real shocks asymmet-

ric themselves. In that case, optimal monetary policy requires a relatively greater
monetary stimulus in the country experiencing the shock, and very possibly, a
change in the international nominal interest-rate differential and in the exchange
rate.
These results may strike the reader as abstract, but they are at the crux of mon-
etary policy decisions and institutions with first-order impacts on peoples’ wel-
fare. Guillermo Calvo has repeatedly uncovered the connections between the
seemingly abstruse theorem and the strikingly relevant policy conclusion. As
John Maynard Keynes famously put it, ‘‘The master-economist . . . must be mathe-
matician, historian, statesman, philosopher. . . . He must understand symbols and
speak in words. . . . He must study the present in the light of the past for the pur-
poses of the future.’’8 Even by Keynes’s exacting standard, Calvo easily qualifies
as one of the great master economists of our day.

Notes

Prepared for the International Monetary Fund conference in honor of Guillermo A. Calvo, April 15–16,
2004. I thank Ricardo Caballero, Mathias Hoffmann, Lorenz Küng, and conference participants for
helpful comments.
1. See Engel (2002) for further discussion. Obstfeld and Rogoff (2000) argue that while the industrial-
country retail or consumer prices of imports indeed appear sticky in domestic currency, wholesale im-
port prices do exhibit some exchange-rate pass-through.
2. Eichengreen and Hausmann (1999). Calvo and Reinhart (2002) identify the widespread foreign-
currency denomination of liabilities as another key factor behind ‘‘fear of floating.’’
3. After writing this paper, I learned that Duarte (2004) had independently made essentially the same
point about the need for exchange-rate flexibility (albeit in a somewhat different setting). See also
Duarte and Obstfeld (2008). These two papers are based on money supply rather than nominal interest
rate policy feedback rules. On some implications of this difference, see note 7 below.
4. Foreign producer i supplies tradable 1 þ i and nontradable i.
5. For j ¼ f, the integration is over the interval ½1; 2
.
6. It is easily verified that these coefficients also maximize the equal-weights ‘‘world planner’’ welfare
function 12 U þ 12 U  , as Devereux and Engel (2003) also find. Thus, the Nash equilibrium in a policy-
rule-setting game between the countries is efficient—there is no coordination failure in this model,
though that is a model-specific result (Obstfeld and Rogoff 2002, Benigno and Benigno 2003). It may
also be checked that at this optimum, the value that c (the response to the price level) takes in ð0; yÞ
is irrelevant for welfare. That is, the optimal choice of the a coefficients fully offsets any welfare effect
of c.
7. Notice from the exchange-rate equation that even when l ¼ 1, and productivity shocks therefore are
permanent, the exchange rate will move in response to ad; t ¼ ad; t1 þ uD; t . When l ¼ 1, the difference
equation governing the exchange rate reduces to the simpler form
2ð1  gÞc Et etþ1
et ¼ ad; t þ ;
1þc 1þc
Pricing-to-Market, the Interest-Rate Rule, and the Exchange Rate 19

which has the standard no-bubbles solution. The exchange rate can change even at an unchanged inter-
national interest-rate differential because relative money supplies adjust endogenously, essentially to
mimic the monetary rule given by Duarte and Obstfeld (2008). In the case l ¼ 1, as shown earlier, there
is no need for an interest-rate change to produce the flexible-price response of consumption. The latter
is automatic. Because overall consumer price levels are rigid, however, a globally asymmetric con-
sumption response implies that an exchange-rate change is still needed to maintain the Backus-Smith
risk sharing conditions. In the Devereux-Engel (2003) paper all shocks are permanent. In the absence
of an appropriate feedback policy rule, however, consumption responses may not equal flexible-price
responses, in contrast to the finding of the present paper. The reason is that it is the money supply
rather than the nominal interest rate that is the policy instrument in the analysis of Devereux and
Engel, and the endogenous adjustment of the interest rate, given money supplies, influences the size of
the consumption response.
8. Keynes 1963, p. 141.

References

Backus, David K., and Gregor W. Smith. 1993. ‘‘Consumption and Real Exchange Rates in Dynamic
Economies with Non-Traded Goods.’’ Journal of International Economics 35: 297–316.
Benigno, Gianluca, and Pierpaolo Benigno. 2003. ‘‘Price Stability in Open Economies.’’ Review of Eco-
nomic Studies 70: 743–764.
———. 2008. ‘‘Exchange Rate Determination under Interest Rate Rules.’’ Journal of International Money
and Finance 27, in press.
Calvo, Guillermo A., and Carmen M. Reinhart. 2002. ‘‘Fear of Floating.’’ Quarterly Journal of Economics
117: 379–408.
Devereux, Michael B., and Charles Engel. 2003. ‘‘Monetary Policy in the Open Economy Revisited:
Price Setting and Exchange-Rate Flexibility.’’ Review of Economic Studies 70: 765–783.
Duarte, Margarida. 2004. ‘‘Monetary Policy and the Adjustment to Country-Specific Shocks.’’ Federal
Reserve Bank of Richmond Economic Quarterly 90: 21–40.
Duarte, Margarida, and Maurice Obstfeld. 2008. ‘‘Monetary Policy in the Open Economy Revisited:
The Case for Exchange-Rate Flexibility Restored.’’ Journal of International Money and Finance 27, in press.
Eichengreen, Barry, and Ricardo Hausmann. 1999. ‘‘Exchange Rates and Financial Fragility.’’ In New
Challenges for Monetary Policy, 329–368. Kansas City, MO: Federal Reserve Bank of Kansas City.

Engel, Charles. 2002. ‘‘Expenditure Switching and Exchange-Rate Policy.’’ In NBER Macroeconomics An-
nual 2002, ed. Mark Gertler and Kenneth Rogoff, 231–272. Cambridge, MA: MIT Press.
Keynes, John Maynard. 1963. ‘‘Alfred Marshall.’’ In Essays in Biography, 125–217. New York: W. W.
Norton & Company.
Obstfeld, Maurice, and Kenneth Rogoff. 2000. ‘‘New Directions for Stochastic Open Economy Models.’’
Journal of International Economics 50: 117–153.
———. 2002. ‘‘Global Implications of Self-Oriented National Monetary Rules.’’ Quarterly Journal of Eco-
nomics 117: 503–535.
Woodford, Michael. 2003. Interest and Prices. Princeton, NJ: Princeton University Press.
2 Optimal Exchange Rate
Regimes: Turning Mundell-
Fleming’s Dictum on Its Head

Amartya Lahiri, Rajesh Singh, and

Carlos A. Végh

2.1 Introduction

One of the most influential results in open economy macroeconomics—which

derives from the Mundell-Fleming model (see Mundell 1968 and Fleming 1962)—
holds that the choice of the exchange rate regime should depend on the type of
shock hitting the economy. If shocks are predominantly of real origin, then flexi-
ble exchange rates are optimal. Instead, if shocks are mainly monetary, fixed (or,
more generally, predetermined) exchange rates are optimal. In fact, as Calvo
(1999, 4) aptly puts it, this is ‘‘a result that every well-trained economist carries on
her tongue.’’ Calvo (1999) himself offers a simple derivation of this result in a
model in which the policy maker’s objective is to minimize output variability.
The intuition is simple enough: in the Mundell-Fleming world of sticky prices
and perfect capital mobility, real shocks require an adjustment in relative prices
which, in the presence of sticky prices, can most easily be effected through
changes in the nominal exchange rate. In contrast, monetary shocks require an ad-
justment in real money balances that can be most easily carried out through
changes in nominal money balances (which happens endogenously under fixed
exchange rates). By and large, this key result has remained unscathed in modern
variations of the Mundell-Fleming model. For instance, Cespedes, Chang, and
Velasco (2004) incorporate liability dollarization and balance sheets effects and
conclude that the standard prescription in favor of flexible exchange rates in re-
sponse to real shocks is not essentially affected.
But rather than tweaking at the margin with variations of the traditional
Mundell-Fleming model, it could be argued that one should take issue with its
most critical assumption: imperfection in goods markets (i.e., sticky prices) but
undistorted capital markets (in other words, perfect capital mobility). Is this the
world we necessarily live in? Far from it. In developing countries, in particular,
asset market frictions appear to be equally, if not more, important than goods
market frictions. In fact, a large segment of the population does not seem to have
22 Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

access to asset markets.1 In this light, it seems worth revisiting the Mundell-
Fleming question in a model with flexible prices but segmented asset markets. This
type of model posits that while a fraction of the population (referred to as traders)
has access to asset markets, the rest of the population (nontraders) does not. In an
earlier paper (Lahiri, Singh, and Végh 2007), we examined this issue in the context
of a stochastic model in which traders have access to incomplete markets. In con-
trast, this paper develops a much starker, perfect-foresight version of the model
that, by avoiding myriad technical complications, allows the essential mecha-
nisms and intuition to shine through. The paper’s punchline is that—contrary to
the Mundell-Fleming prescription mentioned above—if shocks are real, fixed ex-
change rates are optimal, whereas if shocks are monetary, flexible exchange rates
are optimal.
Intuitively, flexible exchange rates allow for a costless adjustment to monetary
shocks by altering the real value of existing nominal money balances. In contrast,
under fixed rates, asset-market segmentation prevents nontraders from rebal-
ancing real money balances by accessing asset markets, which affects the con-
sumption path. Under real shocks, fixed rates allow purchasing power to be
transferred across periods, which results in some consumption smoothing. Under
flexible rates, on the other hand, nontraders are forced to consume their current
endowment.
We thus conclude that the optimal exchange rate regime should depend not
only on the type of shock (real versus monetary)—as rightly emphasized by
Mundell-Fleming models—but also on the type of distortion (goods markets versus
asset markets frictions).2 These ideas can be succinctly summarized in the follow-
ing 2
2 matrix:
Table 2.1
Optimal exchange rate regime
Goods market friction Asset market friction
Real shock Flexible Fixed
Monetary shock Fixed Flexible

The optimal exchange rate regime thus becomes an empirical issue that
depends both on the type of shock hitting a particular economy and on the rela-
tive distortions present in goods and asset markets.
The paper proceeds as follows. Section 2.2 develops the main model—a perfect-
foresight version of Lahiri, Singh, and Végh (2007)—and solves it for the cases of
both flexible and fixed exchange rates. Section 2.3 compares the two regimes for
fluctuating output and velocity paths. Section 2.4 contains some brief concluding
remarks. Some technical issues are relegated to appendices.
Optimal Exchange Rate Regimes 23

2.2 The Model

Consider a discrete-time model of a small open economy perfectly integrated with

the rest of the world in goods markets. There are two types of agents: traders
(who have access to capital markets) and nontraders (who do not have access to
capital markets). The fraction of traders is l while that of nontraders is 1  l.
There is no uncertainty in the model and agents are blessed with perfect foresight.
The law of one price holds for the only good; hence, Pt ¼ Et Pt , where Pt is the
nominal price of the good, E is the nominal exchange rate (in units of domestic
currency per unit of foreign currency), and Pt is the foreign nominal price. For-
eign inflation is assumed to be zero and, for simplicity, Pt is taken to be unity.
Hence, Pt ¼ Et .
Both traders and nontraders are subject to a cash-in-advance constraint. For the
case of traders, we follow Lucas’s (1982) timing and assume that asset markets
open first (say, in the morning) followed by goods markets (in the afternoon). By
assumption, of course, nontraders do not have access to asset markets and hence
only visit goods markets.3
There are two types of shocks: real and monetary. Both traders and nontraders
face identical shocks. Real shocks are captured by fluctuations in the endowment
of the only good, y. Following Alvarez, Lucas, and Weber (2001), we capture
monetary—or velocity—shocks by allowing both traders and nontraders to ac-
cess a fraction v of current period sales ðvt Pt yt Þ and letting vt fluctuate over time.
To fix ideas, it proves useful to keep in mind the following scenario regarding
the model’s timing conventions. Households consist of two individuals: a shopper
and a seller. As is standard, households do not consume their own endowment.
As goods markets are about to open in the afternoon, the seller and the shopper
part and, in the standard model, do not see each other until the end of the day. In
other words, the seller stays in the store selling the endowment to other house-
holds’ shoppers and the shopper visits other stores to purchase goods. In the stan-
dard model, then, the shopper does not return to the store until after goods
markets close and therefore has no access to the money balances accrued to the
seller from the sale of the current-period endowment ðPt yt Þ. In the current paper,
we depart from the standard model by allowing the shopper to come back to the
store once during the goods market session, empty the cash register, and go back
to shopping. We assume that the amount of money in the cash register at the time
the shopper comes back to the store is vt Pt yt , where 0 < v < 1.
Finally, both traders (T) and nontraders (NT) have identical preferences, given by

X
y
Ui ¼ b t uðcti Þ; i ¼ T; NT ð2:1Þ
t¼0

where c i denotes consumption of an agent of type i.

24 Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

2.2.1 Nontraders

Nontraders do not have access to asset markets and hence hold only money. Their
flow budget constraint is given by
NT
Mtþ1 ¼ MtNT þ Et yt  Et ctNT ; ð2:2Þ

where MtNT denotes end-of-period t  1 (and hence beginning of period t) nomi-

nal money balances in the hands of nontraders. The initial level of nominal money
balances, M0NT , is given. Nontraders are subject to a cash-in-advance constraint of
the form

MtNT þ vt Et yt b Et ctNT : ð2:3Þ

The nominal money balances that nontraders can use to purchase goods consist of
the nominal money balances that they bring into period t, MtNT , and a fraction vt
of current-period sales (recall that, by assumption, 0 < vt < 1).
We will only consider equilibrium paths along which the cash-in-advance con-
straint binds.4 If the cash-in-advance binds, then we can solve for ctNT from equa-
tion 2.3 to obtain

MtNT þ vt Et yt
ctNT ¼ ; t b 0: ð2:4Þ
Et

To find out how much money balances nontraders will carry on to the next pe-
riod, substitute 2.4 into 2.2 to obtain
NT
Mtþ1 ¼ ð1  vt ÞEt yt : ð2:5Þ

When the cash-in-advance constraint binds, the nontraders’ problem becomes

completely mechanical. In other words, their opportunity set consists of only one
point in every period—given by equation 2.4—and there is thus no need to carry
out any maximization. Intuitively, nontraders begin their life with a given level of
nominal money balances, M0NT . They augment these cash balances with a fraction
v0 of period 0 sales, v0 E0 y0 . Since the cash-in-advance binds, they spend all of
their money balances, M0NT þ v0 E0 y0 , on consumption in period 0. Their end-of-
period cash balances consist of the cash proceeds from selling their endowment,
E0 y0 , minus the amount of period 0 sales spent in period 0, v0 E0 y0 . They thus
enter period 1 with M1NT ð¼ ð1  v0 ÞE0 y0 Þ and the process begins anew.
Optimal Exchange Rate Regimes 25

2.2.2 Traders

Traders have access to asset markets and thus behave like consumers in any stan-
dard model with perfect capital mobility. The only difference is that, like non-
traders, they have access to a fraction vt of current-period sales.
Let us first look at the flow constraint for the asset market. Traders enter the as-
set market with a certain amount of nominal money balances, MtT , and a certain
amount of bonds, bt . Once in the asset market, they receive/pay interest on the
bonds they carried into the asset markets, Et rbt ; receive transfers from the govern-
ment, T; and buy/sell bonds in exchange for money.5 Traders exit the asset mar-
ket with a quantity M ^ T of nominal money balances and btþ1 of bonds. The flow
t
constraint for the asset market is thus

^ T ¼ M T þ Et ð1 þ rÞbt þ Tt :
Et btþ1 þ M ð2:6Þ
t t
l

Traders are subject to a cash-in-advance constraint:

^ T þ vt E t yt b E t c T :
M ð2:7Þ
t t

What will traders’ nominal money balances be at the end of period t? Traders will
have the money brought from the asset market plus the proceeds from the sale of
their endowment ðEt yt Þ minus the money balances used to purchase goods ðEt ct Þ:
T
Mtþ1 ^ T þ Et yt  Et c T :
¼M ð2:8Þ
t t

By substituting 2.8 into 2.6, we obtain the traders’ flow constraint for period t as a
whole:

T Tt
Et btþ1 þ Mtþ1 ¼ MtT þ Et ð1 þ rÞbt þ Et yt þ  Et ctT : ð2:9Þ
l

2.2.2.1 Utility Maximization

For the purposes of the maximization—and by substituting equation 2.6 into
2.7—we can rewrite the cash-in-advance constraint as

Tt
MtT þ Et ð1 þ rÞbt þ  Et btþ1 þ vt Et yt b Et ctT : ð2:10Þ
l

Traders thus maximize lifetime utility subject to the flow budget constraint 2.9
and the cash-in-advance constraint 2.10, for given values of M0T and b0 . The
Lagrangian is then given by
26 Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

X
y X
y

Tt
b t uðc T Þ þ b t ht MtT þ Et ð1 þ rÞbt þ þ Et yt  Et ctT  Et btþ1  Mtþ1
T

t¼0 t¼0
l

X
y

t Tt
þ b Ct MtT T
þ Et ð1 þ rÞbt þ  Et btþ1 þ vt Et yt  Et ct ;
t¼0
l

where ht and Ct denote the multipliers associated with constraints 2.9 and 2.10,
respectively.
The first-order conditions with respect to c T , Mtþ1
T
, and btþ1 are given, respec-
tively, (assuming, as usual, that bð1 þ rÞ ¼ 1) by

u 0 ðctT Þ ¼ Et ðht þ Ct Þ; ð2:11Þ

ht ¼ bðhtþ1 þ Ctþ1 Þ; and ð2:12Þ

Etþ1 ðhtþ1 þ Ctþ1 Þ ¼ Et ðht þ Ct Þ: ð2:13Þ

The first-order condition with respect to ht naturally recovers the flow con-
straint 2.9. Finally, the Kuhn-Tucker condition for Ct recovers 2.10 and requires
the complementary slackness condition


T Tt T
Mt þ Et ð1 þ rÞbt þ  Et btþ1  Et ct Ct ¼ 0: ð2:14Þ
l

Combining first-order conditions 2.11 and 2.13 yields

u 0 ðctT Þ ¼ u 0 ðctþ1
T
Þ:

As in standard cash-in-advance models with Lucas’s (1982) timing, traders will

fully smooth consumption over time.
1
Combining first-order conditions 2.12 and 2.13 (using b ¼ 1þr ) yields


Etþ1
ht ð1 þ rÞ  1 ¼ Ct : ð2:15Þ
Et

Perfect capital mobility (for traders) implies that the interest parity condition
holds:

Etþ1
1 þ it ¼ ð1 þ rÞ ; ð2:16Þ
Et

which enables us to rewrite condition 2.15 as

h t it ¼ Ct : ð2:17Þ
Optimal Exchange Rate Regimes 27

Since, at an optimum, ht > 0, equation 2.17 says that if it > 0, then Ct > 0, which
implies from the complementary slackness condition 2.14 that the cash-in-
advance constraint binds. Since we will only consider equilibria in which the
nominal interest rate is positive, the cash-in-advance constraint will always bind
and traders’ end-of-period money balances can be obtained by combining 2.7 and
2.8:6
T
Mtþ1 ¼ ð1  vt ÞEt yt : ð2:18Þ

2.2.3 Government

The government’s flow constraint is given by

Et htþ1 ¼ ð1 þ rÞEt ht þ Mtþ1  Mt  Tt ; ð2:19Þ

where ht denotes net foreign bonds held by the government.

2.2.4 Equilibrium Conditions

Money market equilibrium implies that

Mt ¼ lMtT þ ð1  lÞMtNT : ð2:20Þ

NT T
Equations 2.5 and 2.18 imply that Mtþ1 ¼ Mtþ1 . Together with the money mar-
ket equilibrium condition 2.20, this implies that

Mt ¼ MtNT ¼ MtT :

Since there are no differences across agents in terms of the endowment, all agents
hold the same amount of money (on a per-capita basis). Hence, 2.5 and 2.18 to-
gether with the money market equilibrium condition 2.20 yield a quantity theory
equation:

Mtþ1 ¼ ð1  vt ÞEt yt ; t b 0: ð2:21Þ

To make it directly comparable to the quantity theory equation found in text-

books (typically written as MV ¼ Py, where V denotes velocity), we can rewrite
this last equation as

Mtþ1
¼ Et yt ; t b 0:
1  vt
28 Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

Velocity is therefore given by 1=ð1  vt Þ. Thus, a higher v captures an increase in

the velocity of circulation, which rationalizes our term ‘‘velocity shocks’’ when re-
ferring to changes in v.
To obtain the economy’s flow constraint, multiply the nontraders’ flow con-
straint (equation 2.2) by 1  l and the traders’ flow constraint (equation 2.9) by l
and then add them, taking into account the government’s flow constraint (equa-
tion 2.19) and the money market equilibrium condition (equation 2.20):

ktþ1  kt ¼ rkt þ yt  ½lctT þ ð1  lÞctNT ; ð2:22Þ

where k 1 ht þ lbt denotes the economy’s per-capita net foreign assets.

ktþ1
Iterating forward and imposing the transversality condition limt!y ð1þrÞ t
¼ 0,
we obtain the resource constraint

X
y X
y
ð1 þ rÞk 0 þ b t yt ¼ b t ½lctT þ ð1  lÞctNT : ð2:23Þ
t¼0 t¼0

In what follows, we will assume that k 0 ¼ 0.7

2.2.5 Equilibrium Consumption

We will now derive expressions for consumption of both traders and nontraders.
To obtain nontraders’ consumption, substitute the quantity theory equation 2.21
into 2.4 (recall that Mt ¼ MtNT ) to obtain
8
< M 0 þ v0 y0 ; t¼0
E0
ctNT ¼ ð1v ÞE þv
ð2:24Þ
: t1 t1 y t1 t Et y t
; t b 1:
Et

This expression will prove useful when dealing with fixed exchange rates.
When dealing with flexible exchange rates, however, it will prove convenient to
use 2.21 to rewrite 2.24 as

Mtþ1  Mt
ctNT ¼ yt  ; t b 0: ð2:25Þ
Et

To obtain traders’ consumption, substitute 2.24 into 2.23 and solve for the con-
stant level of c T , denoted by c T , to obtain
" !#
1l p Xy
T p r M0 t ð1  vt1 ÞEt1 yt1 þ vt Et yt
c ¼y þ y  þ v0 y0 þ b ;
l 1 þ r E0 t¼1
Et
ð2:26Þ
Optimal Exchange Rate Regimes 29

where

X
y
y p 1 ð1  bÞ b t yt
t¼0

denotes permanent income. Alternatively, substitute 2.25 into 2.23 and iterate to
obtain
 
p r 1  lX
y
t Mtþ1  Mt
cT ¼y þ b : ð2:27Þ
1 þ r l t¼0 Et

Equations 2.25 and 2.27 make clear the redistributive role that monetary policy
plays in this model. If, say, money supply is constant, then nontraders consume
their endowment ðctNT ¼ yt Þ and traders their permanent income ðc T ¼ y p Þ. An
increase in the money supply ðMtþ1 > Mt Þ implies a transfer from nontraders to
traders. The reverse is true in the case of a reduction in the money supply.

2.2.6 Flexible Exchange Rates

Consider a flexible exchange rate regime in which the monetary authority sets a
constant path of the nominal money supply:8

Mt ¼ M; t b 0: ð2:28Þ

Substituting 2.28 into 2.25, we obtain nontraders’ consumption:

ctNT ¼ yt ; t b 0: ð2:29Þ

Two observations are worth making. First, consumption of nontraders will fluctu-
ate one-to-one with fluctuations in the endowment. Flexible exchange rates pro-
vide no insulation whatsoever for nontraders from output fluctuations. Second,
consumption of nontraders is not affected by velocity shocks.
Substituting 2.28 into 2.27, we obtain traders’ consumption

c T ¼ y p: ð2:30Þ
Let us now derive the path of the nominal exchange rate. From the quantity
theory equation 2.21, we obtain

M
Et ¼ ; t b 0: ð2:31Þ
ð1  vt Þyt

It follows that
30 Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

Etþ1 ð1  vt Þ yt
¼ : ð2:32Þ
Et ð1  vtþ1 Þ ytþ1

When output increases (that is, ytþ1 > yt )—and for constant velocity—the nomi-
nal exchange rate will fall (i.e., the domestic currency appreciates). Intuitively,
higher output increases real money demand and hence leads to a fall in the price
level (in other words, in the nominal exchange rate). On the other hand, when
there is an increase in velocity (i.e., vtþ1 > vt )—and for constant output—the
nominal exchange rate will increase (i.e., the domestic currency depreciates).
Intuitively, an increase in velocity implies that more money is available to pur-
chase the same level of output, which will lead to a higher price level (that is, a
higher nominal exchange rate).
Finally, the path of the nominal interest rate follows from combining the inter-
est parity condition 2.16 with 2.32:

ð1  vt Þ yt
1 þ it ¼ ð1 þ rÞ : ð2:33Þ
ð1  vtþ1 Þ ytþ1

2.2.7 Fixed Exchange Rates

Consider now a fixed exchange rate regime in which the monetary authority sets
a constant value of the nominal exchange rate:
Et ¼ E:

To ensure that initial conditions under fixed rates are consistent with those under
flexible rates (in the sense that they generate the same initial level of real-money
balances as in the case of flexible exchange rates), we take initial nominal money
holdings to be M0NT ¼ M0T ¼ M0 ¼ M. Further, we assume that the exchange rate
is fixed at the level given by E ¼ M=ð1  v0 Þy0 . Under these assumptions, initial
real money balances M0 =E are given by ð1  v0 Þy0 , as is the case under flexible
rates (recall 2.31).
Under a fixed exchange rate, we can use equation 2.24 to obtain consumption of
nontraders:
(
M0
þ v0 y0 ; t ¼ 0;
ctNT ¼ E ð2:34Þ
ð1  vt1 Þ yt1 þ vt yt ; t b 1:

Since M0 =E ¼ ð1  v0 Þy0 , it follows that c0NT ¼ y0 .

By the same token, using 2.26, consumption of traders is given by
( !)
1l r Xy
T p p t
c ¼y þ y  y0 þ b ½ð1  vt1 Þ yt1 þ vt yt  : ð2:35Þ
l 1þr t¼1
Optimal Exchange Rate Regimes 31

Let us now derive the path of the nominal money supply, which is endogenous
under fixed exchange rates. M0 ¼ M, as remarked earlier. The path of Mt for t b 1
then follows from the quantity theory equation 2.21:
Mtþ1 ¼ ð1  vt ÞEyt ; t b 0:

2.3 Comparing Flexible versus Fixed Exchange Rates

We are now ready to ask our main question: which exchange regime is better?

2.3.1 Velocity Shocks Only

Suppose that there are only velocity shocks (i.e., set yt ¼ y p ). Then, under flexible
rates, consumption of nontraders is completely flat and equal to y p (as follows
from equation 2.29). Further, as equation 2.30 indicates, traders’ consumption is
also equal to permanent income. Clearly, this equilibrium corresponds to the
first-best. Both traders and nontraders perfectly smooth consumption over time.
Under fixed rates, it follows from equation 2.34 that consumption of nontraders
is given by

y p; t ¼ 0;
ctNT ¼ ð2:36Þ
y p ð1 þ vt  vt1 Þ; t b 1:

In turn, consumption of traders is given by (from 2.21 and 2.27)

" #
T p 1 lXy
t
c ¼ y 1 b ðvt  vt1 Þ : ð2:37Þ
l t¼0

In what follows, it will prove useful to define a ‘‘permanent’’ velocity shock, v p , as

X
y
v p 1 ð1  bÞ b t vt :
t¼0

Under the assumption that v0 ¼ v p , it follows that (see appendix 5.2)9

X
y
b t ðvt  vt1 Þ ¼ 0: ð2:38Þ
t¼1

Substituting 2.38 into 2.37, we obtain traders’ consumption:

c T ¼ y p:
32 Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

Traders’ consumption is therefore the same under flexible and fixed exchange
rates and they are thus indifferent between the two regimes. As for nontraders, it
follows from 2.36 and 2.38 that the present discounted value of nontraders’ con-
sumption under fixed rates is the same as under flexible exchange rates. As a re-
sult, nontraders are clearly better off under flexible exchange rates, in which case
they have a flat path of consumption. Since traders are indifferent, we conclude
that flexible exchange rates dominate.
What is the underlying intuition? The key lies in the role of the exchange rate
as a shock absorber in the presence of velocity shocks. If velocity increases, the
nominal exchange rate also increases (a nominal depreciation of the domestic cur-
rency), thus offsetting the shock. Under fixed exchange rates, the natural adjust-
ment mechanism (the agents’ ability to recompose their nominal money balances
through the central bank) is not fully operative because nontraders cannot access
asset markets. Hence, fluctuations in velocity lead to fluctuations in consumption.
Specifically, an increase in velocity ðvt > vt1 Þ implies that more money balances
are available for consumption; a decrease in velocity ðvt < vt1 Þ implies that fewer
money balances are available for consumption.

2.3.2 Output Shocks Only

Suppose that there are only output shocks (i.e., set vt ¼ v > 0). Then under flexi-
ble rates, consumption of nontraders and traders continues to be given by 2.29
and 2.30. Nontraders absorb the full variability of the endowment path.
Under fixed exchange rates, consumption of nontraders follows from 2.34:

y0 ; t ¼ 0;
ctNT ¼ ð2:39Þ
yt þ ð1  vÞð yt1  yt Þ; t b 1:

Under the assumption that y0 ¼ y p , it follows that (see appendix 5.3)

X
y
b t ð yt1  yt Þ ¼ 0: ð2:40Þ
t¼1

From 2.39 and 2.40, it follows that the present discounted value of ctNT under fixed
rates will be the same as under flexible rates.
Consumption of traders follows from 2.35 and 2.40:

c T ¼ y p: ð2:41Þ
As is the case under velocity shocks, traders’ consumption is the same under flex-
ible and fixed exchange rates. Traders are therefore indifferent between the two
regimes.
Optimal Exchange Rate Regimes 33

For expositional clarity, it is useful to rewrite nontraders’ consumption as


NT y0 ; t ¼ 0;
ct ¼ ð2:42Þ
vyt þ ð1  vÞ yt1 ; t b 1;

which makes clear that from t ¼ 1 onward, nontraders’ consumption is an aver-

age of this period’s and last period’s output. Clearly, consumption of nontraders
will fluctuate under both flexible and fixed exchange rates but will fluctuate less
under fixed rates. Since, as already shown, the present discounted value of ctNT is
the same under both regimes, nontraders’ welfare will be higher under fixed ex-
change rates.
Intuitively, 2.42 states that today’s consumption is a weighted average of last
period’s and this period’s real sales revenues. Fixed exchange rates allow purchas-
ing power to be transferred across periods which, as equation 2.42 makes clear,
results in some consumption smoothing over time. In contrast, under flexible rates
a constant money supply implies that the real value of last period’s sales is equal
to current output. As a result, current consumption depends solely on current
output.
We conclude that since traders are indifferent between the two regimes and
nontraders are better off under a fixed exchange rate, social welfare will be maxi-
mized if, in response to output shocks, a fixed exchange rate is adopted.

2.4 Concluding Remarks

We have shown that the influential Mundell-Fleming result—that the choice of

the optimal exchange rate regime should depend on the type of shock hitting the
economy—critically depends on the assumption that while there are frictions in
goods markets (in other words, sticky prices), asset markets are frictionless. If
we reverse these assumptions—frictionless goods markets and segmented asset
markets—we turn the famous Mundell-Fleming dictum on its head: flexible rates
are called for in the presence of monetary shocks whereas fixed exchange rates are
optimal in the presence of real shocks. We thus conclude that the optimal ex-
change rate depends not only on the type of shock (monetary versus real) but
also on the type of friction (goods market versus asset market).
A more modern approach to exchange rate regimes would view fixed and flexi-
ble exchange rate regimes as two particular cases of a more general monetary
policy rule, which could in turn incorporate a response to current (if observable)
and past shocks. In Lahiri, Singh, and Végh (2006), we follow this more general
approach and show how an optimal monetary policy rule would actually involve
responding to contemporaneous shocks. It is only in the absence of output shocks
34 Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

(that is, a world with only velocity shocks) that a ‘‘pure’’ flexible exchange rate—
as studied in this paper—would be optimal.

2.5 Appendixes

2.5.1 Conditions for a Binding Cash-in-Advance

This appendix derives the conditions needed for the cash-in-advance to bind for
both nontraders and traders and then provides an example of the restrictions that
need to be imposed on the output and velocity processes.

2.5.1.1 When Does the Cash-in-Advance Bind for Nontraders?

NT y
Nontraders choose fctNT ; Mtþ1 gt¼0 to maximize lifetime utility equation 2.1 subject
to the sequence of flow constraints given by 2.2 and the sequence of cash-in-
advance constraints given by 2.3 for a given M0NT . In terms of the Lagrangian:

X
y X
y
L¼ b t uðctNT Þ þ b t lt ðMtNT þ Et yt  Et ctNT  Mtþ1
NT
Þ
t¼0 t¼0

X
y
þ b t Ct ðMtNT þ vt Et yt  Et ctNT Þ;
t¼0

where lt and Ct are the multipliers associated with constraints 2.2 and 2.3,
respectively.
The first-order conditions for ctNT and Mtþ1
NT
are given by

u 0 ðctNT Þ ¼ Et ðlt þ Ct Þ and ð2:43Þ

bðltþ1 þ Ctþ1 Þ ¼ lt : ð2:44Þ

The Kuhn-Tucker condition for Ct is given by

MtNT þ vt Et yt b Et ctNT ; Ct b 0;

ðMtNT þ vt Et yt  Et ctNT ÞCt ¼ 0:

Suppose that Ct > 0; that is, the cash-in-advance constraint binds. Then, it fol-
lows from 2.43 and 2.44 that

u 0 ðctþ1
NT
Þ 1 Etþ1 lt
¼ :
u 0 ðctNT Þ b Et lt þ Ct
Optimal Exchange Rate Regimes 35

Hence, for the cash-in-advance to bind, it must be the case that

Et 0 NT
u 0 ðctNT Þ > b u ðctþ1 Þ: ð2:45Þ
Etþ1

If the cash-in-advance binds, it means that nontraders prefer not to carry over
nominal money balances from one period to the next even though doing so would
provide more consumption tomorrow. In other words, money balances are not
used for saving purposes. In this case—and as condition 2.45 indicates—the con-
sumer is unwilling to save and therefore today’s marginal utility will be higher
than tomorrow’s, adjusted by the discount factor and the return on money.
To fix ideas, consider the case of logarithmic preferences. Condition 2.45 then
reduces to

1 Etþ1
ctNT < ctþ1
NT
: ð2:46Þ
b Et

Using the quantity theory (equation 2.21), we can rewrite this equation as
  
ctNT 1 1  vt yt
NT
< ð1 þ mtþ1 Þ : ð2:47Þ
ctþ1 b 1  vtþ1 ytþ1

Flexible Exchange Rates Consider the case of flexible exchange rates with a con-
stant money supply. In this case, ctNT ¼ yt and mtþ1 ¼ 0. Equation 2.47 then
reduces to

1  vt
b< : ð2:48Þ
1  vtþ1

As long as this condition holds (which implies a restriction on the variability of

the velocity shocks), the cash-in-advance constraint will bind. Clearly, since this
condition involves exogenous variables, one can always choose parameters such
that it will hold.

Fixed Exchange Rates Consider the case of fixed exchange rates. In this case,
Etþ1 ¼ Et ¼ E. Use condition 2.46, taking into account 2.34, to obtain

ð1  vt Þ yt þ vtþ1 ytþ1
b< : ð2:49Þ
ð1  vt1 Þyt1 þ vt yt

Again, since this condition involves only exogenous variables, one can always
choose b and output and velocity processes such that it holds.
36 Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

Intuition To understand the intuition as to why the cash-in-advance (CIA) may

bind for nontraders, consider the case of flexible exchange rates and no velocity
shocks (that is, output shocks only). In that case, condition 2.48 will always hold
because, by assumption, b < 1. Intuitively, suppose that yt > ytþ1 and consider
the nontrader’s choice at time t. Based on the consumption-smoothing motive,
nontraders would want to save in order to consume more next period when out-
put will be low. However, given that mt ¼ 0, periods of high output will coincide
with periods in which the real return on nominal money balances is low. To see
this, notice that using the cash-in-advance the gross real return on holding money
is given by

Et ytþ1
¼ :
Etþ1 yt

Since yt > ytþ1 , then Et =Etþ1 < 1, which means a negative real return on money.
Hence, with logarithmic preferences, the nontrader’s desire to dissave based on
the negative real return on money more than offsets the desire to save based
on consumption-smoothing motives.

2.5.1.2 When Does the Cash-in-Advance Bind for Traders?

For the CIA to bind for traders, we just need to ensure that the nominal interest
rate is positive. The restrictions needed for this depend on the exchange rate
regime.

Flexible Exchange Rates From the interest parity condition 2.16, a positive nom-
inal interest rate requires that

Etþ1 1
> :
Et 1þr

Using the quantity theory equation 2.21, it follows that

Etþ1 1  vt1 yt1

¼ :
Et 1  vt y t

Combining the last two equations—and recalling that bð1 þ rÞ ¼ 1—it follows
that if

1  vt1 yt1
b< ; ð2:50Þ
1  v t yt

then the nominal interest rate will always be positive and the CIA will always
bind for traders as well.
Optimal Exchange Rate Regimes 37

Fixed Exchange Rates Under fixed exchange rates, the interest parity condition
2.16 indicates that the nominal interest rate will always be positive since
1 þ i ¼ 1 þ r.

2.5.1.3 An Example
Let us illustrate the restrictions necessary to ensure a binding cash-in-advance
constraint for cases of only one shock at a time (the case studied in the text). Sup-
pose b ¼ 0:96.

Output Shocks Only Suppose that vt ¼ v ¼ 0:2 > 0 and that yt alternates be-
tween 1.04 and 1. For nontraders, 2.48 holds since b < 1 and condition 2.49
becomes (assuming the most restrictive case, which is yt1 ¼ 1:04, yt ¼ 1, and
ytþ1 ¼ 1:04):

ð1  vÞyt þ vytþ1
b< ;
ð1  vÞyt1 þ vyt

which reduces to b < 0:977 and hence holds. For traders, 2.50 is satisfied since
b < yt =ytþ1 ¼ 0:962, and hence the CIA binds under both flexible and fixed ex-
change rates.

Velocity Shocks Suppose that yt ¼ ytþ1 ¼ y p . The velocity variable alternates

between two values: 0.20 and 0.22. Assume first that vt1 ¼ 0:2, vt ¼ 0:22, and
vtþ1 ¼ 0:2. Then, for nontraders under flexible rates, it must be the case that

1  vt
b< ; ð2:51Þ
1  vtþ1

which holds since b < 0:975. Under fixed rates, it must the case that

1  vt þ vtþ1
b< ; ð2:52Þ
1  vt1 þ vt

which holds—since under the most restrictive case in which vt1 ¼ 0:2, vt ¼ 0:22,
and vtþ1 ¼ 0:2, then b < 0:961.
For traders, the cash-in-advance always holds.
PT
2.5.2 Proof that tF1 b t (vt C vt1 ) F 0 if v0 F v p
Py
Rewrite t¼1 b t ðvt  vt1 Þ as

X
y X
y
b t ðvt  vt1 Þ ¼ bv0 þ ð1  bÞ b t vt :
t¼1 t¼1
38 Amartya Lahiri, Rajesh Singh, and Carlos A. Végh

But, by definition of v p and given that v0 ¼ v p ,

X
y
vp b
b t vt ¼  vp ¼ v p:
t¼1
1b 1b

Hence,

X
y
b t ðvt  vt1 Þ ¼ bðv p  v0 Þ ¼ 0
t¼1

as v0 ¼ v p .
PT
2.5.3 Proof that tF1 b t ( yt1 C yt ) F 0 if y0 F y p

Replace v with y in section 5.2.

Notes

This chapter was originally prepared for the conference in honor of Guillermo Calvo, held at the Inter-
national Monetary Fund in April 2004. We are grateful to Martin Eichenbaum and conference partici-
pants for helpful comments and suggestions.
1. Mulligan and Sala-i-Martin (2000) report that even for the United States, 59 percent of the popula-
tion (as of 1989) did not hold interest-bearing assets. One would conjecture that this figure is even
higher for developing countries.
2. It is worth noting that our results are in the spirit of an older literature that focused on the pros and
cons of alternative exchange rate regimes in models with no capital mobility (see, for instance, Fischer
(1977) and Lipschitz (1978)). See also Ching and Devereux (2003) for a related analysis in the context of
optimal currency areas.
3. Asset market segmentation could be endogenized by assuming that there is a fixed cost of accessing
asset markets. With idiosyncratic fluctuations in endowment, the number of agents that choose to gain
access to asset markets would be endogenously determined.
4. Appendix 2.5.1 derives sufficient conditions for the cash-in-advance constraint to bind. Contrary to
what our intuition would first tell us—that the cash-in-advance constraint would rarely bind because
nontraders would like to save some money for low endowment periods—the cash-in-advance may
bind under very weak conditions because unspent money balances have an opportunity cost that is
positively related to the state of the economy (that is, the opportunity cost is higher in good times). In
good times, therefore, nontraders would like to save for consumption smoothing motives but dissave
for financial reasons.
5. Given the open economy nature of the model, the private sector as a whole must always be able to
exchange money for foreign bonds (and vice versa) in the asset market (even under flexible rates), and
bonds for goods (and vice versa) in the goods market. One can imagine a trading agency that is in
charge of such activities or, alternatively, that the household has a third member, a foreign trader,
whose job is to put aside some of the household’s money or bonds during the asset market and trans-
act with foreigners during the goods market.
Optimal Exchange Rate Regimes 39

6. Appendix 2.5.1 derives the restrictions needed to ensure a positive nominal interest rate.
7. This assumption just ensures that the present discounted value of income is identical across traders
and nontraders when the money supply or the exchange rate is fixed.

8. We will consider only the extreme cases of a constant money supply (under flexible rates) and a
fixed exchange rate (as opposed to time-varying paths of the exchange rate). For an extension of our
main results to more general rules involving a fixed rate of growth of either the money supply or the
exchange rate, see Lahiri, Singh, and Végh (2006).
9. In a stochastic version of the model, the equivalent assumption would be that velocity shocks are
white noise.

References

Alvarez, Fernando, Robert Lucas, Jr., and Warren Weber. 2001. ‘‘Interest Rates and Inflation.’’ American
Economic Review 91: 219–225.
Calvo, Guillermo. 1999. ‘‘Fixed versus Flexible Exchange Rates: Preliminaries of a Turn-of-Millenniun
Rematch.’’ Mimeo, University of Maryland.
Cespedes, Luis, Roberto Chang, and Andres Velasco. 2004. ‘‘Balance Sheets and Exchange Rate Policy.’’
American Economic Review 94: 1183–1193.
Ching, Stephen, and Michael B. Devereux. 2003. ‘‘Mundell Revisited: A Simple Approach to the Costs
and Benefits of a Single Currency Area.’’ Review of International Economics 11: 674–691.
Fischer, Stanley. 1977. ‘‘Stability and Exchange Rate Systems in a Monetarist Model of the Balance of
Payments.’’ In The Political Economy of Monetary Reforms, ed. Robert Z. Aliber, 59–73.
Fleming, J. Marcus. 1962. ‘‘Domestic Financial Policies Under Fixed and Flexible Exchange Rates.’’ IMF
Staff Papers 9: 369–379.
Lahiri, Amartya, Rajesh Singh, and Carlos A. Végh. 2006. ‘‘Optimal Monetary Policy under Asset Mar-
ket Segmentation.’’ Mimeo, Iowa State University.

———. 2007. ‘‘Segmented Asset Markets and Optimal Exchange Rate Regimes.’’ Journal of International
Economics 72: 1–21.

Lipschitz, Leslie. 1978. ‘‘Exchange Rate Policies for Developing Countries: Some Simple Arguments for
Intervention.’’ IMF Staff Papers 25: 650–675.
Lucas, Robert E., Jr. 1982. ‘‘Interest Rates and Currency Prices in a Two-Country World.’’ Journal of
Monetary Economics 10: 335–359.
Mulligan, Casey, and Xavier Sala-i-Martin. 2000. ‘‘Extensive Margins and the Demand for Money at
Low Interest Rates.’’ Journal of Political Economy 5: 961–991.
Mundell, Robert A. 1968. International Economics. New York: MacMillan.
3 Monetary Policy Rules, the
Fiscal Theory of the Price
Level, and (Almost) All that
Jazz: In Quest of Simplicity

Leonardo Auernheimer

Guillermo Calvo’s work has invariably been distinguished not only by deep rele-
vance, but by elegant simpleness. Since imitation (successful or not) is the sincer-
est form of admiration, it seems fitting for this paper to reexamine a group of
closely related questions in the same spirit of simplicity. Readers of Calvo’s work
will also recognize the connection between some of the points touched on in this
paper and some of his contributions—most notably, the relevance of the govern-
ment budget constraint (Calvo 1985), the concern with levels versus rates of
change (Calvo 1981), and the study of interest rate pegging policies (Calvo and
Végh 1990).
The central question I address is the so-called fiscal theory of the price level
(FTPL), and some closely related topics, such as the pegging of the nominal
interest rate and the existence not only of an equilibrium, but of mechanisms
that under different assumptions will or will not allow the attainment of such
equilibrium.
The relevant literature during the last fifteen years or so has been both vast and
diverse, with the ‘‘eye of the storm’’ being the proposition, first advanced in its
most stark form by Woodford (1995), that under certain conditions, and for a
given level of nominal government debt (bonds cum fiat money), the path of the
price level is determined exclusively by fiscal policy (in other words, the antici-
pated path of the government primary surplus and of government debt). Indis-
pensable references in this discussion are, among others, Sims (1994, 1999),
Cochrane (2001, 2002), Kocherlakota and Phelan (1999), McCallum (1997, 2001),
Christiano and Fitzgerald (2000), Buiter (2002, 2005), Leeper (1991), Bassetto
(2002), Niepelt (2004), Dupor (2000), and Daniel (2001) for the case of the open
economy and, of course, Woodford (especially 1994, 1995, and 1998). Although
the natural benchmark for the beginning of the discussion is Woodford (1995),
there are previous pieces anticipating some of the same propositions (Begg and
Haque 1984, Auernheimer and Contreras 1990, 1995). As is usually the case, there
42 Leonardo Auernheimer

are also classical works touching on some of the very basic concepts—Metzler
(1951) being, of course, the primary reference, as well as Tobin (1974) and, more
recently, the already classic Sargent and Wallace’s ‘‘unpleasant monetarist arith-
metic’’ (1981). This paper is certainly not a survey, but the revisitation of the ques-
tion along the lines of a very simple framework first presented in Auernheimer
and Contreras (1990), and the evaluation of some of the propositions for and
against the basic FTPL claims. Even a casual reader will notice that the literature
is extensive, in some instances a bit unfocused, and at times full of sound and
fury.1 Rather than addressing the various propositions and counterpropositions,
we will proceed with the analysis of the simple model, derive some results, and
indicate, in each case, the extent to which these results are or are not in agreement
with those advanced by the various contenders.
We consider an extremely simple model (and will not hesitate to use graphical
representations when helpful), which is rather standard in the literature and, in
fact, basically coincides with models used in most of the discussion. At variance
with most if not all the literature on the subject, we use continuous time, for two
related reasons.2 First, the analysis is simpler and less cluttered. Second, ambigu-
ities and potential confusion among dates and beginning- or end-of-the-period
questions are removed, and the distinction between levels and rates of change is
sharpened—this last consideration being particularly important in this case.3
There is still another, more compelling reason for our choice of continuous time
analysis, which will become clear as we proceed. In the class of problems we dis-
cuss, the fundamental questions are about initial conditions at time t ¼ 0: in the
absence of anticipated future changes in policy or in behavioral parameters, the
whole path of all variables, including of course the price level, is determined at
that initial time. More often than not, and depending on the pre-initial condi-
tions at time t ¼ 0
, those initial conditions require discrete changes in some
variables—the most important one in our case, although not the only one, being
the price level. There is quite a lot of economic analysis and market mechanisms
that can be discussed concerning the realization of those changes, and in continu-
ous time there is a very sharp distinction between the mechanics of those discrete
changes happening at once (as if time could be ‘‘stopped’’ at t ¼ 0
) and the ensu-
ing evolution of the variables once time is ‘‘reinitiated’’ at t ¼ 0þ .
In section 1 we present the bare-bones simple model, and elaborate on some ba-
sic points to be addressed. Section 2 considers the case of a ‘‘monetary rule,’’ in
which the government sets the path of the money supply, and section 3 compares
the model’s conclusions to some of the propositions in the literature. Section 4 dis-
cusses the case of a central bank pegging the nominal interest rate, and section 5
concludes.
Monetary Policy Rules, the Fiscal Theory of the Price Level 43

3.1 The Model

Individuals are identical, infinitely lived, and at each point in time maximize the
present value of their lifetime utility, which depends on their consumption and
the services of their real money stock. We assume the utility function to be separa-
ble in the two arguments. Time is continuous, there is perfect foresight, and prices
are perfectly flexible. Then, at each initial point t ¼ 0, the typical individual
maximizes
ðy
UðcðtÞ; mðtÞÞe
rt dt ð3:1Þ
0

where r is the positive, constant rate of time preference, and c and m are the flow
of consumption and the stock of real cash balances, respectively.
Individuals receive a constant flow endowment of an instantaneously perish-
able single commodity, which we call ‘‘real income,’’ and hold two assets: fiat
money, which yields no interest, and government bonds. The exact characteriza-
tion of these government bonds is crucial for the questions at hand, and requires
some elaboration.4 Here, we define them as ‘‘call bonds,’’ paying the prevailing
nominal interest rate but with instantaneous maturity: in other words, very much
as sight deposits. As deposits, they are denominated in terms of the monetary unit
(say, one dollar), and changes in the nominal interest rate carry no capital gains or
losses. This is, in its discrete-time equivalent form of ‘‘one-period bonds,’’ the as-
sumption used in much of the literature.5 The stock of nominal money is denoted
by M, with P being the money price of the single good (the price level), so
m 1 M=P is the real money stock.
The typical individual’s flow budget constraint expressed in nominal terms, an
identity at each point time t, is then

yPðtÞ þ BðtÞiðtÞ 1 cðtÞPðtÞ þ vðtÞPðtÞ þ dBðtÞ=dt þ dMðtÞ=dt

where y is the constant level of real income, B the level of government bonds, i the
nominal interest rate, and v the flow of a real head tax (subsidy, if negative).
Defining b 1 B=P as the real stock of government bonds and p as the rate of infla-
tion, the constraint can be expressed in real terms as
y þ bðtÞr ¼ cðtÞ þ vðtÞ þ dbðtÞ=dt þ dmðtÞ=dt þ mðtÞpðtÞ: ð3:2Þ

Notice that in writing 3.2 we have used the assumption that the equilibrium
Fisher condition,
iðtÞ ¼ r þ pðtÞ; ð3:3Þ
44 Leonardo Auernheimer

where r is the (constant) real interest rate, holds at all times. Because 3.3 is an
equilibrium condition, then 3.2 is no longer a strict identity, but an equality that
will hold at all times.
The government (the fiscal authority cum the central bank) prints fiat money at
zero cost, borrows or lends via the sale or purchase of bonds, imposes neutral
head taxes (transfer, if negative), and spends on goods which are thrown into the
ocean—a real flow denoted by g. The flow government budget constraint in nom-
inal terms, again an identity, is then

dB=dt 1 BðtÞi þ gðtÞPðtÞ
vðtÞPðtÞ
dM=dt:

Using 3.3 once more, the constraint can be written in real terms as

db=dt ¼ bðtÞr þ gðtÞ
vðtÞ
ðdm=dtÞ
mðtÞpðtÞ

¼ bðtÞr
sðtÞ
ðdm=dtÞ
mðtÞpðtÞ ð3:4Þ

where sðtÞ ¼ vðtÞ
gðtÞ is the government primary surplus (deficit when nega-
tive). This is again an equality that will hold at all times, as in the case of 3.2.
Addition of the two budget constraints yields the resource constraint

y ¼ cðtÞ þ gðtÞ: ð3:5Þ

In what follows we assume that government expenditures remain constant, so
that any changes in the primary surplus s take place via a change in the tax (sub-
sidy). What this means is that the level of consumption will also remain constant.
Maximization of 3.1 by the typical individual, subject to its budget constraint
3.2, yields the familiar Euler necessary conditions for optimality:

dc=dt ¼ ðUc =Ucc Þðr
rÞ ð3:6Þ

and

dc=dt ¼ ðUc =Ucc Þðr þ pÞ
ðUm =Ucc Þ: ð3:7Þ

Expressions 3.6 and 3.7 imply

ðUm =½r þ pÞ ¼ Uc ; ð3:8Þ

and the condition r ¼ r, in turn, implies that

dc=dt ¼ 0; ð3:9Þ

that is, consumption is piecewise constant.6 Note, again, that given the assump-
tion of constant levels of government expenditures and real income, then con-
sumption becomes a constant parameter that, being piecewise constant, can
Monetary Policy Rules, the Fiscal Theory of the Price Level 45

change a finite number of times, as any parameter can—in this case, for example,
as a result of exogenous changes in the endowment.
In addition, optimal plans are to satisfy the transversality conditions7

lim bðtÞe
rt ¼ 0 as t ! y
and

lim mðtÞe
rt ¼ 0 as t ! y:
We are not constraining government debt to be nonnegative—that is, we are
allowing b < 0, so that government can lend to the public and hold positive net
assets.8 We are also ruling out mðtÞ < 0.
Notice next that expression 3.8 is the demand for money. Since in all of what
follows we will specify a constant level of real income and of government expen-
ditures (with whatever changes take place in the primary surplus assumed to be
via a change in head taxes), then consumption will be constant. For a real interest
rate that is then also constant, for convenience we can write expression 3.8 as

m ¼ LðiÞ ¼ Lðr þ pÞ L 0 < 0: ð3:8 0 Þ

Next, and most importantly, ‘‘helicopter drops’’ (or ‘‘annihilations’’ à la Hume)
of fiat money as part of a deliberate policy rule are precluded.9 In particular, we as-
sume that any discrete change in the stock of nominal money takes place via an
equal change in the stock of nominal bonds, with the opposite sign—in other
words, at any point in time t
DMjt 1
DBjt ð3:10Þ

This is, of course, an assumption at the heart of both the fiscal theory and the
correct specification of a constant nominal interest rate pegging. Assumption 3.10
is the equivalent of specifying that all changes in the nominal money stock need to
be accounted for in the government budget constraint’s other asset—the stock of
bonds. Still another assumption will be that the system operates in a range at
which

jðqLðÞ=qpÞðp=LðÞÞj < 1; ð3:10 0 Þ

that is, an increase in the inflation rate will increase the inflation tax—an assump-
tion that not only conforms to empirical estimates (except for some very dramatic,
short-lived hyperinflationary episodes) but that is necessary to make sense of the
analysis.
46 Leonardo Auernheimer

Omitting the time arguments in order to avoid clutter and defining p ¼

L
1 ðmÞ
r ¼ lðmÞ, where L
1 is the inverse of function 3.8 0 , the simple model
can be summarized in the two expressions
db=dt ¼ br
s
mm ð3:11Þ

or its equivalent
db=dt ¼ br
s
dm=dt
mp; ð3:11 0 Þ

and
dm=dt ¼ mðm
lðmÞÞ ð3:12Þ

where m 1 ðdM=dtÞð1=MÞ is the rate of nominal money growth. Notice that, in this
construction, 3.8 holds at all times at which the system is engaged in the optimal
paths implied by the first order conditions.10
We will now consider the first of two cases, in the context of the model, of a
well-defined policy rule—a monetary rule (the second being an interest rate rule,
to be considered later). In every case we will ask the following three questions: (1)
Is there a unique path of the price level satisfying the transversality conditions? (2)
If there is such a unique path, is it accessible for any initial conditions? (3) If the
answer to the two previous questions is in the affirmative, is there a market mech-
anism so that agents are induced to reach such a unique path? This third question
is directly related to the general point raised by Lucas (1978) in a different context,
but very relevant here nonetheless:
One would feel more comfortable . . . with rational expectations equilibria if these equilibria
were accompanied by some form of ‘‘stability theory’’ which illuminated the forces which
move an economy toward equilibrium. (1429)

We will refer to this as the existence, or lack of existence, of a market-induced

stability mechanism, and our answers to the question will be more on the side of
conjectures than on the side of proofs. Finally, notice that in this model, for an ex-
ogenous level of the government surplus and for the control of nominal aggre-
gates, a unique path of the price level is sufficient for determining a unique path
of the level of real government debt and of the real money stock.

3.2 A Monetary Rule

In the system 3.11–3.12, assume that government fixes a preannounced path of

the money supply. For simplicity, we assume that such path implies a constant
rate of monetary growth; that is,

MðtÞ ¼ M0 expðmtÞ tb0 ð3:12 0 Þ

Monetary Policy Rules, the Fiscal Theory of the Price Level 47

Figure 3.1

In addition, we also assume that the anticipated and effective fiscal policy is to
keep constant the level of the primary budget surplus, s. This is, of course, a typi-
cal non-Ricardian policy, as is defined in the literature on the FTPL.11
The system in expressions 3.11 and 3.12 can then be represented in the phase di-
agram in figure 3.1, where
b ¼ ðs þ mmÞ=r ð3:13Þ

lðmÞ ¼ m ð3:14Þ
are the values of b and m satisfying db=dt ¼ 0 and dm=dt ¼ 0, respectively, with b 
and m  being the long-run steady values satisfying both 3.13 and 3.14. Notice
that, of course, 3.14 is not the equilibrium ‘‘money supply equal to money de-
mand’’ (which is given by 3.8 0 and implied in 3.14), but the condition for the rates
of inflation and monetary growth to be equal.
Since at any point in time the stocks of both nominal bonds ðBÞ and nominal
money ðMÞ are given, a discrete increase (decrease) in the price level will decrease
(increase) the levels of both real debt and real money balances by the same propor-
tion.12 In terms of the graph in figure 3.1, this can be represented by movements
along the ray passing through the initial pair b, m. Note, though, that since the
price level is not a state variable, an initial point is strictly defined as a pair Bð0Þ,
Mð0Þ—the nominal stock of bonds and money—so that in this graphical repre-
sentation an initial position is characterized as any point along a ray with a slope
Bð0Þ=Mð0Þ. Notice also that the graph in figure 3.1 is drawn for the more general
case of a positive rate of monetary expansion; in other words, m > 0. The special
but theoretically important case in which the nominal money stock is constant—
m ¼ 0—is depicted in the graph of figure 3.2.
48 Leonardo Auernheimer

Figure 3.2

Consider first the case m > 0. The answer to our first question (is there a unique
path of the price level satisfying the transversality conditions?) is in the affirma-
tive. This is the path mðtÞ ¼ m  and bðtÞ ¼ b  , and the path of prices implied by
3.12 0 . The answer to the second question (if there is a unique such path, is it acces-
sible for any initial conditions?) is in the negative. An initial ratio Bð0Þ=Mð0Þ ¼
b  =m  is a necessary condition for reaching the unique equilibrium m  , b  , via a
discrete change in the price level. For any other initial nominal levels of debt and
money there is no path of the price level satisfying the transversality and the non-
negative money conditions. Notice, nevertheless, that such a general conclusion
would not hold if a nonnegative restriction is imposed on nominal debt (that is, if
government is not allowed to lend and hold positive assets). In this case, any ini-
tial ratio of nominal debt and nominal money could result in paths of the price
level for which both real debt and real money would approach zero.
The third, and most interesting, question is whether there is a market mecha-
nism that, if the necessary initial condition Bð0Þ=Mð0Þ ¼ b  =m  is satisfied, will
generate a price level P  such that Bð0Þ=P  ¼ b  and Mð0Þ=P  ¼ m  . The answer
to this question is given here as the sketch of a conjecture.
Consider a pre-initial case at which the price level is P  , and given the initial
levels of nominal money and bonds the system rests at its steady state equilibrium
b  , m  , with the inflation rate equal to the rate of monetary growth, p ¼ m. Per-
form now the usual experiment of assuming that all of a sudden, for no other rea-
son than a mysterious uniform marking-up of prices, the price level increases to
P1 > P  . The usual story for how an instantaneous equilibrium is restored is
(implicitly assuming that chronological time is stopped) that the rise in prices has
decreased real cash balances, with the result that now, at the same inflation rate,
agents have ‘‘too little money’’—m1 < Lðr þ mÞ—with the stock of real bonds
Monetary Policy Rules, the Fiscal Theory of the Price Level 49

having also fallen to b1 < b  .13 They will then decrease their spending, but since
aggregate output is fixed, successive rounds of ‘‘instantaneous’’ price adjustments
(decreases) will take place, with the price level ending up at the original value P  ,
with the economy back at the initial equilibrium. This story is the usual one in
models where money is the only asset. What about our case, in which agents also
hold government bonds? Wouldn’t individuals, in this instantaneous process, also
seek to perform a portfolio adjustment, exchanging bonds for money? There are
two problems with this version of the story. First, it is true that individuals will
be able to withdraw their call bonds (i.e., not to continue part or all of their roll-
over), but it is also true that if government is committed to the preannounced
path of the nominal money supply, it will immediately restore its previous
level—it would take only an infinitely small increase in the nominal interest rate
to persuade depositors to keep the initial level of those call bonds. The second
problem is that an exchange of bonds for money will not restore the previous
equilibrium. In this conjecture, it seems that the ‘‘monetary story’’ of intended
changes in expenditures is rather persuasive.14
The alternative to the ‘‘instantaneous’’ adjustment process outlined above is, of
course, that at the point b1 , m1 the inflation rate would increase to a level p1 > m
for which m1 ¼ Lðr þ p1 Þ, and for which the system would engage in the dynam-
ics of a path violating the transversality conditions—a path not only formally in-
admissible but one which makes little economic sense.
For initial values outside the ray Bð0Þ=Mð0Þ ¼ b  =m  there is no price level that
would allow the system to attain the unique steady-state equilibrium. A particular
instance of this proposition is the Sargent-Wallace case of an ‘‘unpleasant mone-
tarist arithmetic’’ (Sargent and Wallace 1981) in which, starting from an initial
equilibrium m0 , b0 , government lowers the rate of monetary expansion, keeping
the same level of the government primary surplus.15 It is easy to show that, under
assumption 3.10 0 , a fall in the rate of monetary growth will result in an equilib-
rium at m  > m0 , b  < b0 , and to show (and to verify by mere inspection from the
graph in figure 3.3, in which A is the initial equilibrium, and point A0 the new
steady state after the change in policy) that from the original equilibrium at m0 ,
b0 , there is no possible change in the price level that would generate an immediate
adjustment to the new unique equilibrium. Two things can then happen: there is
no anticipated further change in either fiscal policy (the primary surplus) or mon-
etary policy (the rate of monetary expansion), or there is the generalized anticipa-
tion that at a future time some of those changes will occur so that a new long-run
equilibrium is achieved. In the first case a long-run feasible equilibrium will simply
not exist. In the second case, a future corrective policy change is anticipated—the
possibility envisioned by Sargent and Wallace (1981) being an increase in the rate
of money growth. Either this correction or an appropriate rise in the fiscal surplus
50 Leonardo Auernheimer

Figure 3.3

would generate a unique and feasible adjustment, ultimately leading to a unique

steady-state equilibrium.
Another possibility is, of course, a temporary violation of the monetary rule,
with a once-and-for-all increase in the nominal money stock, leading to a corre-
sponding once-and-for-all increase in the price level—a temporary, short-lived
hyperinflation. It is easy to show (even in simple graphical terms in the context of
figure 3.3), that this can be accomplished either via a ‘‘rain of money’’ (a tempo-
rary violation of assumption 3.10) or a purchase of bonds. These possibilities are
depicted in figure 3.3, where a one-time gift or a rain of money results in a move-
ment from point A to point D, and a purchase of bonds in a movement from A to
C, with a one-time increase in prices then allowing to access the new long-run
equilibrium at point A0 . Notice, parenthetically, that there are two distinct effects
of the increase in the money stock, not always clearly distinguished in the litera-
ture. One is the obvious ‘‘inflating away’’ part of real debt via a fall in the real
value of bonds; the other, the purchase of interest bearing bonds in exchange for
non-interest-bearing money. That is why a rain of money requires a higher in-
crease in the nominal money stock, and a higher one-time increase in prices, than
the purchase of bonds—in the first case, only the first of these effects is at work. If
debt is denominated in real terms, as in Sargent and Wallace’s analysis, then the
required increase in the money stock via purchase of bonds can also do the trick,
with the highest necessary increase in money and the highest increase in prices
(point B in figure 3.3); here, a rise in prices will not decrease real debt, so that the
entire adjustment needs to take place via the second effect, and a rain of money
will not work at all. This temporary hyperinflation ‘‘solution’’ is indeed the man-
ner in which some governments have at times tried to deal with the ‘‘unpleasant
arithmetic’’ problem.16
Monetary Policy Rules, the Fiscal Theory of the Price Level 51

Another, less spectacular, possibility is for government to implement a ‘‘soft’’

monetary rule, in which the rate of growth of the money supply is fixed, but in
which discrete, discretional changes in nominal money are part of the policy.
With the existence of call bonds, it is even possible, if there is a market-induced
stability mechanism, for the public to take the initiative via changes in those
deposits, with government passively allowing for those endogenous changes. We
do not pursue this topic here (although we will refer to it again in the concluding
section 5), except to note first that such a scenario would be quite similar (in fact,
observationally equivalent) to a nominal interest rate pegging or targeting. This is
not surprising: as we will indicate later, a nominal interest rate pegging deter-
mines, at each point in time, the rate of change (but not the level) of prices, and
the possibility alluded to in this paragraph determines at each point in time the
rate of change of nominal money, but not its level.
Consider now the particular case m ¼ 0, a constant nominal money stock, which
is the one considered in much of the literature on the FTPL. Now the two system
equations 3.13–3.14 become
b ¼ s=r ð3:13 0 Þ

lðmÞ ¼ m; ð3:14Þ
which are shown in the graph of figure 3.2. This is a case that, as hinted before, is
of especial theoretical interest, and has been analyzed first by Obstfeld and Rogoff
(1983) in the pure monetary context, and in the context of the fiscal theory of the
price level by, among others, Carlstrom and Fuerst (2000), Christiano and Fitzger-
ald (2000), Kocherlakota and Phelan (1999), and McCallum (2001). Notice that in
this case the link between equations 3.11 and 3.12 would seem to be broken, since
changes in the real money stock brought about by changes in the price level do
not affect the seigniorage component of the government budget constraint, simply
because there is no inflationary finance. But the link still exists for nominal debt,
with discrete changes in the price level affecting real debt, and hence the govern-
ment’s fiscal position.
In this case, the answer to our first question (is there a unique path of the price
level satisfying the transversality conditions?) is in the negative: there is a well-
defined pair of values m  , b  such that bðtÞ ¼ b  and mðtÞ ¼ m  for all t, but there
is also a set of infinite possible paths of the price level (and the inflation rate)
which originate for any initial pair of values bð0Þ ¼ b  , mð0Þ < m  , along which
the transversality conditions are met and the real value of money converges to
zero. The result is the same as in the FTPL literature cited above, but with a ca-
veat. Note that, starting from the initial steady state equilibrium b  , m  , since we
have nominal debt, such a point bð0Þ ¼ b  , mð0Þ < m  cannot be generated by the
52 Leonardo Auernheimer

usual ‘‘inflationary bad dream’’ story, which would result in an immediate in-
crease in the price level and therefore in a point bð0Þ < b  , mð0Þ < m  for which
the transversality conditions would not hold. The condition for hyperinflation,
with bð0Þ ¼ b  , mð0Þ < m  , would result, even in the case of the dream, if debt is
in real terms or, for either nominal or real debt, if a sudden ‘‘annihilation’’ of
money would take place.

3.3 The Fiscal Theory of the Price Level in the Case of a Monetary Rule

We proceed now to compare our results for the case of a monetary rule with some
of the tenets of the FTPL and some of the propositions of its critics. As indi-
cated above, we have performed our analysis in what can be considered a non-
Ricardian policy par excellence, with a fixed constant primary budget surplus and
a fixed path of the money supply—a non-Ricardian policy being, of course, the
case where, according to the FTPL, the theory becomes relevant as an alternative
to the monetary explanation.
Before proceeding, some general comments are in order: first, that in terms of
our simple model, the basic proposition of the FTPL is that the price level is deter-
mined by equation 3.11, or, more specifically, by equation 3.13—that is, equation
3.11 when real debt is not changing, and independent of any monetary considera-
tions. In turn, expression 3.8 0 , the demand for money, is considered to be more
useful when viewed, for an exogenous monetary policy, as determining the inter-
est rate—in other words, the rate of inflation (Woodford 1995, 12, 18).
Second, that a considerable part of the discussion on the FTPL has focused on
whether expression 3.11, or, more specifically, its equilibrium version 3.13, is a
budget constraint or an equilibrium condition. In our view, much of this discus-
sion has been less than illuminating, and has quickly reached the stage of decreas-
ing (or maybe even negative) returns.17 A reference to the theological discussion
of how many angels can dance on the head of a pin is irresistible. We take the
view that 3.11 is a flow budget constraint (and, more specifically, resulting from
an identity), while 3.13 is one of the necessary equilibrium conditions, together
with 3.14.
Finally, we should note that a consistent feature of the FTPL is the treatment of
nominal money and nominal bonds as an aggregate (‘‘total nominal debt’’), and
although there is a careful specification of bonds being interest-bearing debt and
money being, in general, non-interest-bearing debt, we find the practice slightly
misleading. In fact, the neatest result of the FTPL is under the assumption that
money effectively becomes an interest-bearing obligation, via a policy of returning
to the private sector the revenues from money creation—a policy that we analyze
below. Closely related to this point is the definition of an ‘‘adjusted fiscal surplus’’
Monetary Policy Rules, the Fiscal Theory of the Price Level 53

as the usual pure primary surplus plus the revenues from money. Let us consider
now some of our conclusions and their correspondence with some of the proposi-
tions in this literature.
We discuss first, because it is the most straightforward, the case of the hyperin-
flation solution discussed by Carlstrom and Fuerst (2000), Christiano and Fitzger-
ald (2000), Kocherlakota and Phelan (1999), and McCallum (2001), resulting from
an ‘‘annihilation,’’ or a one-time confiscation of part of the nominal money stock.
For the more general case in which m > 0, as we have shown above, such a hyper-
inflationary solution does not satisfy the transversality conditions if we assume
the possibility of negative government debt, as Kocherlakota and Phelan (1999)
do. As shown by Obstfeld and Rogoff (1983), such a solution is indeed an equilib-
rium in a world without nominal government debt, as it would be in our model,
with indexed government debt (i.e., specified in real terms). The criticism of the
FTPL as implying such a hyperinflationary alternative solution is then rather curi-
ous when in fact it is the presence of the fiscal element, with nominal government
debt, that allows us to rule out hyperinflation as an equilibrium when m > 0. In
the case in which m ¼ 0, considered in most of the literature, hyperinflation is
indeed an equilibrium, but we note that the monetarist alternative will not yield
a unique non-hyperinflationary equilibrium either in the presence of a non-
Ricardian policy of an exogenous money stock and a fixed surplus.18 For the mon-
etarist adjustment to work, what is needed is an annihilation of all nominal assets
(money and bonds) by the same proportion. Whether the adjustment back to the
steady-state equilibrium takes place along the lines sketched by Woodford (1995,
12), in which the adjustment is motivated by an increase in ‘‘total government
debt,’’ or via a discrepancy between the demand for money and the actual money
stock (as described in the previous section, or as in Kocherlakota and Phelan 1999)
is almost immaterial.
A tenet of the FTPL is that the price level adjusts to assure db=dt ¼ 0, or br ¼
s þ mm, meaning the price level is determined by expression 3.13—or by 3.13 0 ,
with m ¼ 0, in which case the above equation reduces to br ¼ s. Interpreted in this
form, it would seem clear from the analysis that this would not be the case—such
an adjustment, if it takes place, will in general result in paths of the price level that
will not satisfy the transversality conditions. The only exception is, of course, the
case in which, in terms of our graphical interpretation, the initial ratio of nominal
bonds to nominal money happens to be such that Bð0Þ=Mð0Þ ¼ b  =m  .
The FTPL is often cast in an apparently slightly, but fundamentally different,
form. For the case in which p ¼ m, expression 3.13, the government’s flow budget
constraint that needs to hold for db=dt ¼ 0 can be written as
b ¼ ðs þ mpÞ=r;
54 Leonardo Auernheimer

so that

B=P ¼ ðs þ ðM=PÞði
rÞÞ=r ¼ ðs þ ðM=PÞðiÞ
ðM=PÞðrÞÞ=r

and
ðB þ MÞ=P ¼ ðs þ miÞ=r ð3:15Þ

where ðs þ miÞ is the ‘‘adjusted fiscal surplus,’’ with the term mi rather than the
usual ‘‘inflation tax’’ term mp, to account for the ‘‘interest savings on the govern-
ment’s monetary liabilities’’ (Woodford 1995, 10). In expression 3.15 it is clear
that the price level is uniquely determined by the total of government liabilities
ðB þ MÞ and the adjusted fiscal surplus as posited by the FTPL. But notice that
equation 3.15 is nothing else than the final reduced form of our overall system 3.11–
3.12, since it assumes not only db=dt ¼ 0 but also p ¼ m; that is, dm=dt ¼ 0. It is
then not surprising for the price level to depend on these fiscal magnitudes. It is
trivial that in a system in which there is more than one nominal asset, a corre-
spondence needs to exist between the total of those nominal assets, or an arbitrary
subset of them, and the price level for the economy needs to exhibit neutrality in
the presence of perfect price flexibility. Another problem, of course, is with the
definition of the fiscal surplus as being ‘‘adjusted’’ for the revenues from money
creation; as it is clear from expression 3.15, for a given unadjusted primary fiscal
deficit s, the nominal interest rate (and hence the inflation rate) will depend on
the exogenous rate of money growth. As will be mentioned later, an interest rate
pegging policy, by fixing the term mi, eliminates this objection, and goes a long
way to explain why the FTPL appears as ‘‘especially useful’’ in such a regime.
(Woodford 1995, 4).
Finally, consider the following argument by one of the main proponents of the
FTPL, which in our view is more substantial and goes to the core of the theory.

Furthermore, it is even arguable that the expected path of the money supply does not mat-
ter for price-level determination, except through its consequences for the government’s budget.
(Woodford 1995, 15; his emphasis)

One argument against this proposition could be the following: consider an

economy without government finances, in which the only role of government is
to print fiat money, which is distributed via head subsidies and increases over
time at a fixed proportional rate (which of course could be zero). Will a one-time
change (expected or unexpected) in either the stock or its rate of change have con-
sequences for the price level? The answer is of course in the affirmative. But such
an argument would be misplaced, amounting to throwing away the baby with
the bath water—proponents of the FTPL would immediately reply that in such
an economy all possible policies are Ricardian, and their point would be well
Monetary Policy Rules, the Fiscal Theory of the Price Level 55

taken. What is then that makes money to appear to become so unimportant once
nominal government debt is introduced?
In our view, more than a question of whether an expression such as equation
3.13 determines the price level, the answer rests on the fact that in the presence of
nominal government debt, all monetary actions have a fiscal component and a fiscal
effect, via the effects of changes in prices on real debt, unless those effects are neu-
tralized via changes in the purely fiscal magnitudes, such as the debt or the pri-
mary fiscal surplus (what would, of course, amount to a Ricardian policy for
which the monetarist adjustment becomes relevant). This point can be still better
understood if we consider Woodford’s discussion of a regime which would ‘‘neu-
tralize the fiscal effects of a change in the money supply’’ (1995, 15). Consider, in
terms of the notation of our model, a regime in which the ‘‘revenues from money
creation,’’ defined in Woodford as mi, were returned to the public—of course, on
a per head basis, unrelated to money holdings. In this case monetary changes
have no fiscal effects, and the relevant equations for the government’s flow bud-
get constraint can be written as
db=dt ¼ br þ mi
s
dm=dt
mp

db=dt þ dm=dt ¼ br þ mr
s
and, calling w ¼ b þ m ‘‘total government debt,’’ then

dw=dt ¼ wr
s:
In this case it is obvious that all that matters for the determination of the price
level (and for the overall global adjustment, for that matter) is the nominal aggre-
gate W 1 B þ M. This is probably the purest exhibit of the FTPL, a case in which
exchanges of money for bonds and vice versa are of no consequence. But notice
first that it is derived, as it were, by a policy that in fact converts money into
bonds, and second, that since there is a well-defined demand for money, equation
3.14 is still relevant and needs to be satisfied. In fact, the system becomes identical
to the case analyzed in the previous section for the case m ¼ 0, even with the same
geometrical interpretation in figure 3.3, with the term v instead of b.

3.4 A Nominal Interest Rule

A policy of pegging the nominal interest rate to a predetermined path or single

value has a long history of controversy in the profession, and for quite a time had
a consistently bad reputation. In the seventies and early eighties, in the era of
monetarism with adaptive expectations, the argument was that such policy
results in explosive behavior. After the advent of rational expectations, the
56 Leonardo Auernheimer

argument was, and still is in many quarters, that it would result in price-level
indeterminacy: if the monetary authority pegs the nominal interest rate, then the
nominal money stock needs to react passively in response to changes in the price
level so as to keep constant the real money corresponding to the demand for
money, pinned down by the exogenous value of the rate—meaning that the cen-
tral bank would need to validate any price level.
Explicit consideration of the government budget constraint and the absence of
rains of money (that is, changes in the money stock unmatched by corresponding
changes of opposite sign on government assets, such as government debt) allows
the elimination of indeterminacy—though this is not an opinion shared by most
critics of the FTPL.
There are reasons why the validity of an interest rate pegging is usually dis-
cussed within the context of the FTPL, mainly because in both cases the deter-
minacy or indeterminacy of the price level is at stake, and in both cases the
government budget constraint, and in particular the connection among its compo-
nents, is at the center of the analysis. Yet this need not be necessarily the case.19
In this section we use the elementary framework of section 1 to analyze the
implications of what we can in general call an ‘‘interest rate rule.’’ We will first
present the mechanics involved, perform some experiments, and then discuss
some of the features and implications of a nominal interest rate pegging. We will
argue that not only prices are uniquely determined, but that the policy yields
some attractive features and plausible results, both in the current context and in
the context of an open economy.
At the outset, we need to stress a distinction that, although often made explicit
in some of the literature, is sometimes not kept in mind. The distinction is be-
tween a nominal interest rate ‘‘pegging’’ (or ‘‘fixing’’), and a policy of interest rate
‘‘targeting.’’ The two yield in general the same results, and are therefore observa-
tionally equivalent. But, even at the theoretical level (and perhaps still more so),
they imply very different types of policies. We characterize a pegging policy as
one in which the central bank is willing to lend and borrow, irrespective of quan-
tities, at a fixed nominal interest rate. The central bank is then completely passive,
and the actual purchase or sales of bonds (in our case, call bonds) are entirely left
to the initiative of the private sector, with the government simply supplying (or
demanding) whatever quantities the public demands (or supplies). The analogy
with the pegging or fixing of the exchange rate is immediate. A policy of targeting
a fixed, preannounced level of the interest rate, even when (as in our very simple
model) the targeting is always perfect, is characterized by an active central bank
policy of open market operations so as to keep the rate at the targeted level. The
reason why these observationally equivalent policies are quite different is that
there may not exist a market-induced stability mechanism inducing individuals,
Monetary Policy Rules, the Fiscal Theory of the Price Level 57

despite the fixed level of the interest rate, to take the initiative and engage in
exchanges between money and bonds that would lead to a unique rational expec-
tations equilibrium. In what follows we will simply refer to an interest rate rule as
encompassing both of these observationally equivalent cases.
Consider then the model presented and analyzed in the previous sections,
which results in the reduced form equations 3.11 0 and 3.12, which we reproduce
here for convenience:
db=dt ¼ br
s
dm=dt
mp ð3:11 0 Þ

dm=dt ¼ mðm
lðmÞÞ ð3:12Þ

where, as before, government debt is in nominal terms—more specifically, as call
bonds.
Suppose now that the central bank stands prepared to borrow and lend at a
fixed nominal interest rate i  , and that such commitment is believed by the public
and effectively implemented. Since the nominal interest rate is fixed, so is the real
money stock, say, at a level

m  ¼ Lði  Þ; ð3:16Þ
and, with the Fischer equation holding at all times, arbitrage will assure that

p  ¼ i 
r: ð3:16 0 Þ
Using these values in 3.11 0 and 3.12 yields then the extremely simple formulation

db=dt ¼ br
s
m  ði 
rÞ ð3:17Þ
and 3.16.
A graphical representation, provided in figure 3.4, is immediate and almost
embarrassingly simple. Note that although the real money stock will at all chro-
nological times be constant at the level m  (since during the instantaneous adjust-
ment process taking place at a date at which we stop chronological time in order
to follow the logical steps of the same tatonnement process we considered in sec-
tion 1), we find it useful to use the plane m, b to follow such a process, rather than
in the real line m ¼ m  . From 3.17, there is then only one level of the real govern-
ment debt at which db=dt ¼ 0, and this is

s þ m  ði 
rÞ
b ¼ ; ð3:17 0 Þ
r

with values of the real debt above this level resulting in rising levels of debt, and
vice versa.
58 Leonardo Auernheimer

Figure 3.4

It is obvious from 3.17 that in the absence of anticipated future changes in

parameters and in order to satisfy the transversality conditions the adjustment
needs to be instantaneous. This is reflected in the arrows showing the laws of mo-
tion in figure 3.4.
The answer to our first question (is there a unique equilibrium satisfying the
transversality conditions?) is in the affirmative. In the context of figure 3.4, for
example, this is point A, at b  , m  . Notice that, for any given level of the nominal
aggregate ðB þ MÞ (‘‘government total liabilities,’’ as consistently called by propo-
nents of the FTPL), there is a unique mix of the two components in the aggregate
(that is, a unique ratio B=M) and, given such a ratio, a unique level of prices, for
which the unique steady-state equilibrium is realized. The right B=M ratio and
the right price level are each of them necessary conditions, and, when holding
jointly, they become sufficient. This simple reasoning sheds light on the reason
why the nominal interest rate rule, when the government budget constraint is tak-
en seriously (in other words, considering the aggregate ðB þ MÞ as given at any
point in time), does not lead to price indeterminacy. At the same time, notice that,
as a formal matter, the sole specification of the aggregate ðB þ MÞ is not sufficient, per
se, for determining the equilibrium price level, despite claims by the FTPL. This is the
same problem that we mentioned while considering the case of the monetary rule;
that is, that the condition dm=dt ¼ 0 is consistently neglected by expositors of the
FTPL. In the case of the interest rate rule, it is clear that any aggregate ðB þ MÞ
will determine a unique level of prices for which db=dt ¼ 0 (in other words, equa-
tion 3.17 0 will hold), but in general this will not be the price level for which
dm=dt ¼ 0 will also be satisfied.20 What is necessary for both conditions to hold is
for the ratio of debt to money to be the right one, and an interest rate rule will
generate such a ratio via open market operations. This is so, of course, for the fun-
damental reason that money is a non-interest-bearing government ‘‘liability.’’
Monetary Policy Rules, the Fiscal Theory of the Price Level 59

Our second question is, as in other cases, whether the adjustment leading to
that unique equilibrium is feasible. Consider, in the same figure 3.4, a pre-initial
point anywhere in the b, m plane—say, point B. Given the pegged nominal inter-
est rate policy, the unique equilibrium can be reached via an open market opera-
tion (a sale of bonds to government in exchange for money), leading to point D,
and from there on a fall in the price level, which will translate to an increase in
the real value of both bonds and money leading to the equilibrium point A. An-
other alternative amounting to the same is a fall in prices leading to point C,
and from there on a transformation of bonds into money via exchanges with the
central bank, leading again to the equilibrium A. There are, obviously, infinite
possible sequences of falls in the price level and open market operations leading,
in this instantaneous adjustment process, to the unique, rational steady-state
equilibrium.
Our third question (is there a market-induced stability mechanism such that
agents would motivate the required exchanges of money for bonds and vice
versa, as well as the required changes in prices?) is more complex, and our answer
will be tentative. If the mechanism exists, then the central bank’s passive pegging
will be sufficient. Individuals hold both money and bonds, so that at each point in
time they decide on both the composition of their portfolio and their consumption
expenditures. Thus, we have here two possible adjustment mechanisms, one
which changes the ðB=MÞ ratio and another that changes prices. It seems reason-
able to assume that agents increase their consumption expenditures (leading to a
rise in prices, since aggregate consumption is constant) when they hold ‘‘too
much money’’ ðm > m  Þ and ‘‘too many bonds’’ ðb > b  Þ, and that they exchange
money for bonds, or vice versa, in order to achieve a ratio of their nominal assets
equal to ðB=MÞ ¼ ðb  =m  Þ. If this is the case, then there are reasons to believe, at
the intuitive level, that a market-induced stability mechanism exists, and we take
this to be the case. If this mechanism falters, then the central bank can always
engage in active open market operations in order to complement it—marking a
transition from pegging to targeting.
We perform now three experiments, which should be sufficient for providing
both a sense of the mechanism involved in an interest rate rule as well as some
arguments in favor of a rather positive evaluation of the rule in terms of plausible
results and policy consequences.
Consider, first, an unexpected permanent increase in the level of the pegged
nominal interest rate, a case described in the graph of figure 3.5 (of course, the
same discussion can be performed in analytical terms). Suppose, in order to avoid
clutter, that the initial position was at point A, and that the interest rate increase
results in a new equilibrium at pont C. The adjustment, in this case, requires an
exchange of money for bonds; that is, an open market sale of bonds, which moves
60 Leonardo Auernheimer

Figure 3.5

Figure 3.6

the system from point A to point B, and from there, a fall in the level of prices
resulting in a movement from point B to point C. Notice that in this case the result
is consistent with the usual initial contractionary effect of a rise in the interest
rate—which in this model, since aggregate output and consumption are constant,
translates in a fall of the price level. And this is so despite the fact that, given the
constant real interest rate, the inflation rate will immediately increase to a perma-
nently higher level (expression 3.16 0 ).21
Next, take the case in which an increase in the level of the pegged interest rate is
announced, at time t0 , to take place at a future time t1 , the announcement being
both believed and later implemented. This is depicted in figure 3.6, where point
A shows the initial equilibrium b  , m  at the time of the announcement. With effi-
cient markets, a requirement in the adjustment is that at the time of the effective
increase of the interest rate no discrete changes in the price level are to take place
(the usual ‘‘asset price continuity principle’’), meaning any changes at time t1 will
Monetary Policy Rules, the Fiscal Theory of the Price Level 61

Figure 3.7

be limited to open market operations. The qualitative features of the analysis are
then easy to characterize: an initial fall in the price level, together with a purchase
of bonds by the public at the time of the announcement (immediately positioning
the economy at a point such as C), followed by a gradual increase in the level of
real government debt during the interim period (from C to D) and, at time t1 (at
which the increase in the interest rate is consummated), an additional purchase of
government bonds by the public. The effects are qualitatively similar to the previ-
ous case of a contemporaneous, unanticipated change, but with a smaller initial
increase in the level of prices.
Finally, consider the case of a one-time, unanticipated fall in the level of the fis-
cal primary surplus, s. This is an interesting experiment, because it is a policy
change of the ‘‘unpleasant monetarist arithmetic’’ type: a rise in the primary defi-
cit, with no change in the rate of inflation. This case is depicted in figure 3.7,
where point A corresponds to the initial equilibrium b  , m  . A fall in the primary
surplus determines a lower long-run level of the real government debt, say, at
b  < b  , with a new equilibrium at point A0 —notice that the arrows indicating
the law of motion are drawn for this new lower level b  . The instantaneous ad-
justment involves both a discrete increase in the price level (from A to B) and an
open market operation (a sale of government debt by the public), resulting in a
movement from point B to point A0 .
Some comments on the results of these conceptual experiments are in order.
First, as indicated above, notice that the results of an unanticipated increase in the
nominal interest rate give a rather clear account of the contractionary effects of an
increase in the interest rate and, as a corollary, an explanation for the short- versus
long-run consequences of such an increase: an initial deflationary effect, followed
by a permanently higher rate of inflation. This is not a result that is easy to gener-
ate in models with microeconomic foundations.
62 Leonardo Auernheimer

Second, the adjustment following a fall in the primary surplus implies an auto-
matic response to an occurrence particularly prevalent in developing economies.
In those economies, it is common to encounter situations in which the primary
deficit is incompatible with the inflation rate implied by either the rate of money
growth or, in the case of an open economy, an exogenous path of the nominal ex-
change rate with a ‘‘too-low’’ rate of devaluation and inflation. In all those epi-
sodes the eventual unavoidable accommodation (usually in the central bank’s
rather than in the fiscal authority’s behavior) results in a loss of credibility for the
monetary authority. With a nominal interest rate rule, which requires no change
in the preannounced monetary policy, the credibility question is placed at the
door of the fiscal authority, where it belongs.
An additional comment is also in order, related to the point made in the previ-
ous paragraph concerning the implications of a nominal interest rate rule for the
case of an open economy. It is easy to show that the conclusions in this paper can
be extended, with the same qualitative results, to the case of a small open econ-
omy, and in particular for the nominal interest rate rule. Consider the case of
such a small open economy, with perfect capital mobility. In that case, the nomi-
nal exchange rate becomes the price level, and the rate of devaluation the inflation
rate. Suppose then that the nominal interest rate is pegged at a level i  . Now ex-
pression 3.3 becomes the uncovered interest rate parity condition
iðtÞ ¼ r þ ^eðtÞ;

with r being the rest of the world’s interest rate in terms of foreign currency and ê
the rate of devaluation (for e being the nominal exchange rate). Then, in the same
way that the closed economy pegging the nominal interest rate indirectly pegs the
inflation rate but not the price level, in this case it yields control over the rate of de-
valuation, but not on the level of the exchange rate. Notice the important differ-
ence between this system and the case of an exchange rate rule, in which both the
level and the rate of change of the nominal exchange rate are exogenously con-
trolled. Suppose now, for example, that there is an unexpected increase in the pri-
mary fiscal deficit (that is, a fall in the fiscal surplus). The adjustment then implies
a devaluation, but one which takes place automatically, as part of the normal
mechanism of the rule, rather than one decreed by the central bank when follow-
ing an exchange rate rule. An additional comment for the case of the open econ-
omy is that an increase in the fixed level of the nominal interest rate results in an
initial fall of the nominal exchange rate—in an appreciation of the currency. This
is, of course, equivalent to the initial fall in the price level for the case of the closed
economy. It is a widely accepted result among both economists and policy mak-
ers, but one which is, again, not easy to generate in models with microeconomic
foundations.22
Monetary Policy Rules, the Fiscal Theory of the Price Level 63

3.5 Concluding Remarks

We have presented a very simple continuous-time model with microeconomic

foundations and used it to analyze some of the questions related to the fiscal
theory of the price level and to a policy of pegging the nominal interest rate. We
have shown that the hyperinflationary solution as an alternative rational expecta-
tions path does not exist (in the context of an exogenous monetary rule) for the
quite general case of a positive rate of monetary expansion. Indeed, we see no rea-
son for the usual statement that such a path is the FTPL alternative to the mone-
tary explanation resulting in a steady-state, noninflationary path of prices and
real variables. In fact, the hyperinflationary possibility has been shown to exist
long before the formulation of the FTPL, and it is the presence of nominal govern-
ment debt (fiscal theory or not) that allows the elimination of such a path when
the rate of money growth is positive.
The strict formulation of the FTPL ignores the requirement that in the steady
state, in a stationary economy, the rate of inflation needs to be the same as the
rate of monetary growth—in terms of our expressions for the case of a monetary
rule, not only expression 3.13 needs to hold in the steady state, but so does 3.14.
The problem is eliminated under a scheme in which interest on money is trans-
ferred to the private sector as a lump-sum transfer, in which case, in the steady
state, the price level depends exclusively on the aggregate ðB þ MÞ, but this is
done only at the cost of essentially transforming money into bonds.
In an economy with one or more nominal assets, in which a rational expecta-
tions unique equilibrium is characterized by a unique set of real values of those
assets, it will be true that for each of those assets (or arbitrary group of those
assets) their real value will be their nominal value divided by the price level. One
can then write such an expression, but to assert that from such an equation it fol-
lows that the price level is determined by the level of that nominal asset or any
combination (sum) of those assets would be fallacious and highly misleading. Of
course, nowhere does any proponent of the FTPL assert that, but the use of a
single equation (the long-run equilibrium government budget constraint) conveys
the impression that at times some elements or variations of the fallacy may be
present in the argument.
We find that an interest rate pegging rule not only delivers a unique, well-
determined price level, but that some simple experiments generate results which
are compatible with, and provide a theoretical explanation of, empirical observa-
tions. We have also argued that such a rule seems to have some attractive features
as a policy.
We have explored a couple of other issues not strictly pertaining to the main
purpose of the paper, but which were closely related. In particular, for the case of
64 Leonardo Auernheimer

a monetary rule, we have characterized the short-run temporary hyperinflation-

ary ‘‘solution’’ that has been the response in some countries at various times.
Here, we have distinguished clearly—with even the graphic representation con-
tributing to that clarity—the two effects that a one-time increase in the nominal
money has on government debt: inflating debt away, when in nominal terms,
and the purchase of real debt that is always possible for a positive price of
money, including the case of indexed debt. The latter is often neglected, in par-
ticular when considering the government’s incentives to orchestrate a surprise
hyperinflation.
Finally, a short comment is in order regarding levels and rates-of-change con-
siderations involved in various policy rules. It is clear that the usual strict mone-
tary rule implies an exogenous path of nominal money involving both its level
and its rate of change. In the case of an open economy, an analogue is the stan-
dard exchange rate rule, again fixing both the level and the rate of change (in
other words, the rate of devaluation) of the nominal exchange rate.23 In contrast,
a nominal interest rate rule, indirectly and via the Fisher equation, fixes the rate
of change, but not the level, of prices. In section 2, when considering the case of
the usual monetary rule (which sets the path of both the level and the rate of
change of the nominal money stock), we made reference to the possibility of a
modified monetary rule, fixing only the rate of change of the nominal money
stock. In this case, in the context of our model, at each point in time the level of
both the money supply and government debt could be decided by the public, and
in principle such a rule could yield results very similar to the nominal interest rate
rule, provided either that there would be a market incentive mechanism for indi-
viduals to engage in open market operations, or that the central bank would take
the initiative. What seems to be interesting about this characterization is that both
an interest rate rule and the modified monetary rule, as fixing exogenous rates of
change or either prices or money, leave a degree of freedom for the market to
achieve a steady-state rational expectations unique equilibrium that may not, and
in some cases will definitely not, be feasible with policies which exogenously fixed
both levels and rates of change.

Acknowledgments

I am deeply indebted to Carlos A. Végh for many insightful comments and, more
importantly, for his interest in my work and our invigorating exchanges for many
years. At various times I benefitted from conversations on this topic with Stanley
Fischer and Ben McCallum. The discussion in section 3.4 relies on the analysis
first presented in Auernheimer and Contreras 1990, 1995.
Monetary Policy Rules, the Fiscal Theory of the Price Level 65

Notes

1. Witness some of the statements in the discussion: ‘‘The thesis of this paper is that the ‘fiscal theory of
the price level’ is fatally flawed’’ (Buiter 2002, 459) and, in the abstract of the same paper, ‘‘This paper
argues that the ‘fiscal theory of the price level’ has feet of clay.’’ Niepelt (2004) paper title: ‘‘The Fiscal
Myth of the Price Level.’’
2. For a notable exception, see Daniel (2001).
3. It is of course true that a correctly constructed discrete time model should have, as a limit when the
interval between dates approaches zero, an equivalent continuous time version, and vice versa, but this
is not always easy to secure. See, for example, Carlstrom and Fuerst (2001), where, in the context of a
different question, it is shown how ‘‘seemingly minor modifications’’ (whether beginning-of-the-period
or end-of-the-period real money balances should enter the utility function) can have the consequence of
‘‘dramatic differences’’ in the conclusions. The authors interpret their results as suggesting that ‘‘as
monetary theorists we must be more careful in writing down the basics of our models’’ and that ‘‘there
are concerns with continuous-time analyses that simply sweep this fundamental timing issue under the
rug.’’ We certainly agree with the first of these conclusions, but much less so with the second.
4. See, for example, Cochrane (2001) for a discussion of the importance of debt maturity in the context
of the questions at hand.
5. See Buiter (2002), McCallum (2001), Kocherlakota and Phelan (1999), and Woodford (1995) among
others.
6. The equality r ¼ r is not an arbitrary assumption. If behind the assumption of a ‘‘fixed, instantane-
ously perishable flow endowment’’ there is a fixed, nondepreciable, nonreproduceable aggregate capi-
tal stock—the most sensible economic assumption—capital can be exchanged in the market and will
have a price in terms of its output. Given the fixed rate of time preference, and a fixed marginal prod-
uct of capital, the price of capital will adjust so that the interest rate (the ratio of the marginal product
of capital and its price in terms of output) turns out to be equal to the rate-of-time preference so as to
assure the optimality of the constant level of consumption, dictated by the overall resource constraint.
7. See McCallum (2001), 21.
8. The usual assumption in this literature is b b 0, and this has consequences for the set of admissible
adjustment paths. We see no reason for this limitation, which is at variance with much of the literature
on the open macroeconomy, for example, in which government can hold positive net assets. Kocherla-
kota and Phelan (1999) also allow for negative government debt.
9. This will not necessarily preclude discrete changes in either the nominal money supply or nominal
bonds as ‘‘conceptual experiments.’’
10. See (note 13), though, for a subtle qualification to this statement.
11. More specifically (Kocherlakota and Phelan 1999), a Ricardian policy is one in which government
sets a fiscal and monetary policy compatible with any possible path of prices. A non-Ricardian policy
is one in which government’s setting of the monetary and fiscal variables is compatible with only one
path of prices. A typical example of the latter is the current specification of a monetary rule; a typical
example of the former would be, in the current case, the setting of the rate of money growth to which-
ever value is necessary for the long-run equilibrium to obtain, for any possible level of the fiscal sur-
plus and current government debt. For this last case proponents of the FTPL concede the relevance of
the traditional monetary analysis.

12. This would not be so, of course, in the case of indexed government debt (i.e., debt denominated in
real terms), a case to which we will make reference later.
66 Leonardo Auernheimer

13. The possibility of this instantaneously temporary difference between the demand and the supply of
money is the ‘‘subtle qualification’’ alluded to in note 10.
14. The idea of an adjustment process taking place at a single point in chronological time can be inter-
preted, of course, as describing an adjustment during a very short interval, in a manner that all agents
understand as taking place separately from the laws of motion implied by the overall equations—in
our case, equations 3.11–3.12.
15. Sargent and Wallace (1981) analyze the case of indexed government debt—that is, denominated in
real terms—but their results are also relevant in the case of nominal debt.
16. Although initially triggered via a 150 percent devaluation, followed by the increase in the money
supply (rather than an increase in money followed by a rise in all prices, including the exchange rate,
as in our analysis), readers familiar with the Latin American experience will probably recognize the
analogy of the analysis to the Rodrigazo episode in Argentina in June 1975—the name referring to Cel-
estino Rodrigo, the economics minister at the time.
17. To have an idea of the ‘‘slipperiness’’ of the issue, see, for example, the ‘‘change of heart’’ in the case
of one of the proponents of the FTPL in different versions of the same paper (Cochrane 2002).
18. Notice that Kocherlakota and Phelan (1999), in arguing for the plausibility of the monetarist solu-
tion (meaning a one-time fall in the price level by the same proportion as the nominal money annihila-
tion), assume a departure from the non-Ricardian policy, with government changing the fiscal surplus
to adjust for the increase in real debt as a result of the price fall.
19. In fact, our early work (Auernheimer and Contreras 1990, 1995), which together with Begg and
Haque’s (1984) has been credited by a generous colleague as being ‘‘the original exponents of the
FTPL’’ (Buiter 2002), was exclusively motivated by the analysis of an interest rate policy pegging, with-
out explicit references to a general theory of price determination.
20. Woodford (1995) recognizes this problem, but dismisses it as irrelevant for the proposition that the
aggregate ðB þ MÞ is sufficient to determine the path of prices.
21. It is easy to show that, given assumption 3.10 0 (i.e., the economy is in the range of the monetary de-
mand for which higher inflation translates into higher revenue from money creation), then the new
equilibrium point C will always be above the line connecting points A and B, meaning that a fall in
prices will be required.

22. Calvo and Végh (1990) also generate this result, through a very different mechanism.
23. Notice that this is equivalent, for the close economy, to the government fixing the price level, and
its rate of change, by committing to buy or sell a representative bundle of goods at a preannounced
fixed price.

References

Auernheimer, L., and Contreras, B. 1990. ‘‘Control of the Interest Rate with a Government Budget Con-
straint: Price Level Determinacy, and other Results.’’ Mimeo., Texas A&M University.
———. 1995. ‘‘Control de la tasa de interes con restriccion presupuestaria: determinacion de los precios
y otros resultados.’’ El Trimestre Economico 62, no. 3: 381–396.
Bassetto, M. 2002. ‘‘A Game-Theoretic View of the Fiscal Theory of the Price Level.’’ Econometrica 70, no.
6: 2167–2195.

Begg, D., and Haque, B. 1984. ‘‘A Nominal Interest Rate Rule and Price Level Indeterminacy Reconsid-
ered.’’ Greek Economic Review 6: 31–46.
Monetary Policy Rules, the Fiscal Theory of the Price Level 67

Buiter, W. H. 2002. ‘‘The Fiscal Theory of the Price Level: A Critique.’’ Economic Journal 112: 459–480.
———. 2005. ‘‘New Developments in Monetary Economics: Two Ghosts, Two Eccentricities, a Fallacy,
a Mirage and a Mythos.’’ Economic Journal 115: C1–C31.

Calvo, G. A. 1981. ‘‘Devaluation: Levels versus Rates.’’ Journal of International Economics 11, no. 2: 165–
172.

———. 1985. ‘‘Macroeconomic Implications of the Government Budget: Some Basic Considerations.’’
Journal of Monetary Economics 15, no. 1: 95–112.
Calvo, G. A., and Végh, C. A. 1990. ‘‘Interest Rate Policy in a Small Open Economy.’’ IMF Staff Papers
37, no. 4: 753–776.
Carlstrom, C. T., and T. S. Fuerst. 1998. ‘‘Real Indeterminacy in Monetary Models with Nominal Inter-
est Rate Distortions: The Problem with Inflation Targets.’’ Working Paper No. 9818, Federal Reserve
Bank of Cleveland.
———. 2000. ‘‘The Fiscal Theory of the Price Level.’’ Federal Reserve Bank of Cleveland Economic Review
Q1: 22–32.
———. 2001. ‘‘Timing and Real Indeterminacy in Monetary Models.’’ Journal of Monetary Economics 47:
285–298.
Christiano, L. J., and Fitzgerald, T. 2000. ‘‘Understanding the Fiscal Theory of the Price Level.’’ Federal
Reserve Bank of Cleveland Economic Review 36, no. 2: 1–37.

Cochrane, J. 2001. ‘‘Long-Term Debt and Optimal Policy in the Fiscal Theory of the Price Level.’’ Econo-
metrica 69, no. 1: 69–116.

———. 2002. ‘‘Money as Stock: Price Level Determination With No Money Demand.’’ National Bureau
of Economic Research Working Paper 7498, Cambridge, MA.
Daniel, B. 2001. ‘‘The Fiscal Theory of the Price Level in an Open Economy.’’ Journal of Monetary Eco-
nomics 48: 293–308.
Dupor, B. 2000. ‘‘Exchange Rates and the Fiscal Theory of the Price Level.’’ Journal of Monetary Econom-
ics 45, no. 3: 613–630.
Kocherlakota, N., and C. Phelan. 1999. ‘‘Explaining the Fiscal Theory of the Price Level.’’ Federal Reserve
Bank of Minneapolis Quarterly Review 23, no. 4: 14–23.
Leeper, E. 1991. ‘‘Equilibria Under Active and Passive Monetary and Fiscal Policies.’’ Journal of Mone-
tary Economics 27: 129–147.
Lucas, R. E. 1978. ‘‘Asset Prices in an Exchange Economy.’’ Econometrica 46, no. 6: 1429–1425.
McCallum, B. T. 1997. ‘‘Crucial Issues Concerning Central Bank Independence.’’ Journal of Monetary
Economics 39: 99–112.
———. 2001. ‘‘Indeterminacy, Bubbles and the Fiscal Theory of the Price Level.’’ Journal of Monetary
Economics 47: 19–30.
Metzler, L. A. 1951. ‘‘Wealth, Saving, and the Rate of Interest.’’ Journal of Political Economy 59, no. 2: 93–
116.
Niepelt, D. 2004. ‘‘The Fiscal Myth of the Price Level.’’ Quarterly Journal of Economics 119: 277–300.
Obstfeld, M., and Rogoff, K. 1983. ‘‘Speculative Hyperinflations in Maximizing Models: Can We Rule
Them Out?’’ Journal of Political Economy 91, no. 4: 675–687.
68 Leonardo Auernheimer

Sargent, T., and N. Wallace. 1981. ‘‘Some Unpleasant Monetarist Arithmetic.’’ Federal Reserve Bank of
Minneapolis Quarterly Review 5, no. 3.
Sims, C. 1994. ‘‘A Simple Model for Study of the Determination of the Price Level and the Interaction of
Monetary and Fiscal Policy.’’ Economic Theory 4: 381–399.
———. 2001. ‘‘Fiscal Consequences for Mexico of Adopting the Dollar.’’ Journal of Money, Credit an
Banking 33, no. 2: 597–616.
Tobin, J. 1974. ‘‘Friedman’s Theoretical Framework.’’ In Milton Friedman’s Monetary Framework, ed. R. J.
Gordon, 77–89. Chicago: University of Chicago Press.
Woodford, M. 1994. ‘‘Monetary Policy and Price Level Determinacy in a Cash-In-Advance Economy.’’
Economic Theory 4: 345–389.
———. 1995. ‘‘Price Level Determinacy Without Control of a Monetary Aggregate.’’ Carnegie-Rochester
Series on Public Policy 43: 1–46.
———. 1998. ‘‘Doing Without Money: Controlling Inflation in a Post-Monetary World.’’ Review of Eco-
nomic Dynamics 1: 173–219.
II Monetary and Exchange Rate
Policy in Practice
4 Can Inflation Targeting Work
in Emerging Market Countries?

Frederic S. Mishkin

An important theme in Guillermo Calvo’s work has been that emerging-market

economies are very different from advanced economies, and this has important
implications that need to be factored in when designing macroeconomic policies.
For example, Guillermo (e.g., see Calvo 1999a and 2001 and Calvo and Mendoza
2000) has been quite skeptical of inflation targeting as a monetary policy strategy
for emerging-market countries.1 He worries that allowing policy makers too much
discretion in the weak institutional environment of emerging-market countries
might lead to poor macroeconomic outcomes. I am a known advocate of inflation
targeting, but I think that Guillermo’s concerns about inflation targeting’s efficacy
in emerging-market countries need to be taken seriously.
Before starting it is important to make clear what an inflation targeting regime
is all about. It comprises five elements: (1) the public announcement of medium-
term numerical targets for inflation; (2) an institutional commitment to price
stability as the primary goal of monetary policy, to which other goals are subordi-
nated; (3) an information-inclusive strategy in which many variables, and not just
monetary aggregates or the exchange rate, are used for deciding the setting of
policy instruments; (4) increased transparency of the monetary policy strategy
through communication with the public and the markets about the plans, objec-
tives, and decisions of the monetary authorities; and (5) increased accountability
of the central bank in attaining its inflation objectives. The list should clarify one
crucial point about inflation targeting: it entails much more than a public an-
nouncement of numerical targets for inflation for the year ahead. This is impor-
tant in the context of emerging-market countries because many of them routinely
report numerical inflation targets or objectives as part of their government’s eco-
nomic plan for the coming year, and yet their monetary policy strategy should
not be characterized as inflation targeting, which requires the other four elements
for it to be sustainable over the medium term.
In this paper, I explore what additional issues need to be addressed in
emerging-market countries to make inflation targeting work for them. I start by
72 Frederic S. Mishkin

outlining why emerging-market economies are so different from advanced econo-

mies and then discuss why developing strong fiscal, financial, and monetary insti-
tutions is so critical to the success of inflation targeting. Then I look at two
emerging-market countries that illustrate what it takes to make inflation targeting
work well, Chile and Brazil. I then address a particularly complicated issue for
central banks in emerging-market countries who engage in inflation targeting:
how they deal with exchange rate fluctuations. The next topic focuses on the
IMF’s role in promoting the success of inflation targeting in emerging-market
countries. The conclusion from this analysis is that inflation targeting is more
complicated in emerging market countries and is thus not a panacea. However,
inflation targeting done correctly can be a powerful tool to help promote macro-
economic stability in these countries, as is supported by recent empirical research
that finds that inflation performance in emerging-market countries improves after
the adoption of inflation targeting, with no adverse effect on output (Battini and
Laxton, forthcoming, and Mishkin and Schmidt-Hebbel 2007).

4.1 Why Emerging Market Economies Differ from Advanced Economies

In our paper on the choice of exchange rate regimes (Calvo and Mishkin 2003),
Guillermo and I have outlined five fundamental institutional differences for
emerging-market countries that must be taken into account to derive sound
theory and policy advice. These are
• Weak fiscal institutions
•Weak financial institutions, including government prudential regulation and
supervision
• Low credibility of monetary institutions
• Currency substitution and liability dollarization
• Vulnerability to sudden stops (of capital inflows)

Advanced countries are not immune to problems with their fiscal, financial,
and monetary institutions, but there is a major difference in the degree of the
problem in emerging-market countries. Such weak institutions make emerging-
market countries very vulnerable to high inflation and currency crises, so that the
real value of money cannot be taken for granted. As a result, emerging-market
countries face the threat of domestic residents switching to a foreign currency,
leading to currency substitution (Calvo and Végh 1996). Currency substitution is
likely to be due not only to past inflationary experience, but also to the fact that a
currency like the U.S. dollar is a key unit of account for international transactions.
This phenomenon has induced the monetary authority to allow banks to offer for-
Can Inflation Targeting Work in Emerging Market Countries? 73

eign exchange deposits. In this fashion, a sudden switch away from domestic and
into foreign money need not result in a bank run, since in the presence of foreign
exchange deposits such a portfolio shift could be implemented by simply chang-
ing the denomination of bank deposits. Otherwise, deposits would be drawn
down to purchase foreign exchange, resulting in a bank run.
Foreign exchange deposits induce banks—partly for regulatory reasons that
prevent them from taking exchange rate risk—to offer loans denominated in for-
eign currency, usually dollars, leading to what is called liability dollarization. As
pointed out in Mishkin (1996) and Calvo (2001), liability dollarization is what
leads to an entirely different impact of currency crises on the economy in emerg-
ing market versus advanced countries. In emerging-market countries, a sharp
real currency depreciation raises the value of liabilities in local currency, thus
causing the net worth of corporations and individuals to fall, especially those
whose earnings come from the nontradables sector. This serious negative shock
to corporations’ and individuals’ balance sheets then increases asymmetric infor-
mation problems in credit markets, leading to a sharp decline in lending and an
economic contraction. Thus, liability dollarization (where the currency mismatch
takes place in corporate and household balance sheets) may become a major prob-
lem for economies that are relatively closed and highly indebted (this has typi-
cally been the case in several emerging-market countries after the capital-inflow
episode in the first half of the 1990s; see Calvo, Izquierdo, and Talvi 2002). Under
those circumstances, the monetary authority is likely to display ‘‘fear of floating’’
(see Calvo and Reinhart 2002)—a reluctance to allow free fluctuations in the
nominal exchange rate—placing an additional constraint on emerging market
countries’ monetary policy.2 It should be noted, however, that not all emerging-
market countries suffer from liability dollarization in a serious way (for example,
Chile and South Africa. See Eichengreen, Hausmann, and Panizza 2002).
A dominant phenomenon in emerging-market countries is a sudden stop, a large
negative change in capital inflows, which, as a general rule, appears to contain a
large unanticipated component (see Calvo and Reinhart 2000). This phenomenon
is mostly confined to emerging market countries. It likely hits only such countries
because of their weak fiscal and financial institutions, and it is only recently that
sudden stops have been subject to systematic empirical analysis. Preliminary evi-
dence suggests that there is a high degree of bunching of sudden stops across
emerging market countries. This is especially evident after the 1998 Russian crisis
and the recent Wall Street scandals of Enron and others. This leads to the conjec-
ture that, to a large extent, sudden stops have been a result of factors somewhat
external to emerging market countries as a group. They might have been trig-
gered by crisis in one emerging market country, as the Russian crisis illustrates,
but contagion across emerging-market countries was likely due to difficulties
74 Frederic S. Mishkin

experienced outside these countries, especially in world financial centers (for evi-
dence about that, see Kaminsky, Reinhart and Végh 2003). To illustrate, a possible
contagion mechanism is a liquidity crunch in a financial center, triggered by mar-
gin calls following unexpected capital losses, which leads the financial center to
dump emerging market securities or to at least not bid for new debt instruments
issued by emerging market countries. Calvo (1999b) conjectures that this mecha-
nism could explain the large negative impact that the Russian crisis had on all
emerging markets.
The effect of sudden stops on individual countries is by no means uniform. In
Latin America, for example, Argentina suffered a very serious dislocation, while
neighboring Chile escaped relatively unscathed (although Chile did see its growth
rate fall by more than 50 percent). In Asia, Korea has had a strong recovery while
Indonesia is still reeling from the shock. Tentative analysis suggests that these dif-
ferent outcomes have much to do with initial conditions. Chile had low debt rela-
tive to Argentina and did not suffer from liability dollarization, while at the time
of crisis most debt instruments in Argentina were denominated in U.S. dollars.
On the other hand, even though both Korea and Indonesia suffered from foreign
exchange debt, the former was able to socialize much of the financial problem (as
a result, Korea’s debt climbed from about 12 to about 33 percent of GDP in 1996–
1998). Once again, debt and currency mismatch appear to have played a crucial
role in determining the depth of crisis.
In the following sections, we will see how these institutional features make in-
flation targeting a more complicated exercise in emerging-market economies than
in advanced economies.

4.2 Developing Strong Fiscal and Financial Institutions

Fiscal stability is a fundamental necessary condition for inflation control, and

hence inflation targeting. A key lesson from the ‘‘unpleasant monetarist arithme-
tic’’ discussed in Sargent and Wallace (1981) and the recent literature on fiscal
theories of the price level (Woodford 1994 and 1995) is that irresponsible fiscal
policy puts pressure on the monetary authorities to monetize the debt, thereby
producing rapid money growth and high inflation. If fiscal imbalances are large
enough, monetary policy eventually becomes subservient to fiscal considerations
(so-called fiscal dominance), and an inflation target would have to be abandoned
or seriously modified. Fiscal policy is also found to be highly procyclical in emerg-
ing market countries, which can also contribute to procyclical monetary policy
(Kaminsky, Reinhart, and Végh 2004), and both can contribute to cyclical inflation
fluctuations that make an inflation target harder to achieve.
Can Inflation Targeting Work in Emerging Market Countries? 75

Similarly, a safe and sound financial system is also a necessary condition for the
success of an inflation-targeting regime. A weak banking system is particularly
dangerous. Once a banking system is in a weakened state, a central bank cannot
raise interest rates to sustain the inflation target because this will likely lead to a
financial system collapse. Not only can this directly cause a breakdown of the in-
flation targeting regime, but it can also lead to a currency collapse and a financial
crisis, which will also erode the control of inflation. When markets recognize the
weakness of the banking system, there will be a reversal of capital flows out of
the country (a sudden stop) that will result in a sharp depreciation of the ex-
change rate, which leads to upward pressures on the inflation rate. Moreover, as
a result of the currency devaluation that most likely accompanies the monetary
expansion, the debt burden of domestic firms denominated in foreign currency
rises, while the assets, which are denominated in domestic currency, do so at a
much slower pace, thus leading to a decline in net worth. As described in Mishkin
(1996), this deterioration in balance sheets then increases adverse selection and
moral hazard problems in credit markets, leading to a sharp decline in investment
and economic activity, and ultimately a complete collapse of the banking system.
The subsequent bailout of the banking system leads to a huge increase in govern-
ment liabilities, which will have to be monetized in the future (Burnside, Eichen-
baum, and Rebelo 2001), thus undermining the inflation-targeting regime.
Unfortunately, the scenario outlined here has happened all too often in recent
years as evidenced by the twin crises (currency and financial) in Chile in 1982, in
Mexico in 1994–1995, in East Asia in 1997, in Ecuador in 1999, and in Turkey in
2000–2001.
Fiscal imbalances can also lead to banking and financial crises that will blow
out any monetary regime to control inflation. As outlined in Mishkin and Savas-
tano (2001), large budget deficits may force the government to confiscate assets,
particularly those in the banking system. This has indeed happened often in Latin
America. The suspicion that this might occur would then cause depositors and
other creditors to pull their money out of the banking system, and the resulting
banking crisis would then cause a contraction of lending and the economy. This
has happened several times in Argentina’s checkered history, with the most
recent variant occurring in 2001. The Argentine banking system was generally in
quite good shape until 2000, even though the economy had been in a recession
for several years. The strength of the Argentine banking system was the result of
a sophisticated prudential regulatory and supervisory regime put into place after
the tequila crisis that made Argentina’s prudential supervision one of the best in
the emerging-market world (see Calomiris and Powell 2000). Large budget defi-
cits forced the Argentine government to look for a new source of funds from the
76 Frederic S. Mishkin

banking system, which was primarily foreign-owned. After Domingo Cavallo be-
came minister of the economy in April 2001 and the central bank president, Pedro
Pou, was forced to resign, prudential supervisory standards were weakened and
banks were both encouraged and coerced into purchasing Argentine government
bonds. With the bonds’ value declining as the likelihood of default on this debt
increased, banks’ net worth plummeted. The likely insolvency of the banks then
led to a classic run on the banks and a full-scale banking crisis by the end of 2001.
The result was a collapse of currency, a devastating depression, and an initial
surge in inflation.
The particular inflation control problems arising from fiscal and financial sector
imbalances are of course not unique to emerging-market countries and are also a
concern in advanced economies. However, these problems are of a different order
of magnitude for emerging-market countries and so must be addressed at the
outset if an inflation-targeting regime is to keep inflation under control. Fiscal
reforms, which increase transparency of the government budget, and budget
rules, which help keep budget deficits from spinning out of control, are needed to
prevent the fiscal imbalances that can lead to a collapse of an inflation targeting
regime.3
Avoiding financial instability requires several types of institutional reform.
First, prudential regulation of the banking and financial system must be strength-
ened in order to prevent these types of financial crises.4 Second, the safety net pro-
vided by the domestic government and the international financial institutions set
up by Bretton Woods might need to be limited in order to reduce the moral haz-
ard incentives for banks to take on too much risk.5 Third, currency mismatches
need to be limited in order to prevent currency devaluations from destroying bal-
ance sheets. Although prudential regulations requiring that financial institutions
match up any foreign-denominated liabilities with foreign-denominated assets
may help reduce currency risk, they do not go nearly far enough. Even when the
banks have equal foreign-denominated (dollar) assets and liabilities, if the banks’
dollar assets are loans to companies who themselves are unhedged, then banks
are effectively unhedged against currency devaluations because the dollar loans
become nonperforming when the devaluation occurs.6 Thus, limiting currency
mismatches may require government policies to limit liability dollarization or
at least reduce the incentives for it to occur.7 Fourth, policies to increase the
openness of an economy may also help limit the severity of financial crises in
emerging-market countries. The reason why openness may affect financial fragil-
ity is that businesses in the tradable sector have balance sheets that are less
exposed to negative consequences from a currency devaluation when their debts
are denominated in foreign currency. Because the goods they produce are traded
internationally, they are more likely to be priced in foreign currency. Then a de-
Can Inflation Targeting Work in Emerging Market Countries? 77

valuation that raises the value of their debt in terms of domestic currency is also
likely to raise the value of their assets, thus insulating their balance sheets from
the devaluation. Moreover, as argued in Calvo, Izquierdo, and Talvi (2002), the
more open the economy, the smaller the required real currency depreciation fol-
lowing a sudden stop. Therefore, although firms in the nontradable sector are
exposed to balance sheet shocks if they are liability dollarized, the size of the
shock is smaller when the economy is more open.
One view is that these fiscal and financial reforms must be in place before infla-
tion targeting can even be attempted (Masson, Savastano, and Sharma 1997).
However, although fiscal and financial stability are necessary conditions for infla-
tion control, I think the view that these reforms are prerequisites for attempting
an inflation-targeting regime in emerging-market countries is too strong. Indeed,
Batini and Laxton (forthcoming) find that adoption of inflation targeting in
emerging market countries is followed by better inflation performance even when
the fiscal and financial reforms are not yet in place. Because inflation targeting
commits the government to keeping inflation low, it can be argued (Brash 2000
and Bernanke, Laubach, Mishkin, and Posen 1999) that inflation targeting can
help promote fiscal and financial reforms because it becomes clearer that the gov-
ernment must support these reforms if the inflation-targeting regime is to be suc-
cessful. Also, a commitment to inflation control by the government makes it
harder for the government to advocate loose and procyclical fiscal policy as it is
clearly inconsistent with the inflation target. However, instituting an inflation-
targeting regime by no means insures fiscal and financial reforms. If an inflation-
targeting regime is to be sustainable, a commitment to and work on these reforms
is required. In fact, after adopting inflation targeting, emerging-market coun-
tries do seem to pursue many of the necessary reforms (Batini and Laxton,
forthcoming).

4.3 Developing Strong Monetary Institutions

Two underlying monetary institutions are necessary for monetary authorities

to keep inflation under control. The first is a public and institutional commitment
to price stability as the overriding long-run goal of monetary policy. The institu-
tional commitment can be written into law, as it is in central bank legislation
in many countries (most prominent among which are the Economic and Monetary
Union [EMU] countries). This is clearly helpful in mandating that the central bank
control inflation. Inflation targeting takes this institutional commitment even
further by defining price stability—more specifically, by providing a numerical
inflation target as the objective for monetary policy. However, it is important to
recognize that laws may matter less than the general public’s and politicians’
78 Frederic S. Mishkin

commitment to support price stability. Here past history matters. Many emerging-
market countries have had a history of poor support for the price stability goal
and since laws are easily overturned in these countries, it is not clear that laws
will be sufficient.
The second institutional arrangement necessary for the success of inflation tar-
geting is a public and institutional commitment to the instrument independence
of the central bank.8 Instrument independence means that the central bank is pro-
hibited from funding government deficits and must be allowed to set the mone-
tary policy instruments without interference from the government. Additionally,
the members of the monetary policy board must be insulated from the political
process by giving them long-term appointments and protection from arbitrary
dismissal. There is a large literature on central bank independence and the forms
that it takes (see Cukierman 1992 and the surveys in Forder 2000 and Cukierman
2006), but again what is written down in the law may be less important than the
political culture and history of the country, a point emphasized in Cukierman
(2006). The contrast between Argentina and Canada is instructive here. Legally,
the central bank of Canada does not appear to be very independent because the
government has the ultimate responsibility for the conduct of monetary policy. In
the event of a disagreement between the Bank of Canada and the government, the
minister of finance can issue a directive that the bank must follow. However, be-
cause the directive must be specific and in writing, and because the Bank of Can-
ada is a trusted public institution, a government override of the bank is likely to
be highly unpopular and will rarely occur. Thus, in practice, the Bank of Canada
is highly independent. In contrast, the central bank of Argentina was highly inde-
pendent from a legal perspective. However, this did not stop the Argentine gov-
ernment from forcing the resignation of the highly respected central bank
president Pedro Pou in April 2001 and replacing him with a president who would
do the government’s bidding. Indeed, it is unimaginable in countries like Canada
and the United States or in Europe that the public would tolerate the removal of
the head of the central bank in such a manner, and indeed I do not know of any
case of this happening in recent history.9 Thus a strong legal commitment to cen-
tral bank independence without genuine public and political support for this in-
dependence may not be enough to ensure monetary policies that will focus on
inflation control in many emerging market countries.
An important advantage of inflation targeting is that it allows the monetary
authorities some discretion and flexibility to use monetary policy to cope with
shocks to the domestic economy. Indeed, as argued by Fraga, Goldfajn, and Mine-
lla (2003), this flexibility is even more important in emerging-market countries be-
cause they are subject to larger shocks. Inflation-targeting regimes typically have
built-in flexibility to allow them to achieve their inflation target over longer hori-
Can Inflation Targeting Work in Emerging Market Countries? 79

zons, and are better characterized as ‘‘flexible inflation targeting’’ as described in

Svensson (1997) and Bernanke, Laubach, Mishkin, and Posen (1999). This advan-
tage of inflation targeting is also its weakness. Critics of inflation targeting, of
whom Calvo has been a prominent one (for example, see Calvo 1999a, Calvo
2001, and Calvo and Mendoza 2000), have argued that inflation targeting may al-
low too much discretion to monetary policy and so may not provide a sufficient
nominal anchor. However, as pointed out in Bernanke and Mishkin (1997), infla-
tion targeting should be seen as a regime of ‘‘constrained discretion.’’ Transparent
discussions of the conduct of monetary policy can make it more difficult for the
central bank to follow overly expansionary monetary policy, while accountability
means that the central bank pays a high price if it engages in discretionary policy
that leads to high inflation. However, Calvo is quite correct that inflation target-
ing is no panacea. In order for inflation targeting to work to constrain discretion,
it has to be supported by the public and the political process. Inflation targeting
can help focus the public debate so that it supports a monetary policy focus on
long-run goals such as price stability, as has occurred in many inflation-targeting
countries (Bernanke, Laubach, Mishkin, and Posen 1999). However, these benefits
require even more transparency and excellent communication skills on the part of
the central bank in the more politically complicated environment of emerging
market countries. Even excellent communication may not be enough if the politi-
cal environment is incapable of supporting an independent central bank that fo-
cuses on inflation control.

4.4 Two Case Studies of Inflation Targeting in Emerging Market Countries:

The Chilean and Brazilian Examples

Despite the challenges in getting inflation targeting to work, inflation targeting

has been successful in promoting macroeconomic stability in a number of emerg-
ing market countries. Here we will look at two case studies, Chile and Brazil,
which illustrate that institutional development along the lines discussed earlier
has been critical to inflation targeting’s success in the emerging market context.10
Chile is the poster child for inflation-targeting regimes in emerging market
countries because it has had great success in lowering inflation to levels found in
advanced countries, while at the same time experiencing very rapid economic
growth.11 From 1991, when the inflation-targeting regime started, to the present,
Chile has been able to lower inflation rates from above 20 percent to around 3 per-
cent. Over the same period, output growth has been very high, averaging 6 per-
cent per year from 1991 to 2002. However, it is important to emphasize that the
success of Chile’s inflation-targeting regime has been based on development of
the requisite institutions we have discussed here. Over the 1991–2002 period
80 Frederic S. Mishkin

Chile’s budget surplus averaged a little under one percent of Gross Domestic
Product (GDP). In addition, due largely to the measures taken in the aftermath of
the severe banking crisis in the early 1980s (costing over 40 percent of GDP), Chile
developed banking regulation and supervisory practices that are among the best
in the emerging market world (Caprio 1998) and are viewed as comparable to
those found in advanced countries. As a result, even during the tequila crisis of
1995, the Russian meltdown of 1998, and the current difficulties in Latin America,
the soundness of Chile’s financial system has never come into question. The con-
trols on short-term capital inflows have also been cited as another important fac-
tor behind the relative stability of the Chilean economy in the 1990s. However,
these controls are very controversial and it is not at all clear that they have made
a significant contribution to Chile’s success (de Gregorio, Edwards, and Valdes
2000 and Edwards 1999).
Chile also has worked on developing strong monetary institutions. In 1989
Chile passed new central bank legislation (which took effect in 1990, just before
the start of the inflation-targeting regime) that gave independence to the central
bank and mandated price stability as its primary objective. Indeed, as pointed out
in Landerretche, Morandé, and Schmidt-Hebbel (1999), Chile only gradually
hardened up its inflation-targeting regime over time, with the announced infla-
tion objective initially being interpreted more as official inflation projections
rather than formal or ‘‘hard’’ targets. Only after the central bank already had
some success with disinflation, by 1994, did the inflation projections become hard
targets, with the central bank now accountable for meeting them. Furthermore,
until August 1999 Chile had an exchange-rate band around a crawling peg, and
in 1998 it came close to fixing its exchange rate for a time by narrowing the
exchange-rate band sharply (though this was a mistake that will be discussed in
the next section and was reversed later).12 Thus Chile has exhibited some ‘‘fear of
floating’’ along the lines described by Calvo and Reinhart (2002). Indeed, Chile’s
inflation-targeting regime should be seen as very evolutionary, with Chile going
to a full-fledged inflation-targeting regime only in May 2000 (see Mishkin and
Savastano 2002).
Brazil implemented inflation targeting shortly after the real collapsed in January
1999. After much confusion, in early February the new central bank president,
Arminio Fraga, announced that Brazil would soon be adopting an inflation-
targeting strategy. At the same time, Fraga put his money where his mouth was
by increasing the interbank policy interest rate by 600 basis points, to 45 per-
cent. On June 21, the president of Brazil issued a decree instituting an inflation-
targeting framework. This framework included all the features of a full-fledged
inflation-targeting regime, including (1) the announcement of multiyear inflation
targets (with explicit numerical targets for the twelve-month rate of inflation in
Can Inflation Targeting Work in Emerging Market Countries? 81

the years 1999, 2000, and 2001, and a commitment to announce the targets for
2002 two years in advance; (2) assigning the National Monetary Council the re-
sponsibility for setting the inflation targets and tolerance ranges based on a pro-
posal by the minister of finance; (3) giving the central bank full power to
implement the policies needed to attain the inflation targets; (4) establishing pro-
cedures to increase the central bank’s accountability (specifically, if the target
range is breached, the central bank president would have to issue an open letter
to the minister of finance explaining the causes of the deviation, the measures that
would be taken to eliminate it, and the time it would take to get inflation back in-
side the tolerance range); and (5) taking actions to improve the transparency of
monetary policy (concretely, the central bank was requested to issue a quarterly
inflation report modeled after that produced by the Bank of England).
Brazil’s adoption of inflation targeting was not preceded by prior development
of all the fiscal, financial, and monetary reforms discussed earlier. First, the 1999
collapse of the real can be attributed to the inability of the Brazilian government
to put its fiscal house in order: the currency crisis was prompted by a moratorium
on debt payments instituted at the beginning of January by the governor of the
Brazilian state of Minas Gerais. Brazil did then take steps to improve its fiscal bal-
ances under its program with the IMF, but there were still doubts about the dura-
bility of its fiscal reforms (see Mishkin and Savastano 2002.) On the other hand,
Brazil did have a strong banking system because it had undergone a major
restructuring following the banking crisis of 1994–1996 (see Caprio and Klingebiel
1999.) The independence of Brazil’s central bank and the commitment to price sta-
bility, however, were not clear cut. Both were based on a presidential decree and
confidence in a superb central bank president, but not on a more formal commit-
ment based on legislation.
To the surprise of many, the inflation-targeting strategy seemed to work. The
initial inflation targets were set at 8 percent for 1999, 6 percent for 2000, and 4
percent for 2001, with a tolerance range of G2 percent. There was a remarkably
small pass-through from the large depreciation of the real (which fell by 45 per-
cent on impact and thereafter stabilized at 30 percent below its pre-devaluation
level). For several months, the output contraction was contained (in fact, annual
GDP grew by almost 1 percent in 1999), Brazil was not cut off from external
financing—though there was some arm twisting involved—and there were no
major bank runs. By March 1999, asset prices had started to recover, the real
appreciated, and the central bank found room to lower interest rates—which it
did, quite aggressively (from a high of 45 percent to below 20 percent in a seven-
month period). Inflation and the exchange rate remained subdued through Octo-
ber, when the monthly inflation rate rose to 1.2 percent, the largest monthly
increase since June 1996, and the exchange rate crossed, briefly, the then-critical
82 Frederic S. Mishkin

mark of R$2.00 per U.S. dollar. In the event, inflation in 1999 reached 8.9 percent,
above the 8 percent target for the year but well within the 2 percent tolerance
range; during 2000 inflation continued falling, and closed the year right at the 6
percent mid-point target set by the central bank in mid-1999. However, in 2001,
the inflation rate exceeded the 4 percent inflation target by more than the 2 per-
cent tolerance range, ending up at 7.7 percent.
The weakness of some aspects of the institutional framework for fiscal and
monetary policy did come back to haunt the inflation-targeting regime in Brazil.
In the run-up to the presidential election in October 2002, the market had con-
cerns that the front-runner, Lula, would weaken fiscal and monetary institutions.
Lula had made statements seeming to indicate that once in office he would en-
courage fiscal policy to be highly expansionary and would not take steps to pre-
vent a possible default on Brazil’s foreign debt. He also indicated that he would
not reappoint the highly respected president of the central bank, Arminio Fraga.
Thus Lula’s commitment to the independence of the central bank, to price stabil-
ity, and to the inflation-targeting regime was far from clear. Not surprisingly, the
lack of market confidence in Lula, who was elected President, led to a sharp de-
preciation of the Brazilian real and a sharp upward spike in inflation to 12.5 per-
cent, which substantially overshot the inflation target of 3.5 percent for 2002. The
impact of the Brazilian election on the inflation-targeting regime illustrates that
weak fiscal and monetary institutions are likely to create severe problems for an
inflation-targeting regime.
However, the response of the Brazilian government and central bank to the
overshoots of the inflation targets illustrates that inflation targeting can help keep
inflation under control in the face of big shocks like the 2002 real depreciation. As
noted earlier, under the presidential decree that gave birth to inflation targeting,
the Banco Central do Brasil was required to submit an open letter to the ministry
of finance explaining the causes of the breach of the inflation target and what
steps would be taken to get the inflation rate back down again. Given the large
overshoot of the inflation targets, central bank credibility was on the line and the
Banco Central do Brasil handled it well. It explained how it would modify its in-
flation targets and what inflation path it would shoot for in the future with a very
high degree of transparency (see Fraga, Goldfajn, and Minella 2003). First it
explained why the exchange rate had overshot, and made explicit estimates of
the size of the shocks and their persistence. It estimated the regulated-price shock
to be 1.7 percent and estimated the inertia from past shocks to be 4.2 percent, of
which two-thirds was to be accepted, resulting in a further adjustment of 2.8
percent. Then the central bank added these two numbers to the previously
announced target of 4 percent to get an adjusted inflation target for 2003 of 8.5
percent (¼ 4% þ 1.7% þ 2.8%). The adjusted target was then announced in the
Can Inflation Targeting Work in Emerging Market Countries? 83

open letter sent to the minister of finance in January 2003, which also explained
that getting to the nonadjusted target of 4 percent too quickly would entail far
too high a loss of output. Specifically, the letter indicated that an attempt to
achieve an inflation rate of 6.5 percent in 2003 would be expected to entail a de-
cline of 1.6 percent of GDP, while trying to achieve the nonadjusted target of 4
percent would be expected to lead to an even larger decline of GDP of 7.3 percent.
The procedure followed by the Banco Central do Brasil in this instance was a
textbook case for central bank response to shocks in emerging-market countries.
First, the procedure had tremendous transparency, both in articulating why the
initial inflation target was missed, but also in how the central bank was respond-
ing to the shock and its plans to return to its longer-run inflation goal. This degree
of transparency helped minimize the loss of credibility and the need to adjust the
short-term inflation target. Second, the central bank recognized that not adjusting
the inflation target was just not credible because the market and the public clearly
recognized that inflation would overshoot the initial target. Thus adjusting the
target was absolutely necessary to retain credibility, because to do otherwise
would have just signaled to the markets that the central bank was unwilling to be
transparent. Third, by discussing alternative paths for the inflation rate and why
the particular path using the adjusted target was chosen, the central bank was
able to demonstrate that it was not what Mervyn King (1996) has referred to as
an ‘‘inflation nutter’’ who only cares about controlling inflation and not about out-
put fluctuations. By outlining that lower inflation paths would lead to large out-
put losses, the Banco Central do Brasil demonstrated that it was not out of touch
with the concerns of the public because it indeed does care about output losses,
just as the public and the politicians do. This procedure is then likely to promote
public and political support for central bank independence.
The outcome from this particular episode has been favorable. After the initial
spike, Brazil’s inflation rate and interest rates have been coming down rapidly.
From its level of 12.5 percent in 2002, the inflation rate fell to 9.3 percent by the
end of 2003, which is within the central bank tolerance range for the adjusted in-
flation target of 8.5 percent. Market expectations of inflation also dropped dramat-
ically and suggested that the market expected the central bank to meet its inflation
targets in 2004. The interbank policy interest rate also fell from 26.5 percent in
June 2003 to 16.3 percent in February 2004. Also, after the initial decline in GDP,
the Brazilian economy began growing again.
The Brazilian central bank was not able to do this on its own. As seen in this
discussion, an important component of a successful response to an inflation shock
of this magnitude is growing confidence in government fiscal policy. President
Lula surprised many of his detractors by supporting measures to maintain fiscal
discipline. He supported and got legislation passed in August 2003 to reform the
84 Frederic S. Mishkin

public pension system to make it fiscally sustainable. His government also pur-
sued conservative spending policies that resulted in a primary budget sur-
plus in 2003 of 4.3 percent of GDP, which was above the target of 4.25 percent
requested by the IMF. Of course, there are still concerns about how Brazil will
proceed in the future. A new central bank law that would strengthen the Banco
Central do Brasil’s independence and provide a firmer institutional basis for
inflation targeting has not yet been passed. Also, given the pressures on the Lula
government from the left, continuing fiscal responsibility is not completely as-
sured. The very high real interest rates in Brazil, which are the highest in all
of the inflation-targeting countries, suggests that the credibility of the inflation-
targeting regime in Brazil is still far from complete and so the central bank needs
to pursue a highly restrictive monetary policy in order to keep inflation under
control.
The examples of Chile and Brazil illustrate that inflation targeting is indeed fea-
sible in emerging-market economies, despite their more complicated political and
economic environment. Inflation targeting has been able to provide a strong nom-
inal anchor that can keep inflation expectations in check. However, this requires
not only a focus on good communication and transparency by the central bank,
but also supportive policies to develop strong fiscal, financial, and monetary insti-
tutions. The Brazilian case (and the previously cited evidence in Batini and Lax-
ton, forthcoming) suggests that these policies do not have to be fully in place
when inflation targeting is adopted for it to produce good macroeconomic out-
comes, but it is important to have a strong commitment to develop these institu-
tions and a continuing vigilance that fiscal, financial, and monetary institutions
will continue to be strengthened.

4.5 Dealing with Exchange Rate Fluctuations

The special features of emerging-market economies outlined in the first section of

the paper suggests that emerging-market countries are likely to have greater con-
cerns about exchange rate fluctuations than advanced countries do. Because of
their past history and lower credibility in their ability to keep inflation under con-
trol, emerging-market countries are more likely to find that depreciations lead to a
rise in inflation as a result of the pass-through from higher import prices and
greater demand for exports.
In addition, liability dollarization means that depreciations are particularly
dangerous because they can trigger a financial crisis along the lines suggested in
Mishkin (1996, 1999). These countries have much of their debt denominated in
foreign currency and when the currency depreciates, this increases the debt bur-
den of domestic firms. Since assets are typically denominated in domestic cur-
Can Inflation Targeting Work in Emerging Market Countries? 85

rency and so do not increase in value, there is a resulting decline in net worth.
This deterioration in balance sheets then increases adverse selection and moral
hazard problems, which leads to financial instability and a sharp decline in in-
vestment and economic activity. This mechanism explains why the currency crises
in Chile in 1982, Mexico in 1994–1995, East Asia in 1997, Ecuador in 1999, Turkey
in 2000–2001 and Argentina in 2001–2002 pushed these countries into full-fledged
financial crises, which had devastating effects on their economies.
The potential devastating impact of currency depreciation on the financial sys-
tem and the possibility that it will lead to a burst of inflation provides a reason
for central banks in emerging-market countries not to pursue benign neglect of ex-
change rates even when they are inflation targeting. In order to prevent sharp
depreciations of their currencies, which can destroy balance sheets and trigger a
financial crisis, central banks in these countries may have to smooth exchange
rate fluctuations. However, there is a danger that monetary policy, even under an
inflation-targeting regime, may put too much focus on limiting exchange rate
movements. The possibility of ‘‘fear of floating’’ (Calvo and Reinhart 2002) when
emerging-market countries engage in inflation targeting is a real one.
The first problem with too strong a focus on limiting exchange rate movements
is that it runs the risk of transforming the exchange rate into a nominal anchor
that takes precedence over the inflation target. This has been a problem recently
in Hungary, a transition country that has an exchange rate target as part of its
inflation-targeting regime ( Jonas and Mishkin 2005). For example, when the Hun-
garians adopted inflation targeting in July 2001, they retained an exchange rate
band of G15 percent. Pursuing two nominal objectives could result in a situation
where one objective will need to be given preference over the second, but without
a clear guidance on how such conflict would be resolved. This is likely to make
monetary policy less transparent and hinder the achievement of the inflation tar-
get. Indeed, this is what has happened in Hungary. In January 2003, the forint
appreciated to the upper end of the band, and speculation about the revaluation
of the parity resulted in a sharp acceleration of capital inflows, forcing the Na-
tional Bank of Hungary to respond by cutting interest rates by two percentage
points and intervening heavily in the foreign exchange market. The National
Bank of Hungary is reported to have bought more than five billion euros, increas-
ing international reserves by 50 percent and base money by 70 percent.13 Even
though the National Bank of Hungary subsequently began to sterilize this huge
injection of liquidity, market participants then assumed that maintaining the ex-
change rate band would have a priority over the inflation target and expected in-
flation in 2003 to exceed the National Bank of Hungary’s inflation target by 5
percentage points.14 The outcome was that in 2003 Hungary did indeed overshoot
its target, but by a somewhat smaller 2.2 percentage points.15
86 Frederic S. Mishkin

The second problem resulting from too strong a focus on limiting exchange rate
fluctuations is that the impact of changes in exchange rates on inflation and out-
put can differ substantially depending on the nature of the shock that causes the
exchange rate movement. Different monetary policy responses therefore depend
on the nature of the shock. If the domestic currency depreciates because of a pure
portfolio shock, inflation is likely to rise and the appropriate response to keep it
under control is to tighten monetary policy and raise interest rates. In emerging-
market countries that have substantial liability dollarization, tightening monetary
policy to prevent a sharp depreciation may be even more necessary to avoid fi-
nancial instability (for the reasons mentioned). On the other hand, if the exchange
rate depreciation occurs because of real shocks, the impact is less likely to be infla-
tionary and a different monetary policy response is warranted, but even here it
depends on the nature of the shock. A negative terms-of-trade shock, which low-
ers demand for exports, reduces aggregate demand and is thus likely to be defla-
tionary. In this situation, the correct interest rate response is to lower interest rates
to counteract the drop in aggregate demand, and not to raise interest rates.
The mistakes that the Chilean central bank made in 1998 illustrate how serious
the second problem can be. Chile’s inflation-targeting regime also included a fo-
cus on limiting exchange rate fluctuations by having an exchange rate band with
a crawling peg that was (loosely) tied to lagged domestic inflation. Instead of eas-
ing monetary policy in the face of the negative terms-of-trade shock, the central
bank raised interest rates sharply and even narrowed its exchange rate band. In
hindsight, these decisions were a mistake: the inflation target was undershot and
the economy entered a recession for the first time in the 1990s.16 With this out-
come, the central bank came under strong criticism for the first time since it had
adopted its inflation-targeting regime in 1990, weakening support for the inde-
pendence of the central bank and its inflation-targeting regime. During 1999, the
central bank did reverse course, easing monetary policy by lowering interest rates
and allowing the peso to decline, and in May 2000 it revised its inflation-targeting
regime to reduce its focus on exchange rates.
The conclusion from the discussion here is that there is a rationale for central
banks in emerging-market countries to smooth exchange rates, but they can go to
far. To cope with potential problems of financial instability but preserve the focus
on inflation control, central banks could increase the transparency of any inter-
vention in the foreign exchange market by making it clear to the public that the
intervention’s purpose is to smooth excessive exchange rate fluctuations and not
to prevent the exchange rate from reaching its market-determined level over
longer horizons. However, continuing exchange market interventions, particu-
larly unsterilized ones, are likely to be counterproductive because they are not
transparent. Instead, exchange rate smoothing via changes in the interest rate in-
strument will be more transparent and indicate that the nominal anchor continues
Can Inflation Targeting Work in Emerging Market Countries? 87

to be the inflation target, not the exchange rate. Central banks could also explain
to the public the rationale for exchange rate intervention in a manner analo-
gous to that for interest-rate smoothing—that is, as a policy aimed not at resisting
market-determined movements in an asset price, but at mitigating potentially
destabilizing effects of abrupt and sustained changes in that price.
It is also important for central banks to recognize that the pass-through from ex-
change rate changes to inflation is likely to be regime dependent. After a sus-
tained period of low inflation engineered by an inflation-targeting regime, the
effect of the exchange rate on the expectations-formation process and price-setting
practices of households and firms in the economy is likely to fall.17 Thus, inflation
targeting is likely to help limit the pass-through from exchange rates to inflation
and thus the view that a currently high pass-through is a barrier to successful in-
flation targeting is unwarranted.

4.6 How Can the IMF Help?

A theme in both this and my recent paper with Calvo (Calvo and Mishkin 2003) is
that developing strong fiscal, financial, and monetary institutions is the key to
successful macroeconomic policy. One way the IMF can help emerging-market
countries who choose to inflation target is to provide them with better incentives
to develop these institutions. Instead of trying to impose a large number of condi-
tions on countries, as when the IMF lending program is put into effect, the IMF
can provide the right incentives by being more willing to extend programs ex
ante to countries who are making a serious attempt at reform, while saying no to
governments who are unwilling to do so.
In addition, the IMF conditions for evaluating monetary policy under its pro-
grams can be modified to reflect the special features of inflation-targeting regimes.
In the past, a key element of IMF conditionality was ceilings on the growth rate of
net domestic assets. Unfortunately, net domestic assets conditionality, which is
derived under the IMF’s financial programming framework, is based on an out-
dated theory, the monetary approach to the balance of payments (see Mussa and
Savastano 1999), which requires that the growth rate of monetary aggregates is
closely linked to inflation. However, the link between monetary aggregates and
inflation is almost always found to be very weak when inflation rates are reason-
ably low, as is the case for emerging market countries that have adopted inflation
targeting. As a result, targets for net domestic asset targets are likely to lead to in-
appropriate setting of monetary policy instruments and are likely to decrease
monetary policy transparency.
In an inflation-targeting regime, it seems natural to replace net domestic asset
conditionality with assessment of the country’s inflation performance. Indeed,
this is what the IMF moved to in evaluating monetary policy under its program
88 Frederic S. Mishkin

for Brazil when inflation targeting was adopted in 1999. The IMF program has
conducted quarterly reviews on how well Brazil has done in meeting its infla-
tion targets, but there is still a problem in that the IMF evaluation is essentially
backward-looking (Blejer, Leone, Rabanal, and Schwartz 2001).
Inflation targeting is inherently forward-looking, so the issue arises as to how
IMF conditionality might be modified to be more forward-looking. One approach
would be for the IMF to monitor monetary policy institutions. Specifically, the
IMF conditions could focus on the degree of central bank independence, whether
the central bank mandate focuses on price stability as the long-run overriding
goal of monetary policy, and whether transparency and accountability of the cen-
tral bank is high. As part of this monitoring, the IMF could conduct a careful as-
sessment of central bank procedures including the legitimacy of its forecasting
process and whether it provides adequate explanations for misses of its inflation
targets. In a sense this shift in approach is similar to the shift in approach that has
occurred in bank supervision in recent years. In the past, bank supervision was
also quite backward-looking in that it focused on the current state of banks’ bal-
ance sheets. However, this backward-looking approach is no longer adequate in
today’s world, in which financial innovation has produced new markets and
instruments that make it easy for banks and their employees to make huge bets
easily and quickly. In this new financial environment, a bank that is quite healthy
at a particular point in time can be driven into insolvency extremely rapidly by
trading losses, as forcefully demonstrated by the failure of Barings in 1995. Thus
bank examinations have now become far more forward-looking and place much
greater emphasis on evaluating the soundness of a bank’s management processes
with regard to controlling risk. Similarly, the IMF could shift its conditionality
to focus on the management processes in central banks to keep inflation under
control.

4.7 Concluding Remarks

It has been my honor to have Guillermo Calvo as a close friend for over twenty
years, ever since he helped recruit me to come to Columbia University. Guillermo
has always cared deeply about applying economic analysis to promote better
outcomes in emerging-market countries and his work has been an inspiration to
me. This paper expands on many themes in Guillermo’s research. The bottom
line from the analysis here is that inflation targeting can be an effective tool for
emerging market countries to manage their monetary policy. However, to ensure
that inflation targeting produces superior macroeconomic outcomes, emerging
market-countries would benefit by focusing even more attention on institutional
development, while international financial institutions like the IMF can help by
providing these countries with better incentives to engage in this development.
Can Inflation Targeting Work in Emerging Market Countries? 89

The hope is that this will lead to better economic performance in these countries
over the coming years.

Notes

1. Although I suspect that Guillermo may be more favorable now, but you will have to ask him.
2. The main concern is fluctuations in the real exchange rate but, as is well known (see Mussa 1986),
the real exchange rate shows high correlation with the nominal exchange rate.
3. For a discussion of fiscal reforms, see Perry, Whalley, and McMahon (2000) and Tanzi (2000).
4. See Mishkin (2003) for a description of which financial policies help prevent financial crises in
emerging market countries.
5. For example, recent research, summarized in Demirgüç-Kunt and Kane (2002), suggests that the
spread of deposit insurance in emerging market countries has made banking crises more likely.
Other—rather mixed—evidence about the incidence of moral hazard at the international level is pro-
vided by Dell’Ariccia, Gödde, and Zettelmeyer (2000) and Jeanne and Zettelmeyer (2001).
6. See Mishkin (1996) for how this occurred in Mexico and Garber (1999) for a discussion of how the
prudential regulations for Mexican banks requiring a matched book did not protect them from cur-
rency risk.
7. Recent papers by Caballero and Krishnamurthy (2002) and Jeanne (2002) suggest that de-
dollarization may require a major overhaul of the domestic financial sector.
8. There is an important distinction between goal and instrument independence (Debelle and Fischer
1994 and Fischer 1994). Instrument independence is the ability of the central bank to set monetary pol-
icy instruments without government interference, while goal independence means that the monetary
authorities set the goals for monetary policy. The standard view in the literature is that central banks
should have instrument but not goal independence.
9. However, except for the Great Depression, advanced countries have not been hit by equally large
shocks as in Argentina. Thus the stability of the central bank in advanced countries may be partly
explained by the size of the shocks and, in particular, the general absence of sudden stops.
10. For an overall comparison of the inflation targeting experience in Chile and Brazil, see Schmidt-
Hebbel and Werner (2002).
11. For those who don’t know this expression, a poster child is a beautiful but disadvantaged child put
on a poster (or advertisement) for a charity in order to stimulate donations.
12. See Mishkin and Savastano (2002).
13. See Morgan (2003).
14. Analysts have interpreted this as an evidence that the National Bank of Hungary is determined to
maintain the currency band even at the cost of temporary higher inflation. See IMF (2002).
15. Israel also had a fairly tight exchange rate band during the early years of its inflation targeting re-
gime and this caused it to experience similar difficulties. See Bernanke, Laubach, Mishkin, and Posen
(1999) and Leiderman (2000).

16. Given its location in Latin America, Chile’s central bank did have to worry about a loss of credibil-
ity from exchange rate deprecation. Also, because Chile encountered a sudden stop of capital inflows
at the time, the ability of the Chilean central bank to pursue countercyclical policy was limited.
90 Frederic S. Mishkin

However, although lowering interest rates in 1998 may not have been as attractive an option as it was
in a country like Australia, which responded to the terms-of-trade shock by easing monetary policy,
the sharp rise in the policy interest rate in 1998 was clearly a policy mistake.
17. See Muinhos (2001), Frenkel (2002), and Minella, de Freitas, Goldfajn, and Muinhos (2002, 2003).

References

Batini, Nicoletta, and Douglas Laxton. Forthcoming. ‘‘Under What Conditions Can Inflation Targeting
Be Adopted? The Experience of Emerging Markets.’’ In Monetary Policy Under Inflation Targeting, eds.
Frederic S. Mishkin and Klaus Schmidt-Hebbel. Santiago: Central Bank of Chile.
Bernanke, B., T. Laubach, F. Mishkin, and A. Posen. 1999. Inflation Targeting: Lessons from the Interna-
tional Experience. Princeton, NJ: Princeton University Press.
Bernanke, Ben S., and Frederic S. Mishkin. 1997. ‘‘Inflation Targeting: A New Framework for Monetary
Policy?’’ Journal of Economic Perspectives 11, no. 2: 97–116.
Blejer, Mario I., Alfredo M. Leone, Pau Rabanal, and Gerd Schwartz. 2001. ‘‘Inflation Targeting in the
Context of IMF-Supported Adjustment Programs.’’ Working Paper no. 01/31, IMF, Washington, D.C.
Brash, Donald T. 2000. ‘‘Inflation Targeting in New Zealand, 1988–2000.’’ Speech to the Trans-Tasman
Business Cycle, Melbourne, Australia, February 9.
Burnside, Craig, Martin Eichenbaum, and Sergio Rebelo. 2001. ‘‘Prospective Deficits and the Asian
Currency Crisis.’’ Journal of Political Economy 109, no. 6: 1155–1197.
Caballero, Ricardo J., and Arvind Krishnamurthy. 2002. ‘‘Excessive Dollar Debt: Financial Develop-
ment and Underinsurance.’’ Mimeo., MIT, Cambridge, MA.
Calomiris, Charles, and Andrew Powell. 2000. ‘‘Can Emerging Market Bank Regulators Establish Cred-
ible Discipline? The Case of Argentina, 1992–1999.’’ Working Paper No. 7715, NBEK, Cambridge, MA.
Calvo, Guillermo A. 1999a. ‘‘Capital Markets and the Exchange Rate.’’ Mimeo., University of
Maryland.

Calvo, Guillermo A. 1999b. ‘‘Contagion in Emerging Markets: When Wall Street is the Carrier.’’ Mimeo.,
University of Maryland.

Calvo, Guillermo A. 2001. ‘‘Capital Markets and the Exchange Rate: With Special Reference to the Dol-
larization Debate in Latin America.’’ Journal of Money, Credit, and Banking 33, no. 2: 312–334.
Calvo, Guillermo A., Alejandro Izquierdo, and Ernesto Talvi. 2002. ‘‘Sudden Stops, the Real Exchange
Rate and Fiscal Sustainability: Argentina’s Lessons.’’ Working Paper No. 469, IDB, Washington, D.C.
Calvo, G., and E. Mendoza. 2000. ‘‘Capital-Market Crises and Economic Collapse in Emerging Markets:
An Informational-Frictions Approach.’’ Mimeo., Duke University and University of Maryland.
Calvo, Guillermo, and Frederic S. Mishkin. 2003. ‘‘The Mirage of Exchange Rate Regimes for Emerging
Market Countries.’’ Journal of Economic Perspectives 17, no. 4: 99–118.
Calvo, Guillermo A., and Carmen M. Reinhart. 2000. ‘‘When Capital Flows Come to a Sudden Stop:
Consequences and Policy.’’ In Reforming the International Monetary and Financial System, eds. Peter B.
Kenen and Alexander K. Swoboda, 395–427. Washington, D.C.: IMF.
Calvo, Guillermo A., and Carmen M. Reinhart. 2002. ‘‘Fear of Floating.’’ Quarterly Journal of Economics
117, no. 2: 379–408.
Can Inflation Targeting Work in Emerging Market Countries? 91

Calvo, Guillermo A., and Carlos A. Végh. 1996. ‘‘From Currency Substitution to Dollarization and Be-
yond: Analytical and Policy Issues.’’ In Money, Exchange Rates, and Output, ed. Guillermo A. Calvo,
153–175. Cambridge, MA: MIT Press.
Caprio, Gerald, Jr. 1998. ‘‘Banking on Crises: Expensive Lessons from Recent Financial Crises.’’ Work-
ing Paper No. 1979, World Bank, Washington, D.C.
Caprio, Gerald Jr., and D. Klingebiel. 1999. ‘‘Episodes of Systemic and Borderline Financial Crises.’’
Mimeo., World Bank, Washington, D.C.
Cukierman, Alex. 1992. Central Bank Strategy, Credibility, and Independence: Theory and Evidence. Cam-
bridge, MA: MIT Press.
Cukierman, Alex. 2006. ‘‘Central Bank Independence and Monetary Policymaking Institutions: Past,
Present, and Future.’’ Working Paper No. 360, Central Bank of Chile.
Debelle, Guy, and Stanley Fischer. 1994. ‘‘How Independent Should a Central Bank Be?’’ In Goals,
Guidelines, and Constraints Facing Monetary Policymakers, Conference Series 38, ed. Jeffrey C. Fuhrer,
195–221. Boston: Federal Reserve Bank of Boston.
De Gregorio, Jose, Sebastian Edwards, and Rodrigo O. Valdes. 2000. ‘‘Controls on Capital Inflows: Do
They Work?’’ Working Paper No. w7645, NBER, Cambridge, MA.
Dell’Ariccia, Giovanni, Isabel Gödde, and Jeromin Zettelmeyer. 2000. ‘‘Moral Hazard and International
Crisis Lending: A Test.’’ Mimeo., International Monetary Fund, Washington, D.C.
Demirgüç-Kunt, Asli, and Edward J. Kane. 2002. ‘‘Deposit Insurance around the Globe: Where Does it
Work?’’ Journal of Economic Perspectives 16, no. 2: 175–195.
Edwards, Sebastian. 1999. ‘‘How Effective are Controls on Capital Inflows? An Evaluation of Chile’s
Experience.’’ Mimeo., UCLA.
Eichengreen, Barry, Ricardo Hausmann, and Ugo Panizza. 2002. ‘‘Original Sin: The Pain, the Mystery,
and the Road to Redemption.’’ Paper presented at the conference Currency and Maturity Matchmaking:
Redeeming Debt from Original Sin, IDB, Washington, D.C., November 21–22.

Fischer, Stanley. 1994. ‘‘Modern Central Banking.’’ In The Future of Central Banking: The Tercentenary
Symposium of the Bank of England, eds. Forrest Capie, Charles A. E. Goodhart, Stanley Fischer, and Nor-
bert Schnadt. Cambridge, MA: Cambridge University Press.
Forder, James. 2000. ‘‘Central Bank Independence and Credibility: Is There a Shred of Evidence? Re-
view.’’ International Finance 3, no. 1: 167–185.
Fraga, Arminio, Goldfajn Ilan, and André Minella. 2003. ‘‘Inflation Targeting in Emerging Market
Economies.’’ In NBER Macroeconomics Annual 2003, eds. Mark Gertler and Kenneth Rogoff, 365–400.
Cambridge, MA: MIT Press.
Frenkel, Jacob. 2002. ‘‘The Transition to Inflation Targeting.’’ In Stabilization and Monetary Policy: The In-
ternational Experience, Banco de Mexico, 103–111. Mexico City: Banco de Mexico.
Garber, Peter. 1999. ‘‘Hard-Wiring to the Dollar: From Currency Board to Currency Zone.’’ In Global
Markets Research 2, no. 1: 67–76.

Jeanne, Olivier. 2002. ‘‘Why Do Emerging Economies Borrow in Foreign Currency?’’ Paper presented at
the conference Currency and Maturity Matchmaking: Redeeming Debt from Original Sin, IDB, Washington,
D.C., November 21–22.
Jeanne, Olivier, and Jeromin Zettelmeyer. 2001. ‘‘International Bailouts, Moral Hazard, and Condition-
ality.’’ Economic Policy 33: 409–432.
92 Frederic S. Mishkin

Jonas, Jiri, and Frederic S. Mishkin. 2005. ‘‘Inflation Targeting in Transition Countries: Experience and
Prospects.’’ In Inflation Targeting, Michael Woodford, ed., 353–413. Chicago: University of Chicago
Press.
Kaminsky, Graciela L., Carmen M. Reinhart, and Carlos A. Végh. 2003. ‘‘The Unholy Trinity of Finan-
cial Contagion.’’ Journal of Economic Perspectives 17, no. 4: 51–74.
Kaminsky, Graciela L., Carmen M. Reinhart, and Carlos A. Végh. 2004. ‘‘When it Rains It Pours: Pro-
cyclical Macropolicies and Capital Flows.’’ In NBER Macroeconomics Annual 2004, eds. Ben S. Bernanke
and Mark Gertler, 11–53. Cambridge, MA: NBER.
King, Mervyn. 1996. ‘‘How Should Central Banks Reduce Inflation? Conceptual Issues.’’ In Achieving
Price Stability, Federal Reserve Bank of Kansas City, MO. 53–91.

Landerretche, Oscar, Felipe Morandé, and Klaus Schmidt-Hebbel. 1999. ‘‘Inflation Targets and Stabili-
zation in Chile.’’ Working Paper No. 55, Central Bank of Chile.
Leidermann, Leonardo, and Gil Byfman. 2000. ‘‘Inflation Targeting Under a Crawling Band Exchange
Rate Regime: Lessons from Israel.’’ In Inflation Targeting in Practice: Strategic and Operational Issues and
Application to Emerging Market Economies, eds. Mario I. Blejer, Alainlze, Alfredo M. Leone, and Sergio
Werlang, 70–79. Washington, D.C.: International Monetary Fund.
Masson, P., M. Savastano, and S. Sharma. 1997. ‘‘The Scope for Inflation Targeting in Developing
Countries.’’ Working Paper No. 97/130, IMF, Washington, D.C.
Minella, André, Paulo Springer de Freitas, Ilan Goldfajn, and Marcelo Kfoury Muinhos. 2002. ‘‘Inflation
Targeting in Brazil: Lessons and Challenges.’’ Working Paper 53, Banco Central do Brazil.

———. 2003. ‘‘Inflation Targeting in Brazil: Constructing Credibility under Exchange Rate Volatility.’’
Working Paper 77, Banco Central do Brazil.

Mishkin, Frederic S. 1996. ‘‘Understanding Financial Crises: A Developing Country Perspective.’’ In

Annual World Bank Conference on Development Economics, eds. Michael Bruno and Boris Pleskovic, 29–
62. Washington, D.C.: World Bank.
Mishkin, Frederic S. 1999. ‘‘Lessons from the Asian Crisis.’’ Journal of International Money and Finance 18,
no. 4: 709–723.
Mishkin, Frederic S. 2003. ‘‘Financial Policies and the Prevention of Financial Crises in Emerging Mar-
ket Countries.’’ In Economic and Financial Crises in Emerging Market Countries, ed. Martin Feldstein, 93–
130. Chicago: University of Chicago Press.
Mishkin, Frederic S., and Miguel Savastano. 2001. ‘‘Monetary Policy Strategies for Latin America.’’ Jour-
nal of Development Economics 66, no. 2: 415–444.

Mishkin, Frederic S., and Miguel Savastano. 2002. ‘‘Monetary Policy Strategies for Emerging Market
Countries: Lessons from Latin America.’’ Comparative Economic Studies 44, no. 2: 45–83.

Mishkin, Frederic S., and Klaus Schmidt-Hebbel. 2007. ‘‘Does Inflation Targeting Make a Difference?’’
In Monetary Policy Under Inflation Targeting, eds. Frederic S. Mishkin and Klaus Schmidt-Hebbel, 291–
372. Santiago: Central Bank of Chile.
Morgan, J. P. 2003. Emerging Europe, Middle East & Africa Weekly, Newsletter, January 31.
Muinhos, Marcelo Kfoury. 2001. ‘‘Inflation Targeting in an Open Financially Integrated Emerging
Economy: The Case of Brazil.’’ Working Paper 26, Banco Central do Brazil.
Mussa, Michael. 1986. ‘‘Nominal Exchange Rate Regimes and the Behavior of Real Exchange Rates:
Evidence and Implications.’’ In Real Business Cycles, Real Exchange Rates, and Actual Policies, eds. Karl
Can Inflation Targeting Work in Emerging Market Countries? 93

Brunner and Allan Meltzer, Carnegie-Rochester Conference Series on Public Policy 25, 117–213. Amster-
dam: North-Holland.
Mussa, Michael, and Miguel Savastano. 1999. ‘‘The IMF Approach to Economic Stabilization.’’ In NBER
Macroeconomics Annual 1999, eds. Ben S. Bernanke and Julio Rotemberg, 79–121. Cambridge, MA: MIT
Press.
Perry, Guillermo, John Whalley, and Gary McMahon. 2000. Fiscal Reform and Structural Change in Devel-
oping Countries, Vol. 2. New York: Palgrave Macmillan.
Sargent, Thomas, and Neil Wallace. 1981. ‘‘Some Unpleasant Monetarist Arithmetic.’’ Federal Reserve
Bank of Minneapolis Quarterly Review 5, no. 3: 1–17.
Schmidt-Hebbel, K., and A. M. Werner. 2002. ‘‘Inflation Targeting in Brazil, Chile, and Mexico: Perfor-
mance, Credibility, and the Exchange Rate.’’ Economia 2, no. 2: 30–89.
Svensson, Lars O. 1997. ‘‘Inflation Forecast Targeting: Implementing and Monitoring Inflation Targets.’’
European Economic Review 41: 1111–1146.

Tanzi, Vito. 2000. ‘‘Rationalizing the Government Budget or Why Fiscal Policy Is so Difficult.’’ In Eco-
nomic Policy Reform: The Second Stage, ed. Anne Krueger, 435–452. Chicago: University of Chicago
Press.
Woodford, Michael. 1994. ‘‘Monetary Policy and Price Level Determinacy in a Cash-in-Advance Econ-
omy.’’ Economic Theory 4: 345–380.
Woodford, Michael. 1995. ‘‘Price Level Determinacy without Control of a Monetary Aggregate.’’
Carnegie-Rochester Conference Series on Public Policy 43: 1–46.
5 Why Should Emerging
Economies Give up National
Currencies? A Case for
‘‘Institutions Substitution’’

Enrique G. Mendoza

5.1 Introduction

The severity and contagion of emerging markets crises during the last twelve
years are an unprecedented phenomenon particular to the era of fast-moving, glo-
balized financial markets. Why have emerging economies—many of which
embarked in far-reaching programs of stabilization and market-oriented reform
during the 1990s—fared so poorly? A simplistic explanation is ‘‘bad government,’’
which results in corrupt, incompetent, or dysfunctional political and legal institu-
tions. Indeed, many emerging economies are plagued by problems such as ram-
pant tax evasion, crony capitalism, non-functioning or nonexistent bankruptcy
procedures, widespread financial fraud, corrupt judiciary systems, acute political
strife, and so on. Yet the findings of the growing body of research on emerging-
markets crises show that even ‘‘good’’ governments may have a high degree of
vulnerability to large, sudden reversals of international capital flows. At the core
of this vulnerability lie two fundamental elements: the lack of credibility of do-
mestic policy-making institutions in emerging economies and the imperfections
of the globalized capital markets.
The lack of credibility of economic policies in emerging countries can be attrib-
uted again to bad government, but in these countries even the policies of well-
intentioned and benevolent governments would be questioned. The reason is
that, for all the emphasis on new features of capital markets crises in the era of fi-
nancial globalization, one key aspect of these crises is old news for citizens of
emerging countries: government policy displays a high degree of time inconsis-
tency. The currency pegs and managed exchange rates announced and imple-
mented in the late 1980s and early 1990s were offered as commitments to
stabilization policies that would result in sustainable low inflation. As time
passed, however, the policy adjustments needed to stick to these commitments
became too costly to bear, so adjustments were put off, leading to unsustainable
96 Enrique G. Mendoza

levels of exchange rates and financial crises. These crises were then resolved by
government policies that engineered large redistributions of wealth within the
private sector, and from the private sector to the public sector and/or the foreign
sector. Governments have done this several times before the recent emerging mar-
kets crises (witness, for a short list of examples, Chile 1982, Mexico 1976 and 1982,
or Argentina 1990). At the end of each of these debacles, the repenting govern-
ment promised not to do it again, and to show its commitment it started over
with a new set of monetary arrangements it swore to comply with—until the
next crisis.
The adverse macroeconomic effects caused by lack of policy credibility and by
imperfect world capital markets are the subject of two well-established but largely
disconnected strands of the international macroeconomics literature. Recently,
however, some of the literature on emerging markets crises began to focus on the
interaction between these two elements, showing how they combine to create a
transmission mechanism that can greatly amplify the effects of adverse exogenous
shocks. In an environment in which economic policy is not credible, policy
reforms like privatization of public enterprises and financial institutions, or stabi-
lization policies like the widespread use of exchange rate management to reduce
inflation, may have fueled the period of bonanza that in most cases preceded
balance-of-payments crises (see Calvo and Mendoza 1996 and Mendoza and
Uribe 2000). Financial globalization may have created the possibility of large
swings in capital flows driven by self-fulfilling expectations, imperfect and/or
costly information, and other similar contagion-prone frictions that can cause
large international movements of financial capital despite a country’s ‘‘strong fun-
damentals’’ (see Cole and Kehoe 1996, Calvo and Mendoza 2000a and 2000b,
Mendoza 2002 and 2005, and Chang and Velasco 2000).
The interaction between lack of credibility and capital market imperfections is
reflected in the fact that a noncredible government is typically an implicit feature
of the mechanisms that drive emerging-markets crises caused by financial fric-
tions. The mechanisms used in many theoretical models to date assume that
emerging economies feature short-term dollar-denominated debt, unhedged cur-
rency risks, collateral or liquidity requirements limiting the ability to contract
foreign debt, implicit government guarantees offered to domestic banks or bor-
rowers, and macroeconomic policy environments that are highly unstable and
costly to evaluate. These are all features that can be attributed, at least in part, to
a government’s lack of credibility at home and in world capital markets.
Examples of how the interaction between financial frictions and lack of credibil-
ity can create a business cycle transmission mechanism capable of reproducing
the sudden stop phenomenon are provided in Mendoza (2001, 2002, and 2006).
Mendoza (2002) examines an economy with incomplete insurance markets in
Why Should Emerging Economies Give up National Currencies? 97

which uncertain duration of economic policy has real effects via a mechanism akin
to a stochastic tax distortion, and foreign creditors impose liquidity requirements
on domestic borrowers. If the economy has a sufficiently large external debt, a
sudden policy reversal (even if only a low-probability event) triggers a large re-
versal in capital inflows, a suddenly binding foreign borrowing constraint, and
sharp downward adjustments in economic activity and domestic relative prices.
Mendoza (2001) develops a monetary variant of this model and shows that adopt-
ing a hard currency can yield large welfare gains by eliminating the risk of deval-
uation (in other words, the risk of a reversal of exchange rate policy) and by
relaxing liquidity requirements on foreign borrowing.
The main point of this article is to argue that abandoning national currencies to
adopt a hard currency can be an effective policy for emerging economies to deal
simultaneously with the lack of credibility of domestic financial policies and the
imperfections of globalized capital markets. This two-prong approach to the prob-
lem contrasts with several of the policy recommendations that have been put for-
ward to deal with emerging-markets crises, which typically aim to tackle either
the weaknesses of domestic policy-making institutions or the imperfections of
world capital markets. Those that view weak domestic institutions as the culprit
tend to favor policies that can end pervasive moral hazard problems, including
policies favoring floating exchange rate regimes, committing international finan-
cial institutions to refrain from providing large bail-outs, and allowing for orderly
default procedures by sovereign lenders (see Lerrick and Meltzer 2001). Those
that emphasize the imperfections of global capital markets tend to support poli-
cies that provide coordinated financial assistance and minimize the risk of finan-
cial contagion by aiming to keep countries at their sustainable levels of debt, or
by supporting emerging-market debt prices at or above nonfundamental crash
levels (see Calvo 2002 and Calvo, Izquierdo, and Talvi 2002).
This article builds its case for the adoption of hard currencies in emerging
economies by putting together, in an intuitive manner, arguments developed else-
where in the literature, particularly by Calvo (1998, 2000, and 2002), Calvo and
Mendoza (2000a and 2000b), Mendoza (2001, 2002, 2005, and 2006), and Mendoza
and Smith (2006). The article is intentionally short on the technical details con-
tained in these papers. The central idea is that abandoning national currencies
and adopting a hard currency, when seen from the perspective of emerging
economies facing noncredible policy-making institutions and imperfect globalized
capital markets, can be very beneficial because it does away with the lack of cred-
ibility of monetary and exchange rate policies, and it reduces the informational
frictions behind several mechanisms of financial contagion and sudden reversals
of capital flows (by rendering unnecessary the costly investment of knowledge
and resources that goes into assessing domestic monetary policies).
98 Enrique G. Mendoza

These arguments deviate from conventional arguments in the debate on the

adoption of hard currencies or the formation of currency areas, which tend to fo-
cus on whether it is wise to give up the use of independent monetary policy as a
tool for stabilization policy. (These often boil down to arguments similar to those
in the protracted debate on the significance of nominal rigidities and the effective-
ness of monetary policy tools.) The initial premise here is precisely that indepen-
dent monetary policy in the class of countries that concern this article is a cause
of concern, either because policy-making institutions in these countries are not ef-
ficient vehicles for stabilization policy, or because even if they were, the global
capital market may fail to assess domestic policies with precision and differentiate
them from those of other countries in which policies are mismanaged. The aim is
thus to think of abandoning national currencies as a vehicle for ‘‘institutions sub-
stitution,’’ a paraphrase of the ‘‘imports substitution’’ approach to economic devel-
opment that was dominant in developing countries during the 1960s.
The rest of the paper is organized as follows. Section 5.2 documents key fea-
tures of emerging-markets crises that point to the central role of noncredible pol-
icy and imperfect capital markets in causing the crises. Section 5.3 reviews some
analytical results that suggest that the globalization of (imperfect) financial mar-
kets had endogenous mechanisms that increased the vulnerability of emerging
economies to sudden stops. Section 5.4 summarizes an analytical framework for
studying economic fluctuations in a small open economy that features a transmis-
sion mechanism linking imperfections in world capital markets with lack of credi-
bility, or uncertain duration, of domestic economic policy. Section 5.5 draws
policy conclusions.

5.2 Sudden Stops and Contagion: Facts and Lessons from Emerging-Markets
Crises

Despite heated debate in the aftermath of the 1994 Mexican crash, there is wide
agreement now that the emerging-markets crises of the 1990s signaled the dawn
of a new era in capital-markets crises (see Calvo and Mendoza 1996 and 2000a).
This understanding was reached after observing two phenomena common to
these crises: the sudden stop phenomenon and the phenomenon of financial con-
tagion. A sudden stop consists of a sharp and sudden reversal in capital inflows,
a corresponding abrupt adjustment in the current account, and sharp declines in
production, absorption, and the relative prices of goods and financial assets. Sud-
den stops occurred in all of the emerging-markets crises, perhaps with the excep-
tion of the 1999 Brazilian crisis. Financial contagion took place when financial
markets that were seemingly unrelated to developments in a struggling emerging
Why Should Emerging Economies Give up National Currencies? 99

economy were nevertheless affected and suffered severe effects in terms of price
corrections and liquidity. One example was the ‘‘Tequila Effect,’’ by which the
Mexican 1994 crises triggered a sudden stop in Argentina. The most prominent
cases, however, are the Asian crisis of 1997 and the Russian crisis of 1998. Conta-
gion in the Russian case nearly crashed financial markets in industrial countries
and triggered a liquidity crisis that forced the Federal Reserve to cut interest rates
and coordinate the collapse of hedge fund long-term capital management
(LTCM).
Studies like those of Calvo, Izquierdo, and Talvi (2002), Calvo and Reinhart
(1999), and Milesi-Ferretti and Razin (2000) document in detail the features of the
sudden stops observed in the emerging countries that suffered financial crises in
the 1990s and the last two years. A document by the International Monetary
Fund (1999) describes the collapses in equity prices and the increase in their vola-
tility. Mendoza (2006) and Parsley (2001) show evidence of sharp changes in the
relative price of nontradable goods for Hong Kong, Korea, and Mexico.
Figures 5.1 to 5.3 provide a summary view of the stylized facts of sudden stops
for the cases of Argentina, Korea, Mexico, Russia, and Turkey. Figure 5.1 shows
recent time series data for each country’s current account as a share of GDP. Sud-
den stops are displayed in these plots as sudden, large swings of the current
account that in most cases exceeded five percentage points of GDP. Figure 5.2
shows data on consumption growth as an indicator of real economic activity.
These plots show that sudden stops are associated with a sharp collapse in the
real sector of the economy. Figure 5.3 provides information on two key finan-
cial indicators for each country, the price of domestic equity (valued in U.S. dol-
lars) and the spread of the yield in J. P. Morgan’s Emerging Markets Bond Index
Plus (EMBIþ) for each country relative to U.S. Treasury bills. Sudden stops
feature large declines in equity prices and sudden, sharp increases in EMBIþ
spreads, with equity prices often leading the surge of the spreads at the monthly
frequency.
Empirical studies of contagion disagree on how to define and measure financial
contagion, and on whether there is evidence in the data that contagion was com-
mon in emerging-markets crises (see, for example, Kaminsky and Reinhart 2000
and Rigobon 2002). Nevertheless, casual observation of figure 5.3—keeping in
mind the timing of the different crises that took place during the period plotted in
the chart—suggests that there was indeed some degree of financial contagion.
Moreover, Rigobon’s work shows that even when statistically accurate concepts
and measures of contagion are adopted, the data indicate that there were compo-
nents of contagion across financial markets during the Mexican and Russian
crises.
100 Enrique G. Mendoza

Figure 5.1
Current Account Balances in Percent of Gross Domestic Product. Source: International Financial Statis-
tics, IMF.

Underlying economic causes of the sudden stop and contagion phenomena can-
not be extracted from descriptions of the stylized facts like the one just summa-
rized. In the next two sections, this paper reviews analytical arguments making
the case that the uncertain duration of economic policy in emerging economies
and the imperfections of globalized capital markets could be the culprits. An im-
portant necessary condition for this hypothesis to hold is that global capital
inflows into emerging markets show an important component driven by factors
external to emerging economies. Calvo, Leiderman, and Reinhart (1996) showed
that, indeed, an important statistical predictor of the surge of capital inflows into
Why Should Emerging Economies Give up National Currencies? 101

Figure 5.2
Annual Growth Rates in Real Private Consumption Expenditures. Source: World Development Indica-
tors, World Bank.

emerging markets in the first half of the 1990s was the decline in U.S. interest
rates. More recently, one can observe in figure 5.4 two important changes in the
amount and composition of net private capital inflows into emerging economies
(see also Razin, Sadka, and Yuen 1998). First, there was a large and persistent re-
duction in total net inflows after the Asian and Russian crises. Total net inflows
were seven times higher at the peak of the capital inflows boom in 1996 than in
2000. Second, the composition of net inflows changed dramatically. Bank loans
never surged and became increasingly negative since 1997, while foreign direct in-
vestment (FDI) flows actually increased modestly and remained fairly stable. A
deeper analysis would show two additional key facts. First, that an important
element of the fall in capital inflows was the protracted retrenchment of the
emerging-economies bond market induced by the Russian crisis and Ecuador’s
default on its Brady bonds. Second, that the change in composition of the inflows
102 Enrique G. Mendoza

Figure 5.3
Equity Prices and Country Risk. Source: JP Morgan.
Why Should Emerging Economies Give up National Currencies? 103

Figure 5.3
(continued)

reflected also changes in the country distribution of inflows, as countries like

China or Mexico, recipients of large FDI inflows, fared much better than countries
relying on portfolio or bank loan flows (like Argentina or Ecuador).
The Asian and Russian crises naturally could account for a temporary fall in
capital inflows into the countries affected by these crises, but the trends shown in
figure 5.4 affected all emerging economies and seem to be persistent. From this
perspective, the drying up of capital inflows after Russia 1998 would seem to a
country like Argentina as exogenous as the boom that Calvo, Leiderman, and
Reinhart (1996) attributed to the fall in U.S. interest rates in the early 1990s. Thus,
emerging countries are clearly vulnerable to large and potentially long-lasting
swings in their ability to access global capital markets that are driven to a nontri-
vial extent by forces beyond their direct control.
104 Enrique G. Mendoza

Figure 5.4
Net Private Capital Flows to Emerging Markets

5.3 Rational Contagion in Global Capital Markets

Making the case that contagion and sudden stops are new phenomena in the era
of global capital markets requires an argument for justifying that financial global-
ization can indeed trigger large swings in international capital flows that may not
be justified by a country’s fundamentals. Calvo and Mendoza (2000a and 2000b)
presented three economic models supporting this argument.
The first model is a model of rational contagion in which global investors are
less willing to pay for relevant country-specific information as global capital mar-
ket integration progresses, and hence are more likely to react to rumors. In this
model, the global capital market consists of a large number of identical mean-
variance-optimizing investors who can choose whether or not to pay a fixed cost
and eliminate the idiosyncratic uncertainty of investment in a particular emerging
economy. For simplicity, all investment opportunities are ex ante identical and
their returns are independently and identically distributed.
How does the incentive to pay the fixed information cost (meaning the gain of
paying and designing an optimal portfolio using the updated information versus
investing blindly on the basis of the ex ante distribution of returns) vary as the
Why Should Emerging Economies Give up National Currencies? 105

number of emerging economies in the global capital market grows? We showed

that as long as the size of short positions that can be taken on specific countries is
limited, the incentive to pay the fixed information cost always reaches a point at
which it becomes a decreasing function of the number of countries in the market
(up to a lower bound at which it becomes independent of market size). The intu-
ition is that the return to blind investment is becoming risk-free as the number of
countries increases, while taking advantage of learning for sure about a low re-
turn on a particular country requires taking an increasingly large short position
on that country.
The second model that Calvo and Mendoza proposed makes an argument for
rational herding among mutual fund managers or investment bankers. Assume
that these managers and bankers are again identical mean-variance optimizers,
but now consider that they face an incentive structure that rewards them when
they produce above-market returns and punishes them when the opposite occurs.
We showed that if the marginal reward is larger than the marginal punishment,
there is a range of multiple optimal portfolios. Within this range, all investors
move in a herd to mimic whichever portfolio becomes the day’s preferred portfo-
lio. The reason is that holding this portfolio is optimal because investing more
(less) would yield a marginal cost higher (lower) than the marginal gain. More-
over, increasing the number of countries in which to invest (in the same fashion
as in the first example) widens the multiplicity range up to a point at which it
becomes independent of the size of the global capital market.
The third model starts where the fixed-information-cost analysis ended. That is,
given that information is costly and an individual investor’s incentives to pay for
it fall as the market grows, it makes sense to expect that the market will organize
itself into a group of ‘‘sophisticated’’ informed traders and a group of uninformed
agents that invest their funds with these sophisticated traders. Consider then a
set-up in which the uninformed try to extract information about a country’s fun-
damentals from the noisy information they get about the specialists’ trades. In-
formed investors take an action, say buying Argentine bonds, that is observable
to uninformed investors but that represents a combination of two variables, one
that is an accurate signal of Argentine fundamentals (so that the choice to buy Ar-
gentine bonds is increasing on this signal) and another variable that reflects fac-
tors particular to the informed traders and has a negative effect on the choice to
buy Argentine bonds (like margin calls informed traders may get from their own
creditors). For simplicity, the observed variable is modeled as the difference of the
fundamentals signal minus the investor-specific shock. Uninformed individuals
know the unconditional distributions of both (which are standard normal distri-
butions) while specialists know their exact values.
106 Enrique G. Mendoza

A standard signal extraction problem emerges in this setting: upon observing

the specialists’ trade, the uninformed compute the distribution of the fundamen-
tals signal conditional on the observed trade, and this conditional distribution has
a mean that is decreasing in the noise-to-signal ratio (meaning the ratio of the
variances of the noise variable relative to the fundamentals signal). Hence, if this
ratio is low—that is, if the variance of fundamentals is high relative to the vari-
ance of the noise variable—uninformed investors are more likely to act on the
observed trade because their conditional expectation that the trade reflects strong
fundamentals is higher. In a limiting case in which the noise-to-signal ratio
approaches zero even though the variance of the noise per se is large, uninformed
investors bet that observed trades always signal fundamentals even though they
in fact have an important ingredient of noise.
Unfortunately for emerging markets, a low noise-to-signal ratio is very likely
because of the relatively high cyclical variability of their fundamentals indicators,
such as the terms of trade or gross domestic product (see Mendoza 1995), com-
pared with the rare occurrence of widespread margin calls and liquidity crunches
à la the Russian crisis. The contagion of the Russian default that drove investors
into liquidating positions across emerging markets worldwide can be interpreted
then as the outcome of a low-probability, nonfundamentals shock—in other
words, large margin calls, which induced a sizable cut in the observed positions
of specialized investors in emerging markets that was then largely misinterpreted
by uninformed investors as a shock to the fundamentals of emerging economies.
Informational frictions are crucial for the three models reviewed. For instance, if
all investors are equally and perfectly informed, there would always be investors
willing to purchase the securities that investors who are facing margin calls are
liquidating at prices that cannot deviate much from the fundamentals prices. As-
set prices would then collapse only in the case of a true negative shock to funda-
mentals. Short-selling constraints and margin calls (which themselves can be
interpreted as resulting from informational frictions) are just as crucial. If unlim-
ited short selling were possible, for example, investors would never have reduced
incentives to acquire country-specific information at a fixed cost. Moreover, there
are reasons to believe that informational frictions are more pervasive in the con-
text of global capital flows to emerging markets than in other asset markets. This
is because of the amount and diversity of the information that needs to be gath-
ered on a large set of economic, social, and political factors, and the expertise
required to process and interpret it. There is also little room for economies of scale
in this information-gathering process: collecting information on determinants of
asset returns in Korea may be of some help in predicting returns in southeast
Asia but is of little help in fully understanding returns prospects in Indonesia and
is of no use in assessing returns for Latin America.
Why Should Emerging Economies Give up National Currencies? 107

5.4 Noncredible Policy, Financial Frictions, and Sudden Stops

The macroeconomic implications of the temporariness of noncredible government

policies is the subject of a large literature in international macroeconomics (Rebelo
and Vegh 1996 and Calvo and Vegh 1999 provide detailed surveys). The starting
point of this literature was the observation that in countries in which noncredible
stabilization programs were perceived as temporary in the 1970s and 1980s, con-
sumption boomed in the early stages of the programs and then collapsed at or
around the time at which the programs were abandoned. Calvo (1986) argued
that this could be explained via intertemporal substitution: if the public knows
with full certainty that inflation stabilization will be temporary, it is optimal to
substitute consumption away from future high-inflation periods into present low-
inflation periods. Helpman and Razin (1987) approached the issue focusing on
wealth effects resulting from the fiscal implications of temporary inflation stabili-
zations, modeling these wealth effects as intergenerational wealth, redistributions
that break the Ricardian equivalence principle. A temporary currency peg that
lowers current inflation redistributes wealth, favoring the generations that are
alive while the peg is in place, and working against those that arrive at the time
the peg collapses and in the periods that follow.
Mendoza and Uribe (2000) developed a stochastic model in which a noncredi-
ble currency peg is the source of uninsurable devaluation risk, which in turn is
the driving force of business cycles in a small open economy perfectly integrated
into international capital markets. Their analysis incorporates mechanisms similar
to Calvo’s intertemporal substitution effect and Helpman and Razin’s fiscal-
induced wealth effect, but in the context of an equilibrium business cycle model
in which the private sector formulates optimal plans with regard to consumption
of traded and nontraded goods, investment, labor, and money holdings.
The optimal decisions of the private sector in the Mendoza-Uribe model are dis-
torted by uninsurable devaluation risk in the following manner. First, uncovered,
risk-adjusted nominal interest parity implies that the domestic nominal interest
rate features an endogenous time- and state-dependent risk premium relative to
the world nominal interest rate (determined as a multiple of the rate at which the
public expects the currency to be devalued). The size of this premium depends on
the expected decline in wealth caused by the increase in seigniorage associated
with the collapse of the currency and the surge in inflation that follows. Second,
fluctuations in the nominal interest rate act as a random tax on current labor in-
come, future capital income, and the return on saving, because domestic money
is a risky asset that helps agents economize transactions costs in purchasing con-
sumption and investment goods (a similar outcome obtains in Mendoza 2001,
assuming instead that money enters directly in the utility function).
108 Enrique G. Mendoza

In the Mendoza-Uribe setup, a time-varying J-shaped probability of the aban-

donment of a currency peg yields macroeconomic dynamics consistent with sev-
eral stylized facts of exchange-rate-based stabilizations. A J-shaped pattern of
devaluation probabilities captures the observation that these stabilization plans
start with a very uncertain future (so devaluation probabilities are high when
they are introduced). Credibility then improves gradually (so devaluation proba-
bilities decline) until a certain point in time during which the survival of the plan
is increasingly questioned (so devaluation probabilities begin to rise). Blanco and
Garber (1986) and Klein and Marion (1997) provide econometric evidence in fa-
vor of this argument.
The currency risk premium implies that a J-shaped pattern of devaluation prob-
abilities yields a similarly shaped time path of domestic nominal interest rates for
the economy attempting to stabilize with imperfect credibility. In turn, the declin-
ing interest rate in the early stages of the program favors booms in consumption,
investment, and employment; a widening current account deficit; a reduction in
the expenditure velocity of circulation of money; and a real appreciation of the
currency (meaning an increase in the relative price of nontradable goods in terms
of tradable goods). These are low-tax periods in terms of the random taxes men-
tioned earlier, and they are also periods in which, each date that the peg survives,
the public realizes that actual wealth is higher than expected the date before (since
the surge in seigniorage did not materialize). Hence, the model’s intertemporal tax
distortions and fiscal-induced wealth effect both work to favor the economic
boom and real appreciation. As the nominal interest rate touches bottom and
starts to increase, the tax-like distortions work in the opposite direction and they
can outweigh the fiscal-induced wealth effect (in which case there is a recession
that predates the collapse of the peg, as observed, for example, before the devalu-
ations in Mexico in 1994 and in Argentina in 2002).
It is crucial to note that, in contrast with the perfect-foresight studies common
in the imperfect-credibility or policy-temporariness literature, the Mendoza-Uribe
model features the above cyclical dynamics regardless of whether the currency is
devalued ex post or not. Thus, the essential element of the analysis is the fact that
the stabilization policy anchored on the currency peg is not fully credible, and
hence the public expects that a devaluation may happen with some probability, in
an environment in which currency risk goes unhedged.
The findings of the Mendoza-Uribe model suggest that the analysis of macro-
economic policies of uncertain duration has gone some way in accounting for the
stylized facts of temporary or noncredible exchange-rate-based stabilizations. As a
framework that can account for sudden stops, however, it has an important short-
coming: stabilizing emerging economies are assumed to have access to a perfect,
frictionless international capital market in which they can borrow as much as
Why Should Emerging Economies Give up National Currencies? 109

they want as long as they refrain from running Ponzi games. As will be illustrated
later, it is not possible in this environment to obtain sudden reversals of the cur-
rent account in response to either a policy reversal or other exogenous shocks of
foreign or domestic origin.
Consider now the possibility that as a result of the informational frictions dis-
cussed in section 5.2, international capital markets are highly imperfect. The re-
cent literature on the sudden stops phenomenon has explored the potential of a
large variety of capital-market imperfections to act as triggers of sudden stops
(see the survey by Arellano and Mendoza 2003). For simplicity, follow Mendoza
(2001 and 2002) and Mendoza and Smith (2006) in assuming that these imperfec-
tions take the form of foreign borrowing constraints that impose liquidity require-
ments or collateral constraints on borrowers, and debt contracts that can only be
denominated in units of tradable goods for economies with a large nontradable-
goods industry (a phenomenon referred to as ‘‘liability dollarization’’).
A liquidity requirement is used by lenders as a criterion that helps them man-
age default risk by requiring borrowers to pay a fraction of their current obliga-
tions out of current income, or equivalently by requiring them not to let their
total debt-to-income ratio exceed a certain level. A classic case is the scoring crite-
ria used in mortgage lending, by which qualified borrowers are required to satisfy
specific ratios of nonmortgage debt payments and total debt payments as a frac-
tion of gross income.
What happens with the predictions of the devaluation risk framework of Men-
doza and Uribe (2000) when the structure of international capital markets is modi-
fied to introduce liquidity requirements? There are two essential changes. First,
the liquidity requirement introduces an occasionally binding constraint on foreign
borrowing, by which debt must be less than or equal to a certain fraction of the
value of domestic income in units of tradable goods. Second, whether this con-
straint binds or not at a particular date and state of nature is a combined result of
the underlying exogenous shocks driving the model, the endogenous dynamics of
foreign debt, income generated in the tradables and nontradables sectors of the
economy, and the domestic relative price of nontradables.
In this economy with borrowing constraints, sudden stops occur when the
country’s external debt is sufficiently large and a sufficiently adverse combination
of shocks shifts the economy from a situation in which the liquidity requirement
was not binding to a situation in which it binds. The sufficiently adverse stock of
debt and sufficiently adverse shocks are those such that the amount by which do-
mestic agents would like to enlarge their foreign debt, in the absence of a liquidity
constraint by foreign lenders, exceeds the level that this constraint allows. The
constraint will then be met by a sudden current account reversal and a collapse
in private absorption. Moreover, the adjustments in consumption and the current
110 Enrique G. Mendoza

account are magnified if they trigger a collapse in domestic income and/or in the
relative price of nontradable goods (that is, the real exchange rate) because this
implies that in addition to the liquidity constraint becoming suddenly binding,
the constraint is tightened further by the endogenous collapses of income and
prices (since debt is denominated in units of tradables but partially leveraged on
the sizable income generated by the nontradables sector).
How strong can these effects be? To answer this question, consider some of the
quantitative simulation results reported in Mendoza (2002). These simulations ap-
ply to Mexico as an example of a representative small, open, emerging economy.
Mexico has large industrial sectors producing and generating income in tradable
and nontradable goods, and both goods are also consumed by the domestic pri-
vate sector. Three exogenous random shocks are introduced to drive Mexico’s
business cycles: domestic productivity shocks, shocks to the world real interest
rate, and shocks that reflect the uncertain duration of economic policy (modeled
as regime-switching shocks affecting the rate of depreciation of the currency or a
set of direct and indirect tax rates). Mexican national accounts data are used to
calibrate the model. The calibration also examines historical data on Mexico’s for-
eign asset position to pin down parameter values that allow the model to mimic
Mexico’s average foreign debt-to-output ratio and a reasonable value for the li-
quidity requirement coefficient (in other words, the maximum debt-to-GDP ratio
in units of tradables).
Figure 5.5 shows the effects of a shift from a state with high productivity, low
world interest rate, and low policy distortions to a state with the opposite features
for different levels of initial net foreign assets, or the negative of foreign debt. The
plots show the impact effects for an economy assumed to have access to a perfect
credit market and for the economy that faces the liquidity requirement.
The results shown in figure 5.5 have three key implications. First, if there are no
imperfections in international credit markets, there are no sudden stops. Even at
very high debt levels, the adverse shocks trigger relatively smooth adjustments in
the model’s endogenous macroeconomic aggregates. Second, for very low or very
high debt levels there is also no room for sudden stops. With very high debt the
borrowing constraint binds regardless of whether the shocks are favorable or un-
favorable. With very low debt the constraint never binds and the economy’s re-
sponse to the shocks is nearly identical to that observed in the case in which no
credit-market imperfection is present. These are tranquil times in which shifts
across good and bad shocks produce relatively smooth business cycles and the
economy’s access to world capital markets is not compromised. Third, in the
range of debt positions just above the level at which the range of debt positions
with nonbinding liquidity constraints ends, adverse shocks trigger a suddenly
binding borrowing constraint. This is the sudden stop range. Here, the equilibrium
Why Should Emerging Economies Give up National Currencies? 111

Figure 5.5
Impact Effects of a Shift from ‘‘Best’’ to ‘‘Worst’’ State (as a function of the foreign asset grid)
112 Enrique G. Mendoza

response of the economy to the same size of adverse shocks to productivity, world
interest rate, and policy distortions as in the other ranges features a sharp reversal
of the current account and a major collapse in domestic consumption and produc-
tion. For part of this range there is a large drop in the relative price of nontrad-
ables. (The price can also rise because the liquidity constraint induces supply-
and-demand effects in the market of nontradables, and depending on which effect
is stronger, the price can rise or fall.)
The simulations shown in figure 5.5 are an imperfect approximation to reality
and they follow from a stylized model with many caveats and unrealistic assump-
tions. Yet the magnitude of the sudden stop adjustments that it produces is strik-
ing. Moreover, some of the caveats and unrealistic assumptions of this setting do
not apply to other recent quantitative studies that yield the same main prediction:
capital market imperfections are a powerful mechanism for producing sudden
stops. The literature on quantitative applications of models of emerging-markets
crises has examined alternatives that consider the margin constraints and trading
costs (Mendoza and Smith 2006 and Cavallo, Kisselev, Perri, and Roubini 2002),
costly monitoring of borrowers (Cespedes, Chang, and Velasco 2000), collateral
constraints (Paasche 2001), working capital (Oviedo 2002, Uribe and Yue 2006,
and Neumeyer and Perri 2006), default risk (Arellano 2005), and current account
targets (Valderrama 2002).
A closer look at the analysis of Mendoza and Smith (2006) can help highlight
the interaction between the information costs that distort international capital
markets (as argued in section 5.2), the determination of emerging markets asset
prices, and the business cycle transmission mechanism that drives sudden stops.
This analysis also brings into play a key feature of credit contracts among players
in international capital markets: collateral constraints in the form of margin
requirements. Factual descriptions of the contagion of the Russian crisis and its
connection with the collapse of LTCM describe how widespread margin calls
played a central role in precipitating the systemic sale of emerging countries’
assets in global capital markets (see Dunbar 2000 and the Wall Street Journal article
series published September 22–24, 1998).
Consider a representative foreign securities firm that specializes in trading a
small open economy’s equity. This firm incurs recurrent transactions costs as well
as transactions costs that are a quadratic function of the size of the trades it under-
takes. These two transactions costs represent the informational costs alluded to in
the second section. The per-trade costs are important because they imply an elas-
ticity in the foreign traders’ demand for emerging-markets equity that is inversely
related to the per-trade transactions costs. At the limits, if the per-trade cost is
zero the trader’s demand is infinitely elastic; if the per-trade cost is infinite the de-
mand is perfectly inelastic. The recurrent costs are important because they capture
Why Should Emerging Economies Give up National Currencies? 113

the idea that trading in emerging-market assets requires continuous investment in

gathering and processing country-specific information. These costs need to be in-
curred just to be prepared to trade on a particular emerging country’s assets, re-
gardless of whether there is actually any trade, and the costs themselves are
unrelated to the size of the trades. For example, understanding Chile’s inflation-
targeting system under its framework of widespread indexation, or Mexico’s
unique setup for conducting monetary policy by adjusting cortos (or ‘‘shorts’’)
among banking institutions, is a complex, information-intensive task and its cost
does not depend on how much a proposed trade in Chilean or Mexican bonds is
worth.
In the absence of recurrent trading costs, foreign traders would be willing to
hold on to a given position on emerging-markets assets if and only if the price
equals the fundamentals price (defined as the expected present value of dividends
discounted at the world real interest rate). They would be willing to buy more if
the price is lower than the fundamentals level, and less if the price is above the
fundamentals level. In the presence of recurrent costs, however, traders hold on
to a given emerging-markets asset position only if the price is below fundamen-
tals. Thus, recurrent costs are a necessary condition for the price of emerging-
markets assets to remain lower than justified by fundamentals even in the long
run (when asset positions reach a stationary equilibrium).
The foreign traders trade equity with the corresponding emerging country.
Lenders in the global credit market that agents in this country has access to de-
mand to hold, as collateral, a certain fraction of the market value of the domestic
agents’ equity holdings. This margin constraint is a collateral constraint with two
unique features. First, custody of the collateral is surrendered to lenders when the
debt is contracted, so the typical issues that arise with other collateral constraints,
related to whether lenders can or want to take over the collateral after a defaulting
borrower declares default, are irrelevant. Second, a fall in asset prices tightens the
constraint because it reduces the market value of the assets offered as collateral,
thereby allowing the lender to make a margin call on the borrower to make up
the difference. The margin call is automatic and does not require legal proceed-
ings beyond the margin clause already agreed to in the debt contract. Some mar-
gin requirements exist as a result of government regulation of the financial
industry but they are also widely used as explicit clauses in credit contracts and
can take different shapes (value-at-risk collateralization, for example, is a form of
margin requirement in which the lender makes margin calls based on its estimate
of the potential losses that it could face if a worst-case scenario for the particular
asset exposure in question materializes).
In this setting, again, a sufficiently large initial debt and a mix of sufficiently ad-
verse shocks to productivity, world interest rates, and domestic policies shifts the
114 Enrique G. Mendoza

economy from a state in which the margin constraint did not bind to one in which
it does. But now the process that follows has an extra ingredient: a modernized
version of the debt-deflation mechanism pioneered by Irving Fisher (1933). A
debt-deflation process starts when the sufficiently adverse shocks hit and agents
in the emerging economy get an initial margin call. They then rush to ‘‘fire-sale’’
their equity in the global capital market. However, they meet there with foreign
traders that have a less-than-infinitely-elastic demand for equity because of the
informational costs they incur, and so are only willing to buy the extra equity at a
reduced price. If there were no informational costs, the margin call would simply
result in an equity reallocation from home agents to foreign traders with no
change in emerging-markets equity prices. But with trading costs the asset price
falls, and the lenders see the value of the assets they hold as collateral shrink.
This triggers the margin clauses of debt contracts again, and a new round of mar-
gin calls takes place. The process is repeated until the equilibrium price settles at a
level that satisfies the margin constraint (Mendoza and Smith 2006 ensure that it
does by assuming, as was done in the second section, that agents cannot take un-
limited short positions).
A mechanism for financial contagion also emerges from this model. Consider
that a Russian-style default triggers a ‘‘world liquidity shock,’’ which takes the
form of a large increase in the world interest rate. An emerging economy sitting
in South America and totally unrelated to Russia may experience a sudden stop
and a collapse in the prices of its internationally traded assets as a result of the
combination of margin constraints and trading costs. An interest-rate shock is a
change in fundamentals and as such it should be reflected in what happens to an
emerging country’s asset prices and economic conditions, but the point is that
there can be substantial overreaction to this change in fundamentals because of
the imperfections of the world’s capital market.

5.6 Policy Implications

It may seem odd that a paper on abandoning national currencies reaches this
point without having said much about money and exchange rates (except for the
discussion of devaluation risk as a driving force of business cycles). This is be-
cause the problems that were described with lack of credibility of government
policy, and imperfect international capital markets can exist regardless of the
monetary institutions and currency arrangements. For example, Kumhof, Li, and
Yan (2006) showed that, if fiscal fundamentals are off track, balance-of-payments
crises occur even if a managed exchange rate is replaced with a floating rate un-
der inflation targeting. It also follows from this reasoning, however, that giving
Why Should Emerging Economies Give up National Currencies? 115

up the national currency to adopt a hard currency does not rule out financial crises
that can look like a sudden stop (consider, for example, the Great Depression).
When it comes to recent sudden stops in emerging economies, however, mone-
tary and exchange rate policies are at the core of government policy’s lack of cred-
ibility and the informational frictions behind global capital market imperfections.
An emerging country that completely gives up its national currency for the dol-
lar or the euro, and so reduces its central bank to a bank-supervision agency,
stands to make several gains.

1. The devaluation risk that has played a central role in the chronic boom-bust
cycle associated with the introduction and collapse of managed exchange rates in
this class of countries would be greatly reduced. It can never be fully eliminated,
inasmuch as a sovereign nation can never credibly commit to not try to reverse
course and reintroduce a national currency in the future.
2. Foreign investors that have a long history in investing resources and building
expertise to track monetary policy in Europe and the United States would no
longer need to ponder whether or not to pay for gathering and processing costly
information on the monetary policy of each emerging economy they were inter-
ested in. This can be interpreted as a decline in information costs, which would
translate into better-informed investors and reduced vulnerability to herd
behavior.
3. Reduced information costs also increase the demand elasticity for emerging-
markets equity of foreign traders, which limits the size of asset price declines and
Fisherian debt-deflations that could occur because of frictions like collateral con-
straints and margin calls.
4. Financial assets and liabilities would be matched in terms of currency denomi-
nation, ending the liability dollarization problem. Again, a problem with a similar
flavor will emerge whenever a sudden, large relative price collapse takes place,
but it would no longer be possible for the lack of credibility of domestic monetary
or exchange rate policy to generate it.
5. Considering the contracting environment from which liquidity requirements
and collateral constraints emerge, enhanced credibility and reduced informational
frictions could result in better access to international capital markets in terms of
reduced liquidity coefficients and margin requirements. We know little about
how optimal credit contracts would or should react to the adoption of a hard cur-
rency in replacement of a national currency, but what was argued here on the
basis of findings from the literature is that if liquidity requirements and collateral
constraints did fall, vulnerability to sudden stops and contagion would be greatly
reduced.
116 Enrique G. Mendoza

Regardless of these advantages, emerging countries will not rush to abandon

their currencies. A national currency is a valued symbol of national identity. It is
also a valued and powerful tool for governments to have in difficult times in
which it is a lot easier, and more discreet, to transfer wealth from one set of eco-
nomic agents to another by simply printing money rather than by proposing an
explicit policy to do it. There is also the loss of seigniorage and sovereignty
involved in giving up monetary policy-making powers to a foreign nation, and
the potential loss of not being able to conduct independent monetary policy to
smooth the pains of domestic business cycles. The latter is the less-relevant argu-
ment, even if perhaps it is the one that often occupies academic and policy
debates on exchange rate regimes. The reason is put bluntly in Robert Lucas Jr.’s
1996 Nobel lecture: ‘‘Central bankers and even some monetary economists talk
knowledgably of using interest rates to control inflation but I know of no evidence
from even one economy linking these variables in a useful way . . .’’ (666). Even if
we had such scientific knowledge, the recent historical record of many emerging
economies shows that their policy-making institutions have performed poorly at
using independent monetary and exchange-rate policy to preserve price stability
and avoid large business cycles. Moreover, even if we had both knowledge and
institutions for monetary policy to be useful, the small potential benefit of fine-
tuning business cycles would have to be matched up against the five listed gains
of giving up national currencies.
If abandoning national currencies seems a good but unrealistic idea, what else
can be done to address the underlying causes of emerging-markets crises? One
set of alternatives considers the possibility of establishing support schemes for
the prices of emerging-markets assets (see Lerrick and Meltzer 2001, Calvo 2002,
and Durdu and Mendoza 2006). The premise of these proposals is that either be-
cause of moral hazard problems (Lerrick and Meltzer) or because of capital mar-
ket imperfections (Calvo), asset prices of emerging countries can fall sharply
below levels warranted by fundamentals, and this is a central ingredient of the
process driving sudden stops. Hence the international financial organizations
could be redesigned to set up facilities that prevent prices from reaching those
crash levels. Another alternative is to deepen the internationalization of the finan-
cial system, and either secure support for local bank subsidiaries from their for-
eign owner banks, or limit the extent to which those local banks can provide fuel
for the liability dollarization process by imposing on them high reserve require-
ments (i.e., by moving to narrow banking).
These indirect means of achieving the same goals that adopting a hard currency
can achieve much more easily seem unduly costly to impose on the citizens of
emerging countries, many of whom suffer grave economic consequences as a re-
sult of recurrent episodes of sudden stops and financial contagion. Abandoning
Why Should Emerging Economies Give up National Currencies? 117

national currencies does seem a radical idea, but so did the euro, the European
Union, and the North American Free Trade Agreement not so long ago.
Of course giving up national currencies is not a panacea. It cannot fix many
fundamental economic and institutional problems plaguing emerging countries,
or eliminate forever all forms of financial crises. It is, however, unlike any other
currency arrangement in that it ties as tightly as possible the government’s hands
so as to prevent it from exercising its confiscatory powers via monetary policy,
and in that it simplifies greatly the task of assessing domestic financial policies
that is so critical for driving global capital inflows into emerging economies.

Acknowledgments

This paper borrows heavily from joint work with Guillermo Calvo and Katherine
Smith. Comments and suggestions from Cristina Arellano and Alejandro
Izquierdo are also gratefully acknowledged.

References

Arellano, Cristina. 2005. ‘‘Default Risk and Aggregate Fluctuations in Emerging Economies.’’ Mimeo.,
University of Minnesota.
Arellano, Cristina, and Enrique G. Mendoza. 2003. ‘‘Credit Frictions and Sudden Stops in Small Open
Economies: An Equilibrium Business Cycle Framework for Emerging Markets Crises.’’ In Dynamic
Macroeconomic Analysis, eds. Sumru Altug, Jagjit S. Chadha, and Charles Nolan, 335–405. New York:
Cambridge University Press.
Blanco, Herminio, and Peter M. Garber. 1986. ‘‘Recurrent Devaluation and Speculative Attacks on the
Mexican Peso.’’ Journal of Political Economy 94: 148–166.
Calvo, Guillermo A. 1986. ‘‘Temporary Stabilization: Predetermined Exchange Rates.’’ Journal of Politi-
cal Economy 94: 1319–1329.
———. 1998. ‘‘Capital Flows and Capital-Market Crises: The Simple Economics of Sudden Stops.’’ Jour-
nal of Applied Economics 1: 35–54.
———. 2000. ‘‘Balance-of-Payments Crises in Emerging Markets: Large Capital Inflows and Sovereign
Governments.’’ In Currency Crises, ed. Paul Krugman, 71–98. Chicago: University of Chicago Press.
———. 2002. ‘‘Globalization Hazard and Delayed Reform in Emerging Markets.’’ Mimeo., Center for
International Economics, University of Maryland.
Calvo, Guillermo A., Alejandro Izquierdo, and Ernesto Talvi. 2002. ‘‘Sudden Stops, the Real Exchange
Rate and Fiscal Sustainability: Argentina’s Lessons.’’ Mimeo., Research Department, Inter-American
Development Bank, Washington, D.C.
Calvo, Guillermo A., Leonardo Leiderman, and Carmen M. Reinhart. 1996. ‘‘Inflows of Capital to
Developing Countries in the 1990s.’’ Journal of Economic Perspectives 10, no. 2: 123–140.
Calvo, Guillermo A., and Enrique G. Mendoza. 1996. ‘‘Mexico’s Balance of Payments Crises: A Chroni-
cle of a Death Foretold.’’ Journal of International Economics 41: 235–264.
118 Enrique G. Mendoza

———. 2000a. ‘‘Capital-Market Crises and Economics Collapse in Emerging Markets: An

Informational-Frictions Approach.’’ American Economic Review 90, no. 2: 59–64.
———. 2000b. ‘‘Rational Contagion and the Globalization of Securities Markets.’’ Journal of International
Economics 51, no. 1: 79–113.
Calvo, Guillermo A., and Carmen M. Reinhart. 1999. ‘‘When Capital Inflows come to a Sudden
Stop: Consequences and Policy Options.’’ Mimeo., Center for International Economics, University of
Maryland.
Calvo, Guillermo A., and Carlos A. Vegh. 1999. ‘‘Inflation Stabilization and BOP Crises in Developing
Countries.’’ In Handbook of Macroeconomics, Vol 1C, eds. John B. Taylor and Michael Woodford. Amster-
dam: North-Holland.

Cavallo, Michelle, Kate Kisselev, Fabrizio Perri, and Nouriel Roubini. 2002. ‘‘Exchange Rate Overshoot-
ing and the Cost of Floating.’’ Mimeo., Stern School of Business, New York University.
Cespedes, Luis, Roberto Chang, and Andrés Vélasco. 2000. ‘‘Balance Sheets and Exchange Rate Policy.’’
Mimeo., Department of Economics, New York University.
Chang, Roberto, and Andres Velasco. 2000. ‘‘Banks, Debt Maturity and Crises.’’ Journal of International
Economics 51, no. 1: 169–194.
Cole, Harold L., and Timothy J. Kehoe. 1996. ‘‘A Self-Fulfilling Model of Mexico’s 1994–95 Debt Crisis.’’
Journal of International Economics 41, no. 3: 309–330.
Dunbar, Nicholas. 2000. Inventing Money: The Story of Long Term Capital Management and the Legends Be-
hind It. New York: Wiley and Sons.
Durdu, C. Bora, and Enrique G. Mendoza. 2006. ‘‘Are Asset Price Guarantees Useful for Preventing
Sudden Stops?’’ Journal of International Economics 69, no. 1: 84–119.
Fisher, Irving. 1933. ‘‘The Debt-Deflation Theory of Great Depressions.’’ Econometrica 1: 337–357.
Helpman, Elhanan, and Assaf Razin. 1987. ‘‘Exchange Rate Management: Intertemporal Tradeoffs.’’
American Economic Review 77: 107–123.
International Monetary Fund. 1999. International Capital Markets. Washington, D.C.: International Mon-
etary Fund.
Kaminsky, Graciela L., and Carmen M. Reinhart. 2000. ‘‘On Crises, Contagion, and Confusion.’’ Journal
of International Economics 51: 145–168.
Klein, Michael W., and Nancy P. Marion. 1997. ‘‘Explaining the Duration of Exchange-Rate Pegs.’’ Jour-
nal of Development Economics 54: 387–404.
Kumhof, Michael, Shujing Li, and Isabel Yan. 2007. ‘‘Balance of Payments Crises Under Inflation Tar-
geting.’’ Journal of International Economics 72, no. 1: 242–264.

Lerrick, Adam, and Allan H. Meltzer. 2001. ‘‘Blueprint for and International Lender of Last Resort.’’
Mimeo., Carnegie Mellon University, Pittsburgh, PA.

Lucas Jr., Robert E. 1996. ‘‘Nobel Lecture: Monetary Neutrality.’’ Journal of Political Economy 104: 661–
682.
Mendoza, Enrique G. 1995. ‘‘The Terms of Trade, The Real Exchange Rate and Economic Fluctuations.’’
International Economic Review 36: 101–137.
———. 2001. ‘‘The Benefits of Dollarization when Stabilization Policy Lacks Credibility and Financial
Markets are Imperfect.’’ Journal of Money, Credit and Banking 33, no. 2: 440–474.
Why Should Emerging Economies Give up National Currencies? 119

———. 2002. ‘‘Credit, Prices, and Crashes: Business Cycles with a Sudden Stop.’’ In Preventing Cur-
rency Crises in Emerging Markets, eds. Jeffrey Frankel and Sebastian Edwards, 335–390. Chicago: Uni-
versity of Chicago Press.
———. 2005. ‘‘Real Exchange Rate Volatility and the Price of Nontradables in Sudden-Stop-Prone
Economies.’’ Economia 6, no. 1: 103–148.
———. 2006. ‘‘Lessons from the Debt-Deflation Theory of Sudden Stops.’’ American Economic Review 96,
no. 2: 411–416.
Mendoza, Enrique G., and Katherine A. Smith. 2006. ‘‘Quantitative Implications of a Debt-Deflation
Theory of Sudden Stops and Asset Prices.’’ Journal of International Economics 70, no. 1: 82–114.
Mendoza, Enrique G., and Martin Uribe. 2000. ‘‘Devaluation Risk and the Business Cycle Implications
of Exchange Rate Management.’’ Carnegie-Rochester Conference Series on Public Policy 53: 239–296.
Milesi-Ferreti, Gian Maria, and Assaf Razin. 2000. ‘‘Current Account Reversals and Currency Crises:
Empirical Regularities.’’ In Currency Crises, ed. Paul Krugman. Chicago: University of Chicago Press.

Neumeyer, Pablo, and Fabrizio Perri. 2005. ‘‘Business Cycles in Emerging Economies: The Role of Inter-
est Rates.’’ Journal of Monetary Economics 52: 345–380.

Oviedo, P. Marcelo. 2004. ‘‘Intermediation of Capital Inflows: The Macroeconomic Implications of Neo-
classical Banks & Working Capital.’’ Mimeo., Iowa State University.
Paasche, Bernhard. 2001. ‘‘Credit Constraints and International Financial Crises.’’ Journal of Monetary
Economics 28: 623–650.
Parsley, David. 2001. ‘‘Accounting for Real Exchange Rate Changes in East Asia.’’ Working Paper No.
6/2001, Hong Kong Institute of Monetary Research.
Razin, Assaf, Efraim Sadka, and Chi-Wa Yuen. 1998. ‘‘A Pecking Order Theory of Capital Flows and
International Tax Principles.’’ Journal of International Economics 44, no. 1: 45–68.
Rebelo, Sergio, and Carlos A. Végh. 1996. ‘‘Real Effects of Exchange-Rate-Based Stabilization.’’ In NBER
Macroeconomics Annual 1996, Cambridge, MA: MIT Press.
Rigobon, Roberto. 2002. ‘‘Contagion: How to Measure It?’’ In Preventing Currency Crises in Emerging
Markets, eds. Jeffrey Frankel and Sebastian Edwards, 269–329. Chicago: University of Chicago Press.
Uribe, Martin, and Zhanwei Vivian Yue. 2006. ‘‘Country Spreads and Emerging Countries: Who Drives
Whom?’’ Journal of International Economics 69: 6–36.
Valderrama, Diego. 2002. ‘‘The Impact of Financial Frictions on a Small Open Economy: When Current
Account Borrowing Hits a Limit.’’ PhD dissertation, Department of Economics, Duke University.
Wall Street Journal. 1998. ‘‘Markets Under Siege,’’ Sept. 22–24, p. A1.
6 Hard Currency Pegs and
Economic Performance

Sebastian Edwards and I. Igal

Magendzo

‘‘. . . almost all independent countries choose to assert their nationality by having, to their own in-
convenience and that of their neighbors, a peculiar currency of their own . . .’’
—John Stuart Mill, Principles of Political Economy

6.1 Introduction

The recurrence of currency crises in emerging countries has generated an intense

debate on exchange rate policies. Pegged-but-adjustable exchange rate regimes
have rapidly lost adepts, while hard pegs and freely floating rates have gained in
popularity (See Summers 2000 and Fischer 2001). Recently, a number of econo-
mists have gone as far as arguing that (many) emerging nations should com-
pletely give up their national currencies, and join a currency union.1 Indeed, in a
series of highly influential papers, Guillermo Calvo and his associates have made
some of the most persuasive arguments with respect to the adoption of ‘‘common
currency’’ monetary regimes (see, for example, Calvo 1999, 2000, and 2001; Calvo
and Reinhart 2002; and Calvo, Izquierdo, and Talvi 2003).
In principle, currency unions can take two forms. First, a country may adopt
another nation’s currency as its own. When the other nation is an advanced coun-
try, this monetary arrangement has come to be known by the general name of
dollarization.2 Under dollarization the country in question completely gives up
monetary independence, and monetary policy is run by the advanced nation’s
central bank. Countries can dollarize in a unilateral fashion—in which case they
will lose the revenue from seignorage—or they can sign a monetary treaty with
the advanced country and share seignorage.3 Under the second type of currency
union, a group of countries creates a new currency that is common to the group.4
Under this option, monetary policy is run by a common central bank, the mem-
bers of the currency union share seinorage, and the common currency’s exchange
rate may float relative to other currencies. In the rest of this paper we will refer to
122 Sebastian Edwards and I. Igal Magendzo

this type of monetary regime as an independent currency union or ICU.5 Panama

and Monaco are good examples of dollarized countries. The East Caribbean Cur-
rency Area (ECCA) and the Communante Financiere de l’Afrique (CFA), on the
other hand, are good representatives of what we have called an independent cur-
rency union.
In the last few years some emerging countries have decided to give up their
currencies and officially dollarize their economies. In 2000, for example, in the
midst of a major crisis, Ecuador gave up its currency, the sucre, and adopted the
U.S. dollar. El Salvador adopted the dollar during 2001, and in May 2001, the dol-
lar became legal tender in Guatemala. In other countries, however, politicians
have systematically refused to consider dollarization, even in the face of major
and costly financial crises. This was the case of Argentina, for instance, during
late 2001 and early 2002.
A number of authors have argued that countries—and in particular, emerging
countries—that give up their currency will tend to outperform countries with a
currency of their own. According to this view, not having a domestic currency
will have two major positive effects on economic performance. First, inflation will
be lower in common-currency countries than in nations with their own currency.
Alesina and Barro (2001, 382), for instance, have argued that adopting an
advanced nation’s currency ‘‘eliminates the inflation-bias problem of discretionary
monetary policy.’’ Second, countries that give up their currency will tend to grow
faster than countries with a domestic currency. This growth effect is supposed to
take place through two channels: (1) A common currency will tend to result in
lower interest rates, higher investment, and faster growth (Dornbusch 2001). (2),
by eliminating exchange rate volatility, a common currency is supposed to en-
courage international trade; this, in turn, will result in faster growth. Rose (2000)
and Rose and Van Wincoop (2001), among others, have emphasized this trade
channel.6 Other authors, however, have voiced skepticism regarding the alleged
benefits of common-currency regimes. According to an alternative view that goes
back at least to Meade (1950), countries with a hard peg will have difficulty ac-
commodating external shocks, including terms of trade and world interest rate
disturbances. This, in turn, will be translated into greater instability and, under
some circumstances, will lead to lower economic growth (Fischer 1976, Parrado
and Velasco 2002). Frankel (1999) has taken a more nuanced view, and has
argued that there is no unique recipe on exchange rate policy; while some coun-
tries will benefit from hard pegs, for other countries a floating regime will be
more appropriate. And according to Eichengreen (2001), the evidence on the rela-
tionship between monetary regimes and growth is inconclusive, and does not
support the claim that dollarization—or any exchange rate regime, for that
matter—is an important determinant of growth.
Hard Currency Pegs and Economic Performance 123

Surprisingly, until recently there have been very few formal empirical studies
on the economic consequences of common currencies. In particular, international
comparative studies on alternative exchange rate and monetary regimes have tra-
ditionally ignored common-currency countries. For instance, the comprehensive
study on exchange rate regimes, growth, and inflation by Ghosh, Gulde, Ostry,
and Wolf (1995) does not include nations that do not have a currency of their
own. Likewise, the IMF (1997) study on alternative exchange rate systems
excludes common-currency countries, and the recent paper by Levy-Yeyeti and
Sturzenegger (2001) on exchange rates and economic performance excludes na-
tions that do not have a central bank of their own. This lack of empirical evidence
means that countries contemplating giving up their currency have very little in-
formation on how other countries have historically performed under this mone-
tary regime. Most existing evidence on dollarization is based on the experience of
a single country: Panama, which has used the U.S. dollar as legal tender since
1904.7
Recently, Andrew Rose and a series of collaborators have analyzed in great de-
tail the effects of common currencies on the volume of international trade (see, for
instance, Rose 2000, Engel and Rose 2002, and Frankel and Rose 2002). This inter-
esting and increasingly influential research has concluded that, with other things
given, countries with a common currency tend to trade among themselves more
intensively than countries that have a domestic currency. These analyses, how-
ever, do not make a distinction between the two types of common-currency
regimes discussed above: strictly dollarized and independent currency unions
(ICU). For instance, an inspection of the data sets used by Engel and Rose (2002)
and Frankel and Rose (2002) indicates that they treat dollarized and ICU nations
as a homogeneous group. Moreover, their sample is tilted toward ICU countries,
and has relatively few observations on strictly dollarized nations. From a policy
perspective, however, it is important to make a distinction between these two
common-currency regimes. The reason for this is that the two regimes have im-
portant differences in terms of independence of monetary policy, seignorage, and
capacity to absorb external shocks. Making a distinction between dollarized and
ICU countries is also important from a political economy point of view. As Frie-
den (2003) has argued, adopting another country’s currency is usually perceived
as giving up sovereignty, and has serious political costs. These political costs may
be reduced, however, if the country becomes a partner in an ICU. It is even possi-
ble that by joining an ICU the country reaps most of the benefits of a common cur-
rency without incurring the political costs associated with this measure.
The purpose of this paper is to analyze whether common currency countries—
both dollarized and ICU countries—have outperformed countries that have a cur-
rency of their own. The paper is empirical and proceeds in steps: we first analyze
124 Sebastian Edwards and I. Igal Magendzo

the behavior of all common currency countries and compare them with countries
with domestic currencies. We then turn to a comparison of the dollarized and
ICU countries. Performing these type of international comparisons, however, is
not easy. The problem is how to define an appropriate control group with which
to compare the common-currency nations. Since the adoption of a common cur-
rency is not a natural experiment, using a broad control group of all countries
with a domestic currency is likely to result in biased estimates. In this paper we
tackle this issue by using a treatment effects model that estimates jointly the prob-
ability of being a common-currency country and outcome equations for GDP
growth, inflation, and volatility (Maddala 1983; Heckman, Hidehiko, and Todd
1997; Green 2000; Wooldridge 2002). Some authors—most notably Alesina and
Barro (2000b, 2001)—have analyzed the conditions under which a (small) econ-
omy would benefit from giving up its currency. In contrast, we are interested in
finding out how countries with a long experience with a common-currency re-
gime have performed relative to countries with a currency of their own.
Before proceeding, it is useful to point out the ways in which our analysis dif-
fers from other related work in this general area. First, we have made an effort to
include data on both dollarized and ICU countries. This has not been easy, as
most strictly dollarized countries are very small and their data are not included in
readily available data sets. After significant effort we were able to obtain data on
GDP per capita growth and inflation for twenty strictly dollarized countries. We
also use data on thirty-two countries that are members of an independent cur-
rency union. Our data set, then, is significantly more general than the data set
used by other researchers. Second, we focus directly on the most important mac-
roeconomic variables—real GDP per capita growth, inflation, and growth volatil-
ity. Other studies, in contrast, have analyzed performance in an indirect fashion,
and have focused on ancillary variables such as the level of international trade
and/or interest rates. For instance, Edwards (2001b) and Powell and Sturzenegger
(2003) have investigated the way in which the exchange rate/monetary regime
affects interest rate behavior, and the cost of capital. On the other hand, Frankel
and Rose (2002) have analyzed the way in which currency unions affect bilateral
trade and, through this channel, economic growth.8 Third, we use a ‘‘treatment
effects model’’ to estimate the way in which dollarization affects the macroeco-
nomic variables of interest. And fourth, we make a distinction between strictly
dollarized and ICU countries.
The rest of the paper is organized as follows: In section 6.2 we provide a prelim-
inary analysis of historical experiences with common currencies. In section 6.3 we
use treatment regressions to analyze the effects of common currencies on a group
of macroeconomic variables. In section 6.4 we go a step further and disaggregate
the common currency countries into strictly dollarized and ICU countries. In sec-
Hard Currency Pegs and Economic Performance 125

tion 6.5 we undertake a robustness analysis, and we analyze how different sam-
ples and estimation techniques affect the results. In particular, we report results
on comparative performance obtained from an analysis that uses ‘‘matching esti-
mators’’ techniques, and from using an instrumental variables version of treat-
ment regressions. Finally in section 6.6 we provide some concluding remarks.

6.2 Common-Currency Experiences During 1970–1998: A Preliminary

Analysis

6.2.1 Some Background Discussion

Table 6.1 presents a list of fifty-two common-currency countries and territories

with available data for the period 1970–1998.9 In compiling this list we have
excluded countries that, while having a currency of their own, have had a long
tradition of fixed exchange rates, such as Ireland before 1979 and Bermuda.10 We
have divided our sample into two groups: (1) countries that use an advanced
country’s currency as legal tender, or strictly dollarized countries (in our sample
twenty countries satisfy this criterion); (2) other common-currency countries,
which we call ICUs. The majority of the ICUs use a currency that is common to
the area, but is not issued by any of the individual countries. Five of them, how-
ever, use the currency of another emerging country as legal tender. It is important
to note that our data set is much larger than that used by other studies. For in-
stance, in the influential studies by Rose and Engel (2002) and Frankel and Rose
(2002), there are only twenty-six countries that have data on real GDP per capita.
Of these, only seven use another nation’s currency, and only two—Panama and
Puerto Rico—use a convertible currency as legal tender and are, thus, strictly
dollarized.
The countries and territories that have had a strictly dollarized monetary sys-
tem are very small indeed. Many are city-states well integrated into their neigh-
bors’ economies—Monaco, Liechtenstein, and Andorra are good examples. The
largest strictly dollarized countries are Liberia, Panama, and Puerto Rico. How-
ever, only Panama and Puerto Rico remain dollarized today; Liberia abandoned
the system in the 1980s, when the government of President Samuel Doe decided
to issue local currency as a way of avoiding the constraints imposed by the dollar-
ized system.11 An important characteristic of many strictly dollarized economies
is that they are extremely open. In most of them there are no controls on capital
mobility or on any type of international financial transactions—so much so that
in 2001 six out of the thirteen independent dollarized nations were in the Organi-
sation for Economic Co-operation and Development (OECD) list of ‘‘Unfair Tax
Havens,’’ or countries whose lax financial regulations allow individuals and
126 Sebastian Edwards and I. Igal Magendzo

Table 6.1
Common Currency Countries with Available Macroeconomics Data
CFA Franc Zone France Italy
Benin Andorra (also Spanish Peseta)D San MarinoD
Burkina Faso French GuyanaD
Cameroon French Polynesia Australia
Central African Republic GuadeloupeD KiribatiD
Chad MartiniqueD NauruD
TuvaluD
Comoros MonacoD
Congo New CaledoniaD East Africa
Cote d’Ivoire ReunionD
Kenya (until 1979)
Equatorial Guinea
ECCA Tanzania (until 1979)
Gabon Uganda (until 1979)
Guinea-Bissau Antigua and Barbuda

Mali Dominica India

Niger Grenada Bhutan
Montserrat
Senegal
St. Kitts and Nevis Singapore
Togo
St. Lucia Brunei
USA St. Vincent and the Grenadines
Liberia (until 1989)D Denmark
South Africa GreenlandD
Marshall IslandsD
Lesotho
Micronesia, Fed. States of D Switzerland
D Namibia
Palau LiechtensteinD
PanamaD Swaziland
Puerto RicoD Belgium
New Zealand
Cook IslandsD LuxembourgD

Note:
D
: Corresponds to a dollarized country. We provide in bold either the name of the ICU or the name of
the country whose currency the common-currency countries have adopted.

corporations to evade taxes. Many dollarized territories in table 6.1—French

Guyana, Martinique, for example—have extremely close economic links with the
‘‘home country,’’ including labor mobility and free trade both in goods and in
financial claims. These characteristics of the dollarized economies—very small
and extremely open—suggest that using a broad control group of all nondollarized
countries, which are much larger and not as open, may indeed generate biased
results.12
The ICU countries included in our sample are larger than the dollarized na-
tions. These ICU nations have a common central bank that, in principle, can en-
gage in independent monetary policy. The two most important currency unions
in table 6.1 currently have a pegged exchange rate relative to a convertible cur-
Hard Currency Pegs and Economic Performance 127

rency: the ECCA has had a fixed exchange rate with respect to the U.S. dollar
since 1975. The CFA franc, on the other hand, was pegged to the French franc at
the time of its inception in 1948. In January 1994 the CFA franc was devalued and
repegged with respect to the French franc, and in January 1999, when the euro
was launched, the CFA franc became pegged to the euro. Notice that only a hand-
ful of countries in table 6.1 have adopted a common-currency monetary regime
within the timeframe of our sample. This means that it is not possible to under-
take a diffs-in-diffs analysis.

6.2.2 Unconditional and Unadjusted Comparative Analysis: 1970–1998

In table 6.2 we present comparative data on inflation, per capita GDP growth, and
the standard deviation of growth for our common currency countries.13 In order
to put things in perspective we also present data on these three variables for an
unadjusted control group that includes all countries with a currency of their own.
The table contains three panels: panel A includes data on all fifty-two common-
currency countries; panel B contains data on the strictly dollarized countries, or
countries that have adopted a convertible currency; and panel C includes the
ICU countries. In each panel we include data on the mean and median for the
three outcome macroeconomic variables. In column 3 we present data on mean
and median differences between the common-currency countries and the ‘‘with
currency’’ control group. The numbers in parentheses are t-statistics for the
significance of these differences. The test for the means differences is a standard t-
statistic, while the medians differences test is a t-test obtained using a bootstrap-
ping procedure. In making the computations for inflation differentials we have
followed Rose and Engel (2002) and have excluded countries with hyperinfla-
tions.14 However, excluding these observations only affects the calculation of the
means difference (quantitatively, but not qualitatively); it has no discernible effect
on the computation of median differences.
The results reported in this table indicate that the difference in inflation means
is quite sizable and statistically significant; on average, inflation in common-
currency countries as a group (panel A) has been nine percentage points lower
than in countries with their own currency.15 The difference in inflation medians is
also negative, much smaller (
3.7 percentage points), and still statistically signifi-
cant. These results also show that the rate of inflation in the strictly dollarized
countries (panel B) has been almost one-half that of the ICUs (panel C). In the lat-
ter group, however, inflation has still been significantly lower than in countries in
the control group.
In terms of real per capita GDP growth, the unadjusted comparisons in table 6.2
show that there are no significant differences in the means across any of the two
128 Sebastian Edwards and I. Igal Magendzo

Table 6.2
Inflation, Growth, and Volatility in Common-Currency Countries and Countries with Domestic
Currency
(1) (2) (3)
Dollarized and Other Difference*
union countriesa countriesb (1)
(3)
A. All Common-Currency Countries versus Control Group
Inflation
Mean 7.26 16.20
8.94
(
10.13)
Median 5.30 9.03
3.73
(
11.38)
Per capita GDP growth
Mean 1.29 1.19 0.10
(0.48)
Median 1.29 1.90
0.61
(
4.01)
Volatility of Growth
Mean 4.74 4.17 0.57
(1.91)
Median 3.48 2.89 0.59
(2.67)
B. Strictly Dollarized versus Control Group
Inflation
Mean 4.24 16.20
11.96
(
7.92)
Median 3.58 9.03
5.45
(
13.52)

Per capita GDP growth

Mean 1.25 1.19 0.06
(0.17)
Median 1.26 1.90
0.64
(
5.06)
Volatility of Growth
Mean 4.77 4.17 0.60
(1.40)
Median 3.34 2.89 0.45
(1.53)
Hard Currency Pegs and Economic Performance 129

Table 6.2
(continued)
(1) (2) (3)
Monetary Other Difference*
Unionsa countriesb (1)
(3)
C. ICUs versus Control Group
Inflation
Mean 8.73 16.20
7.47
(
6.95)
Median 6.94 9.03
2.09
(
4.08)
Per capita GDP growth
Mean 1.33 1.19 0.14
(0.52)
Median 1.29 1.90
0.61
(
1.63)
Volatility of Growth
Mean 4.71 4.17 0.54
(1.54)
Median 3.65 2.89 0.76
(2.32)
a
Number of observations with data for inflation is 760, of which 249 are strictly dollarized and 511
belong to an ICU. There are 1,332 observations with data for per capita growth, of which 526 belong
to a strictly dollarized country and 806 to an ICU.
b
Number of observations with data on inflation is 2,732 and there are 3,907 observations with data for
per capita GDP growth.
* Numbers in parentheses are t-statistics.

common-currency groups and the control group. The results also indicate, how-
ever, that the medians difference is significantly negative: the median rate of
growth in the two common-currency groups has been significantly lower—in a
statistical sense—than in the control group of countries with a currency of their
own. Finally, our results show that both groups of common-currency countries
have experienced greater growth volatility than the control group: both means
and medians differences are significantly positive.
Although the comparisons reported in table 6.2 are informative, they are subject
to two potential limitations. First, these are unconditional comparisons, as no ef-
fort has been made to control for other factors potentially affecting macroeco-
nomic performance. Second, the control group may not be the appropriate one. If
this is the case, the results presented in table 6.2 may be subject to a treatment
bias.16 We address both of these problems in the econometric analysis reported in
sections 6.3 through 6.5 of this paper.
130 Sebastian Edwards and I. Igal Magendzo

6.3 Common Currencies and Macroeconomic Performance

We are interested in investigating whether countries with a common-currency re-

gime have had a better macroeconomic performance than countries with a cur-
rency of their own. We focus on three dimensions of performance: GDP per
capita growth, inflation, and growth volatility. In principle, the exchange rate
regime will affect the growth process—both the first and second moments—
through three potential channels. First, a lower cost of capital—usually associated
with ‘‘hard peg’’ economies—will result in a higher rate of physical capital accu-
mulation and a higher growth rate of potential output. Second, a high level of in-
ternational trade—which, as Rose and his coauthors have persuasively shown, is
associated with common-currency regimes—is likely to have a positive effect on
total factor productivity (TFP) growth, and on the growth rate of potential output.
This effect has been emphasized in a number of endogenous growth models, and
operates through the effect of openness on the accumulation of knowledge. And
third, since the exchange rate regime will affect the country’s ability to accommo-
date external terms of trade shocks, it will affect growth volatility and, possibly,
average growth. Indeed, if as Meade (1950) and Corden (2002), among others,
have argued, countries with a hard peg—and in particular countries with a com-
mon currency—have more difficulty accommodating external disturbances, they
will tend to exhibit a more volatile rate of growth than countries with a domestic
currency. The relationship between common currencies and inflation is rather
straightforward. A strictly dollarized system will tend to solve the inflationary
bias associated with discretionary monetary policy, allowing a small country to
share the anchor country’s rate of inflation. A number of studies on Panama—the
strictly dollarized country par excellence—provide support for this view. Indeed,
Panama’s mean rate of inflation for 1970–1997 was 3.4 percent per year; the me-
dian rate during this period was 1.9 percent. Neither of these figures is signifi-
cantly different from U.S. inflation during that period. As Alesina and Barro
(2000b) have argued, the discretionary inflation bias will only be eliminated if the
local currency is (perceived to be) linked irrevocably to the anchor currency, and
it will tend to have a significantly lower rate of inflation than countries with a cur-
rency of their own.

6.3.1 The Empirical Model

Our main interest is to undertake a comparative analysis of the conditional effect

of a common currency on macroeconomic performance. In order to do this, we
estimate jointly an outcome equation and the probability of being a common-
Hard Currency Pegs and Economic Performance 131

currency country. As in the preceding section, we consider three outcome vari-

ables: GDP per capita growth, inflation, and volatility. In the estimation of the
inflation model we use unbalanced panel data for 169 countries, covering 1971–
1998. For the growth model we used two panel data sets: the first one is com-
prised of five-year averages; the second data set is a panel with yearly data.
Finally, for the volatility model we use averages of five-year periods to calculate
the standard deviation of GDP growth.17
The empirical treatment effects model may be written as follows:

yjt ¼ xjt b þ gdj þ m jt ð6:1Þ

1; if djt > 0
djt ¼ ð6:2Þ
0; otherwise

djt ¼ wjt a þ e jt : ð6:3Þ

Equation 6.1 is the macroeconomic performance equation, where yjt stands for
each of the macroeconomic outcome variables of interest in country j and period
t; xjt is a vector of covariates that captures the role of traditional determinants of
economic performance; djt is a dummy variable (the treatment variable) that takes
a value of 1 if country j in period t is a common-currency country, and 0 if the
country has a currency of its own. Accordingly, g is the parameter of interest: the
effect of the treatment on the outcome. The decision to have a common currency is
assumed to be the result of an unobserved latent variable djt , described in equa-
tion 6.2. djt , in turn, is assumed to depend linearly on vector wjt . Some of the vari-
ables in wjt may be included in xjt (Maddala 1983, 120).18 b and a are parameter
vectors to be estimated. m jt and e jt are error terms assumed to be bivariate normal,
with a zero mean and a covariance matrix given by
 
s v
: ð6:4Þ
v 1
If the performance and common-currency equations are independent, the cova-
riance term v in equation 6.4 will be 0. Under most plausible conditions, however,
it is likely that this covariance term will be different from 0.
Greene (2000) has shown that if equation 6.1 is estimated by least squares, the
treatment effect will be overestimated. Traditionally, this problem has been
tackled by estimating the model using a two-step procedure (Maddala 1983). In
the first step, the treatment equation 6.2 is estimated using probit regressions.
From this estimation a hazard is obtained for each j t observation. In the second
132 Sebastian Edwards and I. Igal Magendzo

step, the outcome equation 6.1 is estimated with the hazard added as an addi-
tional covariate. From the residuals of this augmented outcome regression, it is
possible to compute consistent estimates of the variance-covariance matrix 6.4. In-
stead of the traditional two-stage method, in this paper we use a more efficient
maximum likelihood procedure to estimate the model in equations 6.1 through
6.4 jointly.19 As shown by Greene (2000), the log likelihood for observation k is
given by equations 6.5 and 6.5 0 :
( )

wk a þ ðyk
xk b
dÞv=s 1 yk
xk b
d 2 pffiffiffiffiffiffi
Lk ¼ log F pffiffiffiffiffiffiffiffiffiffiffiffiffi

log 2ps; ð6:5Þ
1
v2 2 s

if dk ¼ 1
( )
2

wk a
ð yk
xk bÞv=s 1 yk
xk b pffiffiffiffiffiffi
Lk ¼ log F pffiffiffiffiffiffiffiffiffiffiffiffiffi

log 2ps; ð6:5 0 Þ
1
v2 2 s

if dk ¼ 0:

The model in equations 6.1–6.4 will satisfy the consistency and identifying con-
ditions of mixed models with latent variables if the outcome variable yjt is not a
determinant (directly or indirectly) of the treatment equation—that is, if y is not
one of the variables in w in equation 6.3.20 For the cases of per capita GDP growth
and volatility this is a reasonable assumption. Although the level of GDP per cap-
ita may affect the probability of having a common currency, its rate of change, or
the second moment of its rate of change, is unlikely to have an impact on the deci-
sion to have a domestic currency. This consistency and identifying restriction is
also met in the case of the inflation model. Indeed, in every one of the countries
in our sample, the decision to use a common currency can be traced historically
to variables that are structural in nature, including the country’s size and its cul-
tural and political relation with the anchor country. Moreover, what may affect
the decision to dollarize is the propensity to have a high inflation rate. This propen-
sity, however, is indeed captured by some of the variables in the wjt vector in
equation 6.3. However, in order to check for the robustness of the results obtained
from the estimation of the model in equations 6.1 through 6.5, in section 5 we
present results obtained from the instrumental variables estimation of a treatment
effects model for inflation.
In the estimation of the model 6.1–6.5, we also impose some exclusionary
restrictions; that is, a number of the wjt covariates included in equation 6.3 are
not included in the outcome equation 6.1. These exclusionary restrictions are not
required for identification of the parameters, but they are generally recommended
as a way of addressing issues of collinearity.21
Hard Currency Pegs and Economic Performance 133

6.3.2 Basic Results for Common-Currency Countries as a Group

In this section we report the results obtained from the estimation of the treatment
effects model given by equations 6.1–6.4. The ‘‘treatment group’’ is defined as all
countries without a currency of their own. That is, the dummy variable dtj takes a
value of 1 if in period t country j does not have a domestic currency; no distinction
is made, at this point, between strictly dollarized and ICU countries (see, how-
ever, the results in section 6.4). The data set covers 1970–1998, and includes 199
counties and territories. The number of observations varies depending on the
outcome variable considered. There are 3,122 observations on inflation and 5,233
observations on growth per capita. When using five-year averages—both in the
growth and volatility models—the panel has 950 observations.22

6.3.2.1 The Treatment Equation

In a highly influential article, Mundell (1961, 181) argued that the ‘‘optimum cur-
rency area is the region.’’ By this he meant that regional considerations—
geographical proximity and the existence of factor mobility, among others—were
more important than national (or sovereign) considerations in determining opti-
mal currency areas. This region-based approach has been present in most sub-
sequent work on the subject of optimal currency areas.23 Following Mundell’s
insight, we include a number of regional variables in our empirical analysis of the
probability of being a common-currency country. More specifically, in the specifi-
cation of the treatment equation 6.3, we included the following covariates that en-
capsulate the importance of the region:

1. A dummy variable that takes the value of 1 if the economy in question is an in-
dependent nation and 0 if it is a territory. Since factor mobility is much lower
across independent nations than between a dependent territory and the home
country, we expect the coefficient of this variable to be negative in the estimation
of equation 6.3.
2. A dummy variable that takes the value of 1 if the country in question is an is-
land or archipelago. Since island-archipelago countries are relatively isolated,
they tend to be self-contained regions. We expect this variable to have a negative
coefficient in 6.3.
3. A dummy variable that takes the value of 1 if the country has a common bor-
der with a nation whose currency is defined, by the IMF, as a ‘‘convertible cur-
rency.’’ We call this variable ‘‘border,’’ and we expect its estimated coefficient to
have a positive sign in equation 6.3.
In addition to these regional variables,24 the following covariates were also
included in the specification of the treatment equation 6.3:
134 Sebastian Edwards and I. Igal Magendzo

4. The log of population, measured in millions of people, as an index of the coun-

try’s size. We expect the estimated coefficient of this variable to be negative, indi-
cating that larger countries are less likely to use another nation’s currency.
5. The log of initial (1970) GDP, taken as a measure of the country’s economic
size. We also expect the coefficient of this variable to be negative.
6. An indicator of the degree of openness of the economy. For the majority of
countries and years we used the Sachs and Warner (1995) openness index. We
used data from a variety of sources to supplement the Sachs-Warner index for
those countries and years not covered in their sample.25 As Frieden (2003) has
argued, its estimated coefficient is expected to be positive.
7. A variable that measures the (log of the) distance between each country and
the global markets; in defining this ‘‘distance variable’’ we followed Leamer
(1999). We expect its estimated coefficient to be positive, indicating that countries
that are less integrated into world markets will have a lower probability of being
common-currency countries.

6.3.2.2 The Outcome Equations

We specified (and estimated) three different models, for GDP per capita growth,
inflation, and growth volatility. A difficulty we faced in undertaking this analysis
is that many common-currency countries have limited data availability. For in-
stance, very few of the strictly dollarized countries have data on education attain-
ment or on some other variables traditionally included in growth empirical
analyses (Barro and Sala-i-Martin 1995, Barro 1996). Indeed, popular data sets,
such as the World Bank World Development Indicators (WDI), the International
Financial Statistics (IFS), or the Barro-Lee (1996) data set, include data on only a
handful—three or four—strictly dollarized countries. Nevertheless, and after
searching in a number of alternative data sources, we were been able to include a
number of covariates in the outcome equations 6.1 for per capita growth, infla-
tion, and volatility.26
In the estimation of the GDP growth model we included, as customary, initial
GDP, a measure of openness, a variable that captures the country’s geographical
location, regional dummies (no dummies were included for Asia, which is defined
as the reference region),27 and the common-currency dummy. As Sachs (2000),
among others, has argued, countries located close to the equator tend to grow
more slowly, after controlling for other factors, than nations in other parts of the
world. Our geography variable—which we call ‘‘tropics’’—is defined as the (nor-
malized) absolute distance from each country to the equator. We expect its coeffi-
cient to be negative, capturing the fact that, with other things constant, countries
closer to the tropics will tend to grow at a slower rate than countries in other geo-
graphical areas.28
Hard Currency Pegs and Economic Performance 135

In the inflation model, in addition to the common-currency dummy, we

included openness, lagged inflation (as a measure of persistence), inflation lagged
twice, the tropics geography variable, distance, and the regional dummies. Fi-
nally, in the volatility model we include the following covariates in the outcome
equation: initial GDP, openness, the regional dummies, and the common-currency
dummy. As in the other models, in some of the specifications we introduced re-
gional variables. In the volatility model we expect that the estimated coefficient of
both openness and initial GDP will be negative.

6.3.2.3 Results
In table 6.3 we summarize the results obtained from the estimation of the treat-
ment effects model for GDP per capita growth. Table 6.4 contains the results for
inflation, and table 6.5 those for growth volatility. Each of these tables contains
two panels. The upper panel includes the results from the outcome equation; the
lower panel contains the estimates for the treatment equation.

Probability of Being a Common-Currency Country As may be seen from these

tables, the results are similar across models and are quite satisfactory. The vast
majority of the coefficients have the expected signs and are statistically significant
at conventional levels. They clearly indicate that the probability of being a
common-currency country is higher for very small, not independent countries (or
territories). Being an island reduces the probability of having a common currency,
as does greater distance from world markets. The estimated coefficient of ‘‘bor-
der’’ is not significant at conventional levels and has a negative sign (see, how-
ever, the discussion in section 6.4).29

GDP per Capita Growth In table 6.3 we present the results obtained from the
estimation of the growth model. We report results from six systems: the first three
were estimated using annual data, while the last three were estimated using five-
year averages. As may be seen, the traditional regressors have the expected signs
and are significant at conventional levels. In terms of the monetary regime, these
results show that the coefficient of the common-currency dummy is positive and
statistically significant for all specifications. Its point estimate ranges from 0.749
to 1.204. This suggests that, during the period under consideration and after con-
trolling for other factors, countries with a common-currency regime experienced a
higher rate of growth of GDP per capita than countries with a currency of their
own. These results suggest that the growth advantage of common-currency coun-
tries amounted, on average, to approximately one percentage point per year.
Notice that these results are quite different from the simple means differences
reported in table 6.2: while according to those results there have been no
Table 6.3
Common Currencies and GDP Growth: A Treatment Effects Model* (Maximum Likelihood)
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

0.399
0.449
0.506
0.438
0.464
0.532
Log(GDP0 ) (
5.87) (
5.24) (
5.82) (
4.57) (
3.85) (
3.85)
2.573 2.536 2.465 2.587 2.509 2.429
OPEN (9.86) (9.55) (9.26) (7.04) (6.82) (6.61)

3.668
4.308
TROPIC — — (
3.70) — — (
3.11)
0.872 0.754 1.021 1.079 0.816 1.204
Dummy (2.71) (2.15) (2.86) (2.17) (1.94) (2.17)

0.747
1.813
0.682
1.918
EUROPE — (
2.27) (
4.15) — (
1.48) (
3.16)
0.136 0.234 0.145 0.266
LAC — (0.46) (0.79) — (0.35) (0.64)

0.493
0.849
0.608
1.010
MENA — (
1.30) (
2.17) — (
1.13) (
1.84)
0.126
0.416 0.176
0.460
NORTHAM — (0.16) (
0.52) — (0.16) (
0.41)
0.877 0.536 1.106 0.728
SASIA — (1.73) (1.04) — (1.54) (1.00)

1.630
1.509
1.520
1.362
AFRICA — (
5.40) (
4.98) — (
3.52) (
3.15)
3.297 4.234 3.861 3.361 4.112 3.659
Constant (6.86) (6.93) (6.24) (4.91) (4.77) (4.19)
Treatment equation

0.462
0.463
0.463
0.490
0.493
0.492
Log(POP) (
29.38) (
29.47) (
29.46) (
12.93) (
13.12) (
13.06)

0.129
0.130
0.129
0.122
0.123
0.121
Log(GDP0 ) (
6.88) (
6.91) (
6.88) (
2.73) (
2.73) (
2.70)

1.025
1.026
1.025
0.554
0.535
0.539
INDEP (
12.98) (
12.97) (
12.95) (
3.47) (
3.34) (
3.37)

0.116
0.118
0.118
0.112
0.113
0.122
BORDER (
1.43) (
1.45) (
1.45) (
0.43) (
0.65) (
0.64)
0.115 0.116 0.115 0.240 0.241 0.237
OPEN (1.57) (1.58) (1.57) (1.42) (1.43) (1.40)

0.992
1.006
1.005
0.848
0.898
0.889
ISLAND (
14.01) (
14.27) (
14.25) (
5.09) (
5.45) (
5.40)
0.524 0.521 0.523 0.749 0.739 0.748
DISTANCE (7.16) (7.13) (7.16) (4.50) (4.43) (4.48)
3.531 3.574 3.552 1.390 1.536 1.431
Constant (4.73) (4.78) (4.75) (0.83) (0.92) (0.85)
Number of observations 5233 5233 5233 950 950 950
LR chi2 5.58 5.70 5.44 3.54 3.54 3.54
Prob > chi2 0.018 0.016 0.020 0.060 0.060 0.060
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treat-
ment equation, or equation on the probability of being a common-currency country. The numbers in
parentheses are t-statistics.
Hard Currency Pegs and Economic Performance 137

differences in average rates of growth across the two groups of countries, the esti-
mates in table 6.3 indicate that common-currency countries have grown at a sig-
nificantly faster rate than countries with a currency of their own. The chi square
test for the independence of the treatment and outcome equations indicates that
in all specifications the null hypothesis of independence across the equations is
rejected at conventional levels.30

Inflation The results for the inflation model are reported in table 6.4. As may
be seen from the outcome equation in the upper panel, the common-currency
dummy is negative and significant in every one of the specifications. The point
estimates range from
11.99 to
14.43, not only confirming that inflation has his-
torically been lower in the common-currency countries, but also indicating that
the common-currency advantage is somewhat larger than what the simple mean
differences results reported in table 6.2 suggest. The other covariates in the infla-
tion regressions reported in table 6.4 have the expected signs, and are statistically
significant. In particular, these results indicate that more open countries have
lower inflation, as do countries that are geographically closer to the global mar-
kets. Inflation appears to have some degree of persistence, and the regional dum-
mies indicate that, relative to the benchmark (Asia), Latin America and Africa
have had a significantly higher rate of inflation. As the w 2 statistics show, the null
hypothesis of independent equations is rejected at conventional levels. In order to
investigate the robustness of these results and deal with potential endogeneity
problems, we also estimated the inflation model using an instrumental variables
treatment approach. The results are presented in section 6.5.

Volatility Table 6.5 contains the results for the volatility models. The null hy-
pothesis of independent equations is rejected at conventional levels—the w 2 statis-
tics range from 29.0 to 34.7—and the dummy variables for common currency are
significantly positive, indicating that countries without a domestic currency have
experienced a higher degree of growth volatility than countries with a currency
of their own. Openness reduces volatility—a result that is in line with a number
of theoretical results in international economics.31 In addition, our estimates indi-
cate that, after controlling for other factors, countries that are closer to the equator
have exhibited a higher degree of volatility. Also, countries with a higher initial
level of GDP per capita have had a somewhat higher degree of volatility (the
point estimate of this coefficient is, however, rather low). According to these
results, growth in the countries of the Middle East and North Africa (MENA) has
been particularly volatile. In what appears to be a counterintuitive result, the coef-
ficient for the Europe dummy is positive, although not significant. The reason for
this apparent anomaly is that the Eastern and Central European nations are part
138 Sebastian Edwards and I. Igal Magendzo

Table 6.4
Common Currencies and Inflation: A Treatment Effects Model*
Model 1 Model 2 Model 3

8.591
9.089
8.059

OPEN (
9.14) (
9.31) (
7.84)
1.575 3.256 6.941
DISTANCE (1.82) (2.62) (3.77)

6.252
9.501
TROPIC — (
1.89) (
2.47)
0.089 0.089 0.085
INFLATIONT
1 (18.57) (18.44) (17.93)

0.003
0.003
0.003
INFLATIONT
2 (
2.41) (
2.38) (
2.58)

12.233
11.991
14.434
Dummy (
7.99) (
7.80) (
9.47)
10.226
EUROPE — — (3.88)
10.259
LAC — — (8.72)
3.997
MENA — — (2.08)
2.546
NORTHAM — — (0.80)
1.974
SASIA — — (1.03)
10.631
AFRICA — — (8.76)
4.678
11.381
11.381
Constant (0.62) (1.00) (1.00)
Treatment equation

0.585
0.585
0.585
Log(POP) (
23.75) (
23.75) (
23.75)

0.320
0.320
0.320
Log(GDP0 ) (
9.31) (
9.31) (
9.31)

1.319
1.319
1.319
INDEP (
9.06) (
9.06) (
9.06)

0.671
0.671
0.671
BORDER (
5.29) (
5.29) (
5.29)
0.146 0.146 0.146
OPEN (1.41) (1.41) (1.41)

1.511
1.511
1.511
ISLAND (
14.73) (
14.73) (
14.73)
0.350 0.350 0.350
DISTANCE (3.03) (3.03) (3.03)
14.843 14.843 14.843
Constant (4.73) (4.73) (4.73)
Hard Currency Pegs and Economic Performance 139

Table 6.4
(continued)
Model 1 Model 2 Model 3
Number of observations 3122 3122 3122
LR chi2 6.75 6.75 6.75
Prob > chi2 0.027 0.027 0.027
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treat-
ment equation, or equation on the probability of being a common-currency country. The numbers in
parentheses are t-statistics.

of the World Bank European region. Finally, notice that the estimated coefficients
for the common-currency dummy are significantly larger than the unadjusted
mean differences in volatility presented in table 6.2.

6.4 Strict Dollarization and Independent Currency Unions: Is There a

Difference?

The results reported in the preceding section grouped all common-currency coun-
tries together, implicitly assuming that strict dollarization and ICU are equivalent
monetary regimes. This, however, need not be the case. As was argued in section
1, there are a number of differences between them in terms of independence of
monetary policy, seignorage revenue, and capacity to absorb external shocks. In
this section we break up the common-currency countries into strictly dollarized
countries and ICUs. We then estimate treatment effects models for GDP growth,
inflation, and volatility for each of these two groups separately.32 The specifica-
tion of both the treatment and outcome equations are similar to those reported in
tables 6.3–6.5 for the common-currency nations.
The results for GDP growth are in table 6.6. Those for inflation are in table 6.7
and those on volatility are in table 6.8. In each of these tables we report a limited
number of equations. The results for alternative specifications were similar, and
are available on request. An inspection of the results suggest the following
patterns:
• There are important differences in the results for the treatment equation for
ICU and strictly dollarized countries. More specifically, the results for the proba-
bility of not having a domestic currency differ in the following respects for these
two samples: first, while the coefficient of ‘‘border’’ is positive (and significant
at either the 5 or 10 percent level) in the strict dollarization models in table 6.7
(models 1 and 3), it is significantly negative in the estimates for the ICU coun-
tries (table 6.7, models 2 and 4). This indicates that geographical proximity to a
Table 6.5
Common Currencies and Growth Volatility: A Treatment Effects Model
Model 1 Model 2 Model 3
0.026 0.034 0.024
Log(GDP0 ) (2.78) (3.24) (2.08)

2.471
2.392
2.148
OPEN (
6.78) (
6.55) (
5.92)
1.468 3.289
TROPIC — (1.65) (2.51)
2.456 2.178 2.457
Dummy (5.83) (3.27) (3.80)
0.660
EUROPE — — (1.14)
1.021
LAC — — (1.01)
2.334
MENA — — (4.45)

0.871
NORTHAM — — (
0.81)

1.031
SASIA — — (
1.49)

0.092
AFRICA — — (
0.23)
2.353 2.225 2.225
Constant (3.47) (3.27) (3.27)
Treatment equation

0.464
0.468
0.465
Log(POP) (
12.18) (
12.26) (
12.10)

0.151
0.159
0.160
Log(GDP0 ) (
3.44) (
3.59) (
3.65)

1.336
1.341
1.398
INDEP (
6.82) (
6.82) (
7.11)

0.090
0.078
0.075
BORDER (
0.48) (
0.42) (
0.40)
0.184 0.186 0.196
OPEN (1.96) (1.36) (1.31)

1.170
1.172
1.139
ISLAND (
7.13) (
7.12) (
6.91)
0.586 0.541 0.571
DISTANCE (3.58) (3.23) (3.36)
3.514 4.016 3.752
Constant (2.07) (2.31) (2.12)
Number of observations 950 950 950
LR chi2 34.71 28.98 32.99
Prob > chi2 0.0 0.0 0.0
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treat-
ment equation, or equation on the probability of being a common-currency country. The numbers in
parentheses are t-statistics.
Table 6.6
Independent Currency Unions, Strict Dollarization, and GDP Growth: A Treatment Effects Model
Model 1 Model 2 Model 3 Model 4

0.566
0.545
0.501
0.591
Log(GDP0 ) (
6.46) (
4.40) (
5.18) (
4.72)
3.227 3.319 2.658 3.281
OPEN (10.66) (7.66) (9.68) (7.60)

3.041
3.299
2.601
3.300
TROPIC (
3.03) (
2.31) (
2.47) (
2.31)
2.041 2.723
Dummy ICU (4.31) (3.68) — —
0.267 0.532
Dummy Dollarization — — (0.63) (0.84)

1.644
1.688
1.606
1.666
EUROPE (
3.70) (
3.78) (
3.57) (
2.61)

0.050
0.150
0.102
0.098
LAC (
0.16) (
0.34) (
0.32) (
0.21)

0.861
1.070
0.863
0.998
MENA (
2.17) (
1.90) (
2.14) (
1.73)

0.640
0.719
0.477
0.630
NORTHAM (
0.81) (
0.64) (
0.81) (
0.55)
0.329 0.454 0.413 0.594
SASIA (0.64) (0.62) (0.77) (0.87)

2.113
2.184
1.690
1.580
AFRICA (
6.46) (
3.68) (
4.92) (
3.19)
4.415 4.315 4.181 4.181
Constant (6.93) (4.71) (6.52) (6.52)
Treatment equation

0.345
0.386
0.524
0.542

Log(POP) (
16.62) (
7.83) (
22.71) (
9.44)

0.105
0.095
0.030
0.080
Log(GDP0 ) (
5.32) (
1.94) (
0.80) (
0.90)

0.808
0.238
1.253
1.333
INDEP (
7.73) (
1.70) (
12.78) (
5.77)

1.176
1.032 0.169 0.388
BORDER (
7.64) (
3.05) (1.61) (1.54)

3.831
3.654 1.047 1.276
OPEN (
8.40) (
3.23) (10.35) (5.49)

0.744
0.692
0.561
0.89
ISLAND (
8.47) (
3.30) (
5.68) (
2.01)
1.240 1.407 0.347 0.581
DISTANCE (11.90) (5.83) (2.79) (2.09)

4.826
6.345 4.003 2.434
Constant (
4.70) (
2.69) (3.06) (0.83)
Number of observations 4707 856 4427 856
LR chi2 4.16 2.68 3.97 4.23
Prob > chi2 0.040 0.102 0.050 0.038
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treat-
ment equation, or equation on the probability of being a common-currency country. The numbers in
parentheses are t-statistics. Models 1 and 3 annual data. Models 2 and 4 five year averages.
142 Sebastian Edwards and I. Igal Magendzo

Table 6.7
Independent Currency Unions, Strict Dollarization, and Inflation: A Treatment Effects Model
Model 1 Model 2

10.137
8.337
OPEN (
8.26) (
7.55)
6.983 6.223
DISTANCE (1.94) (3.07)

11.231
11.031
TROPIC (
2.76) (
2.59)
0.084 0.084
INFLATIONT
1 (17.96) (16.75)

0.003
0.003
INFLATIONT
2 (
2.42) (
2.46)

20.970
Dummy ICU (
9.52) —

10.026
Dummy Dollarization — (
4.79)
9.760 9.413
EUROPE (1.31) (3.33)
10.724 10.776
LAC (7.42) (8.45)
3.690 3.800
MENA (2.00) (1.86)
3.467 2.802
NORTHAM (1.05) (0.83)
2.107 1.169
SASIA (1.04) (0.56)
12.232 12.167
AFRICA (8.76) (8.62)

51.319
46.204
Constant (
2.92) (
2.92)
Treatment equation

0.492
0.881
Log(POP) (
16.30) (
14.78)

0.250
0.433
Log(GDP0 ) (
6.85) (
5.82)

0.260
1.967
INDEP (
1.44) (
9.80)
0.146 2.418
OPEN (1.41) (11.05)

1.581
1.352
ISLAND (
12.76) (
7.63)
0.180
0.031
DISTANCE (1.29) (
0.41)
7.443 15.533
Constant (4.85) (4.85)
Hard Currency Pegs and Economic Performance 143

Table 6.7
(continued)
Model 1 Model 2
Number of observations 2925 2676
LR chi2 6.15 4.29
Prob > chi2 0.02 0.04
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treat-
ment equation, or equation on the probability of being a common-currency country. The numbers in
parentheses are t-statistics.

convertible-currency nation is only important in determining the probability of

being a dollarized country. Second, the coefficient for openness is significantly
positive, as expected, in the dollarized models (table 6.7); it is negative in the ICU
models. As we argue in section 6.5, the CAF countries largely drive this result.
These important differences in the probit results support the hypothesis that it is
not correct to consider—as most authors have done until now—the two types of
common-currency regimes as a homogeneous group.
• In the GDP growth models, the dummy for strict dollarization is never signifi-
cant, indicating that after controlling for simultaneity and other factors, there is
no discernible difference in growth performance between dollarized countries
and countries with a currency of their own.
• In the GDP growth models, the dummy for ICUs is significantly positive, with a
point estimate ranging from 1.36 to 2.72. This suggests that, with other things
given, ICU countries have grown at a significantly faster rate than countries with
a currency of their own. Moreover, according to our point estimates the growth
effect of an ICU monetary regime appears to have been quite large.
• In the inflation models the dummies for both the strictly dollarized and the ICU
countries are significantly negative, confirming that both types of common-
currency countries have been able to have a significantly lower rate of inflation
than the with-domestic-currency countries.
• In the inflation models the point estimates for the dummies are quite different
for the two groups of countries. In fact, these results suggest that the low inflation
advantage is greater for the ICU countries. This contrasts with the unadjusted
comparisons reported in table 6.2, which suggested that strictly dollarized coun-
tries had a lower average inflation rate than ICU countries.
•The results for the volatility model show that the dummies’ coefficients are sig-
nificantly positive for both groups of countries. This supports the idea that super-
hard-peg countries’ regimes tend to result in higher real volatility.
Table 6.8
Independent Currency Unions, Strict Dollarization, and Volatility: A Treatment Effects Model
Model 1 Model 2
0.321 0.287
Log(GDP0 ) (2.70) (2.22)

1.771
2.549
OPEN (
4.36) (
7.00)
1.933 4.389
TROPIC (1.54) (3.15)
2.271
Dummy ICU (3.62) —
2.106
Dummy Dollarization — (4.07)
0.654 0.644
EUROPE (1.11) (1.07)

0.053
0.087
LAC (
1.21) (
0.21)
2.567 2.330
MENA (4.87) (4.34)

1.115
0.880
NORTHAM (
1.05) (
0.82)

0.569
1.403
SASIA (
0.83) (
1.95)
0.607 0.305
AFRICA (1.39) (0.66)
2.035 3.756
Constant (2.35) (4.37)
Treatment equation

0.370
0.548
Log(POP) (
7.59) (
9.50)

0.133
0.139
Log(GDP0 ) (
2.91) (
1.55)

1.115
1.608
INDEP (
4.26) (
6.44)

0.824 0.294
BORDER (
2.29) (1.17)

3.795 1.167
OPEN (
3.36) (4.81)

0.969
0.689
ISLAND (
4.73) (
2.86)
1.130 0.338
DISTANCE (4.79) (1.11)

2.928 5.569
Constant (
1.25) (1.75)

Number of observations 855 806

LR chi2 14.19 14.78
Prob > chi2 0.0 0.0
* The upper panel contains the outcome equation. The lower panel contains the estimation of the treat-
ment equation, or equation on the probability of being a common-currency country. The numbers in
parentheses are t-statistics.
Hard Currency Pegs and Economic Performance 145

6.5 Robustness Analysis and Further Results

In this section we deal with some extensions, we investigate the robustness of the
results, we address potential endogeneity problems in the inflation equation, and
we inquire as to what is behind our reported results.

6.5.1 Nonparametric Methods

It is possible that the specification forms chosen for the outcome equations affect
our reported results. In particular, the linearity of these equations may affect the
estimates of the treatment coefficient. In order to investigate whether this is an
important factor, we undertook a nonparametric analysis based on ‘‘matching
estimators’’ (see Blundell and Costa Dias 2000). A general advantage of this non-
parametric method is that no particular specification of the underlying model has
to be assumed. Matching estimators pair each common-currency country with
countries from the with-domestic-currency group.33 If the sample is large enough,
for each treated (common-currency) observation we should find, in principle, at
least one untreated observation with exactly the same characteristics. Each of
these properly selected untreated observations provides the required counterfac-
tual for our comparative analysis.34 The problem is that under most general con-
ditions it is not possible to find an exact match between a treated and untreated
observation. The matching estimator method focuses on estimating an average
version of the parameter of interest.35 That is, the matching estimator consists of
obtaining the difference in outcome as an average of the differences with respect
to similar—rather than identical—untreated outcomes. Rosenbaum and Rubin
(1983) have shown that an efficient and simple way to perform this comparison
is to rely on a propensity score, defined as the probability of participation or
treatment:

PðxÞ ¼ ProbðD ¼ 1=xÞ:

In our case, this is the probability of a country being a common-currency country.
This reduces a multidimensional problem to a one-dimensional problem, pro-
vided that we can estimate PðxÞ. Instead of matching countries directly on all of
their characteristics, we can compare countries with similar probabilities of being
a common-currency country. An additional advantage of this method is that the
estimation of the treatment on the treated is not affected by the lack of data on
some of the other variables affecting the outcome. In that regard, then, this
method provides us with an elegant way of handling potential problems emanat-
ing from omitted variables in the outcome equations.36
146 Sebastian Edwards and I. Igal Magendzo

Table 6.9
Common Currencies, ICUs, and Dollarization: Matching Estimators, Mean Differences*
All common
ICUs Dollarized currency
A. Inflation
6.48
7.06
6.38
(
8.75) (
11.84) (
10.98)
B. GDP per capita growth 0.53
1.11
0.91
(1.97) (
2.84) (
.88)
C. Growth Volatility 0.86 1.18 0.27
(2.95) (1.98) (1.88)

* The results reported in this table correspond to the five nearest neighbors without replacement. The
numbers in parentheses are t-statistics.

In this section we report results obtained from using a simple-average nearest-

neighbor estimator. According to this method, for each treated observation, we
select a predetermined number of untreated nearest neighbors. The nearest neigh-
bors of a particular treated observation i are defined as those untreated observa-
tions that have the smallest difference in propensity score with respect to i. If we
choose to use nn nearest neighbors, we set
1
Wij ¼
nn

for the observations that have been selected; for other observations we set Wij ¼ 0.
We applied this method to both one nearest neighbor and five nearest neighbors.
The results we report in table 6.9 were obtained using a multiple treatments
procedure that assumes that at any point in time there are two possible (and alter-
native) ‘‘treatments,’’ a dollarization treatment, and an ICU treatment. The
untreated group is comprised of all those countries with a currency of their
own.37 The matching results in table 6.9 may be summarized as follows: first, we
confirm that both types of hard peg monetary regimes have resulted in lower in-
flation than regimes with a domestic currency. Second, the GDP growth estimates
confirm that ICU countries have grown at a faster rate than countries with a cur-
rency of their own; the opposite is the case for dollarized countries. In fact, these
results indicate that countries with a dollarized regime have grown at a signifi-
cantly lower rate than countries with a domestic currency. This result contrasts
with the regression results reported in table 6.7, which suggested that there had
been no difference in the rates of growth in the two groups. And third, the esti-
mates for volatility indicate that in both hard peg regimes volatility has been
Hard Currency Pegs and Economic Performance 147

higher than in countries with a currency of their own. These mean differences are
only significant when five nearest neighbors are used.
Overall, we interpret the results from the matching exercise as providing broad
support to the findings reported in the preceding section: countries with both
hard peg regimes have had lower inflation than countries with a currency, only
ICU countries have had higher growth, and both hard pegs appear to have had a
more volatile real economy than countries with a domestic currency.

6.5.2 Redefining ‘‘Very Rapid Inflation’’

In our inflationary analysis we excluded countries with extremely rapid inflation,

or ‘‘hyperinflation countries.’’ In the estimates reported in tables 6.4 and 6.7 the
sample excluded countries with a rate of inflation in excess of 200 percent per
year. It is possible, however, that by still allowing highly inflationary countries in
the sample, the estimates obtained are being driven by extreme or outlier observa-
tions. In order to investigate this issue we re-estimated the inflationary equation
under alternative definitions of ‘‘extremely rapid inflation.’’ More specifically, in
the alternative estimates we first excluded observations with an annual rate of in-
flation in excess of 100 percent; we then repeated the exercise excluding observa-
tions with inflation in excess of 50 percent per year.
The results obtained when these new samples were used confirmed those
reported earlier in the sense that inflation is significantly lower in common-
currency countries. Interestingly, however, under these new definitions of very
rapid inflations the stability advantage of ICU countries appears to be even
greater than that reported in section 6.4.

6.5.3 Instrumental Variables Estimation of the Inflation Model

The results presented in the preceding sections assumed that the treatment did
not depend on the outcome variable.38 That is, in estimating the treatment models
in sections 6.3 and 6.4 we assumed that the treatment (common-currency) dummy
was a strictly exogenous variable. It is possible to argue that this assumption is
not valid in the inflation model; according to this line of argument, countries that
choose a common-currency regime are countries with high inflation rates. In or-
der to deal with this possible source of bias, we also estimated the inflation model
using an instrumental variables (IV) version of the treatment approach.
Under general conditions the outcome equation 6.1 can be rewritten as:

y ¼ a þ gd þ xb0 þ dðx
EðxÞÞb1 þ e0 þ dðe1
e0 Þ; ð6:10 Þ

148 Sebastian Edwards and I. Igal Magendzo

where the subscript indicates the value of d and e0 and e1 are zero mean, normally
distributed error terms, conditional on X and a set of instruments Z. We assume
that the treatment depends on the instruments Z—that is, that the vector w in
equation 6.3 contains variables that are excluded from the outcome equation and
that are orthogonal to the error terms. Woodridge (2002) shows that a consistent
estimator of g, even when d is endogenous, is obtained by a two-step procedure.
In the first step we estimate the parameters in equation 6.3 using a probit regres-
sion of d on X and Z; from this estimation we then obtain the hazard for each ob-
servation. In the second step we estimate the parameters in 6.10 by ordinary least
squares (OLS), including the dummy variable, the variables in X, the hazard func-
tion, and the interaction of the dummy variable with X
EðXÞ, where we use the
sample average of X as an estimate of the expected value. The hazard—estimated
using the instruments Z—together with the interaction terms play the role of a
control function that controls for possible selection bias.39
In the estimation of 6.1 0 for inflation, the following instruments were used: log
of population, log of initial GDP, an independent dummy, an index of the degree
of openness, the ‘‘tropics’’ variable, the regional dummies, the log of distance, the
‘‘border’’ variable, the ‘‘island’’ variable, and the average of inflation in the five
years prior to our first data point. The results obtained from this IV treatment
model confirm those presented in sections 6.3 and 6.4 above. More specifically,
the point estimate of the dummy for strict dollarization is
12.9 and was signifi-
cant. The dummy for the ICU was also significant and its point estimate was

18.8. The detailed results from these IV treatment estimates (not reported here
due to space considerations) are available from us on request.

6.5.4 What is Really Behind these Results? The Role of the CFA and the ECCA

From a policy perspective—and in particular from a ‘‘lessons’’ point of view—an

interesting question is whether a specific group of countries is behind the results
reported in the preceding sections. We are particularly interested in understand-
ing whether there is any pattern behind the results suggesting that, while strictly
dollarized countries have grown at a rate similar to those countries with a cur-
rency of their own, ICU countries have grown significantly faster. In order to in-
vestigate this issue we analyzed the residuals from our regression analysis, and
inquired as to the characteristics of our ICU data set.
As may be seen in table 6.1, the ICU data set may be divided into three groups
of countries: seven members of the East Caribbean Currency Area (ECCA), fifteen
members from the CFA Franc Zone in Africa, and countries from different parts of
the world. A detailed inspection of the raw data suggests that there are very sub-
Hard Currency Pegs and Economic Performance 149

stantial differences in terms of economic growth between the ECCA nations, on

the one hand, and the rest of the ICU countries. Indeed, for the period under con-
sideration, average yearly GDP per capita growth in the ECCA countries has been
3.16 percent. In the other ICU countries it has only been 0.79 percent. The compar-
ison of medians yields a similar result, with the median growth for ECCA coun-
tries at 3.6 percent and that for the rest of the ICUs at 0.72 percent. This
unconditional comparison suggests that the ECCA nations’ performance is behind
our findings (table 6.6) that, with other things given, ICU nations grow at a faster
rate than countries with a domestic currency. In order to investigate this, we esti-
mated a separate treatment effects model for GDP per capita growth for ECCA
and other ICU nations.40 The estimated coefficient for the ECCA common-
currency dummy variable was 2.3, with a t-statistics of 4.46, confirming that
ECCA nations have outperformed by a wide margin countries with a currency of
their own. The results for the non-ECCA ICU countries were quite different, with
a statistically insignificant estimated coefficient for the treatment dummy of 0.5.41
These results, then, confirm the notion that the driving force behind the apparent
superior growth performance of common-currency countries, reported in the pre-
ceding sections, is fully driven by the group of seven ECCA nations.

6.6 Concluding Remarks

In the last few years a number of economists have argued that the emerging
economies should give up their domestic currencies. Interestingly, there have
been very few systematic comparative studies on the performance of countries
that, indeed, do not have a currency of their own. Moreover, there has been no ef-
fort in the empirical literature to make a distinction between strictly dollarized
countries and independent currency union countries. Most of the literature on the
subject has, in fact, been based on case studies of Panama, or on indirect perfor-
mance analyses of groups of countries strongly dominated by ICUs. This lack of
empirical analysis has resulted in policy debates that, until now, have been based
on conjecture and not on hard historical evidence. This has particularly been the
case when the issue under discussion relates to the merits of strict dollarization.
The purpose of this paper has been to analyze, from a comparative perspective,
economic performance in economies that do not have a currency of their own. We
have argued that the main difficulty in performing this type of comparison is
defining the correct control group with which to compare the performance of the
common-currency countries. We tackle this issue by using the treatment effects
model developed in the labor economics literature; this method allows us to joint-
ly estimate the probability of being a common-currency country and the effect of
150 Sebastian Edwards and I. Igal Magendzo

having this particular monetary regime on specific macroeconomic outcomes. Es-

timation using this technique yields results that are substantially different from
those obtained from simple comparisons using a large control group of all with-
domestic-currency countries. More specifically, we found that both ICU and
dollarized countries have had a significantly lower rate of inflation than with-
currency ones. We found that macroeconomic volatility has been higher in both
dollarized and ICU economies than in with-currency countries.
We believe that our results are particularly interesting with respect to GDP
growth. The estimations reported in table 6.6 suggest that while strictly dollarized
countries have had a statistical rate of GDP per capita growth that is not different
from that of countries with a currency of their own, ICU countries have grown
faster than with-currency nations. When we investigated these findings further,
we found that the East Caribbean Currency Area countries were the driving force
behind this estimated superior performance. Indeed, once these seven countries
were excluded from the sample, we found no statistical difference in GDP per
capita growth between the rest of the ICU countries and countries with a currency
of their own. The ECCA countries constitute a very special group: they are very
small indeed, with an average population of fewer than 100,000 inhabitants. They
are all islands geographically close to major markets. Their main industry is tour-
ism, and they have very close economic and cultural ties with the United King-
dom. We believe that their experience with ICU may not be entirely useful for
larger countries planning to reform their exchange rate and monetary regime.
This does not mean, however, that other emerging nations would not benefit
from giving up their currencies. Indeed, as Mundell (1961) argued four decades
ago, it is perfectly possible that countries with certain characteristics will bene-
fit from giving up their currency and either dollarizing or joining a currency
union. In that regard the experiences of Ecuador and El Salvador—two recent
dollarizers—will be very useful in assessing this question in the future. Naturally,
the euro experience will provide a wealth of information on the consequences of
common currencies in advanced nations.

Acknowledgments

This paper was prepared for presentation at a conference in honor of Guillermo

Calvo, held at the International Monetary Fund April 15–16, 2004. In our work
on common currencies we have benefited from discussions with John Cochrane,
Barry Eichengreen, Eduardo Engel, Ed Leamer, Robert Barro, and Andy Rose.
We thank our discussants Lars Svensson and Hang Genberg for many helpful
comments. Marcela Aurelio provided wonderful assistance.
Hard Currency Pegs and Economic Performance 151

Notes

1. See Alesina, Barro, and Tenreyro (2002) for a discussion on the conditions under which emerging
countries will benefit from giving up their currency.
2. Although this option is known as ‘‘dollarization,’’ the advanced country’s currency need not be the
dollar. It could be any other convertible currency.

3. In early 2000 Florida’s then-senior senator Connie Mack introduced legislation into the U.S. Senate
aimed at sharing seignorage with countries that decided to adopt the U.S. dollar as legal tender. The
bill, however, did not move in the legislative process.
4. The euro is perhaps the best-known example of this type of currency union.
5. Strictly speaking there is a third type of currency union: when a small country adopts a nonconver-
tible currency from another country as legal tender. In this case the ‘‘credibility’’ effect of having mone-
tary policy run by an advanced nation’s central bank will not be present. Thus, in the empirical
analysis that follows we group this third category with what we have called independent currency
unions.

6. On analytical aspects of dollarization see Calvo (1999) and Eichengreen and Haussman (1999); for
an interesting theoretical analysis see Cooper and Kempf (2001).
7. Goldfjan and Olivares (2001) use econometrics to evaluate Panama’s experience with dollarization.
Moreno-Villalaz (1999) provides a detailed analysis of the Panamanian system. Bogetic (2000) describes
several aspects of dollarization in a number of countries. As far as we know, Rose and Engel (2000) and
Edwards (2001a) are the first two papers to provide a statistical and econometric analysis of economic
performance in dollarized countries and/or currency unions. See also the papers in Levy-Yeyati and
Stuzenegger (2003).
8. See Klein (2002) for a discussion on dollarization and trade, including a comprehensive bibliogra-
phy on the subject.
9. These countries have data for a long enough period for at least one of two variables: GDP per capita
or inflation. In the rest of the paper we will use the term ‘‘countries’’ to refer both to independent coun-
tries and to territories.
10. Rose and Engel (2002), in contrast, consider both Ireland and Bermuda as common-currency coun-
tries. Ireland’s currency was the Irish pound, or ‘‘punt.’’ From September 1928 through March 1979 it
was linked to the British pound at parity. Bermuda’s currency is the Bermuda dollar. The Bermuda
Monetary Authority issues BD$ notes in several denominations (BD$ 2, 5, 10, 20, 50, and 100). The
BD$ is linked to the U.S. dollar at parity. It should be noted that if these countries are added to the
list of common currency countries in table 6.1, the results reported in this paper remain basically
unaffected.
11. It is not easy to date unequivocally Liberia’s abandonment of the dollarized system. In July 1974
the National Bank of Liberia (NBL) was opened. In 1982 the NBL began issuing five-dollar coins, and
in 1989 it began issuing five-dollar notes. On Liberia’s dollarization experience see Barret (1995) and
Berkeley (1993).
12. The median population of all nondollarized emerging nations is over 100 times larger than that of
the dollarized economies.

13. Our volatility measure is the standard deviation of growth, calculated over a five-year period.
When alterative time frames were used (i.e., seven years) the results did not change in any significant
way.
152 Sebastian Edwards and I. Igal Magendzo

14. More specifically, we excluded from the control group those observations with a rate of inflation in
excess of 200 percent per year. This resulted in eighty observations being dropped from the control
group of countries with a currency of their own. See section 6.5 for results under alternative definitions
of ‘‘high inflation.’’
15. When hyperinflation countries are not excluded the means difference in inflation is a staggering 62
percent.
16. See Maddala (1983).
17. Volatility of GDP growth is measured as the standard deviation of growth in subperiods of five
years. The subperiods correspond to the years 1974–1978, 1979–1983, 1984–1988, 1989–1993 and
1994–1998. Accordingly, the covariates were averaged out for the same subperiods. The union, ICU,
and dollarized dummies were set to one for a particular subperiod if the country belonged to the re-
spective group for at least four of the five subperiod years. In particular, Equatorial Guinea belongs to
the CFA union only since 1985, but it was assigned a value of 1 for the ICU (and the union) dummy for
the subperiod 1984–1988. On the contrary, even though Liberia was dollarized till 1981, it was not con-
sidered a dollarized country for the subperiod 1979–1983 (neither was it included in the control group).
On the other hand, the West African Currency Union was dissolved in 1980, so Kenya, Tanzania, and
Uganda are not considered part of an ICU (nor as part of the control group) for the subperiod 1979–
1983.
18. It is assumed, however, that djt does not depend on yjt . Otherwise, as discussed later, the model
cannot be identified.
19. The two-steps estimates yield similar results, and are available from the authors on request.

20. Details on identification and consistency of models with mixed structures can be found in Maddala
(1983). See also Heckman (1978), Angrist (2001), and Wooldridge (2002).
21. Wooldridge (2002).

22. Remember that the volatility variable is measured as the standard deviation of growth over a five-
year period. When alternative seven-year periods were used, there were no significant changes in the
results.
23. See Alesina, Barro, and Tenreyro (2002) for a recent discussion on the subject.
24. Unfortunately, there are no available data on other regional variables of interest, including a gener-
alized index of factor mobility or of synchronicity of shocks, for all countries in our sample.
25. See the original Sachs-Werner (1995) article for a specific list of requirements for a country to qual-
ify as ‘‘open.’’
26. The data are available from the authors on request.
27. These variables are expected to capture the effect of some variables for which there were no data.

28. See Leamer (1997) and Venables (2002) for discussions on the role of geography and distance on
economic growth.
29. In addition to the variables in tables 6.3–6.6, we considered additional covariates. In particular, we
constructed an index on whether the country in question was a member of a ‘‘deep’’ trading area. This
index, however, identifies almost fully the common-currency countries, reducing the spirit of the
analysis.
30. An important question is whether these results—as well as those in tables 6.4–6.5—are subject to
an ‘‘omitted variables’’ bias, stemming from the fact that there are no data on some of the traditional
Hard Currency Pegs and Economic Performance 153

regressors for the common-currency countries. We address this issue in some detail in section 6.4 of
this paper.
31. See Frenkel and Razin (1987).

32. In doing this we excluded the other common-currency group from the with-domestic-currency
sample. That is, the with-domestic-currency category is strictly made of countries that do have a cur-
rency of their own.
33. If we estimate the equation above using all nontreated observations, the selection bias is given by:
BðxÞ ¼ Eðu0 =x; D ¼ 1Þ
Eðu0 =x; D ¼ 0Þ:

34. In order to guarantee that all treated agents have such a counterpart in the population (not neces-
sarily in the sample), we also need to assume that
0 < ProbðD ¼ 1=xÞ < 1:
35. This averaged version is given by
Ð
Eð y1
y0 =x; D ¼ 1Þ dFðx=D ¼ 1Þ
MðSÞ ¼ S Ð ;
S dFðx=D ¼ 1Þ

where S is a subset of the support of x given D ¼ 1.

36. This assumes that there are no omitted variables problems in the estimation of the propensity used
to select the nearest neighbors.
37. See Lechner (2002) for a discussion on matching methods with multiple treatments. The results in
table 6.9 correspond to the case when ‘‘no replacement’’ is allowed in selecting the nearest neighbors
in the control group. See Edwards and Magendzo (2002, 2006) for details.
38. This assumption is sometimes referred as the ‘‘ignorability-of-treatment assumption.’’ (Wooldridge
2002).

39. For details on these procedures see Woodridge (2002).

40. In the estimation we had to respecify the treatment equation. The reason is that some of the regres-
sors (islands, for example) fully predicted the probability of being a currency union country.
41. These results are from a specification that includes regional dummies in the GDP growth outcome
equation. If these dummies are excluded the ICU dummy becomes negative with a t-statistic of
1.46.

References

Alesina, A., and R. J. Barro. 2000a. ‘‘Currency Unions.’’ Working Paper No. 7927, NBER, Cambridge,
MA.
———. 2000b. ‘‘One Country, One Currency?’’ In Currency Unions, eds. A. Alesina and R. J. Barro, 11–
19. Stanford, CA: Hoover Institution Press.
———. 2001. ‘‘Dollarization.’’ American Economic Review 91, no. 2: 381–385.
Alesina, A., R. J. Barro, and S. Tenreyro. 2002. ‘‘Optimal Currency Areas.’’ Working Paper No. 9072,
NBER, Cambridge, MA.
Angrist, J. 2001. ‘‘Estimation of Limited Dependent Variable Models with Dummy Endogenous Regres-
sors: Simple Strategies for Empirical Practice.’’ Journal of Business and Economic Statistics 19, no. 1: 2–16.
154 Sebastian Edwards and I. Igal Magendzo

Barret, L. 1995. ‘‘The Economic Trials of Liberia.’’ West Africa, March 27–April 2, 461–463.
Barro, R. J. 1996. ‘‘Determinants of Economic Growth: A Cross-Country Empirical Study.’’ Working
Paper No. 5698, NBER, Cambridge, MA.

Barro, R. J., and Lee, J. 1996. ‘‘International Data on Educational Attainment: Updates and Implica-
tions.’’ Working Paper No. 42, CID, Harvard University, Cambridge, MA.

Barro, R. J., and X. Sala-i-Martin. 1995. Economic Growth. New York: McGraw Hill.
Berkeley, B. 1992. ‘‘Liberia’s Warring Currencies.’’ Institutional Investors, September 1992.
Blundell, R., A, and Costa Dias, M. 2000. ‘‘Evaluation Methods for Non-Experimental Data.’’ Fiscal
Studies 21: 427–468.
Bogetic, Z. 2000. ‘‘Official Dollarization: Current Experiences and Issues.’’ Cato Journal 20, no. 2: 179–
213.

Calvo, Guillermo A. 1999. ‘‘Fixed versus Flexible Exchange Rates.’’ Mimeo., University of Maryland.
———. 2000. ‘‘Balance of Payments Crises in Emerging Markets: Large Capital Inflows and Sovereign
Governments.’’ In Currency Crises, ed. Paul Krugman, 71–98. Chicago: University of Chicago Press.
———. 2001. ‘‘Economic Policy in Stormy Waters: Financial Vulnerability in Emerging Economies.’’
Journal of Applied Economics 4, no. 1: 1–25.

Calvo, Guillermo A., Alejandro Izquierdo, and Ernesto Talvi. 2003. ‘‘Sudden Stops, the Real Exchange
Rate, and Fiscal Sustainability: Argentina’s Lessons.’’ Working Paper No. 9828, NBER, Cambridge,
MA.
Calvo, Guillermo A., and Carmen M. Reinhart. 2002. ‘‘Fear of Floating.’’ Quarterly Journal of Economics
117, no. 2: 379–408.
Cooper, R. W., and H. Kempf. 2001. ‘‘Dollarization and the Conquest of Hyperinflation in Divided
Societies.’’ Federal Reserve Bank of Minneapolis Quarterly Review 25, no. 3: 3–12.
Corden, W. M. 2002. Too Sensational: On the Choice of Exchange Rate Regimes. Cambridge, MA: MIT
Press.
Dornbusch, R. 2001. ‘‘Fewer Monies, Better Monies.’’ American Economic Review 91, no. 2: 238–242.
Edwards, S. 2001a. ‘‘Dollarization: Myths and Realities.’’ Journal of Policy Modeling 23, no. 3: 249–265.
———. 2001b. ‘‘Exchange Rate Regimes, Capital Flows, and Crisis Prevention.’’ In Economic and Finan-
cial Crises in Emerging Market Economies, ed. Martin, Feldstein, 31–78. Chicago: University of Chicago
Press.
Edwards, S., and I. I. Magendzo. 2002. ‘‘Dollarization, Inflation and Growth.’’ Working Paper No. 8671,
NBER, Cambridge, MA.
Edwards, S., and I. I. Magendzo. 2006. ‘‘Strict Dollarization and Economic Performance.’’ Journal of
Money, Credit and Banking 38, no. 1: 269–282.
Eichengreen, B. 2001. ‘‘What Problems Can Dollarization Solve?’’ Journal of Policy Modeling 23, no. 3:
267–277.

Eichengreen, B., and R. Haussmann. 1999. ‘‘Exchange Rates and Financial Fragility.’’ Paper presented
at the IDB-OECD Forum, November.

Fischer, S. 1976. ‘‘Stability and Exchange Rate Systems in a Monetarist Model of the Balance of Pay-
ments.’’ In The Political Economy of Monetary Reform, ed. Robert Z. Aliber, 59–73. New York: McMillan.
Hard Currency Pegs and Economic Performance 155

———. 2001. ‘‘Exchange Rate Regimes: Is the Bipolar View Correct?’’ Journal of Economic Perspectives
15, no. 2: 3–24.
Frankel, J. A. 1999. ‘‘No Single Currency Regime is Right for All Countries at at All Times.’’ Working
Paper No. 7338, NBER, Cambridge, MA.
Frankel, J. A., and A. K. Rose. 2002. ‘‘An Estimate of the Effect of Common Currencies on Trade and In-
come.’’ Quarterly Journal of Economics 117, no. 2: 437–466.
Frenkel, J. A., and A. Razin. 1987. Fiscal Policies and Growth in the World Economy. Cambridge, MA: MIT
Press.
Frieden, J. 2003. ‘‘The Political Economy of Dollarization.’’ In Dollarization, eds. E. Levy-Yeyati and F.
Sturzenegger, 305–333. Cambridge, MA: MIT Press.
Ghosh, A., A. Gulde, J. Ostry, and H. Wolf. 1995. ‘‘Does the Nominal Exchange Rate Regime Matter?’’
Working Paper No. 95/121, IMF, Washington, D.C.
Goldfajn, I., and G. Olivares. 2000. ‘‘Is Adopting Full Dollarization the Solution? Looking at the Evi-
dence.’’ Working Paper No. 416, Pontificia Universidade Católica do Rio de Janeiro.
Greene, W. H. 2000. Econometric Analysis. New York: Macmillan.
Heckman, J. 1978. ‘‘Dummy Endogenous Variables in a Simultaneous Equation System.’’ Econometrica
46, no. 4: 931–959.
Heckman, J. J., I. Hidehiko, and P. E. Todd. 1997. ‘‘Matching as an Econometric Evaluation Estimator:
Evidence from Evaluating a Job Training Programme.’’ Review of Economic Studies 64, no. 4: 605–654.
IMF. 1997. ‘‘Exchange Rate Arrangements and Economic Performance in Developing Countries.’’ In
World Economic Outlook. Washington, D.C.: IMF.
Klein, M. W. 2002. ‘‘Dollarization and Trade.’’ Working Paper No. 8879, NBER, Cambridge, MA.
Leamer, E. E. 1997. ‘‘Access to Western Markets and Eastern Effort.’’ In Lessons from the Economic
Transition: Central and Eastern Europe in the 1990s, ed. Salvatore Zecchini, 503–526. Dordrecht:
Kluwer Academic Publishers.

Lechner, M. 2002. ‘‘Program Heterogeneity and Propensity Score Matching: An Application to the
Evaluation of Active Labor Market Policies.’’ Review of Economics and Statistics 84, no. 2: 205–220.
Levy-Yeyeti, E., and F. Sturzenegger. 2001. ‘‘To Float or to Trail: Evidence on the Impact of Exchange
Rate Regimes.’’ Working Paper No. 1/01, Universidad Torcuato di Tella.
Levy-Yeyeti, E., and F. Sturzenegger. 2003. Dollarization. Cambridge, MA: MIT Press.
Maddala, G. S. 1983. Limited-Dependant and Qualitative Variables in Econometrics. New York: Cambridge
University Press.
Meade, J. E. 1950. The Balance of Payments. New York: Oxford University Press.

Moreno-Villalaz, J. L. 1999. ‘‘Lessons from the Monetary Experience of Panama: A Dollar Economy
with Financial Integration.’’ Cato Journal 18, no. 3: 421–439.
Mundell, R. A. 1961. ‘‘A Theory of Optimum Currency Areas.’’ American Economic Review 51, no. 4:
657–665.
Parrado, E., and A. Velasco. 2002. ‘‘Optimal Interest Rate Policy in a Small Open Economy.’’ Working
Paper No. 8721, NBER, Cambridge, MA.
156 Sebastian Edwards and I. Igal Magendzo

Powell, A., and F. Sturzenegger. 2000. ‘‘Dollarization: The Link Between Devaluation and Default
Risk.’’ Working Paper, Universidad Torcuato di Tella, Buenos Aires, Argentina.
Rose, A. 2000. ‘‘One Money, One Market: Estimating The Effect of Common Currencies on Trade.’’
Economic Policy 15, no. 30: 7–46.
Rose, A. K., and C. Engel. 2000. ‘‘Currency Unions and International Integration.’’ Working Paper No.
7872, NBER, Cambridge, MA.
———. 2002. ‘‘Currency Unions and International Integration.’’ Journal of Money, Credit, and Banking 34,
no. 4: 1067–1089.
Rose, A. K., and E. Van Wincoop. 2001. ‘‘National Money as a Barrier to International Trade: The Real
Case for Currency Union.’’ American Economic Review 91, no. 2: 386–390.
Rosenbaum, P. R., and D. B. Rubin. 1983. ‘‘The Central Role of the Propensity Score in Observational
Studies for Causal Effects.’’ Biometrika 70, no. 1: 41–55.
Sachs, J., and A. Warner. 1995. ‘‘Economic Reform and the Process of Global Integration.’’ Brookings
Papers on Economic Activity 1995, no. 1: 1–188.
Sachs, J. D. 2000. ‘‘Globalization and Patterns of Economic Development.’’ Weltwirtschaftliches Archiv/
Review of World Economics 136, no. 4: 579–600.
Summers, L. H. 2000. ‘‘International Financial Crises: Causes, Prevention, and Cures.’’ American Eco-
nomic Review 90, no. 2: 1–16.

Venables, A. J., H. Overman, and S. Redding. 2002. ‘‘The Economic Geography of Trade, Production
and Income.’’ Working Paper No. 0508, CEPR, London.

Wooldridge, J. M. 2002. Econometric Analysis of Cross-Section and Panel Data. Cambridge, MA: MIT
Press.
III Financial Crises
7 Asset Prices and Self-Fulfilling
Macroeconomic Pessimism

Andrés Velasco and Alejandro

Neut

7.1 Introduction

The volatility in asset prices (especially stock prices) over the last few years, both
in the United States and elsewhere, has caused economists to wonder about the
effects of movements in these prices on aggregate demand and economic activity.
One standard link runs from stock prices to private wealth, and then to consump-
tion demand and income. Feeling suddenly poor after a bear market, consumers
can curtail their spending, pushing the economy even further down. But several
econometric studies have looked at the link between financial assets and con-
sumption only to conclude that the elasticities involved are relatively small.1 As
long as fundamentals are in place, volatility in asset prices will have a manageable
effect on output.
Another possible effect runs from asset prices to investment demand. As Guil-
lermo Calvo famously emphasized (and foretold) in 1994, a sudden loss of inves-
tor confidence can reduce the value of assets and trigger pernicious dynamics that
drive a small open economy like Mexico into recession.
This line of reasoning is not applicable only to small open economies. In a
closed economy the link between asset prices and investment is also present.
And, as this paper shows, it may generate multiple equilibria. A bear market
reduces the value of the collateral held by households and firms, which in turn
cuts their ability to borrow and invest, again pushing output down. But that need
not be the end of the story. Changes in activity affect future returns, which in turn
affect current stock prices. Circular causation can occur, which inevitably raises
the question of whether movements in asset prices and economic activity are
based on self-fulfilling beliefs.
Can it be the case that any event (a terrorist attack? a political crisis? a poor
night’s sleep?) causes a bout of depression and a fall in the stock market, which
in turn triggers a fall in investment and a recession, which justifies the initial pes-
simism? This paper provides an extremely simple model that delivers such a
result.
160 Andrés Velasco and Alejandro Neut

Multiple equilibria under rational expectations can arise in very different set-
tings. In Calvo’s seminal paper (1988), the denomination of government bonds
was at the root of the problem. In this case the difficulty lies with a particular fi-
nancial market imperfection: the need to collateralize borrowing. This difference
has policy implications. In Calvo (1988), interest rate caps could rule out the bad
equilibrium. Here, interest rate caps or subsidies may help improve the good
equilibrium but do nothing to eliminate the bad outcome. By contrast, policies
that increase aggregate demand and output may succeed in ruling out the un-
desirable equilibrium.

7.2 The Model

The model has two periods, one good, and two kinds of people, capitalists and
workers. Workers supply labor and consume. Capitalists own the factors of pro-
duction other than labor, which they rent to firms and also consume. They finance
investment in excess of their own resources by borrowing from workers. The key
aspect of the model is that such borrowing is constrained by the need to put up
collateral, and the value of this collateral in turn depends on asset prices.

7.2.1 Domestic Production

Production is carried out by competitive firms. Each firm has access to the Cobb-
Douglas technology

Yt ¼ Kta Ltg Nt1
a
g ; 0<aþg<1 ð7:1Þ

where K denotes capital, L land and N labor. Without loss of generality, we as-
sume capital depreciates completely in one period. Land L is fixed, and we nor-
malize Lt ¼ 1 from now on.
In each one of the two periods cost-minimizing firms choose capital, land, and
labor according to
 
Yt
RtK ¼ a ð7:2Þ
Kt
 
L Yt
Rt ¼ g ð7:3Þ
Lt
 
Yt
Wt ¼ ð1
a
gÞ ð7:4Þ
Nt
Asset Prices and Self-Fulfilling Macroeconomic Pessimism 161

where RtK , RtL , and Wt are the factor returns to capital, land, and labor. The
numéraire in this economy is the price of output Y, which is used for consumption
and investment in capital for the following period.

7.2.2 Workers

Workers consume and they supply one unit of labor inelastically (so that N ¼ 1 by
assumption from now on), for which they are paid labor’s marginal return. As
consumers, they maximize a standard two-period utility function subject to the
budget constraint
C2 W2
C1 þ ¼ D þ W1 þ ; ð7:5Þ
1þr 1þr

where D is accumulated wealth saved by workers (and consequently borrowed

by capitalists) in the past and r is the market rate of interest for bond transactions
between these two groups. Period 1 savings by workers are also channeled to cap-
italists, who invest in either land or capital.
Using equations 7.1 and 7.4, budget constraint 7.5 can be written as

C2 ð1
a
gÞK2a
C1 þ ¼ D þ ð1
a
gÞK1a þ ; ð7:6Þ
1þr 1þr

where K2 is obtained through investment in capital in the first period. Capital K1

and D (assets for the workers, debts for the capitalists) are given by history. The
solution to the maximization problem faced by workers boils down to the savings
function

S1 ¼ f ½r; ð1
a
gÞK1a þ B; ð1
a
gÞK2a : ð7:7Þ

From now on, assume the ‘‘normal’’ case in which f is increasing in r.2

7.2.3 Capitalists

Capitalists are the key players in the model: they finance spending partly with
loans, and their borrowing is subject to frictions. They consume in the closing pe-
riod only. Their objective is to maximize the utility from such consumption, which
boils down to maximizing the amount consumed. They only own land and capi-
tal, but the latter depreciates completely once used. Qt denotes the market value
of land in period t (after returns are paid). In this simple two-period model, Q1 is
endogenous and Q2 ¼ 0.
162 Andrés Velasco and Alejandro Neut

At the beginning of each period, capitalists collect the income from capital and
land and repay debt (to workers). In the first period, their net resources available
for investment are

N1 1 R1K K1 þ R1L L
D ¼ ða þ gÞY1
D ð7:8Þ

where the second equality comes from 7.2 and 7.3. Although individual capitalists
have additional resources in the form of land, the value of land Q1 does not enter
equation 7.8. The reason is that land is only traded among capitalists; hence there
is no additional aggregate net value that can be diverted to acquire capital. Notice
that N1 is exogenous because D is given and so is Y1 : aggregate land and capital
are inherited, and labor is inelastically supplied.
The capital stock available in the second period is K2 , equal to investment in pe-
riod 1.3 Capitalists can invest in additional capital subject to the budget constraint
K2 ¼ N1 þ B1 ð7:9Þ

where B1 is the amount borrowed by capitalists in period 1.

A crucial assumption is that, because of limitations in contract enforcement and
the like, in the case of nonpayment lenders can seize at most the value and returns
to land, net of enforcement costs, equal to L. Hence workers will not lend at the
initial time an amount generating obligations larger than the value of collateral:4,5
B1 a maxf0; Q1
Lg: ð7:10Þ

7.2.4 Market Clearing

Capitalists can only borrow from workers, so that

B1 ¼ S1 : ð7:11Þ

Equations 7.7, 7.9, and 7.11 together yield

K2
ða þ gÞK1a þ A ¼ f ½r; ð1
a
gÞK1a þ D; ð1
a
gÞðK1 þ K2 Þ a : ð7:12Þ

This is an implicit expression for r as a function of K2 and a pair of exogenous

variables

r ¼ fðK2 ; K1 ; DÞ: ð7:13Þ

It is easy to show that, under the assumptions made on f , f is increasing in K2

and D. The partial relationship between the interest rate and K1 is ambiguous.
Asset Prices and Self-Fulfilling Macroeconomic Pessimism 163

7.2.5 Equilibria

Next we define three schedules that jointly determine the possible equilibria of the
model.

1. LK schedule.
Equating the marginal returns of land and capital implies
R2L
¼ R2K : ð7:14Þ
Q1
Using 7.2 and 7.3 in 7.14 one obtains
 
g
Q1 ¼ K2 : ð7:15Þ
a
This schedule, which we term LK (for no arbitrage between land and capital),
gives the equilibrium price of land today as a function of investment in capital to-
day. As K2 rises, tomorrow’s capital-land ratio goes up, increasing the marginal
product of land and hence raising Q1 .
2. KB schedule.
If capitalists are not financially constrained and can borrow as much as they want,
they maximize their next-period consumption by choosing an amount of capital
investment such that

R2K ¼ 1 þ r: ð7:16Þ
Using 7.2 and 7.13, this equation implies

K2 ½1 þ fðK1 ; K2 ; DÞ 1=ð1
aÞ ¼ a 1=ð1
aÞ ð7:17Þ

where the LHS is unambiguously increasing in K2 . We denote this schedule KB
because it corresponds to no-arbitrage between capital and bonds. Let K2 be the
level of capital that solves 7.17. We assume K2 > N1 , so that capitalists will indeed
want to borrow.
But financial constraints may not let capitalists borrow the resources to invest
as much capital as they want, in which case capital-bonds arbitrage does not ob-
tain. In those constrained situations the amount of investment is bounded from
above as follows:

K2 ½1 þ fðK1 ; K2 ; DÞ 1=ð1
aÞ a a 1=ð1
aÞ : ð7:18Þ

3. FC schedule.
The fact that borrowing is constrained can lead investment to be constrained.
Combining 7.10 and 7.9 one has
K2 a maxfN1 ; Q1
L þ N1 g: ð7:19Þ
164 Andrés Velasco and Alejandro Neut

Rearrange 7.19 to read

Q1 b K 2 þ L
N1 ; if K2 b N1 : ð7:20Þ
This inequality shows that for every level of planned investment K2 , the price of
land Q1 must be sufficiently high for that investment to be feasible. We term this
the FC (financial constraint) schedule.6

7.3 Outcomes with and without Crises

The model can be solved quite simply using a diagrammatic representation in Q1 ,

K2 space. Notice that LK always holds because there are no rigidities or con-
straints in arbitraging between land and capital. Therefore, to solve for K2 and
Q1 one needs one additional equation. There are two candidate inequalities,
depicted by schedules KB and FC. At least one of these inequalities must hold
with equality. The reason is straightforward: if KB is not binding, the return to
capital is greater than r. For this to be the case, capitalists must be financially
constrained—in other words, the FC schedule is binding. On the other hand, if
FC is not binding, this means that capitalists are financially unconstrained, and
therefore KB holds with equality.
We therefore have the following possible cases:
• Case where KB is binding:
This is a situation where LK and KB together yield the level of investment K2 that
capitalists would like to undertake if unconstrained (meaning as long as the in-
equality in schedule FC still holds). We show this case in figures 7.1 and 7.2.
• Case where FC is binding:
In this case firms do not have enough collateral to obtain any additional credit.
Schedules LK and FC determine an equilibrium as long as capitalists are willing

Figure 7.1
Asset Prices and Self-Fulfilling Macroeconomic Pessimism 165

Figure 7.2

Figure 7.3

to invest more under those circumstances (that is, as long as the inequality in KB
holds). We show this case in figures 7.2 and 7.3.
Figures 7.1 and 7.3 involve a unique equilibrium, at point A in each case. Figure
7.2 involves multiple equilibria at points A and B. The bad equilibrium implies
no borrowing (K2 ¼ N1 ) and lower asset prices than does the good equilibrium.
A key exogenous variable is N1 . The lower N1 , the farther to the left is FC, possi-
bly taking the economy from the one-equilibrium case in figure 7.1 to the two-
equilibria case in figure 7.2.
A necessary and sufficient condition for a bad equilibrium with no borrowing
to exist (whether unique as in figure 7.3, or non-unique as in figure 7.2) is that the
FC schedule be above the LK schedule at K2 ¼ N1 . That is to say

g
Lb N1 : ð7:21Þ
a
166 Andrés Velasco and Alejandro Neut

Hence, if enforcement costs L are sufficiently small, the no-borrowing bad equilib-
rium disappears. Notice that this condition need not be too restrictive, since given
the definition of net worth (N1 ¼ ða þ gÞY1
D), it could be near zero or even
negative if inherited debts are large.
The intuition of multiple equilibria is simple: there is feedback between asset
prices and aggregate demand. A higher land price allows the capitalists to borrow
and invest more in capital. But at the same time, higher capital investment raises
the marginal product of land and therefore its price. For some parameter values,
the circularity opens the door for multiplicity and self-fulfiling beliefs.

7.4 Policy Alternatives

This model is too simple to say much about policy, but a few points are sugges-
tive. Focus on the case of multiple equilibria only, and look for ways to eliminate
the bad outcome.

Expansionary Fiscal and Monetary Policies

To the extent they can raise current income Y1 and consequently raise N1 , these
policies can move FC to the right until condition 7.21 no longer holds.7 A suffi-
ciently large expansion will cause the FC curve to be below the LK schedule at
K2 ¼ N1 , eliminating the bad equilibrium in figure 7.5. The intuition is that with
larger current output the capitalists’ gross resources available for investment rise
relative to debt due, so they can afford to invest more for every price of land. In
addition, higher Y1 raises asset prices for any level of investment. The combina-
tion can rule out self-fulfilling pessimistic animal spirits.

Debt Forgiveness or Rescheduling

A reduction in D (debt could be written down, paid by the government, or invol-
untarily reprogrammed to period 2) also increases N1 , moving the vertical portion
of the FC schedule to the right while leaving LK unchanged. A sufficiently large
cut in the current debt burden would rule out the bad equilibrium by causing FC
to be below LK at K2 ¼ N1 . This case also corresponds to figure 7.4, with a unique
equilibrium at point A. The intuition is the same as in the previous case: with
improved cash flow, the capitalists can afford to invest even if asset prices are
low.

Financial and Legal Reform

A reduction in L brings about a downward movement in FC, as in figure 7.5. For
a sufficiently small L, the bad equilibrium disappears. The single equilibrium can
Asset Prices and Self-Fulfilling Macroeconomic Pessimism 167

Figure 7.4

Figure 7.5

be either unconstrained (as drawn) or constrained. In the new single equilibrium,

capitalists can borrow and invest as much as they would like.
There are other policies that may have advantages even though they cannot
eliminate a bad equilibrium when one exists. Consider, for instance, subsidies to
savings. To the extent such a policy lowers the value of r ¼ fðK1 ; K2 ; DÞ for any
level of K2 , it shifts up the KB curve (see figure 7.6). But FC is unchanged, so there
is no interest rate that can free the economy from the possibility of a bad equilib-
rium. An alternative way to see this is to note that r does not enter condition 7.21,
so changes in interest rates alone cannot affect the number of feasible equilibria.
The intuition is that a bad equilibrium is the consequence of asset prices that are
so low capitalists end up financially constrained when they do not borrow any-
thing. In that situation, the present value of their collateral is negative.8 Therefore,
no matter how low interest rates go, the present value of that collateral remains
negative and the bad equilibrium does not disappear.
168 Andrés Velasco and Alejandro Neut

Figure 7.6

7.5 Conclusions

The model in this chapter involves perfect competition and well-functioning mar-
kets in all respects but one: there is an imperfection in the capital market that
causes borrowing to be collateralized. Since the value of the collateral depends on
current asset prices, and asset prices depend on current expenditure and future
profits, this opens the door to self-fulfilling expectations. But in certain conditions,
policies can rule out the bad outcome, if one exists.
The mechanism in the paper is quite specific. But the result that small devia-
tions from the perfect financial markets paradigm can generate more than one
equilibrium is not. Other asset prices—for instance, the exchange rate in an open
economy—can play the same role. For an example see Velasco (2001).

Notes

An earlier version of this paper was circulated under the title ‘‘The Macroeconomics of Terrorism.’’
Generous financial support from the National Science Foundation and the Harvard Center for Interna-
tional Development is acknowledged with thanks. We are grateful to the members of LIEP at Harvard
University for useful comments.

1. Housing wealth has a significantly larger effect on consumption as estimated by Case, Quigley, and
Shiller (2001).
2. That is, assume the substitution and wealth effects of movements in interest rates are greater than
the income effect.
3. Recall capital depreciates completely. Introducing a depreciation rate for capital less than one would
change nothing, but would make the algebra more cumbersome.
4. This is as in Kiyotaki and Moore (1997), Krugman (1999), and Aghion, Bachetta, and Banerjee
(2001), among many others.
Asset Prices and Self-Fulfilling Macroeconomic Pessimism 169

5. A possible objection is that the collateral constraint should depend on period 2 variables—that is to
say, on the ‘‘stuff’’ lenders can seize in the event of nonpayment. The same results, but with somewhat
more cumbersome algebra, would obtain if we adopted that specification. For instance, the case in
which lenders can seize the total returns earned by the capitalist in period 2. The constraint would be
ð1 þ rÞB1 a maxf0; ða þ gÞY2
Lg:
See note 6 for details.
6. If instead we had assumed the constraint
ð1 þ rÞB1 a maxf0; ða þ gÞY2
Lg;
the FC schedule would be

Q1 b u½ðK2
N1 Þð1 þ fðK2 ; K1 ; BÞÞ þ L 1=a if K2 > N1 ;

where u is a positive constant and the RHS of the inequality is unambiguously increasing in K2 . The
only difference would be that the FC is no longer linear in K2 .
7. Of course, extending the model to allow for such Keynesian effects would require labor supply to be
endogenous and prices to be sticky. Both can be added without much complication.
8. This is the case for any ‘‘bad’’ equilibrium; for instance, the equilibrium in figure 7.3.

References

Aghion, P., P. Bacchetta, and A. Banerjee. 2001. ‘‘Currency Crises and Monetary Policy in an Economy
with Credit Constraints.’’ European Economic Review 45, no. 7: 1121–1150.
Calvo, G. 1988. ‘‘Servicing the Public Debt: The Role of Expectations.’’ American Economic Review 78:
647–661.
Calvo, G. 1994. ‘‘Comments on Dornbusch-Werner.’’ Brookings Papers on Economic Activity, no. 2: 299–
303.

Case, K., J. Quigley, and R. Shiller. 2001. ‘‘Comparing Wealth Effects: The Stock Market versus the
Housing Market.’’ Working Paper No. 8606, NBER, Cambridge, MA.
Kiyotaki, N., and J. Moore. 1997. ‘‘Credit Cycles.’’ Journal of Political Economy 105, no. 2: 211–248.

Krugman, P. 1999. ‘‘Balance Sheets, the Transfer Problem and Financial Crises.’’ In International Finance
and Financial Crises: Essays in Honor of Robert Flood, eds. P. Isard, A. Razin, and A. Rose, 33–44. Norwell,
MA: Kluwer Academic Publishers.
Velasco, A. 2001. ‘‘The Impossible Duo? Globalization and Monetary Independence in Emerging Mar-
kets.’’ Brookings Trade Forum 2001: 69–99.
8 The Center and the Periphery:
The Globalization of Financial
Turmoil

Graciela L. Kaminsky and Carmen

Reinhart

The first springs of great events, like those of great rivers, are often mean and little.
—Jonathan Swift

8.1 Introduction

A succession of crises in emerging-market economies during the 1990s ignited a

debate in academic and policy-making circles about the transmission of shocks
across national borders. The spreading market strain surrounding the Mexican
peso crisis of 1994, the Asian credit crunch of 1997, and the Russian devaluation
and Long-Term Capital Management’s (LTCM) implosion, both in 1998, have
spawned a body of work that can be summarized under the heading ‘‘contagion.’’
These episodes have also resuscitated the interest in codes and standards, mone-
tary arrangements, the role of international institutions, and securities law—
summarized as the international financial architecture—to construct a barrier to
prevent contagion.
The academic literature on contagion—less pejoratively described as spill-
overs—and the international propagation of shocks has progressed along two
roads in recent years. Early studies attempted to document the existence of conta-
gion. More recent papers have primarily sought to discriminate among the possi-
ble transmission channels of disturbances—that is, whether shocks propagate
through channels established by trade patterns, geography, commonalities among
lenders, or from other sources.
In our view, much of this literature suffers from three serious drawbacks.
First, most studies have not discriminated between the origins of shocks. One
expects, a priori, that the global or regional consequences of a disturbance may
depend importantly on whether the shock—to borrow terms from Sir Arthur
Lewis—originates in the periphery or in the center (Lewis 1977). Were the
172 Graciela L. Kaminsky and Carmen Reinhart

regional consequences of the Thai crisis so severe because of Thailand’s direct

links with other countries in the region or because that shock affected the region’s
largest economy—Japan? Were the paralysis of the bond markets in many parts
of the globe and the persistent equity market volatility due to the Russian default
or to concerns that LTCM’s reach was wider than understood, and that other
firms in other financial centers of the world shared similar failings? And what
about the dog that did not bark: why did Ecuador’s 1999 default not have greater
international consequences? In contrast, this paper attempts to capture the origins
of systemic turbulences and measure the direct and indirect linkages between na-
tional markets.
Second, terms like contagion and spillover can be quite slippery. Some authors
seek to learn about potential linkages by examining correlation patterns across
markets using long time series. Others focus on market behavior during specific
episodes dated a priori from other sources. Our approach is data driven. We de-
fine financial turmoil in terms of the behavior of financial prices and let the data
determine when there were episodes of spillover. In particular, we define days of
turmoil as days with extreme returns.1 Using information on the daily behavior of
stock-market price indexes for thirty-five emerging-to-mature market economies
from January 1997 to August 1999, we examine what happens in stock markets
around the world on days of turmoil in financial centers (Germany, Japan, and
the United States) and in crisis-prone emerging economies (Brazil, Russia, and
Thailand).
Third, most of the studies of financial spillovers rely on an examination of con-
temporaneous and lagged correlations. But correlations alone cannot tell those
systemic turbulences due to a common shock (say, a large change in oil prices or
the announcement of election results in an important country) from true spillover
(a change in one national market directly related to extreme price movements in
another market). To learn about the determinants of systemic financial turmoil,
we turn to newspapers and summarize the key world events associated with sig-
nificant price changes. In many cases, this allows us to identify the source of the
shock—the center or the periphery—and to understand better the temporal di-
mension of the financial market adjustment.
To be more specific about spillovers, we borrow from the literature on financial
market efficiency to distinguish between ‘‘weak-form globalization’’ and ‘‘strong-
form globalization’’ of turmoil. Weak-form globalization occurs when country j
experiences anomalous returns in days of extreme returns in country i, where
anomalous behavior is interpreted as a change in the distribution of returns
assessed by a nonparametric procedure. This definition does not require the coun-
tries suffering the spillovers to have extreme returns (that is, to be in the 5th and
95th percentile). Strong-form globalization occurs when country j experiences tur-
The Globalization of Financial Turmoil 173

moil when country i has extreme returns in the stock market. That is, it is a state-
ment about simultaneity of extreme returns. Using these definitions, we construct
two indices of globalization and examine the patterns of spillover among crisis-
prone emerging markets and financial centers.
While the analysis of more episodes is clearly necessary, one preliminary con-
clusion we draw is that financial shocks often traverse a circuitous route. Prob-
lems occur synchronously in many emerging markets on the periphery because a
shock in one of them first influenced a financial center. If the shock never reaches
the center, it is doubtful it can become systemic, irrespective of the definition of
systemic that is used. For example, in the case of the Asian crisis, Japanese bank
exposure to Thailand—and their subsequent retrenchment from lending to other
Asian countries—played a prominent role in the spread of the crisis. The role
played by the center ( Japan) in this episode was much the same as that played by
U.S. banks in the 1980s during the Latin American debt crisis. Similarly in 1998,
Russia’s default triggered a pervasive widening of spreads that hobbled the weak-
ened LTCM and led to a generalized withdrawal of risk taking. Thus, financial
centers serve a key role in propagating financial turmoil. When financial centers
remain safe, problems in an emerging market stop at the region’s border.
The rest of the paper is organized as follows: section 2 presents a brief discus-
sion of some of the analytical issues relevant to our analysis of the globalization
of financial turmoil. Section 3 constructs the two indices of globalization of finan-
cial turmoil and examines the pattern of spillover within and across regions. Sec-
tion 4 discusses the origins of high degree of contagion. Concluding remarks are
presented in section 5.

8.2 Analytical Issues

For the purposes of our analysis, we divide the world into center and periphery
countries. The former consist of the countries that house the largest financial cen-
ters (such as New York, London, Berlin, and Tokyo) while the latter comprise
everyone else. We distinguish among three patterns in the propagation of shocks.
First, there is the transmission of shocks from one periphery country to another
periphery country, which can take place if the two countries are directly linked
through bilateral trade or finance (figure 8.1). Recent examples of this type of
transmission mechanism include the adverse impact of the 1997–1998 Asian crisis
on Chilean exports and the contractionary impact on Argentina of the Brazilian
devaluation in January 1999. This transmission channel may also be operative if
there are bilateral finance links. For instance, Costa Rican banks were borrowing
from Mexican banks on the eve of the Mexican crisis (see Calvo and Reinhart
1996), but when Mexican banks ran into trouble this source of funds disappeared.
174 Graciela L. Kaminsky and Carmen Reinhart

Figure 8.1
The Transmission of Shocks from One Periphery Country to Another

Figure 8.2
The Transmission of Shocks from one Periphery Country to Another Through a Center Country

Second, there is the transmission of shocks from one periphery country to an-
other via a center country (as shown in figure 8.2). There are several prominent
examples of this type of transmission mechanism in the literature. Corsetti,
Pesenti, Roubini, and Tille (1998) model trade competition among the periphery
countries in a common third ‘‘center’’ market. For instance, Thailand and Malay-
sia export many of the same goods to Japan, Hong Kong, and Singapore. Hence,
when Thailand devalued in mid-1997, the crisis spread to Malaysia, who lost
some of its competitive edge in the common third markets. Calvo (1998) suggests
that Wall Street may have been the carrier of the ‘‘Russian virus’’ in the fall of
1998; he focuses on asymmetric information and liquidity problems in the finan-
cial centers. So, when Russia (a periphery country) defaulted on its bonds, the
leveraged investors that held those bonds in the center country faced margin calls
and needed to raise liquidity. The margin calls caused them to sell their asset
holdings (the bonds and stock of other countries in their portfolio) to an unin-
formed counterpart. Because of information asymmetries, a ‘‘lemons problem’’
arises and the assets are sold at a fire sale price.2
A variant of this financial center story concerns open-end fund portfolio man-
agers who need to raise liquidity in anticipation of future redemptions. As before,
the strategy would be to sell other assets held in the portfolio. The sell-off
The Globalization of Financial Turmoil 175

depresses the asset prices of other countries and the original disturbance spreads
across markets. Frankel and Schmukler (1998) find evidence suggesting that the
crisis in Mexico in late 1994 spread to other equity markets in Latin America
through New York rather than directly. Kaminsky, Lyons, and Schmukler (2001),
who examine the behavior of the mutual fund industry in international equity
markets, support this venue of spillover. They find that in the aftermath of the
Thai crisis, the largest mutual fund withdrawals affected Hong Kong and Singa-
pore, which have the most liquid financial markets. Kaminsky and Reinhart (2000
and 2001) focus instead on the role of commercial banks lenders in the center
country. They stress that following the initial losses due to a crisis in a periphery
country, a bank’s need to rebalance the overall risk of its asset portfolio can lead
to a marked reversal in commercial bank lending across the markets where the
bank has exposure. By calling loans and drying up credit lines to the crisis coun-
try, center banks deepen the original crisis. Through the act of calling loans else-
where, they propagate the crisis to other countries. The debt crisis in the early
1980s and the Asian crisis in 1997 provide two clear examples of this mechanism.
Following Mexico’s default in 1982, U.S. banks with extensive exposure to Mexico
spread the crisis across Latin America. In 1997, Japanese banks, heavily exposed
to Thailand, played the same role in spreading the crisis throughout Asia.
Third, there is the transmission of symmetric shocks from the center country to
the periphery (figure 8.3). This is the type of shock stressed in several papers by
Calvo, Leiderman, and Reinhart (1993, 1996), who analyze the effect of changes
in U.S. interest rates on capital flows to Latin America in the early part of the
1990s. While an obvious example of this type of shock is changes in interest rates
in a financial center country, more subtle ones may include the kinds of regula-
tory changes in the financial centers or structural shifts in financial markets. As
an example of the latter, the closure of Salomon Brothers’ bond arbitrage desk on
July 6, 1998 is thought to have been a factor contributing to the loss of liquidity in
the market for emerging-market bonds, making the markets less resilient and
impairing LTCM’s prospects.

Figure 8.3
Symmetric Shocks from Center to Periphery
176 Graciela L. Kaminsky and Carmen Reinhart

8.3 Financial Globalization: Measures and Determinants

The crises of the 1990s triggered an immense interest in understanding extreme

events. The literature in the late 1990s focused mostly on extreme events in the ex-
change market as captured by exchange rate devaluations, foreign exchange re-
serve losses, and in some cases, spikes in overnight interest rates.3 The goal was
to examine whether deteriorating fundamentals were at the root of these crises.
But when a variety of countries started to topple like dominoes, many authors
began to focus on the characteristics of contagion.4 This literature also deals with
issues of systemic risk. But systemic risk is not just connected to currency crises.
Systemic risk may also be triggered by shocks in banking and stock or bond mar-
kets. Our goal in this section is to present a measure of systemic events triggered
by turmoil in a financial market in one country.5
We understand financial turmoil as an extreme event in a financial market, be it
a rally or a crash. That is, we confine our definition of extreme events to the tail of
the distributions of returns by looking at returns in the 5th and 95th percentile of
the distribution. Because our interest is in systemic events, we have to consider a
substantially large number of markets. Sometimes these systemic events are not
long-lasting (for example, the worldwide stock market crash in October 1987),
implying a need to use high-frequency data. This puts some restrictions on the
markets we can examine. To accommodate these needs, we focus on daily returns
in stock markets.
Our data set spans the period beginning on January 1, 1997 through August 31,
1999. We focus on the daily return on equities in the local currency, based on the
available local bourse indices. The countries in our sample cover mature- and
emerging-market economies, thereby excluding countries with less-developed
capital markets and a significant extent of financial repression. We can classify
the sample into five somewhat arbitrary seven-country groupings: the G7 coun-
tries, which are comprised of Canada, France, Germany, Italy, Japan, the United
Kingdom, and the United States; and the transition economies, which include
Czech Republic, Estonia, Hungary, Poland, Russia, Slovakia, and Ukraine. The
remaining three groups are primarily sorted by region. The Asian cluster includes
Hong Kong, Indonesia, Malaysia, the Philippines, Singapore, South Korea, and
Thailand. The other European group, which excludes those countries that are
part of the G7, includes Finland, Greece, Holland, Norway, Spain, Sweden, and
Turkey. Finally, the Latin American sample consists of the larger economies in
the region, Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela.
Table 8.1 provides summary information on the stock markets examined.
Not surprisingly, the degree of instability of stock returns varies considerably
across countries. Thus, our definition of extreme events is country-dependent. For
example, a drop of 1.8 percent is classified as an extreme event in the United
The Globalization of Financial Turmoil 177

Table 8.1
Stock Market Returns in Domestic Currency: Summary Statistics
Percentiles

Countries Mean 5th 95th

Hong Kong 1.40
3.39 3.17
Indonesia 1.62
3.94 4.51
Korea 1.96
4.43 5.32
Malaysia 1.72
3.63 3.78
Philippines 1.41
3.26 3.46
Singapore 1.30
2.89 3.00
Thailand 1.78
3.49 4.74
Greece 1.59
3.32 3.62
Finland 1.31
2.79 2.82
Holland 1.16
2.42 2.46
Norway 1.00
2.29 2.18
Spain 1.03
2.13 2.29
Sweden 0.98
2.16 2.00
Turkey 2.49
5.75 5.82

Canada 0.71
1.63 1.49

France 0.90
2.02 2.01
Italy 1.26
2.61 2.88
Germany 1.20
2.54 2.27
Japan 0.93
1.99 2.20
UK 0.86
1.90 1.86
USA 0.87
1.80 1.90
Argentina 1.60
4.31 3.41
Brazil 2.13
4.84 4.31
Chile 1.03
2.24 2.42
Colombia 0.83
2.01 2.17
Mexico 1.35
2.78 3.23
Peru 1.01
2.15 2.33
Venezuela 1.55
3.98 3.48
Czech Republic 0.94
2.30 2.10
Estonia 1.84
4.00 4.54
Hungary 1.63
3.48 3.48
Poland 1.35
3.02 3.03
Russia 2.49
5.10 6.48
Slovakia 0.97
2.49 2.38
Ukraine 2.07
5.18 5.42
Notes: The sample extends from January 1, 1997 to August 31, 1999. Mean is the average of one-day
percent returns in absolute values.
178 Graciela L. Kaminsky and Carmen Reinhart

States but it takes a 3.94 percent downfall to qualify as an extreme event in

Indonesia.
Figure 8.4 provides the first glimpse of the bunching of turmoil that we are try-
ing to explain. This figure reports the percentage of countries with simultaneous
extreme changes in stock prices (those movements in the 5th and 95th percentiles)
around the globe or in various regions. The top panel reports the proportion of
countries worldwide simultaneously in the 5th or 95th percentile of the distribu-
tion. The five other panels show the same evidence by region. The globalization
of turmoil is quite evident during the last few days of October 1997, following the
collapse of the Hong Kong stock market. However, the phenomenon was short
lived, underscoring the importance of daily data. The simultaneous turbulences
in the fall of 1998 were more persistent. The events triggering these systemic
crashes originated in Russia, starting on June 1, 1998 with the suspension of trad-
ing in future markets, and were fueled further by the failed auction of Russian
GKO government bills on August 27, 1998. On that day, 74 percent of the stock
markets around the world collapsed. Turbulence persisted until October as the
crisis in Russia was followed by dislocation in G7 financial markets as LTCM spi-
raled downward. The final episode was very short and was linked to the Brazilian
crisis in January 1999. Regional turmoil is far more frequent, with the last half of
1997 plagued by problems in most Asian countries. In the first half of 1998, the
turbulence spread to Latin America, with turmoil in Brazil at its core. Fragility in
mature markets was mostly concentrated in the fall of 1998. In the next section we
use two definitions of globalization to examine the determinants of world and re-
gional instability.

8.3.1 Weak-Form Globalization

This definition focuses on whether turmoil in one country (returns in the 5th or
95th percentile of the distribution) triggers anomalous behavior in other countries.
Anomalous behavior is defined as a change in the distribution of returns. In par-
ticular, we estimate the frequency distribution of returns in country j on days of
turmoil in country i, and the frequency distribution of returns in country j on
days of no turmoil in country i. We compare these two distributions using the
Kolmogorov-Smirnov test of equality of distributions. We classify a country as af-
fected by extreme crashes or rallies in another country when we reject the null hy-
pothesis of equality of the distributions at a five percent significance level or less.
We call this phenomenon weak-form globalization from country i to country j be-
cause it does not impose simultaneous occurrence of returns in the tails for global-
ization to occur.
We first examine the degree of weak-form globalization triggered by turmoil in
three financial centers: Germany, Japan, and the United States. Table 8.2 reports
Figure 8.4
The globalization of turmoil
Note: Numbers in the y-axis represent the percentage of countries experiencing turmoil. Turmoil is
defined as those observations in the 5th and 95th percentiles.
Table 8.2
180

Turmoil in Financial Centers: How Does It Spread?

Empirical Distribution of Stock Market Returns

On all days On days of turmoil in Germany On days of turmoil in Japan On days of turmoil in USA

Percentiles Percentiles Percentiles Percentiles

K&S K&S K&S
Countries Mean 5th 95th Mean 5th 95th p-value Mean 5th 95th p-value Mean 5th 95th p-value

Hong Kong 1.40
3.39 3.17 2.08
4.92 5.07 0.44 2.60
6.61 7.41 ** 2.40
3.92 7.01 **
Indonesia 1.62
3.94 4.51 2.55
5.24 7.41 ** 2.74
4.76 7.30 *** 2.17
4.84 7.12 ***
Korea 1.96
4.43 5.32 2.22
4.97 6.35 0.98 2.81
5.52 7.26 *** 2.41
5.45 6.87 0.76
Malaysia 1.72
3.63 3.78 2.38
4.80 5.89 ** 2.63
5.37 5.71 0.06 2.58
6.15 9.09 0.52
Philippines 1.41
3.26 3.46 2.23
5.54 4.21 *** 2.00
4.95 5.83 ** 1.84
4.86 4.31 0.51
Singapore 1.30
2.89 3.00 1.95
4.66 4.23 0.18 2.27
3.93 7.00 *** 1.85
3.62 3.85 0.23
Thailand 1.78
3.49 4.74 2.09
4.05 6.10 0.69 2.38
3.94 8.16 0.47 1.83
3.54 7.52 0.21
Greece 1.59
3.32 3.62 2.06
6.13 3.60 0.06 2.17
6.12 5.07 0.10 2.39
5.57 4.93 0.06
Finland 1.31
2.79 2.82 3.07
5.67 5.05 *** 2.20
4.85 4.67 *** 2.11
4.79 5.10 **
Holland 1.16
2.42 2.46 2.77
5.21 4.89 *** 1.75
3.85 3.80 ** 1.69
4.54 3.33 **
Norway 1.00
2.29 2.18 2.49
5.39 5.05 *** 1.68
4.95 5.05 ** 1.53
4.64 4.44 **
Spain 1.03
2.13 2.29 2.40
6.03 5.12 *** 1.58
4.79 4.01 ** 1.31
3.69 3.78 0.30
Sweden 0.98
2.16 2.00 2.18
4.34 3.21 *** 1.74
4.07 3.28 *** 1.47
3.71 3.31 **
Turkey 2.49
5.75 5.82 3.74
10.12 9.26 0.06 3.56
10.12 8.78 0.12 3.45
8.85 9.89 **
Canada 0.71
1.63 1.49 1.36
3.72 2.75 *** 0.93
2.04 1.67 0.07 1.83
3.72 2.90 ***
France 0.90
2.02 2.01 2.22
4.29 4.45 *** 1.42
3.94 3.47 *** 1.23
3.34 3.11 0.15
Italy 1.26
2.61 2.88 2.59
5.63 5.29 *** 2.02
5.23 4.48 ** 1.61
4.33 4.96 **
Germany 1.20
2.54 2.27 3.57
5.87 5.17 *** 1.88
4.67 4.21 ** 1.90
5.35 3.94 ***
Japan 0.93
1.99 2.20 1.20
3.07 2.84 ** 2.95
4.15 4.37 *** 1.19
2.79 2.77 0.62
UK 0.86
1.90 1.86 1.79
3.25 3.32 *** 1.23
2.96 2.67 ** 1.19
3.06 3.00 **
USA 0.87
1.80 1.90 1.39
3.42 3.10 0.18 0.97
2.06 2.12 0.28 2.60
3.73 4.02 ***
Graciela L. Kaminsky and Carmen Reinhart
Argentina 1.60
4.31 3.41 2.98
10.44 5.76 0.06 2.14
6.40 4.94 0.52 3.51
9.17 8.13 ***
Brazil 2.13
4.84 4.31 3.73
9.77 8.07 ** 2.73
9.71 4.72 0.32 4.76
10.08 10.34 ***
Chile 1.03
2.24 2.42 1.71
5.04 3.23 ** 1.05
2.79 2.39 0.43 1.78
4.69 3.55 **
Colombia 0.83
2.01 2.17 0.89
2.38 2.30 0.79 1.04
2.28 3.93 0.11 0.88
2.42 1.95 0.88
Mexico 1.35
2.78 3.23 2.67
6.05 5.29 ** 1.98
4.79 4.20 0.09 2.95
6.05 7.17 ***
Peru 1.01
2.15 2.33 1.85
5.75 4.58 *** 1.29
3.02 2.88 0.29 1.66
4.70 3.97 **
Venezuela 1.55
3.98 3.48 2.25
7.51 4.32 0.26 1.72
5.09 2.32 0.06 2.04
6.20 3.77 0.13
Czech 0.94
2.30 2.10 1.62
4.28 2.82 *** 1.18
2.99 2.09 0.14 1.31
3.48 2.33 **
Republic
Estonia 1.84
4.00 4.54 2.80
8.35 8.25 0.16 2.97
11.31 9.09 0.09 2.38
8.35 5.36 0.43
Hungary 1.63
3.48 3.48 3.48
10.03 4.94 *** 3.47
10.76 8.63 ** 3.26
10.76 9.10 ***
Poland 1.35
3.02 3.03 2.34
6.02 5.28 *** 2.42
6.41 4.81 0.09 2.90
6.60 6.53 ***
The Globalization of Financial Turmoil

Russia 2.49
5.10 6.48 4.73
10.80 14.98 *** 3.82
10.11 7.46 0.10 4.19
12.49 14.67 0.07
Slovakia 0.97
2.49 2.38 1.09
3.13 2.35 0.71 0.96
2.56 1.82 0.30 0.83
2.58 1.94 0.37
Ukraine 2.07
5.18 5.42 3.15
12.20 6.06 0.08 3.10
11.07 8.09 0.13 3.02
8.63 7.04 0.26
Notes: Turmoil is defined as those observations in the 5th and 95th percentiles. Mean is the average of one day percent returns in absolute values.
The Kolmogorov-Smirnov test evaluates whether the frequency distribution on days of turmoil in the corresponding financial center is different
from the frequency distribution on all other days. 5th and 95th percentiles report the values of stock market returns at those percentiles. The sam-
ple extends from January 1, 1997 to August 31, 1999. ***, ** represent the significance of the Kolmogorov-Smirnov test at 1 and 5 percent level
respectively.
181
182 Graciela L. Kaminsky and Carmen Reinhart

the spillover of extreme events with a country-by-country detail. To get a high-

resolution picture of anomalous behavior in the stock market, we report the 5th
and 95th percentiles of the distribution of returns for all observations in the sam-
ple and for the observations on days of market turmoil in each of the financial
centers. For example, the 5th and 95th percentile returns for Argentina for the
whole sample are
4.31 and 3.41. When there is turmoil in the United States, the
5th and 95th percentile returns for Argentina become
9.37 and 8.13. As shown in
this table, the Kolmogorov-Smirnov test rejects the null hypothesis of similar dis-
tribution of stock market returns in Argentina on days of financial turmoil in the
United States and the distribution on all other days. Thus, we catalog Argentina
as suffering weak-form globalization from turmoil in the United States. In the
event, extreme movements in equity markets in the United States are transmitted
instantaneously to most Latin American countries—the only exceptions being Co-
lombia and Venezuela. Indeed, 70 percent of the countries in Latin America are,
according to our measure, affected by toil events in the United States. In contrast,
turmoil in the United States triggers an anomalous behavior in only 29 percent of
Asian stock markets. The pattern of the problem-spreading in Japan is in sharp
contrast to that observed for the United States. In this case, Latin American mar-
kets do not react at all to turmoil in Japan, but 71 percent of the Asian countries
experience anomalous returns when Japan posts an outsized return.
Table 8.3 summarizes these results. Overall, shocks in financial centers are
transmitted instantaneously to basically all (70 percent) mature markets (G7 and
European countries), whether the shock occurs in Germany, Japan, or the United
States. These results are suggestive of the higher degree of integration of those
markets. The regional characteristics of the transmission of shocks to emerging

Table 8.3
Weak-Form Globalization of Turmoil: Regional and World Effects
Percentage of Countries with Anomalous Returns when
Turmoil in
Regions Germany Japan USA
Asia 43 71 29
Europe 71 71 71
G7 100 75 75
Latin America 43 0 71
Transition economies 57 14 43
World 59 44 56
Notes: Turmoil is defined as those observations in the 5th and 95th percentiles. An anomalous return is
interpreted as a change in the distribution of returns in country j on days of turmoil in country i.
The Globalization of Financial Turmoil 183

economies are, however, different. U.S. shocks are strongly transmitted across
Latin America; the shocks in Germany simultaneously affect stock markets in
Eastern Europe, Latin America, and Asia; while Japanese turbulences mostly af-
fect other Asian countries. Interestingly, this pattern of transmission matches the
pattern of exposure of financial institutions in Germany, Japan, and the United
States to emerging economies as examined by van Rijckenghem and Weder
(2000). These authors classify bank lending to emerging economies by area of
loan origin. They find that European banks are the largest creditors in all regions,
with North American banks concentrating their lending in Latin America and Jap-
anese banks mostly lending to other Asian countries. In particular, at the onset of
the Asian crisis, 32 percent of all the international loans to Asian countries origi-
nated in Japan, 44 percent originated in Europe, and just 10 percent originated in
North America. Also, during 1997 and 1998, most lending to Eastern European
countries (including Russia) originated in Western Europe (80 percent), while
lending to Latin America originated in Western European banks (60 percent) and
North American banks (30 percent).
Rijckenghem and Weder (2000) also examine the shifts in portfolios of Euro-
pean, North American, and Japanese banks during the Asian and Russian crises.
Japanese banks consistently withdrew from Asia, reducing their lending from
$124 billion in mid-1997 to $86 billion by the end of 1998. North American banks
mainly shifted their lending among emerging markets during the Asian crisis
(from Asia to Latin America and Europe), while they reduced their positions in
all three regions during the Russian crisis. European banks continued to build up
their positions in all regions even after the onset of the Asian crisis. Only during
the first half of 1998 did they reduce their holdings in Asia while increasing them
in Latin America and Eastern Europe. The Russian crisis triggered the end of this
expansionist investment strategy of European banks in emerging markets, with
all banks reducing their exposure to all three regions by about $20 billion.
Table 8.4 examines whether turmoil is transmitted from one country in the pe-
riphery to another country in the periphery or to financial centers. In particular, it
examines the pattern of spillovers on days of turmoil in three crisis-prone coun-
tries in our sample—Brazil, Russia, and Thailand—on a country-by-country basis.
Table 8.5 summarizes the information. The patterns of globalization are similar
for Brazil and Russia. Turbulence in those countries coincides with abnormal
movements around the globe, with the sole exception of Asia. Extreme move-
ments in Thailand are not so far-reaching, in that they spill over only to other
Asian economies. This evidence begs for an answer as to through which channels
these crisis-prone countries with small asset markets have such far-reaching
effects. To answer this question, we examine whether the days of turbulence in a
Table 8.4
184

Turmoil in Emerging Markets: How Does It Spread?

Empirical Distribution of Stock Market Returns

On all days On days of turmoil in Brazil On days of turmoil in Russia On days of turmoil in Thailand

Percentiles Percentiles Percentiles Percentiles

K&S K&S K&S
Countries Mean 5th 95th Mean 5th 95th p-value Mean 5th 95th p-value Mean 5th 95th p-value

Hong Kong 1.40
3.39 3.17 2.41
4.82 7.41 0.32 2.05
3.26 6.84 0.17 2.45
4.52 6.84 ***
Indonesia 1.62
3.94 4.51 2.50
5.15 6.77 ** 2.32
5.15 6.47 0.30 2.89
5.03 8.00 **
Korea 1.96
4.43 5.32 2.60
5.65 6.76 0.37 2.18
4.97 4.91 0.49 2.35
5.66 7.39 0.87
Malaysia 1.72
3.63 3.78 2.52
4.73 9.42 0.68 2.85
4.53 11.80 ** 2.81
5.86 6.91 0.12
Philippines 1.41
3.26 3.46 2.09
5.28 4.80 0.12 2.19
6.01 3.98 *** 2.46
4.84 6.80 **
Singapore 1.30
2.89 3.00 2.13
3.58 5.92 0.09 1.94
3.30 7.00 0.24 2.35
3.84 7.62 **
Thailand 1.78
3.49 4.74 2.18
4.05 7.52 0.31 2.05
3.51 5.96 0.69 5.71
6.33 10.42 ***
Greece 1.59
3.32 3.62 2.20
4.99 5.50 ** 2.23
6.02 3.85 *** 1.89
5.53 4.02 0.26
Finland 1.31
2.79 2.82 2.14
4.85 4.36 *** 2.42
5.59 4.50 *** 1.71
4.81 3.31 0.74
Holland 1.16
2.42 2.46 1.89
5.00 3.38 *** 1.82
5.21 2.68 *** 1.32
3.15 2.50 0.29
Norway 1.00
2.29 2.18 1.99
4.95 4.57 *** 2.11
5.23 3.92 *** 1.61
4.73 3.88 0.08
Spain 1.03
2.13 2.29 1.61
4.74 3.78 ** 1.67
6.03 3.06 *** 1.16
3.09 2.54 0.58
Sweden 0.98
2.16 2.00 1.69
3.61 3.17 *** 1.61
4.07 2.09 *** 1.22
2.79 2.09 0.31
Turkey 2.49
5.75 5.82 3.73
9.25 8.78 *** 4.29
10.99 9.86 *** 3.12
8.23 7.98 0.65
Canada 0.71
1.63 1.49 1.32
3.32 2.29 *** 1.19
3.18 2.46 ** 0.94
1.90 2.13 **
France 0.90
2.02 2.01 1.38
3.34 2.94 *** 1.51
4.29 2.55 *** 0.99
3.08 2.20 0.69
Italy 1.26
2.61 2.88 2.02
4.33 4.96 ** 2.08
5.63 3.22 *** 1.39
3.91 2.86 0.33
Germany 1.20
2.54 2.27 2.01
5.54 3.94 ** 2.05
5.68 3.51 *** 1.58
4.32 3.35 0.38
Japan 0.93
1.99 2.20 1.37
3.00 3.76 0.24 1.18
2.55 3.03 0.61 1.25
3.22 3.56 0.25
UK 0.86
1.90 1.86 1.29
3.13 3.00 *** 1.52
3.14 2.69 *** 0.99
2.03 2.89 0.11
USA 0.87
1.80 1.90 1.72
3.32 3.70 *** 1.31
2.80 4.02 0.23 1.01
2.13 2.18 0.13
Graciela L. Kaminsky and Carmen Reinhart
Argentina 1.60
4.31 3.41 4.58
10.44 8.51 *** 3.32
10.44 7.41 ** 2.23
4.15 6.09 0.12
Brazil 2.13
4.84 4.31 7.67
10.09 12.19 *** 3.89
10.08 7.66 ** 2.56
5.19 6.72 0.20
Chile 1.03
2.24 2.42 2.25
5.04 4.34 *** 1.58
5.04 2.53 0.16 1.28
2.86 3.36 0.09
Colombia 0.83
2.01 2.17 1.10
3.74 2.02 ** 0.91
3.11 1.88 ** 0.88
2.08 2.47 0.40
Mexico 1.35
2.78 3.23 3.21
6.05 8.43 *** 2.58
5.57 6.40 ** 1.80
3.76 4.59 0.26
Peru 1.01
2.15 2.33 1.93
5.64 3.84 *** 1.63
5.11 3.78 ** 1.21
2.47 3.51 **
Venezuela 1.55
3.98 3.48 2.63
7.51 3.85 *** 2.03
6.65 3.08 ** 1.85
4.55 3.67 0.32
Czech 0.94
2.30 2.10 1.61
3.70 2.81 *** 1.83
3.93 3.13 *** 0.98
2.21 2.06 0.96
Republic
Estonia 1.84
4.00 4.54 3.04
10.49 7.01 *** 3.54
10.49 8.64 *** 2.37
6.98 5.74 0.37
Hungary 1.63
3.48 3.48 3.65
10.51 9.32 *** 3.61
10.76 5.91 *** 2.18
6.02 3.47 0.48
Poland 1.35
3.02 3.03 2.61
5.74 6.00 *** 2.45
6.27 4.66 *** 1.98
5.00 3.75 0.06
The Globalization of Financial Turmoil

Russia 2.49
5.10 6.48 4.75
12.49 13.85 ** 9.74
17.49 16.71 *** 3.46
7.35 8.56 0.31
Slovakia 0.97
2.49 2.38 0.97
3.04 2.68 0.94 1.19
3.10 2.52 ** 1.14
3.07 2.90 0.47
Ukraine 2.07
5.18 5.42 4.37
10.94 8.99 *** 3.27
11.56 8.37 0.16 2.37
6.24 8.56 0.96
Notes: Turmoil is defined as those observations in the 5th and 95th percentiles. Mean is the average of one day percent returns in absolute values.
The Kolmogorov-Smirnov test evaluates whether the frequency distribution on days of turmoil in the corresponding emerging market is different
from the frequency distribution on all other days. 5th and 95th percentiles report the values of stock market returns at those percentiles. The sam-
ple extends from January 1, 1997 to August 31, 1999. ***, ** represent the significance of the Kolmogorov-Smirnov test at 1 and 5 percent level
respectively.
185
186 Graciela L. Kaminsky and Carmen Reinhart

Table 8.5
Weak-Form Globalization of Turmoil: Regional and World Effects
Percentage of Countries with Anomalous Returns when
Turmoil in
Regions Brazil Russia Thailand
Asia 14 29 67
Europe 100 100 0
G7 83 67 17
Latin America 100 86 14
Transition economies 86 83 0
World 76 73 18
Notes: A turmoil is defined as those observations in the 5th and 95th percentiles. An anomalous return
is interpreted as a change in the distribution of returns in country j on days of turmoil in country i.

particular crisis-prone emerging market were also days of turbulence in a finan-

cial center with which that particular country was associated. We then examine
whether problems in crisis-prone emerging markets not associated with turmoil
in financial centers also have wide spillover effects. We chose financial centers
according to the evidence discussed in the literature. We pair Brazil with the
United States, Russia with Germany, and Thailand with Japan.
Table 8.6 examines the periphery-to-periphery and periphery-center-periphery
connections. Days of turmoil in crisis-prone emerging economies are divided into
two samples, those on which the corresponding financial center was also roiled,
and those on which the corresponding financial center was not. The results are
dramatically different. Turbulence in Brazil accompanied by turbulence in the
United States is transmitted around the world, with Asia the only untouched re-
gion. In contrast, turbulence in Brazil unaccompanied by turbulence in the United
States only affects stock markets in Latin America. Turmoil in this case only has a
regional reach. The same picture of propagation of shocks is observed in the case
of Russian jitters. Turbulence becomes global if a financial center is affected, but
remains regional when the stock market in the financial center is calm. The evi-
dence from Thailand is somewhat different. Again, simultaneous turmoil in the fi-
nancial center ( Japan) and in Thailand triggers a broader propagation of shocks.
But here this propagation is only regional in nature. There is not even regional
propagation when turbulence affects only the stock market in Thailand. The re-
gional characteristics of some the turbulences in stock markets agree with the evi-
dence from currency crises.6
The question is, what causes this regional pattern of spillovers? Strong bilateral
trade patterns may provide one explanation. For example, Kaminsky and Rein-
The Globalization of Financial Turmoil 187

hart (2000) point to the strong bilateral trade among Mercosur countries, but cau-
tion that turmoil in Brazil is still transmitted rapidly to non-Mercosur Latin Amer-
ican countries. Similarly, shocks from Russia are strongly transmitted to most of
the transition economies even though bilateral trade links among transition
economies diminished drastically in importance following the collapse of the
communist regimes in Eastern Europe in 1989–1991. Third-party trade links may
provide another explanation. For example, Malaysia and Thailand sell similar
goods to Japan and the United States, explaining the contagion from Thailand to
Malaysia following the Thai devaluation in July 1997. But the Mexican crisis in
1994 strongly affected Argentina and Brazil and these countries do not compete
with Mexico in third markets. Again, financial links may help to explain regional
contagion. For example, Kaminsky, Lyons, and Schmukler (2004) examine invest-
ment strategies of U.S.-based mutual funds specializing in Latin America and find
that they were a key element in explaining the reach of the Tequila crisis: as
investors stampeded out of mutual funds specializing in Latin America following
the Mexican devaluation, managers (under the pressure of the massive redemp-
tions) had to sell not just Mexican stocks, but also stocks from Argentina and
Brazil.
Table 8.7 summarizes these results by region and examines the null hypothesis
of financial center irrelevance versus the alternative hypothesis that a financial
center has to be affected for the turmoil to become systemic. To examine this hy-
pothesis, we construct the Wilcoxon, or rank-sum, test. To construct this test, we
look at the results from the Kolmogorov-Smirnov test and construct two samples.
The first sample captures the weak-form globalization pattern following turmoil
in a crisis-prone emerging market coinciding with turmoil in a financial center.
For each j country in the sample, we assign a value equal to 1 if turmoil in the
pair of crisis-prone emerging market and financial center triggers anomalous be-
havior in country j, and 0 otherwise. The second sample captures the weak-form
globalization pattern following turmoil in a crisis-prone emerging market not
coinciding with turmoil in a financial center. Again, for each j country in the sam-
ple, we assign a value equal to 1 if turmoil in just the crisis-prone emerging
market triggers anomalous behavior in country j, and 0 otherwise. Denote the
observations from the first sample by fXg and the observations from the second
sample by fYg. The null hypothesis of financial-center irrelevance implies that
PðX > kÞ ¼ PðY > kÞ for all k. We are interested in the one-sided alternative that
X is stochastically larger than Y; that is, PðX > kÞ > ðY > kÞ for all k. To construct
the rank-sum test, we rank all the observations without regard to the sample to
which they belong. Then the Wilcoxon test statistic is formed as the sum of the
ranks in the first sample:
Table 8.6
Financial Turmoil in Emerging Markets and Financial Centers: How Does It Spread?
Empirical Distribution of Stock Market Returns

On all days On days of turmoil in Brazil

With financial center Without financial center

Percentiles Percentiles Percentiles

K&S K&S
Countries Mean 5th 95th Mean 5th 95th p-value Mean 5th 95th p-value

Hong Kong 1.40
3.39 3.17 2.41
4.82 7.41 0.32 1.67
4.38 5.53 0.74
Indonesia 1.62
3.94 4.51 2.50
5.15 6.77 ** 2.00
4.02 6.26 0.34
Korea 1.96
4.43 5.32 2.60
5.65 6.76 0.37 2.23
5.65 6.76 0.82
Malaysia 1.72
3.63 3.78 2.52
4.73 9.42 0.68 2.49
3.75 14.94 0.92
Philippines 1.41
3.26 3.46 2.09
5.28 4.80 0.12 1.68
4.60 3.11 0.12
Singapore 1.30
2.89 3.00 2.13
3.58 5.92 0.09 1.73
3.21 4.92 0.36
Thailand 1.78
3.49 4.74 2.18
4.05 7.52 0.31 1.97
3.95 6.36 0.77

Greece 1.59
3.32 3.62 2.20
4.99 5.50 ** 1.64
4.00 3.47 0.13
Finland 1.31
2.79 2.82 2.14
4.85 4.36 *** 1.59
4.49 2.83 **
Holland 1.16
2.42 2.46 1.89
5.00 3.38 *** 1.57
2.78 2.92 0.10
Norway 1.00
2.29 2.18 1.99
4.95 4.57 *** 1.66
4.04 3.47 **
Spain 1.03
2.13 2.29 1.61
4.74 3.78 ** 1.43
4.83 2.49 0.15
Sweden 0.98
2.16 2.00 1.69
3.61 3.17 *** 1.41
3.11 2.58 0.16
Turkey 2.49
5.75 5.82 3.73
9.25 8.78 *** 2.88
8.04 7.09 0.26

Canada 0.71
1.63 1.49 1.32
3.32 2.29 *** 0.71
1.81 1.51 0.98
France 0.90
2.02 2.01 1.38
3.34 2.94 *** 1.18
2.95 2.20 0.19
Italy 1.26
2.61 2.88 2.02
4.33 4.96 ** 1.83
4.19 3.38 0.11
Germany 1.20
2.54 2.27 2.01
5.54 3.94 ** 1.52
3.55 2.15 0.41
Japan 0.93
1.99 2.20 1.37
3.00 3.76 0.24 1.36
2.41 3.87 0.20
UK 0.86
1.90 1.86 1.29
3.13 3.00 *** 1.12
2.13 2.59 0.09
USA 0.87
1.80 1.90 1.72
3.32 3.70 *** 0.93
1.77 1.68 0.19

Argentina 1.60
4.31 3.41 4.58
10.44 8.51 *** 3.48
8.82 6.55 ***
Brazil 2.13
4.84 4.31 7.67
10.09 12.19 *** 6.68
9.69 8.81 ***
Chile 1.03
2.24 2.42 2.25
5.04 4.34 *** 1.95
3.87 4.08 ***
Colombia 0.83
2.01 2.17 1.10
3.74 2.02 ** 1.14
3.97 2.23 **
Mexico 1.35
2.78 3.23 3.21
6.05 8.43 *** 2.19
3.45 5.12 0.12
Peru 1.01
2.15 2.33 1.93
5.64 3.84 *** 1.44
5.34 3.37 **
Venezuela 1.55
3.98 3.48 2.63
7.51 3.85 *** 2.33
7.19 3.93 **

Czech Republic 0.94
2.30 2.10 1.61
3.70 2.81 *** 1.44
3.71 2.85 **
Estonia 1.84
4.00 4.54 3.04
10.49 7.01 *** 2.78
8.05 8.57 0.22
Hungary 1.63
3.48 3.48 3.65
10.51 9.32 *** 2.46
6.35 7.99 0.29
Poland 1.35
3.02 3.03 2.61
5.74 6.00 *** 1.84
3.79 4.36 0.26
Russia 2.49
5.10 6.48 4.75
12.49 13.85 ** 3.58
6.54 8.44 0.13
Slovakia 0.97
2.49 2.38 0.97
3.04 2.68 0.94 0.91
3.07 2.93 0.94
Ukraine 2.07
5.18 5.42 4.37
10.94 8.99 *** 3.72
10.94 8.99 0.08

Notes: Turmoil is defined as those observations in the 5th and 95th percentiles. Mean is the average of one-day percent returns in abso-
lute values. The Kolmogorov-Smirnov test evaluates whether the frequency distribution on days of turmoil in the corresponding emerg-
ing market (with or without turmoil in a financial center) is different from the frequency distribution on all other days. 5th and 95th
percentiles report the values of stock market returns at those percentiles. The sample extends from January 1, 1997 to August 31, 1999.
***, ** represent the significance of the Kolgomorov-Smirnov test at 1 and 5 percent respectively.
Empirical Distribution of Stock Market Returns

On days of turmoil in Russia On days of turmoil in Thailand

With financial center Without financial center With financial center Without financial center

Percentiles Percentiles Percentiles Percentiles

K&S K&S K&S K&S
Mean 5th 95th p-value Mean 5th 95th p-value Mean 5th 95th p-value Mean 5th 95th p-value

2.05
3.26 6.84 0.17 1.78
3.10 6.95 0.11 2.45
4.52 6.84 *** 1.96
4.51 5.55 0.13
2.32
5.15 6.47 0.30 1.79
3.44 5.83 0.56 2.89
5.03 8.00 ** 2.13
4.28 5.75 0.72
2.18
4.97 4.91 0.49 2.14
3.94 5.08 0.13 2.35
5.66 7.39 0.87 2.00
5.66 7.39 0.69
2.85
4.53 11.80 ** 2.74
4.68 9.98 ** 2.81
5.86 6.91 0.12 2.15
4.05 4.36 0.48
2.19
6.01 3.98 *** 1.79
4.49 3.95 0.58 2.46
4.84 6.80 ** 1.99
4.43 5.97 0.34
1.94
3.30 7.00 0.24 1.62
3.30 6.22 0.15 2.35
3.84 7.62 ** 1.80
3.82 5.56 0.36
2.05
3.51 5.96 0.69 2.29
3.47 7.44 0.77 5.71
6.33 10.42 *** 5.59
6.76 10.74 ***

2.23
6.02 3.85 *** 1.71
3.86 3.95 0.70 1.89
5.53 4.02 0.26 1.76
5.84 3.67 0.31
2.42
5.59 4.50 *** 1.51
3.17 3.95 0.11 1.71
4.81 3.31 0.74 1.38
2.66 3.06 0.47
1.82
5.21 2.68 *** 1.14
2.53 1.86 0.27 1.32
3.15 2.50 0.29 1.24
1.98 2.93 0.49
2.11
5.23 3.92 *** 1.49
2.96 3.89 0.07 1.61
4.73 3.88 0.08 1.31
2.77 3.51 0.19
1.67
6.03 3.06 *** 0.86
1.78 1.96 0.68 1.16
3.09 2.54 0.58 0.99
2.30 2.39 0.20
1.61
4.07 2.09 *** 1.10
2.77 2.06 0.12 1.22
2.79 2.09 0.31 0.98
2.36 2.02 0.34
4.29
10.99 9.86 *** 3.56
6.66 9.87 0.18 3.12
8.23 7.98 0.65 3.13
8.22 7.96 0.42

1.19
3.18 2.46 ** 0.88
2.20 2.03 0.12 0.94
1.90 2.13 ** 0.92
1.92 2.21 0.08
1.51
4.29 2.55 *** 0.84
2.08 1.30 0.39 0.99
3.08 2.20 0.69 0.90
2.03 2.34 0.99
2.08
5.63 3.22 *** 1.46
2.90 3.05 0.23 1.39
3.91 2.86 0.33 1.16
2.33 2.78 0.10
2.05
5.68 3.51 *** 1.09
2.31 2.10 0.31 1.58
4.32 3.35 0.38 1.44
2.58 3.62 0.82
1.18
2.55 3.03 0.61 1.05
1.93 3.06 0.59 1.25
3.22 3.56 0.25 0.69
1.50 1.87 **
1.52
3.14 2.69 *** 1.15
2.11 2.35 0.10 0.99
2.03 2.89 0.11 0.89
1.65 2.73 0.40
1.31
2.80 4.02 0.23 1.04
1.83 3.49 0.79 1.01
2.13 2.18 0.13 0.99
2.11 2.16 0.16

3.32
10.44 7.41 ** 2.63
4.85 8.49 0.13 2.23
4.15 6.09 0.12 2.18
4.74 6.53 0.33
3.89
10.08 7.66 ** 3.24
8.46 10.79 0.51 2.56
5.19 6.72 0.20 2.57
5.12 7.20 0.24
1.58
5.04 2.53 0.16 1.27
3.07 3.73 0.63 1.28
2.86 3.36 0.09 1.31
2.99 3.53 0.28
0.91
3.11 1.88 ** 0.85
3.79 1.94 0.18 0.88
2.08 2.47 0.40 0.91
1.98 2.51 0.37
2.58
5.57 6.40 ** 1.91
3.75 5.89 0.50 1.80
3.76 4.59 0.26 1.56
3.66 3.62 0.61
1.63
5.11 3.78 ** 1.25
2.83 3.84 0.47 1.21
2.47 3.51 ** 1.16
2.29 2.65 **
2.03
6.65 3.08 ** 1.50
4.19 3.07 0.24 1.85
4.55 3.67 0.32 1.95
4.17 4.12 0.12

1.83
3.93 3.13 *** 1.43
2.33 3.07 0.06 0.98
2.21 2.06 0.96 0.99
2.18 2.35 0.99
3.54
10.49 8.64 *** 3.42
9.09 9.31 *** 2.37
6.98 5.74 0.37 1.76
6.16 4.63 0.76
3.61
10.76 5.91 *** 2.42
7.22 4.22 0.14 2.18
6.02 3.47 0.48 1.69
3.49 3.12 0.72
2.45
6.27 4.66 *** 1.86
4.38 3.86 ** 1.98
5.00 3.75 0.06 1.51
4.28 2.83 0.20
9.74
17.49 16.71 *** 9.37
17.85 15.71 *** 3.46
7.35 8.56 0.31 3.08
7.29 8.32 0.18
1.19
3.10 2.52 ** 1.11
3.06 2.79 0.40 1.14
3.07 2.90 0.47 1.31
4.01 3.19 0.23
3.27
11.56 8.37 0.16 3.04
7.04 8.83 0.19 2.37
6.24 8.56 0.96 1.65
3.68 8.28 0.99
Table 8.7
190

Weak-Form Globalization of Turmoil: Regional and World Effects

Percentage of countries with anomalous returns when turmoil in

Brazil Russia Thailand
With Without With Without With Without
financial financial W financial financial W financial financial W
Regions center center p-value center center p-value center center p-value

Asia 14 0 29 14 67 0
Europe 100 29 100 0 0 0
G7 83 0 67 0 17 14
Latin America 100 83 86 0 14 14
Transition economies 86 14 83 33 0 0
World 76 24 73 12 18 6
Wilcoxon statistic (W) 1320 891 0.00 1452 759 0.00 1188 1023 0.00
Notes: The financial center is respectively USA for Brazil, Germany for Russia, and Japan for Thailand. Turmoil is defined as those observations in
the 5th and 95th percentiles. An anomalous return is interpreted as a change in the distribution of returns in country j on days of turmoil in coun-
try i.
Graciela L. Kaminsky and Carmen Reinhart
The Globalization of Financial Turmoil 191

X
n
W¼ R1i ð8:1Þ
i¼1

where n is the number of countries in each sample. Under the null hypothesis, the
average rank of an observation in sample 1 should equal the average rank of an
observation in sample 2. Using Fisher’s principle of randomization, it is straight-
forward to verify that

nð2n þ 1Þ ns 2
EðWÞ ¼ and VarðWÞ ¼ ð8:2Þ
2 2

where s is the standard deviation of the combined ranks ri for both samples:

1 X 2n
s2 ¼ ðri
rÞ 2 : ð8:3Þ
2n
1 i¼1

The last row of table 8.7 shows the Wilcoxon test statistic for each sample and the
one-sided p value for the null hypothesis of financial-center irrelevance. For exam-
ple, for the case of Brazil, the proportion of all countries affected when both Brazil
and the United States experience turmoil is 76 percent, and the proportion of
countries affected when just Brazil experiences turmoil is 24 percent. For these
two samples, the Wilcoxon p-value under the null hypothesis of financial-center
irrelevance is less than 0.01, leading us to reject the null hypothesis of financial-
center irrelevance. The results for the other two emerging markets are similar. In
all cases, the tests reject the null hypothesis of financial-center irrelevance in favor
of the alternative hypothesis that a financial center has to be affected for turmoil to
become systemic.

8.3.2 Strong-Form Globalization

In the previous section we examined whether turmoil in one country triggers

anomalous behavior in stock markets around the world, with anomalous behav-
ior defined as a change in the distribution of returns. Under this definition of
globalization, other stock markets do not have to experience extreme returns in re-
sponse to extreme returns in one stock market for globalization to occur. A more
stringent concept of the globalization of turmoil would be one of simultaneous
turmoil. We call this definition of the globalization of turmoil ‘‘strong-form global-
ization.’’ A globalization index in this case will just be the proportion of countries
with simultaneous extreme events. Our task in this section is to examine the deter-
minants of this index.
192 Graciela L. Kaminsky and Carmen Reinhart

To examine the causes of systemic events, we use a multinomial logit approach.

We also estimated the model using order logit techniques. The results are quite
similar, so we do not report them to save space. Since we are interested in explain-
ing the degree of globalization, our left-hand variable will be a dummy variable
that can take three values: low, medium, and high globalization.7 Low globaliza-
tion occurs when less than 25 percent of the countries in the sample experience
turmoil; medium globalization occurs when there are more than 25 percent, but
less than 50 percent, of the countries in turmoil. Finally, high globalization occurs
when 50 percent or more of all the countries experience turmoil. Our explanatory
variables are dummy variables capturing days of turmoil in financial centers,
days of turmoil in crisis-prone countries on days of turmoil in financial centers,
and days of turmoil in crisis-prone countries when financial centers are not af-
fected. These dummies will take a value of 1 on days of turmoil and 0 otherwise.
Equation 8.4 is the multinomial logit equation to be estimated.
, j
1
!
X
0 0
Pðy ¼ iÞ ¼ expðx bi Þ 1þ x bi ð8:4Þ
i¼1

The variable y is the globalization index, and the vector x includes the dummy
variables capturing turmoil in the various countries. The variable Pðy ¼ iÞ is the
probability associated with outcome i. The index j refers to the number of out-
comes in our estimation: low, medium, and high globalization. The vector b is the
vector of coefficients to be estimated. As is usual in this type of estimations, for
each explanatory variable we estimate j
1 parameters. The probability that there
is low globalization is our base case and it is equal to
, j
1
!
X
0
Pðy ¼ lowÞ ¼ 1 1þ expðx b i Þ : ð8:5Þ
i¼1

The estimation of equation 8.4 is somewhat problematic because not all the mar-
kets are open at the same time. Thus, a shock leading to turmoil in Brazil can af-
fect all Latin American economies the same day, European economies the same
day or the following day depending on the time at which the shock occurs, and
Asian countries only on the following day. Similarly, if a shock occurs in Russia,
the index of globalization on the left-hand side has to include countries in turmoil
in Europe, the G7, and Latin America on the same day and countries in turmoil in
Asia the next day, but if the turmoil originates in Thailand, the index of globaliza-
tion on the left-hand side has to include the number of countries in turmoil in all
the regions the same day of the shock.
The Globalization of Financial Turmoil 193

We deal with this problem in two different ways. First, we estimate equation 8.4
using only turmoil originating in shocks from one time zone at a time. In this case,
the left-hand-side variable is constructed depending on the origin of the shock,
and we estimate three separate versions of equation 8.4 for financial centers and
three separate versions of equation 8.4 for crisis-prone emerging markets. The
shortcoming of this type of estimation is that we cannot evaluate jointly the effects
of extreme events in the various crisis-prone countries and financial centers.
Second, to account for the effect of turbulence in the three crisis-prone countries
jointly, we perform panel estimations. To deal with the different time zones, the
index of globalization on the left-hand side accounts for low, medium, and high
globalization by region. For each region, we align the explanatory variables on
the right-hand side according to the region they may affect. Since we estimate the
regression for all the regions at the same time, the parameters b provide a some-
what different measure of the effects of turmoil in the various countries on global-
ization. For example, the episodes of high globalization are more confining in the
sense that they require all the regions to have a high degree of globalization
simultaneously. This was not the case in the nonpanel estimation.
Finally, within the panel regression estimates, we jointly evaluate the effects of
coincidence of multiple shocks in emerging markets and financial centers. We
construct two dummy variables. The first one captures days of turmoil in emerg-
ing markets coinciding with turmoil in financial centers. This variable can take
four values, 0 to 3. If this variable takes the value 3, it means that the three crisis-
prone emerging economies experience turmoil and so do their respective financial
centers. The second explanatory variable in this regression will capture the num-
ber of crisis-prone emerging markets in turmoil when there is no turmoil in finan-
cial centers. This variable also takes four values, 0 to 3.
Tables 8.8 and 8.9 examine the effects of turmoil originating in one time zone at
a time. Table 8.8 concentrates on turmoil originating in financial centers. The first
equation has as its explanatory variable a dummy variable equal to 1 when Ger-
many experiences turmoil, and 0 otherwise. The second regression has as its ex-
planatory variable a dummy variable equal to 1 when Japan experiences turmoil,
and 0 otherwise. Finally, the third equation has as its explanatory dummy vari-
able a dummy variable equal to 1 when United States experiences turmoil, and
0 otherwise. Table 8.9 uses the same methodology to evaluate the degree of glob-
alization following jitters in one turmoil cluster at a time: Brazil-U.S., Russia-
Germany, and Thailand-Japan. For each turmoil cluster, the regression has two
explanatory dummy variables. One dummy variable is equal to 1 on days of
turbulences in the emerging market coinciding with days of turmoil in the corre-
sponding financial center, and 0 otherwise. The second explanatory dummy vari-
able is equal to 1 on days of turbulences in the emerging market not accompanied
194 Graciela L. Kaminsky and Carmen Reinhart

Table 8.8 P2
Strong-Form Globalization: Multinomial Logit Estimation Pð y ¼ iÞ ¼ expðx 0 bi Þ=ð1 þ i¼1 expðx 0 bi ÞÞ
Coefficients

Turmoil in Turmoil Turmoil

Degree of globalization Germany in Japan in USA
Medium 2.51*** 1.78*** 1.25***
(7.88) (5.49) (3.74)
High 4.71*** 2.56*** 2.45***
(7.14) (5.72) (6.85)
Pseudo R2 0.19 0.08 0.10
Number of observations 694 694 694

Probabilities conditional on
Turmoil in Turmoil Turmoil
Degree of globalization Germany in Japan in USA

Low 40 58 52
Medium 36 26 22
High 23 16 26

Notes: Numbers in parentheses represent z statistics. ***, **, * represents the significance of the coeffi-
cient at the 1, 5, and 10 percent levels. Turmoil is defined as those observations in the 5th and 95th per-
centiles. The left-hand-side variable captures the degree of globalization. There are three possible
degrees of globalization: low (when less than 25 percent of the countries in the sample experience tur-
moil), medium (when more than 25 percent but less than 50 percent of the countries experience tur-
moil), high (when 50 percent or more of all countries in the sample experience turmoil). In order to be
able to estimate our model, coefficients for the low globalization are set equal to zero (that is our base
case). Interpretation of the reported coefficients has to be done with respect to the base case. Our model
was estimated with a constant but constant coefficients are not reported here for expositional purposes.
Probabilities are given in percent terms and are derived from the multinomial logit estimation
shown in the top panel.

by turmoil in the corresponding financial center, and 0 otherwise. To evaluate

jointly the contribution of these clusters to the globalization of turmoil, we esti-
mate a multivariate turmoil-cluster panel regression. We estimate the model using
panel data because of the time-zone problem. The results are reported in table
8.10. Finally, table 8.11 reports the panel estimation evaluating the effects of multi-
ple coincidence of turmoil in the three crisis-prone emerging markets. The top
panels in all these tables report the estimated coefficients, while the bottom panels
show the conditional probabilities of globalization obtained from the estimations
shown in the top panels.
As for the results, table 8.8 shows that turmoil in financial centers triggers tur-
bulences around the world, with the explanatory power (as captured by the pseu-
do R 2 ) ranging from 8 percent for turmoil originating in Japan to 19 percent for
turmoil originating in Germany. Again, this pattern could be explained, in part,
Table 8.9 P2
Strong-Form Globalization Multinomial Logit Estimation Pð y ¼ iÞ ¼ expðx 0 b i Þ=ð1 þ i¼1 expðx 0 b i ÞÞ
Coefficients

Turmoil in Brazil Turmoil in Russia Turmoil in Thailand

With Without With Without With Without
Degree of financial financial financial financial financial financial
globalization center center p-values center center p-values center center p-values
Medium 2.14*** 1.41*** 0.28 2.70*** 1.36*** * 2.32*** 0.91** **
(3.62) (3.61) (4.33) (3.61) (4.03) (2.3)
High 4.64*** 2.47*** *** 5.53*** 1.18 *** 2.18*** 0.37 *
(8.44) (4.85) (7.87) (1.07) (2.6) (0.63)
Pseudo R2 0.13 Pseudo R2 0.15 Pseudo R2 0.03
The Globalization of Financial Turmoil

Probabilities conditional on
Turmoil in Brazil Turmoil in Russia Turmoil in Thailand
With Without With Without With Without
Degree of financial financial No financial financial No financial financial No
globalization center center turmoil center center turmoil center center Turmoil
Low 21 56 87 23 75 92 47 80 90
Medium 21 27 10 27 23 7 40 17 8
High 57 17 2 50 2 1 13 4 3

Notes: Numbers in parentheses represent z statistics. ***, **, * represents the significance of the coefficient at the 1, 5, and 10 percent levels. Turmoil
is defined as those observations in the 5th and 95th percentiles. The left-hand-side variable captures the degree of globalization. There are three pos-
sible degrees of globalization: low (when less than 25 percent of the countries in the sample experience turmoil), medium (when more than 25 per-
cent but less than 50 percent of the countries experience turmoil), high (when 50 percent or more of all countries in the sample experience turmoil).
In order to be able to estimate our model, coefficients for the low globalization had to equal zero (that is our base case). Interpretation of the
reported coefficients has to be done with respect to the base case. Our model was estimated with a constant but constant coefficients are not
reported here for expositional purposes. P column reports p-values for test of equality between parameters estimated with and without turmoil in
financial centers. The financial center is, respectively, USA for Brazil, Germany for Russia, and Japan for Thailand.
Number of observations for our sample was 694.
195

Probabilities are given in percent terms and are derived from the multinomial logit estimation shown in the top panel.
Table 8.10
196

P2
Strong-Form Globalization: Multinomial Logit Panel Estimation Pð y ¼ iÞ ¼ expðx 0 bi Þ=ð1 þ i¼1 expðx 0 b i ÞÞ

Coefficients
Brazil Russia Thailand
With Without With Without With Without
Degree of financial financial financial financial financial financial
globalization center center p-values center center p-values center center p-values

Medium 0.93***
0.58*** *** 0.35 0.17 0.54 0.63** 0.05 *

(4.23) (
2.12) (1.31) (1.07) (2.34) (0.31)
High 2.67*** 0.85** *** 2.71*** 0.37 *** 2.14*** 0.41 ***
(9.22) (2.11) (8.49) (0.99) (4.68) (1.15)
Pseudo R2: 0.07
Number of observations: 3469
Probabilities conditional on
Turmoil in Brazil Turmoil in Russia Turmoil in Thailand
With Without With Without With Without
Degree of No financial financial financial financial financial financial
globalization Turmoil center center center center center center

Low 77 51 84 61 70 60 76
Medium 21 36 13 24 28 31 22
High 1 13 3 15 2 9 2

Notes: Numbers in parentheses represent z statistics. ***, **, * represents the significance of the coefficient at the 1, 5, and 10 percent levels. Turmoil
is defined as those observations in the 5th and 95th percentiles. The left-hand-side variable captures the degree of globalization. There are three pos-
sible degrees of globalization: low (when less than 25 percent of the countries in the sample experience turmoil), medium (when more than 25 per-
cent but less than 50 percent of the countries experience turmoil), high (when 50 percent or more of all countries in the sample experience turmoil).
In order to be able to estimate our model, coefficients for the low globalization had to equal zero (that is our base case). Interpretation of the
reported coefficients has to be done with respect to the base case. Our model was estimated with a constant but constant coefficients are not
reported here for expositional purposes. P column reports p-values for test of equality between parameters estimated with and without turmoil in
financial centers. The financial center is, respectively, USA for Brazil, Germany for Russia, and Japan for Thailand.
Graciela L. Kaminsky and Carmen Reinhart

Probabilities are given in percent terms and are derived from the multinomial logit estimation shown in the top panel.
Table 8.11 P2
Strong-Form Globalization: Multinomial Logit Panel Estimation Pð y ¼ iÞ ¼ expðx 0 bi Þ=ð1 þ i¼1 expðx 0 b i ÞÞ
Coefficients
Emerging markets
Degree of With financial Without financial
globalization center center p-value
Medium 0.65*** 0.01 ***
(4.76) (0.11)
High 2.64*** 0.57*** ***
(14.64) (2.71)
Pseudo R2 0.07
#Observations 3469
The Globalization of Financial Turmoil

Probabilities conditional on
Turmoil in one emerging Turmoil in two emerging Turmoil in three emerging
market markets markets
With Without With Without With Without
Degree of No financial financial financial financial financial financial
globalization turmoil center center center center center center
Low 77 56 76 19 75 2 73
Medium 21 31 22 20 22 4 21
High 1 13 2 62 4 94 6
Notes: Numbers in parentheses represent z statistics. ***, **, * represents the significance of the coefficient at the 1, 5, and 10 percent levels. Turmoil
is defined as those observations in the 5th and 95th percentiles. The left-hand-side variable captures the degree of globalization. There are three pos-
sible degrees of globalization: low (when less than 25 percent of the countries in the sample experience turmoil), medium (when more than 25 per-
cent but less than 50 percent of the countries experience turmoil), high (when 50 percent or more of all countries in the sample experience turmoil).
In order to be able to estimate our model, coefficients for the low globalization had to equal zero (that is our base case). Interpretation of the
reported coefficients has to be done with respect to the base case. Our model was estimated with a constant but constant coefficients are not
reported here for expositional purposes. P column reports p-values for test of equality between parameters estimated with and without turmoil in
financial centers. The explanatory variable is emerging market. Such variable could equal 0, 1, 2, 3 depending on how many emerging markets (Bra-
zil, Russia, Thailand) experienced turmoil concurrently.
197

Probabilities are given in percent terms and are derived from the multinomial logit estimation shown in the top panel.
198 Graciela L. Kaminsky and Carmen Reinhart

by the higher worldwide exposure of German banks to emerging markets in Asia,

Latin America, and transition economies during the late 1990s. The bottom panel
provides another metric to capture the reach of turmoil in the three financial cen-
ters: the probability of low, medium, and high globalization, conditional on tur-
moil in each financial center and derived from the multinomial estimation. Again
the likelihood of medium-to-high globalization on days of turmoil in Germany is
about 60 percent, but only about 40 percent when shocks originate in Japan, with
market jitters in the United States triggering medium-to-high globalization with a
probability of almost 50 percent.
Table 8.9 reports the results for turmoil originating in emerging markets. The
first three columns report the estimates for Brazil, the next three for Russia, and
the last three for Thailand. As we did when we evaluated weak forms of global-
ization, we pay particular attention to whether days of turmoil in the three crisis-
prone emerging markets coincide with days of turmoil or with days of no turmoil
in financial centers. Financial centers would be irrelevant in explaining high
degrees of globalization of turmoil if the coefficient b attached to the dummy cap-
turing turmoil in the emerging market–financial center cluster is not statistically
different from the coefficient b attached to the dummy capturing turmoil in just
the emerging market. This hypothesis is tested in the third column for each
emerging market. In all cases, we reject this hypothesis at all conventional signifi-
cance levels. To better understand the effects of turmoil in the various countries,
the bottom panel of table 8.9 also reports the conditional probabilities of low, me-
dium, and high globalization obtained from the estimation of equation 8.4. The
results for Brazil indicate that low globalization is the most likely outcome when
turbulence in Brazil does not coincide with turbulence in a financial center. In con-
trast, when the financial center is also experiencing an extreme event, high global-
ization becomes the most likely event, with the probability reaching 57 percent.
Interestingly, if there is no turmoil in Brazil or the United States, the likelihood of
a high clustering of countries with turmoil collapses to 2 percent. The results for
Russia are quite similar. Again, the necessary ingredient for high globalization is
the coincidence of turmoil in Russia and Germany. In this case, the average prob-
ability of high globalization is about 50 percent, but declines to 2 percent when
only Russia experiences turbulence. The results for Thailand indicate that the
reach of the Asian crisis was limited in scope. Still, jitters in Thailand are trans-
mitted to other Asian countries only on days of jitters in Japan. The explanatory
power of these shocks, as captured by the pseudo R 2 , is high for Brazil and Russia
but, as expected, quite small for Thailand.
To evaluate jointly the effects of turmoil in the three emerging markets and fi-
nancial centers, we report the panel estimates in tables 8.10 and 8.11. As shown in
table 8.10, each of the three emerging market–financial center clusters contributes
The Globalization of Financial Turmoil 199

to trigger financial turmoil worldwide, as captured by the statistically significant

b coefficients of the three crisis-cluster dummies for the high globalization event.
Still, the contribution of the Thailand–Japan cluster is somewhat smaller. Our
panel estimation, though, has a smaller predictive power than the nonpanel esti-
mations because of the restrictions imposing similar effects of turmoil of the
various center-periphery clusters across the five regions. Our more stringent defi-
nition of high-globalization episodes is also reflected in lower probabilities of high
globalization following turbulences in the three center-periphery clusters.
Finally, the results in table 8.11 bring to the spotlight the magnification effect of
simultaneous turbulences in several center-periphery clusters. Note that the prob-
ability of high globalization now increases to 94 percent when the three crisis clus-
ters experience turbulences, but just 13 percent when one crisis cluster is in
turmoil. Note that the probability of high globalization on days of no turmoil in
any of the crisis clusters is just 1 percent.

8.4 The Origins of Globalization

In the previous section, we evaluated the odds of simultaneous turbulence around

the world when crisis-prone emerging markets and financial centers were experi-
encing turmoil. We did not explain the origin of these turbulences. To do that, we
have to bring in information beyond that of daily movements in equity prices.
Our source is the written record: we used reports from Bloomberg.com, the Finan-
cial Times, and the Wall Street Journal to construct a chronology of news in those
days. We limited our search to days on which at least 50 percent of countries in
one region had stock market jitters. This chronology is reported in the appendix
table (located at the end of this chapter). The first column dates the days of re-
gional and global turmoil. The next six columns report the proportion of coun-
tries, worldwide and by region, with stock market turmoil. For clarity, we just
report the proportion of markets in turmoil when it reaches more than 50 percent
of the countries worldwide or in each region. The last column reports the news.
To study the onset and propagation of turmoil, it is important to collect all news,
local and foreign, that triggers jitters. This news can be about the state of the econ-
omy, financial institutions, or policies, or may just be rumors. The appendix table
does not report all the news events that moved markets on a particular day, but
reports the most common source of market jitters in the region or around the
world.
As shown in this chronology, the first day of worldwide turmoil is October 27,
1997, with 57 percent of the countries around the globe experiencing turmoil. The
tension started to build up toward the end of August. Until that time, while
several Asian countries experienced turbulences, they did not spread to other
200 Graciela L. Kaminsky and Carmen Reinhart

countries in that region. But on August 28, 1997, financial markets in Indonesia,
Malaysia, the Philippines, Singapore, and Hong Kong collapsed amid a deepen-
ing loss of confidence in the ability of governments to tackle their severe economic
problems. On October 22 turmoil reached Hong Kong and spread in Asia, with
about 60 percent of the Asian countries experiencing market crashes. The crisis in
Hong Kong deepened and on October 23, it triggered a global sell off in Europe,
the G7 countries, and Latin America. By October 27, worldwide globalization
reached about 60 percent of the countries in the sample. This time around, the
globalization of the turmoil was short-lived and within two days markets re-
bounded, with massive rallies around the world.
December 11 is the next day of significant interregional spillover, with Korea at
the center of the debacle in Asia and Europe. Still, repercussions in the G7 coun-
tries were minor. Another day of interregional turmoil was January 12, 1998. At
the heart of the jitters was the collapse of Peregrine (Hong Kong), one of Asia’s
largest investment banks.
The next cluster of global instability started toward the end of May 1998, with
Russian tension spreading to Latin America, transition economies, Asia, Europe,
and the G7 countries. The degree of globalization rapidly rose, reaching about 50
percent of countries worldwide by June 15. Rumors of devaluation in China and
the weakness of the Japanese economy and the yen also contributed to the build-
up of skittishness. The degree of globalization reached 60 percent on August 11.
On August 21, shares of German banks heavily exposed to Russia collapsed, trig-
gering downfalls in other G7 countries. On August 27, the failed auction of Rus-
sian GKOs reignited fears of financial collapse, bringing major downturns in 75
percent of countries worldwide. Financial turmoil griped Latin American markets
following Moody’s downgrade of Brazilian and Venezuelan foreign debt.
Moody’s also put Argentina’s debt and its eleven banks on review for a possible
downgrade on September 3. While markets in some regions rebounded during
the first week of September, financial concerns, brought again to the limelight by
Standard & Poor’s downgrade of Spain’s second-largest bank (with heavy expo-
sure to Argentina) and of Argentina’s two largest banks on September 10,
together with LTCM’s collapse and bail-out on September 24, triggered stock
market crashes around the world. This episode of worldwide financial instability
came to an end with news of credit easing in financial centers; this was related to
the intermeeting reduction in the federal funds interest rate on October 15 in the
United States.
The last episode of financial instability in our sample occurred around the time
of the devaluation of the Brazilian real, which was extremely short-lived. Only on
January 13 did financial markets around the world collapse.
The Globalization of Financial Turmoil 201

Table 8.12 summarizes our findings about the news that rocked financial mar-
kets. But first, the top panel shows the proportion of days of rallies and days of
crashes during episodes of high regional and world globalization (at least 50 per-
cent of countries affected by turmoil). Note that 85 percent of the episodes of high
world globalization involve stock market crashes. Episodes of high regional glob-
alization are more balanced. With the exception of the Asian region, in which
days of joint rallies outnumber days of simultaneous crashes, about 60 percent of
the days of high regional globalization consist of crashes. The middle panel classi-
fies the days of high globalization, both at a regional level and worldwide, accord-
ing to the type of news that seems to have triggered the spillover. Financial
concerns from bankruptcies of large banks or adverse shocks in one or more asset
markets in a center country seem to be at the core of high worldwide globaliza-
tion (40 percent of the episodes). Only 20 percent of the days of high spillovers
seem to be driven by economic, political, and monetary news at the center. An-
other important source of instability is concerns about financial fragility in the
periphery (25 percent of the episodes). In contrast, financial worries in center
countries only account for 26 percent of the episodes of high regional globaliza-
tion. Financial fragility in the periphery seems to be at the heart of regional tur-
bulences (31 percent of the episodes). Finally, international agreements also
contribute to regional turbulences.
One final aspect of globalization that we have still not addressed is whether
high globalization occurs when the magnitude of the shocks in the stock market
is larger. The bottom panel addresses this question. We first divide extreme
returns in three categories according to their size: large (within the 1-percent criti-
cal region on both tails), medium (between the 1-percent and 3-percent critical
regions on both tails), and small (between the 3-percent and 5-percent critical
regions). Afterward, we estimate the average size of the returns for all the coun-
tries in turmoil for each episode of low, medium, and high world globalization.
The bottom panel in table 8.12 shows the proportion of episodes of low, medium,
and high world globalization with small, medium, and large returns. Larger (in
absolute values) returns are more common on days of high globalization: all the
shocks in episodes of high globalization are clustered in, at the most, the 3-percent
critical region, while during episodes of low globalization 46 percent of the shocks
are relatively small (between the 3-percent and 5-percent critical regions).

8.5 Concluding Comments

This paper presents a new approach to measuring and understanding systemic

financial turbulences. We defined two measures of systemic disturbances—weak-
202 Graciela L. Kaminsky and Carmen Reinhart

Table 8.12
Days of Globalization: Asymmetries, Origins, and Size of Shocks
A. Asymmetries

Days of high globalization

Proportion Proportion
Regions of crashes of rallies

World 85 15
Asia 29 71
Latin America 69 31
Europe 61 39
G7 56 44
Transition economies 61 39
B. News on Days of High Globalization
Proportion of Days with News about:

Economy and Monetary International

Financial sector politics policy agreements
Banking Other

Periph- Periph- Periph- Periph- Periph-

Regions Center ery Center ery Center ery Center ery Center ery
Regional 8 8 18 23 11 10 11 3 2 7
World 10 10 30 15 10 10 10 5

C. Degree of Globalization and Size of Returns

Returns
Degree of world
globalization Small Medium Large
Low 46 48 5
Medium 12 86 2
High 0 92 8
Notes: Numbers in the above tables are in percent. In panel C, small returns C are returns between the
3rd (97th) and 5th (95th) percentiles. Medium returns are returns between the 1st (99th) and 3rd (97th)
percentiles. Large returns are returns in the 1st (99th) percentile. The first cell of this panel indicates
that 46 percent of the days of low globalization had countries experiencing on average a small return.
The Globalization of Financial Turmoil 203

and strong-form globalization—and created the corresponding indices of global-

ization. These indices allowed us to capture the routes through which market
jitters in one country reach other countries in the same region, or even worldwide.
They also allowed us to estimate the likelihood of low-to-high globalization fol-
lowing a variety of shocks in crisis-prone emerging markets and financial centers.
One of the preliminary conclusions we draw from this exercise is that financial
centers are at the core of systemic problems: the worldwide globalization of the
turbulences in Asia in fall 1997 only occurred after the stock market crash in the
United States on October 27, while the Russian downfall spread around the globe
only after it triggered fragilities in German banks and helped to provoke LTCM’s
bankruptcy. Without distress in a financial center, disturbances spread at most re-
gionally, with the ‘‘silk road’’ of regional financial distress related in part to trade
links, but also to financial linkages. For example, as documented in Kaminsky,
Lyons, and Schmukler (2004), the 1994 Mexican crisis spread so rapidly to Argen-
tina and Brazil via the massive mutual fund (specialized in Latin America) with-
drawals from those two countries. Finally, our evidence indicates that collapses
and not rallies are at the heart of high-globalization episodes, suggesting the need
for models with asymmetries to explain systemic turmoil.
Our research has focused on explaining the geographical extent of financial tur-
moil. Still, the temporal dimension of high-globalization episodes of turbulences
varies as much, with some episodes lasting just a couple of days (sell-off in stock
markets around the world following the Hong Kong collapse in October 1997)
while others, such as the turmoil during fall 1998, lingered much longer. Also,
our research, like most of the previous literature, has just focused on a particular
asset market. But the degree of systemic problems should not only be understood
as synchronized jitters across a particular asset market in a variety of countries,
but also as simultaneous turmoil across markets in a particular country. Future re-
search should inquire into these differences too.

Acknowledgments

We have received insightful comments and suggestions from Fernando Broner,

Guillermo Calvo, and Takatoshi Ito, as well as participants at presentations held
at the 2002 Third Joint Central Bank Research Conference on Risk Measurement
and Systemic Risk held in Basel, Switzerland; the LACEA 2002 meeting in
Madrid, Spain; the Bank of England, Cornell University, George Washington Uni-
versity, the London School of Economics, the University of Maryland, and the
University of West Virginia. We are grateful to Amine Mati for superb research
assistance.
204 Graciela L. Kaminsky and Carmen Reinhart

Appendix Table
The Globalization of Financial Turmoil: Chronology of News January 1, 1997 to August 31, 1999
DAY ALL ASIA EUR G7 LA TRA NEWS

7 Apr 97 57 EUR: European and US stocks up,

benefiting from comments of EU finance
ministers who indicated the single
currency will begin on time. (FS-
OTHER, CENTER)
28 Aug 97 71 ASIA: Investors flee the region amid a
deepening loss of confidence in the
ability of governments to tackle their
severe economic problems. (E&PN,
PERIPHERY)
22 Oct 97 57 ASIA: Share prices fall sharply in Asian
markets due to a sharp drop in futures
prices in Singapore and fears about
higher interest rates and currency
stability in Hong Kong. (FS-OTHER,
PERIPHERY)
23 Oct 97 57 71 57 G7, EUR, LA: Hong Kong Monetary
Authority was forced to sell US dollars
to support the currency, triggering
interest rate hikes that prompted a
global sell off. (FS-OTHER, PERIPHERY)
27 Oct 97 57 71 86 86 EUR, G7, LA: A $600 billion sell-off shut
down the US market for the first time
since 1981. The sell-off was triggered by
Southeast Asia’s shaky economies and
by a jump in interest rates, as well as by
a stream of weak earning reports. Panic
grips other regions as US market
crashes, especially after the Hong Kong
declines of the past week. (E&PN,
PERIPHERY); (FS-OTHER, CENTER)
28 Oct 97 80 100 86 86 71 57 ASIA, EUR, TRA, G7: still reacting to US
market crash and Hong Kong crash. (FS-
OTHER, CENTER)
LA: Markets soar as US market rallies.
(FS-OTHER, CENTER)
29 Oct 97 63 57 71 86 71 EUR, G7, TRA: Markets soared as US
soared the day before. (FS-OTHER,
CENTER)
ASIA: Asian markets finish lower as
investors fear another steep drop in US
markets. (FS-OTHER, CENTER)
30 Oct 97 71 57 EUR: Stocks soared after Greenspan
eased concern that inflation could be on
the rise. (MP, CENTER)
The Globalization of Financial Turmoil 205

Appendix Table
(continued)
DAY ALL ASIA EUR G7 LA TRA NEWS
LA: Strong declines in the region
stemmed from contagion in Asia. Fears
about Brazil’s real currency and liquidity
crunch of its banking system. (FS-
OTHER, PERIPHERY); (FS-BANKING,
PERIPHERY)
3 Nov 97 57 ASIA: Stocks rally as a financial aid
package to Indonesia restores calm to
the region. China also eases credit. (IA,
PERIPHERY); (MP, PERIPHERY)
7 Nov 97 57 G7: The US dollar surges, reaching a six-
month high as concerns increased in the
market over the ability of the Japanese
government to revive the country’s
economy. (E&PN, CENTER)
12 Nov 97 71 71 LA: Concern about fiscal austerity
package announced by Brazil. Markets
also fall after steep declines in Asian
markets. (E&PN, PERIPHERY), (FS-
OTHER, PERIPHERY)
TRA: Stocks fall after major drops in
Asian markets. (FS-OTHER,
PERIPHERY)
17 Nov 97 86 57 EUR, G7: Stocks up as Japan PM hints
that public spending may be used to
stimulate the economy and protect
depositors following the collapse of the
nation’s largest bank. US reports low
inflation measures. (FS-BANKING
CENTER); (E&PN, CENTER)
24 Nov 97 57 EUR: Shares fall after the collapse of
Japan’s fourth-largest brokerage firm,
Yamaichi Securities. (FS-BANKING,
CENTER)
1 Dec 97 57 G7: Stock markets rally on gains in
Asian markets overnight. (FS-OTHER,
PERIPHERY)
11 Dec 97 100 57 ASIA: Stocks slumped as Moody’s cut
rating of South Korea’s currency.
(E&PN, PERIPHERY)
EUR: Stocks down amid a new wave of
selling in Asian markets and signs of
weakness in the US economy. (FS-
OTHER, PERIPHERY); (E&PN,
CENTER)
206 Graciela L. Kaminsky and Carmen Reinhart

Appendix Table
(continued)
DAY ALL ASIA EUR G7 LA TRA NEWS
9 Jan 98 71 LA: Asian turmoil, especially concerns
about Indonesia, causes market declines.
Central Bank of Chile raises key interest
rate. (FS-OTHER, PERIPHERY); (MP,
PERIPHERY)
12 Jan 98 71 71 EUR, TRA: Peregrine, one of Asia’s
largest investment banks (Hong Kong),
files for liquidation, raising concerns
about emerging markets in general. (FS-
BANKING, PERIPHERY)
13 Jan 98 71 ASIA: Stocks rose on optimism about
IMF-backed reforms for the region. (IA,
PERIPHERY)
14 Jan 98 86 ASIA: Stocks continued to rise on
optimism about IMF-backed reforms for
the region. (IA, PERIPHERY)
19 Jan 98 100 ASIA: Indonesia signaled commitment
to the much-awaited bank reform. Cam-
dessus issues statement of confidence
about Malaysia and countries in the
region. (FS-BANKING, PERIPHERY);
(IA, PERIPHERY)
22 Jan 98 57 ASIA: The plunging Indonesian rupiah
dragged the rest of Asia into a down-
ward spiral. (FS-OTHER, PERIPHERY)
2 Feb 98 71 ASIA: Stocks up as value-oriented funds
flooded back into Asia from Europe and
US. Strength driven by liquidity even
though nothing changed in the funda-
mentals front. (FS-OTHER, PERIPHERY)
27 Apr 98 86 86 EUR, G7: Concern US will raise interest
rates to fight inflation. (MP, CENTER)
26 May 98 57 LA: Concerns about a potential deval-
uation in Russia affecting Brazil and
other emerging markets. (FS-OTHER,
PERIPHERY)
27 May 98 57 EUR: Speculation about Russian
devaluation of the ruble caused fall in
stock prices. (FS-OTHER, PERIPHERY)
1 Jun 98 57 TRA: Russian stock prices plummeted
while the main market for Russian
futures announced that it was sus-
pending trading indefinitely. Unfulfilled
expectations of foreign aid to Russia
contributed to the declines. (FS-OTHER,
PERIPHERY); (IA, PERIPHERY)
The Globalization of Financial Turmoil 207

Appendix Table
(continued)
DAY ALL ASIA EUR G7 LA TRA NEWS
15 Jun 98 51 86 57 57 ASIA: Japanese government announced
that GDP contracted for a second
consecutive quarter. (E&PN, CENTER)
ASIA, LA, G7, TRA: Loss of confidence
in emerging markets in general as
Russian market tumbled for a seventh
straight day. (FS-OTHER, PERIPHERY)
17 Jun 98 71 ASIA: US and Japan coordinated actions
to sell US dollars and buy Japanese yen.
Markets soared due to the stronger yen.
(IA, CENTER)
18 Jun 98 100 ASIA: Countries in the region still
reacting to the US and Japan coordi-
nated actions to prop up the yen. (IA,
CENTER)
14 Jul 98 71 TRA: Russia would receive 22.6 billion
dollars from IMF and other bilateral
donors. (IA, PERIPHERY)
11 Aug 98 60 86 71 57 EUR, G7, LA: Foreign investors seemed
to be the main driving force behind the
market drop. Fears of a weaker yen, and
the prospect of devaluation in China,
sent shock waves throughout the world.
(FS-OTHER, CENTER); (FS-OTHER,
PERIPHERY)
13 Aug 98 71 TRA: Russian shares fell more than 10
percent early on growing fears of a
liquidity crisis among Russian banks.
(FS-BANKING, PERIPHERY)
18 Aug 98 57 EUR: Gains in European markets
following a major Wall Street advance
(FS-OTHER, CENTER)
20 Aug 98 71 LA: Concern Russian banks may fail
and Venezuela may devalue (FS-
BANKING, PERIPHERY); (FS-OTHER,
PERIPHERY)
21 Aug 98 54 71 71 71 LA: Concern about imminent currency
devaluation in Venezuela. (FS-OTHER,
PERIPHERY)
EUR, G7, LA: Russia’s Central Bank
stated that some Russian banks could go
bankrupt accentuating the Russian
financial crisis. In Germany (a major
lender to Russia) stocks plunged,
triggering downfalls in London and
Paris. (FS-BANKING, PERIPHERY); (FS-
OTHER, CENTER)
208 Graciela L. Kaminsky and Carmen Reinhart

Appendix Table
(continued)
DAY ALL ASIA EUR G7 LA TRA NEWS
26 Aug 98 71 EUR: Stocks fall as Russia announces its
debt restructuring plan. (FS-OTHER,
PERIPHERY)
27 Aug 98 74 86 100 100 86 EUR, G7, LA, TRA: Russia’s government
unable to sell its newly restructured
GKO bills, spreading fear that global
crisis will continue. (FS-OTHER,
PERIPHERY)
1 Sep 98 57 57 LA: Stocks end sharply higher mirroring
the DJIA’s rebound. (FS-OTHER,
CENTER)
EUR: Stocks up on optimism about
Europe’s prospects. (FS-OTHER,
CENTER)
2 Sep 98 71 57 TRA: Markets rebound as investors went
for bargains. (FS-OTHER, PERIPHERY)
EUR: Stocks follow rebound in the US
stock market. (FS-OTHER, CENTER)
3 Sep 98 57 57 57 LA: Moody’s downgraded Brazil’s and
Venezuela’s foreign debt and put
Argentina’s foreign currency debt and
11 banks on review for a possible
downgrade. (E&PN, PERIPHERY), (FS-
BANKING, PERIPHERY)
EUR, G7: European stock markets were
hurt by a dollar plunge and worries that
financial troubles are spreading from
Russia and Asia to Latin America. (FS-
OTHER, PERIPHERY)
4 Sep 98 57 TRA: Russia’s parliament delays a vote
on Chernomyrdin’s appointment as
Prime Minister at Yeltsin’s request.
(E&PN, PERIPHERY)
7 Sep 98 57 57 ASIA: Stronger yen and a higher stock
market helps Japanese banks but fund
managers stay skeptical. (FS-BANKING,
CENTER)
EUR: Greenspan hints he would favor
cutting interest rates. (MP, CENTER)
8 Sep 98 57 57 G7, TRA: Renewed confidence was felt
thanks to market-supportive comments
from Fed Chairman Alan Greenspan.
(MP, CENTER)
The Globalization of Financial Turmoil 209

Appendix Table
(continued)
DAY ALL ASIA EUR G7 LA TRA NEWS
10 Sep 98 60 71 71 86 57 EUR, G, TRA, LA: Worries about banks’
exposures as S&P downgrades Spain’s
second-largest bank. Credit ratings for
Argentina’s two largest banks were also
reduced. (FS-BANKING, CENTER); (FS-
BANKING, PERIPHERY)
11 Sep 98 57 LA: Brazilian Government boosted
overnight interest rates by 20 percentage
points to try to stem capital flight, which
reached 2.2 billion dollars the day
before. (FS-OTHER, PERIPHERY)
14 Sep 98 71 71 G7, TRA: Russia’s new PM pledges to
revive the economy. (E&PN,
PERIPHERY)
15 Sep 98 86 57 LA, TRA: G7 meeting hints at financial
aid for Latin America. Argentina may
borrow 5.7 billion dollars from the
World Bank and other international
institutions. (IA, PERIPHERY)
17 Sep 98 54 86 86 EUR, G7: Greenspan states that there is
no move to coordinate interest rates
(MP, CENTER)
21 Sep 98 86 57 G7, EUR: Concern about Japan’s
recession and low growth potential for
OECD countries due to emerging
markets collapse and deepening
financial collapse. Political parties in
Japan remains at odds on how to use
taxpayer money to prop up LTCB of
Japan. (E&PN, CENTER); (FS-OTHER,
EMERGING); (FS-BANKING, CENTER)
22 Sep 98 71 EUR: US markets rebound day after the
Clinton grand jury testimony. (E&PN,
CENTER)
23 Sep 98 57 57 100 LA: President of IDB says Brazil could
receive up to 50 billion dollars in aid
from international institutions. IMF and
US also gave statements of support for
Brazil aid. (IA, PERIPHERY)
G7, EUR: Investors hope that Greenspan
will hint at a possible rate cut when he
testifies before the Senate banking
committee. (MP, CENTER)
24 Sep 98 86 57 ASIA: Stocks up as Greenspan suggests
he may lower interest rates. (MP,
CENTER)
210 Graciela L. Kaminsky and Carmen Reinhart

Appendix Table
(continued)
DAY ALL ASIA EUR G7 LA TRA NEWS
LA: Stocks down as concern over banks
is felt after some of the largest banks put
together a 4 billion dollar bailout of
LTCM, raising concern about credit.
Brazil announces fiscal austerity
measures. (E&PN, PERIPHERY); (FS-
BANKING, CENTER)
25 Sep 98 71 TRA: Russian tax collection continued to
plummet in September, due to the crash
on Russian financial markets and the
country’s ensuing banking crisis
(statement by tax official). (E&PN,
PERIPHERY)
30 Sep 98 57 G7: US cut interest rates and asked other
countries to follow suit. (MP, CENTER)
1 Oct 98 66 86 100 71 G7, EUR, LA: Concerns about global
economic slump. Report US manufac-
turing production weakened for fourth
straight month as exports slumped.
(E&PN, CENTER)
2 Oct 98 57 57 LA: Stock markets soared on hopes of a
financial package for troubled Brazil.
(IA, PERIPHERY)
TRA: Stocks still falling following global
declines of October first. (FS-OTHER,
CENTER)
6 Oct 98 57 G7: Disappointment that the G7 meeting
in Washington failed to adopt a clear
strategy to address global economic
issues drove share prices sharply lower
in world markets. (IA, CENTER)
8 Oct 98 51 86 71 EUR, G7: Speculation the Fed would cut
interest rates. Japan moves to repair its
economy. (MP, CENTER) (E&PN,
CENTER)
9 Oct 98 57 57 57 ASIA, G7: Interest cuts in UK and other
European countries in the preceding
week generated rallies in several
markets. (MP, CENTER)
LA: Brazilian authorities and the Inter-
national Monetary Fund issued a joint
statement on the availability of a rescue
package to help cushion the region from
market turmoil. (IA, PERIPHERY)
The Globalization of Financial Turmoil 211

Appendix Table
(continued)
DAY ALL ASIA EUR G7 LA TRA NEWS
12 Oct 98 54 57 71 71 EUR, G7, ASIA: Japan will substantially
increase the amount of money it will
spend on shoring up its fragile banking
system. (FS-BANKING, CENTER)
16 Oct 98 71 71 EUR, G7: Fed Funds rate cut by a
quarter percentage point on Oct. 15.
(MP, CENTER)
20 Oct 98 57 57 EUR, G7: Suggestions that France and
Germany would lower their interest
rates boosted investor sentiment in
Europe as well as continued gains in the
USA and a rally in Asian markets. (MP,
CENTER); (FS-OTHER, CENTER); (FS-
OTHER, PERIPHERY)
27 Oct 98 71 EUR: Italy makes a surprise cut in
interest rate by a full percentage point.
(MP, CENTER)
30 Oct 98 71 LA: G7 countries said they would back a
new IMF credit line to Brazil, speeding
aid to Brazil. (IA, PERIPHERY)
2 Nov 98 71 57 EUR, TRA: Stocks rallied after the
October 30 US commerce department
report announcing better than expected
third quarter growth rates. (E&PN,
CENTER)
4 Nov 98 71 G7: Democrats increased seats in the US
Congressional elections, the first party
with an incumbent resident to do this
since 1934. Stocks rally after interest rate
cuts in Italy and Sweden in the past
week. (E&PN, CENTER); (MP,
CENTER)
10 Nov 98 57 ASIA: Investors await the release of the
Japanese government’s stimulus
package. (E&PN, CENTER)
11 Nov 98 57 ASIA: Japan’s newest economic stimulus
package is expected to be the largest
ever. (E&PN, CENTER)
20 Nov 98 57 EUR: European stocks finished with
strong gains as bourses benefited from
hopes of further European rate cuts.
(MP, CENTER)
30 Nov 98 57 57 G7: Global markets were given a boost
after the DJIA marked a record high.
(FS-OTHER, CENTER)
212 Graciela L. Kaminsky and Carmen Reinhart

Appendix Table
(continued)
DAY ALL ASIA EUR G7 LA TRA NEWS
LA: Latin American investors were
influenced by heavy profit taking on
Wall Street and Brazil. (FS-OTHER,
PERIPHERY)
1 Dec 98 57 71 G7, EUR: Stocks down on weak dollar.
(FS, CENTER, OTHER)
3 Dec 98 57 LA: The US dollar weakened as
investors were discouraged by the
continuing decline in U.S. stocks and
Wednesday’s defeat in the Brazilian
Congress of an important government
austerity measure. (FS-OTHER,
CENTER); (E&PN, PERIPHERY)
21 Dec 98 57 G7: High expectations on the euro boost
stocks. (FS-OTHER, CENTER)
4 Jan 99 86 57 EUR, G7: Stock prices ended up higher
lured by a weak dollar and start of euro
trading. (FS-OTHER, CENTER)
6 Jan 99 57 71 G7: US rallied on the back of technology
stocks. (FS-OTHER, CENTER)
ASIA: Japanese market followed an
overnight jump in New York stocks lead
by strength in the high-technology
sector. (FS-OTHER, CENTER)
12 Jan 99 57 LA: Markets closed sharply lower due to
rumors of an interest rate hike in Brazil
and a near $200 million outflow. (FS-
OTHER, PERIPHERY)
13 Jan 99 66 100 57 86 57 EUR, LA, TRA: Brazil’s Central Bank
chairman resigns. Brazil devalues its
currency. (E&PN, PERIPHERY); (MP,
PERIPHERY)
14 Jan 99 57 LA: Standard & Poor’s downgraded
certain Latin American banks and some
of Brazil’s foreign currency debt. (FS-
BANKING, PERIPHERY)
15 Jan 99 71 LA: Brazil lets its currency float against
the dollar. (FS-OTHER, PERIPHERY)
18 Jan 99 57 EUR: Bank mergers in France, Spain and
calmer financial markets in Brazil
pushed stocks higher. (FS-BANKING,
CENTER); (FS-OTHER, PERIPHERY)
9 Feb 99 57 G7: There were growing concerns in
Europe about a slowdown in the eco-
nomy. European markets fell following
financial turmoil in emerging markets.
The Globalization of Financial Turmoil 213

Appendix Table
(continued)
DAY ALL ASIA EUR G7 LA TRA NEWS
Japanese investors were waiting for mea-
sures, if any, from the BOJ to curb the
recent sharp rise in bond yields, which
would increase borrowing costs for com-
panies and could stall Japan’s efforts to
revive its battered economy. (E&PN,
CENTER); (FS-OTHER, PERIPHERY)
5 Mar 99 86 G7: Labor department reported hourly
wages rose 0.1 percent in February, less
than the 0.3 percent forecasted. Unem-
ployment went up 0.1 percent point.
(E&PN, CENTER)
16 Apr 99 57 ASIA: Influx of European funds brought
up Asian stocks posting sharp gains
throughout the region. (FS-OTHER,
CENTER)
19 Apr 99 57 ASIA: Investors confident that the global
financial crisis is largely over. (FS-
OTHER, PERIPHERY)
26 May 99 71 LA: Markets rebound as fears concern-
ing Argentina’s ability to maintain its
currency board (as well as fears about a
potential political scandal involving
Brazilian President Cardoso) subside.
(FS-OTHER, PERIPHERY); (E&PN,
PERIPHERY)
29 July 99 57 G7: Investors were relieved when Alan
Greenspan offered nothing new to upset
global markets in a testimony to US
lawmakers. (MP, CENTER)
Notes:
FS: News from the financial sector. They could either originate in the banking sector (BANKING) or
not (OTHER).
MP: News about monetary policy.
E&PN: News about the economy (excluding the financial sector) and political news.
IA: Refers to international agreements or policy coordination actions.
ASIA: Includes Hong Kong, Indonesia, Malaysia, the Philippines, Singapore, South Korea, and Thailand.
EUR: Includes Finland, Greece, Holland, Norway, Spain, Sweden, and Turkey.
G7: Includes Canada, France, Germany, Italy, Japan, United Kingdom, and the United States.
LA: Includes Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela.
TRA: Includes Czech Republic, Estonia, Hungary, Poland, Russia, Slovakia, and Ukraine.
Numbers in cells represent the percentage of countries in their respective region (or world) experiencing
turmoil on that day.
The parenthetical statements after each news event explain the region from which news originated and
our classification of news. For example, on July 29, 1999, 57 percent of the G7 countries were affected by
Alan Greenspan’s testimony. His testimony was classified as Monetary Policy News originating in the
Center (MP, CENTER).
214 Graciela L. Kaminsky and Carmen Reinhart

Notes

1. Extreme returns are those returns in the 5th and 95th percentile of the distribution.
2. Also see Calvo and Mendoza (2000) for evidence suggesting that this mechanism can be important.
3. See, for example, Kaminsky and Reinhart (1999).
4. See, for example, Eichengreen, Rose, and Wyplosz (1996), Glick and Rose (1998), and Kaminsky and
Reinhart (2000).
5. See also Danielsson and de Vries (1997), De Bandt and Hartmann (2000), Hartman, Straetmans, and
Devries (2004), Longin (1996), and Mati (2001) for studies of extreme returns in stock and bond
markets.
6. See, for example, Gelos and Sahay (2000), Glick and Rose (1998), and Kaminsky and Reinhart (2000).
7. In Kaminsky and Reinhart (2000) we constructed a similar index. In that paper, the index was
the proportion of countries with currency crises, which was used to predict currency crises in other
countries. Bae, Karolyi, and Stulz (2000) also look at simultaneous financial strains in Asia and Latin
America and construct a similar index, finding that contagion is predictable using a small set of macro-
economic variables.

References

Bae, Kee-Hong, Andrew G. Karoyli, and Rene M. Stulz. 2003. ‘‘A New Approach to Measuring Finan-
cial Contagion.’’ Review of Financial Studies 16, no. 3: 717–763.

Calvo, Guillermo. 1998. ‘‘Capital Market Contagion and Recession: An Explanation of the Russian
Virus.’’ Mimeo., University of Maryland.
Calvo, Guillermo A., Leonardo Leiderman, and Carmen M. Reinhart. 1993. ‘‘Capital Flows and Real
Exchange Rate Appreciation in Latin America: The Role of External Factors.’’ IMF Staff Papers 40, no. 1.
———. 1996. ‘‘Capital Flows to Developing Countries in the 1990s: Causes and Effects.’’ Journal of Eco-
nomic Perspectives 10: 123–139.
Calvo, Guillermo, and Enrique Mendoza. 2000. ‘‘Rational Contagion and the Globalization of Secur-
ities.’’ Journal of International Economics 51, no. 1: 79–113.
Calvo, Sara, and Carmen M. Reinhart. 1996. ‘‘Capital Flows to Latin America: Is There Evidence of
Contagion Effects?’’ In Private Capital Flows to Emerging Markets After the Mexican Crisis, eds. Guillermo
A. Calvo, Morris Goldstein, and Eduard Hochreitter, 151–171. Washington, D.C.: Institute for Interna-
tional Economics.
Corsetti, Giancarlo, Paolo Pesenti, Nouriel Roubini, and Cedric Tille. 1998. ‘‘Structural Links and Con-
tagion Effects in the Asian Crisis: A Welfare-Based Approach.’’ Mimeo., Yale University, New Haven,
CT.
Danielsson, J., and C. de Vries. 1998. ‘‘Value-at-Risk and Extreme Returns.’’ Working Paper No. 98-
017/2, Tinbergen Institute, Amsterdam.
De Bandt, Olivier, and Philipp Hartmann. 2000. ‘‘Systemic Risk: A Survey.’’ Discussion Paper No. 2634,
CEPR, London.
Eichengreen, Barry, Andrew Rose, and Charles Wyplosz. 1996. ‘‘Contagious Currency Crises.’’ Work-
ing Paper No. 5681, NBER, Cambridge, MA.
The Globalization of Financial Turmoil 215

Frankel, Jeffrey, and Sergio Schmukler. 1998. ‘‘Crises, Contagion, and Country Funds: Effects on East
Asia and Latin America.’’ In Managing Capital Flows and Exchange Rates: Perspectives from the Pacific Ba-
sin, ed. Reuven Glick, 232–266. New York: Cambridge University Press.
Gelos, Gastón, and Ratna Sahay. 2000. ‘‘Financial Market Spillovers in Transition Economies.’’ Working
Paper WP/00/71, IMF, Washington, D.C.
Glick, Reuven, and Andrew Rose. 1998. ‘‘Contagion and Trade: Why Are Currency Crises Regional?’’
Working Paper No. 6806, NBER, Cambridge, MA.
Kaminsky, Graciela L., and Carmen M. Reinhart. 1999. ‘‘The Twin Crises: The Causes of Banking and
Balance-of-Payments Problems.’’ American Economic Review 89, no. 3: 473–500.
———. 2000. ‘‘On Crises, Contagion, and Confusion.’’ Journal of International Economics 51, no. 1: 145–
168.
———. 2001. ‘‘Bank Lending and Contagion: Evidence from the Asian Crisis.’’ In Regional and Global
Capital Flows: Macroeconomic Causes and Consequences, eds. T. Ito and A. Krueger, 73–99. Chicago: Uni-
versity of Chicago Press.
Kaminsky, Graciela, Richard Lyons, and Sergio Schmukler. 2001. ‘‘Economic Fragility, Liquidity, and
Risk: The Behavior of Mutual Funds during Crises.’’ Working Paper, World Bank, Washington, D.C.
Kaminsky, Graciela, Richard Lyons, and Sergio Schmukler. 2004. ‘‘Managers, Investors, and Crises:
Mutual Fund Strategies in Emerging Markets.’’ Journal of International Economics 64, no. 1: 113–134.
Lewis, Arthur. 1977. The Evolution of the International Economic Order. Princeton: Princeton University
Press.
Longin, F. M. 1996. ‘‘The Asymptotic Distribution of Extreme Stock Market Returns.’’ Journal of Business
69: 383–408.
Mati, Amine. 2001. ‘‘Extreme Returns: News and Patterns of Contagion.’’ Mimeo., George Washington
University, Washington, D.C.
Van Rijckeghem, Caroline, and Beatrice Weder. 2000. ‘‘Financial Contagion: Spillovers through Bank-
ing Centers.’’ Mimeo., IMF, Washington, D.C.
9 Why Do Some Countries
Recover More Readily from
Financial Crises?

Padma Desai and Pritha Mitra

9.1 Introduction

The origins of the financial crises of the 1990s around the globe have been studied
extensively. Calvo (1998); Desai (2003a, b); Krugman (1999); Corsetti, Pesenti, and
Roubini (2001); Cespedes, Chang, and Velasco (2000); Rodrik and Velasco (1999);
and Kaminsky and Reinhart (1999) trace them to an environment of stable ex-
change rates, high interest rates, and free, cross-border capital flows in emerging
markets, which prompted foreign lending to their businesses and governments.1
The poorly regulated financial markets of these economies not only led to exces-
sive borrowing from abroad but also to a double mismatch in their borrowing
pattern. The double mismatch consisted of short-term borrowing by banks from
abroad and their long-term lending at home, coupled with their accumulated
debt liabilities (largely unhedged) in foreign currencies and dubious asset acquisi-
tion in domestic currencies. Latin American sovereign borrowers, among them
the governments of Argentina and Brazil, also accumulated a significant foreign
debt burden without concern for their ability to meet their repayment obligations
via export earnings.
The East Asian economies were swept in a capital-outflow-led currency and fi-
nancial crisis that began in Thailand in mid-1997 and spread to neighboring Ma-
laysia, Indonesia, and South Korea. Currencies tumbled at varying rates in Russia
in August 1998 and Brazil in January 1999. In unrelated developments, Argentina
faced similar turmoil leading to its sovereign debt default in December 2001. Soar-
ing interest rates, which were calculated to arrest capital flight and stabilize ex-
change rates, plunged these economies into varying levels of recession.
Following the crisis, the affected country policy makers adopted a triple frame-
work of free capital mobility, floating exchange rates, and high interest rates
aimed at restoring investor confidence. Despite similar policies, their econo-
mies recovered at varying speed. The rapid recovery of the East Asian group
218 Padma Desai and Pritha Mitra

contrasted with the slow pace in the Latin American set. Why is it that in East
Asia, the postcrisis interest rates almost never surpassed 25 percent, whereas
in Latin America they were as high as 91 percent? The dynamics underlying the
relatively fast recovery of some countries can provide new insights into the design
of appropriate post-crisis policy agendas.
However, postcrisis recovery has not been subjected to a rigorous analysis with
a view to drawing policy lessons from the exercise. The relevant literature is pre-
dominantly empirical. For example, Charoenseang and Manakit (2002); Claessens,
Klingebiel, and Laeven (2001); and Koo and Kiser (2001) analyze the role of the
private sector, the reform of corporate governance, and the importance of policy
changes that can improve interaction between banks and corporations in crisis-
prone economies. Park and Lee (2001) come up with a comparative perspective
by affirming that the post-1999 revival of East Asian countries was faster than
could be predicted from previous episodes of crisis elsewhere.2 By contrast, Calvo
(2003) initiates a distinct theoretical departure in financial crisis analysis by mod-
eling the importance of fiscal status and institutions in the growth of emerging
markets, and by highlighting the role of dysfunctional domestic policies and the
resulting financial vulnerabilities in amplifying minor shocks into major turmoil.
Christiano, Gust, and Roldos (2003) also provide a theoretical analysis of post-
crisis recovery. They examine the effects of an interest rate cut in a postcrisis econ-
omy with collateral constraints.3
Our purpose is to explore the largely unexamined topic of the varying recovery
rates of crisis-affected economies by designing a simple model that can be sub-
jected to simulations. Evidently the precrisis health of countries’ macroeconomic
fundamentals—among them balanced government budgets reflecting their fiscal
status, high national saving rates, and strong export performance—can facilitate
a quick recovery of investment potential and output growth in the postcrisis
phase. We will compare the impact of varying these three precrisis macroeco-
nomic fundamentals, one at a time, on the recovery paths of two contrasting per-
formers, Thailand in East Asia and Argentina in Latin America.
We develop an open-economy macroeconomic model that we then use for sim-
ulating the quarterly interest and exchange rates of Argentina for the period from
the third quarter of 2000 to the third quarter of 2003. We refer to this approxima-
tion of the actual Argentine postcrisis profile as the base case. We then simulate
another Argentine scenario, called case 1, by altering a single macroeconomic fun-
damental from the Argentine to the Thai value in the precrisis years, holding all
other parameters and variables constant. The impact of changing a given macro-
economic fundamental for Argentina—say, the saving rate—is then assessed by
comparing the simulated results of each case 1 with the actual values of four indi-
cators for Thailand. More specifically, if Argentina were to be endowed with the
Why Do Some Countries Recover More Readily from Financial Crises? 219

high saving rate of Thailand in the precrisis years, how will its exchange rate de-
cline, interest rate hike, inflation, and GDP growth rates in the recovery phase
compare with the actual magnitudes in Thailand? We then undertake similar sim-
ulations by imposing the Thai fiscal status and export growth performance (pre-
cisely defined later) on precrisis Argentina for assessing its postcrisis outcomes
with respect to the four indicators.
We present, in section 2, the contrasting patterns of the three macroeconomic
fundamentals of relevance to our analysis for Indonesia, Malaysia, South Korea,
and Thailand in East Asia and Argentina and Brazil in Latin America for their re-
spective precrisis years from 1985 to 2003. We then set out our theoretical model
in section 3, discuss our simulation procedure in section 4, and provide our un-
derlying data in section 5. Our simulation results are presented in section 6. We
draw policy conclusions and suggest ideas for further research in section 7.

9.2 Contrasting Macroeconomic Fundamentals: East Asia and Latin America

The speed of recovery in our model will depend on the status of three macroeco-
nomic fundamentals at the onset of financial turmoil. These are balanced or sur-
plus government budgets, high overall saving in the economy, and solid export
performance.4
The relatively healthy fiscal condition of the East Asian Four created the poten-
tial for them to pay back their external debt despite significant decline of their cur-
rencies, and to extend funds to their corporate sector in the midst of a financial
crunch. Their governments were also better poised to extend unemployment ben-
efits to the jobless during the downturn. By contrast, the shaky fiscal health of the
Latin American Two had an opposite effect. They lacked the resources to repay
their external debt, assist their private sector recovery, or extend unemployment
benefits to the unemployed. The contrasting fiscal status of the two groups in the
years prior to the emergence of the crisis—from 1991–1996 for the East Asian
Four, and from 1991–2000 for the Latin American Two (1999 for Brazil)—is
brought out in figures 9.1 and 9.2. Government budgets in the East Asian Four
during 1991–1996 in figure 9.1 were in surplus in most years, whereas they were
negative throughout the period 1991–2000 for Argentina, and hit 10 percent of
GDP in 1999 for Brazil.
Next, high saving rates have a mediating effect on interest rates during a crisis.
High overall saving in the economy implies that more domestic funds may be
available for lending, and thus for supporting investment even during the crisis
phase. Consequently interest rates may shoot up less, and the contraction of in-
vestment and GDP in the postcrisis period may be moderated. Thus saving rates
in the East Asian Four ranged from 25 to 35 percent or higher of GDP (figure 9.3),
220 Padma Desai and Pritha Mitra

Figure 9.1
Government Budget Balance of East Asian Countries. Source: Economist Intelligence Unit.

Figure 9.2
Government Budget Balance of Argentina and Brazil. Source: Economist Intelligence Unit. Data prior to
1994 for Argentina and prior to 1995 for Brazil are not available.
Why Do Some Countries Recover More Readily from Financial Crises? 221

Figure 9.3
Saving Rates of East Asia Countries. Source: Economist Intelligence Unit. Aggregate national saving by
the public and private sectors as a percent of nominal GDP.

whereas the saving rate was stagnant at 15 percent for Argentina and declined to
that level in 1999 for Brazil from a high of 20 percent in 1991 (figure 9.4).
Finally, a strong export sector would provide the East Asian group with the ca-
pability to generate the much-needed foreign exchange in the postcrisis period.
Although a few companies went bankrupt because of their inability to repay for-
eign debt, the survivors benefited from the increased foreign demand resulting
from the lower exchange rate. Moreover, companies that were unable to meet
their external financial obligations had highly developed manufacturing infra-
structure. The depreciated exchange rates made them attractive pickings for for-
eign investors. These factors contributed to a rapid revival of foreign investor
confidence in the East Asian Four. By contrast, the uneven and generally feeble
performance of Argentine and Brazilian export sectors was inadequate to attract
foreign creditors, in the process requiring significantly higher interest rates for the
purpose. According to the contrasting export performance of the two sets of coun-
tries in figure 9.5, the precrisis average annual growth of Thai exports in dollars
was three times that of Argentina in its precrisis years.
The model, which we present in the next section, is designed to assess the im-
pact of these three macroeconomic features on the interest and exchange rates,
and therefore on the inflation and GDP growth rates, in the postcrisis recovery
phase of Argentina in the context of our simulation design.
222 Padma Desai and Pritha Mitra

Figure 9.4
Saving Rates of Argentina and Brazil. Source: Economist Intelligence Unit. Aggregate national saving
by the public and private sectors as a percent of nominal GDP.

Figure 9.5
Total Annual Exports (f.o.b.) in U.S. Dollars. Source: Economist Intelligence Unit.
Why Do Some Countries Recover More Readily from Financial Crises? 223

9.3 The Model

We begin with the classic Dornbusch (1976) exchange rate over-shooting frame-
work with its three components of the asset market, the money market, and the
goods market. The Dornbusch model, however, does not account for investor
expectations, which are crucial in our framework. The Dornbusch model simply
represents the expected depreciation in the exchange rate as a function of its devi-
ation from its equilibrium value. The only shock in the model is a monetary one.
Our innovation in the familiar Dornbusch model is precisely in the asset market
component. Rather than relying on a monetary shock, we introduce exogenous
shocks in the form of investor expectations that affect the economy as it revives
from crisis. The expected depreciation of the exchange rate is reconstructed to re-
flect investor expectations, which, in turn, are a function of the three macroeco-
nomic fundamentals listed earlier. We begin with the asset market.

9.3.1 The Asset Market

r t ¼ r
þ xt ð9:1Þ

xt ¼ yðet  et Þ þ ðet þ eÞð1 þ FÞ ð9:2Þ

where F represents the macroeconomic fundamental that affects investor expecta-

tions. It is defined as
F ¼ expfwððtax  gÞ þ z þ ð1  gÞÞg:

The domestic interest rate, rt , in equation 9.1 is determined by the world inter-
est rate, r
, and the expected rate of exchange rate depreciation, xt . This expected
rate of exchange rate depreciation has two components in equation 9.2. The first
component is the difference between the long-run equilibrium exchange rate and
the current exchange rate, where et is the logarithm of the long-run equilibrium
exchange rate, et is the logarithm of the current exchange rate, and y is an adjust-
ment factor.
The second component of the expected exchange rate depreciation, our innova-
tion in the Dornbusch model, is investor expectations. Investor expectations about
the depreciation of the exchange rate are based on numerous factors such as the
current political and economic climate of a country. For example, if some compa-
nies in an emerging market default on their external debt, investors will expect
other investors to withdraw their funds from this market. This results in a self-
fulfilling speculation against the emerging market currency.
224 Padma Desai and Pritha Mitra

In equation 9.2 of our model, investor speculation against the currency is repre-
sented by et þ e, where et is a transitory component and e is a permanent compo-
nent.5 The more confidence investors have in the macroeconomic fundamentals
of the economy, the lower the effect of speculation on the expected depreciation.
F represents the macroeconomic fundamentals that affect investor expectations in
our model. F has three components, ðtax  gÞ, z, and ð1  gÞ.
• ðtax  gÞ represents the precrisis government budget balance.6
•z represents the precrisis export sector strength. This value is represented by the
annual percentage growth of real exports of goods and services relative to the an-
nual percentage growth of real GDP. The growth rates are averaged over three
years prior to the onset of the crisis.7,8
• ð1  gÞ represents the precrisis national saving rates.9

9.3.2 The Money Market

m  pt ¼ lrt þ fyt ð9:3Þ

The money market equilibrium sets money demand—the right-hand side of equa-
tion 9.3—equal to money supply—the left-hand side of equation 9.3. Here m, pt ,
and yt are the natural logarithms of nominal money, the price level, and real
GDP, respectively.

9.3.3 The Goods Market

yt ¼ u þ dðet  pt Þ þ gyt þ ag  btax  srt ð9:4Þ

p_ ¼ pðyt  yt Þ ð9:5Þ

The goods market equilibrium sets aggregate demand (the right-hand side of
equation 9.4) equal to aggregate supply (the left-hand side of equation 9.4). Here
aggregate demand is represented by a shift factor, u (which includes the exoge-
nous and constant demand for exports10); the relative price of domestic goods
and services, et  pt ; income, yt ; government expenditures relative to revenue,
ag  btax; and interest rates, rt . The consumers’ marginal propensity to consume
in the aggregate demand equation, g, determines the saving rate, 1  g. In equa-
tion 9.5, the rate of increase in prices, p_ , is proportional to the deviation of GDP,
yt , from potential GDP, yt .
Why Do Some Countries Recover More Readily from Financial Crises? 225

9.3.4 Equilibrium

In equilibrium, the interest rate, GDP, and exchange rate must adjust such that the
asset, money, and goods markets are simultaneously in equilibrium. The resulting
equilibrium conditions are:
yt  yt ¼ oð pt  pt Þ ð9:6Þ

et  et ¼ ½ð1  fmdÞ=Dð pt  pt Þ ð9:7Þ

p_ ¼ poð pt  pt Þ ð9:8Þ

where
½mðd þ ysÞ þ mdyl
o¼
D

1
m¼
ð1  gÞ

D ¼ fmðd þ ysÞ þ yl

Following rational expectations, the rate at which the current exchange rate
adjusts to the long-run equilibrium exchange rate is equal to the rate at which the
exchange rates actually adjust. That is, y ¼ po.

9.4 Model Simulations

Initially, we assume that the economy in our model is in equilibrium and investor
speculation against the currency is absent. That is, e0 ¼ e0 ¼ 0. Then the economy
is hit with an external shock such that investors now speculate against the cur-
rency. That is, et > 0, e > 0. The speculation of investors initially affects the
expected exchange rate depreciation, which in turn affects the interest rate. The
goods and money markets are affected through the interest rate. The exchange
rate, GDP, and prices then adjust accordingly.
As time passes, investors speculate less against the currency as the currency
adjusts to its new long-run value. Consequently, the speculation of investors
against the currency is modeled as an exponentially decreasing shock. et ¼
expfktg, k < 0, where t represents time.11
The precrisis values of the macroeconomic fundamentals affecting investor
expectations, F, either moderate or amplify the effects of investor speculation.
When the macroeconomic fundamental variable is strong, investors have more
226 Padma Desai and Pritha Mitra

confidence in the economy, and thus F acts to reduce the effects of speculation.
The opposite occurs when the macroeconomic fundamental is weak.
As already noted, the purpose of our exercise is to assess the impact of precrisis
macroeconomic fundamentals on an economy experiencing an exogenous shock
that induces investors to speculate against the currency. Having selected a spe-
cific macroeconomic fundamental—for example, the government’s precrisis fiscal
status—we use our model to simulate the Argentine economy in its recovery
phase.
In the first step, the model simulation is calibrated to approximately match the
actual interest rate and exchange rate profiles of Argentina. We refer to this simu-
lation as the base case.12 Next, the Argentine government’s precrisis fiscal status is
changed to that of Thailand. All other variables and parameters are kept at their
base case values. We apply the same shock as the one in the base case. In other
words, referring to equation 9.2, et and e remain the same as in the base case. This
simulated model is referred to as case 1.
Finally, the impact of the precrisis government fiscal status is assessed by com-
paring the case 1 simulation results of the interest rate, the exchange rate, the GDP
growth rate, and inflation with the actual values for Thailand. If the case 1 simu-
lated Argentine values are a reasonable match to the actual values for Thailand,
then we can suggest that had the precrisis fiscal status of Argentina been more
like that of Thailand, Argentina’s postcrisis interest rate, exchange rate, GDP
growth rate, and inflation would have been closer to those of Thailand in the post-
crisis period.
We repeat these steps in two additional simulations by bringing in export sector
strength and saving rate in place of government fiscal status. Thus, in the second
simulation, we replace the Argentine precrisis export sector strength with that of
Thailand. In the third and final simulation, the Argentine precrisis saving rate is
replaced with that of Thailand.

9.5 The Data

We apply three sets of data in our model: the actual data series for Argentina and
Thailand, parameter values, and values of macroeconomic fundamentals. The
actual data series includes quarterly data of interest rates, exchange rates, GDP,
and price levels for each country. Details of these data are presented in table 9.1.
We adopt previously estimated parametric values for our money-market and
goods-market parameters. These estimates and their sources are reported in table
9.2. Some of the parameters are calibrated such that the simulated time series of
interest rates and exchange rates matches the actual data for Argentina and Thai-
land. For example, ut , which represents a shift factor in the aggregate demand
Why Do Some Countries Recover More Readily from Financial Crises? 227

Table 9.1
Actual Series
Series Symbol Year(s)

Thailand
Real GDP Y 1997,QI–2003,QI
Lending interest rate (%) p_ 1997,QI–2003,QI
Consumer prices (% average change per annum) r 1997,QI–2003,QI
Exchange rate baht:US$ (period average) E 1997,QI–2003,QI
Argentina
Real GDP Y 2000,QIII–2003,QIII
Lending interest rate (%) p_ 2000,QIII–2003,QIII
Consumer prices (% average change per annum) r 2000,QIII–2003,QIII
Exchange rate peso:US$ (period average) E 2000,QIII–2003,QIII
Note: Source for all is economist intelligence, unit country data. Frequency for all is quarterly.

equation, is adjusted such that the simulated base case time series matches the
actual time series for Argentina. Finally, the values of the precrisis macroeco-
nomic fundamentals of government expenditures and revenues, export sector
strength, and saving rates, and their sources for each country, are presented in
table 9.3.
The data in table 9.4 bring out the striking contrast between the crisis impact on
the economies of Argentina and of Thailand. At its peak, the Argentine interest
rate, at 90.61 percent in 2002, QIII (table 9.4, row 3, column 2), was almost six
times its level of 15.25 percent in Thailand in 1998, QI and QII (table 9.4, row 3,
column 3). The high Argentine interest rate reflects the abysmally low investor
confidence in its recovery prospects.
The difference in crisis severity is also reflected in the exchange rate deprecia-
tion. Prior to the crisis, Argentine and Thai exchange rates were fixed to the dol-
lar, the former more strictly than the latter. When they were allowed to float in
the postcrisis phase, the peso’s maximum plunge was 258.47 percent in 2002, QIII
(table 9.4, row 6, column 2) in contrast to the baht’s maximum decline of 82.10
percent in 1998, QI (table 9.4, row 6, column 3), almost three times less.
The postcrisis GDP growth rates of Argentina and Thailand are also reported in
table 9.4. The two growth rates are similar in their pattern and magnitude al-
though, at its lowest, Argentina’s GDP growth rate in 2002, QI is slightly more
negative at 15.23 percent (table 9.4, row 9, column 2) than Thailand’s in 1998,
QIII at 13.92 percent (table 9.4, row 9, column 3). Argentina’s GDP growth rate,
however, continues to be negative for a longer stretch after the crisis, representing
a more painful recovery.
228 Padma Desai and Pritha Mitra

Table 9.2
Parameter Estimates
Parameter Value Interpretation Source of Estimate

Goods market
d 0.11 Sensitivity of aggregate demand to relative Ghosh and Masson
prices of domestic goods and services (1991)
gArgentina 0.85 Proportion of income spent on purchases of World Development
goods and services (approximately the Indicators
inverse of the saving rate)
gThailand 0.7 Proportion of income spent on purchases of World Development
goods and services (approximately the Indicators, Baharumshah,
inverse of the saving rate) Thanoon, and Rashid
(2003)
s 0.15 Sensitivity of aggregate demand to interest Ghosh and Masson
rates (1991)
a 0.3 Sensitivity of aggregate demand to Calibrated*
government expenditures
b 0.2 Sensitivity of aggregate demand to taxes Calibrated*
p 0.2 Proportion representing the constant Calibrated*
relationship between changes in output and
changes in prices
u 1.472 Shift parameter Calibrated*
yArgentina 69 billion Initial long-run equilibrium output Economist Intelligence
pesos Unit Country Data,
Quarterly Real GDP 2000
average
yThailand 779 billion Initial long-run equilibrium output Economist Intelligence
bahts Unit Country Data,
Quarterly Real GDP 1996
average
Money market
j 1 Sensitivity of real money demand to real Chowdhury (1997),
income Dekle and Pradhan
(1999)
l 0.04 Sensitivity of real money demand to interest Dekle and Pradhan
rates (1999)
MThailand 3726 billion Money supply Economist Intelligence
bahts Unit Country Data, M2
Money Supply, 1996,QIV
MArgentina 91 billion Money supply Economist Intelligence
pesos Unit Country Data, M2
Money Supply, 2000,QIV
Asset market
r* 8% World interest rate World Development
Indicators, approximated
by the average of 1995–
2000 US prime rate
Why Do Some Countries Recover More Readily from Financial Crises? 229

Table 9.2
(continued)
Parameter Value Interpretation Source of Estimate
y 0.2447 Factor for adjustment of current exchange Endogenously
rate to long-run exchange rate value Determined**
w 0.033 Sensitivity of investor speculation to pre- Calibrated*
crisis macroeconomic fundamental
Shock equation
k 0.07 Shock propagation factor Calibrated*
e 0.01 Permanent component of shock Calibrated*
e0 0.178 Initial value of transitory component of Calibrated*
shock
* These parameter values are adjusted so that the simulated Argentine time series match the actual Ar-
gentine time series.
** This parameter value is determined endogenously within the model (refer to section 3).

Finally, inflation soared higher in Argentina than in Thailand in the postcrisis

phase. At its height in 2002, QIV, the rate was 40.31 percent (table 9.4, row 13, col-
umn 2), in contrast to Thailand’s 10.35 percent in 1998, QII (row 13, column 3).
The combination of high inflation, exchange rate depreciation, and soaring in-
terest rate, along with the persistence of a negative GDP growth rate, made
Argentina’s postcrisis recovery more arduous than Thailand’s.

9.6 Simulation Results

In our simulations, the adoption of Thailand’s precrisis macroeconomic funda-

mentals for Argentina should result in an easier recovery path for Argentina,
resembling that of Thailand in terms of interest rate, exchange rate depreciation,
GDP growth, and inflation. Recall that our base case simulates the Argentine
economy from 2000, QIII to 2003, QIII. Each version of the model alters the base
case simulation by changing the value of a precrisis macroeconomic fundamental
from its Argentine value to its Thai value. The resulting simulation is compared to
the actual performance of the recovering Thai economy. A closer match brings out
the effectiveness of the particular precrisis macroeconomic fundamental.
The simulation results are summarized in tables 9.5–9.8, one each for interest
rate, exchange rate depreciation, GDP growth rate, and inflation rate, and pre-
sented in figures 9.6–9.9. We notice similar patterns in these variables across the
simulations: interest rates shoot up, exchange rates decline, GDP growth rates
tumble, and inflation rates go up after the shock and subsequently adjust to take
their equilibrium values. However, the simulation version with the precrisis ex-
port sector strength gives significantly lower values of interest rate, exchange rate
Table 9.3
Macroeconomic Fundamentals
Variable Value Interpretation Source of estimate
GArgentina 15 billion Government expenditures, Economist Intelligence Unit
pesos quarterly data Country Data, Quarterly Real
Government Expenditures,
1999 average
GThailand 137 billion Government expenditures, Economist Intelligence Unit
bahts quarterly data Country Data, Quarterly Real
Government Expenditures,
1996 average
TArgentina 14 billion Government revenues, Economist Intelligence Unit
pesos quarterly data Country Data, Quarterly Real
Government Expenditures,
1999 average
TThailand 142 billion Government revenues, Economist Intelligence Unit
bahts quarterly data Country Data, Quarterly Real
Government Expenditures,
1996 average
Version One: 36.55 4.02/0.11—Annual per- World Development
ZArgentina centage growth of real exports Indicators
of goods and services/
annual percentage growth of
real GDP. Averaged over 3
years prior to the crisis (1998,
1999, 2000)*.
Version One: 1.00 8.06/8.04—Annual percentage World Development
ZThailand growth of real exports of Indicators
goods and services/annual
percentage growth of real
GDP. Averaged over 3 years
prior to the crisis (1994, 1995,
1996).
Version Two: 1.50 1/Total debt service—Exports World Development
ZArgentina of goods and services/Total Indicators
debt payment. Averaged over
3 years prior to the crisis
(1998, 1999, 2000)*.
Version Two: 8.00 1/Total debt service—Exports World Development
ZThailand of goods and services/Total Indicators
debt payment. Averaged over
3 years prior to the crisis
(1994, 1995, 1996).
1  gArgentina 0.15 Saving rate ¼ 1  Proportion World Development
of income spent on purchases Indicators
of goods and services (1999)
1  gThailand 0.30 Saving rate ¼ 1  Proportion World Development
of income spent on purchases Indicators, Baharumshah,
of goods and services (1996) Thanoon, and Rashid (2003)
* The year 2000 is considered a year prior to the crisis since the crisis began at the end of the year 2000.
Why Do Some Countries Recover More Readily from Financial Crises? 231

Table 9.4
Actual Rates of Interest, Exchange Rate Depreciation, GDP Growth Rate, and Inflation Rates for Argen-
tina and Thailand
1 2 3 4
Actual Actual
value for value for Difference
1 Variable* Argentina % Thailand % %

2 Precrisis interest rate 9.66 13.00 3.34

3 Maximum postcrisis interest rate 90.61 15.25 75.36
4 Equilibrium postcrisis interest rate 13.90 6.50 7.40
5 Precrisis exchange rate depreciation 0.00 2.40 2.40
6 Maximum postcrisis depreciation 258.47 82.10 176.37
7 Equilibrium postcrisis depreciation 20.91 2.16 18.75
8 Precrisis GDP growth rate 0.33 1.00 1.33
9 Maximum negative postcrisis GDP growth rate 15.23 13.92 1.31
10 Maximum positive postcrisis GDP growth rate 8.62 8.41 0.21
11 Equilibrium postcrisis GDP growth rate 8.62 6.73 1.89
12 Precrisis inflation rate 0.78 4.37 5.15
13 Maximum postcrisis inflation rate 40.31 10.35 29.96
14 Equilibrium postcrisis inflation rate 5.20 1.97 3.24
* All variables are measured on a quarterly basis. 1. The precrisis rates of all variables refer to their
values in the quarter preceding 2000,QIV for Argentina and 1997,QII for Thailand. 2. The maximum
postcrisis rates represent the highest values of the variables in the postcrisis period. 3. The equilibrium
postcrisis rates represent their final and stable values in the postcrisis period after all factors have fully
adjusted in the simulation. The actual equilibrium values refer to 2003,QIII for Argentina (this was
the most recent value available at the time this analysis was done) and 2003,QI for Thailand. 4. The
Argentine peso was allowed to float in 2002,QI. The baht was allowed to float in 1997,QIII.

depreciation, GDP growth rate decline, and inflation rate than the alternatives.
This version outperforms the simulations in which the precrisis macroeconomic
fundamentals represent the government’s fiscal status and the economy’s saving
rate. We now discuss the details of our simulations, beginning with interest rates.

9.6.1 Interest Rates

In figure 9.6, the base case simulation tracks the actual Argentine interest rate rel-
atively well. The maximum postcrisis interest rate for Argentina was 90.61 percent
(table 9.5, row 2, column 3). The base case simulation, emulating the Argentine
economy, attains a similar maximum postcrisis interest rate of 90.37 percent (table
9.5, row 5, column 3). In contrast, the actual Thai maximum postcrisis interest
rate, 15.25 percent (table 9.5, row 4, column 3), was 75.36 basis points lower than
the corresponding Argentine rate.
Figure 9.6
Interest Rate Response to Shock. Source: The actual interest rates for Thailand and Argentina are lend-
ing rates put together from the Economist Intelligence Unit (EIU). EIU Source: IMP, International Fi-
nancial Statistics.
Notes:
1. The financial crisis began in Argentina in the fourth quarter of 2000 (2000, QIV) and hit Thailand in
the second quarter of 1997 (1997, QII). In the figure, the quarter preceding the shock (Q0) corresponds
to 2000, QIII for Argentina, and 1997, QI for Thailand.
2. The interest rate paths generated by the three simulations of the base case, of Argentina with Thai fis-
cal status, and of Argentina with Thai saving rate almost perfectly overlap, creating only one visible
solid line in the figure. Recall that the base case represents Argentine values generated by our model.
This overlap reflects the weakness of the precrisis fiscal status and saving rate in affecting the postcrisis
interest rate. In other words, even if Argentina had the precrisis fiscal status or saving rate of Thailand,
the postcrisis interest rate of Argentina would match that in the base case in which Argentina retains its
own precrisis fiscal status and saving rate.
3. The interest rate path generated by the simulation of Argentina with Thai export strength produces
postcrisis interest rates that are significantly lower than those resulting from the base case simulation. In
the context of our model, this result implies that the postcrisis interest rates in Argentina would have
been much lower, closer to the postcrisis interest rates of Thailand, if Argentina had the precrisis export
sector strength of Thailand.
4. We undertake the following steps for deriving the interest rates in the figure. Solving the differential
equation for prices (equation 9.8), we obtain the path of prices in the quarters after the shock. Substitut-
ing prices in equations 9.6 and 9.7, we solve for GDP and the exchange rate, respectively. Substituting
the exchange rate in equations 9.1 and 9.2, we obtain the interest rates. Details of the calculations for
solving each series are in the appendix. The solutions are generated in Matlab.
Why Do Some Countries Recover More Readily from Financial Crises? 233

Table 9.5
Interest Rate Analysis
1 2 3 4
Maximum Equilibrium
Precrisis postcrisis postcrisis
interest interest interest
1 Variable* rate % rate % rate %

2 Actual value for Argentina 9.66 90.61 13.90

3 Actual value for Thailand 13.00 15.25 6.50
4 Difference 3.34 75.36 7.40
5 Base case** 8.00 90.37 12.38
6 Scenario 1—Replace Argentine fiscal status with that of Thailand
7 Case 1*** 8.00 90.15 12.37
8 Difference ¼ base case  case 1 0.00 0.22 0.01
9 Scenario 2—Replace Argentine export sector strength with that of Thailand
10 Case 1*** 8.00 44.94 9.97
11 Difference ¼ base case  case 1 0.00 45.43 2.41
12 Scenario 3—Replace Argentine saving rate with that of Thailand
13 Case 1*** 8.00 90.06 12.37
14 Difference ¼ base case  case 1 0.00 0.31 0.01
* All variables are measured on a quarterly basis. **B ase case values refer to simulated Argentine
values. *** Case1 values are simulated Argentine values which result when its precrisis macroeconomic
fundamental is replaced with a precrisis Thai value. 1. The precrisis rates prevailed in the quarter pre-
ceding 2000,QIV for Argentina and 1997,QII for Thailand. 2. The maximum postcrisis rates represent
their highest values in the postcrisis period. 3. The equilibrium postcrisis rates represent their final and
stable values in the postcrisis period after all factors have fully adjusted in the simulation. These actual
values are 2003,QIII value for Argentina (this was the most recent value available at the time this anal-
ysis was done) and 2003,QI value for Thailand. 4. GDP growth simulations assume a relatively stable
GDP level with a zero GDP growth rate for the quarter prior to crisis onset. 5. The Argentine peso was
allowed to float in 2002,QI, the baht in 1997,QIII.

When Argentina’s precrisis fiscal status is changed to that of Thailand, the max-
imum Argentine postcrisis interest rate drops by only 0.22 basis points (table 9.5,
row 8, column 3). By contrast, if the precrisis government fiscal status played a
critical role in postcrisis recovery, the maximum postcrisis interest rate in the
simulation would have dropped by a number closer to 75.36 basis points (the dif-
ference between the actual Argentine and Thai interest rates, table 9.5, row 4, col-
umn 3). Despite a robust precrisis fiscal status closer to Thailand’s, the crisis
would have landed Argentina in exorbitant interest rates.
However, with the imposition of the Thai export sector strength on precrisis
Argentina, the maximum postcrisis interest rate in Argentina is reduced by 45.43
basis points (table 9.5, row 11, column 3). In figure 9.6, this version brings the
234 Padma Desai and Pritha Mitra

Table 9.6
Exchange Rate Depreciation Analysis
1 2 3 4
Maximum Equilibrium
postcrisis postcrisis
Precrisis depreciation depreciation
depreciation of exchange of exchange
1 Variable* % rate % rate %
2 Actual value for Argentina 0.00 258.47 20.91
3 Actual value for Thailand 2.40 82.10 2.16
4 Difference 2.40 176.37 18.75
5 Base case** 0.00 216.71 0.00
6 Scenario 1—Replace Argentine fiscal status with that of Thailand
7 Case 1*** 0.00 215.73 0.00
8 Difference ¼ base case  case 1 0.00 0.98 0.00
9 Scenario 2—Replace Argentine export sector strength with that of Thailand
10 Case 1*** 0.00 67.83 0.00
11 Difference ¼ base case  case 1 0.00 148.88 0.00
12 Scenario 3—Replace Argentine saving rate with that of Thailand
13 Case 1*** 0.00 211.32 0.00
14 Difference ¼ base case  case 1 0.00 5.39 0.00
Notes: See notes for table 9.5.

simulated value of the Argentine interest rate remarkably close to the actual Thai
value, in contrast to the simulation versions employing Thailand’s precrisis fiscal
status and saving rate (to be analyzed later). If Argentina had the precrisis export
sector strength of Thailand, its interest rate would not have soared to 90 percent.
It would have capped around a more manageable 45 percent.
Finally, the augmentation of the precrisis saving rate of Argentina to that of
Thailand pulls down the maximum postcrisis interest rate of the simulated Argen-
tine economy by a minuscule 0.31 basis points (table 9.5, row 14, column 3). The
postcrisis interest rate in Argentina turns out to be insensitive to the precrisis sav-
ing rate.

9.6.2 Exchange Rate Depreciation

The maximum depreciation of the actual Argentine exchange rate was 258.47 per-
cent in 2002, QIII (table 9.6, row 2, column 3). The exchange rate in our base case
simulation attains a maximum depreciation of 216.71 percent (table 9.6, row 5,
column 3). By contrast, the maximum postcrisis depreciation of the Thai baht was
Why Do Some Countries Recover More Readily from Financial Crises? 235

Figure 9.7
Depreciation of Exchange Rates in Response to Shock. Source: Actual baht and peso exchange rates are
from Economist Intelligence Unity (EIU). EIU Source: IMF, International Financial Statistics.
Notes:
1. In the years preceding the onset of the crisis and for some time after that, the Argentine peso and the
Thai baht were pegged to the U.S. dollar, the former under a regime resembling a currency board, and
the latter under a managed exchange rate arrangement. The peso was allowed to float in 2002, QI, and
the baht in 1997, QIII. In the figure, Q0 corresponds to the quarter preceding the currency float 2001,
QIV for Argentina and 1997, QII for Thailand.
2. The depreciation path generated by the simulations of the base case and of Argentina with Thai fiscal
status almost perfectly overlap, creating only one visible solid line in the figure. This result reflects the
weakness of the fiscal status in affecting postcrisis exchange rate depreciation. Even if Argentina had
the precrisis robust fiscal status of Thailand, the postcrisis exchange rate depreciation of Argentina
would remain the same as in the base case in which Argentina retains its own precrisis weak fiscal
status.
3. The exchange rate depreciation path generated by the simulation of Argentina with Thai saving rate
is only slightly lower than the base case simulation path. In other words, even if Argentina had the pre-
crisis high saving rate of Thailand, the postcrisis exchange rate depreciation of the peso would be min-
imally reduced.
4. The depreciation path generated by the simulation of Argentina with Thai export sector strength is
significantly lower than the base case simulation path and the paths generated by the simulations of Ar-
gentina with Thai fiscal status and Argentina with Thai saving rate. This result reflects the strength of
the precrisis export sector in positively affecting postcrisis depreciation.
5. The simulation steps are stated in note 4 of figure 9.6.
236 Padma Desai and Pritha Mitra

relatively modest at 82.10 percent (table 9.6, row 3, column 3), 176.3 percent lower
than that of the peso. Figure 9.7 brings out this striking contrast.
As with the interest rate, the peso exchange rate is relatively insensitive to the
Argentine government’s precrisis fiscal status. When it is replaced by the precrisis
fiscal status of Thailand, the maximum depreciation of the peso is reduced by
only 0.98 percent (table 9.6, row 8, column 3). If the government’s precrisis fiscal
status were a critical factor in affecting its level, the massive actual decline of the
peso, postshock, would have been reduced substantially by 176.37 percent, reach-
ing the baht’s postcrisis maximum decline of 82.10 percent (table 9.6, row 4, col-
umn 3).
By contrast, a change in the precrisis export sector strength of Argentina to that
of Thailand significantly reduces the peso’s depreciation in the postcrisis period in
figure 9.7. The maximum depreciation is reduced by as much as 148.88 percent
(table 9.6, row 11, column 3), closer to the 176.37 percent difference between the
actual peso–baht maximum tumble (table 9.6, row 4, column 3).
Finally, the impact of the precrisis saving rate on the postcrisis peso exchange
rate decline is similar to the impact of the precrisis fiscal status of the government:
the peso’s depreciation is insensitive to the precrisis saving rate in the simulation.
The maximum postcrisis depreciation of the peso is contained by only 5.39 per-
cent (table 9.6, row 14, column 3).
The simulation results suggest substantially moderate postshock interest and
exchange rate movements, associated with a strong precrisis export sector
strength. As a result, Argentina’s recovery path could possibly have been less pro-
longed and painful.

9.6.3 GDP Growth Rates

As noted earlier, our base case simulations understate the magnitudes of Argenti-
na’s GDP growth and inflation rates while tracking their trends.
In figure 9.8, during the postcrisis period, actual GDP in Argentina and Thai-
land experiences a significant decline, with negative growth rates, in the initial
postcrisis quarters. As the economies recover, both GDPs move up. Thus the
actual GDP growth rates of the two economies follow a similar path. The maxi-
mum negative postcrisis actual GDP growth of Argentina is only 1.31 percent
lower than that of Thailand (table 9.7, row 4, column 3).
Our simulations suggest that if Argentina had the precrisis fiscal status of Thai-
land, its maximum negative GDP growth would have matched it in the base case
simulation in which Argentina retains its own precrisis fiscal status (table 9.7, row
8, column 3). Essentially, the precrisis fiscal status has no effect on Argentine GDP
growth as it recovers. Similarly, the postcrisis Argentine economy is insensitive to
Why Do Some Countries Recover More Readily from Financial Crises? 237

Table 9.7
GDP Growth Rate Analysis
1 2 3 4 5
Maximum Maximum
negative positive Equilibrium
Precrisis postcrisis postcrisis postcrisis
GDP GDP GDP GDP
1 Variable* growth % growth % growth % growth %
2 Actual value for Argentina 0.33 15.23 8.62 8.62
3 Actual value for Thailand 1.00 13.92 8.41 6.73
4 Difference 1.33 1.31 0.21 1.89
5 Base case** 0.00 1.25 0.47 0.00
6 Scenario 1—Replace Argentine fiscal status with that of Thailand
7 Case 1*** 0.00 1.25 0.47 0.00
8 Difference ¼ Base case  case 1 0.00 0.00 0.00 0.00
9 Scenario 2—Replace Argentine export sector strength with that of Thailand
10 Case 1*** 0.00 1.06 0.43 0.00
11 Difference ¼ Base case  case 1 0.00 0.19 0.04 0.00
12 Scenario 3—Replace Argentine saving rate with that of Thailand
13 Case 1*** 0.00 1.25 0.47 0.00
14 Difference ¼ Base case  case 1 0.00 0.00 0.00 0.00
Notes: See notes for table 9.5.

the precrisis saving rate. Changing the precrisis saving rate of Argentina to that of
Thailand results in zero change in the postcrisis GDP growth path (table 9.7, row
14, column 3, also figure 9.8).
However, when we replace the precrisis export sector strength of Argentina
with that of Thailand in our simulated Argentine economy, Argentina’s maxi-
mum negative GDP growth rate is pulled up by 0.19 percent (table 9.7, row 11,
column 3). The actual maximum negative postcrisis GDP growth of Argentina is
1.31 lower than in Thailand (table 9.7, row 4, column 3). In figure 9.8, the attribu-
tion of the precrisis export sector strength of Thailand to Argentina is most effec-
tive in reducing fluctuations in GDP growth rate in recovering Argentina.

9.6.4 Inflation

The maximum postcrisis inflation rate in Thailand is 29.96 percent lower than in
Argentina (table 9.8, row 4, column 3). As before, the potentially large impact of a
precrisis macroeconomic fundamental in mediating investor expectations, and
thus reducing the severity of the crisis, would result in a significant reduction of
238 Padma Desai and Pritha Mitra

Figure 9.8
GDP Growth Rates in Response to Shock. Source: Actual GDP growth rates for Argentina from 2000,
QIII to 2003, QIII are from Economist Intelligence Unity (EIU). EIU data are taken from Ministerio de
Economia y Obras Y Servicios Publico. Actual GDP Growth Rates for Thailand from 1997, QI to 2003,
QI are from Economist Intelligence Unity (EIU). EIU data are taken from the National Economic and
Social Development Board of Thailand.
Notes:
1. The financial crisis hit Argentina in 2000, QIV and Thailand in 1997, QII. In the figure, the quarter
preceding the shock, Q0 corresponds to 2000, QIII for Argentina, and 1997, QI for Thailand.
2. The GDP growth rate paths generated by the simulations of the base case, of Argentina with Thai fis-
cal status, and of Argentina with Thai saving rate almost perfectly overlap, creating only one visible
solid line in the figure. This result reflects the weakness of the fiscal status and saving rate in affecting
the postcrisis GDP growth rate. Even if Argentina had the robust precrisis fiscal status or saving rate of
Thailand, the postcrisis growth rate of Argentina would remain the same as in the base case, in which
Argentina retains its own precrisis fiscal status and saving rate.
3. The GDP growth path generated by the simulation of Argentina with Thai export sector strength
fluctuates less than the GDP growth path simulated by the base case. In the context of our model, this
may be interpreted to mean that Argentina’s postcrisis GDP growth rate would have fluctuated less if
its precrisis export sector were as strong as that of Thailand.
4. The simulation steps are stated in note 4 of figure 9.6.
Why Do Some Countries Recover More Readily from Financial Crises? 239

Table 9.8
Inflation Rate Analysis
1 2 3 4
Maximum Equilibrium
Precrisis postcrisis postcrisis
inflation inflation inflation
1 Variable* rate % rate % rate %

2 Actual value for Argentina 0.78 40.31 5.20

3 Actual value for Thailand 4.37 10.35 1.97
4 Difference 5.15 29.96 3.24
5 Base case** 0.00 4.32 0.00
6 Scenario 1—Replace Argentine fiscal status with that of Thailand
7 Case 1*** 0.00 4.32 0.00
8 Difference ¼ Base case  case 1 0.00 0.01 0.00
9 Scenario 2—Replace Argentine export sector strength with that of Thailand
10 Case 1*** 0.00 2.44 0.00
11 Difference ¼ Base case  case 1 0.00 1.88 0.00
12 Scenario 3—Replace Argentine saving rate with that of Thailand
13 Case 1*** 0.00 4.31 0.00
14 Difference ¼ Base case  case 1 0.00 0.01 0.00
Notes: See notes for table 9.5.

the inflation rate in recovering Argentina if its precrisis macroeconomic funda-

mentals were to assume Thai values.
In our simulations, Argentina’s assumption of Thai fiscal status and saving rate
lowers the maximum postcrisis inflation rate of Argentina by a mere 0.01 percent
in both cases (table 9.8, row 8, column 3 and row 14, column 3). The Argentine in-
flation rate is insensitive to its precrisis fiscal status and saving rate. By contrast,
Argentina’s maximum inflation rate in the recovery phase drops by 1.88 percent
when it takes the precrisis export sector strength of Thailand (table 9.8, row 11,
column 3, also figure 9.9).

9.7 Conclusions

We analyze the postcrisis performance of two crisis-affected economies, Argentina

and Thailand, with vastly different precrisis macroeconomic fundamentals, by
applying a simple macroeconomic model. Their precrisis macroeconomic funda-
mentals, among them their governments’ fiscal status, their national saving rates,
and their export sector strength, can be expected to affect postcrisis recovery dif-
ferently. In fact, our model simulations suggest that the precrisis difference in
240 Padma Desai and Pritha Mitra

Figure 9.9
Inflation Rates in Response to Shock. Source: Actual GDP growth rates for Argentina from 2000, QIII to
2003, QIII are from Economist Intelligence Unity (EIU). EIU data are taken from Instituto Nacional de
Estadistica y Censo. Actual GDP growth rates for Thailand from 1997, QI to 2003, QI are from Econo-
mist Intelligence Unity (EIU). EIU data are taken from the IMF: International Financial Statistics.
Notes:
1. The financial crisis hit Argentina in 2000, QIV and Thailand in 1997, QII. In the figure, the quarter
preceding the shock, Q0, corresponds to 2000, QIII for Argentina and 1997, QI for Thailand.
2. The inflation paths generated by the simulations of the base case, of Argentina with Thai fiscal status,
and of Argentina with Thai saving rate almost perfectly overlap, creating only one visible solid line in
the figure. This result reflects the weaknesses of the fiscal status and the saving rate in affecting the
postcrisis inflation rate. Even if Argentina had the robust precrisis fiscal status or saving rate of Thai-
land, the postcrisis inflation rate of Argentina would remain the same as in the base case, in which Ar-
gentina retains its own weak, precrisis fiscal status and saving rate.
3. The inflation path generated by the simulation of Argentina with Thai export strength is notably
lower than the inflation path generated by the base case simulation. In other words, if Argentina had
the precrisis export strength of Thailand, its postcrisis inflation rates would have been significantly
lower.
4. The simulation steps are stated in note 4 of figure 9.6.
Why Do Some Countries Recover More Readily from Financial Crises? 241

export sector strength between Argentina and Thailand (acting through investor
expectations) is large enough to explain most of the difference in postcrisis interest
rate and exchange rate movements between the two countries. The GDP growth
and inflation rate results are quantitatively less strong; however, the precrisis
difference in export sector strength is able to explain some of the postcrisis differ-
ence in these two variables as well. In other words, if Argentina had the precrisis
export sector strength of Thailand, Argentina’s postcrisis recovery path could
have been closer to that of Thailand. On the other hand, a strong fiscal status
and a high national saving rate in the precrisis years could not have spared the
Argentine economy from high interest rates and substantial peso depreciation,
leading in turn to substantial GDP decline and high inflation.
The export earning capacity of an economy reflects its debt repayment potential
by generating foreign exchange earnings and restoring investor confidence. The
contrasting results suggest that investors focus essentially on an economy’s ability
to generate foreign exchange and repay its external debts. However, our model
lacks microeconomic foundations that could adequately capture the interaction
among consumers, producers, foreign investors, and the government. We plan to
design and empirically test such a model with appropriate microeconomic under-
pinnings. Despite this shortcoming, we believe that our simulation results linking
superior postcrisis recovery to precrisis export strength are eminently credible.
Our analysis examines the impact of actual, ex post, precrisis macroeconomic
fundamentals on postcrisis recovery. We are aware that a variety of other factors
played a role in the Thai and Argentine recoveries. One factor that future analysts
may want to introduce in the model is the expectation of postcrisis macroeco-
nomic and financial fundamentals both during and just prior to the crisis. Note
that these expectations tend to be related to the precrisis macroeconomic funda-
mentals. In this sense, our analysis indirectly accounts for them by explicitly
including the macroeconomic impact of the precrisis fundamentals in our model.
On the other hand, expectations of future macroeconomic and financial funda-
mentals will influence forward-looking expectations of servicing sovereign debt,
which in turn will affect the country risk premium, leading to higher domestic in-
terest rates. From this perspective, a more detailed model will require the intro-
duction of the country risk premium.

9.8 Appendix: Derivation of Model

The model is defined by the following five equations:

Asset Market

r t ¼ r
þ xt ð9:A1Þ
242 Padma Desai and Pritha Mitra

xt ¼ yðet  et Þ þ ðet þ eÞð1 þ FÞ ð9:A2Þ

f ¼ expfwððtax  gÞ þ z þ ð1  gÞÞg
Money Market

m  pt ¼ lrt þ fyt ð9:A3Þ

The Goods Market

yt ¼ u þ dðet  pt Þ þ gyt þ ag  btax  srt ð9:A4Þ

p_ ¼ pðyt  yt Þ ð9:A5Þ

We take as given (that is, exogenous) values for all parameters and m, r
, g, tax,
ðet þ eÞ, and yt .

Step 1: Goods and Asset Market Equilibrium

For a given shock, ðet þ eÞ, the economy converges to equilibrium where

y t ¼ yt ; et ¼ et ; pt ¼ pt ) r ¼ r
þ ðe þ eÞð1 þ FÞ:
Substituting equations 9.A1 and 9.A2 into equation 9.A4, we get the equilibrium
value for yt :

yt ¼ m½u þ dðet  pt Þ þ ag  btax  sr
 sðet þ eÞð1 þ FÞ ð9:A6Þ

1
m¼ :
1g
Subtracting equation 9.A6 from 9.A4, we get
yt  yt ¼ m½ðd þ syÞðet  et Þ  mdð pt  pt Þ: ð9:A7Þ
Step 2: Money and Asset Market Equilibrium
For a given shock, ðet þ eÞ, the economy converges to equilibrium where

y t ¼ yt ; et ¼ et ; pt ¼ pt ) r ¼ r
þ ðe þ eÞð1 þ FÞ:
Substituting equations 9.A1 and 9.A2 into equation 9.A3, we get

pt ¼ m þ lr
þ lyðet  et Þ þ lðet þ eÞð1 þ FÞ  fyt : ð9:A8Þ

The equilibrium value for pt in 9.A8 is given by

pt ¼ m þ lr
þ lðet þ eÞð1 þ FÞ  fyt : ð9:A9Þ

Subtracting equation 9.A9 from 9.A8, we get
pt  pt ¼ lyðet  et Þ  fð yt  yt Þ: ð9:A10Þ
Why Do Some Countries Recover More Readily from Financial Crises? 243

Step 3: Combining Goods, Asset, and Money Market Equilibrium

Substituting ðet  et Þ from equation 9.A10 into equation 9.A7, we derive
ð yt  yt Þ ¼ oðpt  pt Þ ð9:A11Þ
o ¼ ½mðd þ ysÞ þ mdyl=D
D ¼ fmðd þ ysÞ þ yl:
Rational expectations dictate that
y ¼ po:
Substituting equation 9.A11 into equation 9.A10, we derive
ð1  fmdÞ
et  et ¼  ð pt  pt Þ: ð9:A12Þ
fmðd þ syÞ þ ly
Substituting 9.A11 into 9.A5, we get
p_ ¼ poð pt  pt Þ: ð9:A13Þ
Step 4: Defining Equilibrium Exchange Rate and Prices
From equation 9.A9, we have

pt ¼ m þ lr
þ lðet þ eÞð1 þ FÞ  fyt : ð9:A14Þ

Substituting equation 9.A6 into 9.9, we get
½ð1  mdfÞ yt  mu þ mdm þ mðdl þ sÞr

et ¼
md
½mðdl þ sÞðet þ eÞð1 þ FÞ  mag þ mbtax
þ : ð9:A15Þ
md
Step 5: Solving the Variables
We take as given (that is, exogenous) values for m, r
, g, tax, ðet þ eÞ, and yt . Sub-
stituting these values into equations 9.A14 and 9.A15, we have

m; r
; g; tax; ðet þ eÞ; yt ) pt ; and et :

We can then solve for the differential equation from 9.A13:
pt ; p_t ) pt :
Using equation 9.A11, we can solve for yt :
pt ; pt ) y t :
Using equation 9.A12, we can solve for et :
p t ; e t ; pt ) e t :
Using equations 9.A1 and 9.A2, we can solve for rt :

r
; et ; et ; ðet þ eÞ ) rt :
244 Padma Desai and Pritha Mitra

Step 6: Simulation
1. Initially we assume e0 ¼ e0 ¼ 0 and we have m, r
, y0 . Applying step 5, we
solve for the initial values of all the variables.
2. A shock hits the economy: et ¼ expfktg, k < 0, et b 0.
3. We define yt ¼ y0  f ðet þ eÞ. Therefore, for each time, t, et ) yt .
4. Next, for each time, t, we have m, r
, ðet þ eÞ, and yt . Applying step 5, we solve
for the time, t, values of all the variables.
5. We repeat parts 3 and 4 of the simulation process until et ¼ 0, thereby reaching
the long-run equilibrium. The permanent component of the shock has perma-
nently reduced the equilibrium value of yt , permanently depreciated et , and per-
manently increased the price level pt .
6. In our simulations, we define the base case as the scenario in which the values
of F, the macroeconomic fundamental that affects investor expectations, reflect
the Argentine values of fiscal status, export sector strength, and saving rate13
½ðtax  gÞ; z; ð1  gÞ. In the alternative cases, the simulation described above is
performed by changing the value of one macroeconomic fundamental: for exam-
ple, z is changed to its Thai value.14

Notes

1. The literature on financial-crisis related issues such as capital account controls, current account defi-
cits, contagion transmission, currency boards, exchange rate regimes, and interest rate policies for
emerging markets, is vast and varied. Caballero and Krishnamurthy (2001), Calvo and Reinhart (2000),
Lahiri and Vegh (2001), Calvo and Mishkin (2003), Kaminsky, Reinhart, and Vegh (2003) are among
the noteworthy references.
2. Park and Lee differ from previous studies by explicitly contrasting the recovery paths of the East
Asian economies from numerous past crises (of as many as 95 previous episodes during the period
from 1970 to 1995). They adopt cross-country regression analysis for the purpose.
3. In the authors’ modeling, foreign borrowing must be collateralized by physical assets. In other
words, the value of physical capital in foreign currency in each period must be greater than or equal to
the value of foreign debt, short-term and long-term as well.
4. Details are in Desai (2003b).
5. The speculation term has transitory and permanent components to account for the fact that a finan-
cial crisis occasionally results in permanent GDP decline. Cerra and Saxena (2003) find evidence of per-
manent losses in GDP levels in the postcrisis phase of East Asian economies. In our model, this implies
that long-term GDP is defined as yt ¼ y0  f ðet þ eÞ. Setting e ¼ 0, and defining y ¼ yt ¼ y0 , we can
also simulate a model without permanent GDP losses.

6. The fiscal data employed in our simulations and their sources are reported in table 9.3. For Thai-
land, the precrisis values for government budget balance correspond to 1996 average quarterly real
government expenditures less revenues. For Argentina, the precrisis values for government budget bal-
ance correspond to 1999 average quarterly real government expenditures less revenues.
7. The export and GDP data employed in our simulations and their sources are reported in table 9.3.
For Thailand, the precrisis values for the export sector strength correspond to the annual percentage
Why Do Some Countries Recover More Readily from Financial Crises? 245

growth of real exports of goods and services relative to the annual percentage growth of real GDP,
averaged over 1994, 1995, and 1996. For Argentina, the precrisis values for the export sector strength
correspond to the annual percentage growth of real exports of goods and services relative to the annual
percentage growth of real GDP, averaged over 1998, 1999, and 2000.
8. A version of the model where export sector strength is modeled as export earnings relative to exter-
nal debt is also analyzed. Our preferred model in which we represent export sector strength via export
growth relative to GDP growth gives stronger results than the alternative in which export sector
strength is modeled as export earnings relative to external debt. We therefore analyze in depth the
results of our preferred model.
9. The saving rate data and their sources are reported in table 9.3. For Thailand, the precrisis saving
rate corresponds to the 1996 annual saving rate. For Argentina, the precrisis saving rate corresponds
to the 1999 annual rate.

10. We assume for simplicity that demand for exports is exogenous. The country being modeled is
assumed to be a small open economy. We assume that there is always a foreign country that demands
its exports. So if the economy produces export goods, these will immediately be purchased by a foreign
country.
11. After the economy is hit with an external shock, the permanent component of the shock, e b 0, re-
mains constant.
12. Our base case simulations of Argentine GDP and inflation track their trends rather than their actual
magnitudes.
13. Argentina: The precrisis fiscal status is measured as the 1999 average quarterly government expen-
ditures less revenue. The precrisis export sector strength is measured as the average 1997–1999 annual
percentage growth of real exports divided by the annual percentage growth of real GDP. The precrisis
saving rate is measured as the 1999 annual saving rate. Details are given in table 9.2.
14. Thailand: The precrisis fiscal status is measured as the 1996 average quarterly government expendi-
tures less revenue. The precrisis export sector strength is measured as the average 1994–1996 annual
percentage growth of real exports divided by the annual percentage growth of real GDP. The precrisis
saving rate is measured as the 1996 annual saving rate. Details are given in table 9.2.

References

Baharumshah, Ahmad Z., Marwan A. Thanoon, and Salim Rashid. 2003. ‘‘Savings Dynamics in the
Asian Countries.’’ Journal of Asian Economics 13: 827–845.
Caballero, Ricardo J., and Arvind Krishnamurthy. 2001. ‘‘A ‘Vertical’ Analysis of Crises and Interven-
tion: Fear of Floating and Ex-ante Problems.’’ Working Paper No. 8428, NBER, Cambridge, MA.
Calvo, Guillermo A. 1998. ‘‘Capital Flows and Capital-Market Crises: The Simple Economics of Sudden
Stops.’’ Journal of Applied Economics 1, no. 1: 35–54.
Calvo, Guillermo A. 2003. ‘‘Explaining Sudden Stops, Growth Collapse and BOP Crises: The Case of
Distortionary Output Taxes.’’ Working Paper No. 9864, NBER, Cambridge, MA.

Calvo, Guillermo A., and Frederic S. Mishkin. 2003. ‘‘The Mirage of Exchange Rate Regimes for Emerg-
ing Market Countries.’’ Working Paper No. 9808, NBER, Cambridge, MA.
Calvo, Guillermo A., and Carmen M. Reinhart. 2000. ‘‘When Capital Inflows Come to a Sudden Stop:
Consequences and Policy Options.’’ In Reforming of the International Monetary and Financial System, eds.
Peter Kenen and Alexandre Swoboda, 175–201. Washington, D.C.: International Monetary Fund.
246 Padma Desai and Pritha Mitra

Cerra, Valerie, and Sweta C. Saxena. 2003. ‘‘Did Output Recover from the Asian Crisis?’’ Working
Paper No. 03/48, IMF, Washington, D.C.
Cespedes, Luis F., Roberto Chang, and Andres Velasco. 2000. ‘‘Balance Sheets and Exchange Rate Pol-
icy.’’ Working Paper No. 7840, NBER, Cambridge, MA.
Charoenseang, June, and Pornkamol Manakit. 2002. ‘‘Financial Crisis and Restructuring in Thailand.’’
Journal of Asian Economics 13: 597–613.
Chowdhury, Abdur R. 1997. ‘‘The Financial Structure and the Demand for Money in Thailand.’’ Applied
Economics 29: 401–409.
Christiano, Lawrence J., Christopher Gust, and Jorge Roldos. 2003. ‘‘Monetary Policy in a Financial Cri-
sis.’’ Journal of Economic Theory 119, no. 1: 64–103.
Claessens, Stijn, Daniela Klingebiel, and Luc Laeven. 2001. ‘‘Financial Restructuring in Systemic Crises:
What Policies to Pursue?’’ Paper presented at NBER Conference on Management of Currency Crises,
Monterey, March 28–31.

Corsetti, Giancarlo, Paolo Pesenti, and Nouriel Roubini. 2001. ‘‘Fundamental Determinants of the
Asian Crisis: The Role of Financial Fragility and External Imbalances.’’ In Regional and Global Capital
Flows: Macroeconomic Causes and Consequences, eds. Takatoshi Ito and Anne O. Krueger, 11–41. Chicago:
University of Chicago Press.
Dekle, Robert, and Mahmood Pradhan. 1999. ‘‘Financial Liberalization and Money Demand in the
ASEAN Countries.’’ International Journal of Finance and Economics 4: 205–215.
Desai, Padma. 2003a. ‘‘Explorations in Light of Financial Turbulence from Asia to Argentina.’’ Paper
presented at Conference on the Future of Globalization: Explorations in Light of Recent Turbulence,
Yale Center for the Study of Globalization, Yale University, October 10–11.

Desai, Padma. 2003b. Financial Crisis, Contagion, and Containment: From Asia to Argentina. New Jersey:
Princeton University Press.
Dornbusch, Rudiger. 1976. ‘‘Expectations and Exchange Rate Dynamics.’’ Journal of Political Economy 84,
no. 6: 1161–1176.
Ghosh, Atish, and Paul R. Masson. 1991. ‘‘Model Uncertainty, Learning and the Gains from Coordina-
tion.’’ American Economic Review 81, no. 3: 465–479.
Kaminsky, Graciela L., and Carmen M. Reinhart. 1999. ‘‘The Twin Crises: The Causes of Banking and
Balance-of-Payments Problems.’’ American Economic Review 89, no. 3: 473–500.
Kaminsky, Graciela L., Carmen M. Reinhart, and Carlos A. Vegh. 2003. ‘‘The Unholy Trinity of Finan-
cial Contagion.’’ Working Paper No. 10061, NBER, Cambridge, MA.
Koo, Jahyeong, and Sherry L. Kiser. 2001. ‘‘Recovery from a Financial Crisis: The Case of South Korea.’’
Economic and Financial Policy Review, Federal Reserve Bank of Dallas 4: 24–36.
Krugman, Paul. 1999. The Return of Depression Economics. New York: W.W. Norton & Company.
Lahiri, Amartya, and Carlos A. Vegh. 2001. ‘‘Living with the Fear of Floating: An Optimal Policy Per-
spective.’’ Working Paper No. 8391, NBER, Cambridge, MA.
Park, Yung C., and Jong W. Lee. 2001. ‘‘Recovery and Sustainability in East Asia.’’ Paper presented at
Conference on Management of Currency Crises, NBER, Cambridge, MA, March 28–31.
Rodrik, Dani, and Andres Velasco. 1999. ‘‘Short-Term Capital Flows.’’ Working Paper No. 7364, NBER,
Cambridge, MA.
IV Debt, Taxation, and Reforms
10 Government Debt: A Key Role
in Financial Intermediation

Michael Kumhof and Evan Tanner

10.1 Introduction

Optimally, governments should finance their expenditures such that losses from
distortionary taxation are minimized. Many authors have emphasized that such
losses might be substantially reduced through the use of state-contingent capital
levies on government debt—in bad states of nature, the government defaults out-
right and/or engineers a debt devaluation through a price level increase.1 How-
ever, real-world policy debates are not typically cast in such terms. First, bonds
with explicit state-contingent returns are rarely available to governments. Second,
default or debt devaluation is typically considered to be a policy of last resort, and
certainly not a desirable way to balance the budget.2 For example, according to
the International Monetary Fund’s (2003) recent ‘‘stress test’’ approach to fiscal
policy, a fiscal adjustment should be large enough to preempt inflation, default,
or additional adjustment in the future.
Why do policy makers think of debt devaluation so differently from the aca-
demic literature? As economists, we typically address such questions by consider-
ing various market imperfections. The most popular approach in the literature is
the sticky price friction, as in Siu (2004), Schmitt-Grohe and Uribe (2004), and
Angeletos (2003). Another is a cash-in-advance constraint on consumption, as in
Nicolini (1998) and Dı́az-Giménez, Giovanetti, Marimon, and Teles (2003). But
among policy makers, the reason that is often advanced is the damage that a debt
devaluation would do to the domestic financial system. We therefore hypothesize
that the difference between academic theory and policy reality is due to several
features of financial markets that are absent from existing models, and that are
particularly pronounced in developing countries.
Most important, financial markets suffer from problems associated with asym-
metric information,3 and legal or institutional imperfections can make it pro-
hibitively costly and time-consuming to take security interests in real estate or
movable property that could help to overcome these asymmetries. In this case
250 Michael Kumhof and Evan Tanner

fiscal policy, and especially predictable debt management that provides a safe
asset, can be crucial in supporting at least some intermediation, thereby helping
overcome barriers between borrowers and lenders. In essence, safe government
debt facilitates financial intermediation by serving a collateral-like function. In
developing countries, as we will show, this is reflected in the presence of large
amounts of government debt on bank balance sheets. In some cases, government
debt is also used as explicit collateral in repurchase agreements—a transaction
requiring that government debt be safe. And finally, as is stressed in a large litera-
ture on the development of emerging bond markets, government debt plays an
important role as a benchmark for private-sector bond markets, which are key to
successful overall financial development.
Consider how all this could be incorporated into a simple theoretical model.4
Assume that the government issues nominal debt and levies a proportional tax
on labor income to finance a given expenditure flow. Producers need to borrow
capital from an intermediary, who requires that a certain minimum fraction of
the loan must be covered by government debt as collateral. Then fiscal policy cre-
ates two distortions. First, the positive labor tax rate implies a suboptimal work
effort. Second, the collateral requirement implies a suboptimal level of capital.
The first distortion usually makes a highly volatile inflation rate desirable in order
to engineer state-contingent real returns on debt. But the second distortion does
the opposite, because an inflation-induced erosion of the public debt stock would
affect the economy’s ability to intermediate capital. A welfare-maximizing gov-
ernment therefore faces a trade-off. And that trade-off may suggest why defaults
and debt devaluations are so rarely used in practice.5
International Monetary Fund (2004) illustrates the practical importance of this
reasoning. This study finds that in each of the four recent government debt
restructurings it examines, the key consideration in delaying a restructuring for
as long as possible has been fear of the resulting damage to domestic banks, espe-
cially given their potential to spread and amplify the negative effects of a restruc-
turing throughout the economy.
This line of thinking can draw on several pieces in Guillermo Calvo’s inspiring
body of work. A key element of Calvo (1988) is that surprise inflation is not cost-
less, thereby giving rise to a trade-off between taxation and inflation. The paper
implements this by way of an exogenous cost function for inflation. And Calvo
and Vegh (1990, 1995) assume that bonds enter the utility function. While the ap-
plication of that technology in those papers was different, this holds the key to a
more formal modeling of costly unanticipated inflation, such as the approach sug-
gested in the previous paragraph.
In this paper, we provide evidence relevant to the outlined discussion. We
show that the type of debt that is important for the effect of default and debt de-
Government Debt 251

valuation on domestic financial systems, domestically held debt, is of greater and

growing importance than external debt for almost every single emerging mar-
ket. This in itself is interesting, given that most of the interest in the context of
emerging-market debt has focused on external debt, to which very different eco-
nomic criteria apply in a default scenario.
Next, to address the question of government debt in financial intermediation,
we show that in most of these countries banks hold an extremely high proportion
of their assets in government debt. To relate this finding to financial market
imperfections, we show that there is a positive relationship between such hold-
ings and a comprehensive index of legal and institutional imperfections, contain-
ing measures of difficulties in contract enforcement, in registering property, and
in obtaining credit. Finally, we provide detailed evidence that suggests that the
high bank holdings of government debt are no longer due to financial repression
in all but a handful of the larger developing countries—they are indeed the result
of a choice to hold the safe asset.

10.2 Government Debt in Developing Countries

10.2.1 Domestic versus External Debt

The optimal fiscal policy literature has mainly used closed-economy models, and
the analysis has mainly been applied to large industrialized countries such as the
United States. On the other hand, when it comes to government debt in develop-
ing countries, the literature has used open-economy models, and has been almost
exclusively concerned with external or sovereign debt. External debt default is
clearly considered costly in practice,6 but in the literature that cost is not generally
assumed to include direct negative effects on the domestic financial system. The
question arises as to whether this exclusive preoccupation with external debt is
still justified in today’s developing countries. As we will now show, with very
few exceptions—including, most notably, Argentina—the answer to that question
is no.
Our main data source for this exercise is the Bank for International Settlements
(BIS) (2003), which aggregates comprehensive data on individual securities issues
obtained from financial markets sources.7 As explained in the data appendix, the
split of the BIS data into domestic and international securities is very conservative
in what it classifies as a domestic security. The only securities classified as domes-
tic are issues by residents, targeted at resident investors, in domestic currency. As
the BIS data set excludes Brady bonds, we merge it with a JP Morgan data set on
outstanding Bradies.
252 Michael Kumhof and Evan Tanner

The results are presented in figures 10.1–10.6, which plot quarterly data from
1994, Q1 through 2003, Q4. For each group of countries we first present the do-
mestic share of marketable government debt and then the ratio of marketable
government debt to GDP. We first show various groups of developing countries,
and then industrialized countries for comparison.
The broad trends for the marketable debt-to-GDP ratios for industrialized and
developing countries are quite different. Most industrialized countries started
with relatively high ratios (around 50 percent but with wide deviations in either
direction) that have not exhibited a long-term growth trend, a recent slight in-
crease having been preceded by a decline in the 1990s. The well-known exception
is Japan. Developing countries typically had much lower marketable debt-to-GDP
ratios of around 20 percent in the mid-1990s, but with few exceptions they have
since exhibited a positive growth trend. More importantly, the growth trend in
the domestic part of that debt has been even stronger, as figures 10.1a, 10.2a, and
10.3a illustrate. In almost all major developing countries the share of domestic
debt has risen and now represents in excess of 70 percent of overall debt.8 This is
remarkable in view of the conservative criteria used for classifying debt as domes-
tic. There is no clear trend in industrialized countries (see figures 10.4–10.6). The
only major exceptions to this trend are countries that have experienced severe
financial crises accompanied by large devaluations and consequent reductions in
the real value of domestic debt. But the only cases where this has kept the domes-
tic debt share well below 50 percent until the present day are Russia and Argen-
tina.9 Mexico (in 1995) and Thailand (in 1997) also experienced such episodes but
have since returned to domestic debt shares of over 70 percent. In these four cases
the large devaluations that eroded nominal domestic debt were accompanied by
severe banking crises and output contractions, with recoveries taking several
years (see Kaminsky and Reinhart 1999). This provides ample evidence to justify
our interest in domestically held debt in developing countries.10
As a natural extension of the above, we also briefly examine the cyclical proper-
ties of domestic debt. Does domestic debt act as a countercyclical ‘‘shock
absorber,’’ a consequence of the positive relationship between tax revenue and
GDP, or of countercyclical government expenditures? Or are governments able to
borrow more and therefore run larger deficits during cyclical upswings, consis-
tent with findings for developing countries by Kaminsky, Reinhart, and Végh
(2004)?
Table 10.1 presents results from a simple bivariate regression of the cyclical
components of the domestic debt-to-GDP ratio on real economic activity. Domes-
tic debt moves procyclically (countercyclically) as the slope coefficient ðb1 Þ is
greater than (less than) zero.
Government Debt 253

Figure 10.1
Latin America (a) Domestic share of marketable government debt, (b) Marketable government debt to
GDP ratios
254 Michael Kumhof and Evan Tanner

Figure 10.2
Asia (a) Domestic share of marketable government debt, (b) Marketable government debt to GDP
ratios
Government Debt 255

Figure 10.3
Other developing countries (a) Domestic share of marketable government debt, (b) Marketable govern-
ment debt to GDP ratios
256 Michael Kumhof and Evan Tanner

Figure 10.4
Small industrialized countries (a) Domestic share of marketable government debt, (b) Marketable gov-
ernment debt to GDP ratios
Government Debt 257

Figure 10.5
Scandinavia (a) Domestic share of marketable government debt, (b) Marketable government debt to
GDP ratios
258 Michael Kumhof and Evan Tanner

Figure 10.6
Large industrialized countries (a) Domestic share of marketable government debt, (b) Marketable gov-
ernment debt to GDP ratios
Government Debt 259

Table 10.1
Domestic Government Debt and the Business Cycle Estimated Regression:
[Domestic Debt/GDP]tCYC ¼ b0 þ b1 Econ ActivtytCYC þ error
Quarterly Data, 1994:3–2004:2
Industrialized Countries Emerging/Developing Countries
Country b1 t-stat Country b1 t-stat
Australia 0.21 0.56 Argentina 0.03 0.38
Austria 0.19 1.47 Brazil 0.05 0.21
Belgium
0.06
0.15 Chile 0.26 2.56
Canada
1.14
3.28 Czech Rep.
0.03
0.07
Denmark
0.18
0.30 Hungary
0.11
1.01
Finland
0.34
1.73 Malaysia 0.03 0.54
France
0.12
0.23 Mexico
0.03
0.36
Greece
0.27
0.72 Peru 0.01 0.35
Iceland
0.13
0.87 Philippines 0.34 1.56
Ireland 0.09 0.45 Poland 0.10 1.51
Italy
0.47
0.68 South Africa
0.72
1.37
Japan 0.60 2.91 Thailand 0.00 0.02
Netherlands
0.24
0.62 Turkey
0.41
2.92
New Zealand 0.52 2.07
Norway 0.05 0.44
Portugal
0.52
1.92
South Korea 0.12 3.92
Spain
0.14
0.42
Sweeden 0.34 0.63
Switzerland
0.40
1.65
United Kingdom 0.06 0.13
United States 0.05 0.20
Notes: 1. Data sources: Domestic Debt/GDP, Bank for International Settlements; Economic Activity
(Industrial Production, Real GDP), International Monetary Fund—International Financial Statistics.
2. Trend/Cycle decomposition uses Hodrick–Prescott filter.
3. For the following countries, the industrial production index is used as an indicator for quarterly eco-
nomic activity: Austria, Belgium, Brazil, Finland, Greece, Hungary, Japan, Korea, Malaysia, Mexico,
Netherlands, Norway, Poland. In all other cases, real GDP is used.
260 Michael Kumhof and Evan Tanner

In most cases, the slope coefficient is statistically indistinguishable from zero.

However, in four industrialized countries (Canada, Finland, Portugual, Switzer-
land) and one emerging country (Turkey) domestic debt appears to be counter-
cyclical (b 1 < 1, jt-statj > 1:5). By contrast, in three industrialized countries
( Japan, New Zealand, South Korea) and three emerging/developing countries
(Chile, Phillipines, Poland) domestic debt appears to move procyclically (b 1 > 1,
jt-statj > 1:5).
In the following subsections, we discuss several economic roles played by gov-
ernment debt in a developing economy. We emphasize throughout how such
roles would be diminished if the government regularly employed state-contingent
capital levies on debt. One key idea is that stable government debt management
may facilitate bank-based financial intermediation, especially in an environment
with a weak legal and institutional infrastructure. In addition to that, practitioners
emphasize that stable government debt markets are the backbone for further
development of financial markets beyond a bank-based system. Given that well-
developed financial intermediation may have positive effects on capital accumula-
tion and growth, this implies important real effects of fiscal and debt policy.

10.2.2 Financial Institutions Hold a Large Part of Their Assets in Government

Debt

Holders of government debt in developing countries include banks and pension

funds. The role of the latter is still relatively minor, with a few notable exceptions
such as Chile, as is the role of private securities markets. Financial intermediation
is therefore highly dependent on banks. As we will now show, banks hold a very
large proportion of their assets in government debt, which means that stable debt
management becomes critical for financial stability.
Figures 10.7–10.12 show the exposures of national banking systems to govern-
ment debt in developing and developed economies. We define that exposure as
the average ratio (1998–2002) of financial institutions’ net credit to the govern-
ment11 to their total assets. (See the data appendix for further detail.) In indus-
trialized countries (figures 10.11–10.12), where the fiscal situation is mostly very
robust and government debt therefore extremely safe, that ratio is typically
around 10 percent, whereas in developing countries (figures 10.7–10.10) ratios of
20 percent–40 percent are very common, and even ratios above 50 percent are
observed in some of the largest developing countries. In such an economy, a gov-
ernment that contemplates a debt devaluation is faced with the insolvency of its
banking system, because the typical bank capitalization ratio is less than 10 per-
cent of assets. This is indeed exactly what happened in Argentina, where Perry
and Serven (2002) conclude that ‘‘the roots of the [Argentinean] crisis lie in
Government Debt 261

Figure 10.7
Latin American financial institutions—credit to public sector/total assets

Figure 10.8
Asian financial institutions—credit to public sector/total assets
262 Michael Kumhof and Evan Tanner

Figure 10.9
Eastern European financial institutions—credit to public sector/total assets

Figure 10.10
Middle Eastern/African financial institutions—credit to public sector/total assets
Government Debt 263

Figure 10.11
Industrialized countries’ financial institutions—credit to public sector/total assets

Figure 10.12
Small European countries’ financial institutions—credit to public sector/total assets
264 Michael Kumhof and Evan Tanner

the . . . rigid exchange rate regime, the fragile fiscal position, and the hidden vul-
nerability of the banking system behind its strong facade.’’ During the protracted
negotiations that eventually led to Argentina’s massive default, it was a key con-
sideration that a large government default would seriously harm the banking
system.

10.2.3 The Reason for Large Government Debt Holdings: Legal and
Institutional Weaknesses

Large holdings of government debt by banks in developing countries have in the

past been the result of financial repression. However, as first stressed by Reinhart,
Rogoff, and Savastano (2003), forced holdings of government debt by captive fi-
nancial institutions are gradually disappearing and being replaced by govern-
ment debt securities issued at market interest rates.12 The evidence presented in
table 10.2 confirms that view.13 Reserve requirements no longer exist in several
key developing countries, and have been falling in others, to the point that they
are actually no longer binding. Binding reserve requirements seem to mostly af-
fect pension funds in some countries, because they are prevented from investing
abroad and have only limited domestic investment possibilities (see also World
Pension Association 2002).
The large holdings of government debt by banks must therefore be largely vol-
untary. On the other hand, as we have seen, government debt in developing
countries is in fact mostly lower than in industrialized countries. The main prob-
lem is, therefore, that banks do not lend very much, or, in other words, that im-
portant segments of an economy are unable to borrow, or can only obtain credit
with great difficulty. De Soto (2000) suggests that the culprit is poorly developed
laws and institutions.14 Potential borrowers cannot collateralize their loans with
physical assets, while lenders cannot enforce lending contracts and security provi-
sions in the courts. Fleisig (1996, 1998) quantifies some of the resulting differences
in lending behavior between industrialized and developing countries, and em-
phasizes that the major problem is the inability to use movable capital as collateral,
with mortgages being subject to fewer problems. He mentions that in the U.S. 70
percent of all loans are secured by collateral, and almost 50 percent of all credit is
secured by movable collateral. Movable capital represents 67 percent of the U.S.
industrial capital stock and 75 percent of gross business investment (Asian Devel-
opment Bank 2000). It is obviously critical for businesses to be able to use such
capital as collateral. But in Argentina, for example, they are unable to do so. Only
around 10 percent of all bank credit is secured at all, and practically none of it is
secured by movable collateral. This is because, as described by Fleisig (1996,
Government Debt 265

Table 10.2
Reserve Requirements
Country Year Description (RR ¼ reserve requirement)

Brazil 1996 RR remains over 80% on demand deposits.

1999 April: RR on demand deposits up to 100%, on other acc. claims up to 60%.
October: RR on bank deposits reduced from 65% to 45%.
2001 RR on demand deposits 45%, RR on time deposits 10%.
2004 No de jure RR (source: IMF Staff Economists).
China 2004 No de jure RR (source: IMF Staff Economists).
Czech Republic 2004 No de jure RR (source: IMF Staff Economists).
India 1990 Cash reserve ratio (CRR) 10%.
1998 CRR raised to 10.5%, lowered to 10%, raised to 11%.
1999 CRR lowered to 10.5%, then to 10%, and to 9% in November.
2000 CRR lowered to 8%.
2001 CRR lowered to 7.5%.
2004 Statutory Liquidity Ratio (SLR): RR of 25% in the form of government
bond holdings.
Indonesia 1999 No RR.
2004 No de jure RR (source: IMF Staff Economists).
Jamaica 2004 Actual holdings of government bonds far exceed RR (source: IMF Staff
Economists).
Korea 2001 RR ¼ 2.9%.
Lebanon 2004 No de jure RR (source: IMF Staff Economists).
Mexico 1991 Liquidity coefficients on foreign currency deposits up to 50% dep. on
maturities.
1995 Change from a daily-zero RR system to cumulative-zero RR over 28-day
period.
1999 RR ¼ 0%.
Philippines 1991 RR rationalized.
1993 RR for bank deposits lowered from 25% to 22%.
1994 RR lowered to 17%.
1995 RR lowered to 15%.
1997 RR lowered to 13%.
1998 RR lowered to 10%.
1999 RR ¼ 11% for commercial banks, RR ¼ 0% for rural and cooperative banks.
2002 RR ¼ 9%.
Poland 2004 No RR on banks (source: IMF Staff Economists).
Russia 2004 RR ¼ 3.5%.
Turkey 1994 Comprehensive reform of RR. RR ¼ 10.8%.
1995 RR ¼ 9.3%.
1996 RR ¼ 8% for domestic currency deposits, RR ¼ 11% for foreign currency.
2004 Actual holdings of government bonds far exceed RR (source: IMF Staff
Economists).
266 Michael Kumhof and Evan Tanner

1998), creation of security interests, especially over movable property, is difficult,

costly, and uncertain, perfection and registration is not effectively possible, and
enforcement is slow and expensive. As a result, U.S. banks relying predominantly
on movable collateral are seen as safe by both bank examiners and the public,
while Argentinean banks that do so are not. They will therefore have to curtail
the amount of such lending and seek to bolster their balance sheet through safer
investments, including holdings of government debt. Asian Development Bank
(2000) reports that when debtors can offer acceptable collateral, private creditors
offer six to eight times more credit, two to ten times longer maturity, and 30 to 50
percent lower interest rates. The effect on interest rates is also described by Fleisig
(1998), who compares loans against movable property in the United States and
Argentina. Because in Argentina movable security is considered almost irrelevant
to the recoverability of a loan, a loan for agricultural machinery (for example)
therefore has lending rates of around 60 percent. Of the enormous difference of
that rate to a corresponding U.S. lending rate, only about one-sixth is explained
by macroeconomic risk and one-tenth by the general difficulty of collecting
against security (including mortgages). A full three-quarters of the difference is
due to problems with the legal framework fr secured transactions against mov-
able property. This is a more general problem for both Latin America, as dis-
cussed in Garro (1998), and for Asia, as discussed in Asian Development Bank
(2000) and Cranston (1998). These authors find that banks generally prefer real
estate (and personal guarantees). The macroeconomic consequences of such
problems include an inefficient savings allocation, an inefficient allocation of
capital (with a preference for assets that can be used as collateral), fragile banks
highly exposed to credit risks, an inability to use securitization for risk diversifi-
cation (thereby increasing the dependence on fragile banks), and negative effects
on equity because small and start-up businesses, who do not own land, find it
almost impossible to obtain credit. All of this has negative effects on real economic
activity. Fleisig (1998) cites country studies that put the cost of defective systems
of lending security at 5 percent–10 percent of annual GNP, with the cost of
reforming such systems a small fraction of that. The major international financial
institutions have therefore in recent years paid increasing attention to these
issues, as exemplified by the European Bank for Reconstruction and Development
(Simpson and Menze, 2000),15 Asian Development Bank (2000), and World Bank
(2004).
It should be added that, in developing countries, even real estate security is
often subject to numerous problems, including cost, time delays, difficulties in
establishing priority in the absence of efficient registries, and other issues. Cran-
ston (1998) explains that in several Asian countries much of the land remains
Government Debt 267

communally held according to customary law, ruling out its use as collateral. Flei-
sig (2002) mentions that in Latin America most of the land is not titled, and exist-
ing registries are rudimentary.
Legal and institutional imperfections are therefore of critical importance for
business, but are at the same time hard to quantify. The work of La Porta, Lopez
de Silanes, Shleifer, and Vishny (1998) has therefore been extremely valuable in
creating cross-sectional country indices for the quality of law and institutions af-
fecting businesses. It is now available through World Bank (2004) in the form of
updated indicators,16 on the basis of, among others, the papers by Djankov,
McLiesh, and Shleifer (2004) and Djankov, La Porta, Lopez de Silanes, and Shlei-
fer (2003). For the purpose of this study we are interested specifically in the factors
responsible for segmented credit markets, which are represented by three factors.
The first is ‘‘getting credit,’’ describing the quality and accessibility of credit infor-
mation and the quality of collateral and bankruptcy laws. The second is ‘‘enforc-
ing contracts,’’ measuring the cost and time delays in the collection of an overdue
debt. And the third is ‘‘registering property,’’ measuring the cost and time delays
of transferring title to real estate.
We compute an overall raw index for all countries in our sample by attaching
equal weights to all three categories.17 Our final index ranges between 0 and 100,
with 0/100 for the countries with the worst/best raw index. We create another
index for the share of debt on financial institutions’ balance sheets, as discussed
earlier, with 0/100 for the lowest/highest government debt-to-total assets ratio
among all countries in the sample. Figure 10.13 plots these two indices against
each other. The results are striking. There is a very strong negative relationship
between the quality of law and institutions and the amount of government debt
that financial institutions choose to hold on their balance sheets. Almost all the
countries in the top left corner of figure 10.13 are developing countries, and al-
most all in the bottom right are industrialized (Chile does the best among devel-
oping countries).
Financial institutions with highly risky and largely unsecured lending books
are very vulnerable to macroeconomic shocks. Their high holdings of government
debt are generally an attempt to offset the credit risk inherent in private credit
with government debt, which ideally bears no credit risk. This is recognized by
the Basle rules for capital adequacy, which give a much lower risk weighting
(zero) to such debt, even if the debtor is an emerging-market government. The
safety of banks’ government debt portfolio is a precondition for even the limited
amount of private sector lending that does take place. In such countries prudent
management of public debt is critical for the health of the banking system, and us-
ing state-contingent returns on government debt would be highly damaging.
268 Michael Kumhof and Evan Tanner

Figure 10.13
Government debt on balance sheets and legal/institutional quality

10.2.4 Government Debt and Further Financial Development: Private Bond

Markets

As we have argued, prudent management of government debt is important to

safeguard fragile domestic banking systems. But financial development in the
long run benefits greatly from moving beyond a purely bank-based system to in-
clude not only a stock market, but also a long-term private bond market. And
here again, as emphasized by World Bank-IMF (2001),18 the government bond
market plays a critical role because it is the backbone of most fixed income secu-
rities markets.
Herring and Chatusripitak (2000) present the key arguments for why private
bond markets matter for financial development, stability and growth,19 and show
how legal and institutional factors have impeded the development of private
bond markets in Asia. The lack of developed bond markets in developing coun-
Government Debt 269

tries has in recent years also been a major theme for the international financial
institutions, including the Inter-American Development Bank (Castellanos 1998,
Reinstein 2002, and del Valle 2002), Asian Development Bank (2002), World
Bank-IMF (2001), and policy makers in individual countries.20 The Asian Bond
Market Initiative (Rhee 2004) is a concerted effort to develop bond markets across
that region.
Underdeveloped bond markets make the pricing of credit risks and equities
harder because of the absence of a benchmark yield curve. Derivatives markets
cannot develop, making the diversification of risk exposures harder. Fewer sav-
ings are utilized, and borrowers face higher borrowing costs and shorter matu-
rities (potentially leading to foreign borrowing and thereby resulting in foreign
exchange risks), and banks become too large relative to the overall financial sec-
tor, thereby making the economy more vulnerable to crises because of its depen-
dence on a small number of institutions.
Herring and Chatusripitak (2000) identify two major prerequisites for the devel-
opment of bond markets. The first is the legal and institutional infrastructure
mentioned in section 2.3. The second, already mentioned as a key concern in
World Bank-IMF (2001), is a deep, liquid, government bond market. As we have
seen, legal problems restrict even bank-based private sector lending, and clearly
they do the same to bonds-based lending. But while it is important to remove
those weaknesses, this is a time-consuming process, and in the meantime a stable
government bond market is all the more important, and can help overcome at
least some of the problems caused by those weaknesses. In several developing
countries that have still not tackled their legal frameworks effectively, develop-
ment of the government bond market has nevertheless started to support a pri-
vate bond market, at least to strong borrowers that are relatively less dependent
on collateral.
A government debt market does this first by putting in place a basic financial
infrastructure including laws, institutions, products, services, repo, and deriva-
tives markets, and second by playing a role as an informational benchmark. A sin-
gle private issuer of securities would never be of sufficient size to generate a
complete yield curve, and his securities would not be nominally riskless because
only the government has the power to print domestic currency. The government,
through the government debt market, can therefore provide a public good to fi-
nancial markets, but only under two further conditions. First, macroeconomic vol-
atility, especially inflation volatility, must be low so that a nominal yield curve is
informative about the real cost of borrowing. State-contingent inflation would in-
terfere with this goal, and experience confirms this—the volatility in Brazil in
1999 and in Argentina in 2001–2002 were major impediments to the further devel-
opment of their local financial markets (del Valle 2002). Second, the government
270 Michael Kumhof and Evan Tanner

must issue a sufficient volume of debt. For this latter reason, during the era of
shrinking U.S. public debt (the late 1990s and early 2000s), some observers
expressed concern that it would be more difficult to conduct monetary policy in a
smaller government debt market (Reinhart, Sack, and Heaton 2000 and Fleming
2000).21 Similarly, for developing countries Herring and Chatusripitak (2000) con-
clude that the goal of developing a robust bond market may conflict with the goal
of minimizing the cost of government borrowing if a government with fiscal sur-
pluses decides to issue government debt and invest the proceeds in foreign secu-
rities in order to provide liquidity, as Hong Kong has done in recent years.
We have seen that government debt plays two key roles. It provides infrastruc-
ture and acts as an informational benchmark in securities markets. And on the
balance sheets of financial institutions it is, to depositors, a form of security that
increases their willingness to have their funds intermediated in a generally risky
environment. As such, it informally acts as collateral. However, government debt
can also play a more direct role as collateral in wholesale securities markets. In
particular, it plays a critical role in managing risks in derivatives markets, pay-
ment and settlement systems, and in the market for repurchase agreements. Bank
for International Settlements (2001) shows that there has been an enormous in-
crease in such collateralized transactions in recent years.
Under a repurchase agreement, a market participant sells a security while
simultaneously agreeing to repurchase it in the future. Such a transaction func-
tions as a secured loan. The party purchasing the security makes funds available
to the seller while holding that security as collateral. In virtually every financial
market worldwide, the only form of security acceptable in repo transactions is
government debt, due to its low or zero risk and high liquidity. Because repo mar-
kets require quite well-developed financial markets, they generally appear at a
later stage of development. We nevertheless found sizeable repo markets in Brazil
and Mexico, using as a criterion Bankscope balance-sheet data to express the stock
of outstanding repo loans as a fraction of total assets. In Brazil that ratio is around
7 percent, and in Mexico around 5 percent in 2002–2003. Other sizeable markets
exist in the Philippines, Poland, and India.

10.3 Summary and Conclusions

Is state-contingent inflation (or default) on government debt an optimal way to

conduct fiscal policy in response to shocks? This paper makes an attempt to un-
derstand the very different replies to this question given by the theoretical litera-
ture on optimal fiscal policy on the one hand, and by experienced practitioners
on the other. Many contributions to the theoretical literature advocate what we
have referred to as debt devaluations (in response to negative shocks), while prac-
Government Debt 271

titioners only think of it as a last resort, to be avoided at almost all costs. The em-
phasis in this paper is on understanding the thinking of practitioners, particularly
decision makers in developing countries, by presenting some pertinent data. We
show that debt devaluations are likely to have very negative effects on financial
intermediation in developing countries, as their banks are highly exposed to gov-
ernment debt and at the same time face much higher risks in private sector lend-
ing because of weak legal and institutional infrastructures. Keeping government
debt a safe investment for banks is critical to support the already low financial
intermediation that does exist. It is also critical to support further improvements
in financial intermediation, away from banks and toward securities markets.
Many countries that have not followed, or have been unable to follow, this advice
have experienced serious lending crunches and recessions, and have set back the
development of their financial markets by years.
It would be beneficial to further develop the theory of optimal fiscal policy
against the background of these arguments, as this would greatly enhance its use-
fulness to applied policy analysis. In doing so, great inspiration can be drawn
from the work of Guillermo Calvo, who has made several key contributions to
this literature.

10.4 Data Appendix

Domestic versus External Debt

The data were obtained from Bank for International Settlements (2004), which
follows the following classification system for domestic versus external/inter-
national securities:

Issues by Issues by
Residents Nonresidents
In domestic currency, targeted Domestic International
at resident investors
In domestic currency, targeted International International
at nonresident investors
In foreign currency International International

Note that the classification covers the borrower side of securities issues, not the
actual ownership. This system is very conservative in what it classifies as a do-
mestic security.22 Issues by nonresidents targeted at domestic investors, whether
in domestic or foreign currency, are invariably classified as international,
272 Michael Kumhof and Evan Tanner

although many if not most holders will be domestic. The same is true for issues in
domestic currency targeted at foreign investors, because it is well known that the
ultimate holders of developing countries’ domestic currency debt are in many
cases domestic residents.23 Bank for International Settlements (2004) also states
that notes and money market instruments issued by nonresidents in domestic cur-
rency and in domestic markets are not included due to lack of data, which again
biases results against domestically held debt. The opposite bias results from an-
other missing debt class, Brady bonds, but we remedy this by merging the BIS
data set with a JP Morgan data set on outstanding Brady bonds, using the Merrill
Lynch (1994) guide for the classification of securities.

Bank Holdings of Government Debt

We define the exposure of banks to government debt as the ratio of (Financial

Institutions’ Net Credit to Government) / (Financial Institutions’ Net Total
Assets).
The percentage figures are computed from International Financial Statistics,
and are averages for 1998–2002, except for countries in the euro area, where
1999–2003 was used. The numerator is the sum of all entries representing net
credit to the public sector by deposit money banks, other banking institutions,
and nonbank financial institutions (the latter series do not exist for all countries).
The denominator is the sum of the net total assets of these three groups, after can-
celing out credit items between them. To obtain the net figures we deduct from
both numerator and denominator the sum of all entries representing credit by the
public sector to these institutions. Crisis years are excluded from the computation
of averages as follows: Argentina 2002, Indonesia 1998.

Acknowledgements

This paper was prepared for the festschrift in honor of Guillermo Calvo, held at
the International Monetary Fund on April 16, 2004. The authors would like to
thank the editor for his patience. We are grateful to Abdul Abiad and Ashoka
Mody, who generously shared their data set on financial repression with us. Ex-
cellent research assistance was provided by Nicolas Amoroso and Wolfgang
Harten. All errors are our own.

Notes

1. See Lucas and Stokey (1983), Chari, Christiano, and Kehoe (1994), and a large subsequent literature
surveyed, for example, in Chari and Kehoe (1999). More recent contributions include Angeletos (2002)
and Buera and Nicolini (2004).
Government Debt 273

2. Such a policy must be unanticipated and credibly on a one-time basis. If it does become unavoidable
to resolve a fiscal problem in this way, then achieving that credibility is a key objective of a well-
designed IMF fiscal adjustment program. The time-inconsistency problem associated with such a pol-
icy discussed by Calvo (1978) and Calvo and Guidotti (1993) is not considered in this paper.
3. A related literature employs models of segmented asset markets; for example, Alvarez, Lucas, and
Weber (2001). In such models a subset of agents is unable to trade assets and therefore to smooth con-
sumption over time and states of nature. Instead, state-dependent taxes and transfers (including the in-
flation tax, in that paper) help them do so. In a similar vein, Shin (2003) develops a model of fiscal
policy with heterogeneous agents. To us it seems that the key issue is the ability of firms to continue to
obtain financing. Studies such as Kaminsky and Reinhart (1999) have shown that (debt) devaluations
in developing countries are almost invariably accompanied by a banking collapse that causes a severe
output contraction.
4. See Kumhof (2004) for more details.
5. Depending on the parameterization of the model, it may also address another problem of the opti-
mal fiscal policy literature, which is that in many of those models only very small stocks of government
debt can be sustained in equilibrium.
6. It also involves an effect that does not arise under a domestic debt default, and that makes default
beneficial: a wealth transfer from foreigners.
7. This data set excludes nonmarketable debt. For more comprehensive measures of government debt
than the one used in this study, a disaggregation into domestic and international debt is not available
at the frequency and country coverage required.

8. The overall size of emerging local bond markets at the end of 2002 was four times the size of those
countries’ foreign currency external debt (see Mathieson et al. 2004, IMF 2004).
9. The latter is a special case because its commitment to a currency board left little incentive to develop
a domestic currency bond market.
10. Similar evidence for many smaller developing countries is presented in Cabbar and Jonasson
(2004) for South Asia, del Valle (2002) and Batlay and del Valle (2002a) for the Middle East and North
Africa, and Batlay and del Valle (2002b) for Eastern Europe and Central Asia.
11. Unlike the BIS measure used above, this includes nonmarketable government debt.
12. Caprio (1999) also discusses evidence to this effect.
13. We are extremely grateful to Abdul Abiad and Ashoka Mody for sharing with us the historic infor-
mation contained in table 10.2, which can also be found in Abiad (2004). The 2004 entries were col-
lected by contacting IMF staff economists.
14. Pagano (2001) contains similar arguments.
15. The EBRD’s Model Law on Secured Transactions of 1994 has been used to help several Eastern Eu-
ropean transition economies in the redesign of their legal systems.
16. The data set is available at www.doingbusiness.org.
17. Our qualitative results are not dependent on the details of the weighting scheme.
18. See chapter 1, page 3, table 1.1 in World Bank-IMF (2001).
19. Similar arguments are presented by Fry (1997).

20. For example, Kang, Kim and Rhee (2004) discuss Korea, and Kim (2001) discusses a number of
Asian countries in great detail.
274 Michael Kumhof and Evan Tanner

21. Not everyone shared this concern for the U.S., where financial markets are so deep that some sub-
stitutes may be found, especially the abundant issues by government-sponsored enterprises. But Her-
ring and Chatusripitak (2000) stress that the same is not true for developing countries. This is precisely
because there the historic lack of developed public debt markets has been an impediment to the growth
of private financial markets.

22. Notice that limiting the attention to securities is necessary for practical purposes, but a more com-
plete picture would also consider nonmarketable government obligations. These include on the domes-
tic side a variety of spending commitments and outstanding debt balances, and on the international
side concessional lending from governments and international financial institutions, as well as syndi-
cated bank loans.
23. Another source of bias is that local issues in foreign currency are generally classified as interna-
tional. But here exceptions are made by BIS for Argentina, Peru, and the Philippines.

References

Abiad, A. 2004. ‘‘A New Database of Financial Reforms.’’ Working Paper, IMF, Washington, D.C.
Alvarez, F., R. E. Lucas, Jr., and W. E. Weber. 2001. ‘‘Interest Rates and Inflation.’’ American Economic
Review 91, no. 2: 219–225.
Angeletos, G.-M. 2002. ‘‘Fiscal Policy with Noncontingent Debt and the Optimal Maturity Structure.’’
Quarterly Journal of Economics 117, no. 3: 1105–1131.
Angeletos, G.-M. 2003. ‘‘Optimal Fiscal and Monetary Policy.’’ NBER Macroeconomics Annual 2003,
Cambridge, MA: MIT Press.
Asian Development Bank. 2000. ‘‘Secured Transactions Law Reform in Asia: Unleashing the Power of
Collateral.’’ In Law and Policy Reform at the Asian Development Bank, Vol. II. Manila, Philippines: Asian
Development Bank.

Bank for International Settlements. 2001. ‘‘Collateral in Wholesale Financial Markets: Recent Trends,
Risk Management and Market Dynamics.’’ Publication No. 17, Committee on the Global Financial Sys-
tem, BIS, Basel, Switzerland.
Bank for International Settlements. 2003. ‘‘Guide to the International Financial Statistics.’’ Paper No. 14,
BIS, Basel, Switzerland. Available at http://www.bis.org/publ/bispap14.pdf.
Batlay, M., and C. del Valle. 2002a. ‘‘Regional Snapshot of Government Bond Markets in the Middle
East and North Africa.’’ Financial Sector Development, The World Bank. Powerpoint presentation.
Batlay, M., and C. del Valle. 2002b. ‘‘Regional Snapshot of Government Bond Markets of Eastern Eu-
rope and Central Asia.’’ Financial Sector Development, The World Bank. Powerpoint presentation.
Buera, F., and J. P. Nicolini. 2004. ‘‘Optimal Maturity of Government Debt without State Contingent
Bonds.’’ Journal of Monetary Economics 51, no. 3: 531–554.
Cabbar, D. and T. Jonasson. 2004. ‘‘Regional Snapshot of Government Bond Market Development in
South Asia Economies.’’ Paper presented at Seacen-World Bank-IMF Seminar on Strengthening the De-
velopment of Domestic Debt Securities Markets, Colombo, Srilanka, June 8–11.
Calvo, G. A. 1978. ‘‘On the Time Consistency of Optimal Policy in a Monetary Economy.’’ Econometrica
46, no. 6: 1411–1128.
———. 1988. ‘‘Servicing the Public Debt: The Role of Expectations.’’ American Economic Review 78: 647–
671.
Government Debt 275

Calvo, G. A., and P. E. Guidotti. 1993. ‘‘Optimal Maturity of Nominal Government Debt: An Infinite
Horizon Model.’’ International Economic Review 33, no. 4: 895–919.
Calvo, G. A. and C. A. Vegh. 1990. ‘‘Interest Rate Policy in a Small Open Economy: The Predetermined
Exchange Rates Case.’’ IMF Staff Papers 37, no. 4: 753–776.
———. 1995. ‘‘Fighting Inflation with High Interest Rates: The Small Open Economy Case under Flexi-
ble Prices.’’ Journal of Money, Credit and Banking 27: 49–66.
Caprio, G. Jr. 1999. ‘‘Review of S. Fry: Emancipating the Banking System and Developing Markets for
Government Debt.’’ Journal of Economic Literature 37, no. 4: 1723–1725.
Castellanos, J. 1998. ‘‘Developing Government Bond Markets.’’ Working Paper IFM-111, Inter-
American Development Bank, Washington, D.C.
Chari, V. V., L. Christiano, and P. J. Kehoe. 1994. ‘‘Optimal Fiscal Policy in a Business Cycle Model.’’
Journal of Political Economy 102, no. 4: 617–652.
Chari, V. V., and P. J. Kehoe. 1999. ‘‘Optimal Fiscal and Monetary Policy.’’ In Handbook of Macroeconom-
ics, Vol. 1C, eds. J. B. Taylor and M. Woodford, 1671–1745. Amsterdam: Elsevier.
Cranston, R. 1998. ‘‘Credit Security and Debt Recovery: Law’s Role in Reform in Asia and the Pacific.’’
In Emerging Financial Markets and Secured Transactions, International Economic Development Law Volume
6, eds. J. J. Norton and M. Andenas. Boston: Kluwer Law International.
del Valle, C. 2002. ‘‘Developing Government Securities Markets—A Framework.’’ Mimeo. Financial
Sector Development, The World Bank.
De Soto, H. 2000. The Mystery of Capital: Why Capitalism Triumphs in the West and Fails Everywhere Else.
New York: Basic Books.
Dı́az-Giménez, J., G. Giovanetti, R. Marimon, and P. Teles. 2003. ‘‘Nominal Debt as a Burden on Mone-
tary Policy.’’ Working Paper No. 8. Centre de Referència en Economia Analitica, Barcelona.
Djankov, S., C. McLiesh, and A. Shleifer. 2004. ‘‘Private Credit Around the World.’’ Working Paper, De-
partment of Economics, Harvard University.
Djankov, S., R. La Porta, R. Lopez-de-Silanes, and A. Shleifer. 2003. ‘‘Courts.’’ Quarterly Journal of Eco-
nomics 118: 453–517.
Fleisig, H. W. 1996. ‘‘Secured Transactions: The Power of Collateral.’’ Finance and Development, The
World Bank. Available at http://worldbank.orgfandd/english/0696/articles/0150696.htm.
———. 1998. ‘‘Economic Functions of Security in a Market Economy.’’ In Emerging Financial Markets
and Secured Transactions, International Economic Development Law Volume 6, eds. J. J. Norton and M.
Andenas. Boston: Kluwer Law International.
———. 2002. ‘‘Microenterprises and Collateral.’’ Issues Brief No. 2, Center for the Economic Analysis of
Law, Washington, D.C.
Fleming, M. J. 2000. ‘‘Financial Market Implications of the Federal Debt Paydown.’’ In Brookings Papers
on Economic Activity 2: 221–251.
Fry, M. J. 1997. Emancipating the Banking System and Developing Markets for Government Debt. New York:
Routledge.
Garro, A. M. 1998. ‘‘Difficulties in Obtaining Secured Lending in Latin America: Why Law Reform Re-
ally Matters.’’ In Emerging Financial Markets and Secured Transactions, International Economic Development
Law Volume 6, eds. J. J. Norton and M. Andenas. Boston: Kluwer Law International.
276 Michael Kumhof and Evan Tanner

Herring, R. J. and N. Chatusripitak. 2000. ‘‘The Case of the Missing Market: The Bond Market and Why
It Matters for Financial Development.’’ Working Paper 11, ADB Institute, Tokyo, Japan.
International Monetary Fund. 2003. ‘‘Sustainability Assessments—Review of Application and Method-
ological Refinements.’’ Policy Development and Review Department. Available at http://www.imf.org
/external/np/pdr/sustain/2003/061003.pdf.
International Monetary Fund. 2004. ‘‘Sovereign Debt Restructurings and the Domestic Economy—
Experience in Four Recent Cases.’’ Policy Development and Review Department. Available at http://
www.imf.org/external/NP/pdrsdrm/2002/022102.pdf.
Kaminsky, G., and C. M. Reinhart. 1999. ‘‘The Twin Crises: The Causes of Banking and Balance of Pay-
ments Crises.’’ American Economic Review 89, no. 4: 473–500.

Kaminsky, G., C. M. Reinhart, and C. Végh. 2004. ‘‘When It Rains It Pours: Procyclical Capital Flows
and Macroeconomic Policies.’’ In NBER Macroeconomics Annual 2004, eds. M. Gertler and K. Rogoff,
11–53. Cambridge, MA: MIT Press.
Kang, K., G. Kim, and C. Rhee. 2004. ‘‘Developing the Government Bond Market in Korea: History,
Challenges, and Implications for Asian Countries.’’ Paper presented at Asia Economic Panel (AEP), Co-
lumbia University, October 7–8.
Kim, Y.-H., ed. 2001. Government Bond Market Development in Asia. Manila, Philippines: Asian Develop-
ment Bank.
Kumhof, M. 2004. ‘‘Fiscal Crisis Resolution: Taxation versus Inflation.’’ Working Paper, Stanford
University.

La Porta, R., F. Lopez-de-Silanes, A. Shleifer, and R. Vishny. 1998. ‘‘Law and Finance.’’ Journal of Politi-
cal Economy 106, no. 6: 1113–1155.

Lucas, R. E. Jr., and N. Stokey. 1983. ‘‘Optimal Fiscal and Monetary Policy in an Economy Without
Capital.’’ Journal of Monetary Economics 12: 55–93.
Mathieson, D. J., J. E. Roldos, R. Ramaswamy, and A. Ilyina. 2004. ‘‘Emerging Local Securities and
Derivatives Markets.’’ Selected Topics Paper, IMF, Washington, D.C.
Merrill Lynch. 1994. ‘‘The 1995 Guide to Brady Bonds.’’

Nicolini, J. P. 1998. ‘‘More on the Time Inconsistency of Optimal Monetary Policy.’’ Journal of Monetary
Economics 41: 333–350.
Pagano, M., ed. 2001. Defusing Default-Incentives and Institutions. Washington, D.C.: Johns Hopkins Uni-
versity Press.
Perry, G., and L. Serven. 2002. ‘‘La Anatomia de Una Crisis Multiple: Que Tenia Argentina de Especial
y Que Podemos Aprender de Ella (with English summary).’’ Desarrollo Economico 42, no. 167: 323–375.
Reinhart, C. M., K. S. Rogoff, and M. A. Savastano. 2003. ‘‘Debt Intolerance.’’ Brookings Papers on Eco-
nomic Activity 2003, no. 1: 1–74.
Reinhart, V., B. Sack, and J. Heaton. 2000. ‘‘The Economic Consequences of Disappearing Government
Debt.’’ Brookings Papers on Economic Activity 2: 163–220.
Reinstein, A. 2002. ‘‘Issues in Building Corporate Money and Bond Markets in Developing Market
Economies.’’ Working Paper, IFM Division, Inter-American Development Bank, Washington, D.C.
Rhee, C. 2004. ‘‘Asia Bond Market Initiative: Progress and Future Directions.’’ Presentation at the Inter-
national Monetary Fund.
Government Debt 277

Schmitt-Grohe, S., and M. Uribe. 2004. ‘‘Optimal Fiscal and Monetary Policy under Sticky Prices.’’ Jour-
nal of Economic Theory 114, no. 2: 198–230.
Shin, Y. 2003. ‘‘Optimal Fiscal Policy with Incomplete Markets.’’ Working Paper, Department of Eco-
nomics, University of Wisconsin, Madison.
Simpson, J., and J. Menze. 2000. ‘‘Secured Transactions: Ten Years of Legal Reform.’’ Law in Transition
(European Bank for Reconstruction and Development) Autumn: 20–27.
Siu, H. E. 2004. ‘‘Optimal Fiscal and Monetary Policy with Sticky Prices.’’ Journal of Monetary Economics
51, no. 3: 575–607.
World Bank-IMF. 2001. Developing Government Bond Markets: A Handbook. Washington, D.C.: World
Bank.
World Bank. 2004. Doing Business in 2004: Understanding Regulation. Washington, D.C.: World Bank.
World Pension Association. 2002. ‘‘Investment Restrictions to Pension Funds.’’ Available at http://
www.world-pensions.org/docus/tres16.doc.
11 Capital Income Taxation in the
Globalized World

Assaf Razin and Efraim Sadka

11.1 Introduction

These days globalization across various economies is a universal force to reckon

with. Guillermo Calvo’s framework for his prolific research was, indeed, the con-
straints on economic policy imposed by the integrated, and fluctuating, world
capital market. Maurice Obstfeld and Alan M. Taylor (2003) examine the his-
torical development of globalization (in particular, international capital mobility)
by political-economy forces. After World War I, ‘‘newly or better-enfranchised
groups such as the working classes’’ contributed to severely impede capital mobil-
ity. The peace and prosperity that emerged following World War II, and that
intensified after the end of the Cold War, unleashed political forces for freer
capital mobility. Nevetheless, the aging population (through falling birth rates
and increased longevity) raises the need for tax revenues by the welfare state.
This chapter focuses on capital income taxation. Can high domestic capital
taxes survive international tax competition brought about by recent widespread
globalization?
Evidently, the answer is negative. As put succinctly by The Economist (1997):

Globalization is a tax problem for three reasons. First, firms have more freedom over where
to locate. . . . This will make it harder for a country to tax [a business] much more heavily
than its competitors. . . . Second, globalization makes it hard to decide where a company
should pay tax, regardless of where it is based. . . . This gives them [the companies] plenty
of scope to reduce tax bills by shifting operations around or by crafting transfer-
pricing. . . . [Third], globalization . . . nibbles away at the edges of taxes on individuals. It is
harder to tax personal income because skilled professional workers are more mobile than
they were two decades ago.

In fact, Oates (1972, 192), in his classic treatise on the related issue of fiscal fed-
eralism, states:
280 Assaf Razin and Efraim Sadka

The result of tax competition may well be tendency towards less than efficient levels of out-
puts of local public services. In an attempt to keep tax rates low to attract business invest-
ments, local officials may hold spending below those levels for which marginal benefits
equal marginal costs.

The literature on tax competition has developed a great deal since this early
‘‘race-to-the-bottom’’ prediction. Specifically, Razin and Sadke (1991) point out
that this race is halted when residence-based taxation can be effectively enforced.
Baldwin and Krugman (2000) invoke agglomeration effects in order to explain
why capital may be a quasifixed factor on which the tax does not have to race to
the bottom. Indeed, Krogstrup (2002) indicates that capital taxes did not fall in
the 1960s and 1990s in the European Union (EU), and the average tax revenues
from the corporate sector even increased, both as percentages of GDP and of total
tax revenues. Nevertheless, residence-based taxation is rarely enforced effectively
and uniformly. For instance, a member country of the EU-15 can more effectively
tax its resident on income originating elsewhere in the EU-15, but less effec-
tively on income originating elsewhere in the EU-27, in this transition stage, and
even less effectively on income originating in tax havens elsewhere in the world.
We illustrate in this chapter how strong international tax competition in the era
of globalization imposes severe constraints on capital income taxation, and there-
by put into question its standing in the public finance of the welfare state.
We develop a political economy model to assess how the forces of globalization
affect the taxation of capital income. The chapter is organized as follows: section 2
provides a simple analytical framework for the study of capital taxation in the
presence of international capital mobility. In particular, we analyze the tax struc-
ture in the political-economy equilibrium. In section 3 we apply the model for the
analysis of international tax competition. Section 4 concludes.

11.2 International Capital Mobility: A Stylized Political-Economy Tax Model

We present a stripped-down model of international capital mobility, which en-

ables us to explore key issues of international taxation without being sidetracked
by irrelevant complications. We consider an economy that lives for two periods,
indexed by t ¼ 1; 2. There is one aggregate, all-purpose good in each period, serv-
ing for both consumption and investment. See Ruzin, Sadka, and Swagel (2002,
2004) for a related analysis.

11.2.1 Households

There are two types of workers: skilled workers, who have high productivity and
provide one efficiency unit of labor per unit of labor time, and unskilled workers,
Capital Income Taxation in the Globalized World 281

who provide q < 1 efficiency units of labor per unit of time. Workers have one
unit of labor time during each one of the two periods of their life. They are born
without skills and thus with low productivity. In the first period, each worker
chooses whether to get an education and become a skilled worker, or instead re-
main unskilled.
There is a continuum of individuals, characterized by an innate ability parame-
ter, e, which is the time needed to acquire a skill. By investing eb units of labor
time in education during the first period, a worker becomes skilled, after which
the remaining ð1
eÞ units of labor time in the first period provide an equal
amount of efficiency units of labor in the balance of the first period. We assume
that the individual also provides one efficiency unit of labor in the second pe-
riod. We also assume a positive pecuniary cost of acquiring skills, g, which is not
tax-deductible.
Given these assumptions, there exists a cutoff level, e  , such that those with
education cost parameters below e  will invest in education and become skilled,
whereas everyone else will remain unskilled. The cutoff level is determined by
the equality between the present value of the payoff to education and the cost of
education (including foregone income):


w2
ð1
tL Þð1
qÞ w1 þ ¼ ð1
tL Þw1 e  þ g; ð11:1Þ
1 þ ð1
tD Þr

where wt is the wage rate per efficiency unit of labor in period t ¼ 1; 2; r is the do-
mestic rate of interest; tL is the tax rate on labor income (constant over time); and
tD is the tax rate on capital income of residents from domestic sources (see next
paragraph). Rearranging terms, equation 11.1 yields


 w2 =w1 g
e ¼ ð1
qÞ 1 þ
: ð11:2Þ
1 þ ð1
tD Þr ð1
tL Þw1

Note that the two taxes, the tax on labor income and the tax on capital income,
have opposite effects on the decision to acquire skill. The tax on labor income
reduces the foregone (net of tax) income component of the cost of education. It
also reduces the payoff to education by the same proportion.1 Were the pecuniary
cost g equal to zero (or else tax-deductible), the labor income tax would have no
effect on the decision to acquire skill. However, with a positive pecuniary cost of
education, the labor income tax has a negative effect on acquiring skills: it reduces
e  and, consequently, the proportion of the population who becomes skilled
(namely, Gðe  Þ). On the other hand, the tax on capital income has a positive effect
on education because it reduces the (net-of-tax) discount rate, thereby raising the
present value of the future payoff to education.
282 Assaf Razin and Efraim Sadka

We assume for the sake of simplicity that the individual’s leisure time is exoge-
nously given. Nevertheless, total labor supply is distorted by the taxes, as can
be seen from equation 11.2. Note that there are Gðe  Þ skilled individuals and
1
Gðe  Þ unskilled individuals in each period. The labor supply of each one of
the unskilled individuals, in efficiency units, is q in each period. Therefore, total la-
bor supply in efficiency units of the unskilled individuals is q½1
Gðe  Þ in each
period. However, a skilled individual devotes e units of her time in the first period
to acquiring education, and hence works only 1
e units of time in the first pe-
riod. Thus, the individual labor supply in the first period varies over e. The labor
supply of skilled individuals is equal to
ð e
ð1
eÞ dG:
0

Any skilled individual supplies as labor all of her unit time in the second period.
Thus, total labor supply ðLt Þ in efficiency units in period t ¼ 1; 2, is given by
ð e
L1 ¼ ð1
eÞ dG þ q½1
Gðe  Þ ð11:3Þ
0

and

L2 ¼ Gðe  Þ þ q½1
Gðe  Þ: ð11:4Þ

For the sake of simplicity, assume that all individuals have identical preferences
over first- and second-period consumption [c1 ðeÞ and c2 ðeÞ, respectively], repre-
sented by a common, concave utility function u½c1 ðeÞ; c2 ðeÞ. Each individual has
initial income (endowment) in the first period of I1 units of the consumption-
capital good. The total amount of the initial endowment (I1 , because the size of
the population is normalized to one) serves as the stock of capital employed in
the first period. (This initial endowment is generated by past savings or is inher-
ited.) Because taxation of the fixed initial endowment is not distortionary, we
may assume that the government could efficiently tax away the entire value of
the initial endowments. Thus, an individual of type e faces the following budget
constraints in periods one and two, respectively:

c1 ðeÞ þ sD ðeÞ þ sF ðeÞ ¼ E1 ðeÞ þ T1 ð11:5Þ

and
C2 ðeÞ ¼ T2 þ E2 ðeÞ þ SD ðeÞ½1 þ ð1
tD Þr

þ SF ðeÞ½1 þ ð1
tF
tN Þr  ; ð11:6Þ

Capital Income Taxation in the Globalized World 283

where Et ðeÞ is after-tax labor income (net of the cost of education), t ¼ 1; 2, and
where Tt is a uniform lump-sum transfer (demogrant) in period t ¼ 1; 2. That is,

ð1
tL Þð1
eÞw1
g for e a e 
E1 ðeÞ ¼ ð11:7Þ
ð1
tL Þqw1 for e b e 

and

ð1
tL Þw2 for e a e 
E2 ðeÞ ¼ ð11:8Þ
ð1
tL Þqw2 for e b e  :

An individual can channel savings to either the domestic or foreign capital mar-
ket, because the economy is open to international capital flows. We denote by
SD ðeÞ and SF ðeÞ savings channeled by an e-individual to the domestic and foreign
capital markets, respectively. We denote by r and r  the real rate of return in these
markets, respectively.2 The government levies a tax at the rate tD on capital (inter-
est) income from domestic sources. Capital (interest) income from foreign sources
is subject to a nonresident tax at the rate of tN , levied by the foreign government.
The domestic government may levy an additional tax on its domestic residents,
on their foreign-source income, at an effective rate of tF . Note that tF þ tN is the
effective tax rate on foreign-source income of residents.
For the sake of brevity, we consider only the case of a capital-exporting
country—that is, its national savings exceed domestic investment, with the differ-
ence (defined as the current account surplus) invested abroad.3 (The analogous
case of a capital-importing country can be worked out similarly.) By arbitrage
possibilities, the net-of-tax rates of interest, earned at home and abroad, are equal-
ized; that is,

ð1
tD Þr ¼ ð1
tF
tN Þr  : ð11:9Þ

Employing 11.9, one can consolidate the two one-period budget constraints 11.5
and 11.6 into one lifetime budget constraint:

c1 ðeÞ þ Rc2 ðeÞ ¼ E1 ðeÞ þ RE2 ðeÞ þ T ð11:10Þ

where

R ¼ ½1 þ ð1
tD Þr
1 ð11:11Þ

is the net-of-tax discount factor (which is also the relative after-tax price of
second-price consumption), and
T 1 T1 þ RT 2 ð11:12Þ
284 Assaf Razin and Efraim Sadka

is the discounted sum of the two transfers (T1 and T2 ).4 As usual, the consumer
maximizes her utility function, subject to her lifetime budget constraint. A familiar
first-order condition for this optimization is that the intertemporal marginal rate
of substitution is equated to the tax-adjusted interest factor

MRSðeÞ 1 u1 ½c1 ðeÞ; c2 ðeÞ=u2 ½c1 ðeÞ; c2 ðeÞ ¼ 1 þ ð1
tD Þr ¼ R
1 ; ð11:13Þ

where ui denotes the partial derivative of u with respect to its ith argument,
i ¼ 1; 2. Equations 11.13 and 11.10 yield the consumption-demand functions
c1 ½R; E1 ðeÞ þ RE2 ðeÞ þ T and c2 ½R; E1 ðeÞ þ RE2 ðeÞ þ T of an e-individual. The
maximized value of the utility function of an e-individual, v½R; E1 ðeÞ þ
RE2 ðeÞ þ T, is the familiar indirect utility function.
Denote the aggregate consumption demand in period t ¼ 1; 2 by
Ct ½R; ð1
tL Þw1 ; ð1
tL Þw2 ; T
ð1
1 ct ½R; E1 ðeÞ þ RE2 ðeÞ þ T dG
0

ð e
¼ ct ½R; ð1
tL Þð1
eÞw1 þ Rð1
tL Þw2 þ T
g dG
0

þ ½1
Gðe  Þct ½R; ð1
tL Þqw1 þ Rð1
tL Þqw2 þ T; ð11:14Þ

where use is made of equations 11.7 and 11.8. Note that e  is a function of
ð1
tL Þw1 and of Rw1 =w2 (see equation 11.2).

11.2.2 Producers

All firms are identical and possess constant-returns-to-scale technologies, so that

with no further loss of generality we assume that there is only one firm, which
behaves competitively. Its objective, dictated by the firm’s shareholders, is to max-
imize the discounted sum of the cash flows accruing to the firm. We assume that
the firm finances its investment by issuing debt. In the first period, it has a cash
flow of
ð1
tD Þ½FðK1 ; L1 Þ
w1 L1 
½K2
ð1
dÞK1 þ tD dK1 ;

where f ðÞ is a neoclassical, constant-returns-to-scale production function. In the

second period, the firm has an operating cash flow of
ð1
tD Þ½FðK2 ; L2 Þ
w2 L2  þ ð1
dÞK2 þ tD dK2 :
Capital Income Taxation in the Globalized World 285

We denote by d both the physical and the economic rate of depreciation (assumed
for the sake of simplicity to be equal to each other). This depreciation rate is also
assumed to apply for tax purposes. We essentially assume that the corporate in-
come tax is fully integrated into the individual income tax. With such integration
of the individual income tax and the corporate tax, there is no difference between
debt and equity finance. Specifically, we assume that the individual is assessed a
tax (at the rate tD ) on the profits of the firm, whether or not they are distributed,
and that there is no tax at the firm level. The firm’s discounted sum of its after-tax
cash flow is therefore
p ¼ ð1
tD Þ½FðK1 ; L1 Þ
w1 L1 
½K2
ð1
dÞK1  þ tD dK1

þ fð1
tD Þ½FðK2 ; L2 Þ
w2 L2  þ tD dK2 þ ð1
dÞK2 g=½1 þ ð1
tD Þr: ð11:15Þ

Note that K1 is the preexisting stock of capital at the firm, carried over from pe-
riod zero. Maximizing 11.15 with respect to K2 , L1 , and L2 yields the standard
marginal productivity conditions
FL ðK1 ; L1 Þ ¼ w1 ; ð11:16Þ

FL ðK2 ; L2 Þ ¼ w2 ; ð11:17Þ

and
FK ðK2 ; L2 Þ
d ¼ r: ð11:18Þ

Note that although taxes do not affect the investment rule of the firm, the taxes
are nevertheless distortionary. To see this distortion, consider the intertemporal
marginal rate of transformation ðMRTÞ of second-period consumption (namely,
c2 ) for first-period consumption (namely, c1 ). It is equal to ð1
dÞ þ FK ðK2 ; L2 Þ.
When the economy gives up one unit of first-period consumption in order to in-
vest it, then it receives in the second period the depreciated value of this unit
(namely, ð1
dÞ), plus the marginal product of capital (namely, FK ). From equa-
tion 11.18, we can see that

MRT ¼ 1 þ r:
However, from equation 11.13 we can see that the common intertemporal mar-
ginal rate of substitution of all individuals is equal to
MRS ¼ 1 þ ð1
tD Þr:

Hence, the MRT need not equal the MRS; in fact, the MRT is larger than the MRS
when the tax rate on capital income from domestic sources ðtD Þ is positive. This
violates one of the Pareto efficiency conditions.
286 Assaf Razin and Efraim Sadka

Note that the firm has pure profits (or surpluses) stemming from the preexist-
ing stock of capital, K1 . We denote this surplus by p1 which is equal to
p1 ¼ ð1
tD Þ½FðK1 ; L1 Þ
dK1
w1 L1  þ K1 : ð11:19Þ

The surplus consists of the after-tax profit of the first period, plus the level of the
preexisting stock of capital. Given the constant-returns-to-scale technology, the
firm’s after-tax cash flow consists entirely of this surplus; that is, p ¼ p1 . This
equality follows by substituting the Euler’s equation

FðK2 ; L2 Þ ¼ FK ðK2 ; L2 ÞK2 þ FL ðK2 ; L2 ÞL2

and the marginal productivity conditions—equations 11.17 and 11.18—into equa-

tion 11.15. Naturally, the government fully taxes away the surplus p1 before
resorting to distortionary taxation (via the various tS0 )

11.2.3 Policy Tools: Taxes, Transfers, and Debt

The government has a consumption demand of CtG in period t ¼ 1; 2. We assume

that the government can lend or borrow at market rates. With no loss of general-
ity, we assume that the government operates only in the foreign capital market—
that is, its first-period budget surplus is invested abroad; for concreteness,
suppose that this is positive. Therefore, the government does not have to balance
its budget period by period, but only over the two-period horizon

C1G þ R  C2G þ T1 þ R  T2 ¼ tL w1 L1 þ tL R  w2 L2 þ tD R  rSD þ tF R  r  SF

þ p1 þ tD ½FðK1 ; L1 Þ
dK1
w1 L1 ; ð11:20Þ

where
ð1
SD ¼ sD ðeÞ dG ð11:21Þ
0

is the aggregate private savings, channeled into the domestic capital market;
ð1
SF ¼ sF ðeÞ dG ð11:22Þ
0

is the foreign aggregate private savings, channeled into the foreign capital market;
and

R  ¼ ½1 þ ð1
tN Þr  
1 ð11:23Þ

Capital Income Taxation in the Globalized World 287

is the foreign discount rate faced by the domestic economy. Note that the foreign
government levies a tax at the rate tN on interest income from the home govern-
ment budget surplus invested abroad.
The left-hand side of equation 11.20 represents the present value of the govern-
ment expenditures on public consumption and transfers, discounted by the factor
R  , which is the interest factor at which the domestic economy can lend. The
right-hand side of equation 11.20 represents the present value of the revenues
from the labor income taxes, the interest income taxes, and the pure surplus of
the firm.
Market clearance in the first period requires that

CA þ C1 þ C1G þ K2
ð1
dÞK1 þ Gðe  Þg ¼ FðK1 ; L1 Þ; ð11:24Þ

where CA is the current account surplus.5 Market clearance in the second period
requires that

C2 þ C2G ¼ FðK2 ; L2 Þ þ ð1
dÞK2 þ CA½1 þ ð1
tN Þr  : ð11:25Þ

Note that the tax at the rate tN is levied by the foreign country on the interest in-
come of the residents of the home country, and must therefore be subtracted from
the resources available to the home country.
In order to get one present-value resource constraint, we can substitute the cur-
rent account surplus, CA, from equation 11.24 into equation 11.25

C1 þ R  C2 þ C1G þ R  C2G þ K2
ð1
dÞK1 þ Gðe  Þg

¼ FðK1 ; L1 Þ þ R  FðK2 ; L2 Þ þ R  ð1
dÞK2 : ð11:26Þ

11.2.4 Median Voter Equilibrium

In this model, e b is the only characteristic that distinguishes one individual from
another. Recall that the lower is e, the more able is the individual and more objec-
tionable she is to tax hikes. We therefore take the median voter to be the decisive
voter. Thus, the political-economy equilibrium tax rates maximize the (indirect)
utility of the median voter. Policy tools at the government’s disposal are, inter
alia, labor income taxes and capital income taxes. The derivation of the equilib-
rium is relegated to the appendix.
The equilibrium tax on capital income is implicitly given by the following
condition:

1
d þ FK ðK2 ; L2 Þ ¼ 1 þ ð1
tN Þr  : ð11:27Þ

288 Assaf Razin and Efraim Sadka

Figure 11.1
The Political-Equilibrium Stock of Capital ðK2 Þ

The political-economy equilibrium stock of capital (implicitly determined from

equation 11.27, ascertains the aggregate production efficiency theorem (Diamond
and Mirrlees 1971): the intertemporal marginal rate of transformation (which is
1
d þ FK ðK2 ; L2 Þ) must be equated to the world intertemporal marginal rate of
transformation faced by the domestic economy (which is equal to 1 þ ð1
tN Þr  ).
This rule can be illustrated in figure 11.1, where first-period total (private and
public) consumption ðC1 þ C1G Þ is plotted on the horizontal axis and second-
period total consumption ðC2 þ C2G Þ on the vertical axis. Suppose that L1 , L2 , and
e  were already set at their political-economy equilibrium levels. The production
possibility frontier is described by the curve ABD, whose slope is equal (in abso-
lute value) to 1
d þ FK ðK2 ; L2 Þ. The political equilibrium stock of K2 is HD, which
gives rise to the consumption possibility frontier given by MBN. Any other level
of K2 , say H 0 D, must generate a lower consumption possibility frontier—the
curve M 0 B 0 N 0 .
Employing the firm’s investment rule (the marginal productivity condition
11.18) and the arbitrage condition (equation 11.9), we can conclude from equation
11.27 that:

r ¼ ð1
tN Þr  : ð11:28Þ
Capital Income Taxation in the Globalized World 289

That is, the pretax domestic rate of interest ðrÞ must be equated to the world rate
of interest faced by the domestic economy, which is the world rate of interest, net
of the source taxes. Equations 11.9 and 11.28 yield the political-economy equilib-
rium tax on foreign-source income:

tF ¼ tD ð1
tN Þ: ð11:29Þ

Thus, in the political-economy equilibrium, the home country imposes the same
tax rate ðtD Þ on foreign-source income from capital as on domestic-source income
from capital, except that a deduction is allowed for foreign taxes paid (and levied
at source): one dollar earned abroad is subject to a tax at source at the rate tN ; the
after-foreign-tax income, which is 1
tN , is then taxed by the home country at the
rate tD . The total effective tax rate paid on foreign-source income is therefore

tF þ tN ¼ tD þ tN
tN tD :

11.3 International Tax Competition and Capital Taxation

A critical issue of taxation in the era of the globalization of the capital markets
is the ability, or the inability, of national governments to tax their residents on
foreign-source capital income. An editorial in the New York Times (May 26, 2001)
underscores the severity of this issue:
From Antigua in the Caribbean to Nauru in the South Pacific, offshore tax havens leach bil-
lions of dollars every year in tax revenues from countries around the world. . . . The Internal
Revenue Service estimates that Caribbean tax havens alone drain away at least $70 billion
per annum in personal income tax revenue. The OECD suspects the total worldwide to be
in the hundreds of billions of dollars . . . the most notorious tax havens do not even extend
their minimal tax rates to their own citizens or domestic enterprises. Their primary aim is
to encourage and profit from individuals and businesses seeking to evade taxes in their
own countries.

It is fairly safe to argue that tax havens, and the inadequacy of cooperation
among OECD national tax authorities in information exchanges, put binding ceil-
ings on how much foreign-source capital income can be taxed. What then are the
implications for the taxes on domestic-source capital income?
Consider the extreme situation where the home country cannot effectively en-
force any tax on foreign-source capital income of its residents. That is, suppose
that tF ¼ 0. Then we can see from the political-equilibrium tax rule applying to
foreign-source capital income, equation 11.29, that the tax rate on domestic-source
capital income, tD , would be set to zero too. Thus the capital income tax vanishes
290 Assaf Razin and Efraim Sadka

altogether. And even if some enforcement of taxation on foreign-source capital in-

come is feasible so that tF does not vanish altogether, it still follows from equation
11.29 that tD ¼ tF =ð1
tN Þ, so that a low tF generates a low tD . Indeed, a poor en-
forcement of international taxes would generate political processes that would re-
duce significantly the domestic-source capital income taxation.
The unwillingness of foreign tax authorities to cooperate with the home tax au-
thority in helping to enforce capital taxation on home country residents’ capital
income originating abroad usually stems from their desire to lure capital to their
countries. This is what is meant by tax competition. They further compete with
the home country by lowering the source tax ðtN Þ they levy on the capital income
of the home country residents. We thus capture formally the effect of tax competi-
tion by assuming that tN falls as foreign governments lure capital to their coun-
tries. Then we can see from equation 11.28 that r, the net (of depreciation d)
marginal product of domestic capital, must rise. With diminishing marginal prod-
uct, this must happen when the stock of domestic capital falls and more capital
flows abroad. Hence, the tax base for the domestic-source capital income shrinks,
thereby turning the enforcement of foreign-source capital income all the more
acute.

11.4 Conclusion

The behavior of taxes on capital income in the recent decades points to the notion
that international tax competition that follows globalization of capital markets
puts strong downward pressures on the taxation of capital income—a race to the
bottom. This behavior has been perhaps most pronounced in the EU-15 following
the single market act of 1992. (See, for example, Razin and Sadka 2005.) The 2004
enlargement of the EU with ten new entrants put a strong downward pressure on
capital income taxation for the EU-15 countries. Table 11.1 describes the corporate
tax rates in the twenty-five EU countries in 2004. It reveals a marked gap between
the original EU-15 countries and the ten accession countries. The latter have sig-
nificantly lower rates. Estonia, for instance, has no corporate tax, the rates in Cy-
prus and Lithuania are 15 percent, and Latvia, Poland, and Slovakia are at 19
percent. In sharp contrast, the rates in Belgium, France, Germany, Greece, Italy,
and the Netherlands range from 33 percent to 40 percent.
Europe’s new constitution, adopted by the European Summit in Brussels, June
19, 2004 (the text, however, has yet to be ratified in all twenty-seven member
states), cannot stop the tax competition process because the new constitution
retains the national veto in tax cooperation and harmonization. But, there is flexi-
bility for some countries that want to push ahead tax harmonization to do so. In-
Capital Income Taxation in the Globalized World 291

Table 11.1
Statutory Corporate Tax Rates in the Enlarged EU, 2003
Country Tax Rates (%)

Austria 34
Belgium 34
Cyprus* 15
Czech Republic* 31
Denmark 30
Estonia* 0
Finland 29
France 33.3
Germany 40
Greece 35
Hungary* 18
Ireland 12.5
Italy 34
Latvia* 19
Lithuania* 15
Luxembourg 22
Malta* 35
Netherlands 34.5
Poland* 27
Portugal 30
Slovakia* 25
Slovenia* 25
Spain 35
Sweden 28
UK 30
* New entrants.

deed, Germany and France are currently pushing some of the new entrants to
raise their corporate tax rates. Tax competition within the EU is in sharp contrast
to the U.S. federal fiscal system, where the capital income tax (on individuals and
corporations) is federal and not state specific. But both the EU and the U.S. are
subject to severe tax competition from the rest of the world.

11.5 Appendix: Derivation of the Political-Economy Equilibrium

The political-economy equilibrium tax rates maximize the (indirect) utility of the
median voter. Denoting the indirect utility function of the median voter by V, it is
given by:
292 Assaf Razin and Efraim Sadka


v½R; ð1
eM Þw1N þ Rw2N þ T
g if eM < e 
VðeM ; R; w1N ; w2N ; TÞ ¼
v½R; qðw1N þ Rw2N Þ þ T if eM > e  ;

where wtN ¼ ð1
tL Þwt is the after-tax wage per efficiency unit of labor in period
t ¼ 1; 2, and eM is the innate ability parameter of the median voter. Policy tools at
the government’s disposal are, inter alia, labor income taxes and capital income
taxes. We therefore assume that the government can effectively choose the after-
tax wage rates (w1N and w2N ) and the after-tax discount factor ðRÞ. The govern-
ment can choose also T, the discounted sum of the lump-sum transfers (T1 and
T2 ). Once w1N , w2N , R, and T are chosen, then private consumption demands
[C1 ðR; w1N ; w2N ; TÞ and C2 ðR; w1N ; w2N ; TÞ] are determined. The cutoff level, e  , and
labor supplies, L1 and L2 , are also determined as follows:

e  ðR; w1N ; w2N Þ ¼ ð1
qÞ½1 þ Rw2N =w1N 
g=w1N ð11:2 0 Þ

ð e  ðR; w N ; w N Þ
1 2
L1 ðR; w1N ; w2N Þ ¼ ð1
eÞ dG þ qf1
G½e  ðq; w1N ; w2N Þg ð11:3 0 Þ
0

L2 ðR; w1N ; w2N Þ ¼ G½e  ðR; w1N ; w2N Þ þ qf1
G½e  ðR; w1N ; w2N Þg: ð11:4 0 Þ

In choosing its policy tools (R, w1N , w2N , and T) and its public-consumption
demands (C1G and C2G ), the government is constrained by the economy-wide
‘‘budget’’ constraint 11.26, where C1 , C2 , L1 , L2 , and e  are replaced by the func-
tions C1 ðÞ, C2 ðÞ, L1 ðÞ, L2 ðÞ, and e  ðÞ, given by equations 11.14 and 11.2 0 –11.4 0 ,
respectively. Note that the capital stock in the first period ðK1 Þ is exogenously
given. The capital stock in the second period ðK2 Þ must satisfy the investment
rule of the firm (equation 11.18). Note that because the economy is financially
open, the individuals, by the arbitrage condition (equation 11.9), are indifferent
between channeling their savings domestically or abroad. This means that the
government can choose K2 , and then r and the pretax wages (w1 and w2 ) are de-
termined so as to clear the capital market and labor market in each period
through equations 11.18, 11.16, and 11.17, respectively. This does not mean that
the government actually chooses the stock of capital ðK2 Þ for the firm, or the pre-
tax wage rates (w1 and w2 ), or the domestic interest rate ðrÞ. Rather w1 , w2 , and r
are determined by market clearance, and the firm chooses K2 so as to maximize
its value. What we did is to determine K2 , w1 , w2 , and r at levels that are compati-
ble with firm-value maximization and market clearance in the presence of taxes.
To sum up, the government in a political-economy equilibrium chooses C1G , C2G ,
R, w1N , w2N , T, and K2 so as to maximize the utility of the median voter (as given
by equation 11.29), subject to the economy-wide ‘‘budget’’ constraint, equation
Capital Income Taxation in the Globalized World 293

11.26. Note that C1 , C2 , L1 , L2 , and e  in the latter constraint are replaced by the
functions C1 ðÞ, C2 ðÞ, L1 ðÞ, L2 ðÞ, and e  ðÞ, respectively. We can ignore the gov-
ernment budget constraint 11.20 by Walras Law.
Note that in this maximization, K2 appears only in the economy-wide ‘‘budget’’
constraint, equation 11.26. Thus the first-order condition for the political-economy
equilibrium level of K2 is given by

1
R  FK ðK2 ; L2 Þ
R  ð1
dÞ ¼ 0:

Note that this choice does not depend on whether the median voter is skilled or
unskilled.
Substituting the firm’s investment rule, equation 11.18, and rearranging terms
yields

1
d þ FK ðK2 ; L2 Þ ¼ 1 þ ð1
tN Þr  :

Notes

This chapter is based on a keynote address to the annual congress of the International Institute of Pub-
lic Finance (IIPF), Prague, August 25–28, 2003.
1. Evidently, if the tax is progressive, the payoff would be reduced proportionally more than the
foregone-income cost.

2. These rates (r and r  ) hold in essence between periods one and two and we therefore assign no time
subscript (one or two) to these rates.

3. Evidently in a nonstochastic set-up like ours, the country is either capital exporter or capital
importer.
4. Note that even though T may seem at first glance to be dependent on tD (through the discount fac-
tor R), we may nevertheless assume that these are two independent policy tools because the govern-
ment can always change either T1 and T2 in order to keep T constant when it changes tD .
5. For notational simplicity, we assume that the net external assets are initially equal to zero, so that
there is no initial external debt payment term in the current account.

References

Baldwin, Richard and Paul Krugman. 2000. ‘‘Integration and Tax Harmonization.’’ Discussion Paper
No. 2630, CEPR, London.
Diamond, Peter A., and James A. Mirrlees. 1971. ‘‘Optimal Taxation and Public Production.’’ American
Economic Review 61, no. 1: 8–27 and no. 3: 261–278.
Krogstrup, Signe. 2002. ‘‘What Do Theories of Tax Competition Predict for Capital Taxes in EU Coun-
tries? A Review of the Tax Competition Literature.’’ Working Paper No. 05/202, Graduate Institute of
International Studies, Geneva, Switzerland.
294 Assaf Razin and Efraim Sadka

Oates, William E. 1972. Fiscal Federalism. New York: Harcourt Brace Jovanovich.
Obstfeld, Maurice, and Alan M. Taylor. 2003. ‘‘Globalization and Capital Markets.’’ In Globalization in
Historical Perspective, eds. Michael D. Bordo, Alan M. Taylor, and Jeffrey G. Williamson, 121–124. Chi-
cago: University of Chicago Press.
Razin, Assaf, and Efraim Sadka. 1991. ‘‘International Tax Competition and Gains from Tax Harmoniza-
tion.’’ Economics Letters 37, no. 1: 69–76.
Razin, Assaf, and Efraim Sadka, with the cooperation of Chang Woon Nam. 2005. The Decline of the
Welfare State: Demography and Globalization. Cambridge, MA: MIT Press.
12 Can Public Discussion Enhance
Program ‘‘Ownership’’?

Allan Drazen and Peter Isard

12.1 Introduction

IMF programs have a mixed record of success. A significant share of programs

fail to reach their originally scheduled completion dates, many others require sub-
stantial modifications along the way,1 and many countries remain under IMF pro-
grams for prolonged periods of time.2
An oft-cited reason that Fund programs get off track is a lack of ownership on
the part of the program country, where ‘‘ownership’’ refers loosely to the extent
to which a country is committed to the general reform process and the conditions
specified by the program independently of the incentives provided by multilateral
institutions. Genuine ownership on the part of a country, combined with sufficient
lending from the Fund and good program design, is believed to yield program
success (or at least a high probability of success). And indeed, when the Fund
refers to ownership it defines the concept in a circular way, based implicitly on
the vision of reaching program completion dates: ‘‘National ownership refers to a
willing assumption of responsibility for a program of policies, by country officials
who have the responsibility to formulate and carry out those policies, based on
an understanding that the program is achievable and is in the country’s best
interest.’’3
The main challenges that arise when wrestling with the concept of ownership
do not stem primarily from disagreement over the most appropriate definition,
but rather reflect the difficulties in making the concept of ownership operational.4
How can the Fund determine whether a country is likely to remain committed to
the reform effort and comply with the conditions of its program through the com-
pletion date? And how can the Fund design programs to enhance the prospect
that countries will remain committed and completion dates will be reached?
These questions identify and distinguish between two central issues: demonstrat-
ing that ownership is present and building or creating ownership when it is
thought to be weak. The former issue is perhaps most relevant for countries that
296 Allan Drazen and Peter Isard

have failed to comply with program conditions in the past and must demonstrate
that, contrary to past performance, they are now truly interested in and com-
mitted to reform.5 The latter issue recognizes that ownership is endogenous, not
an innate or exogenous characteristic. Ownership can be improved and fostered
through various devices. That is why we argue that a key challenge faced by a
reform-minded government is building ownership—that is, creating support for
the process of reform and for the specific measures embodied in a Fund program.
Among the general public and especially critics of the IMF (and the World
Bank), the lack of ownership and the failures of Fund programs are often attrib-
uted to a lack of public discussion of the programs during their formulation
stages. Our use of the term public discussion in this paper refers broadly to mecha-
nisms through which policy makers disseminate and exchange information with
the public, including discussions both to formulate programs and to explain pro-
gram design. It appears to be similar what others have referred to as ‘‘participa-
tion’’ in the recent literature on economic development and poverty and social
impact analysis.
Critics of the IMF often characterize the process of formulating a program as
one of negotiations between the IMF (seen as having its own agenda) and a thin
layer of high-level government officials, with little or no input from the public,
from nongovernmental organizations (NGOs), or from other interested parties.
The failure of these programs to lead to a significant bettering of the general pop-
ulation’s economic condition is often seen as reflecting this lack of public input.
Demonstrations (or riots) against the IMF and its programs are taken as evidence
of the seriousness of the problem. Public discussion is considered an important
vehicle for establishing ownership, and lack of public discussion is perceived as
part of the reason that some programs are never adopted or fail to reach
completion.
The need for public discussion is recognized within the IMF and World Bank.6
To date, however, the case for public discussion is not much more developed
than the cases for other terms that sound unambiguously positive. In particular,
the discussion of ‘‘public discussion’’ seems quite unfocused and lacking in coher-
ent analysis of what form it should take and how it can contribute to strengthen-
ing Fund programs or demonstrating and building ownership. Such lack of
careful, analytic argument about the role of public discussion can leave the unfa-
vorable impression that the IMF and World Bank simply view public discussion
as politically correct rather than as a vehicle that can significantly improve pro-
gram design—or that they engage in public discussion primarily to make the
public feel more comfortable, or warm and fuzzy, about their programs. A more
serious concern is that the lack of a clear understanding of how public discussion
works can leave the impression that public discussion strengthens ownership only
Can Public Discussion Enhance Program ‘‘Ownership’’? 297

by leading counterproductively to the adoption of weaker programs. At worst,

public discussion is seen as making strong program design very difficult or im-
possible by giving too much influence to groups that are either against real reform
or hold views strongly at odds with generally accepted economic wisdom. Thus,
it is not surprising that public discussion has a bad name in some circles (at least
when it is not being discussed publicly).
Both the association of public discussion with fuzzy thinking and the view that
it detracts from successful program design are especially unfortunate. In our
view, careful public discussion of programs can increase necessary public support
and in any case is often unavoidable. The purpose of this paper is to distinguish a
number of different functions that public discussion can serve and to illustrate its
potential for improving collective choice.
This paper does not break new theoretical ground, but it uses the tools of eco-
nomic theory and political science to develop a better understanding of how pub-
lic discussion can contribute to building and demonstrating ownership. The first
task is to develop a better understanding of both ownership and public discus-
sion. In section 12.2, we argue that ownership is more complex than many discus-
sions of it would suggest. Ownership must include not only willingness, but also
the technical capacity and political ability to carry out a program (especially the
political ability.) In section 12.3, we argue that public discussion can serve a num-
ber of purposes—including educating the public, revealing public preferences
and constraints (or educating policy makers), demonstrating the unbiasedness of
policy makers, and finding common ground among heterogeneous interests—
each of which can be better understood by moving to a more formal treatment.
What public discussion can or cannot achieve depends on what the ownership
problem is, and the different purposes of public discussion may interact with one
another. Here too, understanding the political constraints affecting a government
may help us understand what to expect from public discussion. In section 12.4,
we illustrate our arguments using simple examples relevant to actual program de-
sign. In section 12.5 we consider some of the drawbacks of public discussion, es-
pecially as applied to Fund programs.
After having developed a clearer sense of ownership and a more formal under-
standing of what public discussion can achieve in different types of circumstances,
we focus in section 12.6 on the design of IMF programs and the role that public
discussion can play in enhancing program ownership. We argue that the diffi-
culties in attaining program ownership depend on the nature of the conditions
involved, that this should be taken into account in designing programs and seek-
ing to build overall program ownership, and that the recent emphasis on stream-
lining Fund conditionality can be counterproductive if it results in the elimination
of conditions for which ownership is potentially very high. We also suggest that
298 Allan Drazen and Peter Isard

the effectiveness of both Fund programs and the Fund’s surveillance of nonpro-
gram countries might be enhanced through more explicit focus ex ante on the
obstacles to building ownership for policies and programs, and on how those
obstacles can be overcome or mitigated through public discussion. Section 12.7
provides a summary of our arguments and conclusions.

12.2 Defining Ownership

Ownership is a more complex concept than it may seem. As noted above, the de-
gree to which a country ‘‘owns’’ a Fund program can be broadly defined as the
extent to which the country is oriented toward meeting the program conditions,
independent of any incentives provided by multilateral lenders. Alternatively,
and from the perspective of seeking to draw meaningful implications for Fund
program design and the role of public discussion, the degree of ownership can be
defined as the probability that a Fund program will be adopted and reach com-
pletion without a significant weakening of the reform effort.
The latter definition points to a number of factors on which ownership depends.
One is the willingness of the government to meet the conditions of the program.
Without such willingness, the probability of reaching completion is unlikely to be
high. Suppose, however, that there is willingness, but a lack of technical capacity
to collect taxes or carry out other measures on which the program crucially de-
pends. (‘‘The spirit is willing, but the infrastructure is weak.’’) It would not seem
meaningful to suggest that a country owns a program that it has no hope of carry-
ing out: ownership also depends on technical capacity. Conversely, ownership for
its own sake is not a desirable objective. Just as it is not very meaningful to design
diets that individuals are willing and able to undertake because they require only
minimal changes in eating habits, it would not be very desirable to design pro-
grams that countries are willing and technically able to carry out simply because
they require only minimal changes in policy.
A third crucial factor for ownership is the political ability of the government to
carry out a meaningful program. To illustrate the point, suppose policy makers in
a country with continuing large fiscal deficits want to implement a program of fis-
cal austerity, but face powerful special interest groups (SIGs), either inside or out-
side government, who can block fiscal reform. One could not argue that there is
country ownership if the SIGs do not agree to the fiscal austerity program, even
if the government is willing and technically able to carry the program out.7 The
same argument suggests that ownership can be strengthened by weakening either
the power of or the incentives for SIGs to block the program. The nature of the po-
litical system can very much determine the power of SIGs, while both the design
Can Public Discussion Enhance Program ‘‘Ownership’’? 299

and public discussion of programs can affect their incentives. We return to the lat-
ter point in our analysis of how public discussion can affect ownership.
Willingness, technical capacity, and political ability are not independent of each
other as determinants of program ownership. The distinction between the willing-
ness of the government to carry out reforms and its political ability is often un-
clear, as is the distinction between its technical capacity and its political ability to
collect taxes or discipline public spending. In some cases, distinguishing between
technical and political capacity may be unimportant or irrelevant. One may want
to think simply of ‘‘institutional capacity,’’ broadly defined to reflect the technical
characteristics of fiscal processes as well as the nature of the political institutions
that influence what comes out of fiscal processes.
The Fund is unlikely to find itself involved in a program in which country
ownership is ex ante complete. Were ownership complete—that is, if a country
had the willingness, technical capacity, and political ability to pursue serious
reforms—it would not have to approach the Fund for support.8 In this sense, the
challenge faced by the Fund in negotiating and designing programs and in con-
sidering how to make effective use of public discussion is not simply to try to de-
termine the degree of ownership, but also to try to increase—that is, to build and
strengthen—ownership. This leads to the question of the role of public discussion
in designing effective programs and especially in enhancing ownership.

12.3 A Framework for Understanding the Role of Public Discussion

Public discussion has an image problem: calls for public discussion are often per-
ceived to be based primarily on the desire to increase ownership by making the
public feel more comfortable about a program in some general way. Our aim in
this paper is to move away from the warm and fuzzy and to try to analyze public
discussion more rigorously.
From a more analytical perspective, public discussion of a program may be
seen as having a number of functions, which may be grouped into four categories:
•To educate the public about the nature of the program and, more generally, about
macroeconomic realities, including the trade-offs between short-run costs and
medium-run benefits and the implications of the program for relevant interest
groups. It can also provide a vehicle for the government (and/or the Fund) to
make the public aware of information on which the program is based;
• To learn public preferences and constraints, and to convince the public that program
design takes this information into account, which (along with educating the pub-
lic) may enhance compliance and cooperation with the program;
300 Allan Drazen and Peter Isard

•To demonstrate unbiasedness—that is, to convince the public that the program is
designed for the general good, rather than to serve the interests of the authorities
or the Fund;
• To find common ground among heterogeneous interests.

‘‘Public discussion’’ might appear to be a topic not easily formalized or

discussed with any rigor. However, there are two highly relevant strands of
literature—one in economics, the other in political science—that are rigorous
and, we believe, potentially quite relevant for developing a more concrete aware-
ness of the different ways that public discussion can affect policy outcomes. As we
discus in section 3.1, the economics literature on cheap talk is useful for illustrat-
ing how discussion can help country governments and/or the Fund learn about
preferences and constraints, demonstrate their unbiasedness, or find common
ground among different interest groups. And the political theory literature on de-
liberative democracy is helpful in thinking about the educational role of public
discussion.

12.3.1 Cheap Talk

Economic theory distinguishes between costly signaling mechanisms and ‘‘cheap

talk,’’ where the latter is defined as a signal an agent can send that is neither costly
nor binding and that does not directly affect the payoffs associated with any
given outcome.9 As an example of costly signaling, suppose that two jobs require
different levels of ability, but that individual ability is unobserved. If innately
high-ability individuals find it less costly to become educated than do low-ability
individuals, observable educational attainment may serve as a signal of unob-
served ability. The signal works because it is differentially costly for different
types of individuals to send it.
Suppose, however, that individuals simply announce to employers what their
ability is, it being costless for an individual to make any announcement he wants
(hence the term ‘‘cheap talk’’). One might think that such costless signals are com-
pletely uninformative. However, under certain conditions, cheap talk can in fact
convey information about unobserved characteristics. Cheap public discussion
may enable the government to convey the information that its objective is maxi-
mizing social welfare (rather than having its own agenda), and cheap talk may
enable the public to convey information about its preferences to those who design
programs. We return later to this application of public discussion.
Cheap talk may also be a useful coordination mechanism; that is, a device for
finding common ground. The challenge of program design presents the govern-
Can Public Discussion Enhance Program ‘‘Ownership’’? 301

ment not only with the difficulties of convincing the populace of its good inten-
tions, but also with the need to forge an agreement among different special
interest groups. We also consider how public discussion applies to the latter task.
Note that in considering the importance of talking we have not made any refer-
ence to bargaining or negotiations per se. The issue of how bargaining affects
outcomes can, of course, be analyzed formally, but that is not our intention in
studying public discussion.

12.3.2 Deliberative Democracy

Political science has generated quite a bit of literature on communication between

political leaders and the public, and on the role of such communication in shaping
policy decisions and generating support for policies and political leaders. Much of
this literature is empirical or illustrative, arguing the importance of communica-
tion but without providing a theoretical or formal way to understand why public
discussion may be important.
There is, however, a rigorous strand of literature focusing on why public dis-
cussion may be important in a democracy.10 This is the analysis of ‘‘deliberative
democracy,’’ which can be roughly defined as ‘‘decision making by discussion
among free and equal citizens’’ (Elster 1998, 1), with the stress on ‘‘discussion.’’
According to this literature, a crucial part of the democratic process is discussion
and deliberation before a collective choice is made.
The central question in assessing discussion is put clearly by Fearon (1998):
What good reasons might a group of people have for discussing matters before making
some collective decision, rather than simply voting on the issue or using some other deci-
sion rule that does not involve discussion? In other words, what is the point or value of dis-
cussing things before making political decisions? (44, italics in original)

Advocates of deliberative democracy argue that decision-making in a democracy

is not simply the aggregation of (existing) policy preferences among heteroge-
neous agents in a society. That is, it is not simply a collective choice rule (such as
majority voting) for reaching an aggregate policy decision when individual citi-
zens do not agree on their most preferred policy. Instead, and perhaps more im-
portantly, democratic decision-making involves the transformation of individual
preferences through deliberation and discussion (Habermas 1987).11 In common
parlance, if two individuals have different opinions, the free and open discussion
of issues that is part of the democratic process enables each side to attempt to con-
vince the other of the correctness of its own position. Thus, discussion provides a
mechanism for trying to narrow differences of opinion. In the limit, deliberation
302 Allan Drazen and Peter Isard

and discussion would bring about a consensus of opinion so that there would be
no need to use voting or some other mechanism to aggregate preferences.
In applying this perspective to Fund programs, the relevant question is, how
can public discussion lead to a consensus, or at least a narrowing of differences of
opinion, on what policy should be followed? It is crucial to note that the focus
here is on policies and not on outcomes. Two constituencies may differ in their
desires for narrowing income inequality, for example, as an outcome, or, they
may agree on the desired degree of income inequality, but differ over the best
way to achieve it. Hence, their disagreement is not about goals, but about the best
way to achieve them.12 They disagree about the connection between policies and
outcomes, or more generally, about how the world works. We term this an issue
of causal policy-to-outcome relations, or simply ‘‘causal relations.’’
We note this point since—in the context of Fund programs—the possibility of
educating the public about the effects of policies is a key consideration. That is,
public discussion can make the public aware of the payoff matrix they face, and
thus perhaps lead them to choose to comply with the proposed program (to
choose ownership) as leading to better outcomes for all parties than the outcomes
associated with noncompliance.
In addition to the positive question of whether deliberation can lead to owner-
ship, there is a normative question of whether deliberation leads to welfare-
improving outcomes. As already mentioned in the introduction, some observers
stress the possible negative implications, a point we consider in detail in section 5.

12.3.3 Voting versus Program Design

Many discussions of deliberative democracy focus on the role of public discussion

prior to voting on policy. By contrast, we are interested in the value of public dis-
cussion of policy proposals before a government commits to a policy program
with the IMF.
The fact that public discussion is followed by the government adopting a pro-
gram, rather than voting, is a key distinction between our framework and most
applications in the political theory literature. This difference may have implica-
tions for the functions of public discussion, as well as for the pathologies (the
circumstances in which public discussion may be counterproductive). A second
factor distinguishing our framework, which may influence the relative impor-
tance of various arguments, is that IMF programs—especially difficult ones that
need the most public discussion—are typically designed when the economic out-
look is particularly bad. As we shall see, these features of our framework intro-
duce some considerations not present in a voting setup.
Can Public Discussion Enhance Program ‘‘Ownership’’? 303

12.4 Illustrating the Functions of Public Discussion

Public discussion is a vehicle for transmitting information. This includes both eco-
nomic information about the state of the economy and the links between policies
and economic outcomes and sociopolitical information about the preferences and
agendas of the government, relevant national constituencies, or the Fund itself.
This section focuses on how the design and ownership of programs can be
strengthened by the transmission of such information. Our purpose is not to break
new theoretical ground, but to illustrate these points by means of simple exam-
ples as applied to actual Fund programs.
In analyzing the various functions that public discussion may serve, it is useful
to address three separate questions sequentially. First, as a reference case, what
are the functions of public discussion when the government is known to maxi-
mize social welfare (rather than being thought to have its own agenda) and the
public is homogeneous (rather than having heterogeneous interests)? Second,
how can public discussion help convince the public that the government’s objec-
tive is in fact the maximization of social welfare (thus presumably increasing pub-
lic support and compliance)? Third, what additional functions might public
discussion have in addressing the conflicting interests of groups in society?
Throughout this section we restrict attention to communication between two
parties, focusing on communication either between the government and the pub-
lic or between two interest groups. We assume that the Fund’s objectives coincide
with those of the government.13 This assumption implies that the argument that
the Fund’s interests do not coincide with those of countries entering programs
can be captured by case 2 (section 4.2) on the role of public discussion when the
authorities in a country may not be acting in the interests of social welfare.

12.4.1 Case 1: Government Known to Maximize the Social Welfare of a

Homogeneous Public

Consider first the case where it is common knowledge that the government’s
objective is the maximization of social welfare and the public is homogeneous
in its preferences and beliefs.14 These two assumptions are admittedly extreme
and unrealistic, but this case is useful in understanding the importance of public
uncertainty about the government’s objectives and of heterogeneity in agents’
objectives.
In this reference case we ask, since the public knows that the government is
maximizing its welfare, why isn’t it effective for the government simply to an-
nounce its program without any prior discussion? Our answer is that achieving
304 Allan Drazen and Peter Isard

the optimal outcome may depend on a number of functions that public discussion
can serve: (1) revealing the public’s preferences; (2) convincing the public that the
government is aware of what it values; (3) educating the public; (4) revealing in-
formation about available resources and/or efficient choices; and (5) making the
government’s commitment to a program more credible.
When the government has imperfect information about what the public wants,
public discussion can reveal information. This can be a serious problem when pro-
gram negotiation involves only the Fund and a thin layer of high-level govern-
ment officials. Since preferences are necessarily complex, it can be argued that the
government and the Fund cannot possibly design a program that is socially opti-
mal (or perceived as such) without consulting the public. Compared to other
ways of forming judgments about ownership, public discussion has the attraction
of being a relatively quick and resource-efficient way to seek feedback.15
We may illustrate our points in simple matrices of payoffs to the government
and a homogeneous public. Consider matrix 12.1. (In this case the use of a formal
model is perhaps trivial, but it sets the stage for later examples.) The public’s type
is left, but the government does not know this. (Formally, the game is either left
or right, which public knows but government does not.) For example, left means
relatively large tax-financed social welfare programs and right means relatively
small tax-financed social welfare programs. A policy of high (government spend-
ing) is optimal when the public prefers left and a policy of low (government
spending) is optimal when the public prefers right. (Since the government is
known to maximize social welfare, the payoffs are identical to both public and
government in each cell.)

Matrix 12.1
Public
Government
Program Left Right

High 3,3 1,1

Low 1,1 3,3

The government must choose the optimal policy when it does not know what
the public wants. In the absence of communication, the government must choose
on the basis of its prior beliefs about the public’s preferences (where we assume
that once government chooses a policy it cannot change). Suppose government
Can Public Discussion Enhance Program ‘‘Ownership’’? 305

assigns probability of 1/2 to each possibility (actually, any interior probabilities

will suffice for this example), so that the expected payoff is ð2; 2Þ.
Here, the problem is solved if the public simply announces that it is left, which
it is optimal for it to do. Public discussion increases welfare since it allows the
public to credibly signal its preferences even though such discussion is cheap
talk. (Formally, the public’s preferences are self-signaling.)16
A related role for public discussion is to convince the public that the govern-
ment is really aware of what the public values. This is relevant when the public
knows that the government’s objective is maximization of social welfare, but feels
the government may not be aware of society’s preferences. This is clearly related
to the previous argument, but it is not identical: the previous argument applies
when the government does not have information on the public’s preferences; the
second role recognizes that even a perfectly informed government must some-
times convince people that it is informed. Put another way, in the first case public
discussion is used to shape a program; in the second, to help sell a program. For-
mally, the public’s preferences must not simply be known, but must also be com-
mon knowledge. This possibility suggests one interpretation of the notion that
people want to be included in program formulation—in this case, enough so to
be convinced that the decision maker really does know their preferences. In other
words, public discussion may serve to increase the legitimacy of policy.
Another function of public discussion is to address the possibility that the gen-
eral public knows the government’s preferences (or at least knows that the gov-
ernment’s objective is the maximization of its welfare) but is uninformed about
how the world works; that is, about what we earlier termed causal relations.17 In
other words, the function of public discussion is to educate the public about eco-
nomic realities. Przeworski (1998), for example, argues that transmitting knowl-
edge about such causal relations is central to what deliberation is supposed to do:
The preferences on which people act are contingent on their beliefs about the consequences
of their actions. And, indeed, parties and candidates competing for office do not offer
merely policies: they explain to the electorate how these policies will affect the outcomes,
trying to persuade citizens that their policies, as distinct from those of their opponents, will
lead to the outcomes they want. (144)

We may represent this case by making a number of changes to matrix 12.1.

First, left and right now refer not to the public’s type, but to the state of the world.
Second, it is now the government that has superior information—it knows the
state of the world and the associated optimal policy, while the public does not.
For purposes of the example, let the state be right, calling for a policy of low gov-
ernment spending. (That is, in matrix 12.1, the true ‘‘game’’ is only the right-hand
column.) Third, a government program can be enacted only if the public favors it.
306 Allan Drazen and Peter Isard

(Hence, it is the public that chooses high or low.) Finally, the public does not
know the true payoffs connected with different policies. As a concrete example,
one may think of the state of the world as overheated, with the public not fully
understanding the longer-run effects of countercyclical and procyclical policies.
Though the true payoffs are as in matrix 12.1, the public believes that the payoffs
associated with different policies are as in matrix 12.2.

Matrix 12.2
State of the World
Government
Program Left Right

High 2,2 1,1

Low
1,
1
2,
2

In this case, the public would favor high government spending even if the gov-
ernment announced that the state was right and that announcement was believed.
Hence, though the government has superior information, it must convince the
public not only about the state of the economy, but also about what works. Once
it does that (that is, once it convinces the public that the payoffs are those in ma-
trix 12.1 rather than matrix 12.2), then it need simply announce the state of the
world and a policy of low spending would be supported.
The educational function of public discussion may be especially important in
countries where IMF programs have historically been viewed as ‘‘bitter medicine
prescribed by an unsympathetic international financial community,’’ as repre-
sented by the public belief in payoffs described by matrix 12.2. In these cases,
public discussion has the potential to help identify the truth by sifting through
whatever valid or invalid inferences have been drawn from historical experience.
Educating the public about causal relations may also be important to ensure
compliance with a program. A program may only be successful if the public com-
plies with it, which depends on the public understanding why it is beneficial. To
illustrate, suppose left and right in the previous two matrices now refer to actions
the public may take, high and low to actions the government takes. The govern-
ment moves first, but knows how the public will react (that is, whether the public
will comply with a government program). Both the public’s and the government’s
actions are fully observable. Suppose the true payoffs are as in matrix 12.3, but the
public believes that matrix 12.2 represents the true payoffs.
Can Public Discussion Enhance Program ‘‘Ownership’’? 307

Matrix 12.3
Public Actions
Government
Program Left Right

High 2,2 1,1

Low 1,1 4,4

The public would then choose left, no matter what the government chose. The
government would be induced to choose high. If the public knew the true payoffs
to their actions, they would choose right, and comply with a program of low gov-
ernment spending.
Consider, for example, a program of fiscal restraint that requires replacing gov-
ernment supply of goods (high government spending) with reliance on the pri-
vate sector (low government spending) for supply. Success of the market solution
requires public compliance. Public willingness to rely on the market rather than
government is represented by right rather than left. Matrix 12.2 thus might repre-
sent public distrust of the market’s ability to supply services, combined with an
unwillingness to use the market. Public education thus means educating the pub-
lic on the value of greater reliance on the market when there is a need to reduce
the size of government, with matrix 12.3 showing the value of the public acting in
accordance with this view.
The functions of public discussion in revealing information to the government
and in eliciting optimal public responses interact when it enables the government
to learn about the resources the public has. That is, public discussion can also play
an important role in educating the government and identifying policy choices that
are likely to be most efficient, or elicit the greatest ownership, when there are a
number of options. (We leave it to the reader to combine the previous cases to
demonstrate this formally.) Governments are rarely (if ever) fully informed of the
effects of their policies on society, and there are endless examples of situations in
which public discussion has made governments aware of adverse policy effects
on certain groups of society and of the possible desirability of policy adjustments
or compensating policy actions. This may be seen in the broad consultation that
takes place among stakeholders and development partners during the process of
preparing country strategy papers under the IMF and World Bank’s poverty re-
duction initiative. For example, public discussion made governments aware of
the regressive nature of local graduated taxes in Uganda and of commune-level
308 Allan Drazen and Peter Isard

revenue-raising measures in Vietnam, of the effects of energy price liberalization

on the livelihoods of fishing communities and of low-paid casual laborers who
were heavily affected by transportation costs in Mozambique, and of the scope
for building ownership by addressing these public concerns.18 In addition to mak-
ing the government more aware of the effects of its policies, public discussion and
participation sometimes identify resources that can be mobilized to pursue policy
objectives more efficiently.19
Yet another function of public discussion is to induce the government to make a
greater commitment to a program, and hence make it more credible that it will be
carried through. One way to think of this is in terms of solving a time consistency
problem. By taking publicly announced positions, the government changes the
payoff structure associated with following a given policy in order to commit itself
to a specific course of action.
For example, following a severe banking and currency crisis that led to near
hyperinflation in early 1997, and in the context of negotiations with the IMF, the
newly elected president of Bulgaria engaged in discussions with major political
parties to forge a commitment to an economic program that included, inter alia,
the introduction of a currency board. The commitment resulting from the discus-
sions was formalized in a declaration proclaiming the readiness of the major polit-
ical parties to cooperate in fulfilling the agreement with the Fund. Elections in
April 1997 shifted political power to a new majority, but the negotiated program
continued to enjoy broad national ownership and was endorsed by the new
parliament.20

12.4.2 Case 2: Public Uncertainty about Government Objectives

We now focus on the issue of uncertainty about the government’s objectives in the
design of a program. That is, the government faces the problem of convincing the
public it is acting in the public’s best interests, rather than having its own agenda.
Though the imperfect information issues discussed in the previous case may still
be present (that is, where the government does not know the public’s prefer-
ences), we assume these away in order to focus on the problem of the government
making it credible to the public that it is acting in its best interests. We also con-
tinue to assume that the public is homogeneous (though not necessarily informed
about what policies would maximize their welfare).
If the public believes that the government’s preferences are not correlated with
its own, then cheap talk may fail to convey information. Specifically, suppose that
only the government knows the state of the economy, high or low, which the pub-
lic cannot observe, though it knows how the world works (that is, the true payoff
matrix). Let us interpret more and less as the public’s action (say, the level of
Can Public Discussion Enhance Program ‘‘Ownership’’? 309

wage demands). Public discussion (cheap talk), where the government announces
the state of the economy, precedes any actions.
We now assume that the government’s and the public’s objectives can differ,
which might reflect, for example, the weight government puts on the reaction of
global capital markets. Hence, suppose that instead of a payoff matrix such as ma-
trix 12.1, payoffs were described by matrix 12.4 (where the government’s payoff is
listed first).

Matrix 12.4
Public Behavior
State of
Economy More Less

High 1,3 5,1

Low
6,
3 3,2

In this case, the government would like the public to play less, no matter what
the economic situation is. Hence, government would like the public to believe that
the state of the economy is low. But the public, knowing the payoff matrix, knows
government has this incentive and so does not believe any government announce-
ment. Talking doesn’t reveal the government’s information because it is believed
to have an objective other than the maximization of social welfare—that is, to
have different preferences than the public.
However, even if the preferences of government and public are not perfectly
correlated, information can be conveyed, but the type of message sent will be cru-
cial in determining whether it is successfully conveyed. Suppose, more realistically,
that the state of the economy is not discrete as in matrix 12.4, but is continuous.
Suppose the payoffs are such that government wants the public to believe that
the state of the economy is somewhat worse than it really is (perhaps to give
them an incentive to accept lower wages as part of a tough program). For simplic-
ity, suppose that if the state of the economy was S, the government would want
the public to believe that the state was S
x, where this incentive is known to the
public. One might think that cheap talk can convey no information here, since if
the public discounts the government’s announcement by some amount y (that is,
in response to an announcement of S
x, public believes the state is S
x þ y),
government will simply announce S
x
y.
A key result in the cheap talk literature (Crawford and Sobel 1982) is that
imprecise messages can convey information as long as the incentive to distort
310 Allan Drazen and Peter Isard

information (in the example above, the amount x by which the government wants
to understate S) is not too large. That is, while a supposedly precise announce-
ment of the state of the economy will not be believed, a crude announcement of
the state of the economy, such as simply high or low, even when the state is con-
tinuous, will convey information.
This result has applications not only to cases in which the public is uncertain
about the government’s preferences, but also to cases in which the public knows
that the government has different preferences, as long as the difference in prefer-
ences is not too large. In such situations, trying to get the public to comply with
government-recommended policies may be difficult when the choice set includes
a continuum of policies (for example, different degrees of adjustment). However,
presenting a choice between two discretely different alternatives may allow the
government to credibly convey its superior information to a skeptical public and
thus gain acceptance of a program (roughly) tailored to the economic situation.
Such a situation is often encountered in Fund programs—say, in fiscal tighten-
ing in response to the state of the economy. Let the state of the economy be con-
tinuous, with high and low being very good or very bad states. The degree of
possible fiscal tightening is also continuous, but with the government and public
differing on what is optimal in each state of the economy, due to different dis-
count rates for evaluating the trade-offs between current costs and future benefits.
Both parties know that the public discounts the future more than does the govern-
ment. The government has more information than the public about the economic
situation, both ex ante or ex post, and the public’s support is required for fiscal
tightening.
The differences in discount rates give the government an incentive to overstate
the seriousness of the economic situation by a discrete amount, and the public in
turn is fully aware of government’s incentive. Suppose however that the public is
presented with a discrete choice—either government enacts no policy change or a
specific amount of fiscal tightening. (One may think of the choice of either accept-
ing or rejecting a specific Fund program). Accordingly, if the difference between
the discount rates of government and public is not too large, government’s char-
acterization of the economic situation (cheap talk) in combination with a choice
restricted to two options can be successful in inducing the public to accept the
proposed amount of fiscal tightening.
On the other hand, if the difference between government and public preferences
is too large relative to the outcomes associated with the two choices, or—to return
to the case where the public is uncertain about the government’s preferences—if
the public believes, rightly or wrongly, that the difference between the govern-
ment’s objectives and its own is sufficiently large (that is, that government has
Can Public Discussion Enhance Program ‘‘Ownership’’? 311

too strong an incentive to mislead the public), then cheap talk cannot convey
information.
When cheap talk provides no information, a government may need to use
costly signals to convey information about its preferences. For example, a govern-
ment whose objective is social-welfare maximization may reveal this by taking
actions that would not be mimicked by a government with other objectives, be-
cause the latter would find them too costly. In formal terms, this is the issue of
unobserved type, where one can think of two types of governments: a ‘‘good,’’
or social-welfare-maximizing government, and a ‘‘bad,’’ or non-social-welfare-
maximizing government. How can a government whose objective is social-
welfare maximization separate itself from one that has other objectives? For the
good government to care (and for the bad government to have an incentive to try
to masquerade as a good government), there must be a favorable response to the
government, or its program, being perceived as in the public’s interest. Such a re-
sponse is present when public support is required for programs to succeed and
when the public only complies with programs that it either perceives to be associ-
ated with a good government (that is, accepts because it is associated with a gov-
ernment it trusts) or is otherwise convinced to support.
Even when costly signals are used to reveal type, public discussion may play a
role. Jamaica provides an interesting example in which the government estab-
lished good credentials by taking the costly action of relinquishing IMF financial
support and by employing transparency (public discussion) to win support for a
homegrown macroeconomic program. Following three decades of heavy reliance
on Fund support from the 1960s through 1996, and while still facing very large
adjustment problems, the government declared its independence from the IMF,
essentially rejecting the Fund’s push for a devaluation to restore competitiveness
along with work-out measures for the financial sector. It opted instead for its
own strategy of tight fiscal and monetary policies, guarantees for bank deposits
and other liabilities, and a gradual approach to dealing with problem financial
institutions. The government also made a strong commitment to transparency by
agreeing to the publication of the Fund’s fairly critical Public Information Notice
following the 1997 Article IV consultation, along with subsequent Fund reports,
and by issuing its own commentaries in which it emphasized both points of
agreement and disagreement, thereby fostering public debate. Although it is dif-
ficult to judge whether the Jamaican government’s homegrown approach was
more appropriate than the Fund’s approach would have been in the absence of
political commitment, it is clear that the government succeeded in building a
strong degree of ownership for its homegrown program by putting its credibility
on the line, incurring the cost of foregoing Fund credit, and fostering public dis-
cussion through its transparency policy.
312 Allan Drazen and Peter Isard

When the government demonstrates its commitment to social welfare by choos-

ing policies favored by the public, another problem may arise. What if the author-
ities believe that the public does not necessarily know what the best policy is?
(The notion that a key function of public discussion is to educate the public about
economic realities presumes that this is often the case.) A government that cares
about social welfare faces a problem. Choosing the policy that the public says it
prefers enables the government to demonstrate that it has ostensibly the same
objectives as the public, but often at the cost of not actually maximizing the pub-
lic’s welfare! (This distinction is sometimes referred to as the difference between a
government being responsive to citizens’ expressed preferences and representative
of public preferences. See Manin, Przeworski, and Stokes 1999.) This suggests
that the only effective strategy that a ‘‘good’’ government can use to convince an
uncertain public of its type is to educate the public that it is indeed advocating a
good program—that is, a set of policies likely to generate attractive payoffs for
the public relative to the payoffs associated with other courses of action (or
inaction).
Genuine public discussion may serve a purpose here that cannot be achieved
through a simple announcement of policy or assertions about the economic real-
ities on which policy is based.21 The reason is that the requirement that positions
be argued and justified to the public (either directly or by convincing experts who
have credibility with the public) will eliminate some arguments as convincing
rationales for policy. Exposing the (supposed) rationale for policy to the cold light
of open public debate and scrutiny can often reveal whether or not the rationale is
really social-welfare maximization. (Fearon 1998 provides a nice discussion of this
point.) Public debate can thus serve the purpose of demonstrating unbiasedness
or revealing government type. The willingness to engage in public discussion
may itself signal that the government’s objective is the maximization of social wel-
fare, independent of any information that public discussion reveals. Good pro-
grams may face a lower cost of exposure to public discussion as they are more
likely to survive public scrutiny. Public discussion may thus have different costs
for different types of governments, or different types of programs.22

12.4.3 Case 3: The Public is Heterogeneous

In the third case, we focus on the role of public discussion when groups in society
have conflicting interests. Conflict among members of a society is central to politi-
cal economy in general (Drazen 2000), and is a key problem in formulating reform
and stabilization programs. Though the problem of the public having imperfect
information about the government’s preferences may be important here as well,
we abstract from it to focus more exclusively on how public discussion might
Can Public Discussion Enhance Program ‘‘Ownership’’? 313

help hammer out an acceptable program when the public is heterogeneous. For
the same reason, we also assume that the problem is not that the government
must make credible that its objective is social-welfare maximization. (Of course,
there are interesting questions here—for example, the government must make
credible it is not too closely aligned with one interest group, such as the financial
sector.)
Suppose the public is divided into three groups: the general public, and two
special interest groups (SIGs) denoted ‘‘Green’’ and ‘‘Blue,’’ who must consent for
a program to be adopted. The general public is unorganized and lacks the power
that the SIGs have to block a Fund program. The government may be seen as
maximizing the welfare of the general public. Here we focus on public discussion
as revealing information about SIGs or coordinating their actions. It might also
weaken their ability to achieve selfish aims by making it necessary for them to jus-
tify their demands, analogous to the discussion at the end of section 4.2.
Getting the approval of SIGs may be seen as a coordination problem when it is
not known exactly what they might be willing to concede (though their general
preferences are known) or when there is initial disagreement about how the
world works. Let’s start with a simple reference case. Each group must simultane-
ously choose an action, and each perceives that the payoffs depend on the differ-
ent combinations of actions. As in matrix 12.4, public discussion—that is, an
announcement by both Blue and Green—precedes any actions.

Matrix 12.5
Green

Blue Left Right

Up 3,3 1,1

Down 1,1 3,3

Here, cheap talk works as a coordination device because there is no real conflict
of interest. If Blue announces that his interest is down, then Green will clearly
want to say right and we get to a better solution than we could expect without
communication. In formal terms, Blue’s message is both self-signaling and self-
committing. It is self-signaling in that Blue wants to announce down if and only
if it is best. It is self-committing in that if Green believes Blue (which we just
argued it should), the announcement creates the incentive for Blue to actually
carry it out. Public discussion has the function of simply informing groups of
314 Allan Drazen and Peter Isard

where there may be common ground. If there is, then the program adopted will
be superior to what would be adopted in its absence.
Of course, this example is too simple, because there is no conflict of interest,
only a problem of coordination. Farrell and Rabin (1996) point out that when the
interests of SIGs are not so potentially well aligned, cheap talk messages are less
likely to be self-signaling or self-committing. A simple example is the well-known
‘‘prisoner’s dilemma,’’ where the payoffs are as in matrix 12.6. Cheap talk pre-
cedes simultaneous action by the two groups.

Matrix 12.6
Green

Blue Left Right

Up 7,7 4,8

Down 8,4 5,5

Here, for Blue it is optimal to choose down for any choice Green makes, while
for Green it is optimal to choose right for any choice Blue will make. Cheap talk
will not allow them to coordinate. Messages are not self-committing. Though one
may argue that repeated play of the game by the same players may lead to a co-
operative solution, and that is what is seen in experimental studies,23 the basic
message is clear: if SIGs see their interests as sufficiently in conflict in relation to a
stabilization program, public discussion in the sense of simply exchanging infor-
mation may have limited value.
An argument put forward in discussing case 2 is relevant here as well. Public
discussion that requires interest groups to justify their demands may eliminate
some claims because they are unjustifiable in a setting of genuine, open discus-
sion. Forums to elicit public input and reaction to proposed programs may be
effective in revealing useful information only to the extent that they are not too
warm and fuzzy. More exactly, eliciting credible information about preferences of
groups is not easy once the groups know they have strongly conflicting objectives.
The situation is somewhat different when government and IMF actions (includ-
ing actions other than public discussion) can influence the payoffs that SIGs asso-
ciate with different outcomes. When the interaction is among SIGs, but the
authorities can change the payoff matrix, they could, for example, transform a
prisoner’s dilemma as in matrix 12.6 to a coordination game as in matrix 12.5.
More concretely, the ability to modify proposed programs and sweeten the deal
Can Public Discussion Enhance Program ‘‘Ownership’’? 315

for interest groups can be seen in this light. In the process of drafting legislation
that can get through national parliaments, it is common to add amendments to
win the support of specific legislators or groups. Fund programs are often struc-
tured with similar political constraints in mind. When the ownership problem is
one of getting the political consent of powerful groups, building ownership may
in fact be identified with cases in which the cooperative outcome is made credible,
as distinct from cases in which it is not. We return to this issue in section 6.

12.4.4 Public Discussion and Democratic Values

A very different, but important point is that public discussion can be instrumental
in strengthening democratic values, independent of the outcome of specific dis-
cussions. Encouraging public discussion on one type of issue may increase the
likelihood that public discussion is used in other contexts as well. If the political
culture is such that government decisions are not generally subjected to public
discussion—if the paradigm of deliberative democracy is not well developed—
then public discussion of economic policy can have a valuable learning effect.
This may well be the case in new democracies.24 An outside body, such as the
Fund, may be effective in fostering such a discussion in countries where this tradi-
tion is weak or nonexistent.

12.5 Drawbacks of Public Discussion

Public discussion may have significant drawbacks as well. Przeworski (1998) and
Stokes (1998) have argued, for example, that when public discussion is followed
by voting, lobbyists and interest groups may have incentives to mislead voters
during public discussion, such that voting actually becomes less informed rather
than more informed. The possibility that public discussion could cause people to
be misled by SIGs is also present in the context of program design, but we don’t
think that this is the central problem of public discussion of Fund programs. In
fact, the key drawbacks of public discussion in the context of adopting a Fund
program may be quite different than the drawbacks of discussion as a prelude to
voting.
In the case of public acceptance of a Fund program, one problem is that public
discussion may essentially give interest groups more veto power over a proposed
program, or a heightened awareness of adverse implications and a stronger incen-
tive to use their veto power. To the extent that public discussion simply informs
governments about the public’s true preferences, as in case 1 (where the public as
a whole is seen as homogeneous), discussion improves program design. But when
the public includes strong SIGs whose interests are not those of the general public
316 Allan Drazen and Peter Isard

(case 3), there is a danger that a policy-making process that gives them too large a
voice will result in a program that does not serve the public good. Social welfare
may thus be higher when the authorities make take-it-or-leave-it offers rather
than allowing public discussion to give SIGs too much influence. This has been a
criticism, for example, of giving NGOs a large say in program design, where the
implicit assumption is that their interests do not necessarily reflect the broader
public interest.
Second, public discussion can be time-consuming and may slow down the pro-
cess of shaping a program. This is potentially quite a serious problem when the
economy is in crisis and a program needs to be put in place quickly. The problem
of having exactly the right amount of discussion is obviously a very difficult one
that probably is not conducive to any general rules. The right amount of public
discussion will be very much situation-specific.
Third, since programs may go through a lot of changes, making discussion pub-
lic will involve groups (and the government) in espousing positions that may sub-
sequently be rejected and even become seen as incorrect. The costs of being put in
such a situation may induce groups to refrain from putting forward public posi-
tions at all, or to take positions in public that differ from the positions they would
put forward in private. Hence, public discussion could actually greatly hinder in-
formation transmission relative to more confidential means of discussion.
Fourth, since the discussion of Fund programs tends to focus attention on a
country’s bad economic situation, making information public may in itself make
the situation even worse. Some types of information may spook financial markets
and lead to major capital outflows, making it even harder to design a successful
program. This appears to be the main concern expressed in relevant IMF Board
meetings by those executive directors who oppose full transparency—that is, full
revelation of the Fund’s information and/or concerns about countries (IMF
2003c).
None of these concerns is easy to address, since they are all both genuine and
situation-specific (and hence hard to address in any generality). They should not
be taken as arguments against public discussion per se, but only as cautions in
considering how public discussion of a program should be structured.

12.6 The Design of IMF Programs

The importance of national ownership is clearly recognized in the principles gov-

erning the design of Fund programs. The Fund’s Guidelines on Conditionality
(IMF 2002), as revised in September 2002, indicate that national authorities have
the lead role in shaping program documents25 and that the Fund should encour-
age countries to build broad support for sound policies. Moreover, in association
Can Public Discussion Enhance Program ‘‘Ownership’’? 317

with the Guidelines, the set of general operational instructions provided to Fund
staff engaged in program design is based on five key principles: ‘‘national owner-
ship of reform programs; (ii) parsimony in program conditions; (iii) tailoring of
programs to a member’s circumstances; (iv) effective coordination with other mul-
tilateral institutions; and (v) clarity in the specification of conditions’’ (IMF 2003a,
paragraph 2).
In moving from principles to procedures, the operational instructions empha-
size that Fund staff ‘‘should seek proposals from national authorities at an early
stage in the policy dialogue.’’ They also stress that Fund staff ‘‘should encourage
the authorities to engage in a transparent participatory process in developing a
policy framework, and should continue to be prepared to assist the authorities in
this process by giving seminars, meeting with various interest or political groups
(parliamentary committees, trade unions, business groups, etc.) and by being
available to the media. . . . [while being mindful] of the authorities’ views on staff
contact with domestic groups. . . . ’’ In addition, the operational guidance note
stresses that documents prepared by Fund staff in the course of briefing the
Fund’s management and reporting formally to the Executive Board should assess
the challenges to broad ownership, including key capacity weaknesses and issues
relating to political structures. It also clarifies the principle of parsimony in pro-
gram conditions: performance criteria, prior actions, and other program condi-
tions must be limited to those that, if excluded, ‘‘would seriously threaten the
achievement of program goals or the Fund’s ability to monitor implementation’’
(IMF 2003a, paragraphs 6 and 7).
The emphasis on ownership and transparent participatory processes in the
recently revised Fund Guidelines and guidance notes is consistent with the posi-
tion of the Fund’s Independent Evaluation Office (IEO). In its evaluation of cases
in which Fund programs have not succeeded, the IEO concluded that the extent
and structure of program conditions was much less important than securing an
underlying commitment to core policy adjustments. It consistently suggested that
the aim in program design should be to move as quickly as possible to a situation
in which the core elements of a program are subject to a policy debate within the
country’s own policy-making institutions, and that the Fund staff should actively
seek to present policy options, analyze the tradeoffs between them, and encourage
open debate on the alternatives (Independent Evaluation Office 2002).
These recommendations are supported by case studies of Pakistan, the Philip-
pines, and Senegal, where lack of political commitment was a major factor in pro-
gram failures, and of Morocco and Jamaica, where a ‘‘real difference seems to
have been made by strong domestic ownership’’ (Independent Evaluation Office
2002, 16–18).
318 Allan Drazen and Peter Isard

In the case of Senegal, despite significant progress during the 1980s in moving
the economy away from excessive state intervention, roughly one-third of the
measures envisioned under World Bank structural adjustment loans were not
implemented as scheduled. These measures were concentrated in areas such as
labor regulations, where there was strong opposition from vested interests and
where the government apparently made little prior effort to generate public dis-
cussion and reach consensus on key issues. Senegal’s failure to pursue adequate
public discussion also contributed to a lack of clarity and progress in its efforts to
restructure the groundnut sector during the 1980s, which was a source of income
for the majority of the rural population and a key sector in the effort to reduce
poverty.26
By contrast, in Morocco, where evaluation reports have found no major differ-
ences in the approach to economic program design from that followed in other
countries with IMF programs, ‘‘[i]ncreasing transparency in putting information
and policies out for public discussion . . . appears to have helped develop a
broader consensus [for reform]’’ and was a critical factor in weaning the country
from prolonged reliance on IMF loans (Independent Evaluation Office 2002, 196).
Despite the heightened awareness of the importance of country ownership and
the relevance of public discussion (participation), most references to public dis-
cussion or participation in Fund documents are cast in broad terms, with little or
no explicit recognition of the different functions that public discussion can serve,
and little or no explicit focus on which specific functions it can usefully serve in
specific circumstances. Of course, Fund staff rarely engage in public discussion
without advance brainstorming or other forms of preparation, which often is not
reflected in any written documents. And it would be misleading to suggest that
Fund staff engage in public discussion without a fairly clear implicit sense of
what they are trying to achieve. Nevertheless, the effectiveness of public discus-
sion in building and assessing ownership of Fund programs might be signifi-
cantly enhanced by greater awareness of the range of functions that public
discussion can serve and by a more systematic focus on strategies for engaging in
public discussion in specific circumstances.
Although public discussion can enhance ownership in various ways, it would
be misleading to suggest that discussion alone can induce complete ownership of
Fund programs. Many of the countries that seek to negotiate Fund programs re-
quire fiscal adjustment to restore and maintain macroeconomic stability. Because
such adjustment necessarily leaves some interest groups worse off in the short
run, complete ownership of the fiscal components of Fund programs is generally
difficult to achieve. Nevertheless, public discussion in this context—with sufficient
efforts to keep it well focused—can be particularly important for building owner-
ship by educating the public about the overall economy-wide benefits of fiscal ad-
Can Public Discussion Enhance Program ‘‘Ownership’’? 319

justment in the short run, about prospects that the benefits will be widely shared
over the medium run, and about the gains from a cooperative solution that elic-
its financial support from the Fund and other sources. It can also contribute to
building ownership by helping governments find the type of cooperative solu-
tion or policy mix that stands the best chance of sustaining political support—
particularly so when the Fund gives policy makers broad freedom to design the
specific details of the fiscal adjustment effort. In this context, it may be very im-
portant for the Fund to refrain from pressuring governments to adopt policies
that are appealing on efficiency grounds but typically provoke strong public re-
sentment, such as raising or eliminating ceilings on the price of necessities.
Despite these various ways that public discussion can enhance ownership of fis-
cal adjustment policies, it cannot overcome the fact that various interest groups
may continue to seek to benefit in the short run by undermining the fiscal adjust-
ment effort. This is why the effectiveness of many Fund programs depends criti-
cally on appropriate prior conditions, fiscal performance criteria, and the phased
provision of Fund credit.27
Although public discussion on its own cannot achieve complete ownership of
programs for countries in which the restoration of macroeconomic stability
requires fiscal adjustment, Fund programs need not be (and generally are not)
limited to policy actions that impose short-run costs. Indeed, one approach to
enhancing the overall ownership of Fund programs is to try to counteract condi-
tions that impose short-run costs with conditions that the public correctly per-
ceives to convey major benefits. In this context, public discussion, through many
of the functions it can serve (eliciting information about what the public wants,
educating the public, educating the government and the Fund, revealing that the
government is indeed oriented toward maximizing social welfare, and leading
heterogeneous interests to a cooperative solution), may well be able to strengthen
the overall ownership of programs by building virtually complete support for
various types of structural measures (such as clarification of property rights or
strengthening of tax-collection mechanisms and accounting standards) that do
not impose costs on any powerful interest groups, at least in the context of a
good government where there is little or no inherent incentive to oppose impor-
tant growth-enhancing structural reforms.

12.7 Conclusions

There has been growing recognition in recent years that the effectiveness of IMF-
supported programs depends heavily on the degree of country ownership, and
that ownership can be promoted by seeking to broaden and deepen the base of
support for sound policies among a country’s domestic interest groups. These
320 Allan Drazen and Peter Isard

perceptions are reflected both in the September 2002 revision of the Fund’s Guide-
lines for Conditionality and in the international community’s decision to link debt
relief for highly indebted poor countries (HIPCs) to the formulation of poverty
reduction strategies through processes that involve the broad participation of
stakeholders.
This paper has focused on public discussion as one potentially important vehi-
cle for enhancing program ownership (the functions of public discussion are sum-
marized at the beginning of section 3). The motivation comes from our sense that
economists do not yet have a very clear understanding of the various channels
through which public discussion can work, or of the circumstances in which pub-
lic discussion can be effective. This lack of understanding has contributed to the
view—held in some quarters—that public discussion strengthens ownership pri-
marily by leading counterproductively to the adoption of weaker programs.
More seriously, failure to appreciate and distinguish between the range of func-
tions that public discussion can serve and the circumstances that determine its ef-
fectiveness or ineffectiveness implies that country governments and the Fund may
not be exploiting the potential of public discussion in the most effective ways. A
better understanding of public discussion can contribute to the effort to design
strong programs that can command broad country ownership. We have argued,
primarily by way of example in illustrating the functions that public discussion
may serve, that it can in fact be an important tool in raising the probability of pro-
gram success.
It would be wishful thinking, however, to suggest that public discussion alone
can lead to complete ownership of a program that imposes significant costs in
the short run. Those interest groups that stand to incur the short-run costs pose
significant risks to the adjustment effort in the absence of appropriately structured
conditionality, which provides financial incentives for countries to remain in com-
pliance with their policy commitments. Moreover, as we argued in section 5, there
are drawbacks to public discussion, especially in the presence of strong interest
groups.
By contrast, public discussion may be able to achieve virtually complete owner-
ship of growth-enhancing structural reforms that impose no significant costs in
either the short run or the longer run. Accordingly, public discussion has the po-
tential to contribute importantly to growth by inducing the public to want to pur-
sue more growth-enhancing structural reforms as part of their policy programs. In
principle, program conditionality is not necessary for achieving compliance when
there is complete ownership of reforms, although in practice it may be useful for
spurring progress.
Ironically, the efforts that the Fund has made to build program ownership by
streamlining conditionality in recent years appear to have shifted the focus of
Can Public Discussion Enhance Program ‘‘Ownership’’? 321

Fund programs away from the types of growth-enhancing structural policies on

which ownership would seem relatively easy to build. We attribute this in part to
the limited attention that economists have paid to the concept of ownership and
to the process of building ownership through public discussion.

Acknowledgments

This paper was prepared for the IMF Conference in honor of Guillermo Calvo,
held in Washington, D.C., April 15–16, 2004. We wish to thank our discussant
Raquel Fernandez, as well as Jim Boughton, Eddie Dekel, Thomas Dorsey, Mar-
cela Eslava, Judith Gold, Caroline Kende-Robb, and Nuno Limão for helpful com-
ments and discussions. Parts of this paper were written while the first author was
visiting the IMF, which he thanks for its hospitality. The views expressed in the
paper are those of the authors and should not be attributed to the IMF.

Notes

1. For readers unfamiliar with Fund jargon, the term completion date refers to the date at which the bor-
rowing country becomes eligible for the last installment of the IMF loan, based on demonstrated com-
pliance with, or agreed waivers or modifications of, the performance criteria and other conditions of
the program. Out of the 615 Fund programs approved between 1973 and 1997, only 70 percent
achieved their originally agreed completion dates (in many cases with modifications along the way).
An additional 12 percent were extended beyond their original durations, another 11 percent were can-
celled early but followed promptly by successor arrangements, and 7 percent were effectively sus-
pended; see Mussa and Savastano (2000).
2. Under a definition that treats a country as a prolonged user of IMF resources if it has been under
IMF programs for seven or more years within a ten-year period, forty-four countries were prolonged
users during some part of the 1971–2000 period; see Independent Evaluation Office (2002).

3. IMF (2002).
4. See Boughton and Mourmouras (2002).
5. It is not only to the IMF that ownership must be demonstrated. Demonstration of ownership may
also have a catalytic effect in inducing both official bilateral lending and private capital flows, which
exceed or add substantially to official multilateral loans.
6. See, for example, IMF (2003a), paragraph 7.

7. This distinction between country and government (or authority) ownership is central to the argu-
ments on conditionality presented in Drazen (2002) and is explored in detail there.

8. Fund involvement might be crucial in inducing complete country ownership (as in the use of condi-
tionality in Drazen 2002), so that one could observe Fund programs where country ownership appears
to be complete (ex post).
9. Farrell and Rabin (1996) present an excellent, easily accessible introduction to the subject.
10. From a formal point of view, political scientists have also used cheap talk models to study, for ex-
ample, communication in legislatures.
322 Allan Drazen and Peter Isard

11. Some critics argue that this transformation may actually be a distortion of individual preferences.
See section 12.5 below.
12. The reader will probably quickly point out that individuals may view income inequality not as it-
self the outcome, but rather in terms of its effect on other variables, such as growth. That is, individuals
may both favor high growth, but differ on the connection between income inequality and growth. For
illustration, see the discussion in chapter 11 of Drazen (2000).
13. Financial constraints on the IMF, for example, may lead its objectives to differ from those of a coun-
try’s authorities. See Drazen (2002) for a discussion of some of the issues involved in whether the gov-
ernment and the IMF have the same or different objectives.
14. ‘‘Common knowledge’’ means not only that something is known by all parties, but that everyone
knows that everyone knows, everyone knows that everyone knows that everyone knows, and so on.
15. By contrast, the methods that Wimmer, de Soysa, and Wagner (2002) suggest for assessing the
feasibility and sustainability of reforms would appear to require considerable resources and time to
implement.
16. When the public consists of groups with conflicting interests, then public discussion meant to elicit
information on preferred policy may not serve the same function, since groups have the incentive to re-
veal information strategically. We return to this problem in section 12.4.3.

17. Since programs are economically complex, the public will never fully understand all the details.
Hence, there must be a degree of trust in the government. This can be attained by generating a feeling
of inclusion in the process.
18. See Robb (2003), box 3.
19. See IMF (2003b).

20. See IMF (2001).

21. In the cheap talk examples, there is no real discussion or deliberation—the informed party sends a
single message to the uninformed party. This is not unique to these examples. As Aumann and Hart
(2003, 1619) point out, ‘‘Most formal models in this area [strategic information transmission] allow for
at most one message from each player.’’ They consider multistage cheap talk (‘‘long cheap talk’’), but
the application of their results to public discussion is an issue yet to be explored.
22. However, the signal is less than perfect, both because some regimes seem adverse to public discus-
sion per se, and because public discussion may have adverse effects, as discussed in the next section.
23. It is argued that players in situations like this often play cooperatively, and that communication
may have a coordinating effect. See footnote 14 in Farrell and Rabin (1996).

24. Brender and Drazen (2005) find that political deficit cycles are present only in new democracies
and are not statistically significant in established democracies as a group. They suggest that this
observed difference may reflect in part the lack of availability and analysis of information about the
fiscal process. They also find that as a country gains experience with democracy, the political cycle dis-
appears. See also Akhmedov and Zhuravskaya (2004) for evidence of this effect in Russia after the tran-
sition to democracy.
25. There is no requirement that country authorities draft program documents, but rather the directive
that the Fund staff be responsive when the authorities desire a greater role in the drafting process.
26. Independent Evaluation Office (2002), chapter 11, which cites reports of the World Bank’s Opera-
tional Evaluation Department.
Can Public Discussion Enhance Program ‘‘Ownership’’? 323

27. The time dimension of Fund programs provides an interesting direction for further thought on the
role of public discussion. In particular, given that the Fund is often criticized for being too willing to
waive or modify performance criteria (not being a sufficiently ‘‘tough cop’’), it is interesting to consider
whether public discussion that starts at the time of the initial request for financing can strengthen com-
pliance with program conditions by making it more credible that the Fund will say ‘‘no’’ on test dates
for subsequent installments of financing when governments breach the conditions established at the
outset of their programs.

References

Akhmedov, A., and Zhuravskaya, E. 2004. ‘‘Opportunistic Political Cycles: Test in a Young Democracy
Setting.’’ Quarterly Journal of Economics 119: 1301–1338.
Aumann, R., and S. Hart. 2003. ‘‘Long Cheap Talk.’’ Econometrica 71: 1619–1660.

Boughton, J., and A. Mourmouras. 2002. ‘‘Is Policy Ownership An Operational Concept?’’ Working Pa-
per WP/02/72, IMF, Washington, D.C.

Brender, A., and A. Drazen. 2005. ‘‘Political Budget Cycles in New versus Established Democracies.’’
Journal of Monetary Economics 52: 1271–1295.
Crawford, V., and J. Sobel. 1982. ‘‘Strategic Information Transmission.’’ Econometrica 50: 1431–1451.

Drazen, A. 2000. Political Economy in Macroeconomics. Princeton, NJ: Princeton University Press.
———. 2002. ‘‘Conditionality and Ownership in IMF Lending: A Political Economy Approach.’’ IMF
Staff Papers 49 (Special Issue).
Elster, J., ed. 1998. Deliberative Democracy. Cambridge, UK: Cambridge University Press.
———. 1998. ‘‘Introduction.’’ In Deliberative Democracy, 1–18. Cambridge, UK: Cambridge University
Press.
Farrell, J., and M. Rabin. 1996. ‘‘Cheap Talk.’’ Journal of Economic Perspectives 10: 103–118.
Fearon, J. 1998. ‘‘Deliberation as Discussion.’’ In Deliberative Democracy, J. Elster, ed., 44–68. Cambridge,
UK: Cambridge University Press.
Habermas, J. 1987. Theory of Communicative Action. Boston: Beacon Press.
Independent Evaluation Office. 2002. Evaluation of the Prolonged Use of Fund Resources. Washington: In-
ternational Monetary Fund. Available at www.imf.org.
International Monetary Fund. 2001. ‘‘Strengthening Country Ownership of Fund-Supported Pro-
grams.’’ Policy Development and Review Department. Available at http://www.imf.org/External/
np/pdr/cond/2001/eng/strength/120501.pdf.

International Monetary Fund. 2002. ‘‘Guidelines on Conditionality.’’ Legal and Policy Development
and Review Departments. Available at http://www.imf.org/External/np/pdr/cond/2002/eng/
guid092302.pdf.
———. 2003a. ‘‘Operational Guidance on the New Conditionality Guidelines.’’ Available at http://
www.imf.org/External/np/pdr/cond/2003/eng/050803.htm.
———. 2003b. ‘‘Chad: Poverty Reduction Strategy Paper.’’ Country Report No. 03/209, IMF, Washing-
ton, D.C.
324 Allan Drazen and Peter Isard

———. 2003c. ‘‘IMF Reviews the Fund’s Transparency Policy—Issues and Next Steps.’’ Public Infor-
mation Notice (PIN) No. 03/122. Available at http://www.imf.org/external/mp/sec/pn/2003/
pn03122.htm.
Manin, B., A. Przeworski, and S. C. Stokes. 1999. ‘‘Elections and Representation.’’ In Democracy, Ac-
countability, and Representation, eds. A. Przeworski, S. C. Stokes, and B. Manin, 29–54. Cambridge, UK:
Cambridge University Press.
Mussa, M., and M. Savastano. 2000. ‘‘The IMF Approach to Economic Stabilization.’’ NBER Macroeco-
nomics Annual 1999. Cambridge: MIT Press: 79–122.
Przeworski, A. 1998. ‘‘Deliberation and Ideological Domination.’’ In Deliberative Democracy, ed. J. Elster,
140–160. Cambridge, UK: Cambridge University Press.

Robb, C. M. 2003. ‘‘Poverty and Social Impact Analysis—Linking Macroeconomic Policies to Poverty
Outcomes: Summary of Early Experiences.’’ Working Paper WP/03/43, IMF, Washington, D.C.
Stokes, S. C. 1998. ‘‘Pathologies of Deliberation.’’ In Deliberative Democracy, ed. J. Elster, 123–139. Cam-
bridge, UK: Cambridge University Press.
Wimmer, A., I. de Soysa, and C. Wagner. 2002. ‘‘Political Science Tools for Assessing Feasibility and
Sustainability of Reforms.’’ Paper prepared for the Independent Evaluation Office, International Mone-
tary Fund, Washington, D.C.
V Transition and Growth
13 Sources and Obstacles for
Growth in Transition
Countries: The Role of Credit

Fabrizio Coricelli, Bostjan Jazbeč,

and Igor Masten

13.1 Introduction

One of the distinguishing features of planned economies was the total irrelevance
of credit and finance for the determination of output. Market reforms gave credit
and finance a central role in economic activity. Guillermo Calvo was among the
first economists to emphasize the role of credit and financial markets in the transi-
tion process (Calvo and Coricelli 1992, 1993), arguing that the initial collapse in
output could be interpreted as a trade implosion due to the sudden dry-up of
financing for firms, induced by the initial liberalization and stabilization pro-
grams. Following such initial fall in output, transition economies moved along a
path of relatively fast growth, although with significant differences across coun-
tries. This phase of growth has been accompanied by progress in financial devel-
opment. However, the depth of credit markets remained low, even though
during the period 2000–2005 credit has grown rapidly in several transition coun-
tries, leading in some cases to credit booms.
In this paper we identify some causes for the continuing underdevelopment of
financial markets and explore the implications of such underdevelopment for
growth in transition countries. Through an econometric analysis of microdata for
a relatively large set of countries, including both advanced European and transi-
tion countries, we find that financial sector development, through an increase in
the depth of both bank credit and stock markets, would induce large growth
effects. We also find that the impact of financial sector development on growth
seems to be larger in transition rather than in mature industrial countries. This
suggests that growth effects are larger at lower levels of development of a finan-
cial sector, a condition characterizing most transition countries. We interpret this
evidence as suggesting that financial sector development affects growth mainly
through a softening of liquidity constraints on enterprises. This belief is strength-
ened by the additional finding that trade credit plays a much more relevant role
in substituting for official lines of financing in transition countries.
328 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

The paper is structured as follows: section 13.2 provides some evidence on the
level of financial development of transition countries and emphasizes the role of
institutional variables in explaining underdevelopment of financial markets in
transition countries. Section 13.3 reviews the channels through which credit mar-
kets could affect output growth, discussing the role of bank credit, trade credit,
and self-financing. Section 13.4 contains the empirical results of econometric anal-
ysis on industry-level data for a sample of countries including both advanced Eu-
ropean market economies and transition countries, for a ten-year period starting
in 1995. We show that progress in credit market development, through the con-
vergence of the depth and efficiency of credit markets toward the level prevailing
in advanced market economies, could have a major impact on growth in transi-
tion countries. The impact of financial market development on transition countries
turns out to be much larger than for more advanced market economies. We show
that this result may be due to the presence of threshold effects, related to the ini-
tial level of financial sector development. Indeed, financial sector development
has a stronger effect in countries starting from underdeveloped financial markets.
Furthermore, trade credit is found to be a relevant substitute for bank credit in
countries at the lower end of the financial sector development Spectrum. Section
5 concludes.

13.2 Financial Sector Development in Transition Countries

In spite of rapid growth, especially in the Baltic countries and in Southern-Eastern

European countries (SEE), financial markets remain shallow in transition coun-
tries (table 13.1). However, there are significant differences across countries, with
CEB displaying much higher levels of financial market development with respect
to CIS countries, with SEE in the middle.1 If we measure financial development
against the level of economic development, summarized by the level of GDP per
capita at purchasing power parity, we can identify the presence of a transition
gap. Indeed, most transition countries are placed below the regression line that
predicts the levels of financial development as a linear function of per capita
incomes (figure 13.1). Similar results have been obtained in studies that take into
account other factors, in addition to incomes per capita, that may affect financial
development (Cottarelli, Dell’Ariccia, and Vladkova-Hollar 2003).
The reasons for the underdevelopment of financial markets probably have to
do with the initial design of liberalization and reform policies and with objective
difficulties in developing financial markets in the midst of enormous structural
change and transformation of the economy.
Partly stimulated by the literature on transition, it is now acknowledged that
institutional development plays a key role in macroeconomic performance, both
Sources and Obstacles for Growth in Transition Countries 329

Table 13.1
Domestic Credit to Private Sector and Stock Market Capitalization, 2005 (in percent of GDP)
Domestic credit to Stock market
private sector capitalization
2000 2005 2000 2005
Albania 3.0 10.3 na na
Armenia 7.1 8.0 1.3 0.9
Azerbaijan 5.9 9.5 0.1 na
Belarus 8.6 1.0 4.1 na
Bosnia and Herzegovina 5.6 22.6 na na
Bulgaria 11.6 26.0 4.8 20.1
Croatia 36.0 55.6 14.5 35.2
Czech Republic 35.7 16.8 19.3 31.8
Estonia 24.0 61.8 32.4 26.5
FYR Macedonia 10.5 18.6 0.2 11.4
Georgia 6.4 9.5 0.8 5.5
Hungary 30.1 51.1 25.8 31.9
Kazakhstan 11.2 26.7 7.5 21.6
Kyrgyz Republic 11.2 8.0 0.3 1.8
Latvia 18.1 67.4 7.4 17.4
Lithuania 10.0 34.0 13.9 31.8
Moldova 12.6 21.2 30.3 na
Poland 20.8 23.4 17.9 31.6
Romania 7.2 10.2 3.4 22.3
Russian Federation 13.3 25.7 15.3 71.9
Slovak Republic 33.6 26.8 3.5 9.5
Slovenia 38.5 56.9 13.6 23.8
Tajikistan 19.2 17.1 na na
Turkmenistan 2.1 1.4 na na
Ukraine 11.2 31.2 6.0 31.3
Uzbekistan 27.9 20.4 1.0 0.3

CEB 26.4 42.3 16.7 26.0

SEE 12.3 23.9 5.8 23.0
CIS 8.4 9.4 6.7 17.0
World 45.7 55.8 53.2 57.7
European Union 74.4 85.8 78.7 67.0
Brazil 34.9 33.7 37.6 54.6
Philippines 40.6 32.7 68.0 33.5
Source: World Bank, EBRD banking survey, capital markets survey.
330 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

Figure 13.1
Development of Credit Markets and Incomes Per Capita

on growth rates and on their volatility (Roland 2000 and Acemoglu et al. 2002,
among others). One of the main channels through which institutions affect growth
is that of financial markets. The latter are indeed extremely sensitive to institu-
tional design.2
How can we explain such a transition gap? A recent study by Djankov,
McLiesh, and Shleifer (2007) provides a useful reference for computing the causes
of such a gap, as it analyzes the determinants of credit-to-GDP ratios for the larg-
est sample of countries considered so far in the literature (129 countries). The
study considers macroeconomic factors (such as inflation and GDP growth), fixed
costs in setting up financial markets (which give an advantage to large economies
in terms of the absolute size of their GDP), and institutional variables, related to
information available, through credit registries, the protection of creditor rights,
and inefficiency in the legal procedures for recovering credit in cases of default.
These institutional variables, especially information and creditor rights, explain
much of the cross-country differences in credit ratios. Therefore, further progress
in institution building in these areas appears to be the main avenue to reducing
the gap in financial sector development in transition countries.
Sources and Obstacles for Growth in Transition Countries 331

Transition countries offer a unique vantage point for analyzing the role of insti-
tutions as they went through a process of radical institutional change, as most
institutions necessary for the functioning of a market economy were absent in the
previous regime. Liberalization policies left the determination of output to decen-
tralized decisions in a market with free prices. At the same time, firms had to ob-
tain in the market the required financing for their operations. Financial markets,
however, were completely absent in the previous regime and could not be devel-
oped instantaneously.
Transition countries went through a process that reversed the developments
typical for market economies. In the history of capitalism, the development of fi-
nance preceded the development of industry. Most financial contracts were devel-
oped to support the activity of merchants. When the industrial revolution took
place, there was financial capital available to sustain the growth of industry. By
contrast, planned economies were characterized by very developed industrial sys-
tems. Indeed, most planned economies had an excessive weight of industry in the
economy. In contrast, finance and money were a ‘‘veil’’ for the enterprise sector.
They did not play any active role. The real equilibrium was determined by plan-
ning on real variables, and banks served as pure accounting institutions. There
was no accumulation of information and skills to assess credit-worthiness. The
same concept of credit-worthiness was irrelevant, as there was no possibility of
bankruptcy. This was the financial sector that most transition countries had when
they started their reforms. Whatever the final verdict on what caused the initial
collapse in output, it is hard to dispute that there was a fundamental friction in
the economy between the possibility of determining a decentralized equilibrium
and the total absence of financial markets. Transition countries had to simultane-
ously develop financial markets and effect an enormous reallocation of resources
in the economy. Following the initial shock, financial markets had to be created
through a combination of policy intervention and spontaneous developments in
the economy. Several features of the adjustment process after the initial shock
were linked to the essence of finance. In some countries private credit markets
developed through trade credit, in others barter trade became predominant, and
in others generalized payment default through interenterprise arrears emerged.
During this second phase, the experience of transition countries was highly heter-
ogeneous. To simplify, CEB were able to drive through the initial crisis and estab-
lish a relatively well-functioning market economy.
In the countries of the former Soviet Union, the output collapse was more per-
sistent, resembling a bad equilibrium, a distant relative of a market economy.
However, in recent years several CIS countries, including Russia, have begun to
grow very rapidly.
332 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

Whether these two different paths were caused mainly by different macroeco-
nomic policies or by different paths of institutional change and different initial
conditions has been debated. In the review of the debate in Campos and Coricelli
(2002), different initial conditions in terms of a minimal structure of market insti-
tutions and the subsequent impetus given by the attraction of Europe played a
crucial role.
Before turning to the empirical analysis of the effect of finance on growth in
transition countries, we next briefly sketch a simple framework that may help
to highlight the main channels through which finance may affect growth in
transition.

13.3 Credit and Growth During Transition: A Simple Framework

We sketch a simple framework to assess the relationship between output and

credit during transition. The model may help to organize our analysis of the dif-
ferent stages of transition.
Assume there is a cash-in-advance constraint for the purchase of inputs by
firms. The cash-in-advance constraint for the purchase of nonlabor inputs (N) can
be expressed as follows, with C denoting cash flow and A an unanticipated shock:

N a gðCðAÞÞ
1 with g b 1: ð13:1Þ

When g ¼ 1, the firm has no access to external financing, a condition of extreme

credit constraint that was not uncommon in several countries in the initial phases
of transition (Coricelli 1998). Even in 2005, in several CIS countries, most enter-
prises were completely excluded from official credit markets (table 13.2).
In general, the share of both working capital and investment that was financed
through internal funds in transition countries was much larger than in advanced
market economies (table 13.3).
Assume a simple production function with intermediate goods (N) as the only
variable input, and labor (L) as the fixed factor,

Q ¼ f ðL; sNÞ; with f 0 ðNÞ > 0 and f 00 ðNÞ < 0: ð13:2Þ

The parameter s denotes the importance of intermediate inputs in production. We
also assume that payments to factor L are made at the end of the period and thus
do not affect the cash-in-advance constraint in the current period, although they
are important for the accumulation of liquidity by firms over time.3 We can thus
continue our discussion without considering the labor costs in following equa-
tions. In the pre-reform regime, bank credit covered all purchases of inputs. After
reforms, input prices change, interest rates are imposed on loans, and taxes are
Sources and Obstacles for Growth in Transition Countries 333

Table 13.2
Firms without Bank Loans and Financially Constrained (in percent of the sample)
Small firms Medium firms Large firms

Without Without Without

bank bank bank
Without loans and Without loans and Without loans and
bank financially bank financially bank financially
loans constrained loans constrained loans constrained
CEB 60.8 27.3 41.9 13.1 29.2 7.4
SEE 59.7 29.1 39.8 15.9 32.1 11.0
CIS 67.5 34.9 51.5 24.4 45.8 13.4

Germany 37.2 14.6 24.6 9.8 15.3 4.8

Source: BEEPS 2005.
Notes: A firm is classified as ‘‘small’’ if it has less than 49 employees, ‘‘medium’’ if it has between 50
and 249 employess, and ‘‘large’’ if it has 250 employees or more.

Table 13.3
Financing Structure of Firms, 2005 (in percent of total financing)
Working capital financing
Internal Borrowing Trade
finance from banks Equity credit Other
CEB 68.0 10.1 6.9 6.2 6.6
SEE 73.2 12.9 1.0 5.6 5.8
CIS 77.3 10.1 2.0 4.0 6.0
Fixed investment financing
CEB 62.4 14.3 6.5 1.9 12.0
SEE 70.8 17.7 0.9 2.4 6.8
CIS 77.2 11.6 1.9 1.8 6.9
Source: BEEPS.

raised on enterprise profits. Let us assume that banks keep their loans to enter-
prises constant in nominal terms. Define Q as output, P the price of output, B
bank debt, t enterprise tax rate, and P after-tax profits. We define firm liquidity
in real terms (nominal variables deflated by input prices); that is, the real liquidity
necessary to buy inputs. Furthermore, let us assume that in the pre-reform regime
the zero-profit condition holds: PQðNÞ ¼ Pi N ¼ B, or, at relative prices equal to 1,
Q ¼ B.
Thus, enterprise liquidity in the new regime is
ðB þ PÞ=Pi ¼ ðB þ ðPQðNÞ
BÞð1
tÞÞ=Pi : ð13:3Þ
334 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

We also assume that the firm is viable in the new regime—specifically, its reve-
nues at full capacity utilization (pre-reform level of output) are sufficient to cover
its costs, inclusive of interest payments. It is clear from 13.3 that at unchanged
nominal bank credit, the purchase of inputs necessary to produce full-capacity
output is affordable only if t ¼ 0 and relative prices P=Pi remain equal to 1. In
such a case 13.3 implies Q ¼ B. However, when t > 0 and there is an adverse shift
in relative prices, with input prices increasing more than output prices (adverse
supply shock), real liquidity declines and with it input purchases and thus output.
Thus, such adverse changes can be thought of as bad realization of the shock A. In
such a case constraint 13.2 is binding and firm’s output is liquidity constrained.
An implication of the above model is that firms with higher s—that is, firms that
are more liquidity-dependent for their production—will display a larger decline
in output following a liquidity squeeze. This is indeed what Calvo and Coricelli
(1993) found for the case of Poland at the outset of the reforms of the 1990s. In ad-
dition, given the assumption of decreasing marginal product from intermediate
inputs, the impact of credit expansion on output would be larger for more con-
strained activities. From 13.3 we can also note that higher profit rates allow firms
to decouple their production from external finance, as they can use their cash flow
for purchases of inputs. The high profit rates in transition countries may thus ex-
plain the coexistence of high growth rates of output and low external finance.
Let us now reinterpret the analysis in terms of the constraint 13.2. The parame-
ter g (greater or equal to 1) identifies the development of financial markets: the
higher is g, the less dependent firms are on their cash flow to finance purchase of
inputs. It should be noted that in addition to bank credit, other forms of financing
such as trade credit would reduce the dependence of output on cash-flow fluctua-
tions. As discussed later, trade credit could be a substitute for bank credit. How-
ever, a large use of trade credit in a system with an underdeveloped banking
system increases sharply the risk of contagion and the magnification of local
shocks to the entire system, and thus is likely to increase output volatility. What
is special in the experience of transition countries? There were two special features
in transition countries: first, the degree of underdevelopment of financial markets
was extreme, as credit did not play any role in the planned economy. Second,
transition countries are perhaps the only example in which one could identify li-
quidity needs at the start of market reforms, without risking confusing supply
and demand for credit.4
The paradox is that the equilibrium in the pre-reform period was observa-
tionally equivalent to having perfect financial markets. Indeed, output decisions
and the distribution of inputs among firms were decided by the planner. Credit
played just an accounting role, accommodating real decisions. There was no con-
nection between financial markets and liquidity needs by firms. Decisions at t
1
Sources and Obstacles for Growth in Transition Countries 335

were not constrained by availability of liquidity. These constraints emerged at t

when a decentralized market system was introduced. With the benefit of hind-
sight, the downplay of the role of financial markets was one of the major draw-
backs of stabilization and reform programs in formerly planned economies. We
stress this point not for pure intellectual reasons, but because we believe this ini-
tial sin conditioned the experience of transition economies and explains some of
the persistent problems characterizing even the more advanced among the transi-
tion countries—namely, the new members of the European Union and those that
are candidates to entry.
Of course, we do not want to underestimate the importance of imposing hard
budget constraints on enterprises. In the planned economy, incentives were heav-
ily distorted by lack of discipline in the use of resources that can derive by having
to repay bank loans. Eliminating automatic accommodation of real decisions was
thus a key component of market reforms. However, the difficult task is to create
the right incentives but at the same time ensure that liquidity needs for viable
and growing firms are available. Combining imposition of right incentives and
provision of liquidity is a feature of well-developed financial markets and of effec-
tive institutional mechanisms for contract enforcement (Coricelli and Djankov
2001). Indeed, such institutional preconditions were a fundamental aspect for the
creation of private markets, especially trade credit, that could compensate the
slow progress in developing an efficient banking sector.

13.3.1 Trade Credit and Inter-enterprise Arrears

A distinctive feature of successful reformers—namely the CEB group—in contrast

to CIS countries is that in the former private credit markets developed rather
quickly, while in the latter market liberalization was accompanied by a de-
monetization of the economy or the explosion of payment default.
Calvo and Coricelli (1996) developed a simple model of credit chains within a
circular system in which each firm is both buyer and seller of inputs, thus poten-
tially both creditor and debtor in the inter-enterprise trade credit market.5 Here
we illustrate the main mechanism through a simple example.
There are n firms producing inputs, using labor and inputs from other firms.
Assume that trade credit involves a proportion y of input transactions. Each firm
exchanges one unit of good. The proportion ð1
yÞ is the share paid in cash. This
cash can be obtained from banks or simply be liquidity accumulated by firms. If
the same proportion of trade credit and cash payments is used by every firm,
ð1
yÞ rather than ð1
yÞn is the amount of cash needed to support transactions
for an amount n of goods. In the simplest case of identical value of purchases and
sales of inputs, each firm has a balanced net position.6
336 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

Cash is needed only at one point of the chain. However, if we assume that the
cash constraint is binding, a contraction of bank credit by ð1
yÞ induces a decline
of sales by ð1
yÞn. The same effect would occur if there is a negative shock to the
cash flow of an individual firm that cannot pay its suppliers any more, inducing a
chain reaction of reductions in sales.
Consider again the cash-in-advance constraint 13.1

N a gðCðAÞÞ
1 : ð13:4Þ

This now becomes

ð1
yÞN a gðC 0 ðAÞÞ
1 ; ð13:5Þ

where C 0 ðAÞ now takes into account that the cash flow of a firm is affected by its
credit provided to other firms for its sales. In our example of zero net trade credit,
CðAÞ
C 0 ðAÞ ¼ ð1
yÞN. Thus, the cash-in-advance constraint remains the same,
but it affects a smaller amount of input transactions. The main difference between
a situation with trade credit and one without it is that the amount of cash needed
to effect transactions would be higher without trade credit. In our scheme, it
would be 1 rather than y.
Alternatively, trade credit could replace bank credit, playing ex post the same
role of the netting-out operation carried out by banks in the pre-reform regime.
Even though it could be argued that trade credit is a way to overcome infor-
mational problems associated with bank credit, it is nevertheless reasonable to
assume that trade credit requires a minimum set of institutions ensuring enforce-
ment of private contracts. Lacking these institutions, liquidity shocks induce an
equilibrium of generalized default, in the form of inter-enterprise arrears (or in-
voluntary credit). While in the very short run arrears can reduce output losses, as
firms can acquire inputs, such an equilibrium is bound to create large costs. As a
result of payment arrears, credit risk becomes so large as to impede any transac-
tions based on credit.
Indeed, it is not an accident that following a burst in inter-enterprise arrears one
observes a boom in barter trade, as was the case in several CIS countries. In Calvo
and Coricelli (1996) it is shown that institutional factors affecting the penalty im-
posed to bad behavior are key in determining whether the bad equilibrium with
generalized default takes place. In the case of the good equilibrium that applies
to CEB, one can note that trade credit was an important factor for financing inter-
enterprise transactions.7 As noted earlier, there is a view that explains trade credit
on the basis of informational advantages and of better enforcement of contracts
than bank credit. However, this purely microeconomic view downplays a major
drawback of trade credit, the bilateral nature of the contracts. Credit positions are
Sources and Obstacles for Growth in Transition Countries 337

not diversified and thus local shocks are transmitted to the whole system. In addi-
tion, while direct information on the customer is valuable, the presence of trade
credit complicates the analysis of financial positions of firms, and thus of credit-
worthiness. A perfectly viable firm can run into liquidity problems if its customer
does not pay. As a result, the firm cannot pay its supplier. In the bilateral contract
the assessment of the customer’s capacity to repay depends on the whole chain.
Thus, the information necessary to establish the capability of an individual firm
to repay its debt is much higher.
All these factors may imply that a predominance of trade credit with respect to
bank credit increases output volatility by magnifying the aggregate impact of id-
iosyncratic shocks. However, the Kiyotaki and Moore (1997) story and its general
equilibrium extension used by Cardoso-Lecourtois (2003) to explain higher vola-
tility of output in emerging markets fail to recognize that the credit chain reaction
is relevant only when there are large net positions in trade credit. In such an
event, problems of cash-poor firms are transferred to cash-rich firms, inducing an
inefficient magnifying effect to the whole system of liquidity shocks to cash-poor
firms. The more circular the system is, the less relevant is this channel. Circularity
means that net positions tend to be close to zero. This is indeed the case in most
countries, including transition countries. For instance, in Romania, Calvo and
Coricelli (1996) found that trade credit was almost perfectly circular, with net
debt positions accounting for a marginal fraction of total input transactions.
Furthermore, trade credit tends to be a major source of enterprise transaction
financing in advanced market economies as well. We argue that the overall devel-
opment of the financial sector is a fundamental factor in determining whether
trade credit is a source of greater volatility.
Consider again the cash-in-advance constraint 13.1 that would still apply to a
model with circular trade credit, but let us think of an economy evolving over
time. The constraint implies that the higher the level of financial sector develop-
ment is, the higher is g and the lower are the effects on output of shocks to cash
flow. In contrast, if g is close to 1, cash-flow shocks affect the capability of firms of
effecting their cash payments and carrying out production at its full capacity. In
such a situation, it is likely that firms will try to raise their liquidity by reducing
the amount of trade credit provided to their customers. In other words, firms will
try to get more cash from their sales. If all firms do the same as the overall amount
of cash available in the enterprise sector remains the same, a chain reaction with
lower amounts of inputs purchased will emerge, as in the credit chain foreseen
by Kiyotaki and Moore (1997). The ensuing equilibrium is analogous to one of a
chain reaction of default. This mechanism is particularly relevant if one takes into
account that there are forms of payments, like those to workers or the govern-
ment, that are rigid.
338 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

The discussion so far has emphasized the liquidity channel through which fi-
nancial markets may affect output in transition countries. Development of finan-
cial markets affects output mainly by softening the liquidity constraints on firms.
Under the maintained assumption of decreasing marginal product for inputs in
production, it turns out that the effect of financial development should be larger
the more severe are the liquidity constraints, and thus the underdevelopment of
financial markets. In our view, this channel is the main one in transition countries,
as opposed to the channel associated with financing innovation that has been
identified in the literature (see Levine 2005). In the next section we analyze the im-
pact of financial development on growth by looking at microdata for a relatively
large sample of countries, including several transition economies.

13.4 Finance and Growth during Transition: An Empirical Analysis

Having established that transition countries are still lagging behind in their devel-
opment of financial markets, the issue is to give a measure of how relevant such a
gap is for the growth rate of transition economies. We analyze this issue on a large
enterprise data set. Firm- or sector-level observations seem necessary to assess the
impact of financial development on growth. Analyses at the country level of ag-
gregation are affected by serious drawbacks in terms of identification of the effect
of finance on growth.

13.4.1 The Data Set

Our sample of data covers thirty European countries and twenty-six three-digit
International Standard Industrial Classification Revision 2 (ISIC R2) manufactur-
ing industries for the period 1996–2004, which are obtained by aggregating firm-
level data from the Amadeus database. The countries in the sample are EU-25
countries to which we added data for Iceland, Norway, Croatia, Bulgaria, Roma-
nia, the Russian Federation and Ukraine. Since the time span of data is not uni-
form across countries, we are dealing with an unbalanced panel of data.
Data on external finance dependence at industry level (three- and four-digit
ISIC R2 level) are taken from Rajan and Zingales (1998) (hereafter RZ). They de-
fine external finance dependence as the share of capital expenditure that a given
industry cannot finance through internal cash flow. Data on financial market de-
velopment (market capitalization of listed firms, domestic credit, and bank credit
to the private sector, all expressed as share in GDP) are taken from the World De-
velopment Indicators database (WDI) of the World Bank. These variables are then
interacted with the RZ measure of external finance dependence to obtain the vari-
Sources and Obstacles for Growth in Transition Countries 339

able that measures the effect on growth of external financial funds provided
through financial markets. Similarly, Fisman and Love (2003) (hereafter FL) con-
struct a variable that measures the dependence on trade credit in the benchmark
U.S. case as another source of external finance. We use their indicator in addition
to the one by Rajan and Zingales. Not just trade credit is likely to play a major
role in transition countries, but generally it seems more appropriate to consider
external finance not only for capital expenditure, but also for working capital—
that is, the main determinant of enterprise debt.
Growth of industry output is calculated from firm-level data on sales drawn
from the Amadeus database of the Bureau Van Dijck, which includes also small
and medium-sized firms. Sales are deflated with the producer price index
obtained from the IMF IFS database.8 The final dataset consists of 4,449 observa-
tions, comprising 638 country–industry units with seven years of time observa-
tions on average.

13.4.2 The Methodology

One of the major challenges in the literature on financial development and growth
has been how to address the potential endogeneity problem between the growth
rate of firm-level output and the degree of financial development. Using industry-
level data, Rajan and Zingales (1998) proposed a solution to the problem by using
the dependence on external finance by different sectors in the United States as the
benchmark. The idea is that the financial market in the United States can be
assumed to be close to perfect and thus the financial structure of firms is deter-
mined by an optimal choice that is not constrained by supply factors. In addition,
Rajan and Zingales argue that differences across firms of the same sectors are mi-
nor, and thus sectoral indicators are a good proxy for firm-level dependence on
external finance. The U.S. indicators can be considered exogenous indicators of
financing needs. Cross-country analysis of growth of real sales of firms, excluding
the United States, can then be used to determine the role of financial development
on growth. The sectoral U.S. financial dependence indicator is multiplied by the
level of financial sector development in different countries to construct what is by
now a familiar indicator in the literature, the Rajan-Zingales indicator. In our esti-
mations we interact the RZ measure of external finance dependence with the
share of total finance (market capitalization of listed firms and credit to the pri-
vate sector) in GDP. We concentrate on this measure of financial development be-
cause it is the most general among the measures used in the literature.
If the coefficient on the RZ indicator in a cross-country regression with the
growth of real sales as a dependent variable turns out to be positive, this indicates
340 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

that financial sector development affects the growth rates of firms. A positive co-
efficient on the RZ indicator implies that firms that need more external finance
grow faster in countries with a more-developed financial sector.
As Fisman and Love (2003) pointed out, this raises the issue of sufficient financ-
ing for the firms with high returns in the countries with less-developed financial
markets. From RZ it follows that the additional financing needed could be col-
lected from internal financing. Petersen and Rajan (1997) argue that alternative
funds could be raised by borrowing from suppliers. FL made a natural extension
of Petersen and Rajan’s approach by constructing a measure of trade credit using
a methodology similar to RZ. In order to obtain an industry-level measure of
trade credit usage, they employ the ratio of accounts payable to total assets, calcu-
lated for the U.S. firms for different sectors. Similarly to RZ, this measure is also
multiplied by the level of financial sector development in different countries. A
negative sign of the coefficient is consistent with the hypothesis that firms that are
more dependent on trade credit have a relative advantage in countries with less-
developed financial intermediaries, which implies a substitutability between trade
credit and bank credit. In contrast, if the coefficient is positive, there is a comple-
mentarity between the two forms of financing.
Even if RZ solve the problem of endogeneity of the financial indicator, there is
still a problem of possible reverse causality from growth of output to the level of
financial development. As emphasized by Guiso et al. (2004), a potential problem
of RZ is that financial development may affect both the growth rate of firms and
industries and the pattern of industry specialization. As a consequence, firms in fi-
nancially less-developed markets may adopt technologies that make them less de-
pendent on external finance. When estimating the effect of financial development
on growth using industry-level data, RZ tackle this endogeneity problem by
including in the estimated equations the beginning-of-period industry share in
value added. This has been used by other authors as well, including Guiso et al.
(2004). In our panel data set with time-varying data, initial period industry shares
in total value added are simple fixed effects. This means that a simple within esti-
mator corrects for the potential bias induced by the correlation between industry
specialization pattern and financial development. However, it must be noted that
by allowing for time variation in the panel, we may induce an additional potential
bias simply because financial development may be demand- and not only supply-
determined. Indeed, we use the contemporaneous relation between financial
development and growth, in contrast to RZ who use the initial-period level of fi-
nancial development. In our approach this problem is unlikely to be too damag-
ing, as our original units of observation on sales are firms who may have only a
very limited effect on aggregate supply of finance. Furthermore, the aggregate ef-
fect of all firms together may be limited since we concentrate on manufacturing
Sources and Obstacles for Growth in Transition Countries 341

that normally accounts for less than half of aggregate value added. In addition, a
significant share of credit may be supplied to households that even in most transi-
tion countries account for a large share of total credit. These are all indications
that the endogeneity problem in our estimations may be very limited, if present
at all. We formally tested this hypothesis with the Hausman test, which revealed
that our within-panel estimates do not suffer from an endogeneity problem.9 This
allowed us to treat the RZ and FL indicators as exogenous.
Our baseline empirical model is

Dyict ¼ aic þ bðRZi  FDct Þ þ gðFLi  FDct Þ þ dt þ uict ð13:6Þ

where Dyict denotes growth of real sales in industry i, country c, and year t. RZi
represents the Rajan and Zingales (1998) measure of external finance dependence,
while FLi stands for the corresponding measure of the use of trade credit as-
sembled by Fisman and Love (2003). FDct is a measure of financial development
(the sum of stock market capitalization and private credit as a percent of GDP).
aic is a full set of industry–country fixed effects, while dt denote common time
effects.10
The baseline model was extended in two directions in order to capture poten-
tially nonlinear effects of financial development on growth. First, explanatory
variables were interacted with a dummy variable for transition countries, DTr ,
and its complement, DNTr . In other words,
Dyict ¼ aic þ bTr ðRZi  FDct Þ  DTr þ gTr ðFLi  FDct Þ  DTr

þ bNTr ðRZi  FDct Þ  DNTr þ gNTr ðFLi  FDct Þ  DNTr þ dt þ uict : ð13:7Þ

The transition dummy de facto divides our sample between countries with low
and countries with high financial development. Significant differences in coeffi-
cients between the two groups of countries may thus reflect nonlinearities in the
effects of financial development on growth.
The second extension adopts a more systematic approach to modeling non-
linearities in the effect of financial development conditional on the level of finan-
cial development itself, without resorting to the use of country dummies. This
was achieved by allowing for explicit threshold effects. Following Hansen (1999),
we allow for a multiple threshold model. In particular, the model can be com-
pactly written as

X
4
Dyict ¼ aic þ dt þ ðbj ðRZi  FDct Þ
j¼1

þ gj ðFLi  FDct ÞÞIðtj
1 < FD < tj Þ þ uict : ð13:8Þ

342 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

This corresponds to a triple threshold model with t0 and t4 unspecified. For com-
parison, we also report the estimates of the double-threshold model. The thresh-
old variable is the measure of financial development.11

13.4.3 Empirical Results

Empirical results are reported in tables 13.3–13.6. Table 13.3 replicates the original
RZ and FL approaches on our data; that is, with no time variation in growth of in-
dustry output and financial development, but with the full set of country and sec-
tor dummies. Growth is averaged over the period under investigation, while the
initial level of financial development is taken in 1995. The RZ variable is correctly
signed, but insignificant. Including the FL variable also shows insignificant effects
of trade credit, while leaving the estimate of the RZ coefficient virtually un-
changed. Similar conclusions also emerge if we allow the coefficient to differ be-
tween transition and developed European countries.
Results significantly change when we allow for time variation in the panel.
Table 13.4 shows that, when inserted individually, the RZ variable is again insig-
nificant. Inclusion of the FL variable and allowing for the effects to differ between
transition and other European countries, however, produce meaningful results.
The RZ turns out to be positive and significant in both groups of countries, and
considerably higher in transition countries. What is more important, the size
of coefficients is much higher in the panel analysis relative to the original RZ
approach. Moreover, trade credit appears to be a substitute to official financing,
with stronger effects in transition countries.

Table 13.4
Effect of Financial Development on Growth: Original Rajan and Zingales Specification (1995 Financial
Development)
RZ 0.0012 0.0011
(0.0021) (0.0022)
FL 0.0092
(0.168)
RZ transition
0.0017
0.0024
(0.0066) (0.0066)
RZ nontransition 0.0007 0.0006
(0.0023) (0.0024)
FL transition 0.642
(0.489)
FL nontransition 0.107
(0.183)
Note: Standard errors in parentheses. Constant not reported.
Sources and Obstacles for Growth in Transition Countries 343

These results confirm the original RZ result that financial development posi-
tively affects growth. However, it is worth noting that this result obtains only
after we control for trade credit as an alternative source of external finance and,
more importantly, after we allow for time variation in the data and nonlinearity
of the effect. In this respect, we find that financially less-developed countries ben-
efit considerably more from financial development, in line with our discussion in
section 3.
Evidence of nonlinearity in the effect of financial development on growth is
more refined in the estimated threshold models. Formal testing supports the
triple-threshold model as more appropriate than a double-threshold model. As
can be seen by comparing the upper panels of tables 13.5 and 13.6, allowing for

Table 13.5
Effect of Financial Development on Growth—Within Estimates
RZ
0.0215
0.0430
(0.0149) (0.0373)
FL
0.461*
(0.190)
RZ transition 0.1442* 0.2447*
(0.0054) (0.0757)
RZ nontransition
0.0352* 0.0140*
(0.0155) (0.0017)
FL transition
0.688*
(0.375)
FL nontransition
0.395*
(0.103)
Note: Standard errors in parentheses. Constant not reported. Tests reveal significant presence of fixed
effects in all specifications. Time dummies included in the model. * denotes statistical significance.

Table 13.6
Threshold Effects of Financial Development on Growth—Double Threshold Model

Threshold 52.93 70.29

95% conf. int. [40.94, 52.99] [64.69, 256.10]
F2 (p-value) 25.84 (0.03)
Parameters FD < t1 t1 < FD < t2 FD > t2
RZ 0.308* 0.052 0.020
(0.113) (0.161) (0.018)
FL
1.53* 2.78*
0.43*
(0.56) (0.78) (0.09)

F2 is the test statistic for presence of thresholds using 300 bootstrap replications. Standard errors of
coefficients in parentheses. Time dummies included in the model. * denotes statistical significance.
344 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

Table 13.7
Threshold Effects of Financial Development on Growth—Triple-Threshold Model
Threshold 52.93 62.92 70.29
95% conf. int. [40.94, 52.99] [62.11, 92.97] [70.29, 70.34.10]
F2 (p-value) 55.12 (0.00)
Parameters FD < t1 t1 < FD < t2 t2 < FD < t3 FD > t3
RZ 0.308? 0.133
0.154 0.020
(0.112) (0.187) (0.316) (0.018)
FL
1.56* 0.83 8.02*
0.44*
(0.56) (0.90) (1.50) (0.09)
F2 is the test statistic for presence of thresholds using 300 bootstrap replications. Standard errors of
coefficients in parentheses. Time dummies included in the model. * denotes statistical significance.

the third threshold, found to lie between the first two, considerably reduces the
estimated confidence interval of the second-threshold model (Table 13.7). We
find thresholds at levels of financial development at 53, 63, and about 70 percent
of GDP. The lower and upper thresholds are very precisely estimated, while the
confidence intervals for the middle threshold prove to be much wider.
In line with our priors, the largest effect of financial development on growth is
found for countries that have a financial development ratio below the first thresh-
old. Indeed, the coefficient is even larger than found for the transition countries in
the previous panel analysis. Above that threshold the coefficient declines and
becomes insignificant and close to zero for levels of financial development above
the upper threshold.12 All developed Western European countries in the sample
were above that threshold throughout the period under investigation. It is impor-
tant to note that the most advanced transition countries passed the threshold in
recent years as well, implying that considerably smaller growth dividends than
those observed in the past can be expected from further financial development.13
Regarding the effects of trade credit, we found that at lower levels of financial
development trade credit acts as a strong substitute for official finance. Going
through insignificance in the second-threshold region, trade credit becomes a
strong and significant complement in the third-threshold region. Above the third
threshold, trade credit becomes a substitute for official finance, even though the
effect is much smaller than the one prevailing below the first threshold.
In summary, the effect of financial development appears to be much
stronger for transition countries. Trade credit serves as a relevant substitute
for official financing, softening the adverse impact on the growth of financial
underdevelopment.
Sources and Obstacles for Growth in Transition Countries 345

13.5 Conclusions

Credit markets play a crucial role in the transformation from centrally planned to
market economies. Our empirical analysis suggests that financial sector develop-
ment has a significant independent effect on output growth. For transition coun-
tries, the growth dividend from financial sector development appears to be much
larger than for more advanced market economies, indicating the potentially larger
effects of financial development on growth in countries with low levels of devel-
opment. Trade credit plays a relevant role in substituting for lack of official fi-
nance, especially in transition countries.
The policy challenge is how to foster the efficient development of financial mar-
kets, which continue to be shallow in many transition countries. The process of
financial sector development crucially hinges on institution building. CEB had a
great opportunity for strengthening their institutions by joining the European
Union. Further progress is likely to come from joining the Eurozone. Contrary to
the traditional optimal currency area theory, countries with large structural asym-
metries with respect to the Eurozone countries (because of much lower GDP per
capita) and with still-underdeveloped financial markets can benefit significantly
from joining the Eurozone. Joining a currency union may lead to large benefits in
terms of diversification of risk and access to smoothing instruments. Often empir-
ical analyses measure the readiness to join a monetary union in terms of structural
asymmetries, measuring the overall volatility of output and the correlation be-
tween individual country shocks and those of the currency area to be joined.
These analyses neglect the importance of transmission of shocks through the fi-
nancial sector. If the financial sector of a union is more efficient than the national
one, joining the union reduces the transmission of nominal shocks to output, or,
for that matter, the transmission of temporary shocks to output, whatever their
origin.

Notes

1. CEB denotes Central-Eastern European countries and the Baltic States: Czech Republic, Estonia,
Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia. SEE stands for Southern-Eastern European
countries: Albania, Bosnia-Herzegovina, Bulgaria, Croatia, Macedonia, Romania, and Serbia. CIS
stands for Commonwealth of Independent States: Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan,
Kyrgyz Republic, Moldova, Russia, Taijikistan, Turkmenistan, Ukraine, and Uzbekistan.
2. See, for instance, EBRD (2003).
3. See Calvo and Coricelli (1992) for a discussion of the possibility of ‘‘borrowing from workers.’’
4. Rajan and Zingales (1998) discuss the issue and, to avoid these problems, assume that the liquidity
needs of a sector are those observed in the United States, defined as a benchmark for perfect financial
markets (see discussion in section 4).
346 Fabrizio Coricelli, Bostjan Jazbeč, and Igor Masten

5. See Kiyotaki and Moore (1997) for a model of credit chains.

6. If we had a model with nonzero net positions, the amount of liquidity needed would be equal to the
sum of these net positions plus the amount ð1
yÞ.

7. In microdata for Hungary for the mid-1990s, for instance, it was found that many firms were com-
pletely cut off from access to bank loans and financed their purchases of inputs through trade credit
(Coricelli 1998).
8. All observations with growth of real sales that exceeded 100 percent were treated as outliers and
thus excluded from the database. Industry-level growth of output was calculated as a simple average.
9. In the construction of the Hausman test, we compared within estimates with the instrumental vari-
ables within estimates using lagged values of growth of real sales, the RZ indicator, and FL indicators.
The Hausman test statistic is w 2 ð8Þ ¼ 10:61 with a corresponding p-value of 0.22. The test thus suggests
the use of the normal within estimator as the most appropriate since it is more efficient under the null
hypothesis that cannot be rejected.
10. Note that the Rajan and Zingales (1998) in their original specification estimated a different model.
Growth of output was measured as the average over a period, while financial development was taken
from the initial period.
11. In line with the results of the Hausman test in the basic panel analysis, we treat it as exogenous,
which means that the model setting fits the assumption in Hansen (1999). Estimation of threshold lev-
els and their confidence regions follows the multistep procedure described in Hansen (1999). Hansen’s
method is designed for balanced panels, while we operate with an unbalanced panel. In such a case it
must be noted that it is unknown whether all Hansen’s results regarding inference carry completely
through.
12. Virtually the same findings as regards the RZ variable emerge from both the double- and triple-
threshold model.

13. This statement is confined to measures of development that we use in the sample; that is, the depth
of financial markets. The effects of other, mainly institutional, characteristics related to the efficiency of
financial markets (that may or may not be correlated with the depth of financial markets) were not
accounted for in present empirical analysis.

References

Acemoglu, D., S. Johnson, J. Robinson, and Y. Thaicharoen. 2002. ‘‘Institutional Causes, Macroeco-
nomic Symptoms, Crises and Growth.’’ Mimeo., Massachusetts Institute of Technology, Cambridge,
MA.
Calvo, G., and F. Coricelli. 1992. ‘‘Stagflationary Effects of Stabilization Programs in Reforming Social-
ist Countries: Enterprise-side and Household-side Factors.’’ World Bank Economic Review 6, no. 1: 71–90.
Calvo, G., and F. Coricelli. 1993. ‘‘Output Collapse in Eastern Europe: The Role of Credit.’’ International
Monetary Fund Staff Papers 40, no. 1: 32–52.
Calvo, G., and F. Coricelli. 1996. ‘‘Credit Market Imperfections and Low-output Equilibria in Econo-
mies in Transition.’’ In Financial Factors in Economic Stabilization and Growth, eds. M. Blejer, Z. Eckstein,
Z. Hercowitz, and L. Leiderman, 75–102. New York: Cambridge University Press.
Campos, N. F., and F. Coricelli. 2002. ‘‘Growth in Transition: What We Know, What We Don’t, and
What We Should.’’ Journal of Economic Literature 40, no. 3: 793–836.
Sources and Obstacles for Growth in Transition Countries 347

Cardoso-Lecourtois, M. 2003. ‘‘Chain Reactions, Trade Credit and the Business Cycle.’’ Mimeo., Univer-
sidad Carlos III de Madrid.
Coricelli, F. 1998. Macroeconomic Policies and Development of Markets in Economies in Transition. Budapest:
Central European University Press.
Coricelli, F., and S. Djankov. 2001. ‘‘Hardened Budgets and Enterprise Restructuring: Theory and the
Application to Romania.’’ Journal of Comparative Economics 29: 749–763.
Cottarelli, C., G. Dell’Ariccia, and I. Vladkova-Hollar. 2003. ‘‘Early Birds, Late Risers, and Sleeping
Beauties: Bank Credit Growth to the Private Sector in Central and Eastern Europe and the Balkans.’’
Working Paper No. WP/03/213, IMF, Washington, D.C.
Djankov, S., C. McLiesh, and A. Shleifer. 2007. ‘‘Private Credit in 129 Countries.’’ Journal of Financial
Economics, 84(2), 299–329.
European Bank for Reconstruction and Development (EBRD). 2003. Transition Report 2003: Integration
and Regional Cooperation. London: EBRD.

Fisman, R., and Love, I. 2003. ‘‘Trade Credit, Financial Intermediary Development and Industry
Growth.’’ Journal of Finance 58: 353–374.

Guiso, L., T. Jappelli, M. Padula, and M. Pagano. 2004. ‘‘Financial Market Integration and Economic
Growth in the EU.’’ Economic Policy 19, no. 40: 523–577.
Hansen, B. E. 1999. ‘‘Threshold Effects in Non-dynamic Panels: Estimation, Testing and Inference.’’
Journal of Econometrics 93: 345–368.
International Monetary Fund. International Financial Statistics. Various issues, Washington, D.C.

Kiyotaki, N., and J. Moore. 1997. ‘‘Credit Chains.’’ Mimeo., London School of Economics.
Levine, R. 2005. Finance and Growth: Theory and Evidence. In P. Aghion and S. Durlauf, eds. Handbook
of Economic Growth. Amsterdam: Elsevier Science.

Petersen, M., and R. Rajan. 1997. ‘‘Trade Credit: Theories and Evidence.’’ The Review of Financial Studies
10(3): 661–691.

Rajan, R., and L. Zingales. 1998. ‘‘Financial Dependence and Growth.’’ American Economic Review 88:
559–586.
Roland, G. 2000. Transition and Economics: Politics, Markets, and Firms. Cambridge, MA: MIT Press.

World Bank. World Development Indicators. Various issues. Washington D.C.

14 Growth in Transition
Economies: Domestic Policies,
External Assistance, and
Institution Building

Stanley Fischer and Ratna Sahay

14.1 Introduction

The transition process in the economies of the former Soviet bloc is essentially at
an end, as the twenty-five transition economies go their own ways and their pol-
icy problems become less distinguishable from those confronting other econo-
mies. After a little more than a decade, some countries are headed for Europe,
with different expected dates of arrival; some in the CIS that, for a while, looked
as if they would never progress, most notably Russia, are growing rapidly; and a
few remain in severe difficulties.
It is unlikely there will ever be another natural economic experiment on a scale
as large as that of the transition process, with twenty-five economies changing
policies radically at almost the same moment.1 What have we learned from the ex-
perience?2 That depends on what was expected. The consensus was that there
would be an initial decline in output due to disorganization as the price and insti-
tutional structure of the economies changed and macroeconomic stability was
established, and that after that period, countries would begin to grow more rap-
idly than the advanced countries, toward whose levels of income they would
eventually converge. The initial decline was probably expected to last one to two
years, but as we know now the decline—in magnitude and length of time—was
grossly underestimated, especially in the CIS countries.
Although the transition economies are now all growing, and although our read-
ing of the history of the period, and the statistical representation of it, convinces
us that the basic strategy recommended by international financial institutions
(IFIs) was right, that is a controversial view. The accusation that the IFIs lost
Russia and the charge that shock treatment and too-rapid privatization produced
unnecessary output losses, disorganization, corruption, and misery have been fa-
miliar parts of the indictment of the approach recommended by Western officials
and other advisers.3
350 Stanley Fischer and Ratna Sahay

The paper is organized as follows: in section 14.2 we take stock of the evolution
of output in the twenty-five transition economies since the dismantling of the cen-
tral planning system. In section 14.3, we extend the data set used in our previous
work and confirm our earlier results that both macroeconomic policies and struc-
tural reforms help growth to revive. The role of initial conditions and policy per-
formance are explored in greater detail in sections 14.4 and 14.5. We then focus
on institutional development in the next three sections. In particular, we focus on
an additional element in the indictment brought by critics of the advice offered to
the transition economies by the IFIs and other mainstream economists; namely,
that their advice ignored the need to build the institutional framework for a mar-
ket economy. We argue that this charge that the IFIs did not take into account the
importance of institutional development, especially the rule of law, is without
merit. In section 14.9, we present empirical analyses that highlight the role of
institutions, broadly defined. We present concluding comments in section 14.10.

14.2 Evolution of Output During Transition

This study is based on data through the end of 2003.4 Figure 14.1a shows the evo-
lution of GDP since transition began in four groups of countries: those of Central
and Eastern Europe (CEE); the Baltics; the CIS-5,5 which includes Russia; and the
CIS-7,6 which consists of the poorest countries of the CIS. Transition year, T, is
defined as the year in which central planning was abandoned in the former com-
munist countries (table 14.1). Transition began in 1992 in the former Soviet Union
(Baltics, the CIS-5, and the CIS-7), and somewhat earlier in the CEE countries. The
chart shows unweighted averages of output for the countries in each of the
groups. Output began to fall toward the end of the 1980s, even before transition
began in most countries. As is well known, the output fall was much larger in the
former Soviet Union countries, including the Baltics, than in the central and east-
ern European countries. It is also true that output took longer to recover in the
CIS countries than in eastern and central Europe and the Baltics.
As is evident from table 14.1 and figure 14.1b, once output began to grow, the
average growth rate in the CIS-5 and CIS-7 was higher by nearly four and more
than two percentage points, respectively, than in the CEE countries, which grew
at about 3.5 percent. However, because of their late start and much larger output
decline, measured output in most CIS countries still had not significantly sur-
passed its pretransition levels by 2003.7
In the event, the initial output declines looked more like collapses—which be-
gan before the end of the old regime—than the more measured declines that were
expected. Growth began to revive at different times in each transition economy.
The rate of growth in the initial recovery phase was probably below expectations.
Figure 14.1
Transition Economies (a) Real GDP Index, (b) Since Lowest Level. Source: International Monetary Fund.
Authors’ calculations.
Table 14.1
Transition Economies: Output Performance, 1989–2003

Average
output
Cumula- growth
GNP Year in tive since
(PPP) which Maximum output lowest
per output output growth, level
Transition capita was decline lowest until
year (T) in 1989 lowest since T-1 to 2003 2003a
(1) (2) (3) (4) (5) (6)
Baltics 1992 7973 1993/94 38.1 52.3 4.87
Estonia 1992 8900 1994 29.4 54.1 4.80
Latvia 1992 8590 1993 44.2 53.0 5.32
Lithuania 1992 6430 1994 40.6 49.8 4.49

CEE 1990/91 5760 1991/97 24.7 44.5 3.54

Albania 1991 1400 1992 33.2 95.5 6.09
Bulgaria 1991 5000 1997 39.3 27.5 4.05
Croatia 1990 6171 1993 37.6 54.8 4.37
Czech Republic 1991 9000 1992 12.1 24.4 1.98
Hungary 1990 6810 1993 18.1 40.4 3.39
Macedonia, FYR 1990 3394 1995 27.2 15.1 1.76
Poland 1990 5150 1991 13.7 62.9 4.07
Romania 1991 3470 1992 20.6 22.3 1.83
Slovak Republic 1991 8000 1993 24.4 51.0 4.12
Slovenia 1990 9200 1992 20.4 51.3 3.76

CIS-7 1992 4191 1993/99 46.0 58.2 5.72

Armenia 1992 5530 1993 14.1 100.7 6.97
Azerbaijan 1992 4620 1995 57.9 88.5 7.93
Georgia 1992 5590 1994 65.4 62.1 5.37
Kyrgyz Republic 1992 3180 1995 44.8 45.4 4.68
Moldova 1992 4670 1999 62.2 24.1 5.40
Tajikistan 1992 3010 1996 58.8 59.3 6.66
Uzbekistan 1992 2740 1995 17.5 27.2 3.01
CIS-5 1992 5954 1995/99 42.0 54.4 7.00
Belarus 1992 7010 1995 31.5 58.4 5.75
Kazakhstan 1992 5130 1995 31.1 53.3 5.34
Russia 1992 7720 1998 45.6 35.5 6.08
Turkmenistan 1992 4230 1997 45.9 93.5 11.00
Ukraine 1992 5680 1999 55.2 31.3 6.81
Sources: GNP per capita in 1989 (DDGT 1998, World Bank), real GDP (World Economic Outlook 2004).
Note: Transition year is defined as the year in which central planning was dismantled. T-1 refers to the
year before transition began. Through columns 4 to 6, real GDP index (T-1 ¼ 100) is used. The calcula-
tions for Armenia based on time T, as there are no data for T-1 and 1989. The transition year and the
year in which output is lowest for each group corresponds to the modal average.
a
Geometric averages.
Growth in Transition Economies 353

However, in more recent years, the poorer countries of the CIS-7 as well as the
CIS-5 have been growing faster than CEE and the Baltics, and if high rates of
growth continue, there is hope for convergence—though that could take a very
long time.

14.3 Revisiting the Earlier Empirical Studies

In our earlier work (Fischer, Sahay, and Végh 1996a, 1996b, and 1998), we con-
cluded that the transition experience confirmed the view that both macroeco-
nomic stabilization and structural reforms contribute to growth, and that the
more structural reform that took place, the more rapidly the economy grew.
Table 14.2 presents results obtained from running the panel data specifications
used in our previous work on a data set that has been extended to 2001.8 Note
that the number of observations nearly tripled with this extension of the data set.

Table 14.2
Transition Economies: Growth Regressions—Fischer-Sahay-Vegh Specifications with Expanded Data
Sets (t-statistics in parentheses)
Fixed effects regressions: Dependent variable: GDP growth
(1) (2) (3) (4) (5) (6)
Extended Extended Extended
FSV 96 data FSV 97 data FSV 98 data

Fixed exchange rate 11.35 5.50 — 7.08 4.53

(2.00)* (2.95)* (2.15)* (2.27)*
Fiscal balance 0.30 0.49 0.24 0.42 0.29 0.41
(1.42) (7.03)* (1.90) (5.32)* (2.60)* (5.28)*
CLI 7.42 2.78 — — —
(3.54)* (9.68)*
Inflation — —
2.73
2.50 — —
(
5.27)* (10.03)*
Official assistance — — 0.26
0.02 — —
(2.32)* (0.14) — —
LIP — — 12.97 16.26 — —
(2.86)* (4.17)*
LI — — — — 30.72 38.44
(2.70)* (7.69)*
R-Squared 0.72 0.48 — 0.57 0.66 0.40
Total Observations 75 216 — 208 100 216
Source: Cumulative liberalization index (CLI), liberalization index of privatization (LIP), and liberal-
ization index (LI) from DDGT (1998)/World Bank, inflation and fiscal balance from WEO 2002, aid
share of gross national income from WDI 2002.
Note: Fixed exchange rate regime is a dummy, it takes 0 for floating exchange rate regime and 1 for
fixed exchange rate regime. Fiscal balance is general government balance as share of GDP.
354 Stanley Fischer and Ratna Sahay

The earlier results showed that lower inflation or fixed exchange-rate regimes
(possibly because they brought inflation down much faster), more foreign assis-
tance, and faster and deeper structural reforms helped raise growth rates. The
results on fiscal balance (the tighter the fiscal policy, the faster the growth rates)
were weak.
The new results (with the extended data set) are consistent with our earlier
results. In fact, confirming the IFI view, the fiscal balance variable that was not
significant in the earlier data set is now significant across all three regressions.9
Moreover, the coefficient on the fiscal policy variable (the budget balance)
increases with the longer-period data set. The coefficients on the reform indices—
whether the cumulative liberalization index (CLI), or simply the liberalization
index (LI), or the liberalization index of privitization (LIP), which captures priva-
tization and enterprise reforms—are significant throughout the period, irrespec-
tive of the time period considered.
An interesting result relates to the foreign assistance variable: with the extended
data set, the coefficient on foreign assistance is no longer significant, indicating
possibly that aid mattered more during the initial years of transition than later, or
reverse causation, with the slow growers receiving more aid in later years.
In summary, the extension of the data set strengthens our confidence in the em-
pirical results obtained a few years ago by us and other researchers, as well as in
the policy implications of those results.10 We elaborate more on these aspects in
later sections.

14.4 Initial Conditions, History, and Geography

Our earlier work (Fischer, Sahay, and Végh 1996a, 1996b, and 1998) found a sig-
nificant effect of initial conditions on growth in the early years of the transition.
For example, growth was generally worse, ceteris paribus, for those countries
that were farther to the east (as measured by distance from Brussels), or that had
spent a longer period under communism.
Table 14.3 presents different measures of initial conditions, history, and geogra-
phy of the transition economies. At the start of the transition, per capita income
levels in the CIS countries were generally lower than those in the Baltics and the
CEE countries. Compared to the Baltics and the CEE countries, the CIS countries
had also endured a longer period of communism and are farther away from the
more advanced countries of Europe (as measured by the distance from Düssel-
dorf). Of the four groups of countries, the CIS-5 are better endowed with natural
resources than the other regions. While on average, the Baltics and the CIS coun-
tries are much more fragmented ethnically than the CEE countries, there is large
diversity on this score among the countries within the groups.11
Growth in Transition Economies 355

Table 14.3 also includes an index of initial conditions computed by the Euro-
pean Bank for Reconstruction and Development (EBRD). A higher value of the
initial conditions index indicates a more favorable starting position. This index is
derived by the EBRD from factor analysis on a set of measures for the level of de-
velopment, trade dependence on the Council for Mutual Economic Assistance
(CMEA), macroeconomic disequilibria (such as repressed inflation and black mar-
ket premium), distance from the EU, natural resource endowment, market mem-
ory, and state capacity. As expected, economic distortions as measured by the
EBRD index were much higher in the CIS countries than in the CEE. We use this
index in the regression exercises to examine whether and for how long initial con-
ditions mattered.
The role of initial conditions was examined by De Melo, Denizer, Gelb, and
Tenev (1998); Berg et al. (1999); Havrylyshyn and van Rooden (2001); and others.
The general conclusion was that the effect of initial conditions, while strong at the
start of transition, wears off over time. The diminishing role of initial conditions is
confirmed by the results presented in table 14.4. The coefficients are insignificant
in two of three regressions. In the regression where it is significant (equation (7)
in table 14.4), the initial condition index has the right sign, indicating that those
with more favorable initial conditions grew faster. When the index is interacted
with the time-in-transition dummy, the coefficient remains significant but turns
negative, indicating that the longer the period, the less the significance of the ini-
tial conditions. These results are reassuring for countries aiming to overcome their
adverse starting conditions by implementing good policies.

14.5 Policy Performance

As reported in section 14.3, in our earlier papers (Fischer, Sahay, and Végh 1996a,
1996b, and 1998) the results on the macroeconomic policy variables were not ro-
bust across specifications. But before we report in greater detail the results with
the expanded data set, it is worthwhile to look at the policy performance that
accompanied the revival of growth in the later years of the transition.
Table 14.5 reports on policy performance. We look at two key variables—the
inflation rate and budget balance. The inflation performance across each of the
subgroups—the Baltics, CEE, CIS-7, and CIS-5—is impressive. During the first
three years of transition virtually all the countries averaged three-digit inflation
rates. By 2003, inflation rates were reduced to single- or low double-digit levels.
In fact, in the Baltics and the CEE, the inflation performance by 2003 was similar
to that in the OECD countries. Pegging the exchange rate to a strong currency in
the initial years helped most countries to bring inflation down very rapidly. This
performance was sustained by the expectation of EU accession in later years in
Table 14.3
356

Transition Economies: History and Geography

Per capita
income Per capita First Initial Distance Ethnic
Transition (PPP) in GDP (PPP) year of condition from fraction- Natural
year (T) 1989 in T communism index Düsseldorf alizationa resources
Baltics 1992 7973 6119 1940 C0.2 1347 0.50 Poor
Estonia 1992 8900 6320 1940
0.4 1449 0.53 Poor
Latvia 1992 8590 5335 1940
0.2 1293 0.61 Poor
Lithuania 1992 6430 6703 1940 0 1299 0.35 Poor
CEE 1990/91 5760 7242 1947 2.6 1134 0.26 Poor
Albania 1991 1400 2186 1945 2.1 1494 0.46 Poor
Bulgaria 1991 5000 5632 1947 2.1 1574 0.26 Poor
Croatia 1990 6171 7133 1945 2.5 913 0.37 Poor
Czech Republic 1991 9000 10801 1948 3.5 559 0.11 Poor
Hungary 1990 6810 9447 1948 3.3 1002 0.17 Poor
Macedonia, FYR 1990 3394 5011 1945 2.5 1522 0.57 Poor
Poland 1990 5150 5684 1948 1.9 995 0.04 Moderate
Romania 1991 3470 ... 1948 1.7 1637 0.20 Moderate
Slovak Republic 1991 8000 7938 1948 2.9 824 0.25 Poor
Slovenia 1990 9200 11345 1945 3.2 815 0.17 Poor
CIS-7 1992 4191 2887 1924 C0.6 3704 0.46 Poor
Armenia 1992 5530 2160 1920
1.1 3143 0.12 Poor
Azerbaijan 1992 4620 3046 1921
3.2 3270 0.31 Rich
Georgia 1992 5590 4650 1921
2.2 3069 0.55 Moderate
Kyrgyz Republic 1992 3180 2978 1921
2.3 5047 0.66 Poor
Moldova 1992 4670 3311 1940
1.1 1673 0.55 Poor
Tajikistan 1992 3010 1866 1921
2.9 4938 0.58 Poor
Uzbekistan 1992 2740 2195 1921
2.8 4788 0.48 Moderate
Stanley Fischer and Ratna Sahay
CIS-5 1992 5954 6501 1920 C1.9 2924 0.45 Rich
Belarus 1992 7010 6660 1918
1.1 1435 0.37 Poor
Kazakhstan 1992 5130 5615 1921
2.5 5180 0.68 Rich
Russia 1992 7720 9077 1917
1.1 2088 0.31 Rich
Turkmenistan 1992 4230 5154 1921
3.4 4254 0.46 Rich
Ukraine 1992 5680 5998 1918
1.4 1664 0.42 Moderate
Source: Per capita income (World Bank), per capita GDP (World Economic Outlook, 2002), initial condition index (EBRD reports), ethnic fractional-
ization (Kok Kheng, 2001).
Note: GDP data in transition year were not available for Uzbekistan, Slovenia, and Czech Republic. For these countries GDP data in the first year
after transition are reported.
a
Growth in Transition Economies

Range 0–1; 0 represents no ethnic fractionalization.

357
358 Stanley Fischer and Ratna Sahay

Table 14.4
Transition Economies: Output Performance, Initial Conditions, Policies, and Institutional Develop-
ment, 1991–2001 (t-statistics in parentheses)
2SLS Panel regressions1 dependent variable: GDP growth
(7) (8) (9)
a
Initial conditions index 1.439 C0.013 C0.441
(2.08)* (0.03) (0.94)
Initial conditions index *year C0.213 C0.039 0.027
(3.45)** (0.99) (0.67)
Exchange rate regimeb C4.293
(3.05)**
First lag of inflation C0.017 C0.017
(7.30)** (7.33)**
Change in fiscal balance 0.387 0.583 0.685
(4.55)** (7.59)** (8.70)**
Reform indexc 15.382 4.77 3.608
(10.58)** (7.19)** (5.91)**
State capture indexd C0.126
(2.85)**
Intercept C36.963 C9.319 C4.797
(11.13)** (6.11)** (2.79)**
R-Squared 0.41 0.4 0.47
Probability value 0.000 0.000 0.000
Total observations 265 296 252
Source: Initial conditions index, reform index, and state capture index (Transition Report); exchange
rate regime (EAER); fiscal balance and inflation (WEO 2002).
1
One star (*) indicates significance at the five-percent level, while two stars (**) indicate significance at
the one-percent level.
a
Initial conditions index is derived from factor analysis and represents a weighted average of measures
for the level of development, trade dependence on CMEA, macroeconomic disequilibria, distance to
the EU, natural resource endowments, market memory, and state capacity. The higher values of the
index relate to more favorable starting positions.
b
Exchange rate regime dummy. Takes 0 for floating exchange rate system and 1 for fixed exchange rate
system.
c
Reform index is average of liberalization index, market reform index, financial reform index, and
privatization. The reform index is instrumented by the sum of reform indices in other 24 transition
countries excluding the country for which it is instrumented.
d
State capture index measures the extent to which businesses have been affected by the sale of govern-
ment decisions and policies to private interests. A higher value indicates more ‘‘capture.’’ It is based
on the business environment and enterprise performance survey implemented in 1999 by the EBRD in
collaboration with the World Bank.
Growth in Transition Economies 359

Table 14.5
Transition Economies: Policy Performance
Average Average
inflation budget
rate, first balance Budget
three years Inflation (% GDP), balance
Transition since price rate in first three (% GDP)
year (T) liberalization 2003 years since T in 2003
Baltics 1992 298 1.0 C1.5 C0.8
Estonia 1992 150 1.3 0.1 0.6
Latvia 1992 395 2.9
1.5
1.8
Lithuania 1992 350
1.2
3.2
1.4
CEE 1990/91 181 4.4 C5.2 C4.0
Albania 1991 116 2.3
20.1
5.1
Bulgaria 1991 163 2.3
10.3 0.0
Croatia 1990 455 1.5
3.0
4.5
Czech Republic 1991 30 0.2 1.8
7.0
Hungary 1990 27 4.7
3.5
5.9
Macedonia, FYR 1990 113 2.5
3.9
2.0
Poland 1990 302 0.8
3.6
6.6
Romania 1991 209 15.3
0.6
2.5
Slovak Republic 1991 31 8.5
9.0
4.7
Slovenia 1990 363 5.6 0.5
1.4
CIS-7 1992 434 8.2 C19.7 C1.3
Armenia 1992 462 4.8
36.7
2.0
Azerbaijan 1992 509 2.2
8.2
1.5
Georgia 1992 483 4.8
34.0
2.3
Kyrgyz Republic 1992 570 2.7
14.2
4.7
Moldova 1992 475 11.7
13.9 1.6
Tajikistan 1992 112 16.4
19.8
1.8
Uzbekistan 1992 430 14.8
11.3 1.6
CIS-5 1992 298 12.9 C5.6 0.5
Belarus 1992 526 28.4
2.4
1.2
Kazakhstan 1992 91 6.4
5.3 3.2
Russia 1992 485 13.7
12.1 1.0
Turkmenistan 1992 298 11.0 6.7 0.0
Ukraine 1992 91 5.2
14.9
0.7
Source: Inflation and general government budget balance (World Economic Outlook 2004). Authors’
calculations.
360 Stanley Fischer and Ratna Sahay

many of those countries. The development of domestic financial markets and ac-
cess to international capital markets over time also helped in financing budgets
without resorting to central bank money creation.
There has been a substantial fiscal consolidation across all groups since the start
of the transition. The CIS-7 countries stand out for their impressive feat in reduc-
ing the budget deficit from nearly 20 percent of GDP in the first three years to just
over 1 percent in 2003. Due to financing constraints and lack of access to capital
markets, these countries faced severe expenditure compression throughout the
period. Helbling, Mody, and Sahay (2004) show that foreign aid, especially grants,
came on too small a scale to provide significant support to their budgets; this
appeared to have both hindered their recovery process and led to a rapid build-
up of external debt-to-GDP ratios.
The CIS-5 and the Baltics also managed to nearly balance their budgets by 2003,
although the starting points in the Baltics were already quite favorable. Interest-
ingly, the consolidation in the CEE countries was much smaller, but less was
needed, given their more favorable starting points. Moreover, most of these coun-
tries invested in public sector reforms that initially contributed little to deficit re-
duction. So the quality of spending improved over time even though the trend
fall in the level of the deficit was small. The CEE countries also had greater market
access that enabled them to finance their budgets by borrowing from abroad.
The empirical results with the extended data in table 14.2 reinforce the message
that good output performance was accompanied by good macroeconomic policy
performance. While policy measures of inflation stabilization (exchange-rate
dummy or inflation itself) remained significant irrespective of the size of the data
set, the fiscal variable gains significance consistently in the extended data set.
Hence, the focus on macroeconomic stabilization by domestic policy makers, rein-
forced by advice from multilateral agencies, seemed to have helped the transition
countries to grow—although the possibility of reverse causation has to be taken
into account.

14.6 The Role of Institutions

We now examine the role of institution-building in the transformation of the cen-

trally planned economies to market economies. In its 1998 volume Institutions
Matter,12 the World Bank (p. 11) defines institutions as ‘‘formal and informal rules
and their enforcement mechanisms that shape the behavior of individuals and
organizations in society.’’ They are distinguished from organizations, which are
‘‘entities composed of people who act collectively in pursuit of shared objec-
tives.’’13 Examples of formal institutions presented by the authors are laws, regula-
tions, and contracts; among informal institutions are trust, ethics, and political
Growth in Transition Economies 361

norms. Organizations include political (legislatures, political parties, government

agencies, the judiciary), economic (private firms, trade unions, business associa-
tions), and social (NGOs, schools, PTAs).
A similar definition is used in the new institutional growth literature, which
claims that institutions dominate policies. For instance, Acemoglu, Johnson, Ro-
binson, and Thaicharoen (2003, 52) define institutions as ‘‘a cluster of social arrange-
ments that include constitutional and social limits on politicians’ and elites’ power,
the rule of law, provisions for mediating social cleavages, strong property rights
enforcement, a minimum amount of equal opportunity, and relatively broad-
based access to education, etc.’’ (italics in original). This seems to be a description
of a well-run country, or a measure of governance. Similarly, but less normatively,
Rodrik, Subramanian, and Trebbi (2002, 132) specify as institutions ‘‘the rules of
the game in a society and their conduciveness to desirable economic behavior,’’ in
particular ‘‘the role of property rights and the rule of law.’’ The institutional vari-
able in Dollar and Kraay (2003) is a measure of the rule of law. Hall and Jones
(1999, 114) attribute national differences in economic performance to differences
in social infrastructure, defined as ‘‘the institutions and government policies that
make up the economic environment within which individuals and firms make
investments, create and transfer ideas, and produce goods and services.’’
Most of these definitions are consistent with those set out by the World Bank
(2002): institutions are rules and modes of behavior, including the rule of law;
organizations are ‘‘entities composed of people who act collectively in pursuit of
shared objectives.’’ The new institutional literature is sometimes taken as saying
that not much matters for the development of a country beyond its institutions,
which are themselves determined by a more or less immutable history, such as those of Eu-
ropean settlement. The italicized component mistakes econometric convenience in
choosing instruments for a substantive argument: for instance, no one observing
modern Zimbabwe would take the institutional structure—in the sense of rule of
law—as immutable. We also doubt the first statement, having seen too many pol-
icy changes that produced changes in real behavior—for example, successful eco-
nomic stabilizations. But that is not the subject of this paper.
Those who argued that the IFIs and Western advisers were missing the point in
their efforts to help the transition economies referred mainly to the need to pro-
mote the rule of law, property rights, and the investment climate (which are
closely linked). They were thus talking about institutions as we have just defined
them. However, we shall be examining both institutions and organizations, for
we find the distinction a difficult one to draw, and in particular believe that insti-
tutions in the World Bank sense require organizations to be effective, and that
in seeking to develop institutions, it is also necessary to develop organizations.
It may well be, though, that it is easier to set up organizations than to develop
362 Stanley Fischer and Ratna Sahay

institutions such as the rule of law, which may require changes in modes of social
interaction that take a long time to be implanted in social behavior.
Apart from the complication that arises in defining institutions clearly, it is also
difficult to measure them meaningfully and without subjectivity. Moreover, many
measures (such as those based on surveys conducted in a particular year) are not
available on a time series basis, which complicates our ability to run regressions
that explain short-run growth (the latter because it would be more meaningful to
see the impact of institutional development, rather than the institutions themselves,
on how output evolves in the short run). Hence, in our approach, we use qualita-
tive information and simple statistics to make our assessment on institution-
building, and interpret institution-building in as broad a sense as possible.

14.7 Institution Building During the Transition

Table 14.6 and figures 14.2 and 14.3, based primarily on indices constructed by
the EBRD, present several measures of the extent and success of the institution
building that took place in the last decade.14 The reform index is a composite mea-
sure of EBRD subindices: the financial reform index, market reform index, the
liberalization index, and a privatization index. The financial reform index, in
turn, has two subindices: banking and nonbanking reforms. The market reform
index comprises the enterprise reform index and competition policy. The liber-
alization index is made up of price, trade, and exchange-regime liberalization
indices. Finally, the privatization index measures small- and large-scale privatiza-
tion. Liberalization indices increased very rapidly, as they could be adopted and
implemented quickly. But progress is evident even in the other categories; for in-
stance, commercial law and the legal environment, where changes take longer to
implement. Table 14.6 and figure 14.2 show that while all three regions (the CEE
and Baltics, CIS-5, and CIS-7) have progressed, the CEE and the Baltics are more
advanced, and have advanced faster, than the others.
Regarding laws, indices on the legal framework have only recently (since 1997)
begun to be constructed. Available data indicates that as expected, the CEE and
the Baltics are more advanced than the others. However, as seen in table 14.6 and
figure 14.3c, the legal environment measures in the CIS-5 advanced rapidly,
reaching levels closer to those of the CEE by 2000, even though they were behind
both the CIS-7 and the CEE in 1997. Financial regulations (figure 14.3b), by con-
trast, appear to have progressed faster in the CIS-7 than in the CIS-5, even though
they were lagging in 1997. In the three categories of commercial law, financial reg-
ulations, and legal environment (company law), all three groups of countries are
more advanced in the legal environment than in commercial law or financial reg-
ulations. However, while the charts show improvements, the absolute levels for
the variables shown in figure 14.3 are low in the CIS countries. The regulatory
Growth in Transition Economies 363

quality (figure 14.3d) is good in the CEE and the Baltics but poor in the other
countries.
Table 14.7 shows high—though not perfect—correlations among all measures
of reform. In particular, if we take the legal variables (the last three rows in the
matrix) as representing institutional reform, their average correlation with the
overall reform index is about 0.7. A surprising feature of the table is that the corre-
lation between the commercial law index and measures of price liberalization,
competition policy, and privatization is relatively low—but recall that the legal
indices are available for only the later part of the period. Table 14.8 shows that
the legal variables are negatively correlated with the time a country spent under
communism—another way of saying that these reforms are more advanced in
the CEE and the Baltics. We should note also the strong correlations among the
many reform measures in tables 14.7 and 14.8: once a country is reforming, it typ-
ically advances on many fronts.
In sum, there are no major surprises here, except perhaps that there has been
progress in all areas of institutional reform, narrowly and broadly interpreted.
Nor is it a surprise that countries that have been more successful in implementing
policy reforms have also been more successful in implementing structural or insti-
tutional reforms.

14.8 Role of IFIs in Institution Building

The charge that the IFIs were unaware of the importance of institutional develop-
ment, especially of the rule of law, cannot be sustained. An early awareness of the
importance of institutions is evident in Fischer and Gelb (1991). The conditional-
ities in IFI programs generally included elements of institutional and organiza-
tional development. Moreover, much of the technical assistance was focused on
the building-up of institutions; for instance, in the case of the IMF, the develop-
ment of central banks, treasuries, tax systems, financial systems, and statistical
systems. Further, there were many bilateral non-IFI efforts at providing technical
assistance, including, for example, technical assistance by the American Bar Asso-
ciation to develop legal systems.15
Table 14.9 indicates the extent to which restrictions on foreign direct investment
and portfolio flows were lifted in the transition countries, which often reflected
key recommendations by the IMF-supported programs in these countries. In gen-
eral, the countries liberalized their capital accounts substantially. As expected,
there were fewer restrictions in the Baltics and the CEE countries than in the CIS-
7 or the CIS-5. The table also shows the extent to which these countries imple-
mented IMF-supported programs. Here the record is the best in the Baltics, where
almost all conditions were met (95 percent), followed by CEE and the CIS-7 with
similar records at around 85 percent. The CIS-5 has a lower implementation rate
Table 14.6
364

Transition Economies: Institutional Development

Banking Nonbanking Enterprise Commercial Financial Legal

Reform index reform reform reform law regulations environment

1991 2001 1991 2001 1991 2001 1991 2001 1997 2001 1998 2000 1997 2000
Baltics 1.17 3.27 1.00 3.33 1.00 2.67 1.00 3.00 3.33 4.00 3.00 3.33 3.67 4.00
Estonia 1.3 3.5 1 4 1 3 1 3 4 4 3 3 4 4
Latvia 1.1 3 1 3 1 2 1 3 3 4 3 3 3 4
Lithuania 1.1 3.3 1 3 1 3 1 3 3 4 3 4 4 4
CEE 1.80 3.11 1.40 3.10 1.30 2.60 1.40 2.60 3.20 3.50 3.00 3.00 3.30 3.60
Albania 1.1 2.6 1 2 1 2 1 2 2 2 2 2 2 3
Bulgaria 1.6 3 1 3 1 2 1 2 3 4 3 3 3 4
Croatia 1.8 3 1 3 1 2 1 3 4 4 3 3 4 4
Czech Republic 2.1 3.5 2 4 2 3 2 3 4 3 3 3 4 3
Hungary 2.3 3.6 2 4 2 4 2 3 4 4 4 4 4 4
Macedonia, FYR 1.8 2.9 1 3 1 2 1 2 2 4 2 2 2 3
Poland 2.1 3 2 3 2 4 2 3 4 3 4 .. 4 4
Romania 1.2 2.9 1 3 1 2 1 2 3 4 3 4 3 4
Slovak Republic 2.1 3.3 2 3 1 2 2 3 3 3 3 3 3 3
Slovenia 1.9 3.3 1 3 2 3 1 3 3 4 3 3 4 4
CIS-7 1.00 2.50 1.00 1.86 1.00 1.86 1.00 2.00 2.00 2.67 1.67 2.71 2.67 2.86
Armenia 1 2.8 1 2 1 2 1 2 3 2 2 2 3 3
Azerbaijan 1 2.4 1 2 1 2 1 2 1 2 2 2 2 3
Georgia 1 2.8 1 2 1 2 1 2 2 3 1 3 3 3
Kyrgyz Republic 1 2.8 1 2 1 2 1 2 2 2 3 3 3
Moldova 1 2.6 1 2 1 2 1 2 2 4 2 2 3 3
Tajikistan 1 2.3 1 1 1 1 1 2 .. 2 1 4 .. 2
Uzbekistan 1 2.3 1 2 1 2 1 2 2 3 .. 3 2 3
Stanley Fischer and Ratna Sahay
CIS-5 1.00 2.10 1.00 1.80 1.00 1.80 1.00 1.75 2.25 3.00 1.80 2.50 2.25 3.25
Belarus 1 1.3 1 1 1 2 1 .. 2 3 1 2 2 2
Kazakhstan 1 2.8 1 3 1 2 1 2 2 4 2 3 2 4
Russia 1.1 2.6 1 2 1 2 1 2 3 3 3 3 3 4
Turkmenistan 1 1.3 1 1 1 1 1 1 .. 2 1 2 .. ..
Ukraine 1 2.5 1 2 1 2 1 2 2 3 2 .. 2 3
Source: Transition Report (various issues), EBRD, London.
Note: The indices are ranked from 1 to 4, where 4 indicates the highest level of reform.
Growth in Transition Economies
365
366 Stanley Fischer and Ratna Sahay

Figure 14.2
Transition Economies: Institutional Development—Financial and Private Sectors. Source: EBRD, vari-
ous issues.

at nearly 75 percent—it is possible that the institutional capacity of these countries

to implement reforms may have been overestimated.
We examine IMF conditionality more closely in table 14.10. This table shows the
number of conditions in IMF-supported programs with transition countries dur-
ing 1993–1997.16 Performance criteria, usually imposed on macroeconomic vari-
ables, are more stringent conditions than structural benchmarks. In addition to
these two types of conditions, prior actions (not reported here) are also often
needed before a country can enter an IMF-supported program.17
Many of the structural benchmarks cover a broad definition of institution-
building, and include conditions that relate to the building of organizations; for
instance, in financial sector reform. Table 14.10 shows that countries that were
less advanced institutionally had, on average, almost twice as many structural
benchmarks as compared to the more advanced ones. These benchmarks were
fairly comprehensive, covering several areas in trade and exchange systems, tax
policy and systems, expenditure management systems, public enterprise reforms,
Growth in Transition Economies 367

Figure 14.3
Transition Economies: Legal Institutional Development. Source: EBRD, various issues.

privatization, and financial sector reforms. There were fewer conditions relating
to macroeconomic variables in the CIS countries than those relating to structural
changes or institution-building; in the CEE countries, the number of macroeco-
nomic conditions was roughly equal to the nonmacroeconomic conditions. Dur-
ing the early years, there were many benchmarks to prevent the accumulation of
arrears by the state budget, enterprises, and the central bank—these benchmarks
were imposed to curb the culture of soft budget constraints that was common in
the communist regime. Budget constraints generally hardened over time.
In Table 14.11 we present data on IMF technical assistance to the twenty-five
transition economies over the period 1989–2003. The volume of technical assis-
tance to each of the three groupings of countries in the table peaked in the period
1992–1995, but was maintained at a high level through the end of 2003, especially
to the twelve CIS countries. This assistance was aimed at building the institutional
infrastructure for the macroeconomic management of the economy.
We conclude that if there was a problem in the slow development of institu-
tions, it was not for lack of effort on the part of outside advisers. Rather, it was
Table 14.7
368

Institutions and Reforms

Finan- Non- Compe- Price Trade

Reform cial Market Lib. Banking banking Enterpr. tition liberal- liberal- Small Large Comm. Finan. Legal
index reform reform index Priv. reform reform reform policy ization ization priv. priv. law law envi.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16)

Reform index 1.00

Financial reform 0.95* 1.00
Market reform 0.92* 0.91* 1.00
Liberalization index 0.94* 0.82* 0.80* 1.00
Privatization 0.97* 0.90* 0.86* 0.89* 1.00
Banking reform 0.92* 0.95* 0.85* 0.79* 0.84* 1.00
Nonbanking reform 0.80* 0.92* 0.82* 0.57* 0.70* 0.74* 1.00
Enterprise reform 0.92* 0.90* 0.94* 0.76* 0.84* 0.90* 0.77* 1.00
Competition policy 0.77* 0.77* 0.93* 0.55* 0.68* 0.67* 0.77* 0.75* 1.00
Price liberalization 0.74* 0.56* 0.56* 0.88* 0.68* 0.61* 0.42* 0.58* 0.45* 1.00
Trade liberalization 0.90* 0.76* 0.72* 0.97* 0.83* 0.81* 0.59* 0.77* 0.55* 0.74* 1.00
Small privatization 0.91* 0.78* 0.75* 0.84* 0.96* 0.80* 0.64* 0.79* 0.61* 0.70* 0.83 1.00
Large privatization 0.90* 0.81* 0.81* 0.73* 0.94* 0.81* 0.70 0.82* 0.68* 0.59* 0.74* 0.81* 1.00
Commercial law 0.63* 0.67* 0.54* 0.42* 0.54* 0.66* 0.56* 0.60* 0.39* 0.32* 0.43* 0.43* 0.56* 1.00
Financial law 0.78* 0.81* 0.77* 0.54* 0.68* 0.73* 0.79* 0.74* 0.67* 0.54* 0.51* 0.62* 0.65* 0.65* 1.00
Legal environment 0.71* 0.73* 0.62* 0.53* 0.64* 0.71* 0.63* 0.68* 0.43* 0.42* 0.53* 0.54* 0.64* 0.92* 0.69* 1.00

Note:
Column (1) ¼ Column (2) þ Column (3) þ Column (4) þ Column (5); Column (2) ¼ Column (6) þ Column (7); Column (3) ¼ Column (8) þ Column (9); Column
(4) ¼ Column (10) þ Column (11); Column (5) ¼ Column (12) þ Column (13).
Stanley Fischer and Ratna Sahay
Table 14.8
Initial Conditions, Institutions, and Endowments
Time
Legal Initial Distance under Ethnic
Reform Commercial Financial environ- condition from commu- fraction-
index law law ment index Düsseldorf nism alization
Reform index 1.00
Commercial law 0.63* 1.00
Financial law 0.78* 0.65* 1.00
Legal environment 0.71* 0.92* 0.69* 1.00
Initial conditions index 0.40* 0.39* 0.46* 0.45* 1.00
Growth in Transition Economies

Distance from Düsseldorf
0.46*
0.35*
0.45*
0.42*
0.75* 1.00

Time under communism
0.46*
0.53*
0.56*
0.60*
0.84* 0.72* 1.00
Ethnic fractionalization
0.20*
0.29*
0.45*
0.40*
0.62* 0.49* 0.44* 1.00
369
370 Stanley Fischer and Ratna Sahay

Table 14.9
Transition Economies: Foreign Investment Indices
Foreign direct
investment Portfolio investment
restrictions indexa restrictions indexb IFI indexc
1993–1999 1996–1999 1993–1997
Baltics 1.4 0.0 95.2
Estonia 0 0.0 100
Latvia 1.4 0.0 93.1
Lithuania 2.8 0.0 92.6

CEE 1.3 0.6 84.0

Albania 1.8 1 90.9
Bulgaria 1.3 0.4 50
Croatia 0.9 0.6 80
Czech Republic 0.3 0.1 96
Hungary 1.1 0.4 98.1
Macedonia, FYR 0.8 0.9 83.5
Poland 1.6 0.5 96.4
Romania 2.8 1 76.7
Slovak Republic 0.8 0.6 ...
Slovenia 1.8 0.7 ...
CIS-7 1.6 0.7 86.8
Armenia 0.4 0 94.4
Azerbaijan 0.8 0.6 93.4
Georgia 0.8 0.5 94.6
Kyrgyz Republic 1.4 1 78
Moldova 3.1 0.6 89.5
Tajikistan 1.8 1 ...
Uzbekistan 2.8 1 70.6
CIS-5 2.6 0.9 73.8
Belarus 3.4 1 62.2
Kazakhstan 2.6 1 75.4
Russia 2.6 0.6 88.5
Turkmenistan 2.8 1 ...
Ukraine 1.8 1 69.1
Source: Foreign direct investment restriction index and portfolio investment restriction index (Gari-
baldi, Mora, Sahay, and Zettelmeyer (GMSZ), 2002), IFI index (MONA database, IMF).
a
Calculated by GMSZ 2002 based on Annual Report on Exchange Arrangements and Restrictions
(IMF). The index ranges from
0.2 to 6, where 6 reflects most restrictions.
b
Calculated by GMSZ 2002 based on Annual Report on Exchange Arrangements and Restrictions.
Ranges from 0 to 2, where 2 indicates outright prohibition of portfolio flows.
c
Measures percentage of IMF performance criteria met (100 equals all performance criteria met).
Table 14.10
Transition Economies: IMF Conditionality on Macroeconomic Policies, Reforms, and Institution Building, 1993–1997
Performance Criteria (PC) Structural Benchmark (SB)

Trade/ Tax/ Priva-

Arrangement Macro- Laws, Arrangement exchange Pricing Public expendi- Financial tization Total
type economic others type systems marketing enterprise ture sector reform Other SB

Baltics 26 — 2 1 0 2 3 1 1 11
Estonia SBA 16 — SBA 2 3 1 6 2 3 1 18
Latvia SBA 42 42 SBA 1 0 0 1 0 0 1 3
Lithuania EFF, SBA 21 8 EFF 2 1 0 0 6 1 1 11

CEE 23 2 1 2 7 7 5 2 25
Growth in Transition Economies

Bulgaria SBA 18 — SBA 3 3 2 3 7 6 4 28

Croatia EFF, SBA 35 — SBA, EFF 1 0 0 4 8 6 1 20
Czech Republic SBA 4 — — — — — — — — — —
Hungary SBA 33 — SBA 1 0 0 12 0 2 0 15
Macedonia, SBA, ESAF 5 — SBA, ESAF 0 1 0 13 10 4 8 36
FYR
Poland SBA 11 — SBA 3 0 1 5 4 3 1 17
Romania SBA 52 — SBA 1 1 8 7 11 6 0 34

CIS-7 21 — 5 4 4 12 8 7 7 46
Armenia SBA, ESAF 48 — SBA, ESAF 3 4 4 19 13 2 8 53
Azerbaijan SBA, EFF, 11 — EFF, ESAF, 8 9 4 24 13 21 19 98
ESA SBA
Georgia SBA, ESAF 9 — SBA, ESAF 1 2 4 13 7 7 5 39
Kyrgyz Rep. SBA, ESAF 20 4 ESAF, SBA 1 4 7 10 8 5 6 41
Moldova EFF, SBA 19 — SBA, EFF 12 2 2 5 3 9 1 34
Uzbekistan SBA 4 SBA 4 3 2 2 1 0 0 12

CIS-5 34 — 3 2 3 9 6 8 4 34
Belarus SBA 11 — SBA 1 1 5 5 2 3 4 21
Kazakhstan EFF, SBA 18 — SBA, EFF 2 4 3 9 11 15 3 47
Russia EFF, SBA 80 — EFF, SBA 5 1 1 19 7 6 4 43
Ukraine SBA 27 13 SBA 3 2 2 2 3 6 5 23
371

Source: MONA Database, International Monetary Fund.

372 Stanley Fischer and Ratna Sahay

Table 14.11
Transition Economies: IMF Assistance in Institution Building, 1989–2003 (Person-years)
Fiscal area Financial sector Statistics

Central Europe (10)

1989–1991 7.1 11.6 1.2
1992–1995 20.7 25.2 4.9
1996–1999 13.1 20.1 4.5
2000–2003 19.2 19.7 8.4
Baltics (3)
1989–1991 — 0.3 —
1992–1995 7.6 9.6 5.7
1996–1999 2.8 7.6 —
2000–2003 1.4 1.7 0.0
CIS Countries (12)
1989–1991 2.9 0.6 0.8
1992–1995 57.0 66.3 22.1
1996–1999 53.9 51.5 21.6
2000–2003 42.3 31.7 10.5

Source: International Monetary Fund.

Note: Number of countries in parentheses.

the sheer difficulty of developing some institutions (in the new institutional litera-
ture sense) such as the rule of law, which require changes in societal norms and
ways of doing business. Those changes inherently take a long time. Other institu-
tions, or organizations, such as a central bank, can be created more easily and rap-
idly, and a great deal of technical assistance was directed to building central
banks and fiscal systems.
It could also be argued that there was no point in trying to move ahead with
economic policy changes and organization-building until the right institutions
had been created. This is not a point we agree with, for we believe that changes
in institutions, in the sense of modes of behavior, are strengthened and sustained
by an appropriate organizational framework.

14.9 Regression Analysis: Including Institutions

In table 14.4, we add measures of organizational and institutional development to

the basic regressions reported in table 14.2. Also included in the regressions is an
index of ‘‘state capture’’—or corruption—based on work by the EBRD and World
Bank, which attempts to measure the extent to which businesses have been af-
fected by the sale of government decisions and policies to private interests. In a
Growth in Transition Economies 373

broader sense, this index is likely to capture the extent to which laws are in prac-
tice respected.
There are some changes in the regressions between tables 14.2 and 14.4. In par-
ticular, instead of using the CLI, which runs only through 1999, we replace it with
the reform index (described earlier), which runs through 2001, and which is a
more comprehensive measure of institution-building than the CLI. In addition,
the regressions are two-stage least-squares panel regressions, with the reform in-
dex in each case being instrumented by the sum of the reform indices for the other
countries in the regression.
Regressions (7) and (8) in table 14.4 in essence repeat the results of table 14.2,
and show strong statistical significance of both macroeconomic variables and
the reform index. Inflation and the exchange-rate-regime variable are strongly
correlated, and they are not separately significant if both are entered into the re-
gression. Initial conditions lose their statistical significance once the exchange-
rate-regime variable is replaced by a measure of inflation.
Regression (9) in table 14.4 adds the state capture index to the regression. The
coefficient on that variable is statistically significant, indicating that even after ac-
counting for the extent of structural reforms and institution-building reflected in
the reform index, state capture has an independent negative impact on growth.
The reform index varies within the panel of data from close to 0 to 4. This
means that the quantitative impact of the reform index on growth implied by the
regression equations is very high, possibly too high to be plausible. The state-
capture index has a range within the sample of about 25, suggesting an impact on
growth between the most and the least state-captured countries of about 3 per-
centage points—a number that we find plausible.

14.10 Conclusions

In a few years, history will have recorded the rapid transformation of centrally
planned economies into market economies, indistinguishable from other market
economies, irrespective of their level of development. It will also have recorded
the speed and magnitude of change, which were remarkable by any standards.
By and large, countries that are closer to industrialized Europe are at a more
advanced stage, but the pace of reforms and the catching up in the lesser-
developed countries of the former Soviet Union has been swift since their eco-
nomic and political liberalization.
We have demonstrated in this paper that after the large fall in output at the
start of the transition process, the resumption of growth occurred at different
times and paces, accompanied by a tightening of macroeconomic policies, rapid
structural reforms, and institutional development. We also documented that the
374 Stanley Fischer and Ratna Sahay

IFIs were well aware of the need for institutional and organizational development
in the transition process, and that a great deal of effort was devoted to helping
build institutions in the transition economies.
One final word: we noted earlier that it is sometimes assumed that a country’s
institutions are determined by a more or less immutable history, and thus must
be slow to change. Institutions can change, particularly at a time of crisis. And
our ability to predict which institutions are immutable is low. So, at the same
time as we emphasize the role of institutions in growth and development, we
should also recognize that institutions can change. And they have changed during
the transition process in the former Soviet bloc.

Notes

Stanley Fischer is Governor of the Bank of Israel, and Ratna Sahay is Senior Advisor in the Finance De-
partment at the International Monetary Fund. This paper was presented at a conference in honor of
Guillermo Calvo at the IMF on April 14–15, 2004. An earlier version was presented at the AEA Meet-
ings in Washington, D.C., in 2003. We would like to thank Andrew Feltenstein, Rodney Ramcharan,
David O. Robinson, Andrei Shleifer, and Antonio Spilimbergo for their insights and helpful comments.
The authors are grateful to Hulya Ulku, Manzoor Gill, and Ping Wang for their excellent research assis-
tance. The views expressed in this paper are those of the authors and do not necessarily reflect the
views or policies of their institutions.
1. In terms of what can be drawn from the data, radical changes in direction in an economy like
China’s or India’s, where more than twenty-five economic units (states or provinces) collect data and
differ in aspects of economic policy, geography, and other conditions, also provide a great deal of in-
formation. But in those cases, there is only one national policy at a given time.
2. See, for instance, Aslund (2001), EBRD Transition Reports, Stiglitz (2002), and World Bank (1999/
2000).
3. Many different economists and other advisers were active in the transition economies, and their ad-
vice was by no means uniform. However the mainstream advice was generally for rapid change where
possible (for example, price and trade reform) and as rapid change as possible in other areas, such as
privatization. See Aslund (2001).
4. In analyzing some questions, we used the latest available data that end earlier than 2003.
5. Belarus, Kazakhstan, Russia, Turkmenistan, and Ukraine.
6. Armenia, Azerbaijan, Georgia, Kyrgyzstan, Moldova, Tajikistan, and Uzbekistan.
7. Of course, the pretransition data have to be interpreted with some caution in that they were mea-
sured at distorted prices and the quality of products was most likely inferior. GDP is likely to be under-
estimated during the initial transition years, more so in the countries of the former Soviet Union.
8. Given data constraints, regressions with the liberalization indices (Cumulative Liberalization Index
[CLI], Liberalization Index of Privatization [LIP], Liberalization Index [LI]) have been run with data
updated only until 1999, while those with aid as a share of GDP were updated until 2001.
9. Aside from the currency boards in the Baltics and Bulgaria, most of the transition economies are
now operating with flexible exchange-rate regimes (albeit ‘‘managed’’ in most cases). Early in the tran-
Growth in Transition Economies 375

sition period, countries that pegged the exchange rate generally were able to stabilize more rapidly, but
as the Russian case vividly illustrates, adoption of the peg sometimes led to setbacks later. In other
cases (central and eastern Europe) there was a gradual transition to exchange-rate flexibility as capital
accounts were opened.
10. See, for instance De Melo, Denizer, and Gelb (1996), Berg, Borensztein, Sahay, and Zettelmeyer
(1999), and Garibaldi et al. (2002).
11. While the results are not reported in this paper, we examined empirically whether these additional
factors mattered for growth. We found that, controlling for other factors, resource-rich countries per-
formed worse than others and, as expected, the more ethnically diverse countries grew more slowly.
12. Burki and Perry (1998).

13. The distinction is attributed to the new institutional economics.

14. The top grade for any index is 4.
15. We believe it would be useful to construct a full listing of all the technical assistance provided to
any given transition country. One such listing, for the CIS-7, can be found in table 6 in Hare (2003).
Hare estimates that on average about 20 percent of the financial aid provided to the CIS-7 was for pur-
poses of institution-building. We are not aware of such information having been collected for other
transition economies.
16. The numbers are based on data collected by IMF economists Valerie Mercer-Blackman and Anna
Unigovskaya.
17. The number of conditions are counted by the test dates that were set for each of the measures.
Hence, if there were four test dates for ceilings on reserve money, the total would include these four
numbers even though they were on the same variable.

References

Aslund, Anders. 2001. Building Capitalism: The Transformation of the Former Soviet Bloc. New York: Cam-
bridge University Press.
Acemoglu, Daron, Simon Johnson, James Robinson, and Yunyong Thaicharoen. 2003. ‘‘Institutional
Causes, Macroeconomic Symptoms: Volatility, Crises and Growth.’’ Journal of Monetary Economics 50,
no. 1: 49–123.

Berg, Andrew, Eduardo Borensztein, Ratna Sahay, and Jeromino Zettelmeyer. 1999. ‘‘The Evolution of
Output in Transition Economies: Explaining the Differences.’’ Working Paper No. WP/99/73, Interna-
tional Monetary Fund, Washington, D.C.
Burki, Shahid Javed, and Guillermo Perry, eds. 1998. Beyond the Washington Consensus: Institutions Mat-
ter. Washington, D.C.: World Bank.
De Melo, Martha, Cevdet Denizer, and Alan Gelb. 1996. ‘‘From Plan to the Market: Patterns of Transi-
tion.’’ World Bank Economic Review 10: 397–424.
De Melo, Martha, Cevdet Denizer, Alan Gelb, and S. Tenev. 1998. ‘‘Circumstances and Choice: the Role
of Initial Conditions and Policies in Transition Economies.’’ Working Paper No. 1866, World Bank,
Washington, D.C.

Dollar, David, and Aart Kraay. 2003. ‘‘Institutions, Trade, and Growth.’’ Journal of Monetary Economics
50, no. 1: 133–165.
376 Stanley Fischer and Ratna Sahay

EBRD Transition Report, several issues. London.

Fischer, Stanley, and Alan Gelb. 1991. ‘‘The Process of Socialist Economic Transformation.’’ Journal of
Economic Perspectives 5, no. 4: 91–105.

Fischer, Stanley, Ratna Sahay, and Carlos Végh. 1996a. ‘‘Stabilization and Growth in Transition Econo-
mies: The Early Experience.’’ Journal of Economic Perspectives 10, no. 2: 45–66.

Fischer, Stanley, Ratna Sahay, and Carlos Végh. 1996b. ‘‘Economies in Transition: The Beginnings of
Growth.’’ American Economic Review 86, no. 2: 229–233.
Fischer, Stanley, Ratna Sahay, and Carlos Végh. 1998. ‘‘From Transition to Market: Evidence and
Growth Prospects.’’ In Lessons from the Economic Transition: Central and Eastern Europe in the 1990s, ed.
Salvatore Zecchini, 79–101. Norwell, MA: Kluwer Academic Publishers.
Garibaldi, Pietro, Nada Mora, Ratna Sahay, and Jeromino Zettelmeyer. 2002. ‘‘What Moves Capital to
Transition Economies?’’ IMF Staff Papers 48, no. 4: 109–145.
Hall, Robert E., and Charles I. Jones. 1999. ‘‘Why Do Some Countries Produce So Much More Output
Per Worker than Others?’’ The Quarterly Journal of Economics 114, no. 1: 83–116.
Hare, Paul. 2003. ‘‘Institution Building in the CIS-7: Role of the International Community.’’ Paper pre-
sented at the Lucerne Conference of the CIS-7, January. Available at http://www.cis7.org.
Havrylyshyn, Oleh, and Ron van Rooden. 2003. ‘‘Institutions Matter in Transition, But So Do Policies.’’
Comparative Economic Studies 45, no. 1: 2–24.

Helbling, Thomas F., Ashoka Mody, and Ratna Sahay. 2004. ‘‘Debt Accumulation in the CIS-7 Coun-
tries: Bad Luck, Bad Policies, or Bad Advice?’’ In The Low-Income Countries of the Commonwealth of Inde-
pendent States, eds. C. Shiells and S. Sattar, 15–50. Washington, D.C.: International Monetary Fund.
Longer version issued as Working Paper No. WP/04/93, IMF, Washington, D.C.
Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi. 2004. ‘‘Institutions Rule: The Primacy of
Institutions Over Geography and Integration in Economic Development.’’ Journal of Economic Growth 9:
131–165.

Stiglitz, Joseph. 2002. Globalization and Its Discontents. New York: W. W. Norton.
World Bank. 1999/2000. ‘‘Entering the 21st Century: World Development Report 1999/2000.’’ Avail-
able at http://www.worldbank.org/wdr/2000/fullreport.html.
World Bank. 2002. ‘‘Building Institutions for Markets.’’ World Development Report. 2002. The World
Bank, Washington, D.C.
15 Growth Collapses

Ricardo Hausmann, Francisco

Rodrı́guez, and Rodrigo Wagner

15.1 Introduction

According to the 2006 World Development Indicators, no developed economy

attained its peak per capita GDP before 2000. By contrast, 53 percent (61 out of
116) of developing economies saw their best times before that year. Of these,
more than 40 percent (26) saw their best peak GDP before 1980. This number is
even higher in particular regions of the developing world: in Latin America, 56
percent (15/27) of economies saw peak output before 2000, while in sub-Saharan
Africa the corresponding figure is 74 percent (25/34).
Recessions in the developing world are much deeper and longer than in the
developed world. Figure 15.1 presents histograms of peak-trough ratios and dura-
tion of output contractions for developing and developed countries.1 The com-
parisons are striking. Of recessions in the developed world, 87.8 percent have
peak-trough ratios less than 5 percent of peak GDP. The remaining 12.2 percent
have peak-trough ratios between 5 and 15 percent of peak GDP. For developing
countries, the corresponding figures are 48.3 percent and 25.4 percent, with the
remaining 26.3 percent having contractions where the peak-trough ratio exceeds
15 percent of peak GDP. In terms of duration, 88.9 percent of recessions in devel-
oped economies last less than four years. In the developing world, the corre-
sponding figure is 63.6 percent.
The phenomenon of deep and prolonged recessions constitutes a challenge to
macroeconomic theory. At the very least, it suggests that a vision of economic
fluctuations as trends around a stable and growing level of potential GDP is prob-
lematic for understanding growth in the developing world. It also suggests that
understanding the causes of low-frequency fluctuations may be key in accounting
for differences in development performance.
Guillermo Calvo has been one of the pioneers of the study of deep recessions in
developing economies. In a series of seminal papers,2 Calvo and his coauthors
have strived to achieve a complete characterization of a specific type of economic
378 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Figure 15.1
Comparison of Crisis Characteristics

contraction: output collapses that occur in the context of sudden stops in capital
flows in developing economies that are highly integrated into world financial
markets. Among their most important conclusions is that these output collapses
tend to be followed by rapid recoveries of output despite the lack of recoveries of
either domestic or foreign credit.
Our paper goes one step further and attempts to tackle a broader question. In-
stead of focusing on particular types of output collapses, we ask what can be
learned from studying the distribution of economic contractions across countries
and over time. In other words, we investigate whether it is possible to go beyond
the aggregate characterization shown in figure 15.1 to a deeper understanding of
the causes behind prolonged recessions. In this paper we will study both the
events that coincide with the onset of crises and the determinants of the duration
of crises. Among our main results, we find that a number of events—wars, export
collapses, sudden stops in capital flows, and high levels of inflation—coincide
with the onset of crises. However, we find that the duration of crises is particu-
larly difficult to predict. Aside from region- and time-specific effects, we find that
Growth Collapses 379

a measure of export flexibility—given by the distance-weighted density of un-

exploited export products—which may capture the flexibility of the economy to
adapt to external shocks, is an important predictor of crisis recovery. We also find
that both conditional and unconditional hazard rates are declining in time, sug-
gesting that countries have a harder time exiting crises the longer that they spend
in them.
Francis Galton once criticized his statistician colleagues because they ‘‘limited
their inquiries to averages, and do not seem to revel in more comprehensive
views’’ (1889, 62). Sir Galton could have just as well been referring to modern
growth empirics. Ever since Barro’s (1989) seminal contribution, empirical work
on economic growth has to a great extent been concerned with explaining differ-
ences in average growth among countries. Despite Pritchett’s (2000) call to think
closely about the ‘‘hills, plateaus, mountains, and plains’’ characterizing the
growth data, until recently very little work had been carried out attempting to ex-
plain the substantial differences in patterns taken by growth series with similar
first and second moments.3
Among the notable exceptions are the work of Calvo, Izquierdo, and Mejı́a
(2004); Calvo, Izquierdo, and Loo-Kung (2005); and Calvo, Izquierdo, and Talvi
(2006). These authors have concentrated on studying the characteristics of output
collapses that are associated with sudden stops in capital flows in emerging-
market economies. They have found that some of these sudden stops—particu-
larly those associated with systemic international capital market turmoil—are
followed by rapid output recoveries arising in spite of the limited reconstruction
of credit markets.
By concentrating on a very specific form of output collapses, Calvo and his
coauthors have been able to take advantages of the similarities between compara-
ble episodes. Other works in this literature have taken a more general approach,
attempting to exploit the differences arising in broader samples. Ben-David and
Papell (1998), for example, study episodes of growth slowdowns, defined by sta-
tistically significant breaks in the time-series trends, and find differences between
the magnitude and timing of these slowdowns among developed and developing
countries. Pritchett (2000), in contrast, studied declines of at least two percentage
points in trend growth rate—without distinguishing between significant or insig-
nificant trends. Neither of these papers makes a serious attempt to understand the
causes behind the onset or magnitude of growth slowdowns.
Two recent papers have made a more direct attempt to understand the
dynamics of growth decelerations. Cerra and Sexena (2005) study the long-run
implications of economic recoveries. In essence, they show that after economic
contractions, GDP growth does not typically return to trend growth. Additionally,
they show that the incidence of crises is an important reason for unconditional
380 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

divergence in the postwar growth data.4 Reddy and Miniou (2006) study ‘‘real
income stagnations,’’ which they define as long and sustained periods of negative
growth. Their definition is perhaps the one that comes closest to ours in the litera-
ture.5 They find that countries that suffered spells of real income stagnation were
more likely to be poor, located in Latin America or Africa, undergoing armed con-
flict, and highly dependent on primary exports.6 The main characteristics of these
and other studies dealing with long-run economic contractions are summarized in
table 15.1.
One common feature of all these studies is the lack of any explicit method to
deal with unfinished episodes of contraction (meaning those that have not ended
by the last year of the dataset). Of the ten papers listed in table 15.1, five take an
explicit decision to omit or truncate the unfinished contraction episodes; two
additional ones (Ben-David and Pappell 1998 and Pritchett 2000) adopt a method-
ology that is incapable of handling these breaks, thus also dropping them in prac-
tice. An eighth paper (Calvo, Izquierdo, and Talvi 2006) adopts a very restrictive
definition of crises—as drops in aggregate output that occur in the context of sys-
temic capital market turmoil—with the result that recoveries are very rapid and
in their main sample there are no episodes of unfinished crises. This decision
comes at a cost: there is no obvious reason from a growth theory perspective why
one should concentrate on aggregate instead of per capita or per worker output.
The two remaining papers analyze the change in growth rates among two prede-
termined periods and thus do not have to deal directly with this issue. In those
papers, the issue of unfinished crises is avoided by choosing an arbitrary cut-off
date to calculate changes in growth rates.
The issue of dealing with unfinished episodes is important because a substan-
tial number of crisis episodes have generally not finished by the end of the period
for which continuous data is available. Using the definition of crisis that we will
adopt in this paper (periods of continuous negative average growth), we find
that 16.07 percent of the events defined as crises are censored. While this number
may not seem large, what is problematic is that it is asymmetrically composed of
long crises episodes. Only 4.95 percent (18/363) of crisis episodes lasting less than
5 years are censored, while 78.5 percent (22/28) of those lasting more than 24
years are censored. To take a simple example of how this can affect even the most
basic conclusions of research, if we had decided to drop the censored observations
we would have calculated the mean duration of contractions to be 3.96 years. If
we include the censored observations, we find that the mean duration is 6.05
years. Even this estimate is surely an underestimate of the expected duration of a
crisis, because the real duration of censored episodes is unobserved. If we were to
fit the simplest possible duration function—with an exponentially distributed
hazard rate—to this data, we would estimate an expected duration of 7.21 years.
Growth Collapses 381

Inferences about the behavior of countries during crises will thus end up being
automatically biased to reflect the performance of economies that are more suc-
cessful at dealing with crises.
The key contribution of this paper is to analyze the determinants of the dura-
tion of economic contractions using econometric methods designed explicitly to
deal with censored observations. These methods, broadly grouped under the
labels ‘‘duration analysis’’ and ‘‘survival analysis,’’ have gained increasing promi-
nence in modern economics primarily through their application in microecono-
metric settings.7 Their key defining characteristic is the joint use of information
on duration of censored and uncensored episodes in derivation of the likelihood
function. Despite their obvious appeal for the study of many macroeconomic phe-
nomena, their application in macroeconomic analysis has been limited. The larg-
est proportion of papers that use duration analysis in macroeconomic settings
study duration dependence of business cycles in developed countries (Mills 2001;
Bodman 1998; Di Venuto and Layton 2005; Diebold and Rudebusch 1990;
Mudambi and Taylor 1991; Sichel 1991). Applications to developing countries are
scant. One recent exception is Mora and Siotis (2005), who estimated a conditional
duration model of recessions in a sample of twenty-two emerging markets. Their
specification is limited by the small size of their sample and the fact that they con-
sider only external factors.
The rest of the paper is organized as follows: section 15.2 discusses our defini-
tion of crises and our data set. Section 15.3 takes a first look at the summary statis-
tics of crises. Section 15.4 examines the factors that coincide with the onset of
crises, while section 15.5 discusses the results of regressions to account for the du-
ration of crises. Section 15.6 concludes.

15.2 Defining the Event

For the purposes of the analysis in this paper, we define a crisis as an interval that
starts with a contraction of output per worker and ends when the value immedi-
ately preceding the decline is attained again. Thus a crisis that occurs between
times t and t þ j has by definition an average growth rate equal to 0 during that
period, and negative during the period between t and any t þ j
e for any e < j.
A crisis cannot start if the country is already in crisis. In other words, if yt < yt
1 ,
a crisis will start only if either (1) there is no yt
j > yt
1 , or, (2) if there is a
yt
j > yt
1 , there is a yt
p > yt
j with p < j.
This definition is illustrated in figure 15.2. This definition has several advan-
tages for our object of study. First, it covers a selected group of growth decelera-
tions. In particular, it covers growth decelerations where growth goes from being
positive (as it must be before reaching a peak) to being negative between t and
Table 15.1
382

Summary of Existing Literature

Crisis that
Paper Time series Countries Database Freq Breaks GDP used Taxonomy did not end
Ben-David and 1955–1993/ 74 PENN 5.5 Annual One or zero Real per capita Slowdown (if Method does
Papell (1998) 1950–1993 breaks, PPP D
growth but not apply; in
depending on growth > 0). practice they
significance Meltdown (if are ommitted
growth < 0)
Rodrik (1999) 1960–1989 110 WDI 15 years One and the Real per capita Collapse Not pertinent
same for all growth, PPP
countries in 1975
Ali and 1965–1996 62 WDI 10–21 One and the Real per capita Collapse Not pertinent
Elbadewi (1999) developing years same for all GDP growth
countries countries (1965–
1974)—(1975–
1996)
Pritchett (2000) 1960–1992 111 PENN 5.6 — One or zero Real GDP, PPP Plateau, Method does
(some until breaks, 1985 mountain, hill, not apply; in
1985) depending on plains practice they
significance are ommitted
Ros (2005) 1960–1999 70 PENN 6.1 Annual Not breaks, but Real per capita Collapse Dichotomous
developing events of crisis GDP classification;
countries in practice
equivalent to
truncation
Cerra and 1960–2001/ 192/154 WDI/ Annual Not breaks, but Real GDP Recessions Omit
Sexena (2005) 1960–2000 PENN 6.1 events of crisis growth
Unbalanced
panel
Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner
Cespedes and de Data 1980– 71 WDI/IMF Annual Not breaks, but Real GDP Product Omit
Gregorio 2005 2004 (But Y WEO events of crisis contraction
trend built episode or
since 1960) recession
Reddy and 1960–2001 119 WDI Annual Not breaks, but Moving average Real income Truncate
Minoiu 2006 events of crisis ð1=3fðt
1Þ þ ðtÞ stagnation
Growth Collapses

þ ðt þ 1Þg of Per
capita GDP,
constant LCU
Blyde, Daude, 1960–2003 71 PENN 5.6 Annual Not breaks, but Real PPP Collapse Omit
and Fernandez events of crisis
Arias (2006)
Calvo, Izquierdo, 1980–2003 31 (emerging WDI/IMF Annual Not breaks, but Real GDP Episode of No cases in
and Talvi (2006) markets WEO events of crisis output main sample
covered by contraction
EMBI)
383
384 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Figure 15.2
Graphical Example of Crisis Definition

any t þ j
e. But it does not cover all growth decelerations. In particular, we

wanted to exclude the growth decelerations that one would find naturally in neo-
classical growth models, such as those generated by convergence to the steady
state. Our definition allows us to restrict ourselves to episodes where growth
decelerates to a negative rate.
Our interest in negative growth rates is related to our specialization to duration
analysis. Although episodes of negative growth are a standard area of interest
in macroeconomics, sustained episodes of negative growth are much harder to
explain. Although it is certainly possible to account for negative growth using a
neoclassical model, this requires that the economy suffer strong and sustained
deteriorations in its fundamentals that are sufficiently strong to offset the effects
of technical progress for periods of one or more decades. Thus long, sustained epi-
sodes of negative growth are only consistent with very large adverse changes in
fundamentals, or dynamics in which those fundamentals do not recover.
In order to calculate our crisis indicator, we used GDP per person of working
age—henceforth GDPW—in constant local currency units from World Bank
(2006). We use working-age population because it is the best widely available
proxy for the size of the labor force, which most closely matches the concept of la-
bor input in growth theory. We use local currency units because we are not inter-
ested in comparisons of levels among countries, and differences in purchasing
power parity (PPP) adjustment factors across times can be arbitrary and have
unintended consequences (see Reddy and Minoiu 2006).
Our crisis definition immediately suggests several measures of crisis intensity
that we calculate and use in the rest of the analysis. These are:
1. Duration: the number of years elapsed between the beginning and the end of
the crisis.
Growth Collapses 385

2. Peak-trough ratio: the ratio between the value of GDPW immediately preceding
the crisis and the lowest value it attains during the crisis, expressed as a ratio of
the peak value.
3. Integral measure of years of lost output: this is the sum of all the gaps between
peak and the GDP for each year of the crisis. It is an approximation to the integral
above the output series and below a horizontal line drawn at the peak output,
expressed as a fraction of peak output.

15.3 Collapses Big and Small

We start out by examining the general characteristics of crises and how they vary
across regions. These summary statistics are presented in table 15.2. The first strik-
ing fact that we observe is that there is a wide dispersion of crisis characteristics in
the world sample. The majority of crises observed do indeed appear to be of the
typical business-cycle type: the median crisis duration is just two years, while the
median peak-trough ratio and integral measure of years lost are respectively four
and six percent of precrisis GDP. These median values, however, come from a
highly skewed distribution. The average duration of crises, at 6.05 years, is three
times as high as the median duration, while the mean-to median ratios of the
peak-trough ratio and product-years lost are respectively 2.6 and 15.9.
High dispersion of crises characteristics is also a feature of interregional varia-
tions. Although all regions appear to be hit by some short-lived recessions, long-
lived recessions are much more prevalent in the developing world. Thus, while
the mean duration of a recession in industrialized countries is only 2.52 years, in
Latin America it is 6.88 years and in Central and Eastern Europe it is 9.74 years.
The median peak-to-trough ratio of a crisis, for example, is 6 times as large in Af-
rica and 29 times as large in Central and Eastern Europe as in the industrialized
world.
These differences are striking, but what is even more striking is that they almost
surely underestimate the magnitude of the differences in crisis duration, both
around the world median and across regions. The reason is a simple one, and
will form the foundation for much of our analysis. Shorter crises are much less
likely to be censored than long crises. Almost by definition, the longer a crisis is,
the more likely that it will be interrupted, either by the end of the sample or by
an interruption in data reporting. Therefore we can never actually observe the
complete duration of very long crises. It is hard to exaggerate the magnitude of
this difference. In our sample, only 19 out of 387 crises (4.9 percent) that last less
than five years are censored, while 67 of 148 crises (45.3 percent) lasting more
than five years are censored. Take the example of Venezuela, whose GDP per
working-age population was 39.4 percent lower in 2004 than in its peak in 1970.
Table 15.2
Summary Statistics of Crises by Region
386

Number
of obser- Standard 25th 75th
vations Mean deviation Minimum percentile Median percentile Maximum
Duration of crisis
Africa 151 8.14 10.09 1 1 3 13 43
Asia 42 4.79 5.71 1 1 2 7 24
Central and Eastern Europe 34 9.74 6.38 1 1 12.5 15 19
East Asia and Pacific 46 3.78 4.23 1 1 2 5 20
Industrialized Countries 90 2.52 2.88 1 1 2 3 19
Latin America and Caribbean 109 6.88 8.92 1 1 3 7 34
Middle East and North Africa 63 5.13 7.56 1 1 2 4 27
Total 535 6.05 8.02 1 1 2 7 43
Peak to trough
Africa 151 0.13 0.16 0.00 0.02 0.06 0.19 0.95
Asia 42 0.08 0.12 0.00 0.02 0.04 0.08 0.52
Central and Eastern Europe 34 0.29 0.24 0.00 0.06 0.29 0.45 0.77
East Asia and Pacific 46 0.08 0.08 0.00 0.02 0.05 0.11 0.35
Industrialized Countries 90 0.02 0.03 0.00 0.01 0.01 0.03 0.13
Latin America and Caribbean 109 0.10 0.13 0.00 0.01 0.05 0.15 0.62
Middle East and North Africa 63 0.12 0.20 0.00 0.01 0.04 0.12 0.91
Total 535 0.11 0.16 0.00 0.01 0.04 0.13 0.95

Product-years lost (integral)

Africa 151 1.42 2.94 0.00 0.02 0.10 0.93 16.40
Asia 42 0.51 1.16 0.00 0.02 0.07 0.26 5.51
Central and Eastern Europe 34 2.71 2.81 0.00 0.06 2.15 4.03 9.79
East Asia and Pacific 46 0.31 0.54 0.00 0.02 0.07 0.26 2.46
Industrialized Countries 90 0.07 0.16 0.00 0.01 0.02 0.06 1.16
Latin America and Caribbean 109 1.05 2.35 0.00 0.01 0.06 0.62 13.78
Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Middle East and North Africa 63 1.16 2.88 0.00 0.01 0.06 0.20 12.73
Total 535 1.00 2.36 0.00 0.01 0.06 0.44 16.40
Growth Collapses 387

Even if Venezuela were to experience extremely high growth after 2004, its crisis
is likely to last considerably longer than 34 years.
One possible mechanism for dealing with this problem would be to drop crises
that have not ended from the sample. As discussed in our review of the literature,
this is the standard choice in much of the literature dealing with this problem.
However, it is a highly inefficient solution, as it entails throwing out a very high
proportion of the crises that are most interesting for our purposes: those that are
very long. Therefore we adopt the alternative solution in this paper, which is to
use the estimation techniques of duration analysis (also called survival analysis in
the biometric literature), which are explicitly designed to deal with censored dura-
tion times.
The essence of duration analysis is to explicitly consider unfinished crises as
arising out of the same distribution as finished crises. All countries are assumed to
be characterized by a (possibly time-dependent) probability of leaving a crisis at
any moment of time given that they are still in the crisis state. That probability—
called the hazard rate—is affected by country-specific characteristics that can be
summarized by a vector of independent variables that may include country-
specific effects. The behavior over time of this rate is very interesting. If the hazard
rate is increasing over time, it means that as time elapses, the probability that a
country will recover to its precrisis GDPW level increases. This is the type of be-
havior that one would expect if the precrisis level of GDPW was an equilibrium
out of which the economy was perturbed by a temporary shock. However, if
we see that the longer an economy spends in a crisis the harder it is for it to get
out of it—as would be implied by a declining hazard rate—this would suggest
either that the economy suffered strong blows to its fundamentals—so that its
steady-state level of GDPW shifted downward—or that it jumped to an inferior
equilibrium.
Furthermore, in the presence of positive technological change one would al-
ways expect that, if enough time has elapsed after the initial shock, an economy
would return to its precrisis level of GDP even if it initially suffered an adverse
shock to its fundamentals. The reason is that for a given level of fundamentals,
the probability that an economy hits any level of GDPW with positive technologi-
cal change must tend to 1 as time increases. Declining hazard rates over the very
long run would thus be consistent not only with an initial adverse shock, but
rather with continuously deteriorating fundamentals as may occur because of the
political system’s endogenous reaction to the adverse shock.
A first intuitive way to summarize the information in our data set regarding the
characteristics of crises is thus to plot the unconditional hazard functions. These
hazard rate estimates are plotted in figure 15.1. They are derived as smoothed ker-
nel density estimates of the Nelson-Aalen cumulative hazard function and track
388 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Figure 15.3
Smoothed Hazard Function by Region

the probability of exiting a crisis conditional on being in the set of countries that
have not exited the crisis when t periods have elapsed since falling into it.
The remarkable result that emerges from figure 15.3 is that hazard rates, both
within regions and for the world as a whole, do not appear to be increasing.
Rather, they are either flat or decreasing over time. Two distinct patterns appear
to emerge. The first one is that of industrialized countries and East Asia and the
Pacific, which have two humps. Within these groups, countries either get out of
their crises quickly, or get out later. Even the second group, however, tends to get
out of crises much earlier than those in many other regions. The rest of the
regions—including the pooled world sample—have hazard rates that are gener-
ally flat or decreasing, although confidence intervals obviously become wider as
one reaches higher durations. Hazard rates also tend to be much lower in non-
industrial countries, reaffirming the conclusion that crises are much more likely to
be of the short duration business-cycle type in industrialized countries than in un-
derdeveloped regions.
The confidence intervals drawn around the unconditional region-level hazard
rates in figure 15.1 are quite wide. This is simply a reflection of the fact that very
long crises are sufficiently rare within each region so as to make it difficult for us
Growth Collapses 389

Figure 15.4
Smoothed Hazard Function: Industrial and Nonindustrial Countries

to precisely estimate the hazard rate. In figure 15.4 we show the result of pooling
the sample of nonindustrial countries and comparing it with that of industrial
countries. The hazard rate for developing countries is clearly declining up at least
to a period of thirty years since entry into crisis. It is also considerably lower than
that of industrial countries. Whereas most industrial countries have a probability
higher than 20 percent of leaving the crisis during each of the first few years in it,
for developing countries that probability stays below 10 percent.
The fact that substantial interregional differences across survival functions exist
can be tested systematically through several standard tests for equality of survi-
vor functions. These tests are reported in table 15.3. Column 1 shows the result of
testing the null hypothesis that all regions have the same survival function. It is
thus the statistical counterpart of figure 15.3. Column 2 tests the null that indus-
trial countries have the same hazard function as developing countries, forming
the statistical counterpart of the comparison in figure 15.2. Lastly, column 3 eval-
uates the null hypothesis that all groups of nonindustrial countries have the same
survival function. All three homogeneity hypotheses are easily rejected. The data
thus indicates that there is substantial interregional heterogeneity in the recovery
from adverse shocks.
The logical question that this analysis leads us to regards the source of these dif-
ferences across regions. Is it a reflection of the fact that different regions are hit by
390 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Table 15.3
Tests for Equality of Survivor Functions
Only
Industrialized vs. nonindus-
All regions nonindustrialized trialized
Log-rank 54.49*** 33.78*** 21.31***
Cox 38.24*** 20.78*** 16.72***
Wilcoxon 37.36*** 20.05*** 17.84***
Tarone-Ware 45.38*** 25.95*** 19.82***
Peto-Peto 43.76*** 24.58*** 19.72***

different types of crises? Or is it rather a consequence of the fact that regions differ
in their capacity to react to adverse shocks? In order to answer these questions,
we must understand what factors drive countries to fall into crises and what fac-
tors determine the duration of crises once a country has entered into them.8 These
are the questions that we tackle in the next two sections.

15.4 Why Do Countries Fall Into Crises?

We approach the study of the determinants of crisis occurrence through the esti-
mation of random effects probit regressions on a panel of countries. The basic
idea of this specification is to allow us to understand the potential relative triggers
of a country falling into a crisis. In essence, we investigate whether and how dif-
ferent possible instigators correlate with the incidence of the crisis. We look at
a battery of potential causes of crises, ranging from the usual suspects—natural
disasters, wars, sudden stops, and export collapses—to other less conventional
factors.
Any exercise of this type may be subject to several types of specification bias.
Simultaneity bias is one—though not necessarily the most important one—of
them. Others include omitted variables, inadequacy of the linear specification,
and incorrect assumptions about the error covariance structure. In the case of
some potential explanatory variables—such as natural disasters—endogeneity
may be less of a problem than some of these other biases. We do not make an
effort to search for appropriate instruments for all of our explanatory variables be-
cause we view our exercise as a primarily exploratory attempt to investigate what
factors coincide with the onset of crises, rather than to test tightly specified causal
hypotheses. If our exercise is successful, it would help build a typology of crises
according to the key factors that occur at the same time as the crisis.
Our baseline specification will thus be

Pit ¼ Fðb 0 Xit þ hi Þ ð15:1Þ

Growth Collapses 391

where Pit is the probability that country i falls into a crisis at time t, Xit is a k  1
vector of conditioning variables (generally including a constant term), hi is a
country-specific effect, b is the k  1 vector of parameters to be estimated, and
Fð:Þ is the standard normal distribution. The random effects probit specification
models hi as following a Nð0; s 2 Þ distribution.
Note also that our event of interest is whether a country enters a crisis or not.
According to our definition of crises, a country is obviously a candidate for enter-
ing a crisis only if it is not already in one. Therefore we exclude from the sample
all country-years in which the country is in the midst of a crisis. This decision is
based on the fact that these country-years contain no relevant information about
the process of entering into crises.
Neoclassical growth theory views output collapses as arising out of adverse
shocks that either move the steady-state level of income or alter the per capita
stock of physical and human capital. It is thus logical to start by looking at signifi-
cant disruptions of an economy’s productive framework that may either affect its
capacity to convert inputs into outputs or directly affect its stock of accumulated
productive assets. Several candidates come to mind. Perhaps the first two are
natural disasters and wars. These tend to constitute large, generally exogenous
shocks that generate significant disruptions to a society’s capacity to produce.
They will also commonly directly affect the capital stock. The speed of some
postwar recoveries is indeed a commonly cited observation in defense of the con-
ditional convergence hypothesis. Two other potential candidates are export col-
lapses and sudden stops in capital flows. The latter has been well developed in
the literature, particularly through the pioneering work of Guillermo Calvo. Ex-
port collapses, while much less studied, tend to crop up in the analysis of many
episodes of collapse (see Hausmann and Rodrı́guez 2006).
Our data for natural disasters is drawn from the International Disaster Data-
base, maintained by the Office of U.S. Foreign Disaster Assistance (OFDA) and
the Center for Research on the Epidemiology of Disasters (CRED), which has in-
formation on the physical and human damage caused by 14,877 natural disasters
that occurred between 1960 and 2006. We define a substantial occurrence of natu-
ral disasters if either the number of people affected by all natural disasters occur-
ring in a given year is greater than 1 percent of the population, or the number of
people killed by all natural disasters occurring in a given year is greater than 0.1
percent of the population. Our proxy for natural disasters will be an indicator
variable that will be 1 if there was a substantial occurrence of natural disasters in
t, t
1, or t
2.
Regarding the occurrence of wars, we draw our data from Kristian Gleditsch’s
(2004) Expanded War Data Set, which covers all inter- and intrastate wars between
and within independent states since 1816. We build an indicator variable that
equals 1 if the country was involved in an interstate or civil war in t, t
1, t
2,
392 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

or t
3. We also build separate dummies for civil and interstate wars respec-
tively. Declines in exports are measured using data on merchandise exports from
World Bank (2006). We use the log difference in exports between t and t
5 as
our indicator of an export performance.
For the purposes of defining sudden stops in capital flows, we closely follow
the definition of Calvo, Izquierdo, and Mejı́a (2004). According to their definition,
a sudden stop is a year-on-year decline in capital inflows containing at least one
year in which the decline exceeded two standard deviations from its sample
mean. The sudden stop starts when the fall exceeds one standard deviation from
the sample mean and ends when it is above one standard deviation. Our measure
of private capital flows comes from World Bank (2006) and consists of private
debt and nondebt flows. Note that our measure differs from two other measures
used by Calvo, Izquierdo, and other coauthors in some of their work. In particu-
lar, it differs from the systemic sudden stops (Calvo, Izquierdo, and Talvi 2006)
measure given by the episodes of sudden stops that coincide with increases in the
aggregate EMBI spread. Aside from the difficulty in obtaining a measure of capi-
tal market turmoil relevant for nonemerging-market economies, our key reason
for using the broader category is our interest in the broader phenomenon of
declines in capital flows. Our measure also differs from the definition found in
Calvo, Izquierdo, and Talvi (2006) that combines falls in capital flows with output
collapses. The rationale for this is quite simple: we are attempting to understand
the capacity of sudden stops to predict output collapses, whereas most of the
work of Calvo and his coauthors is concentrated on understanding the dynamics
of output collapses that coincide with a decline in capital flows.9 We shall discuss
the sensitivity of our results to alternative measures of sudden stops later.
Aside from these natural candidates, we try a number of additional explanatory
variables that may be associated with the onset of crises. We measure the level of
a country’s democracy by its score on the polity index (Marshall, Jaggers, and
Gurr 2004). We also measure political transitions by the change over time in its
polity index. We use a measure of the log of 1 plus the inflation rate as a proxy
for macroeconomic instability. We also attempt to control for a set of additional
potential explanatory variables such as years of primary, secondary, and total
schooling (from Barro and Lee 2000), the rule of law (ICRG 1999), life expectancy
at birth, percent of the population that is urban, and number of telephone main-
lines per capita (from World Bank 2006). All of our estimates include region and
decade dummies.
One additional variable of interest that we will study is the measure of the
value-weighted density of the unexploited product space elaborated by Haus-
mann and Klinger (2006). This measure is designed to capture the sophistication
of the goods that an economy could produce—but is not producing—with its
productive assets. It is built as a weighted average of the sophistication of all po-
Growth Collapses 393

tential export goods, where the weights are given by the distance between these
goods and the economy’s present export basket. The measure of distance in the
product space is calculated based on the frequency with which particular good-
pairs are exported by the same country, while the measure of sophistication is
given by the average income of the countries that export that good, which we call
PRODYjt , as originally proposed by Hausmann, Hwang, and Rodrik (2007). More
formally, let the proximity between two goods in the product space be given by
the minimum of the conditional probabilities of exporting each one of those goods
given that you are exporting the other one:
jijt ¼ minfpðxit j xjt Þ; pðxjt j xit Þg; ð15:2Þ

where pðxit j xjt Þ is the probability that you have revealed comparative advantage
in good i at time t given that you have revealed comparative advantage in good j
at time t. Let xcjt be an indicator variable that takes the value 1 if country c has a
revealed comparative advantage greater than 1 in good j at time t, and 0 other-
wise. Then the option value of a country’s unexploited export opportunities can
be measured by
X X jijt
open_ forestct ¼ P ð1
xcjt Þxcit PRODYjt : ð15:3Þ
i j i jijt

open_ forest thus captures the flexibility of an economy’s export basket in that it
measures the value of the goods that it could be producing with the inputs that it
currently devotes to its export production. open_ forest is particularly appropriate
for thinking about an economy’s capacity to react to adverse export shocks. To fix
ideas, suppose that an economy’s exports of good i were to disappear overnight.
This could happen, for example, as a result of the exhaustion of a natural re-
source, of the emergence of a new lower-cost supplier in international markets, or
of the invention of a cheap substitute for that good. We know that this economy
must shift resources into a new export sector. jijt can be interpreted as our best
guess of the probability that a country will shift resources into good j, and
jijt ð1
xcjt Þ can be seen as our best guess of the probability that it will export a
good j that it is not already exporting. jijt ð1
xcjt ÞPRODYjt is the expected value
(measured in terms of the sophistication of exports) from exporting that good,
making open_ forest the weighted average of that expected value over all goods
that the economy currently exports. In other words, open_ forest reflects the ex-
pected value of an economy’s next-best export basket if it moved out of its current
basket of exports.
To the extent that many crises are precipitated by declines in an economy’s key
export sectors, we expect open_ forest to be a good indicator of the economy’s
flexibility in moving to a new export basket in the face of those declines. We
394 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

thus expect open_ forest to be a significant determinant of the duration of crises.

Open_ forest may also be important in stopping crises before they materialize. The
reason is that export declines in traditional sectors may occur at the same time as
new sectors are moving in to absorb unused resources. The higher the productiv-
ity of the newer export sectors, the less likely that the initial export collapse will
cause the economy to enter a period of negative growth. In order to test these
hypotheses, we will include open_ forest in the probit regressions of this section as
well as the duration regressions of the next section.
Our baseline results are presented in table 15.4. Column 1 presents the result of
regressing the probability of falling into crisis on the log of PPP-adjusted real GDP
per working-age population and a set of continent and time dummies. The GDP
term is negative, indicating that richer countries are less prone to economic crises,
although the coefficient is not significant, in contrast to several of the continent
dummies, which are highly significant. In column 2 we add the log change in real
merchandise exports. Its coefficient is strongly significant with the expected nega-
tive sign, while the coefficient on GDP remains insignificant. The next column
adds wars, natural disasters, and sudden stops. While wars and sudden stops are
highly significant with the expected sign, the result on natural disasters is surpris-
ing. The coefficient is far from statistical significance (p ¼ .36) and furthermore
has the wrong sign. This is particularly surprising since natural disasters are the
one variable in the data set about whose endogeneity we are less worried. In or-
der to confirm that its coefficient is not being distorted by the endogeneity of other
explanatory variables, we reestimated the equation, dropping all variables except
for natural disasters and time and continent dummies. This exercise (not shown in
the table) still gives an insignificant, though positive, coefficient (p ¼ .47). Column
4 adds four additional variables: inflation, political transitions, open_ forest, and
the level of democracy. The first two are strongly significant, open_ forest is border-
line significant, and democracy is clearly insignificant. All have the expected sign:
inflation and political change are associated with greater propensity toward cri-
ses, while open_ forest and democracy are associated with lower crisis prevalence.
Wars now drop to borderline insignificance. This appears to be more the result of
reduced sample size: if we reestimate the equation of column 3 for the same num-
ber of observations as in column 4, we get a very similar coefficient as when we
include the additional variables (.451, t-stat ¼ 1.7). Finally, in the last column, we
drop the clearly insignificant natural disasters and democracy variables. The co-
efficient signs and significance tests are unaffected, with the exception of wars,
which goes from borderline insignificance to borderline significance. Declines in
merchandise exports, sudden stops, high inflation episodes, and political transi-
tions are individually strong predictors of the onset of crises, while wars and
open forests have a weaker but still significant association with crisis onset.
Table 15.4
Random Effects Probit Regressions, All Countries
Dependent variable: Probability of
falling into a crisis (1) (2) (3) (4) (5)
Log GDP per working-age person
0.014 0.051 0.002 0.044 0.015
(0.24) (0.85) (0.03) (0.44) (0.15)
Latin America 0.491 0.590 0.610 0.250 0.238
(4.01)*** (4.61)*** (4.48)*** (1.23) (1.2)
Africa 0.409 0.547 0.479 0.072 0.047
(2.58)*** (3.32)*** (2.64)*** (0.24) (0.16)
South and Central Asia
0.136 0.025
0.022
0.303
0.351
(0.73) (0.13) (0.1) (1.08) (1.28)
East Asia and Pacific 0.073 0.217 0.187
0.122
0.162
(0.53) (1.54) (1.11) (0.54) (0.72)
Central and Eastern Europe 0.190
0.236
0.171
0.105
0.185
(1.24) (0.96) (0.55) (0.26) (0.46)
Middle East and North Africa 0.430 0.547 0.492 0.149 0.124
(3.04)*** (3.78)*** (3.19)*** (0.64) (0.58)
1970s 0.310 0.494
(2.59)*** (1.47)
1980s 0.319 C0.177
0.049 0.527 0.026
(2.93)*** (1.65)* (0.42) (1.64) (0.19)
2000s C0.486 C0.298
0.495
(3.77)*** (2.12)** (1.48)
1990s 0.231 C0.287 C0.206 0.394
0.104
(2.1)** (2.67)*** (1.77)* (1.23) (0.72)
Log change in real merchandise exports C0.446 C0.450 C0.433 C0.416
(5.01)*** (4.73)*** (3.04)*** (2.95)***
War 0.655 0.415 0.468
(2.97)*** (1.56) (1.81)*
Natural disaster C0.121 0.070
(0.92) (0.41)
Sudden stop 0.167 0.236 0.227
(2.1)** (2.32)** (2.24)**
Log of inflation 0.999 1.009
(3.23)*** (3.28)***
Change in polity indicator 0.317 0.365
(2.48)** (2.92)***
Open forest C0.148 C0.161
(1.73)* (1.91)*
Democracy
0.003
(0.27)
Constant C1.293 C1.382
1.073
0.007 0.944
(2.14)** (2.26)** (1.56) (0) (0.67)
Observations 2001 1877 1688 1054 1062
Countries 158 147 136 83 83
Percent crises predicted 0.0% 0.3% 1.4% 5.7% 6.1%
Pseudo-R 2 3.0% 5.3% 6.0% 7.5% 7.9%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%,
***
1%.
396 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

What can we say about the global significance of these variables? There are var-
ious ways to address this question. One is by noting that the addition of the
explanatory variables drives down the significance of the continent and time
dummies. All continent dummies except for South and Central Asia and Central
and Eastern Europe are significant and positive in column 1. Since the omitted
category is industrialized countries, this indicates a higher unconditional proba-
bility of falling into crisis for nonindustrialized countries. By column 5, those
effects have disappeared—indeed, the South and Central Asia dummy has turned
significantly negative. It thus appears that our explanatory variables account for
the differences between the developing and developed worlds in the incidence of
crises.10
A second way to address this question is by looking at some goodness-of-fit
indicators. These are reported in the bottom two rows of table 15.4. The result of
this exercise is not as encouraging. The first column shows the percentage of crises
that are accurately predicted by our models. Even our most satisfactory model of
column 5 only predicts 6.11 percent of crises adequately. This is also reflected in
the pseudo-R 2 measure of McFadden (1974), which is given by

Gu
Rp2 ¼ 1
; ð15:4Þ
G0

with Gu and G0 respectively denoting the log-likelihood of the unrestricted model

and the model with only a constant. Rp2 hovers between .03 and .079, indicating a
poor capacity of the model to predict the onset of crises.
The goodness-of-fit results, however, should be interpreted with caution (see
Wooldridge 2001, 465–466 for a discussion). The timing of a crisis is likely to be a
very uncertain event, and it would indeed be surprising if we were able to cor-
rectly predict it a very high number of times. Indeed, the flip side of our inability
to adequately predict the onset of crises is that the models are very good at pre-
dicting the nonoccurrence of crises. Since the entry into a crisis is a rare event
(compared to staying out of the crisis), almost all prediction rules are naturally go-
ing to be very conservative. An alternative way to ask about the quality of our
model’s predictive capacity is by looking at the times that the model correctly pre-
dicts a crisis (that is, when the estimated probability exceeds 0.5) as a percentage
of the total number of times it predicts a crisis. In our baseline model of column 5,
this percentage is a much more reasonable 52.4 percent.
The economic significance of our coefficient estimates can best be interpreted by
studying the marginal effect of changes in the explanatory variables on the esti-
mated probability of a crisis. These effects are displayed in table 15.5, which also
shows the effect of a one-standard-deviation increase in each of the explanatory
variables. By this metric, the largest single effect comes from inflation: a one-
Growth Collapses 397

Table 15.5
Marginal Effects of Explanatory Variables, Baseline Estimation
dp/dx sd(x) (dp/dx)  sd(x)

Continuous variables
Log of per worker GDP 0.0035 1.0401 0.0036
Log change in manufacturing exports
0.0972 0.5699
0.0554
Lof(1 þ Inflation) 0.2360 0.3347 0.0790
Open forest
0.0376 0.9898
0.0372
Indicator variables
War 0.1338 0.1337 0.0179
Sudden stops 0.0549 0.4374 0.0240
Political transitions 0.0955 0.4416 0.0422
Latin America 0.0605 0.3864 0.0234
Africa 0.0113 0.4273 0.0048
Asia
0.0713 0.2587
0.0185
East Asia and Pacific
0.0354 0.3307
0.0117
Central and Eastern Europe
0.0392 0.3307
0.0130
Middle East and North Africa 0.0305 0.2948 0.0090
1980s 0.0062 0.4163 0.0026
1990s
0.0239 0.4163
0.0099
2000s
0.0896 0.3150
0.0282

standard-deviation increase in our inflation indicator increases the probability of a

crisis by 7.9 percent. The effect of manufacturing exports, however, is also sub-
stantial: a one-standard-deviation decline in the rate of growth of manufacturing
exports causes an increase of 5.54 percentage points in the probability of a crisis.
For indicator variables, a more natural metric is to think of the effect of the
variable changing from one to zero. By this metric, a war is by far the most de-
structive single event, causing an increase in the probability of a crisis of 13.38
percentage points. By contrast, a political transition costs 9.55 percentage points
and a sudden stop costs a 5.49 percentage point-increase in the probability of cri-
ses. These coefficients should be interpreted with the already mentioned caveat
about causality.
In section 15.4 we showed that there were substantial interregional differences
—particularly between industrial and developing regions—in the characteristics
of crises. In tables 15.6 and 15.7 we look at this issue more systematically by split-
ting the sample between developing economies and industrial economies. We
indeed find important differences in the results in the two subsamples. Export
declines, inflation, political transitions, and open_ forest retain their effect in the
subsample of developing countries. Curiously, wars and sudden stops lose some
Table 15.6
Random Effects Probit Regressions, Developing Countries
Dependent variable: Probability of
falling into a crisis (1) (2) (3) (4) (5)
Log GDP per working-age person
0.013 0.057 0.012 0.112 0.067
(0.21) (0.85) (0.15) (0.98) (0.61)
Latin America 0.329 0.871 0.841 0.426 0.496
(2.02)** (3.3)*** (2.54)** (1) (1.18)
Africa 0.246 0.841 0.735 0.259 0.307
(1.43) (3.04)*** (2.12)** (0.53) (0.64)
South and Central Asia C0.335 0.288 0.182
0.128
0.088
(1.68)* (0.98) (0.5) (0.27) (0.19)
East Asia and Pacific
0.107 0.500 0.446 0.044 0.088
(0.62) (1.84)* (1.28) (0.1) (0.2)
Middle East and North Africa 0.261 0.847 0.751 0.323 0.383
(1.46) (3.05)*** (2.18)** (0.72) (0.87)
1970s 0.236 0.422 0.239 0.153 0.156
(1.67)* (2.78)*** (1.45) (0.39) (0.4)
1980s 0.330 0.285 0.208 0.216 0.212
(2.56)** (2.1)** (1.43) (0.57) (0.56)
1990s 0.292 0.210 0.091 0.153 0.134
(2.24)** (1.51) (0.61) (0.41) (0.36)
Log change in real merchandise exports C0.498 C0.509 C0.565 C0.531
(5.17)*** (4.87)*** (3.42)*** (3.27)***
War 0.645 0.326 0.403
(2.44)** (1.03) (1.33)
Natural disaster
0.072 0.144
(0.52) (0.79)
Sudden stop 0.105 0.205 0.185
(1.06) (1.57) (1.44)
Log of inflation 0.817 0.826
(2.49)** (2.54)**
Change in polity indicator 0.251 0.309
(1.84)* (2.34)**
Open forest C0.163
0.178
(1.69)* (1.88)*
Democracy
0.005
(0.5)
Constant C1.129 C2.160 C1.637 C0.207 0.351
(1.85)* (3.11)*** (2.02)** (0.12) (0.2)
Constant
3.953
3.801
3.527
3.273
3.341
(3.21)*** (3.37)*** (3.55)*** (2.9)*** (2.9)***
Observations 1379 1277 1129 686 694
Countries 133 122 111 64 64
Percent crises predicted 0.0% 0.8% 1.9% 8.4% 6.6%
Pseudo-R 2 2.4% 5.8% 6.1% 7.4% 7.7%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%,
***
1%.
Growth Collapses 399

Table 15.7
Random Effects Probit Regressions, Industrialized Countries
Dependent variable: Probability of
falling into a crisis (1) (2) (3) (4) (5)
Log GDP per working-age person
0.227
0.104 0.164 0.483 0.476
(1.15) (0.47) (0.58) (1.02) (1)
Log change in real merchandise exports 0.178
0.001
0.385
0.349
(0.69) (0) (0.86) (0.78)
War 0.730 0.696 0.621
(1.71)* (1.24) (1.15)
Natural disaster
5.672
5.310
(0) (0)
Sudden stop 0.373 0.407 0.403
(2.49)** (2.09)** (2.09)**
1970s 0.409 4.988 4.969
(1.28) (0) (0)
1980s 0.363 5.023 5.014
(1.53) (0) (0)
1990s 0.021 5.225 5.174
(0.09) (0) (0)
Log of inflation 11.970 11.345
(3.98)*** (3.98)***
Political transitions 0.356 0.195
(0.6) (0.34)
Open forest
0.184
0.144
(0.55) (0.44)
Democracy (polity index) 0.323
(0.78)
Constant 1.146
0.156
3.247
12.656
9.905
(0.56) (0.07) (1.08) (0.01) (0)
Observations 622 600 559 368 368
Countries 25 25 25 19 19
Percent crises predicted 0.0% 0.0% 2.5% 5.3% 6.6%
Pseudo-R 2 1.4% 3.4% 5.7% 7.3% 7.7%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%,
***
1%.
400 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

significance in this exercise—but in the case of wars the effect again seems to
come from the restriction of the sample. By contrast, exports, political transitions,
and open_ forests appear not to be relevant in industrial countries. Inflation retains
strong significance, with a much higher absolute coefficient estimate that reflects
the much smaller ranges of variation of this variable in developed economies.
Wars—which in the case of developed economies are almost always overseas
interstate conflicts such as the Gulf War—are also insignificant in this subsample.
While these differences are certainly interesting, the evidence that the data-
generating process is fundamentally different across regions is not all that strong.
The developed country sample is smaller, so it is logical to expect broader confi-
dence intervals. All of the variables that we found to be significant in the broader
sample have the same sign in both subsamples, and in most cases—with the nota-
ble exception of the inflation rate—the coefficient estimates are strikingly similar.
The similarity of these coefficient estimates suggests that the key reason for the
difference in the frequency of crises across regions comes not so much from differ-
ences in the way in which these crises are generated, but rather from differences in
the distribution of the underlying determinants.
Another potential source of structural differences may come from the fact that
very lengthy or costly crises may be associated with different factors from those
that generate shorter crises. Table 15.8 examines this hypothesis by splitting the
sample between short and long crises. We split these in two dimensions: crisis du-
ration of five years (columns 1 and 2) and crisis duration of more than 75 percent
of the last precrisis year’s GDP (columns 3 and 4). This exercise provides some
very interesting results. Regardless of whether one uses the duration or the lost
output splits, one is struck by the similarity of the coefficient estimates for exports,
wars, sudden stops, inflation, and political transitions. There are, however, strik-
ing differences between the effects of open forests and the Latin America dummy
across subsamples. These suggest that open forests and some unobserved charac-
teristics of Latin America may not be so much a predictor of crisis onset (for
which its significance in the whole sample is at best weak), but rather of crisis du-
ration. The bulk of the estimates, however, suggest that one can get thrown into
short and long crises for very similar reasons. The substantial difference, therefore,
may be in how one recovers from these crises.
The next three tables include a series of additional robustness tests for our base-
line specification. Table 15.9 studies the effect of adopting alternative definitions
of capital flows. One problem with the capital flows window measure (which is
discussed at length in Calvo, Izquierdo, and Mejia 2004) is that the decline in cap-
ital flows may be caused by an increase in export capacity. In column 1 we use a
measure that combines the decline in capital stock with the condition that imports
must also have declined. The coefficient on this measure, while positive, is not
Growth Collapses 401

Table 15.8
Random Effects Probit Regressions by Intensity of Crisis
(1) (2) (3) (4)
Dependent variable: Probability Duration < 5 Duration > 5 Integral < 0:75 Integral > 0:75
of falling into a crisis years years GDP-years output years
Log GDP per working-age
0.088 0.193
0.105 0.277
person (0.79) (1.33) (0.98) (1.67)*
Log change in real C0.275 C0.569 C0.358 C0.568
merchandise exports (1.73)* (2.59)*** (2.33)** (2.35)**
War 0.428 0.419 0.345 0.652
(1.49) (1.07) (1.2) (1.56)
Sudden stop 0.203 0.150 0.235 0.061
(1.82)* (0.91) (2.19)** (0.31)
Log of inflation 0.868 0.899 0.965 0.838
(2.57)** (2.08)** (2.97)*** (1.79)*
Political transitions 0.308 0.468 0.290 0.494
(2.16)** (2.54)** (2.16)** (2.31)**
Open forest
0.020 C0.411 0.023 C0.512
(0.19) (3.5)*** (0.23) (3.96)***
Latin America
0.121 0.947 0.050 1.192
(0.52) (2.85)*** (0.23) (2.61)***
Africa
0.108 0.473
0.003 0.705
(0.31) (1.02) (0.01) (1.22)
South and Central Asia C0.548 0.182
0.473 0.163
(1.75)* (0.39) (1.61) (0.25)
East Asia and Pacific
0.373 0.489
0.280 0.833
(1.49) (1.25) (1.2) (1.59)
Central and Eastern Europe
0.370 0.471
0.477 1.125
(0.82) (0.75) (1.05) (1.62)
Middle East and North Africa 0.076 0.362 0.107 0.496
(0.32) (0.93) (0.48) (0.94)
1970s 0.366 0.409 4.937
(1.09) (1.21) (2.1)**
1980s 0.302 0.179 0.364 5.140
(0.94) (0.8) (1.13) (2.16)**
1990s 0.198 0.027 0.350 4.569
(0.62) (0.11) (1.09) (1.9)*
2000s
5.374
(0)
Constant
0.393 1.459
0.872
3.216
(0.23) (0.71) (0.53) (0)
Observations 1004 933 1023 921
Countries 81 83 81 83
Percent crises predicted 0.8% 3.9% 2.1% 2.6%
Pseudo-R 2 4.8% 20.6% 5.1% 27.2%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%,
***
1%.
Table 15.9
Random Effects Probit Regressions, Alternative Sudden Stop Definitions
Dependent variable: Probability of
falling into a crisis (1) (2) (3) (4)
Log GDP per working-age person 0.027 0.036 0.055 0.049
(0.28) (0.38) (0.54) (0.48)
Log change in real merchandise exports C0.409 C0.400 C0.360 C0.369
(2.91)*** (2.82)*** (2.3)** (2.37)**
War 0.457 0.459 0.540 0.564
(1.76)* (1.77)* (1.97)** (2.06)**
Sudden stop 1 0.118
(1.02)
Sudden stop 2 0.060
(0.62)
Sudden stop 3 0.164
(1.59)
Sudden stop 4 0.221
(1.9)*
Log of inflation 1.014 1.010 0.985 0.954
(3.3)*** (3.29)*** (3.13)*** (3.03)***
Political transitions 0.363 0.357 0.369 0.368
(2.91)*** (2.86)*** (2.82)*** (2.81)***
Open forest C0.160 C0.165 C0.201 C0.197
(1.89)* (1.95)* (2.27)** (2.22)**
Latin America 0.226 0.225 0.204 0.193
(1.14) (1.14) (0.99) (0.93)
Africa 0.073 0.087 0.002
0.022
(0.25) (0.29) (0.01) (0.07)
South and Central Asia
0.327
0.315
0.278
0.298
(1.2) (1.15) (0.99) (1.06)
East Asia and Pacific
0.127
0.107
0.117
0.144
(0.57) (0.48) (0.52) (0.63)
Central and Eastern Europe
0.218
0.234
0.134
0.129
(0.54) (0.58) (0.38) (0.37)
Middle East and North Africa 0.116 0.110 0.129 0.094
(0.54) (0.51) (0.59) (0.43)
1970s 0.477 0.467 0.608 0.625
(1.44) (1.41) (1.67)* (1.72)*
1980s 0.491 0.478 0.627 0.639
(1.54) (1.5) (1.8)* (1.83)*
1990s 0.366 0.353 0.461 0.495
(1.15) (1.11) (1.32) (1.42)
Constant 0.395 0.374 0.505 0.548
(0.27) (0.25) (0.33) (0.36)
Observations 1061 1061 1004 1002
Countries 83 83 83 83
Percent crises predicted 0.0611 0.0556 0.0599 0.0539
Pseudo-R 2 0.0760 0.0753 0.0836 0.0845
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%,
***
1%.
Growth Collapses 403

significantly so. This may be because the combination of these two criteria is very
stringent. In column 2 we relax it by defining a sudden stop to be any decline in
capital flows that coincides with an import decline. This measure appears to be
very poorly correlated with the onset of crises (p ¼ .538). Column 3 uses a defini-
tion based on total (as opposed to just private) capital flows, which we measure as
the sum of the trade balance and the decline in reserves, combined with a decline
in imports. This measure does somewhat better, nearing statistical significance
(p ¼ .113). If we make this last definition somewhat more stringent by requiring
declines in total capital flows to exceed 3 percent of GDP and import declines to
exceed 5 percent of initial import values, significance increases slightly (p ¼ .058).
We have carried out a substantial number of additional tests with many alterna-
tive definitions, and find that, while it is certainly possible to come up with defini-
tions of sudden stops that are significantly associated with the onset of crisis, such
a conclusion is not robust to changes in the way in which we define the event. An
additional conclusion that can be drawn out of table 15.8 is that the incidence of
our significant explanatory variables does not change with different choices of
sudden stop indicators. Changes in exports, political transitions, high inflation,
wars, and open_ forest maintain their patterns of association with the onset of cri-
ses in all alternative specifications.
To this moment we have assumed that the country-specific effect hi in equation
15.1 is uncorrelated with the explanatory variables, giving rise to the random
effects probit specification. This assumption is of course questionable, but relaxing
it is problematic because of the well-known incidental parameters problem. An al-
ternative is to use the fixed effects logit specification, where the coefficient vector b
p
can be estimated with n-consistency. This specification, however, is not without
its cost. As discussed in detail by Wooldridge (2001, 492), it requires the condi-
tional independence of the dependent variable given the explanatory variables. In
our context, this implies assuming that the probability of a crisis is independent of
the number of crises that have occurred in the past.
Table 15.10 shows the results of this specification. There are important changes
as well as similarities with the random effects probit specification. The key simi-
larity lies in the coefficients for merchandise exports, the capital-flow window def-
inition of sudden stops (which retain strong statistical significance), and wars
(which are significant in the broader sample and borderline significant once one
adds additional controls). The effects of inflation and political transitions are pre-
served, although with lower p-values than under the probit specification. The
striking difference, however, lies in the changes in the log of per capita GDP,
which is now strongly negative and significant—indicating that richer countries
have less propensity to experience crises—and open_ forest, which is now posi-
tively, though insignificantly, related to the onset of crises.
404 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Table 15.10
Fixed Effects Logit Specification
Dependent variable: Probability of
falling into a crisis (1) (2) (3) (4) (5)
Log GDP per working-age person C2.137 C1.704 C1.611 C3.593 C3.607
(3.77)*** (2.89)*** (2.43)** (3.51)*** (3.53)***
Log change in real merchandise exports C0.925 C0.931 C1.245 C1.242
(4.64)*** (4.41)*** (4.04)*** (4.02)***
War 1.037 0.840 0.844
(2.54)** (1.72)* (1.73)*
Natural disaster 0.031 0.210
(0.11) (0.6)
Sudden stop 0.352 0.528 0.528
(2.2)** (2.54)** (2.54)**
Log of inflation 1.334 1.333
(1.95)* (1.95)*
Change in polity indicator 0.460 0.454
(1.8)* (1.78)*
Open forest 0.347 0.352
(0.81) (0.82)
1970s C0.660
(1.83)*
1980s
0.119 0.125 0.394 0.599 0.606
(0.42) (0.55) (1.6) (2.01)** (2.04)**
1990s 0.176 0.335 0.454 1.089 1.103
(0.73) (1.14) (1.44) (2.61)*** (2.65)***
2000s 0.232 0.457 0.651 0.660
(0.62) (1.14) (0.82) (0.83)
Observations 1906 1761 1571 986 986
Countries 145 126 116 75 75
Percent crises predicted 0.00% 0.00% 0.00% 0.00% 0.00%
Pseudo-R 2 1.74% 3.32% 4.55% 8.47% 8.41%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%,
***
1%.

Our last battery of robustness tests is displayed in table 15.11, where we study
the effect of adding additional potential explanatory variables to the probit speci-
fication. In this table, we include average years of primary, secondary, and total
schooling as measures of human capital’s effect on propensity to fall in crises. Nei-
ther of these measures is significant (columns 1–3). Neither is a measure of institu-
tions (the rule of law), of physical infrastructure (telephone mainlines per capita),
of urbanization, or of life expectancy.
We can summarize the results of this section as follows: a number of variables
appear to be associated with the onset of crises. In terms of robustness, the vari-
Growth Collapses 405

able that comes out on top is the log change in merchandise exports, which has
come out as significant in all the specifications in which it is included, with the ex-
ception of the subsample of developed countries. In terms of economic signifi-
cance, a one-standard-deviation increase in inflation appears to be associated
with much more damage than a similar increment in any other variable. Most
specifications coincide in a significant effect of the capital-flows window defini-
tion of sudden stops, as well as political transitions, on the probability of a crisis
occurring. The effects of wars, initial income, and residual continent or time dum-
mies are much more variable to specification. Particularly, while open_ forest
comes out as a significant predictor in some specifications, its coefficient tends to
be weak and its sign is reversed in the conditional logit specification.

15.5 How Do Countries Get Out of Crises?

In this section we analyze the determinants of crisis duration. Most existing con-
tributions in the literature do not deal with the problem of censoring that natu-
rally arises in the analysis of contractionary episode duration. As discussed in the
introduction, the standard solution taken in the literature addressing this issue is
to drop or truncate those observations. Either solution is inappropriate. Dropping
the observations biases the sample toward short duration episodes, while trun-
cating them inadequately represents crises as having shorter durations than in
reality.
The results presented in this section deal with the problem of censored observa-
tions by adopting a duration analysis approach. Specifically, if we have n coun-
tries with t1 . . . tn crises duration, we concentrate on finding the estimate of the
probability density function f ðtÞ with associated survival time SðtÞ that maxi-
mizes the likelihood function
Y
L¼ f ðti Þ di Sðti Þ 1
di ð15:5Þ
i

where di is an indicator variable that takes the value 0 if the peak per worker GDP
has not been reached by the last observation in the sample. Broadly speaking,
there are two approaches in the literature to estimating equation 15.5. One is to
specify a parametric functional form for f ðtÞ and to estimate the parameters of
that form. The second one is to use a nonparametric approach to estimation of
f ðtÞ. The latter is commonly associated with estimation of the Cox proportional
hazards model. Although the nonparametric approach is more flexible, it can
lead to more imprecise estimates of the hazard function than a correctly specified
parametric form. We will present versions of both models in this section.
Table 15.11
406

Random Effects Probit Regressions: Alternative Controls

Dependent variable: Probability of

falling into a crisis (1) (2) (3) (4) (5) (6) (7) (8)

Log GDP per working-age person 0.199 0.131 0.206 0.182 0.104
0.061 0.077 0.174
(1.41) (0.98) (1.41) (1.23) (0.85) (0.44) (0.34) (0.36)
Log change in real merchandise exports C0.539 C0.533 C0.548 C0.459 C0.447 C0.407 C0.487 C0.872
(3.2)*** (3.16)*** (3.24)*** (2.15)** (2.96)*** (2.88)*** (2.04)** (2.35)**
War 0.767 0.765 0.768 0.658 0.475 0.476 0.910 0.968
(2.78)*** (2.77)*** (2.78)*** (2.03)** (1.82)* (1.84)* (2.15)** (2.09)**
Sudden stop 0.244 0.254 0.246 0.248 0.202 0.230 0.443 0.654
(2.27)** (2.37)** (2.29)** (1.81)* (1.95)* (2.27)** (2.93)*** (3.28)***
Log of inflation 1.152 1.062 1.096 0.720 0.930 0.971 1.128 0.975
(3.59)*** (3.3)*** (3.44)*** (1.95)* (2.99)*** (3.12)*** (2.28)** (1.66)*
Change in polity indicator 0.312 0.314 0.312 0.531 0.334 0.367 0.038
0.083
(2.23)** (2.24)** (2.23)** (3.13)*** (2.59)*** (2.94)*** (0.16) (0.22)
Open forest
0.137
0.138
0.134
0.066
0.123 C0.181 C0.374
0.283
(1.27) (1.28) (1.24) (0.47) (1.41) (2.04)** (2.25)** (1)
Years of primary schooling
0.083 0.023
(1.49) (0.04)
Years of secondary schooling
0.059
0.110
(0.73) (0.21)
Total years of schooling
0.052
0.012
(1.4) (0.03)
Rule of law 0.013 0.288
(0.16) (1.72)*
Telephone mainlines (per 1,000 people)
0.001
0.002
(1.27) (1.2)
Urban population (% of total) 0.004 0.016
(0.77) (1.63)
Life expectancy at birth, total (years)
0.024
0.044
(1) (0.94)
Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner
Latin America 0.312 0.259 0.264 0.738 0.129 0.178 0.040 0.334
(1.48) (1.18) (1.24) (2.47)** (0.56) (0.83) (0.12) (0.54)
Africa
0.027
0.047
0.064 0.497
0.043 0.006 C0.940
1.114
(0.08) (0.15) (0.2) (1.05) (0.14) (0.02) (1.73)* (0.97)
South and Central Asia
0.174
0.214
0.202
0.055
0.428
0.375
0.741
0.180
(0.58) (0.71) (0.67) (0.14) (1.47) (1.35) (1.55) (0.27)
Growth Collapses

East Asia and Pacific 0.024
0.067
0.027 0.013
0.237
0.208
0.278
0.218
(0.1) (0.28) (0.12) (0.05) (1) (0.9) (0.77) (0.44)
Central and Eastern Europe 0.203
0.080 0.064 0.054
0.347
0.173
0.022
0.288
(0.43) (0.18) (0.14) (0.12) (0.81) (0.43) (0.05) (0.38)
Middle East and North Africa
0.023
0.054
0.049 0.531 0.043 0.063 0.141 0.162
(0.09) (0.22) (0.2) (1.83)* (0.18) (0.27) (0.39) (0.26)
1970s 0.497 0.548
(1.48) (0.94)
1980s
0.022
0.026
0.019 0.148 0.054 0.524 0.643 0.043
(0.14) (0.17) (0.13) (1.07) (0.38) (1.62) (1.17) (0.18)
1990s
0.173
0.173
0.156
0.021 0.388 0.476
(1.12) (1.11) (1) (0.13) (1.21) (0.88)
2000s
0.377
(1.1)
Constant
0.805
0.324
0.880
2.351
0.193 1.258 4.587 2.652
(0.49) (0.2) (0.53) (1.04) (0.12) (0.69) (1.73)* (0.56)
Observations 951 951 951 672 1037 1062 522 369
Countries 74 74 74 61 80 83 77 50
Percent crises predicted 6.54% 7.19% 5.88% 6.54% 5.81% 6.11% 7.41% 6.00%
Pseudo-R 2 8.45% 8.25% 8.42% 10.64% 7.73% 8.01% 10.83% 13.12%

Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%, ***
1%.
407
408 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Another key issue has to do with how to handle country level heterogeneity in
this framework. For analogous reasons to those of panel data estimation with bi-
nary dependent variables, fixed effects estimators are not consistent for duration
models (Andersen, Klein, and Zhang 1999). Two alternative approaches exist.
One is to assume that countries have differing propensities to experiencing crises.
These propensities—called frailties—are analogous to the random effects of panel
data models. An alternative approach is to use the fact that in the presence of re-
peated events, the Cox proportional hazards model parameter estimates converge
to a value that can be interpreted meaningfully, but the estimated covariance ma-
trix is inappropriate for hypothesis testing (Lin and Wei 1989 and Struthers and
Kalbfleisch 1984). Variance-corrected models modify the covariance matrix of the
Cox model in order to be able to carry out appropriate tests.
Before proceeding to the statistical tests, we start out by looking at the charac-
teristics of crises according to the events associated with them. The summary sta-
tistics associated with these different types of crises, as well as their associated
unconditional hazard functions, are shown respectively in table 15.12 and figure
15.5. If different types of crises correspond to different types of shocks, then we
would expect that the patterns of recoveries associated with different crises would
also differ. In particular, this would be true if we believe that some crisis triggers
have nonpermanent effects on the determinants of steady-state income. As we
have discussed previously, in this case hazard functions should be clearly increas-
ing. This is precisely the feature that Calvo and his coauthors have argued charac-
terizes some sudden stops of capital flows. Similarly, to the extent that the level of
democracy appears to be irrelevant for crisis onset, political regime transitions
should have a transitory effect on the level of income: they should create havoc

Table 15.12
Characteristics of Crises by Coinciding Initial Events
OLS trend
Number Peak to Lost growth
of obser- trough years (end to
vations Duration ratio of GDP endþ5)
Substantial change in 95 6.51 11.6% 0.98 1.1%
merchandise exports
Wars 32 6.53 16.5% 1.50 2.1%
Natural disasters 51 7.51 13.0% 1.30 1.0%
High inflation 62 6.60 10.2% 0.97 1.7%
Political transition 118 6.69 11.0% 1.04 1.7%
Sudden stop 109 5.83 8.8% 0.86 1.6%
High open forest (>14) 75 3.08 4.6% 0.23 2.3%
Growth Collapses 409

Figure 15.5
Hazard Function by Type of Crisis: Alternative Specifications

during the time of the transition, but after they occur one would not expect there
to be a permanent effect.
What is interesting about figure 15.5 is that it shows that declining hazard rates
appear to characterize many different types of crises. Indeed, all splits appear to
be characterized by the same overall pattern: a short initial period of increasing
hazard rates, and a much longer period of strongly declining rates. The relative
magnitudes are also similar across characteristics. The one striking difference is
open_ forest. Countries with very high open forests have crises of much lower du-
ration and consequently display a higher probability of exiting the crisis at any
one moment. Recall that of the variables that we have used to carry out the splits,
open_ forest (along with natural disasters) was not robustly associated with the
onset of crisis. This figure suggests that it may be directly associated with crisis
duration.
This hypothesis can be tested more systematically by looking at the effect of
these factors in duration regressions. We do this in the rest of the section. We start
out by looking at the most common parametric duration model, which is the Wei-
bull distribution. Essentially, we estimate the hazard rate for country i as a func-
tion of a baseline hazard and the covariates
410 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

hi ðt j XÞ ¼ h0t vi expðbXÞ; ð15:6Þ

where h0t ¼ pt p
1 . The parameter p characterizes the shape of the Weibull distri-
bution, with p < 1 corresponding to a decreasing hazard function. The ni are the
country-level frailty terms that are assumed to follow a Gamma distribution. Esti-
mates are displayed in table 15.13.
Column 1 shows the effect of simply running a regression of duration on log of
per worker GDP. The hazard specification is presented, so that a positive coeffi-
cient implies that increasing the variable in question leads to a higher probability
of leaving the crisis. Not surprisingly, per worker income appears to be positively
related to the probability of exiting a crisis. However, as column 2 shows, this re-
sult is not robust to controlling for region and time dummies. In column 3 we in-
troduce open_ forest into the regression. We find that it is very strongly correlated
with crisis duration. A one-standard-deviation increase in open_ forest implies an
increase of 84.5 percent in the probability of leaving the crisis (expð.543Þ 
sdðopen_ forestÞ). Column 4 introduces controls for the level of democracy and
sudden stops. The democracy control can be taken as a naive test of Rodrik’s
(1999) hypothesis that countries with better institutions for conflict management
have an easier time adjusting to negative shocks. Framed this way, the hypothesis
gets weak support in our data: the effect of democracy is positively related with
the probability of leaving the crisis, though the coefficient is borderline significant
and not very robust. However, Rodrik’s hypothesis is somewhat more nuanced—
see the additional results in table 15.13. Calvo and his coauthors have suggested
that the recoveries associated with some types of sudden stops may be associated
with faster recoveries. The estimates in column 4 show that this is not the case for
collapses in capital flows, generally speaking.
The next columns of table 15.13 introduce other potential determinants of crisis
duration. The first logical candidates for this are the different determinants that
we used in the probit analysis of the previous sections. If how you fall into the cri-
sis matters for postcrisis behavior, then we should expect crisis durations to differ
significantly for crises that were initiated by different events. We have already
seen in figure 15.3 that there is little indication of this being the case in uncondi-
tional hazard rates by group, but we now verify this within the framework of the
parametric Weibull specification. In effect, the estimates in column 4 show that
different factors that were significantly associated with the onset of crises are not
associated with crisis duration. This is the case of wars, inflation, and political
transitions—as well as of natural disasters, which did not prove significant in the
probit analyses.
One may expect that both open forests and democracy may make the society
more capable of responding to particular types of shocks rather than uniformly
Growth Collapses 411

Table 15.13
Duration Regressions, Weibull Specification with Frailty
Dependent variable: Years
in crisis (1) (2) (3) (4) (5) (6) (7)
Log GDP per working-age 0.254
0.238
0.263 C0.492 C0.744
0.258 C0.497
person (2.61)*** (1.45) (1.31) (1.93)* (2.61)*** (1.29) (1.98)**
Open forest 0.543 0.578 0.731 0.457 0.530
(3.61)*** (3.03)*** (3.37)*** (2.54)** (2.46)**
Democracy (polity index) 0.039 0.035 0.043
(1.66)* (1.38) (1.64)
Sudden stop 0.004
0.139
(0.02) (0.62)
Log change in real merchan- 0.429 C1.288 C1.964
dise exports (1.34) (0.5) (0.55)
War
0.621
(1.15)
Natural disaster
0.099
(0.25)
Log of inflation
0.189
(0.29)
Change in polity indicator C0.290
(1.01)
Change in exports*open 0.112 0.157
forest
(0.54) (0.56)
Polity*change in merchan-
0.026
dise exports (0.53)
Polity*sudden stops 0.000
(0.01)
Constant C4.102 2.061 C5.764 C4.245 C3.171 C4.656 C2.653
(4.51)*** (1.22) (1.91)* (1.18) (0.78) (1.44) (0.7)
N 330 330 229 188 175 227 188
Time Dummies No Yes Yes Yes Yes Yes Yes
Continent Dummies No Yes Yes Yes Yes Yes Yes
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%,
***
1%. Representation, hazard rate with region and decade dummies (not shown).
412 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

increasing the probability of exiting all types of crises. Open_ forest, for example,
should have a positive effect on an economy’s capacity to adapt to export col-
lapses, while democracy may make societies more capable of adapting to external
shocks such as sudden stops. The last two columns of table 15.13 evaluate these
hypotheses, and find little support. Neither an interaction between open forests
and changes in exports nor the interaction term between democracy and the two
external shock indicators (changes in exports and sudden stops) are significant.
This may of course reflect the relative coarseness of these multiplicative terms to
capture complex nonlinearities.
Table 15.14 tests a battery of additional possible correlates of crisis duration.
These include an alternative indicator of nonsystemic sudden stops in column 1
(the same one in column 4 of table 15.8, which captures large decreases in total
capital flows that coincide with substantial import declines), an indicator of trade
policy (column 2), a measure of human capital (total years of schooling, column
3), a measure of financial deepening (liquid liabilities in GDP, column 4), a general
measure of openness (trade/GDP ratio, column 5), and a measure of social mod-
ernization (life expectancy, column 6). We also try two measures of the idea that
institutions may have an effect on crisis duration. The first one introduces an in-
teraction between our indicator of democracy and the terms of trade shock, thus
capturing the idea that democracies are better able to adapt to adverse terms of
trade shocks. This effect is insignificant. Another specification uses the interaction
between the Gini index and 1, minus a scaled democracy variable. This is closest
to Rodrik’s (1999) precise specification. It gets strong support in the data, with a
significant negative coefficient, suggesting that when there are high levels of social
conflict, better institutions for conflict management are associated with shorter cri-
ses. In one last specification we include the regressors of columns 1–7 together
(column 9). Most of these terms are significant, while the open_ forest indicator re-
mains strongly significant.
In the next table we adopt as our baseline specification a regression including
time and region dummies, open_ forest, and the log of initial GDP, and we evaluate
whether the strength of the coefficient on open forests is at all dependent on the
parametric specification of the hazard function. We use four alternative parame-
terizations: the exponential, the Gompertz, the log-logistic, and the log-normal. In
reading table 15.15, it is important to bear in mind that the last two specifications
do not accept a hazard rate interpretation and are thus reported in accelerated-
failure time modes, so that the dependent variable is the duration of the crisis.
Thus, a positive effect in the hazard representation is analogous to a negative ef-
fect in the accelerated-failure time representation. That is, in fact, what we find:
open_ forest has a positive, significant effect in the hazard rate representations and
a negative, significant effect in the failure time representations. The result that
Growth Collapses 413

open_ forest decreases the time necessary to escape a crisis is robust to the parame-
terization adopted.
Figure 15.6 displays the conditional hazard rates that emerge from the five
parametric specifications that we have estimated (with the controls of table
15.13). These are the estimated hazard rates for an observation with the expected
random effect ni ¼ 1. In contrast to the unconditional hazard rates of figures 15.2–
15.5, they are not affected by the changing composition of the population and re-
flect the estimated probability that a particular country will exit the crisis. For
comparison purposes, we also report the unconditional hazard rate of a Cox
model without shared frailties. Except for the exponential form, which is con-
strained to be constant, all our estimates of the hazard rates give declining or
roughly flat functions in time.
In the next two tables we turn toward estimation in the framework of a
variance-corrected Cox proportional hazards model. In particular, we specialize
to the conditional risk-set model of Prentice, Williams, and Peterson (1981, hence-
forth PWP). The basic idea of the PWP model is to stratify by event number, so
that the conditional risk set for experiencing crisis k is the number of countries
that have experienced k
1 crises in the past. The model is stratified by number
of crises in order to obtain the corrected variance estimates. Tables 15.15 and
15.16 repeat the estimation exercises of tables 15.12 and 15.13 using the PWP spec-
ification. The results are broadly similar. Open_ forest is always significant, with
the only exception being the last column of table 15.17, which has a very small
number of observations (here it has a borderline p value of .142 with 60 observa-
tions). None of the other potential covariates emerge as significant.
As we have argued above, declining hazard rates may indicate the presence of
multiple equilibria—with negative shocks leading countries to shift to inferior
equilibria—or adverse permanent productivity shocks. A valid question to ask at
this stage is whether there is evidence that this phenomenon is due to changes in
fundamentals. One way to tackle this question is by asking whether the triggers
that appear to have sent the economy into the crisis have returned to precrisis lev-
els at the time periods during which we are observing declining hazard rates. Fig-
ure 15.7 presents one such exercise. In it we calculate the average paths for
countries with crisis duration greater than or equal to ten years for five of the vari-
ables we have found to be significantly associated with the onset of crises: wars,
exports, capital flows, inflation, and political transitions. The evidence is mixed,
and thus interesting. By the tenth year of the crisis, the fraction of countries in the
midst of a war has returned to precrisis levels. The number of countries undergo-
ing political transitions has also declined, though not to precrisis levels. Capital
flows have actually gone up in comparison to their precrisis levels. This is interest-
ing not only because it suggests that the decline in capital flows is not the cause
Table 15.14
414

Weibull Specification, Alternative Controls

Dependent variable: Years in crisis

Representation—Hazard (1) (2) (3) (4) (5) (6) (7) (8) (9)

Log GDP per working-age person
0.172
0.291
0.351
0.168
0.335 C1.518 C0.366
0.568 C5.013
(0.74) (1.18) (1.28) (0.85) (1.64) (2.49)** (1.75)* (2.25)** (3.43)***
Open forest 0.535 0.590 0.582 0.576 0.638 0.475 0.578 0.728 2.058
(3.06)*** (3.13)*** (3)*** (3.47)*** (4.07)*** (1.7)* (3.39)*** (3.48)*** (2.64)***
Sudden stop 4
0.051
(0.24)
Latin America C1.714 C1.661 C1.461 C1.458 C1.670 C2.083 C1.635
1.307 C2.609
(3.68)*** (3.68)*** (3.45)*** (3.62)*** (3.93)*** (2.46)** (3.77)*** (2.80)*** (2.22)**
Africa
1.134
1.129
0.763 C1.141 C1.181
0.196 C1.370
0.546
1.389
(1.51) (1.53) (1.19) (1.87)* (1.8)* (0.13) (2.16)**
0.76 (0.7)
South and Central Asia
0.430
0.747
0.348
0.364
0.909 C2.101
0.828
0.879 C4.622
(0.58) (1.08) (0.52) (0.59) (1.4) (1.83)* (1.32)
1.32 (2.55)**
East Asia and Pacific
0.443
0.949
0.324
0.467
0.746
0.737
0.688
0.431 C3.911
(0.76) (1.59) (0.68) (0.93) (1.35) (0.92) (1.32)
0.7 (2.24)**
Central and Eastern Europe C1.809 C2.177
1.105
1.093 C1.878 C2.481
1.009
1.442 C3.947
(2.15)** (2.36)** (1.42) (1.33) (2.6)*** (2.75)*** (1.33) (1.99)** (2.19)**
Middle East and North Africa
0.428
0.145
0.218
0.307
0.528
1.350
0.386 0.323 0.107
(0.85) (0.29) (0.43) (0.66) (1.06) (1.63) (0.82)
0.57 (0.09)
1970s C1.460 0.465 0.476
1.117
0.762 1.083
(1.97)** (1.69)* (1.6) (1.22) (0.69)
1980s C1.960 C0.399 C0.478 C1.865 C1.593
0.438 C0.808
0.619 C1.402
(2.81)*** (1.66)* (1.9)* (2.33)** (2.74)*** (0.29) (2.95)*** (2.34)** (2.42)**
1990s C1.587 C1.512 C1.204
0.132
0.373
0.472
0.664
(2.23)** (1.84)* (2.02)** (0.09) (1.34) (1.67)* (0.88)
2000s 1.199 0.421
(1.07)
0.51
Tariff rate 2.426 C95.204
(0.31) (2)**
Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner
Total years of schooling 0.100 0.418
(1.26) (2.03)**
Liquid liabilities/GDP 0.002 0.012
(0.35) (0.75)
Openness 0.003 0.025
(1.08) (1.46)
Growth Collapses

Life expectancy at birth, total (years)- 0.126 0.090

(1.66)* (0.69)
Interaction between polity and terms of 0.169
1.653
(0.52) (1.38)
Gini*(1-Democracy)
0.025
(2.54)**
Constant
4.948 C6.156 C6.223 C5.707 C5.332
1.657 C4.799
4.963 9.409
(1.42) (1.85)* (1.83)* (1.85)* (1.74)* (0.31) (1.59)
1.53 (0.99)
N 189 177 179 191 223 97 183 178 60
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%, ***
1%. Regression (9) comes from log file of
table 15.17.
415
416 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Table 15.15
Duration Regressions, Alternative Specifications
(1) (2) (3) (4)
Dependent variable: Years in crisis Exponential Gompertz Cox Log-normal
Distribution representation hazard hazard AFT AFT
Log GDP per working-age person
0.236
0.105 0.300 0.323
(1.27) (0.77) (1.62) (1.58)
Open forest 0.517 0.396 C0.368 C0.346
(3.68)*** (3.66)*** (2.97)*** (2.55)**
Latin America C1.504 C1.051 0.777 0.769
(3.83)*** (3.48)*** (2.2)** (2.07)**
Africa C1.143
0.668 0.720 0.774
(1.86)* (1.55) (1.24) (1.25)
South and Central Asia
0.817
0.509 0.518 0.506
(1.32) (1.24) (0.88) (0.82)
East Asia and Pacific
0.203
0.174
0.133
0.122
(0.45) (0.53) (0.35) (0.33)
Central and Eastern Europe C1.711 C1.297 1.108 1.240
(2.49)** (2.32)** (1.96)** (1.98)**
Middle East and North Africa
0.437
0.173 0.064
0.047
(0.96) (0.53) (0.18) (0.13)
1970s 0.462 0.239 C0.538 C0.565
(1.83)* (1.07) (2.5)** (2.57)**
1980s
0.340
0.222 0.256 0.272
(1.6) (1.16) (1.39) (1.41)
2000s 1.218 1.229
0.604
0.610
(2.06)** (2.21)** (1.18) (1.3)
Constant C5.623 C5.194 2.922 2.290
(1.98)** (2.55)** (1.16) (0.87)
N 229 229 229 229

Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%,
***
1%.

for declining hazard rates, but also because it is suggestive that the marginal
product of capital in equilibrium has not gone down.11 On the other hand, we
find that the average inflation levels have gone up while the share of exports in
GDP has gone down during the crisis. Both of these are consistent with the hy-
pothesis that the economy enters some type of economic tailspin both in its export
capacity and in the quality of its macroeconomic policy during prolonged crises.
There are several ways in which we can interpret the strength of the open_ forest
variable. At a general level, open_ forest is an indicator of an economy’s flexibility.
It measures the possibilities that an economy has of moving to the production of
other goods, weighted by the sophistication of these goods. It thus combines a hy-
pothesis that flexibility is important with the hypothesis that countries develop
Growth Collapses 417

Figure 15.6
Conditional Hazard Function: Alternative Specifications

producing rich-country goods, as suggested by Hausmann, Hwang, and Rodrik

(2007). In table 15.18 we use a measure of open forests that does not weigh
goods by their PRODY, thus implicitly assigning goods equal value. The absolute
value of the coefficient and its significance are very similar to those obtained by
open_ forest. When both variables are introduced in the regression (not shown) nei-
ther of them is significant at 5 percent, suggesting that they are too collinear to
distinguish between the alternative hypotheses they represent. To the extent that
pure density is a simpler explanation, Occam’s razor would suggest sticking with
it. Column 2 looks at other potential measures of flexibility. One is a Herfindahl
index of export concentration. The idea is that countries with more concentrated
export sectors will have a harder time reacting to adverse shocks as it will be
more difficult for other industries to expand. It is possible that open_ forest is sim-
ply capturing the effects of having a diversified export structure. The result shown
in columns 2–4 are surprising: export concentration seems to be associated with a
higher, not lower, probability of exiting the crisis.
A similar fact appears to be true about the log of pPopulation, another measure
of the size of the economy and of its possible flexibility (column 3). In any case, the
coefficient on open forests is robust to the inclusion of these alternative indicators
Table 15.16
418

Duration Regressions, PWP Stratified Cox Model

Dependent variable: Years in crisis

Representation, hazard (1) (2) (3) (4) (5) (6) (7)

Log GDP per working-age person 0.186
0.108
0.098
0.247 C0.341
0.093 C0.269
(2.74)*** (1.09) (0.83) (1.47) (2.06)** (0.8) (1.66)*
Latin America C1.184 C0.989 C0.793 C0.728 C1.010 C0.831
(4.97)*** (3.51)*** (2.29)** (2.17)** (3.53)*** (2.34)**
Africa C1.298 C0.690
0.408
0.407 C0.747
0.477
(4.31)*** (1.66)* (0.66) (0.64) (1.75)* (0.75)
South and Central Asia C0.585
0.384
0.425
0.535
0.390
0.471
(1.85)* (1.14) (1.27) (1.38) (1.17) (1.39)
East Asia and Pacific C0.576
0.127
0.410
0.373
0.161
0.395
(2.2)** (0.45) (0.92) (0.83) (0.57) (0.9)
Central and Eastern Europe C1.531 C1.002 C0.605
0.489
0.745 C0.575
(4.2)*** (1.83)* (2.28)** (1.29) (1.5) (1.73)*
Middle East and North Africa C0.731
0.249 0.278 0.198
0.254 0.236
(2.43)** (0.79) (0.67) (0.44) (0.81) (0.54)
1970s 0.280 0.107
0.529 0.240 C0.675
(1.27) (1.34) (1.03) (1.98)**
1980s
0.183
0.030
0.091
0.597
0.057 C0.880
(1.03) (0.15) (0.45) (1.46) (0.29) (2.88)***
1990s
0.107
0.564 C0.779
(0.57) (1.41) (2.33)**
2000s
0.178 1.274 0.780 1.230
(0.6) (3.76)*** (2.41)** (3.57)***
Open forest 0.355 0.350 0.388 0.283 0.307
(3.17)*** (2.5)** (2.75)*** (2.08)** (1.92)*
Democracy (polity) 0.029 0.032 0.032
(1.71)* (1.63) (1.49)
Sudden stop 0.044 0.000 0.010
(0.24) (0) (0.05)
Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner
Log change in real merchandise exports 0.288
1.357
1.939
(0.8) (0.63) (0.67)
War 0.022
(0.05)
Natural disaster 0.040
(0.12)
Growth Collapses

Log of inflation
0.070

(0.11)
Change in polity indicator
0.138
(0.57)
Change in exports*open forest 0.108 0.155
(0.64) (0.69)
Polity*change in merchandise exports
0.027
(0.79)
Polity*sudden stops 0.007
(0.28)
N 535 535 233 191 175 230 191
Pseudo-R2 0.1% 2.2% 4.3% 4.2% 4.4% 3.9% 4.2%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%, ***
1%.
419
Table 15.17
420

PWP Stratified Cox Model, Alternative Controls

Dependent variable: Years in crisis

Representation, hazard (1) (2) (3) (4) (5) (6) (7) (8)

Log GDP per working-age person
0.093
0.202
0.207
0.147
0.196 C0.573 C0.260 C1.890
(0.51) (0.07) (0.14) (0.40) (0.17) (0.47) (0.41) (0.12)
Open forest 0.381 0.394 0.389 0.345 0.430 0.736 0.434 1.272
(2.86)*** (3.03)*** (2.24)** (2.5)** (3.71)*** (2.97)*** (2.87)*** (1.47)
Sudden stop 4
0.050
(0.29)
Latin America C1.114 C1.142 C0.869 C0.941 C1.133 C1.593 C1.017 C1.998
(3.35)*** (3.68)*** (2.83)*** (3.22)*** (3.83)*** (2.94)*** (3.13)*** (2.49)**
Africa
0.666 C0.821
0.485 C0.804 C0.795
1.186 C0.868
1.803
(1.37) (1.83)* (1.09) (1.88)* (1.9)* (0.94) (1.84)* (0.91)
South and Central Asia
0.303
0.528
0.244
0.221
0.483 C1.748
0.536 C2.658
(0.87) (1.51) (0.64) (0.6) (1.46) (2.47)** (1.6) (1.92)*
East Asia and Pacific
0.307 C0.728
0.139
0.411
0.519
0.409
0.485
2.351
(0.89) (1.93)* (0.4) (1.28) (1.57) (0.81) (1.22) (1.52)
Central and Eastern Europe C1.195 C1.245 C0.499 C0.656 C1.047 C1.132 C0.483
1.392
(2.31)** (1.88)* (1.66)* (2.33)** (1.97)** (2.03)** (1.89)* (1.24)
Middle East and North Africa
0.240
0.120
0.041
0.184
0.375
0.848
0.160
0.270
(0.64) (0.37) (0.11) (0.52) (1.13) (1) (0.4) (0.24)
1970s C1.278 0.263 0.255 C1.006 C0.948
0.545
(3.28)*** (1) (0.9) (2.9)*** (0.67)
1980s C1.374
0.143
0.183 C1.303 C1.295 C1.554 C0.400
0.568
(4.45)*** (0.6) (0.74) (3.55)*** (4.38)*** (2.31)** (1.85)* (0.94)
1990s C1.231 C1.143 C1.227 C1.211
0.198
0.523
(3.44)*** (2.56)** (3.56)*** (1.93)* (0.82) (0.74)
2000s 0.797
(1.96)*
Tariff rate 3.602
50.989
(0.65) (0.85)
Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner
Total years of schooling 0.054 0.049
(0.94) (0.56)
Liquid liabilities/GDP 0.003 0.005
(1.07) (0.43)
Openness 0.003 0.008
(1.43) (0.81)
Growth Collapses

Life expectancy at birth, total (years)- C0.015
0.003

(0.28) (0.03)
Interaction between polity and terms of 0.057
0.125
(0.22) (0.13)
N 190 177 179 191 224 100 183 60
Pseudo-R2 5% 5% 5% 5% 5% 8% 5% 12%
Note: z-statistics in parentheses. Asterisks denote level of significance as follows: *
10%, **
5%, ***
10.
421
422 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Figure 15.7
Evolution of Covariates during Crises

of flexibility, as well as of land area (column 4). In column 5 we present an addi-

tional robustness test, which is to control for the magnitude of the initial crisis by
introducing measures of the magnitude of the initial shock. In particular, we con-
trol for initial and lagged GDP growth, initial and lagged export growth, and the
initial terms of trade shock. Since some of these measures will be correlated with
the dependent variable by definition, we do not use this set of controls more
broadly in this paper. Open_ forest is robust to the inclusion of these controls.

15.6 Concluding Comments

This paper has analyzed episodes during which economic growth decelerates to
negative rates in a sample of 180 developing and developed economies. We iden-
tify 535 episodes of output contractions. The distribution of these episodes is
highly skewed: while the median duration is 2 years, more than a quarter of them
last more than 7 years and roughly 14 percent last at least 15 years. Developing
countries are much more likely to experience prolonged contractions than are in-
dustrial countries.
Table 15.18
Alternative Hypotheses
Dependent variables: Years in crisis (1) (2) (3) (4) (5)
Log GDP per working-age person
0.249
0.23
0.365
0.38
0.606
(1.23) (1.15) (1.75)* (1.78)* (1.66)*
Open forest 1.1
(2.89)***
Open forest (Density only) 0.586 0.917 1.286 1.334
(3.68)*** (3.54)*** (4.27)*** (4.15)***
Herfindahl 1.891 2.491 2.772
(1.65)* (2.15)** (2.28)**
Log of population
0.245
0.254
(2.70)*** (2.42)**
Area (sq. km) 0.018
(0.25)
Growth in terms of trade at t ¼ 1 7.445
(4.54)***
Growth in GDP at t ¼ 1 30.394
(4.69)***
Growth in merchandise exports at t ¼ 1
0.638
(0.64)
Growth in GDP at t ¼ 0 5.519
(0.78)
Growth in merchandise exports at t ¼ 0 2.642
(2.48)**
Constant 0.266
1.916 1.513 0.325
9.602
(0.11) (0.72) (0.51) (0.11) (1.80)*
Latin America
1.557
1.603
1.768
1.798
1.054
(3.61)*** (3.72)*** (4.04)*** (4.03)*** (1.59)
Africa
1.187
1.244
1.167
1.201
0.613
(1.76)* (1.86)* (1.73)* (1.73)* (0.58)
South and Central Asia
0.825
0.783
0.453
0.475
1.428
(1.22) (1.17) (0.67) (0.68) (1.54)
East Asia and Pacific
0.177
0.164
0.278
0.271 0.985
(0.37) (0.35) (0.58) (0.56) (1.27)
Central and Eastern Europe
1.757
1.713
1.686
1.576
2.426
(2.41)** (2.36)** (2.23)** (1.97)** (2.33)**
Middle East and North Africa
0.428
0.63
0.695
0.738 0.16
(0.87) (1.26) (1.37) (1.41) (0.21)
1970s
0.879
0.875
0.918 0.255
0.615
(1.43) (1.43) (1.51) (0.98) (0.72)
1980s
1.701
1.648
1.71
0.542
1.096
(2.93)*** (2.83)*** (2.93)*** (2.43)** (1.44)
1990s
1.31
1.221
1.177
0.716
(2.20)** (2.05)** (1.97)** (0.91)
2000s 1.161
(1.93)*
Observations 229 229 229 225 130
Number of groups 100 100 100 97 62
Note: Absolute value of z statistics in parentheses.
* significant at 10%; ** significant at 5%; *** significant at 1%.
424 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

We have studied the factors that coincide with the onset of these crises. In terms
of statistical significance, we find that the change in exports is the variable most
strongly associated with the probability of suffering a crisis—at least in develop-
ing countries. A one-standard-deviation decrease in the growth rate of merchan-
dise exports implies a 5.47 percentage point increase in the probability of a crisis.
In terms of economic significance, a one-standard-deviation increase in inflation
appears to be slightly more damaging, though the coefficient is somewhat less
precisely estimated. Wars, sudden stops, and political transitions also tend to co-
incide with the onset of crises.
The duration of crisis is somewhat more difficult to predict. Surprisingly, the
variables that we find to be significantly associated with the probability of a crisis
occurring do not appear to be related to crisis duration. One variable that we find
to be robustly associated with crisis duration—aside from continent and decade
effects—is a measure of the density-weighted value of a country’s alternative ex-
port basket suggested by Hausmann and Klinger (2006). We take this measure to
be an indicator of the flexibility of an economy’s productive apparatus in adapt-
ing to external shocks. This intuition is confirmed by case studies of collapse epi-
sodes in developing countries that emphasize the role that poor performance in
the nontraditional export sector plays in deepening growth collapses.12 We also
find evidence that an interaction between democracy and inequality, as suggested
by Rodrik (1999), affects the duration of crises.
Our results leave open several avenues for future analysis. On the one hand, it
would be desirable to refine the duration model’s predictive capacity. One possi-
ble avenue for doing this would be to adopt a model of time-varying covariates.
We have shied away from that alternative because we find it easier to believe in
the exogeneity of changes that took place before the onset of the crisis than we
would in changes that occur during the crisis. Another avenue is to explore the
possible channels through which open_ forest is correlated with crisis duration.
Our first tentative attempts to get at this issue failed to find a significant interac-
tion between export collapses and open_ forest. Several explanations could account
for this fact. Open_ forest may be a more general measure of the economy’s adapt-
ability to several types of productivity shocks, a multiplicative interaction may be
too coarse to capture generalized nonlinearities, or open_ forest may be proxying
for some unmeasured country-specific effect. Further research could help discern
among these potential competing hypotheses.
We also find that decreasing conditional and unconditional hazard rates are a
pervasive characteristic of our estimation. While this is not necessarily surprising
for unconditional rates, as it may be a consequence of the changing composition
of the population, it is definitely counterintuitive when these hazard rates are
conditioned on estimated country random effects and covariates. Even though
Growth Collapses 425

decreasing hazard rates can be accounted for within a neoclassical model as a re-
sult of substantial, permanent shocks to output, the depth and duration of some
recessions in this sample appear hard to explain. To take just one example, there
is widespread agreement among many observers of the Bolivian economy that its
institutional, political, and macroeconomic framework was more solid at the be-
ginning of this century than in the mid-seventies.13 However, GDP per working-
age person was 14.9 percent lower in 2004 than it was in 1978, despite the fact
that world productivity surely increased during this period. Further investigation
of the characteristics of recessions may allow us to find ways to discriminate be-
tween alternative interpretations of this type of phenomenon.

Notes

We would like to thank Masami Imai, Cameron Shelton, Sanjay Reddy, Eduardo Zambrano, and semi-
nar participants at Harvard University for their valuable comments and suggestions. Alejandro
Izquierdo and Bailey Klinger kindly provided data used in this analysis. The usual disclaimer applies.
1. The formal definition of these concepts will be introduced in section 3.

2. Calvo, Izquierdo, and Mejı́a (2004), Calvo, Izquierdo, and Loo-Kung (2005), Calvo (2005), Calvo,
Izquierdo, and Talvi (2006).

3. An extensive literature has developed attempting to explain growth volatility. See Ramey and
Ramey (1995), Imbs (2002), and Aghion and Banerjee (2005) for discussions.
4. The definition of recoveries used by Cerra and Saxena (2005) differed from ours in that they define a
trough as any year in which growth changes from negative to positive. In practice, this implies that
they underestimate the length of ‘‘double dip’’ and ‘‘n-tuple dip’’ crises.
5. Reddy and Minoiu (2006) also use the first ‘‘turning point’’ after a crisis to date the end of the crisis,
which is subject to the same objection made regarding the Cerra and Saxena technique.
6. Hausmann, Pritchett, and Rodrik (2005), in contrast, have looked at episodes during which growth
accelerates. Their main finding is that accelerations are not well explained by macroeconomic policy
reforms.

7. Two useful recent surveys are Hosmer and Lemeshow (1999) and Box-Steffensmeier and Jones
(2004).
8. Analytically, it is important to distinguish between the causes that lead countries to fall into crises
and the reasons that their recovery speeds differ. Although there could be similarities between both
processes—and crises generated by large shocks may be more difficult to get out of—they may well
be very different. For example, in a related analysis, Collier and Hoeffler (2004) have found that the
causes that lead countries to fall into civil wars are very different from those that determine the dura-
tion of those wars.
9. Another difference with our definition is that we use annual data, while Calvo and his coauthors,
who study the short-run dynamics of sudden stops, use monthly data.
10. This result is not an artifice of sample reduction either: running the regression of column 1 for the
sample in column 4 gives significantly positive continent dummies for Latin America, Africa, and the
MENA region.
426 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

11. Note that with this specification, we cannot tell if that is because of lower levels of the capital stock
or because an improvement in productivity.
12. See, for example, Hausmann and Rodrı́guez (2006) on Venezuela and Auty (2001) on Saudi Arabia.

13. See Morales and Sachs (1992) and Jimenez Zamora, Candia, and Mercado Lora (2005).

References

Aghion, P., and A. Banerjee. 2005. Volatility and Growth. Oxford: Oxford University Press.
Ali, Ali A., and Ibrahim A. Elbadewi. 1999. Inequality and the Dynamics of Poverty and Growth. Center for
International Development at Harvard University, Working Paper 32.
Andersen, P. K., J. P. Klein, and M. J. Zhang. 1999. ‘‘Testing for Centre Effects in Multi-centre Survival
Studies: a Monte Carlo Comparison of Fixed and Random Effects Tests. Statistics in Medicine 18, no. 12:
1489–1500.
Auty, R. M. 2001. The Political Economy of Resource-driven Growth. European Economic Review 45:
839–846.
Barro, R. 1989. ‘‘A Cross-country Study of Growth, Saving, and Government.’’ Working Paper No.
2855, NBER, Cambridge, MA.
Barro, Robert J., and Jong-Wha Lee. 2000. ‘‘International Data on Educational Attainment: Updates and
Implications.’’ Working Paper No. 42, Center for International Development, Harvard University,
Cambridge, MA.
Ben-David, D., and D. H. Papell. 1998. ‘‘Slowdowns and Meltdowns: Postwar Growth Evidence from
74 Countries.’’ Review of Economics and Statistics 80, no. 4: 561–571.
Blyde, Juan, Christian Daude, and Eduardo Fernandez Arias. 2006. ‘‘Growth Collapses and Productiv-
ity Destruction.’’ Mimeo., Inter-American Development Bank, Washington, D.C.
Bodman, Philip M. 1998. ‘‘Asymmetry and Duration Dependence in Australian GDP and Unemploy-
ment.’’ Economic Record 74, no. 227: 399–411.
Box-Steffensmeier, J. M., and B. S. Jones. 2004. Event History Modeling: A Guide for Social Scientists. New
York: Cambridge University Press.
Calvo, G. A. 2005. ‘‘Crises in Emerging Market Economies: A Global Perspective.’’ Working Paper No.
10520, NBER, Cambridge, MA.
Calvo, G. A., A. Izquierdo, and R. Loo-Kung. 2006. ‘‘Relative Price Volatility under Sudden Stops: The
Relevance of Balance-Sheet Effects.’’ Journal of International Economics 69, no. 1: 231–254.
Calvo, G. A., A. Izquierdo, and L.-F. Mejia. 2004. ‘‘On the Empirics of Sudden Stops: The Relevance of
Balance-Sheet Effects.’’ Working Paper No. 10520, NBER, Cambridge, MA.
Calvo, Guillermo A., Alejandro Izquierdo, and Ernesto Talvi. 2006. ‘‘Phoenix Miracles in Emerging
Markets: Recovering without Credit from Systemic Financial Crises.’’ Working Paper No. W12101,
NBER, Cambridge, MA.
Cerra, Valerie, and Sweta Chaman Saxena. 2005. ‘‘Growth Dynamics: The Myth of Economic Recov-
ery.’’ Working Paper No. WP/05/147, IMF, Washington, D.C.
Céspedes, Luis Felipe, and Jose de Gregorio. 2005. ‘‘Recuperaciones Cı́clicas: Evidencia Internacional.’’
Mimeo., Banco Central de Chile. Available at http://www.bcentral.cl/esp/estpub/otrasconf/otras/
pdf/Cespedes_De_Gregorio.pdf.
Growth Collapses 427

Collier, P., and P. Hoeffler. 2004. ‘‘Greed and Grievance in Civil War.’’ Oxford Economic Papers 56, no. 4:
563–595.
Diebold, Francis X., and Glenn D. Rudebusch. 1990. ‘‘A Nonparametric Investigation of Duration De-
pendence in the American Business Cycle.’’ Journal of Political Economy 98, no. 3: 596–616.
Di Venuto, Nicholas, and Allan Layton. 2005. ‘‘Do the Phases of the Business Cycle Die of Old Age?’’
Australian Economic Papers 44, no. 3: 290–305.
Galton, Francis. 1889. Natural Inheritance. New York: MacMillan.
Gleditsch, Kristian Skrede. 2004. ‘‘A Revised List of Wars Between and Within Independent States,
1816–2002.’’ International Interactions 30: 231–262.
Hausmann, R., J. Hwang, and D. Rodrik. 2007. ‘‘What You Export Matters.’’ Journal of Economic Growth
12, no. 1: 1–25.
Hausmann, R., and B. Klinger. 2006. ‘‘Structural Transformation and Patterns of Comparative Advan-
tage in the Product Space.’’ Working Paper No. 128, Center for International Development, Harvard
University, Cambridge, MA.
Hausmann, R., L. Pritchett, and D. Rodrik. 2005. ‘‘Growth Accelerations.’’ Journal of Economic Growth
10, no. 4: 303–329.
Hausmann, R., and F. Rodrı́guez. 2006. ‘‘Why Did Venezuelan Growth Collapse?’’ Mimeo., Center for
International Development, Harvard University, Cambridge, MA.

Hosmer, D. W., and S. Lemeshow. 1999. Applied Survival Analysis: Regression Modeling of Time to Event.
New York: Wiley.

International Country Risk Guide. 1999. PRS Group. CD-ROM.

Imbs, J. 2002. ‘‘Why the Link between Volatility and Growth is Both Positive and Negative.’’ Discus-
sion Paper No. 3561, CEPR, London.

Jimenez Zamora, Elizabeth, Gaby Candia, and Marcelo Mercado Lora. 2005. ‘‘Economic Growth, Pov-
erty and Institutions: A Case Study of Bolivia.’’ Mimeo., Global Development Network, World Bank,
Washington, D.C.
Lin, D. Y., and L. J. Wei. 1989. ‘‘The Robust Inference for the Cox Proportional Hazards Model.’’ Journal
of the American Statistical Association 84: 1074–1078.
Marshall, M. G., K. Jaggers, and T. R. Gurr. 2004. ‘‘Polity IV Dataset.’’ Electronic File. College Park: Uni-
versity of Maryland.
McFadden, D. 1974. ‘‘The Measurement of Urban Travel Demand.’’ Journal of Public Economics 3: 303–
328.
Mills, Terence C. 2001. ‘‘Business Cycle Asymmetry and Duration Dependence: An International Per-
spective.’’ Journal of Applied Statistics 28, no. 6: 713–724.
Mora, Ricardo, and Georges Siotis. 2005. ‘‘External Factors in Emerging Market Recoveries: An Empiri-
cal Investigation.’’ European Economic Review 49, no. 3: 683–702.
Morales, Juan Antonio, and Jeffrey Sachs. 1990. ‘‘Bolivian Debt Management.’’ In Developing Country
Debt and Economic Performance, ed. Jeffrey Scahs, 252–255. Chicago: University of Chicago Press.

Mudambi, Ram, and Larry W. Taylor. 1991. ‘‘A Nonparametric Investigation of Duration Dependence
in the American Business Cycle: A Note.’’ Journal of Political Economy 99, no. 3: 654–656.
428 Ricardo Hausmann, Francisco Rodrı́guez, and Rodrigo Wagner

Prentice, R. L., B. J. Williams, and A. V. Peterson. 1981. ‘‘On the Regression Analysis of Multivariate
Failure Time Data.’’ Biometrika 68: 373–379.
Pritchett, Lant. 2000. ‘‘Understanding Patterns of Economic Growth: Searching for Hills among Pla-
teaus, Mountains, and Plains.’’ World Bank Economic Review 14, no. 2: 221–250.
Ramey, G., and V. A. Ramey. 1995. ‘‘Cross-Country Evidence on the Link between Volatility and
Growth.’’ American Economic Review 85, no. 5: 1138–1151.
Reddy, Sanjay G., and Camelia Minoiu. 2006. ‘‘Real Income Stagnation of Countries, 1960–2001.’’
Mimeo., Columbia University, New York.
Rodrik, D. 1999. ‘‘Where Did All the Growth Go? External Shocks, Social Conflict, and Growth Col-
lapses.’’ Journal of Economic Growth 4, no. 4: 385–412.
Ros, Jaime. 2005. ‘‘Divergence and Growth Collapses: Theory and Empirical Evidence.’’ In Beyond
Reforms: Structural Dynamics and Macroeconomic Vulnerability, ed. J. A. Ocampo, 211–232. Palo Alto:
Stanford University Press.

Sichel, Daniel E. 1991. ‘‘Business Cycle Duration Dependence: A Parametric Approach.’’ Review of Eco-
nomics and Statistics 73, no. 2: 254–260.

Struthers, C. A., and J. D. Kalbfleisch. 1986. ‘‘Misspecified Proportional Hazard Models.’’ Biometrika 73,
no. 2: 363–369.
Wooldridge, J. G. 2001. Econometric Analysis of Cross-Section and Panel Data. MIT Press, Cambridge, MA.

World Bank. 2006. World Development Indicators 2006. Electronic File. Washington, D.C.: The World
Bank.
VI The Man Behind the Mind
16 The Columbia Years

Edmund S. Phelps

So, Guillermo: this is your life!

I remember someone advising me in a similar situation a while back that in a
speech of this type accuracy about the subject’s life is not important. Style is im-
portant. So it’s often better to simplify here and embroider there.
There is a lot of bad advice out there, too. You may have seen the Orson Welles
movie in which the Charlton Heston character, south of the border on his honey-
moon, accomplishes prodigious feats that save the town from bad guys, deeds his
wife and others know nothing of.1 Commenting, the Marlene Dietrich character
asks, ‘‘What does it matter what you say about someone?’’ Let me tell you, in Fest-
schrift speeches it matters a helluva lot!
I won’t go over with you the story line of that movie. Today it’s the story line
of Guillermo Calvo we want to stay focused on. But I do remember what I think
was the movie’s theme: some important stuff that some people do doesn’t get
recognized.
Well, for today, anyway, that won’t be true of the stuff Guillermo has done.

The Early Years

We all know that Guillermo grew up in Buenos Aires, hometown of his wife
Sara—and, as a matter of fact, of my wife Viviana. After attending the University
of Buenos Aires, he went for a doctorate in economics at Yale, where I got my
doctorate. After a spell of teaching in South America, he joined the Economics De-
partment at Columbia, which I had done a year earlier. (Later he went to Penn,
which I had also done.)
Not a great deal is known about Guillermo’s Yale period. Legend has it that,
one day, Guillermo raised his hand in James Tobin’s class to point out a difficulty
with something in the lecture. According to the legend, Guillermo was right—and
their relationship went nowhere from there. Curiously, around the same time I
432 Edmund S. Phelps

also had a clash with Jim in a crowded seminar room over a book of mine, Fiscal
Neutrality toward Economic Growth. Maybe all these parallel experiences were
omens that Guillermo and I were destined to come together in a common cause.
In fact, Guillermo also took a graduate course of mine on growth economics. It
was my first graduate course and I would arrive at every meeting with about
forty equations. (I needed to have them in front of me because I was unable to re-
member them.) I thought it was clear from the look on his face that he understood
everything. He sat by the east wall of the Cowles Foundation classroom, near
Susan Rose (later Susan Rose Ackerman). Though not bound to accuracy, as I
said, I confess that I don’t actually recall any chats with him, and yet I know that
we knew each other somehow. Little did I know that in six or seven years’ time
we would become allies in writing a new chapter of macroeconomics and close
friends.
I left Yale in January 1966, starting my work on unemployment and inflation at
LSE [London School of Economics], Penn, and CASBS [Center for Advanced
Study in the Behavioral Sciences] at Stanford. Guillermo, after writing his disser-
tation with Tjalling Koopmans, left to teach in South America under a program of
the Ford Foundation.

The Years at Columbia

I reached Columbia in September 1971, knowing that the Economics Department

was in a deep hole but hoping there was some way of climbing out. Three of us—
Kelvin Lancaster, the chairman, Ronald Findlay, and I—were determined to turn
the department around. But how to do it? The emergence of Guillermo was our
first big break—though I did not appreciate how big it was going to be, particu-
larly for me.
As I remember it, I received one day that fall an international telegram from
Guillermo:

am leaving the andes stop need to find position in research university stop can
you provide assistant professorship? stop guillermo calvo

I set about persuading my colleagues that Guillermo was extremely bright and
highly talented. I’m not sure I tried very hard to spread the word to other univer-
sities. In any case, it appeared that no other university had bid, at least not one
Guillermo wanted to go to, and we landed him for delivery in January 1973. (At
the same time I worked strenuously to recruit John Taylor, even to the point of
hanging out at a Stanford party to get to know him better.) John was signed up
too, for delivery six months after Guillermo. Someone told me that one of them, I
am not sure which, was paid $500 more than the other for some reason, and this
The Columbia Years 433

differential continued untouched as long as they were both at Columbia. It was an

extraordinarily bureaucratic place in those days.
The years with Guillermo and John at Columbia were enormously productive. I
had wondered whether my future would be something of a let-down after the fer-
ment of the 1960s and the great luck I had in that decade. And when they arrived
at Columbia nothing much happened at first: I was doing optimal taxation in a
Rawlsian vein, and Guillermo was finishing up old work. John, however, drew
me back into wage and price setting, and when I got stuck I would sometimes
chat about the matter with Guillermo. By 1975, we were all on the same page. We
did nonsynchronous price and wage setting, a modern version of contract theory,
optimal monetary stabilization theory, and some nonmonetary modeling, too. The
main effort was the building of New Keynesian theory (in Michael Parkin’s term.)
Toward the end of the decade Stanley Fischer told me he felt Columbia had the
best department for macroeconomics in the country. Robert Lucas exclaimed,
‘‘Ned, you have a whole school here.’’ Maybe there was no day in the 1970s when
I had a mountain high like the feeling I had a couple of times in the 1960s. But
there was an extraordinary atmosphere of continual challenge, accomplishment,
and the sense of a carpet of new problems and discoveries ahead. When John Tay-
lor left in 1979 it was sad, and when Guillermo left too, in 1986, it was painful.

Some of the Papers

Of the many papers that Guillermo wrote or cowrote, I particularly treasure four
that I was tacitly or explicitly involved in. The most conspicuous of them was his
paper on price stickiness, which became an instant classic. At my urging, John
had looked at my 1968 paper on money-wage dynamics and wrote down in that
spirit a difference-equation system for the wage rates of the various vintages—
this quarter, last quarter, and the two before that; meanwhile I was studying a
similar system with which to analyze the problem of optimal disinflation. I be-
came almost obsessed with trying to get this system into a continuous-time frame-
work but failed because I hadn’t thought of what would be Guillermo’s trick of
supposing that the old prices fade away with a half-life. It was a perfect paper be-
cause you had to read it carefully and think about it a while before you could feel
that you understood it. The model had the property that a step-increase of effec-
tive demand, such as a fiscal stimulus, would cause both employment and the in-
flation rate to jump; then both would recede back to their steady-state levels, as
the price level approached its new steady-state level. An implication was that
the stronger the stimulus, and thus the higher the level to which employment
jumped, the faster the inflation rate would be falling (following its jump)—this de-
spite the accelerationism inherent in the natural rate construction contained in the
434 Edmund S. Phelps

model. (Some thought Guillermo had got the sign wrong.) It is true that this theo-
rem could have been proven by John and me using our period model of discrete
price or wage setting, but we had not seen this delicious implication.
Some policy economists, understanding that the price level tends to restabilize,
given an inflation stabilization policy in the central bank, conclude that the natu-
ral rate is no bar to fiscal stimulus after all, since the inflation rate cannot explode;
but this is an evident non sequitur. It is still impossible to sustain by effective-
demand stimulus an employment path bounded above its natural path. The
Calvo solution is a demonstration that, when the economy is driven off its natural
course, the equilibrium path immediately heads toward home.
Of the other three papers, one was the paper on time consistency, which I hope
the Phelps-Pollak paper helped to inspire, though perhaps it did not. Another was
the paper on general equilibrium in a customer-market economy, which I finally
persuaded Guillermo to do with me. That paper turned out to be a forerunner to
the many nonmonetary models of structural slumps and structural booms that I
began constructing in the last years of the 1980s. By that time, to my embarrass-
ment, I had forgotten about the earlier paper with Guillermo, although I can say
that I did it differently the second time around.
The joint effort I would like especially to recount is the one on contract theory. I
had been invited to write a paper on indexation of money wages and prices. It
was suspected that, with enough sophistication, every price and wage would be
indexed to the consumer price level and, when that is done, a step-decrease in ef-
fective demand would trigger an immediate drop in all prices and money wages,
which would save employment from a Keynesian recession. Some of my rumina-
tions led me to criticize what I called the neo-neoclassical theory of wage con-
tracts. I argued that the states of the world in that theory are not of this world: in
the real world, at least the capitalist world, the effective state at the firm is to an
important degree in the employer’s mind. And the employees cannot see into the
head of their employer. So if he tells them the new state is such that they should
accept a wage cut without striking or cutting back on their performance, they will
not be able necessarily or generally to see all or even much of the evidence that
has led the employer to declare the new state; the employer himself may have
very little evidence, and be unable to convey much of his observations and
impressions. This led me to discussions with Guillermo aimed at a modern con-
tract theory in which the wage has to be a function of observables, such as the lay-
off level and the level of profits at the firm. We sketched such a model in the
appendix to the paper I wrote. Guillermo did the analysis required to derive
what was in effect a wage curve relating the wage called for by the contract to the
number of employees in the laid-off status. In retrospect, I saw that this was the
wrong place to launch such a basic innovation.
The Columbia Years 435

In his Columbia years Guillermo also found time to write plenty of other
papers. A favorite of mine is his reworking of Tobin’s ‘‘dynamic aggregative
model,’’ which had the distinguishing feature that it dispensed with Keynes’s
investment demand function (whose inverse gives the marginal efficiency of in-
vestment), replacing it with the marginal productivity of capital function. The
paradox of this Keynesian landmark was its tacit implication that consumer de-
mand was powerless to stimulate output and employment—there was 100 per-
cent crowding out. In the Calvo version, a positive shift in the propensity to
consume becomes expansionary through its impact on rational expectations of
resulting inflation over the future. I wondered at the time whether the paper was
a helping hand stretched out to Keynesians or a sort of joke. For if a consumer de-
mand stimulus boosts aggregate demand and employment only to the extent it is
expected to cause prolonged inflation (and against a backdrop of fixed-money
wage rates), it is not apt to be a popular policy tool.

The Personal Side

I haven’t forgotten that during our years together at Columbia Guillermo and I
were pretty close—especially when I was between marriages and later when he
was between marriages. Whether it was about opera or about life, it was usually
Guillermo I talked to. And he often had some insight or perception to share.
I remember one evening when he and I went to Mozart’s Magic Flute at the Met.
At some point the talkative birdman, Papageno, who had been gagged for awhile,
is ungagged and says, ‘‘Ah, now I am Papageno again.’’ That line hugely amused
Guillermo. His response to it made me realize that it expressed the theme of the
opera—that there is nothing as central to us as the ideas and values that we hold
and need to profess. I also realized that there was nothing more important to
Guillermo than developing and getting out his thoughts. He is outstandingly seri-
ous about his ideas and his values. I think we are alike in that way. Someone
wrote recently that I am mild-mannered. Guillermo is also mild-mannered. We
are like Clark Kent. Beneath Guillermo’s genial exterior is a rigorous and fierce
mind that has no patience with sloppy thinking and clueless modeling. I think I
am the same—at least I hope so. This must be why we became very close.
When he was speaking at my own festschrift celebration in 2001, Guillermo
said, introducing his junior coauthors, that they were my ‘‘grandchildren.’’ That
makes him my son. Guillermo: I am a proud father.

Note

1. The film is A Touch of Evil, with Charlton Heston, Janet Lee, Orson Welles, Marlene Dietrich, and
Akim Tamiroff.
17 The Practitioner

Roque B. Fernández

It is a great honor for me to be present at this event in recognition of the contribu-

tions of Guillermo Calvo to economic theory, as well as to practical policy-
making. As have most of the members participating in this conference, I had the
privilege of interacting with him in different moments of my professional life.
First, when I was student at the University of Chicago; second, at the time where
we were creating the University of CEMA in Buenos Aires; and third, when I
assumed a place on the Central Bank of Argentina Board of Directors, and later
on as Economic Minister. Most participants are acquainted with his very sig-
nificant theoretical work, so I will focus my presentation on the role of Guil-
lermo Calvo as an economic practitioner or adviser rather than as a theoretical
economist.
At the time of my being in the Central Bank, I asked the IMF for technical assis-
tance, suggesting that I would need the technical advice of Guillermo Calvo, and
the IMF conceded my request. That was the end of the eighties and beginning
of the nineties when Argentina was suffering a hyperinflation. In short, there
were two fundamental problems: one, the conventional fiscally ridden monetary
expansion, and the other, the Central Bank support of the financial system. As the
dynamics of hyperinflation evolved, monetary expansion was mostly devoted to
preserving the financial system, and the Central Bank acted as a lender of last
resort to commercial banks whose liabilities were short-term deposits indexed to
wholesale prices, consumer prices, and a special index contructed by averaging
nominal interest rates paid to depositors. Reserve requirements were at the mar-
gin almost 100 percent, and were remunerated by the government at at a rate that
kept pace with the prevailing rate of inflation. The peso devaluated erratically,
sometimes overshooting the rate of inflation, and depositors often ran against
commercial banks to withdraw pesos to buy dollars.
Jointly with other members of the Central Bank Board I had worked out
some ideas for rescheduling deposits, but we were entangled with several com-
plex issues on bank regulation. Guillermo Calvo helped to disentangle the
438 Roque B. Fernández

complexities to reach a very simple proposal. The idea consisted of offering

depositors a government bond denominated in dollars in exchange for short-term
peso deposits (most deposits were seven-day deposits). This idea was enacted in
1990. The government bond was a ten-year bond (BONEX 89), and the exchange
was not a voluntary exchange; except for special exceptions (small deposits, old
people, or people affected by serious illness) the exchange was compulsory. As a
result of the financial reform, in a few months the rate of inflation was signifi-
cantly reduced and Argentina started a period of sustained growth with price
stability that lasted for more than ten years. The dollar-denominated bonds were
fully paid, and depositors reestablished the confidence in the financial sector.
Although the idea helped solve a chronic problem of more than twenty years of
high and volatile inflation, and it was accepted over time as a reasonable solution,
initially there was some resistance, and Guillermo Calvo had to endure a personal
situation totally unexpected in his role of economic adviser. After participating in
a lunch meeting with the Argentine Bankers Association (ADEBA) where he did
not talk or answer any question related to financial reform, one of the bankers—
who had anticipated that the rescheduling of bank deposits was a potential policy
decision—decided to present a complaint to the Argentine IMF representative. Es-
sentially the banker blamed the IMF for sending an adviser who had not thought
enough about the complexity of the situation and was giving policy makers mis-
leading technical assistance that could jeopardize the stability of the Argentinean
banking system.
Was the banker right? Policy makers could be wrong, and could be responsible
for taking quick decisions, perhaps without full information, and perhaps without
enough time to think about what they are doing, but economic advisers are sup-
posed to know everything beforehand. They are supposed to have studied and
evaluated a similar situation in the past, and they are supposed to have the rele-
vant economic theory at hand to answer all the questions policy makers might
have. Was Guillermo Calvo really qualified for the job? Had he really thought
enough about what he was recommending? Some of his thoughts at the time
were documented in his papers ‘‘A Delicate Equilibrium: Debt Relief and Default
Penalties in an International Context’’ (1989) and ‘‘Debt Relief and Debt Resched-
uling: The Optimal-Contract Approach’’ (Calvo and Kaminsky 1991). In what fol-
lows I will comment on some of the topics on policy-making covered in those
papers as well as on some personal recollections on several policy issues, but
limiting myself just to the topic of emerging-markets sovereign debt.
Rescheduling of deposits in Argentina at the end of the eighties and beginning
of the nineties was essentially a rescheduling of government debt. Assets of com-
mercial banks were mostly government obligations, either because commercial
banks financed government spending, or because high reserve requirements (fully
The Practitioner 439

remunerated by the government) were enacted to sterilize government-related

monetary expansion. That was not the only government debt; there was also
defaulted debt on syndicated loans from international commercial banks. There-
fore rescheduling of deposits was the first step in dealing with the problem of the
debt that started in 1989 and ended with a Brady bonds exchange for old debt in
1993.
Argentina was not alone; several other countries were having a debt problem
and Guillermo Calvo observed two fundamental types of events: first, the interna-
tional debt problem that arose at the start of the eighties was pretty much around
in the early nineties, with the financial community reacting to the first signs of dis-
tress by launching a coordinated action through multilateral institutions under
the assumption that high interest rates and debtors’ term-of-trade deterioration
were only passing difficulties, which a payments’ rescheduling would help to
overcome. Second, after impressive adjustment efforts, debtor countries were
hardly able to collect the resources necessary to resume normal payment of their
debt obligations. Attempting to collect higher revenues led in some cases to nega-
tive growth and record-high inflation. He conjectured that, sooner or later, a feel-
ing would start to emerge that it could be unrealistic to expect debtor countries to
repay their debts in full.
At the end of the eighties the debt-reduction view gained momentum with the
Brady Proposal, according to which debtors and creditors were encouraged to
find mutually acceptable ways to reduce the value of contractual debt. At the
same time several market operators warned that debt reduction could become a
tug-of-war process, giving the wrong incentives to future debtors. If that were the
case, it would be very hard for new lending to resume in a significant way.
To address this difficult issue Calvo and Kaminisky tried to go behind the veil
of formal contracts and obtain information about implicit contract clauses that
might account for the possibility of debt reduction. On sovereign debt contracts,
they assumed that the premium over a risk-free interest rate was equivalent to an
insurance premium the debtor had to pay in the good states of nature in order to
cover the lender against partial repayment in bad states. They envisioned debtors
and lenders as engaging in optimal consumption-smoothing contracts, which
were not fully state-contingent only because there were ‘‘verification’’ costs (as
defined by Townsend 1979). That is, in ‘‘good’’ states of nature not subject to veri-
fication, the debtor is obligated to pay the principal and contractual interest rate,
which is given by a risk-free rate plus the prespecified interest-rate (or risk) pre-
mium. On the other hand, in ‘‘bad’’ states of nature where verification is activated,
a fixed cost is paid by the borrower, in exchange for which he is only obligated to
pay just a reduced amount of principal and interest. In this view, debt reduction
could not be rationalized as an incentive to play an incompetent macroeconomic
440 Roque B. Fernández

policy. The implicit contract allowed for rescheduling or sizable debt reductions
totally consistent with rational full precommitment. Strategic defaults were ex
ante ruled out by sizable verification costs assuring debtors’ willingness to pay.
Most policy makers, as well as high-ranking multilateral officials, thought that
modeling implicit rescheduling and default with the standard debt contract
seemed highly stylized and unrealistic. But now we could argue, for example,
that at the beginning of the nineties with a hyperinflation state of nature, resched-
uling deposits from seven days to ten years was contemplated in a sort of an im-
plicit contract. Today we have additional information we did not have at the time
of the policy action. The rescheduling was legally challenged, and the Supreme
Court of Argentina ruled in favor of the rescheduling given the prevailing (state-
of-nature) financial conditions of Argentina.
Of course it would be hard to tell that Brady bonds, or global bonds (as were
denominated later on), had implicit clauses for rescheduling or debt reduction.
On the contrary, collective action clauses allowing for rescheduling, interest, or
debt reduction were excluded in emerging-market bonds under New York gov-
erning law. But it is nevertheless true that during the nineties emerging markets
were able to issue a significant amount of debt under New York governing law
that in several cases was eventually defaulted, rescheduled, and significantly
reduced.
If debt reduction is explicitly ruled out, how can economists assume implicit
covenants for debt reduction? Is there any provision in the legal tradition or gov-
erning law of financial centers to make explicit what we consider implicit in mod-
eling optimal contracts? Again, today we can show the existence of implicit
covenants for debt reduction, and not only that, today those implicit covenants
have been accepted in the courts under the New York governing law.
Under the law of the state of New York (see Choi and Gulati 2003) there are
two fundamental groups of legal provisions that are particularly relevant in sov-
ereign debt contracts. One group provides for unanimous action clauses (UACs)
requiring approval of a hundred percent of bondholders before any haircut is
enacted. This means that just one holdout may cause efforts to modify the pay-
ment obligations of a bond to fail. The other group are generally denominated
exit consents, covering matters such as negative pledges, governing law, submis-
sions to jurisdiction, and listing provisions, that can be used to implicitly circum-
vent the unanimity requirement. Exit consents can be modified with simple
majority (51 percent), or in some instances with special majority (66 percent) of
the outstanding bonds.
To illustrate, imagine the case of outstanding bonds having UACs covering just
payments dates for principal and interest, and the sovereign in financial distress
The Practitioner 441

has managed to work out a coalition with a majority of domestic bondholders to

enact a restructuring deal. Suppose that external bondholders, either original
bondholders or in a coalition with some domestic bondholders and vultures (in-
ternal and/or external), refuse to go along with the deal because they have
expectations of a better deal. Exit consents could be used by the sovereign in coali-
tion with domestic bondholders to change old bond provisions prior to their exit
from the old bonds. Through exit consents, domestic bondholders exchange their
old bonds for the new restructured bonds, simultaneously consenting to changes
in the terms of the old bonds not covered under UAC provision. Through changes
in the governing law, listing provisions, and other terms, the exit consent proce-
dure can diminish dramatically the value of the old bonds. Changing the govern-
ing law from that of New York to that of the home country, which could be less
sympathetic to holdout behavior, would make it harder for the holdouts to sue.
Alternatively, exit consent may be used to rescind the issuer’s waiver of sovereign
immunity or its consent to jurisdiction in New York, both of which would com-
plicate the ability of holdouts to bring suit to enforce their rights under the old
bond covenant (See Fernández 2003 and Fernández and Fernández 2004). Obvi-
ously, implicitly there could be a lot more than one can imagine in legal contracts,
and Calvo was right in his theoretical assumption of an implicit debt reduction
possibility.
Either implicitly or explicitly, covenants with debt reduction are not possible in
good states of nature; debt reduction is only possible in bad states of nature. I will
not enter into the very relevant criticism that most of the time bad luck is endoge-
nous, or that bad luck is the natural outcome of wrong economic policies. I will
not enter because I mostly agree with that criticism, and, obviously, policy makers
should be very careful to apply the right economic policy to avoid endogenous
bad luck. But some times exogenous bad luck happens, and debt rescheduling or
debt reduction is an unwanted result. And this brings me to the last topic of this
presentation. Is there any other possibility to deal with true bad luck?
As the nineties went through, bad states of nature were renamed as sudden
stops. And capital flows volatility became an important item in the research
agenda of Guillermo Calvo. If someone asks anyone knowledgeable from the
media who is Guillermo Calvo, the most probable answer is ‘‘an economist that
predicted the Tequila Crisis.’’ Although being able to predict has its own merits,
sometimes it becomes the sad part of the economic practitioner. On some occa-
sions, for a professional economist, being able to predict a sudden stop is equiva-
lent to a physician being able to predict an incurable illness. The adviser would
never know if a policy maker will be able to survive if the prediction of a sudden
stop is accurate.
442 Roque B. Fernández

In a recent work, analyzing the default crisis of 2001–2002, Calvo, Izquierdo,

and Talvi (2003) pointed out that Argentina, being an economy extremely closed
to international trade, was extremely vulnerable to a sudden stop such as the one
that followed the Russian crisis. But again, in this respect Argentina was not
alone, and it is relevant to mention that sudden stops have also affected other
emerging economies producing important financial crises, as was the case in East
Asian countries in 1997.
As reported by Andrew Crockett (2002), the international experience in the af-
termath of the East Asian crisis exhibits two new features. First, East Asia has be-
come a very large net capital exporter. The external demand generating the export
surplus has undoubtedly facilitated the region’s recovery from the crisis. And
countries with current account deficits, particularly the United States, have bene-
fited from capital inflows to finance their deficits. Nonetheless, it is hard to believe
that East Asia should be an exporter of savings in the longer term, or that the U.S.
deficit is indefinitely sustainable. The second feature is that the region is exporting
safe capital while importing risky capital. This international exchange of risk is
restoring and strengthening national and corporate balance sheets in Asia and is
thereby rendering the region’s economies more resilient, and perhaps less vulner-
able to sudden stops. Still, this pattern has drawn the criticism that it has impeded
the development of East Asia’s own bond markets.
East Asian countries are small open economies from a trade account point of
view, while Argentina and other Latin American countries are small open econo-
mies from a capital account point of view. Nevertheless, I believe that exporting
safe capital and importing risky capital is a common feature to East Asia and Lat-
in America after the crisis of the nineties. Therefore it is interesting to evaluate the
situation of a post-financial crisis or post-default dynamics of capital flows in
these emerging economies.
First, one should ask the question of what is wrong in exporting safe capital
and importing risky capital? As I said, the criticism is that it impedes the develop-
ment of a domestic bond market. To some extent this is not totally true, as we
have observed that countries in open default (Argentina at the time of this confer-
ence) have managed to have new bonds with a significantly lower country risk
than old (defaulted) bonds. It is also true that there are expectations of successful
negotiations with defaulted bondholders, and that the new bonds survive thanks
to restrictions to the local financial system. But even in the case that new local
bonds could not be issued, it is not clear what is the welfare loss (if any) of
importing the services of financial intermediation. One could argue that local sav-
ers are better off importing the services of a governing law under a foreign juris-
diction until structural reforms assuring stability of property rights are enacted in
The Practitioner 443

the home country. One could also argue that nonsubsidized private local bor-
rowers could also be better off taking international loans collateralized abroad
with their own savings or export goods. In sum, one could imagine very realistic
situations where a fully open financial system with unrestricted capital account
flows is the best possible outcome, while institutional reforms are enacted to as-
sure a sound development of local financial and capital markets.
Second, one should also ask the question if there are fundamental reforms in
the international financial institutions that could help to alleviate the volatility of
capital flows to emerging economies. Should sudden stops require a fundamental
review in the explicit or implicit legal provisions governing sovereign debt con-
tracts? This is a question that has received considerable attention in later years
and there have been two confronting views. One would argue in favor of a more
active role for international financial institutions along the traditional lines of fi-
nancial liquidity assistance, and other would argue the opposite, in the sense that
international financial assistance could make things worse. For example, Dooley
(2000) has argued that liquidity assistance of the type proposed by supporters of
the idea of an ‘‘international lender of last resort’’ would not eliminate sudden
stops; on the contrary, it would promote larger sudden stops. He has persuasively
argued that models of corporate finance discussing the standard private debt con-
tract (similar, but not quite the same as Calvo and Kaminisky) are the appropriate
theoretical frameworks if sovereign immunity is accounted for. Output losses fol-
lowing default (or ‘‘verification costs’’ in Calvo terminology) are the result of a
second-best optimum of an incomplete contract and not a mistake in contract
design.
In this debate Calvo has been very careful in his advice and has not endorsed
the view of an international lender of last resort. In his work ‘‘Explaining Sudden
Stops, Growth Collapse, and BOP Crisis,’’ Calvo argued that policy makers
should aim at improving fiscal institutions to avoid endogenous bad luck, as
argued before. Lowering the fiscal deficit is highly effective in the medium term,
but could be counterproductive in the short run if it relies on higher taxes. The
key assumption is that higher taxes lower the after-tax marginal value productiv-
ity of capital, pushing the economy to a low-growth equilibrium. International fi-
nancial institutions could help to break the stalemate by offering loans for reform.
If successful, the loans will be fully repaid because fiscal reform would place the
economy on a high-growth path.
To conclude I want to express my gratitude to all the people involved in pre-
paring this conference and also to Guillermo Calvo. Over the years I benefited
not only from his technical advice, but also from his friendship and support at
times when I needed it most.
444 Roque B. Fernández

References

Calvo, Guillermo. 1989. ‘‘A Delicate Equilibrium: Debt Relief and Default Penalties in an International
Context.’’ In Analytical Issues in Debt, eds. J. A. Frenkel, M. P. Dooley and P. Wickmam. Washington,
D.C.: International Monetary Fund.
Calvo, Guillermo. 2002. ‘‘Explaining Sudden Stops, Growth Collapse, and BOP Crisis: The Case of Dis-
tortionary Output Taxes.’’ Mimeo., IADB, University of Maryland, and NBER.
Calvo, G., A. Izquierdo, and E. Talvi. 2003. ‘‘Sudden Stops, The Real Exchange Rate and Fiscal Sustain-
ability: Argentina’s Lessons.’’ Working Paper No. 9828, NBER, Cambridge, MA.
Calvo, Guillermo, and Graciela Kaminsky. 1991. ‘‘Debt Relief and Debt Rescheduling: The Optimal-
Contract Approach.’’ Journal of Development Economics 36: 5–36.
Choi, Stephen, and Mitu Gulati. 2003. ‘‘Why Lawyers Need to Take a Closer Look at Exit Consents.’’
International Financial Law Review, September: 15–18.
Crockett, Andrew. 2002. ‘‘Capital Flows in East Asia Since the Crisis.’’ Bank for International Settle-
ments. Speech given at the ASEAN Plus Three Meeting, Beijing, October 11.
Dooley, Michael. 2000. ‘‘Can Output Losses Following International Financial Crisis be Avoided?’’
Working Paper No. 7531, NBER, Cambridge, MA.
Fernández, Roque. 2003. ‘‘Crisis de Liquidez y Default Estratégico en el Servicio de la Deuda Sober-
ana.’’ Academia Nacional de Ciencias Económicas, Diciembre. Available at http://www.cema.edu.ar/
u/rbf/.

Fernández, Katherina, and Roque Fernandez. 2004. ‘‘Willingness to Pay and the Sovereign Debt Con-
tract.’’ Universidad del CEMA, March. Available at http://www.cema.edu.ar/u/rbf/.
Townsend, R. M. 1979. ‘‘Optimal Contracts and Competitive Markets with Costly State Verification.’’
Journal of Economic Theory 21: 265–293.
18 In His Own Words: An
Interview with Guillermo
Calvo

Enrique G. Mendoza

Guillermo Calvo is one of the most influential economists in the field of interna-
tional macroeconomics in the last thirty years. He has produced seminal articles
in every area of macroeconomics and international economics he has worked in,
including his early classic articles on capacity utilization and time inconsistency;
his 1980s works on efficiency wages, price stickiness, and policy credibility; and
his recent studies on sudden stops and emerging-market crises. Yet the defining
feature of Guillermo Calvo’s contribution to our profession is not the depth and
wide scope of the economic theories he has developed, but the central emphasis
he puts in all his work on the role of economics as a tool for understanding reality
and improving the quality of human life.
Guillermo Calvo’s passion for the policy implications of economic theory is ob-
vious to anyone that has met him since his days as Senior Advisor of the Research
Department of the IMF in the mid-1980s. This feature of his professional interests
was much less obvious to those who interacted with him during his early years as
an important figure of the rational expectations revolution. It was probably hard
to see that behind the highly technical treatment presented in his articles at that
time was an author who had his feet soundly set on the ground and focused on
understanding how society could benefit from the renaissance of macroeconomic
theory that was taking place. Interestingly, Michael Rothschild did figure out the
true nature of Guillermo Calvo in those early years. When the University of Cali-
fornia in San Diego tried to hire Calvo in the mid-1980s, Rothschild explained to
Calvo that he was an excellent fit for San Diego because he was a particular type
of theoretician: ‘‘Most theoreticians make theory out of theory,’’ Rothschild noted,
but Calvo was different because he made ‘‘theory out of reality, taking what is
really out there in a much wider and complex form than a well-developed but
narrow theory.’’
The following pages are excerpts from three interview sessions that Guillermo
Calvo and I had at his office in the Inter-American Development Bank in spring
2003. These interviews provide a clear picture of Calvo as the theoretician of
446 Enrique G. Mendoza

reality that Michael Rothschild described. The recollection of the conversation

with Rothschild is one of many fascinating memories of Calvo’s personal life and
professional career that emerged during our meetings. We taped and transcribed
the three meetings in Spanish, and then Calvo and I together edited that material
to produce this much shorter interview in English.

The Beginnings

EM: Guillermo, you were born and raised in Argentina. I imagine that the expe-
riences you went through growing up in turbulent Argentina played a central role
in developing your interest in economics, and in a particular class of economic
problems. Would you like to say a few words about how growing up in Argen-
tina shaped your interest in the economics profession?
GC: When I was finishing high school in Buenos Aires there was no school of
economics to speak of. However, I was very lucky because my father, who
worked at the central bank, brought me economics books from the library there,
under the advice of some of his friends who had worked with Raúl Prebisch. I
found the material strange and fascinating at the same time. One day he brought
me the General Theory and I almost decided that economics was not for me! Fortu-
nately, in school there was a course called economics. As programmed, it con-
sisted of some kind of history of economic thought, but, once again, I got lucky.
The course was taught by Julio Olivera who had a passion for Walras. As a result,
we spent the whole semester discussing the Walrasian system. I could hardly be-
lieve that a subject that appeared impossible to tackle when reading Keynes sud-
denly became so clear and elegant. One year later I joined the central bank and
worked under my teacher’s son, Julio H. G. Olivera, whose orders to me were, es-
sentially, go to the library, get a copy of Allen’s Mathematics for Economists and
Hicks’ Value and Capital and don’t leave your room until you are done with them.
I felt like I had reached nirvana! In addition, there was a seminar series in which
we discussed some key papers coming out in journals like the JPE [ Journal of Polit-
ical Economy]. One of the first such papers was Samuelson’s overlapping genera-
tions model. Quite frankly I must say that I became a little dizzy trying to read
papers on the frontier of economics, but I found it very encouraging that every-
thing we read had a solid mathematical basis. Thus, even though the deeper eco-
nomics often escaped me, I felt I had a firm grasp of the reasoning in each one of
the papers, and that, I felt, was good enough. Did growing up in such a fascinat-
ing economic laboratory, Argentina, influence my decision to go into economics?
Maybe, because before I learned economics it was very hard for me to follow the
public debate, which was heavily peppered with economic terms. But I believe
In His Own Words: Interview with Guillermo Calvo 447

that what really hooked me at the beginning was the beautiful theory, and finding
out that emotion could be subject to mathematical analysis.
EM: Let’s talk about your decision to go to graduate school. Can you describe
how early in your college years you decided that you would go for an economics
PhD abroad, and how you ended up choosing to attend Yale?
GC: I had practically no college years because, as I told you, when I started there
was no school of economics. Thus, I enrolled in accounting, which I found quite
uninteresting. However, at the University of Buenos Aires, where I attended, Pro-
fessor Julio H. G. Olivera ran a series of seminars that his best students attended,
and I was accepted because I worked with him at the central bank. (By the way,
Rolf Mantel and Miguel Sidrauski also attended those seminars). In those semi-
nars we read Hick’s Value and Capital, Samuelson’s Foundations, and Koopmans’
Three Essays on the State of Economic Science. With a group of adventuresome class-
mates I also independently read the recently published Debreu’s Theory of Value,
which required learning some basic topology. The Koopmans-Debreu duo (both
of whom were at Yale at the time) put Yale among my favorite places to go, and
made me think of graduate school, not as the next step in a professional career,
but as a door to heavenly intellectual delights! However, scarcely having one-
third of college under my belt made my chances of acceptance in graduate school
extremely slim. Moreover, my willingness to learn the intricacies of accounting
was declining at an alarming speed. Thus, there came a time when I felt that I
had fallen into a cruel trap, no exit in sight. However, once again my good fortune
gave me a hand. This time in the form of the USAID [United States Agency for In-
ternational Development], which offered scholarships in a program at Yale! This
was an MA program but the best students had a chance of being considered for
their PhD program. I was accepted on the basis of recommendation letters, and
nobody seemed to care that I was far from graduation. As I was later told, this
had been simply an oversight of the admissions committee!
EM: What about your years at Yale? Can you give us a quick review of the basic
facts (what years where you there, who were some of your classmates, who were
the leading members of the faculty), and more importantly, can you tell us about
your mentors and your dissertation, and reflect back on how the experience at
Yale matched your pre-graduate school aspirations?
GC: When I got to Yale in 1964, Debreu had already left for Berkeley, but Koop-
mans and Herb Scarf were there. Moreover, Ned Phelps taught growth theory,
and David Cass was a young assistant professor who paced around the depart-
ment like a caged animal, full of ideas and projects. In my second year, Joe Stiglitz
showed up and the place caught fire! At the same time, we had the strong pres-
ence of Tobin and several other famous names. In my particular case, the presence
448 Enrique G. Mendoza

of Carlos Dı́az-Alejandro was very important, because he was in the process of

writing his economic history of Argentina, a masterpiece. It was very good to
have someone with whom to talk plain, but deep, economics. Besides, Carlos was
a great guy who knew how difficult it is to plunge into a new culture and lan-
guage, and did his best to lighten the burden. Thus, the straight answer to your
question is that I was fully satisfied with my experience, but I had not realized
how lonely you could become as a foreigner. Classmates? Edmar Bacha was my
buddy. We struggled together to disentangle the mysteries of a language that
none of us had a good command of, a problem that became especially acute
when we had to learn about the inventions of the Industrial Revolution in an eco-
nomic history course (the Flying Shuttle? . . . we still don’t know what it is!). Ted
Truman was another classmate, but I must admit that I socialized very little (I
was already married), and there was not a lot of scientific interaction among
students.
EM: So the year is now 1967 (you were 26 years old), but you finish your PhD at
Yale at the age of 33. Why did it take you so long? What were the placement
options that you contemplated for the beginning of your career, and which one
did you take?
GC: In 1967 I went to a summer program in mathematical economics run by
Hirofumi Uzawa, thanks to Koopmans’ recommendation. That was a wonderful
three months, when I started to work on my dissertation on optimal growth in a
vintage capital model. Uzawa had a lighter touch than Koopmans, and thus
encouraged me to attack problems without, at first, being too careful about the
mathematical details. That was all very positive. However, the Vietnam War was
still raging, and campuses had become ground zero for antiwar demonstrators.
That plus my own misgivings about the Vietnam War made me feel more and
more at a disconnect with the world around me. As a result, I started to look for
options outside the U.S. That is why in 1968 I left without finishing my thesis and
headed for Lima, Peru, where the Ford Foundation sponsored a postgraduate
program at the central bank. Soon after I arrived, there was a military coup, and
in 1969 I headed off to Bogotá, Colombia, another Ford Foundation teaching post.
I stayed there until December 1972. I wrote several papers inspired by many inter-
esting issues in those countries . . . but no thesis. Remember, this is way before the
Internet, and way before mathematics became so easy to type! Thus, sending a
draft to your advisor via snail mail was a painful and not very promising en-
deavor. In a way, I had fallen into another trap. But my good fortune prevailed.
Carlos Dı́az-Alejandro heard that Columbia was looking for assistant professors
without success, and mentioned my name to his friend (and now mine) Ron
Findlay. In short, I visited Columbia and my good rapport with wonderful people
In His Own Words: Interview with Guillermo Calvo 449

like Ron Findlay and Kel Lancaster was immediate. Two or three days later I got a
telegram (no faxes, no e-mails, mind you) in Bogotá with an offer as a lecturer,
which would be transformed into assistant professor on writing my dissertation.
That I did in 1974, at the ripe age of 33! With the benefit of hindsight, however,
this long gestation period was very beneficial because I learned a lot of math to
tackle an existence proof in my dissertation. (Such proof, by the way, never saw
the light of day because it had become so technical that Koopmans thought he
would have to ask the math department for an adviser. To make sure that there
were no serious existence problems, he did some informal checking with some
mathematicians there who felt my proof was okay, which led Koopmans to give
me the green light, allowing me to assume existence). Equally important, in the
years prior to actually writing my dissertation I was exposed to a set of very rich
economic issues firsthand, which served me as a source of inspiration for many
years to come. Thanks to that detour in the real world, I realized how relevant
economics was for understanding, and occasionally solving, important problems.
At that point, my ‘‘marriage’’ to economics had become complete: beauty and rel-
evance converged!

Time Consistency and Other Early Works (The Columbia Days)

EM: I have the impression that your Columbia days were very important for
your personal and professional life. Can you describe for us the environment that
you were exposed to there and how it affected your developing career?
GC: Columbia was a crucial step in my career. I had a dream team of colleagues:
Ned Phelps, Phil Cagan, Ron Findlay, Bob Mundell, Jagdish Bhagwati, Carlos
Rodriguez, John Taylor, Stan Wellisz, Maury Obstfeld, Carlos Dı́az-Alejandro,
and more. This outstanding group of scholars provided the right atmosphere to
develop abstract ideas. You see, the real world is very good for inspiration, but
real-world people, no matter how intelligent they are, have very little patience for
theory. I am sure that if I had stayed in the real world my career would have been
very different.
EM: Together with the classic article by Kydland and Prescott, your article on
time inconsistency is credited with making one of the most substantial contribu-
tions of the rational expectations revolution to economics, both theory and policy.
Can you tell us about the gestation process of the idea itself and also about the
process leading to the publication of the paper?
GC: My friend Assaf Razin claims that time inconsistency was an obsession with
me as far back as 1967 (during a seminar organized by Uzawa in Chicago that
Assaf also attended). In fact, I used to tease my younger siblings when we were
450 Enrique G. Mendoza

children with time-inconsistency tricks, so Assaf is likely to be right (although I,

quite honestly, do not recall. Was I so sickly obsessive that I did not notice?). In
any case, soon after I arrived at Columbia, and prompted by Ned Phelps, I be-
came interested in the theory of justice (Utilitarians, Rawls, and all that). I wrote
a paper on the time inconsistency of Rawlsian maximin principle, for example.
Simultaneously, I spent many hours with Carlos Rodriguez trying to disentangle
the implications of Auernheimer’s recently published JPE paper on an honest gov-
ernment rule for money creation where, it turns out, one can find the seeds of
macro time inconsistency. First I wrote an open-economy time inconsistency ex-
ample, which was published in the Journal of Money, Credit and Banking [ JMCB],
and later the better known closed-economy example published in Econometrica.
Motivation for writing the latter I owe to Alvin Marty who, learning about the
open-economy paper, told me that if I wanted to reach mainstream macroecono-
mists, I should make the case in terms of a closed-economy model. He was cer-
tainly right. It shows, incidentally, how U.S.-oriented macro was at the time (and
still is in some places). In any case, when I was working on macro time inconsis-
tency, Ed Prescott showed up at Columbia to give one of his seminal papers on
the subject, but I was out of town. When I came back, I asked a senior colleague
about Ed’s paper, and he said that he could not quite figure out. Thus, I placed
the paper on my to-read list, and did not read it until, during the publication pro-
cess, the referee (who turned out to be Ed, on his own recognition) pointed an
accusing finger at my oversight. This I quickly fixed, the paper was published,
and the rest is history!
EM: It is interesting that at times a seminal idea in economics leads to interpreta-
tions that are in conflict with those of the originator of the idea, particularly when
it comes to economic policy. For example, Mundell’s arguments on currency areas
and exchange rates have been and are still used by a large number of economists
to favor flexible exchange rates, while Mundell himself sees the same arguments
as a strong argument in favor of fixed exchange rates. Do you feel that some-
thing similar happened with the notion of time inconsistency? To be more precise,
what is your own view on the key policy lessons one should take from time
inconsistency?
GC: It shows that a policy maker could be a saint and still be a damned liar dur-
ing the election campaign. In other words, time inconsistency cannot be simply
dismissed as a character failure, or a political party failure, which the electorate
should be able to weed out. Moreover, the examples show that the negative effects
of time inconsistency could be reduced, and even eliminated, by devising new
policy instruments or institutions. Hence, there are potentially important lessons
for policy and institutions. I believe that the independence of central bank issue
In His Own Words: Interview with Guillermo Calvo 451

got a big boost from this literature. I would say the same thing about the political
economy literature.
EM: On the theoretical side, I find it very interesting that while some researchers
went on to study time-consistent policy in dynamic games, a large chunk of the
research you did in the second half of the 1980s focused more on the macroeco-
nomic consequences of having to live with time-inconsistent policies (or with the
‘‘lack of credibility’’ of policies, as you referred to it). This use of the word ‘‘credi-
bility’’ is sometimes questioned because the credibility literature that your work
started is not based on time-consistent dynamic games. Can you tell us why your
interests flowed in a different direction, and what your opinion is of the proper
use of the word ‘‘credibility?’’
GC: Good question! I believe there is such a thing as premature formalization,
and also believe something like that happened with time inconsistency. Once the
profession grabbed hold of time inconsistency, researchers started to play games
with it, games not necessarily inspired by the real world or policy relevance, but
games mostly inspired by the need to become noticed and be published. Don’t
take this as a criticism of the literature. I am a firm believer in pure research. But
volume does not make relevance, or even indicate the priorities set by the profes-
sion, especially the seriously committed side of the profession. Speaking of my-
self, after finishing my papers on time inconsistency, I felt that I wanted to go
back to simpler grounds where I could better understand the dynamics implied
by time inconsistency. Thus, knowing that in a full-fledged model I could gener-
ate time inconsistency, I decided to simplify that part by inserting phenomena
that resembled time inconsistency in an exogenous manner and focus more
sharply on their dynamic implications. That is partly why my policy temporari-
ness papers came to life. But another important reason was that I was trying to ex-
plain why orthodox stabilization programs in Latin America in the late 1970s and
early 1980s did not work out. I was afraid that public opinion would swing to a
crazy heterodox extreme and all fiscal discipline would be thrown out the win-
dow. The credibility papers helped to show that when credibility is imperfect,
even good orthodox policies could lead to unsatisfactory results.
Let me add that my sense is that time inconsistency is not an everyday problem
but something that policy makers are drawn to in extreme circumstances. Leo Lei-
derman and I, for instance, showed in an AEA [American Economic Association]
paper that one could rule out time inconsistency from the monetary policy of sev-
eral developing countries. We did that by showing that the first-order restrictions
of time-consistent optimal policy could not be rejected. This suggests that time in-
consistency as a daily phenomenon may not be relevant, but does not rule out
that one can find it once in a while—and in a big way! I don’t think the formal
452 Enrique G. Mendoza

literature has explored this avenue, with the exception of a few papers, like one
by Bob Flood and Peter Isard (‘‘Monetary Policy Strategies,’’ IMF Staff Papers 37:
612–632, 1989), which assumes the existence of a fixed cost in exploiting time
inconsistency.
EM: There were other important areas you worked on during the early part of
your professional career. From my own work, I remember in particular your clas-
sic AER [American Economic Review] paper on capacity utilization. Can you tell us
more about this paper and some of the other work that you were involved with in
the 1970s?
GC: This paper was inspired by the debate in Colombia about capacity utiliza-
tion. Several empirical studies suggested that capacity utilization was very low
there, and I wanted to check what basic theory would say about it. However, if
anything, my paper thickened the plot because the model implies that capacity
utilization should increase with the real exchange rate. Thus, if underdevelop-
ment goes hand in hand with high interest rates (which was the conventional wis-
dom in that world of low capital mobility), then Colombia should rather display
high capacity utilization.
On the other work, I already mentioned the theory of justice, but what really
absorbed my attention in the early 1970s was the theory of supervision. I wrote
several of those papers on the subject with Stan Wellisz. Stan, by the way, is one
of my admired figures at Columbia University, and at the time greatly helped me
to put relevance into my math scribbling. The focus in these papers was to explain
hierarchical ladders in an organization, and structural unemployment. Based on
the imperfect supervision paradigm, I wrote one of the first papers on what was
later called efficiency wage hypothesis to explain unemployment. However, I
eventually stopped working on this field because I felt that the next necessary
step was empirical analysis at the micro level. At the time, I could not find a part-
ner to help me plunge into that uncharted territory, and I felt that doing it by my-
self would have distracted me too much from the macro issues that still captured
my imagination.

Sticky Prices and Noncredible Policy (Penn and the IMF Transition)

EM: To continue on the track of your professional career, you moved from Co-
lumbia to Penn in 1986. What were the motivations for your move? Can you sum-
marize for us your experience at Penn?
GC: Maury Obstfeld and I left for Penn at the same time. The department of eco-
nomics at Columbia was in a shambles; young assistant professors like Maury
In His Own Words: Interview with Guillermo Calvo 453

were overburdened with thesis projects and got little administrative support.
There were fights among the faculty, some of which became public and made the
department a butt of jokes in the profession. Leaving was a very hard decision for
my wife and me because we both loved New York and had many good col-
leagues at Columbia. My stay at Penn was pleasant but short, however. Along
came an attractive offer to visit the IMF Research Department, and later a perma-
nent offer, right after the collapse of the Berlin Wall. Being at the Fund meant hav-
ing a front seat to observe this colossal drama.
EM: Your 1983 JME [ Journal of Monetary Economics] article on staggered prices is
one of the most widely cited articles in the recent literature on general equilibrium
models with nominal rigidities. Readers will be very interested in learning what
motivated you to develop the ideas you proposed in this paper.
GC: Ned Phelps and John Taylor had made big strides on that kind of model for
several years (as I recall they were already hard at work on these issues in 1973).
However, I did not get very interested in it, because I was trying to understand
high-inflation countries where, I thought, price stickiness should not be a big
issue. All of that changed in 1981 when Argentina abandoned its ‘‘Tablita’’ stabili-
zation plan and underwent a series of large devaluations. To my surprise, unem-
ployment took a big jump, and the real exchange rate suffered a sizable increase. I
could not understand this in terms of flexible-price models, which led me to pay
more attention to the Phelps-Taylor approach. (Let me add that nowadays I
would first turn my attention to imperfections in the credit market, rather than
price stickiness, but that was the early 1980s and the conventional wisdom in
theory circles—although not necessarily among development economists—was
that the capital market was not a source of problems in developing countries).
Since, once again, my main interest was to understand the dynamics of imperfect
credibility, I made simplifying assumptions in order to be able to formulate the
sticky-prices model in continuous time (where one can use phase diagrams) and
start from utility functions. The latter was important for my purpose, because
credibility analysis becomes too ad hoc if you start from demand functions. Let
me add a vignette. I started working on this approach when I visited Chicago
during the spring of 1981. However, I never made a presentation on it in the
Money Workshop, where price stickiness was not part of the conventional wis-
dom, to put it mildly. I guess I was afraid of being butchered on the spot! You
can imagine my surprise when much later I saw mainstream macroeconomists
using the framework as a matter of fact. Let me add, that the paper you are referring
to is a closed-economy macro paper that I wrote after writing the open-economy
version motivated by Argentina’s experience. I wonder what would have hap-
pened to my citation count if I had stopped at the open-economy paper!
454 Enrique G. Mendoza

EM: What is your opinion of the recent literature on dynamic general equilib-
rium models with nominal rigidities?
GC: It is motivated by the failure of pure real business cycle models, and it is a
natural development. I have contributed to it also with Oya Celasun and Michael
Kumhof (trying to account for inflation inertia). Like cash-in-advance, price-
stickiness models fill a vacuum in general equilibrium theory without which one
cannot even begin to address some basic policy issues in monetary economics.
Unfortunately, the microfoundations are still weak.
EM: We arrive at the mid-1980s. This is the time when two of your most famous
papers on the macrodynamics of the imperfect credibility of economic policy
appeared, one in the JPE in 1986 and one in the JMCB in 1987. The JPE paper
aimed to explain the consumption booms associated with failed disinflation pro-
grams based on exchange-rate management, and the JMCB paper integrated this
idea together with a Krugman-style model of currency crisis in an intertemporal
optimization framework. Together with the paper that Helpman and Razin pub-
lished in the AER around the same time (approaching the issue from a different
perspective), these papers were the originators of the large literature on the real
effects of exchange-rate-based stabilizations. Can you describe what sparked your
interest to begin writing about this particular issue?
GC: Once again, Argentina’s devaluation showed to me very clearly that nomi-
nal devaluations could result in real devaluation, and all kinds of important real
effects could follow. That was the motivation. In my work, real effects followed
from lack of full credibility, and that is how, in a way, I was tying up this line of
research with time inconsistency. I insisted on bringing in utility and production
functions because, once again, it is very hard to analyze credibility problems start-
ing from ad hoc demand functions.
EM: I would like to switch now for a moment to your transition from Penn to
the Research Department of the IMF, which took place in 1987–1988. I wonder,
what motivated you to consider leaving academia for the IMF? Can you describe
the duties you performed at the Fund and give us your impressions about how
this experience influenced your research?
GC: I first went to the Fund as a visiting scholar, and later was offered the posi-
tion left vacant by the departure of Max Corden. The head of research was Jacob
Frankel, and he gave me full rein to concentrate on research and visit the field as
much as I deemed necessary. The unexpected bonus was guys like you, Carlos
Végh, Pablo Guidotti, and Carmen Reinhart, with whom I did many research
projects. It was hard to go back to academia under those circumstances, when it
appeared that I had been granted a NSF [National Science Foundation fellowship]
In His Own Words: Interview with Guillermo Calvo 455

for life, an airline ticket that would never expire to see the world after the collapse
of the Berlin Wall, and an army of first-rate collaborators!
EM: You and I met at the IMF when I arrived there in 1989. My recollection, in
line with what you just said, is that at that time you were the head of an active
group of researchers in international macroeconomics. I recall two important re-
search programs in which you were involved. One of them was your joint work
with Carlos Végh, in which you were pursuing further exchange-rate-based stabi-
lization models to analyze the real exchange rate and introduce your staggered
pricing setup; the other was your solo work and your joint work with Pablo on
self-fulfilling expectations models of public debt. Can you give us a short over-
view of these two research programs?
GC: Carlos Végh gave me the extra kick I needed to bring my open-economy
price-stickiness models to full fruition. He brought a lot of enthusiasm and cre-
ativity to this endeavor, and I am glad he did because the final product could be
called Mundell-Fleming Mark II. With Pablo we worked on the optimal currency
denomination and term structure of public debt. This stemmed partly from my
AEA 1988 paper on public debt (Servicing the Public Debt: The Role of Expectations,
647–671), which, in turn, was inspired by the repeated failures of Brazil and Ar-
gentina to lower inflation to reasonable levels. The original conjecture was that
peso-denominated debt was behind this kind of problem because lack of credibil-
ity kept nominal interest high, resulting in unsustainable fiscal deficit (if inflation
had actually been lowered). This research led us naturally to explore dollarization
and term structure. Interestingly, a solution to the original high-interest problem
was dollarization of public debt, while the current evidence and research makes
one seriously question such a dollarization policy. Missing from those papers,
however, is the sudden stop phenomenon that has been prominent in recent fi-
nancial crises, and leads one to take dollarization with a great deal more caution.
EM: At the beginning of the 1990s the nature of the problems affecting interna-
tional capital markets was changing dramatically. We went from the debt crisis
and the problems of disinflation to the bonanza of the first half of the 1990s and
the surge of inflows of foreign capital into what would be called later the ‘‘emerg-
ing markets.’’ Amid the chorus of optimistic voices praising this phenomenon as
an outcome of the painful years of stabilization and reform in developing coun-
tries, a paper that you coauthored with Leo Leiderman and Carmen Reinhart
argued that this was probably not the case. Would you like to elaborate on the
details of your argument?
GC: This line of research surged almost fortuitously from several trips I took
around Latin America for the Fund at the beginning of the 1990s. In country after
456 Enrique G. Mendoza

country that I visited people talked about capital inflows and explained them by
some change in domestic policy. The conventional wisdom at the Fund also
attached primary responsibility to domestic policy, particularly the fact that sev-
eral countries were beginning to undertake structural reform, and the implemen-
tation of the Brady Plan. I became suspicious about this line of explanation as I
realized that capital inflows were taking place in countries that were pursuing
very different policies and, besides, that some beneficiaries of those inflows were
still dealing with serious domestic security issues (for example, the Shining Path
in Peru). Thus, I conjectured that there must be a common factor in all of these
episodes, and that it was likely to lie outside the region; for example, the U.S. Car-
men Reinhart (who was working at the Fund’s Research Department at the time),
Leo Leiderman (who was a visiting scholar there), and I joined forces to explore
this conjecture and, sure enough, we found strong evidence of the relevance of ex-
ternal factors. Nowadays several other papers have replicated our results but, at
the time, we got a lot of flak, particularly at the Fund where the prevailing view
was that if a country did its homework, the capital market would reward it with
stable capital inflows. Our concern was that the flows might be reflecting low in-
terest rates in the U.S., which could be reversed very rapidly (like in the early
1980s) and cause financial chaos. Unfortunately, we were right. The Tequila crisis
in Mexico 1994–1995 took place after U.S. interest rates started to rise.
EM: While at the IMF you also worked on the issues related to the transition
economies moving out of socialism. In your opinion, what were the main chal-
lenges that these economies were facing? If you had to grade the contribution
that the West and the international financial organizations made to the process of
transition, what grade would you give it? Did you have the impression that the
tools of economic theory that we had at our disposal were useful in this context,
or did you find yourself looking for a new toolbox?
GC: Transition economies provided us a colossal experiment where everything,
all aspects of the economy, had to be thought out at the same time. The policy
issues were at the polar opposite of fine-tuning. Even new institutions had to be
established. My impression, however, is that the Fund treated some of these cases
as if they were economies with well-running capitalist institutions. Thus, for ex-
ample, bank credit was drastically curtailed without paying much attention to
the resulting credit crunch, given that banks in those countries were, by far, one
of the main credit suppliers (coupled with inter-enterprise credit). This led Fabri-
zio Coricelli and me to write a paper on Poland, and a couple of other papers on
inter-enterprise credit (the main alternative to bank credit) in transition econo-
mies. But these pieces had little impact. Fortunately, several transition economies
have grown new institutions, and whether or not we were right is now a largely
irrelevant issue.
In His Own Words: Interview with Guillermo Calvo 457

EM: Now let’s talk about your own transition from the IMF to the University of
Maryland in 1994. What motivated your departure from the Fund, and what fac-
tors made you decide to move to the University of Maryland?
GC: The atmosphere at the Fund had deteriorated since Jacob Frenkel’s depar-
ture. This was partly due to losing his protective umbrella, but also to a phenome-
non that is quite common in bureaucratic institutions whose main focus is not
research: an occasional tendency to devote more resources to the main tasks of
the organization, and away from research. Besides, the offer at Maryland was
very tempting.

The New World after the Mexican Crash of 1994 (Maryland and the IDB)

EM: Early in 1994 you prepared some comments for a Brookings Papers article
that the late Rudi Dornbusch coauthored with Alex Werner on Mexico’s slow
growth performance despite its impressive reform record. Your comments went
on a different track and argued that there were causes of grave concern over the
possibility of a balance-of-payments crisis in Mexico because of large and growing
financial imbalances in the Mexican economy. These comments became the seeds
of your well-known paper ‘‘Varieties of Capital Market Crises’’ and of our two
1996 joint papers on the Mexican crisis (one published in AER Papers & Proceed-
ings and the other in the JIE [ Journal of International Economics]). The comments
also led to the New York Times article on your work under the heading ‘‘The
Prophet of Doom That Was Right.’’ Can you describe the key points of your argu-
ment and the thought process that led you in this direction?
GC: Rudi and Alex seemed to believe that Mexico could be put on the right track
by devaluing 15 percent, and I thought they were leaving out of the picture a key
ingredient: financial stocks. In particular, I feared that devaluation would breed
distrust in the minds of investors and lead to a run, which is what happened.
When it happened I was so excited that I worked around the clock on the varieties
paper, and a first version was ready in about three days. The NYT note brought
me a level of notoriety that I had never known, expected, or looked for. But I can-
not deny that it was fun—and profitable! In Argentina when I get an interview,
they still refer to me as the guy who anticipated the Tequila. This is funny, of
course. But even funnier is the fact that although I may have anticipated the Mex-
ican crisis, I did not anticipate the Tequila—because what distinguishes the
Tequila from a garden-variety balance-of-payments crisis is the contagion that
the Mexican crisis provoked around the globe.
EM: In our 1996 papers, we argued strongly that Mexico’s crisis was not an iso-
lated phenomenon and instead ought to be seen as the first case of a new breed of
458 Enrique G. Mendoza

global capital-markets crises. We wrote that these crises could occur even if the
‘‘observable’’ fiscal accounts and other so-called fundamentals were in good
shape, and that instead phenomena like the anticipation of banking crises and fi-
nancial contagion in international capital markets could play a central role. My
recollection is that authors like Tim Kehoe and Hal Cole or Andres Velasco and
Roberto Chang shared our views, but for the majority of the profession it took the
Asian crisis of 1997 and the Russian debacle of 1998 to accept this notion. With the
large number of crises that have now occurred, do you think that the facts vali-
dated the views we expressed in 1996?
GC: Yes. Capital markets are incomplete, especially so in emerging markets. In
that context, capital inflow episodes do not necessarily generate efficient solutions
but at least do not generate catastrophes while inflows last. As credit increases on
average, borrowing from a new source can easily offset a cut in a line of credit
from another. Thus, under those conditions it is unlikely that serious systemic
problems would develop. On the other hand, as capital flows dry up or are
reversed, there are few efficient ways to deal with the situation. For the private
sector we have bankruptcy regulations, which may be efficient for big firms, but
create havoc everywhere else. Moreover, there is no equivalent to Chapter 11 for
sovereign countries. Thus, there is a strong element of self-fulfilling expectations
when capital goes out—which need not be correlated with traditional fundamen-
tals like fiscal deficits. For the capital outflow to cause damage, the country must
display significant debt levels. This explains why all of these crises have followed
periods of significant capital inflows.
EM: It appears that, compared to the 1990s, we have now entered a very differ-
ent period in global capital markets in which non-FDI [foreign direct investment]
inflows into emerging markets have all but dried up. What lessons should we
draw from the crises of emerging markets? For example, do you think developing
countries should try to manage or control capital inflows? What role should inter-
national financial institutions play in the new era of globalized capital markets?
GC: Controls on capital inflows should be a policy of last resort. I believe there is
a role for the G7 here, a ‘‘traffic control’’ role, in which a new institution or facility
is created to ensure that the price of emerging market debt is not subject to large
swings, one way or the other. In a recent paper in Economia, the journal of the
Latin American and Caribbean Economic Association (LACEA), I called the new
institution EMF, Emerging Market Fund. In this fashion, there will be some con-
trol at a global level, helping to prevent global systemic problems, but capital
would still flow freely across emerging markets. The alternative of each country
establishing its own regulation could give rise to problems akin to competitive
devaluations. The end result could be a worldwide investment allocation that
In His Own Words: Interview with Guillermo Calvo 459

could hardly be called efficient. Besides, without international cooperation, it is

very hard for an individual country to establish effective controls on capital flows.
EM: From the standpoint of economic theory, explaining the phenomenon that
you labeled the sudden stop (a sudden, large reversal of the current account and
capital inflows coupled with Great-Depression-size collapses in domestic produc-
tion, absorption, and relative prices) remains a major challenge. We published a
short article in the 2000 AER Papers & Proceedings suggesting some lines of re-
search based on various forms of informational and financial-market imperfec-
tions that seemed promising venues to account for sudden stops. In your opinion,
how far are we in developing a sound theory of sudden stops, and are you confi-
dent at this point that we can use this theory to propose policy measures that can
be useful to prevent sudden stops?
GC: Developing a sound theory of sudden stops requires making assumptions
on the nature of market imperfections. Thus, it will be hard, if not impossible, to
develop a general theory of sudden stops. Therefore, the goal I have set for myself
is to develop a set of examples (a bit what I did in my varieties paper), continue to
look closely at emerging markets, and conduct some econometric testing.
Unfortunately, the typical theorist in the profession, who feeds himself on theory,
will likely have a tendency to try to find explanations in complete-markets theory,
because developing that theory can be done without crossing Rio Grande. Thus, if
crises cease to be as frequent as they have been in the last decade or so, I am
afraid that the profession will tend to commit sudden stop theory based on mar-
ket imperfections to the trash can. I said ‘‘I am afraid’’ because, as a result, future
policy makers (including the IMF) are likely to make the same mistakes they
made in the 1990s.
EM: In 2001 you took a two-year leave of absence from your Maryland position
to take on the job of Chief Economist and head of the newly created Research De-
partment of the Inter-American Development Bank. Can you tell us, what lessons
have you learned from your experience as Chief Economist and head of research
of an international financial organization?
GC: What is exciting about RES at IDB is that it represents the only policy-
oriented research unit in all of Latin America that has a serious regional and
global view. Moreover, through a series of networks, RES is in close contact with
key policy makers in the region. I believe this kind of activity is very productive
because, to the extent that countries choose the globalization route, they have
to engage in serious, professional dialogue with other countries pursuing the
same goals at a regional and global level. Moreover, providing countries with a
global view is very useful because in a globalized environment countries are
highly vulnerable to external shocks. In the absence of a global view, typically
460 Enrique G. Mendoza

the domestic-policy debate tends to ignore global factors and focus mostly on
domestic-issues. Politicians love grabbing each other’s throats!

Looking Ahead

EM: In this last section of the interview, I want to ask you some frequently
asked questions about the economics profession and its future. In your opinion,
which are the most influential contributions to macroeconomics and international
economics over the last thirty years that you have been an active researcher in
these fields?
GC: I, for one, benefited greatly from the so-called rational expectations revo-
lution. The reason is that it gave me a framework in which I could tackle
credibility-type issues. Prior to that, one has to recall that macroeconomists
thought that monetary theory could not be formulated under rational expecta-
tions. In his famous and very influential hyperinflation paper, for example, Cagan
shows that as the economy converges to RE (‘‘perfect foresight’’ in his paper
because uncertainty was not explicitly modeled), the steady-state equilibrium
became unstable. In particular, the model could display hyperinflation with a
constant money supply! As a result, until the early 1970s, most papers assumed
adaptive expectations with sufficiently long lags. It could be argued that AE
reflects some kind of lack of credibility—people don’t pay attention to policy
announcements, and prefer to extrapolate from the past. But aside from the fact
that AE would be just one particular form of lack of credibility, there is the funda-
mental issue that one simply would not know the implications of full credibility,
because the latter would imply rational expectations! On the other hand, in the
field of international finance, I believe the work of Fleming and Mundell was
very important because it showed the macroeconomic relevance of the capital ac-
count of the balance of payments for small economies. In addition, it helped to ex-
tricate the field from its obsession with not looking outside the womb, and only
focusing on closed-economy models.
EM: If I ask you to name one or two economists that you consider your favor-
ites, or the ones that have influenced your own work the most, who would you
name and why?
GC: Here is my short list: Mundell, Phelps, Lucas-Sargent-Wallace. I already
explained why Mundell is on the list. Phelps was my teacher at Yale, and I found
him fascinating. He really helped me grow when we were colleagues at Colum-
bia. The last trio I have also implicitly alluded to before. The three of them, how-
ever, have not only shown that rational expectations is an interesting hypothesis
but have done fundamental research that helped to establish RE as a valid scien-
In His Own Words: Interview with Guillermo Calvo 461

tific discipline. By doing that, incidentally, they also opened the door for eventu-
ally rejecting the RE hypothesis.
EM: In terms of the social value of our profession, what do you think have been
the most important contributions of the fields you work on to improving social
welfare?
GC: It is very hard to tell. I sometimes think that countries learn from their own
mistakes (and sometimes not even that!), and there is little an economist can do to
help. Having said that, however, I believe the main contribution of macro has
been to teach basic issues like the price level has a lot to do with monetary vari-
ables; expectations are key and, thus, for policy effectiveness, credibility is central;
external factors are important; the financial sector plays a critical role for macro
stability, and financial vulnerabilities are behind major catastrophes in recent
emerging market crises, and so on.
EM: If you had a genie that offered to give you the answer to two unresolved
economic problems, which ones would you choose? Or to put it differently, in
which areas do you think we should direct our efforts as researchers in the next
few years?
GC: The two areas where we have made some progress but I feel there is a lot
ahead of us are growth and poverty alleviation. Maybe triggering growth will
take care of the latter, but poverty alleviation is so important that we should not
wait any longer and should put lots of research effort in that direction.
EM: Finally, and since you are one of the masters of time inconsistency, if you
were given the chance to start over from any of the different periods of your ca-
reer that we covered here, would you reoptimize and move in a different direction
than you did?
GC: Not really. I am glad that I have straddled academia and policy circles. I
have learned a lot and, above all, I’ve had a lot of fun!
EM: Thanks a lot for this fascinating review of your career.

Reprinted from ‘‘Toward an Economic Theory of Reality: An Interview with Guillermo A. Calvo,’’ by
Enrique G. Mendoza. Macroeconomic Dynamics 9, 1 (February 2005): 123–145.
 Publications of Guillermo A.
Calvo

Books

Debt, Stabilization, and Development. With R. Findlay, P. Kouri, and J. B. de Macedo. 1989. Oxford: Basil
Blackwell.
Eastern Europe in Transition: From Recession to Growth? Proceedings of a Conference on the Macroeconomics
of Adjustment. With M. I. Blejer, F. Coricelli, and A. H. Gelb. 1993. Washington, D.C.: World Bank.
Money, Exchange, Rates and Output. 1996. Cambridge, MA: MIT Press.
Private Capital Flows to Emerging Markets After the Mexican Crises. 1996. Washington, D.C.: Institute for
International Economics.
The Debt Burden and its Consequences for Monetary Policy. Proceedings of a Conference held by the Inter-
national Economic Association at the Deutsche Bundesbank. 1998. London: Macmillan.
Money, Capital Mobility, and Trade: Essays in Honor of Robert Mundell. Edited with Rudi Dornbusch and
Maurice Obstfeld. 2001. Cambridge, MA: MIT Press.
Emerging Capital Markets in Turmoil: Bad Luck or Bad Policy? 2005. Cambridge, MA: MIT Press.

Articles

‘‘The Mirage of Exchange Rate Regimes for Emerging Markets Countries.’’ With F. Mishkin. Journal
of Economic Perspectives, 2003. Also in Emerging Capital Markets in Turmoil: Bad Luck or Bad Policy? ed.
G. Calvo. Cambridge, MA: MIT Press 2005.
‘‘Sudden Stops, the Real Exchange Rate and Fiscal Sustainability: Argentina’s Lessons.’’ With A.
Izquierdo and E. Talvi. In Monetary Unions and Hard Pegs, eds. V. Alexander, J. Mélitz, G. M. von
Furstenberg. Oxford: Oxford University Press, 2003. Also in Emerging Capital Markets in Turmoil: Bad
Luck or Bad Policy? ed. G. Calvo. Cambridge, MA: MIT Press, 2005.
‘‘Relative Price Volatility under Sudden Stops: The Relevance of Balance-Sheet Effects’’ With A.
Izquierdo and R. Loo Kung. Journal of International Economics 69, no. 1, June 2006.

‘‘The resolution of global imbalances: Soft landing in the North, sudden stop in emerging markets?’’
With E. Talvi. Journal of Policy Modeling 28, 2006.
‘‘Sudden Stops and Phoenix Miracles in Emerging Markets.’’ With A. Izquierdo and E. Talvi. American
Economic Review 96, no. 2, May 2006.
464 Publications of Guillermo A. Calvo

‘‘Inflation Inertia and Credible Disinflation—The Open Economy Case.’’ With O. Celassum and M.
Kumhof. Journal of International Economics, forthcoming.
‘‘Efficient and Optimal Utilization of Capital Services.’’ American Economic Review 65, no. 1, March 1975.

‘‘Optimal Growth in a Putty-Clay Model.’’ Econometrica 44, no. 5, September 1976.

‘‘Optimal Maximin Accumulation with Uncertain Future Technology.’’ Econometrica 45, no. 2, March
1977.
‘‘A Model of Exchange Rate Determination under Currency Substitution and Rational Expectations.’’
With C. A. Rodriguez. Journal of Political Economy 85, no. 3, April 1977.

‘‘The Stability of Models of Money and Perfect Foresight: A Comment.’’ Econometrica 85, no. 5, October
1977.
‘‘Some Notes on Time Inconsistency and Rawl’s Maximin Criterion.’’ Review of Economic Studies 45, no.
1, February 1978.
‘‘Urban Unemployment and Wage Determination in LDCs: Trade Unions in the Harris-Todaro Model.’’
International Economic Review 19, February 1978.
‘‘Supervision, Loss of Control and the Optimum Size of the Firm.’’ With S. Wellisz. Journal of Political
Economy 86, October 1978.
‘‘On the Optimal Acquisition of Foreign Capital through Investment of Oil Export Revenues.’’ With R.
Findlay. Journal of International Economics 8, no. 4, November 1978.
‘‘On the Indeterminacy of Interest Rates and Wages with Perfect Foresight.’’ Journal of Economic Theory
19, no. 2, December 1978.
‘‘Ex-post Behavior of Firms Offering Optimal Employment Contracts.’’ Economic Letters 1, no. 3, 1978.
‘‘Optimal Seigniorage from Money Creation—An Analysis in Terms of the Optimum Balance of Pay-
ments Deficit Problem.’’ Journal of Monetary Economics 4, 1978.
‘‘On Models of Money and Perfect Foresight.’’ International Economic Review 20, no. 1, February 1979.
‘‘Optimal Population and Capital Over Time: The Maximin Perspective.’’ Review of Economic Studies 46,
no. 1, February 1979.
‘‘Fiscal Policy, Welfare and Employment with Perfect Foresight.’’ Journal of Macroeconomics 1, Spring
1979.
‘‘Quasi-Walrasian Theories of Unemployment.’’ American Economic Review 69, no. 2, May 1979.
‘‘The Incidence of a Tax on Pure Rent: A New (?) Reason for an Old Answer.’’ With L. J. Kotlikoff and
C. A. Rodriguez. Journal of Political Economy 87, August 1979.
‘‘Hierarchy, Ability and Income Distribution.’’ With S. Wellisz. Journal of Political Economy 87, no. 5,
October 1979.
‘‘Employment-Contingent Wage Contracts.’’ With E. S. Phelps. Journal of Monetary Economics, Carnegie-
Rochester Conference Series no. 5, 1977. Reprinted in Studies in Macroeconomic Theory: Employment and In-
flation, ed. E. S. Phelps. New York: Academic Press, 1979.

‘‘Capital Accumulation and Welfare: A Note.’’ Economic Letters 4, no. 2, 1979.

‘‘Tax-Financed Government Spending in a Neoclassical Model with Sticky Wages and Rational Expec-
tations.’’ Journal of Economic Dynamics and Control 2, no. 1, February 1980.
Publications of Guillermo A. Calvo 465

‘‘Apertura Financiera, Paridad Móvil y Tipo de Cambio Real’’ (Financial Opening, Flexible Parity and
the Real Exchange Rate). Ensayos Económicos, Central Bank of Argentina, December 1980.
‘‘Technology, Entrepreneurs and Firm Size.’’ Quarterly Journal of Economics 95, no. 4, December 1980.

‘‘Capitalización de las Reservas y Tipo Real de Cambio.’’ (Capitalization of Reserves and Real Ex-
change Rate). Cuadernos de Economia, Chile, April 1981.

‘‘Devaluation: Levels vs. Rates.’’ Journal of International Economics 11, 1981.

‘‘On the Time Consistency of Optimal Policy in a Monetary Economy.’’ Econometrica 46, no. 6, Novem-
ber 1978. Reprinted in Rational Expectations and Econometric Practice, eds. R. E. Lucas Jr. and T. J. Sar-
gent. Minneapolis: University of Minnesota Press, 1981.
‘‘Involuntary Unemployment and Inventories: An Exploratory Model of Equilibrium and Pure Compe-
tition.’’ Journal of Macroeconomics 4, no. 3, Summer 1982.
‘‘Growth and Inflationary Finance: Variations on a Mundellian Theme.’’ With D. P. Peel. Journal of Polit-
ical Economy 91, no. 5, September 1983.
‘‘A Model of Non-Walrasian General Equilibrium: Its Pareto Inoptimality and Pareto Improvement.’’
With E. S. Phelps. In Macroeconomics, Prices and Quantities, ed. J. Tobin. Washington D.C.: The Brook-
ings Institution, 1983.
‘‘International Factor Mobility and National Advantage.’’ With S. Wellisz, Journal of International Eco-
nomics 14, 1983.

‘‘Staggered Contracts and Exchange Rate Policy.’’ In Exchange Rates and International Macroeconomics,
ed. J. A. Frenkel. Chicago: University of Chicago Press, 1983.

‘‘Staggered Contracts in a Utility-Maximizing Framework.’’ Journal of Monetary Economics 12, Septem-

ber 1983.
‘‘Trying to Stabilize: Some Theoretical Reflections Based on the Case of Argentina.’’ In Financial Policies
and the World Capital Market: The Problem of Latin American Countries, eds. P. Aspe A., R. Dornbusch, and
M. Obstfeld. Chicago: University of Chicago Press, 1983.
‘‘Demand Fluctuations, Inventories and Capacity Utilization.’’ With F. E. Thoumi, Southern Economic
Journal 50, no. 3, January 1984.
‘‘Macroeconomic Implications of the Government Budget: Some Basic Considerations.’’ Journal of Mone-
tary Economics 15, no. 1, January 1985.
‘‘Reserves and the Managed Float: A Search for the Essentials.’’ Journal of International Money and Fi-
nance 4, no. 1, March 1985.
‘‘The Inefficiency of Unemployment: The Supervision Perspective.’’ Quarterly Journal of Economics 100,
no. 2, May 1985.

‘‘Currency Substitution and the Real Exchange Rate: The Utility Maximization Approach.’’ Journal of In-
ternational Money and Finance 4, 1985.

‘‘Fractured Liberalism: Argentina Under Martinez de Hoz.’’ Economic Development and Cultural Change
34, no. 3, April 1986.
‘‘Temporary Stabilization: Predetermined Exchange Rates.’’ Journal of Political Economy 94, no. 6, De-
cember 1986.
‘‘On the Costs of Temporary Liberalization/Stabilization Experiments.’’ In Economic Reform and Stabili-
zation in Latin America, eds. M. Connolly and C. Gonzalez. New York: Praeger Publishers, 1986.
466 Publications of Guillermo A. Calvo

‘‘Welfare, Banks, and Capital Mobility: The Case of Predetermined Exchange Rates.’’ In Structural Ad-
justment and the Real Exchange Rate in Developing Countries, ed. Sebastian Edwards. Chicago: University
of Chicago Press, 1986.
‘‘Balance of Payments Crises in a Cash-in-Advance Economy: Current Account and Real Exchange
Rate Implications with Perfect-Foresight Dynastic Families.’’ Journal of Money, Credit and Banking 19,
February 1987.
‘‘Real Exchange Rate Dynamic with Fixed Nominal Parities: Structural Change and Overshooting.’’
Journal of International Economics 22, no. 1-2, February 1987.
‘‘On the Costs of Temporary Policy.’’ Journal of Development Economics 27, no. 1-2, December 1987.
‘‘On the Economics of Supervision.’’ In Incentives, Cooperation and Risk Sharing, ed. H. R. Nalbantian.
New Jersey: Rowman and Littlefield, 1987.
‘‘Costly Liberalizations.’’ International Monetary Fund Staff Papers 35, September 1988.
‘‘Servicing the Public Debt: The Role of Expectations.’’ American Economic Review 78, no. 4, September
1988.
‘‘Optimal Time-Consistent Fiscal Policy with Uncertain Lifetimes.’’ With Maurice Obstfeld, Econometr-
ica 56, March 1988. An extended version was published in Economic Effects of the Government Budget,
ed. Elhanan Helpman, Assaf Razin, and Efraim Sadka. Cambridge, MA: MIT Press, 1988.
‘‘Anticipated Devaluations.’’ International Economic Review 30, no. 3, August 1989.

‘‘Is Inflation Effective for Liquidating Short-Term Nominal Debt?’’ International Monetary Fund Staff
Papers 36, no. 4, December 1989.

‘‘Controlling Inflation: The Problem of Non-Indexed Debt.’’ In Debt, Adjustment and Recovery: Latin
America’s Prospect for Growth and Development, eds. S. Edwards and F. Larrain. New York: Basil Black-
well, 1989.
‘‘Perspectives on Foreign Debt.’’ In Spanish, with Eduardo Borensztein. El Trimestre Económico, Decem-
ber 1989.
‘‘A Delicate Equilibrium: Debt Relief and Default Penalties in an International Context.’’ In Analytical
Issues in Debt, ed. Jacob A. Frenkel. Washington, D.C.: International Monetary Fund, 1989.
‘‘Incredible Reforms.’’ In Debt, Stabilization and Development, eds. Guillermo Calvo, Ronald Findlay,
Pentti Kouri, and Jorge Braga De Macedo. New York: Basil Blackwell, 1989.
‘‘Credibility and Nominal Debt: Exploring the Role of Maturity in Managing Inflation.’’ With Pablo
Guidotti. International Monetary Fund Staff Papers 37, no. 3, September 1990.
‘‘Time Consistency of Fiscal and Monetary Policy.’’ With M. Obstfeld. Econometrica 58, no. 5, September
1990.

‘‘Interest Rate Policy in a Small Open Economy: The Predetermined Exchange Rates Case.’’ With Carlos
Végh. IMF Staff Papers 37, December 1990.

‘‘Indexation and Maturity of Government Bonds: An Exploratory Model.’’ With Pablo Guidotti, in Cap-
ital Markets and Debt Management, eds. R. Dornbusch and M. Draghi. New York: Cambridge University
Press, 1990.
‘‘From Centrally-Planned to Market Economy: The Road from CPE to PCPE.’’ With Jacob Frenkel. IMF
Staff Papers 38, June 1991.
Publications of Guillermo A. Calvo 467

‘‘Debt Relief and Debt Reschedule: The Optimal Contract Approach.’’ With Graciela Kaminsky. Journal
of Development Economics 36, July 1991.
‘‘Credit Markets, Credibility and Economic Transformation.’’ With Jacob A. Frenkel. Journal of Economic
Perspectives 5, Fall 1991.
‘‘The Perils of Sterilization.’’ IMF Staff Papers 38, no. 4, December 1991.

‘‘Financial Aspects of Socialist Economies: From Inflation to Reform.’’ In Adjustment and Growth in
Reforming Socialist Economies: Lessons from Experience, eds. Vittorio Corbo, Fabrizio Coricelli, and Jan
Bossak. Washington, D.C.: World Bank, 1991.
‘‘Optimal Maturity of Nominal Government Debt: The First Tests.’’ With Pablo Guidotti and Leonardo
Leiderman. Economic Letters 35, no. 4, 1991.
‘‘Temporary Stabilization Policy: The Case of Flexible Prices and Exchange Rates.’’ Journal of Economic
Dynamics and Control 15, 1991.
‘‘Optimal Inflation Tax under Precommitment: Theory and Evidence.’’ With Leonardo Leiderman.
American Economic Review 82, no. 1, March 1992.
‘‘Stabilizing Previously-Centrally-Planned Economy: Poland 1990.’’ With Fabrizio Coricelli. Economic
Policy 0, no. 14, April 1992.
‘‘Optimal Maturity of Nominal Government Debt: An Infinite-Horizon Model.’’ With Pablo Guidotti.
International Economic Review 33, no. 4, November 1992.

‘‘Are High Interest Rates Effective for Stopping High Inflation? Some Skeptical Notes.’’ The World Bank
Economic Review 6, no. 1, 1992.

‘‘Obstacles to Transforming Centrally-Planned Economies: The Role of Capital Markets.’’ With Jacob A.
Frenkel. In Transition to a Market Economy in Central and Eastern Europe, Proceedings of the OECD-
World Bank Conference, Paris, 1992.
‘‘Stagflationary Effects of Stabilization Programs in Reforming Socialist Countries: Enterprise-Side vs.
Household-Side Factors.’’ With Fabrizio Coricelli. World Bank Economic Review 6, no. 1, 1992.
‘‘Transformation of Centrally-Planned Economies: Credit Markets and Sustainable Growth.’’ With
Jacob A. Frenkel. In Central and Eastern Europe: Roads to Growth, ed. G. Winkler. Washington, D.C.:
IMF, 1992.
‘‘Capital Inflows and Real Exchange Rate Appreciation in Latin America: The Role of External Factors.’’
With L. Leiderman and C. Reinhart. IMF Staff Papers 40, no. 1, March 1993.

‘‘Output Collapse in Eastern Europe: The Role of Credit.’’ With Fabrizio Coricelli. IMF Staff Papers 40,
no. 1, March 1993.
‘‘On the Flexibility of Monetary Policy: The Case of the Optimal Inflation Tax.’’ With P. Guidotti. Re-
view of Economic Studies 60, no. 3, June 1993.
‘‘Exchange-Rate-Based Stabilization under Imperfect Credibility.’’ With Carlos Végh. In Proceedings
from I.E.A. Conference on Open Economy Macroeconomics. New York: MacMillan, 1993.
‘‘Management of Nominal Public Debt: Theory and Applications.’’ In The Political Economy of Govern-
ment Debt, eds. Harrie Verbon and Frans van Winden. Amsterdam, Holland: North Holland, 1993.
‘‘Trade Reforms of Uncertain Duration and Real Uncertainty: A First Approximation.’’ With Enrique G.
Mendoza. IMF Staff Papers 41, no. 4, December 1994.
468 Publications of Guillermo A. Calvo

‘‘Inflation Stabilization and Nominal Anchors.’’ With C. Végh. Contemporary Economic Policy 12, no. 2,
April 1994.
‘‘Money Demand, Bank Credit, and Economic Performance in Former Socialist Economies.’’ With M.
Kumar. IMF Staff Papers 41, no. 2, July 1994.
‘‘The Capital Inflows Problem: Concepts and Issues.’’ With L. Leiderman and C. Reinhart. Contemporary
Economic Policy 12, no. 3, July 1994.
‘‘Capital Inflows to Latin America: The 1970s and the 1990s.’’ With C. Reinhart and L. Leiderman. In
Economics in a Changing World, Proceedings of the Tenth World Congress of the International Economic
Association, Vol. 4, ed. E. L. Bacha. London: Macmillan, 1994.
‘‘Credibility and the Dynamics of Stabilization Policy: A Basic Framework.’’ In Advances in Econome-
trics, Sixth World Congress, Vol. II, Proceedings from the VI World Meeting of the Econometric Society,
ed. C. A. Sims. Cambridge, UK: Cambridge University Press, 1994.
‘‘Interenterprise Arrears in Economies in Transition.’’ With F. Coricelli. Empirica 21, 1994.

‘‘Stabilization Dynamics and Backward-Looking Contracts.’’ With C. Végh. Journal of Development Eco-
nomics 43, no. 1, 1994.

‘‘Targeting the Real Exchange Rate: Theory and Evidence.’’ With C. Reinhart and C. Végh. Journal of
Development Economics 47, 1995.
‘‘Fighting Inflation with High Interest Rates: The Small-Open-Economy Case under Flexible Prices.’’
With C. Végh. Journal of Money, Credit, and Banking 27, no. 1, March 1995.
‘‘The Management of Capital Flows: Domestic Policy and International Cooperation.’’ In The Interna-
tional Monetary and Financial System: Developing-Country Perspectives, ed. G. K. Helleiner. New York: St.
Martin’s Press, 1996.
‘‘Inflows of Capital to Developing Countries in the 1990s.’’ With L. Leiderman and C. Reinhart. Journal
of Economic Perspectives 10, no. 2, Spring 1996.
‘‘Capital Flows and Macroeconomic Management: Tequila Lessons,’’ International Journal of Finance Eco-
nomics 1, 1996.
‘‘What Role for the Official Sector?’’ With M. Goldstein, in Private Capital Flows to Emerging Markets after
the Mexican Crisis, eds. G. Calvo, M. Goldstein, and E. Hochreiter. Washington, D.C.: Institute for Inter-
national Economics, 1996.
‘‘Capital Flows in Central and Eastern Europe: Evidence and Policy Options.’’ With R. Sahay and C.
Végh, in Private Capital Flows to Emerging Markets after the Mexican Crisis, eds. G. Calvo, M. Goldstein,
and E. Hochreiter. Washington, D.C.: Institute for International Economics, 1996.
‘‘Disinflation and Interest-Bearing Money.’’ With C. Végh. Economic Journal 106, no. 439, November
1996.
‘‘Monetary Policy and Interenterprise Arrears in Post-Communist Economies: Theory and Evidence.’’
With F. Coricelli. Journal of Policy Reform 1, 1996.
‘‘Petty Crime and Cruel Punishment: Lessons from the Mexican Debacle.’’ With Enrique G. Mendoza.
American Economic Review 86, no. 2, May 1996.
‘‘Reflections on Mexico’s Balance of Payments Crisis: A Chronicle of Death Foretold.’’ With E. Men-
doza. Journal of International Economics 41, September 1996.

‘‘Growth, Debt, and Economic Transformation: The Capital Flight Problem.’’ In New Theories in Growth
and Development, eds. F. Coricelli, M. DiMatteo and F. H. Hahn. Cambridge, UK: Macmillan, 1997.
Publications of Guillermo A. Calvo 469

‘‘Varieties of Capital-Market Crises.’’ In The Debt Burden and its Consequences for Monetary Policy, eds. G.
Calvo and M. King. New York: Macmillan, 1998.
‘‘Why is ‘the Market’ so Unforgiving? Reflections on the Tequilazo.’’ In Financial Reform in Developing
Countries, eds. J. M. Fanelli and R. Medhora. New York: Macmillan, 1998.
‘‘Uncertain Duration of Reform: Dynamic Implications.’’ With A. Drazen. Macroeconomic Dynamics 2,
no. 4, December 1998.
‘‘Capital Flows and Capital-Market Crises: The Simple Economics of Sudden Stops.’’ Journal of Applied
Economics I, November 1998.
‘‘Inflation Stabilization.’’ With C. Végh. In Handbook of Macroeconomics, eds. J. Taylor and M. Woodford.
Amsterdam, Netherlands: North Holland, 1999.
‘‘Empirical Puzzles of Chilean Stabilization Policy.’’ With Enrique G. Mendoza, in Chile: Recent Policy
Lessons and Emerging Challenges, eds. G. Perry and D. Leipziger. Washington, DC: World Bank, 1999.
‘‘Capital-Market Crises and Economic Collapse in Emerging Markets: An Informational-Frictions
Approach.’’ With Enrique G. Mendoza. American Economic Review 90, May 2000.
‘‘Contagion, Globalization, and the Volatility of Capital Flows.’’ With Enrique G. Mendoza, in Capital
Flows and the Emerging Economies, ed. Sebastian Edwards. Chicago: University of Chicago Press, 2000.
‘‘When Capital Flows Come to a Sudden Stop: Consequences and Policy.’’ With Carmen Reinhart,
in Reforming the International Monetary and Financial System, eds. Peter B. Kenen and Alexander K.
Swoboda. Washington, D.C.: International Monetary Fund, 2000.
‘‘Balance of Payments Crises in Emerging Markets: Large Capital Inflows and Sovereign Govern-
ments.’’ In Currency Crises, ed. Paul Krugman. Chicago: University of Chicago Press, 2000.
‘‘Betting Against the State.’’ Journal of International Economics, June 2000.
‘‘Rational Contagion and the Globalization of Securities Markets.’’ With Enrique G. Mendoza. Journal of
International Economics 51, no. 1, June 2000.
‘‘Fixing For Your Life.’’ With C. Reinhart. Brookings Trade Forum, 2000.

‘‘Capital Markets and the Exchange Rate: With Special Reference to the Dollarization Debate in Latin
America.’’ Journal of Money, Credit and Banking 33, no. 2, May 2001.
‘‘Economic Policy in Emerging Markets’’ (in Spanish). Moneda y Crédito, 212, 2001.

‘‘On Dollarization.’’ Economics of Transition Journal 10, no. 2, 2002.

‘‘Contagion in Emerging Markets: When Wall Street is a Carrier.’’ Proceedings from the International
Economic Association Congress, Vol. 3, Buenos Aires, Argentina, 2002.

‘‘Fear of Floating.’’ With C. Reinhart. Quarterly Journal of Economics 117, no. 2, 2002.
‘‘A Theory of Inflationary Inertia.’’ Volume in honor of Edmund S. Phelps, eds. P. Aghion, R. Frydman,
J. Stiglitz and M. Woodford. Princeton, N.J.: Princeton University Press, 2003.
‘‘Explaining Sudden Stops, Growth Collapse and BOP Crises: The Case of Distortionary Output
Taxes.’’ IMF Mundell-Fleming Lecture, IMF Staff Papers, 2003. Also in Emerging Capital Markets in Tur-
moil: Bad Luck or Bad Policy? ed. G. Calvo. Cambridge, MA: MIT Press 2005.
‘‘The Mirage of Exchange Rate Regimes for Emerging Markets Countries.’’ With F. Mishkin. Journal of
Economic Perspectives 17, no. 4, 2003.
 Index

Acemoglu, D., 330, 361 Asian Bond Market Initiative, 269

Ackerman, Susan Rose, 432 Asian Development Bank, 264, 266, 269
Alesina, A., 122, 124, 130 Asset market, 223–224, 241–244
Allen, R. G. D., 446 Asymmetry, 5, 16–17, 249–250
Amadeus database, 338 Auernheimer, Leonardo, xv, 41–68
American Economic Association, 451, 455
American Economic Review, 452, 454, 457, 459 Bacha, Edmar, 448
Andersen, P. K., 408 Backus, David K., 7
Andorra, 125 Baldwin, R., 280
Angeletos, G.-M., 249 Banco Central do Brasil, 83–84
Arellano, Cristina, 109, 112 Bank for International Settlements (BIS), 251,
Argentina, xiv, xix, 74, 103, 105, 446, 448 270–272
Calvo policy for, 437–438 Bank of Canada, 78
Cavallo and, 76 Bank of England, 81
central bank independence and, 78 Banks
credibility and, 108 advanced vs. emerging economies and, 72–74
debt rescheduling and, 440–442 Argentina crisis and, 75–76
deposit rescheduling in, 438–439 central, 71, 81 (see also Central banks)
dollarization and, 122 financial intermediation and, 249–252, 260, 264,
double mismatch and, 217 266–272
financial intermediation and, 252, 260, 264, 272 foreign exchange and, 73
globalization and, 173, 176, 182, 187, 203 inflation targeting and, 79–84
government debt and, 251 institutional development and, 74–79
inflation policy and, 75–76, 437, 455 runs on, 73
movable collateral and, 266 sudden stops and, 75
nominal rigidities and, 454 Barings, 88
private bond markets and, 269–270 Barro, R. J., 122, 124, 130, 134, 379, 392
recovery and, 218–219, 221, 226–241 Barro-Lee model, 134
Tablita plan and, 453 Bassetto, M., 41
Tequila Effect and, 99 Batini, Nicoletta, 72, 77
wealth redistribution and, 96 Begg, D., 41
Argentine Bankers Association (ADEBA), 438 Ben-David, D., 379–380
Asia. See also Specific country Berg, A., 355
crisis in, xvii, 99, 101, 103, 173, 183, 200, 442, 458 Bermuda, 125, 151n10
double mismatch and, 217 Bernanke, Ben S., 77, 79
East Asian Four and, 219, 221 Bhagwati, Jagdish, 449
fiscal health of, 217–222 Blanco, Herminio, 108
rapid recovery of, 218 Blejer, Mario I., 88
472 Index

Bloomberg.com, 199 optimal policy rules and, 18

Blundell, R., 145 output collapses and, 380, 392, 400, 408
Bodman, P. M., 381 papers of, 433–435, 438, 443, 445, 453–455, 457
Brady bonds, 101, 251, 272 pegging and, 121
Brady Plan, 439, 456 Penn and, 452–453
Brash, Donald T., 77 rational expectations (RE) and, 460–461
Brazil, xvi, 455 recessions and, 377–378
central bank independence and, 82–84 recovery and, 218
double mismatch and, 217 shock transmission and, 175
globalization and, 172–173, 176, 178, 183, 187, sticky prices and, 433, 445, 452–457
198, 200 sudden stops and, 99–101, 104–105
inflation targeting and, 79–84 switching effects and, 3–4
Lula and, 82–84 as theoretician of reality, 445–446
private bond markets and, 270 time consistency and, 449–452
recovery and, 219, 221 transition economies and, 456
Brookings Papers ( journal), 457 Calvo, Sara, 431
Buiter, Willem, xv, 41 Canada, 78, 176, 260
Bulgaria, 338 Capital flows, 95, 456
Burnside, Craig, 75 balance-of-payment crises and, 96
capital mobility and, 279–293
Cagan, Phillip, 449, 460 contagion and, 98–106
Call bonds, 43 continuous time analysis and, 42–46
Calomiris, Charles, 75 convergence hypothesis and, 391
Calvo, Guillermo, 97, 116 crises recovery and, 217–241
Argentina crisis and, 437–438 financial intermediation and, 249–252, 260, 264,
background of, 431, 446–449 266–272
balance-of-payment crises and, 96 fiscal theory of the price level (FTPL) and, 42–
capital income taxation and, 279 64
Columbia University and, xiii–xix, 432–435, globalization and, 173–175, 279–293 (see also
448–453 Globalization)
contract theory and, 434, 438–440 inflation targeting and, 71–90
credibility and, 107, 452–457 Lahiri-Singh-Végh model and, 23–38
credit’s role and, 327 liability dollarization and, 73
crisis recovery and, 217 macroeconomic pessimism and, 159–168
debt reduction and, 438–443 Mundell-Fleming model, 21–38
elegant simplicity of, 41 output collapses and, 377–425 (see also Output
financial intermediation and, 250, 272 collapses)
Fraga and, 80 perfect mobility and, 21
future policy and, 460–461 sudden stops and, 73–75, 96–114, 378, 392, 394,
hard currencies and, 97 443, 445
inflation targeting and, 71–74, 77, 79–80, 85, 87– taxes and, 279–293
89 transition economies and, 327–345, 349–376 (see
influence of, xiii–xix, 3, 433–435, 437–443, 445 also Transition economies)
Inter-American Bank and, 459–460 Capital mobility
International Monetary Fund (IMF) and, 437– arbitrage condition and, 288–289
438, 445, 453–456 globalization and, 279–293
liability dollarization and, 73 households and, 280–284
macroeconomic pessimism and, 159–160 labor and, 279–286
mathematics and, 446, 448, 452 median voter equilibrium and, 287–289
Mendoza interview with, 445–461 Pareto optimality and, 285
Mexican crisis and, 173, 457–460 policy tools and, 286–287
Mundell-Fleming model and, 21 political economy and, 279–293
Index 473

production and, 284–286 Claessens, S., 218

taxes and, 279–293 Cobb-Douglas production function, 160
Caprio, Gerald, Jr., 80–81 Cochrane, J., 41
Cardoso-Lecourtois, M., 337 Cold War, 279
Carlstrom, C. T., 51, 53 Cole, Harold L., 96, 458
Carstens, Agustin, xiii Collateral, 264, 266
Cash-in-advance (CIA) constraint Colombia, 182, 176
credit markets and, 332–338 Columbia University, xiii, xix, 432–435, 448–453
financial intermediation and, 249 Common currency
fixed exchange rates and, 35–37 comparative analysis of, 127–129
flexible exchange rates and, 35–36 effects of, 123–127
intuition and, 36 empirical model and, 130–132
Lahiri-Singh-Végh model and, 24–25, 34–38 estimator techniques and, 125
Nontrader’s binding constraint and, 34–36 growth and, 121–124
Trader’s binding constraint and, 36–37 inflation and, 132, 135, 137, 147–148
Cass, David, 447 macroeconomic performance and, 130–139
Castellanos, J., 269 moral hazard and, 131–132
Cavallo, Domingo, 76 nonparametric analysis and, 145–147
Cavallo, Michelle, 112 outcome equations and, 134–135, 147–148
Celasun, Oya, 454 robustness analysis and, 145–149
Center for Advanced Study in the Behavioral treatment equation and, 133–134, 146
Sciences (CASBS), 432 volatility and, 137, 139
Center for Research on the Epidemiology of Communante Financiere de l’Afrique (CFA), 122,
Disasters (CRED), 391 127, 148–149
Central Bank of Argentina, 437 Consumption
Central banks capital mobility and, 279–293
accountability and, 71, 81 cash-in-advance (CIA) constraint and, 249
advanced vs. emerging economies and, 72–74 equilibrium and, 28–29 (see also Equilibrium)
bank runs and, 73 financial intermediation and, 249–252, 264, 266–
dollarization and, 121–122 272
fiscal theory of the price level (FTPL) and, 43–46 Lahiri-Singh-Végh model and, 23–38
independence of, 78, 82–84, 450–451 local-currency pricing (LCP) and, 4–7
independent currency unions (ICUs) and, 126– macroeconomic pessimism and, 159–168
127 Obstfeld model and, 5–18
inflation targeting and, 71, 79–84, 87 velocity shocks and, 31–32
local-currency pricing (LCP) and, 4–7 Contagion
pegging and, 55–62 causes of, 100–101
Pou and, 76, 78 credibility and, 97–104
Cerra, V., 379 ex ante distribution and, 104–105
Cespedes, Luis, 21, 112, 217 global capital markets and, 104–106, 171–172
Chang, Roberto, 21, 96, 112, 217, 458 information costs and, 104–105
Charoenseang, J., 218 measurement of, 99
Chatsuripitak, N., 268–270 rational, 104–106
Cheap talk, 300–301, 309–310 sudden stops and, 98–106
Chile, xvi, 73–75, 173 Continuous time analysis, 42–46
financial intermediation and, 260, 267 Contract theory, 434, 438–440
globalization and, 176 Contreras, B., 41
inflation targeting and, 79–80, 84–86, 113 Convergence hypothesis, 391
wealth redistribution and, 96 Corden, W. M., 130, 454
China, 103, 200 Coricelli, Fabrizio, xviii, 327–347
Choi, S., 440 Corsetti, G., 174, 217
Christiano, L. J., 41, 51, 53, 218 Costa Dias, M., 145
474 Index

Costa Rica, 173 causes of, 390–405

Cottarelli, C., 328 Chile, 79–80, 84–85, 96
Council for Mutual Economic Assistance chronological data on, 204–213
(CMEA), 355 deteriorating fundamentals and, 176
Covariance matrix, 131 double mismatch and, 217
Cowles Foundation, 432 Ecuador, 85, 103, 172
Cox proportional hazards model, 413 emerging market economies and, 98–104 (see
Cranston, R., 266–267 also Emerging market economies)
Crawford, V., 309 event definition and, 381–385
Credibility, 445, 460 financial intermediation and, 249–252, 260, 264,
Calvo and, 452–457 266–272
contract enforcement and, 267 globalization and, 171–175 (see also Globaliza-
emerging economies and, 95–97, 107–114 tion)
financial frictions and, 107–114 Japan, 172–173
Mendoza-Uribe model and, 107–108 literature on, 217
policy implications and, 114–117 macroeconomic pessimism and, 159–168
property registration and, 267 Mexico, xvi, 85, 96, 98–99, 103, 187, 441, 457–460
sticky prices and, 452–457 output collapses and, 377–425 (see also Output
sudden stops and, 96–104, 107–114 collapses)
time consistency and, 449–452 recovery and, 217–241 (see also Recovery)
Credit markets regression analysis and, 390–405
aggregate effects and, 340–341 Russia, 73–74, 80, 99, 101, 103, 174, 458
cash-in-advance (CIA) constraints and, 332–338 shock transmission and, 171–172
chain system and, 335–338 systemic risk and, 175–176
credit-to-GDP ratios and, 330 Tequila, 99, 187, 441, 457–460
cross-country regressions and, 339–345 Thailand, 172–173
depth of, 327 time inconsistency and, 95–96
empirical analysis of, 338–345 Turkey, 85
financial sector development and, 328–332 wealth redistribution and, 96
growth during transition and, 332–338 Croatia, 338
inter-enterprise arrears and, 335–338 Crockett, Andrew, 442
interest rates and, 332–333 Cukierman, Alex, 78
labor and, 332–333, 337 Cumulative liberalization index (CLI), 354
liberalization and, 327, 331 Currency, xvi
liquidity and, 333–338 bank runs and, 73
literature on, 328, 330 common unions and, 121–124 (see also Common
output collapses and, 377–425 (see also Output currency)
collapses) credibility and, 95–104, 107–117
reform and, 327, 335–338 credit markets and, 327–345
shocks and, 336–337 crises recovery and, 217–241
stability and, 327 devaluation and, 62
threshold model and, 341–344 dollarization and, 21, 121–122, 455 (see also
trade credit and, 335–338 Dollarization)
transition economies and, 327–345 Dornbusch model and, 223
underdevelopment of, 327–332 duration analysis and, 384
volatility and, 330, 337 fiscal theory of the price level (FTPL) and, 46–52
World Development Indicators (WDI) and, 338 floating fears and, 73, 85
Crises, xvi–xvii, 443 globalization and, 200
Argentina, 103 (see also Argentina) hard, 97–117, 121–153 (see also Pegging)
Asian, xvii, 99, 101, 103, 173, 183, 200, 442, 458 independent currency unions (ICUs) and, 121–
balance-of-payments, 96, 114–115 129, 133, 139, 143, 146–150
Brazil, 79–84 inflation targeting and, 72–73
Index 475

information costs and, 107–117 Dell’Ariccia, G., 328

institutional development and, 74–79 del Valle, C., 269
International Monetary Fund (IMF) and, 123 De Melo, M., 355
liability dollarization and, 73–74, 109 Democracy
macroeconomic fundamentals and, 219–222 deliberative, 301–302
Mendoza-Uribe model and, 107–108 output collapses and, 394, 410, 412
optimum currency area and, 133 values and, 315
rational herding and, 105 Denizer, C., 355
seignorage and, 121 Depreciation. See Currency
substitution and, 72–73 Desai, Padma, xvii, 217–246
sudden stops and, 107–114 De Soto, H., 264
treatment equation and, 133–134 Devereux, Michael, xiv, 4, 9, 13, 16–17
Czech Republic, 176 Dı́az-Alejandro, Carlos, 448–449
Dı́az-Giménez, J., 249
Daniel, B., 41 Diebold, F. X., 381
Debreu, G., 447 Dietrich, Marlene, 431, 435n1
Debt, xiv, xvii, 97 Di Venuto, N., 381
asymmetric information and, 249–250 Djankov, S., 267, 330, 335
Bank for International Settlements (BIS) and, 251 Doe, Samuel, 125
Calvo policy and, 438–443 Dollar, D., 361
capital mobility and, 286–287 Dollarization, xvi, 455. See also Common currency
collateral and, 113 advanced economies and, 121–122
as countercyclical shock absorber, 252 CFA franc and, 127
crises recovery and, 217–241 (see also Crises) description of, 121–122
developing countries and, 251–270 Ecuador and, 150
domestic vs. external, 251–260, 271–272 El Salvador and, 150
double mismatch and, 217 growth and, 122
exchange rate fluctuations and, 84–87 increase of, 122
fiat money and, 41 independent currency unions (ICUs) and, 124,
financial intermediation and, 249–252, 260, 264, 139, 143, 150
266–272 interest rates and, 122
fiscal theory of the price level (FTPL) and, 41–64 labor mobility and, 126
forgiveness of, 166 liability, 73–74, 109
GDP ratios and, 252–260 Mundell-Fleming model and, 21
globalization and, 174–175 (see also Globaliza- Panama and, 125, 130
tion) Puerto Rico and, 125
highly indebted poor countries (HIPCs) and, 320 strict, 139–144
inflation targeting and, 73–74 tax evasion and, 125–126
institutional development and, 74–79 treatment equation and, 146
institutions and, 260–267 treatments and, 146
legal weaknesses and, 264–267 volatility and, 143
lemons problem and, 174 Dornbusch, Rudiger, 122, 223, 457
movable collateral and, 264, 266 Double mismatch, 217
pegging and, 55–62 Double threshold model, 343–344
private bond market and, 268–270 Drazen, Allan, xviii, 295–324
procyclicality and, 252 Dupor, B., 41
rescheduling of, 440–442 Duration analysis, 424
transition economies and, 327–345, 360 (see also correlates for, 412
Transition economies) crises recovery and, 405–422
unanimous action clauses (UACs) and, 440–441 equilibrium and, 387–390
de Gregorio, Jose, 80 event definition and, 381–385
Deliberative democracy, 301–302 exponential parameter and, 412–413, 416–417
476 Index

Duration analysis (cont.) output collapses and, 387–390, 422–425 (see also
Gini index and, 412 Output collapses)
Gompertz parameter and, 412–413, 416–417 ownership issues and, 295–321
hazard rates and, 387–390, 408–422 policy implications and, 114–117
Herfindahl index and, 417 signaling and, 106
human capital and, 410 stability and, 95
log-logistic parameter and, 412–413, 416–417 sudden stops and, 98–104
log-normal parameter and, 412–413, 416–417 Emerging Market Fund (EMF), 458
multiple equilibria and, 413, 416 Emerging Markets Bond Index (EMBI), 99, 392
Nelson-Aalen cumulative function and, 387–388 Engel, Charles, xiv, 4, 9, 13, 16–17, 123, 125
output collapses and, 378–381, 394–405, 408–422 Enron, 73
Prentice-Williams-Peterson model and, 413, Equations
418–421 baseline crisis specification, 390
regional effects and, 378–379, 385–390 capital mobility model, 281–289, 292–293
time-specific effects and, 378–381, 394–405 Cobb-Douglas, 160
Weibull specification and, 410 covariance matrix, 131
Durdu, C. B., 116 domestic production, 160
Dynamic aggregative model, 435 duration analysis, 405
empirical growth model, 131–132
East Asian Four, 219, 221 enterprise liquidity, 333
East Caribbean Currency Area (ECCA), 122, 127, equilibrium Fisher condition, 43
148–150 Euler, 12, 17, 44
Economia ( journal), 458 export proximity, 393
Economic and Monetary Union (EMU) countries, fiscal theory of the price level (FTPL), 43–47, 51,
77 54, 57
Economist, The ( financial magazine), 279 flexible-price equilibrium, 8–9
Ecuador, 75, 85, 101, 103, 150, 172 inflation model, 147
Education Lahiri-Singh-Végh, 23–37
labor and, 280–284 local-currency pricing (LCP), 5, 7
public discussion and, 299–302, 312 macroeconomic pessimism, 160–165
Edwards, Sebastian, xvi, 80, 121–156 maximum likelihood function, 405
Eichenbaum, Martin, 75 Obstfeld model, 5, 7–14
Eichengreen, Barry, 73, 122 pseudo R2 measure, 396
Elasticity, 6, 112, 114–115, 159–162 recovery model, 223–225, 241–243
El Salvador, 122, 150 sticky-price equilibrium, 10–14
Elster, J., 301 strong-form globalization, 192
Emerging market economies. See also Transition threshold model, 341
economies transitional credit growth, 332–333, 336, 341
advanced economies and, 72–74 weak-form globalization, 191
borrowing constraints and, 109–110 Wilcoxon test, 191
capital markets and, 457–460 worker consumption, 161
contagion and, 98–106 Equilibrium, 113, 434, 438, 460
credibility and, 95–104, 107–117 arbitrage condition and, 288
crises lessons from, 98–104 asset price continuity principle and, 60–61
crises recovery and, 217–241 consumption, 28–29
financial intermediation and, 251–270 duration analysis and, 387–390, 413, 416
globalization and, 171–213 (see also Globalization) exchange rates and, 16–18 (see also Exchange
independent currency unions (ICUs) and, 121– rates)
127, 133, 139, 143, 146–150 fiscal theory of the price level (FTPL) and, 43–62
inflation targeting and, 71–90 Fischer condition and, 43
institutional development and, 74–77 flexible-price, 8–9
liability dollarization and, 109 Kolmogorov-Smirnov test and, 178, 182
Index 477

Lahiri-Singh-Végh model and, 27–29 optimal regimes for, 21–38

leisure time and, 282 pegging and, 55–62, 121 (see also Pegging)
local-currency pricing (LCP) and, 4–10 switching effects and, 3–4
macroeconomic pessimism and, 159–168 transition economies and, 355–360
median voter, 287–289 Expanded War Data Set (Gleditsch), 391
money market, 27
multiple, 413, 416 Fearon, J., 301, 312
Nash, 18n6 Fernández, Roque B., xiv, xix, 437–444
nominal rigidities and, 453–454 Fiat money, 41, 45
Obstfeld model and, 5–18 Financial intermediation
optimal policy rules and, 14–18 asymmetric information and, 249–250
output collapses and, 387–390 Bank for International Settlements (BIS) and, 251
Pareto optimality and, 285 Calvo and, 250, 272
pegging and, 55–62 cash-in-advance (CIA) constraint and, 249
political economy model and, 280–293 contract enforcement and, 267
preset price, 9–14 default scenario and, 250–251
recovery and, 225–231, 241–244 government debt and, 249–252, 260, 264, 266–
reform and, 166–167 272
steady-state, 49–52, 63 International Monetary Fund (IMF) and, 249–
sudden stops and, 110, 112 250
transversality and, 28 movable collateral and, 264, 266
velocity shocks and, 27–28 obtaining credit and, 267
Equity production and, 250
market issues and, 99, 112–115, 172, 175, 182, property registration and, 267
199, 266, 285 real estate and, 249–250
open economies and, 112–115 stress test approach and, 249
volatility and, 172 Financial Times, 199
Estonia, 176 Findlay, Ron, 432, 448–449
Euler equations, 12, 17, 44 Finland, 176, 260
European Bank for Reconstruction and Fiscal Neutrality toward Economic Growth (Phelps),
Development (EBRD), 266, 355, 362, 372 432
European Union (EU), 280, 290–291, 335 Fiscal theory of the price level (FTPL), 74
Exchange rates, xiv–xv, 97, 217, 438 adjusted fiscal surplus and, 52–55, 66n14
asymmetric responses and, 16–17 asset price continuity principle and, 60–61
cash-in-advance constraints and, 24–25, 34–38 call bonds and, 43
crises recovery and, 217–241 continuous time analysis and, 42–46, 65n3
Devereux-Engel model and, 4, 13 equilibrium and, 43–62
Dornbusch model and, 223 Euler equation for, 44
fixed, 9–14, 30–36 fiat money and, 41, 45
flexible, 3, 8–9, 16–18, 29–33, 35 government bonds and, 43–52, 63
floating fears and, 73, 80, 85 growth and, 43–52
fluctuation issues and, 84–87 inflation and, 43–45, 48–50
Friedman model and, 3–4 instantaneous adjustments and, 59
globalization and, 176 (see also Globalization) literature on, 41–42
hard currencies and, 97–117 monetary rule and, 46–55
inflation targeting and, 79–87 nominal interest rule and, 55–62
institution building and, 362–372 nominal money treatment and, 52–53
Lahiri-Singh-Végh model and, 23–38 once-and-for-all increase and, 50
liability dollarization and, 73 pegging and, 41–42, 51, 55–63
local-currency pricing (LCP) and, 4–10 primary surplus and, 49–50
Mundell-Fleming model and, 21–38 taxes and, 44
Obstfeld model and, 5–18 transversality and, 45–54, 58
478 Index

Fischer, Stanley, xviii, 121–122, 349–376, 433 deteriorating fundamentals and, 176
Fisher, I., 114 determinants and, 176–199
Fisher condition, 43 equity and, 199
Fisher’s principle of randomization, 191 financial centers and, 173, 175
Fisman, R., 339–342 Fisher’s principle of randomization and, 191
Fitzgerald, T., 41, 51, 53 Kolmogorov-Smirnov test and, 178, 182, 187
Fixed effects logit analysis, 403 lemons problem and, 174
Fleisig, H. W., 264, 266–267 liquidity and, 174–175
Fleming, M. J., 21, 270, 460 literature on, 176
Flood, Robert, 452 long-term capital management (LTCM) and,
Forder, James, 78 171–172, 175, 178, 200, 203
Ford Foundation, 448 measurement and, 176–199
Foreign aid, 349–376 nongovernmental organizations (NGOs) and,
Foreign direct investment (FDI), 101, 458 296
Foundations of Economic Analysis (Samuelson), origins of, 199–201
447 political-economy model for, 280–293
Fraga, Arminio, 78, 80, 82 prices and, 174–175, 199
Frailties, 408 shock transmission and, 173–175, 182–183, 192–
France, 176, 291 194, 198–199
Frankel, J. A., 122–123, 125, 175 spillover and, 172–173, 182, 186–187, 200
French Guyana, 126 strong-form, 172–173, 203
Frieden, J., 134 systemic risk and, 175–176
Friedman, Milton, 3 trade competition and, 174
Fuerst, T. S., 51, 53 weak-form, 172–173, 201
Wilcoxon test and, 187
G7 countries, 176, 178, 192, 200 World War I era and, 279
Galton, F., 379 Goldfajn, Ilan, 78, 82
Game theory Gompertz parameter, 412–413, 416–417
heterogeneous public and, 312–315 Goods market, 224, 241–244
homogeneous public and, 303–308 Government. See also Debt
public uncertainty and, 308–312 asymmetric information and, 249–250
social welfare maximization and, 303–308 bonds and, 43–52, 63, 76, 268–270, 272, 438,
special interest groups (SIGs) and, 298–299, 440–441
313–315 central bank independence and, 78, 82
time consistency and, 449–452 cheap talk and, 300–301, 309–310
Garber, Peter M., 108 crises recovery and, 217–241
Garro, A. M., 266 deposit rescheduling and, 438–439
Gelb, A., 355 fiat money and, 41, 45
Germany, financial intermediation and, 249–252, 260, 264,
globalization and, 172, 176–178, 182–183, 186, 266–272
193–194, 198, 200 fiscal theory of the price level (FTPL) and, 41–64
tax harmonization and, 290–291 GKO bills and, 178
Ghosh, A., 123 Lahiri-Singh-Végh model and, 27
Gini index, 412 macroeconomic pessimism and, 159–168
Giovanetti, G., 249 nongovernmental organizations (NGOs) and,
Gleditsch, K., 391 296, 316
Globalization, 202 pegging and, 55–62
capital income taxation and, 279–293 promises of, 96
central vs. peripheral countries and, 173–175 public uncertainty about, 308–312
chronological market demographics for, 204– social welfare maximization and, 303–308
213 special interest groups (SIGs) and, 298–299
contagion and, 171–172 time inconsistency and, 95–96
Index 479

unanimous action clauses (UACs) and, 440–441 Gompertz parameter and, 412–413, 416–417
wealth redistribution and, 96 Nelson-Aalen cumulative function and, 387–388
Great Depression, 459 nonindustrial countries and, 388
Greece, 176 output collapses and, 387–390, 405, 408–425
Greene, W. H., 124, 131–132 pooling and, 388–389
Gross domestic product (GDP). See Growth regional analysis and, 387–390
Growth, xiv, xviii, 432, 443, 447. See also Debt Heaton, J., 270
common currencies and, 121–124, 130–139 Heckman, J. J., 124
continuous negative, 380 Helbling, T. F., 349–376
dollarization and, 122, 139–144 Helicopter drops, 45
empirical model for, 130–132 Helpman, Elhanan, 107, 454
export flexibility and, 405–422 Herding, 105
financial intermediation and, 249–252, 260, 264, Herfindahl index, 417
266–272 Herring, R. J., 268–270
fiscal theory of the price level (FTPL) and, 43–52 Heston, Charlton, 431, 435n1
independent currency unions (ICUs) and, 121– Hicks, J. R., 446–447
124, 127–129 Hidehiko, I., 124
International Monetary Fund (IMF) and, 87–88, Highly indebted poor countries (HIPCs), 320
295–321 Holland, 176
liberalization and, 331 Hong Kong
outcome equations and, 134–135 globalization and, 174–176, 200, 203
output collapses and, 377–425 (see also Output institution substitution and, 99
collapses) private bond markets and, 270
peak levels in, 377, 381, 384–385 Households
private bond markets and, 268–270 capital mobility and, 280–284
recovery and, 219–241 labor and, 280–284
Russia and, 349 Lahiri-Singh-Végh model and, 23–38
total factor productivity (TFP), 130 Human capital. See Labor
transition economies and, 327–345, 349–376 (see Hungary, 85, 176
also Transition economies) Hwang, J., 393, 417
treatment equation and, 133–134 Hyperinflation, 50, 52–53, 308, 437, 440, 460
volatility and, 330 (see also Volatility)
Guatemala, 122 Iceland, 338
Guidotti, Pablo, xvii, 454 Independent currency unions (ICUs)
Guiso, L., 340 analysis of, 125–127
Gulati, M., 440 central banks and, 126–127
Gulde, A., 123 comparative analysis of, 127–129
Gulf War, 400 dollarization and, 121–122, 124, 139–144, 150
Gurr, T. R., 392 estimator techniques and, 125
Gust, C., 218 growth and, 121–124
interest rates and, 122
Habermas, J., 301–302 macroeconomic performance and, 133
Hall, R. E., 361 nonparametric analysis and, 146–147
Hansen, B. E., 341 robustness and, 146–149
Haque, B., 41 treatment equation and, 146
Hausmann, Ricardo, xviii–xix, 73, 377–428 India, 270
Havrylyshyn, O., 355 Indonesia, 176, 200, 217, 272
Hazard rates Industrial Revolution, 448
Cox proportional model and, 413 Inflation, xvi
developing countries and, 389 bad dream story and, 52
duration analysis and, 387–390 common currency and, 132, 135, 137
frailties and, 408 cyclical, 74
480 Index

Inflation (cont.) financial intermediation and, 249–250

devaluation and, 62 policy implications and, 114–117
disinflation and, 80, 433, 454–455 public discussion and, 288–316
financial intermediation and, 270–272 signaling and, 305
fiscal imbalances and, 75–76 social welfare maximization and, 303–308
fiscal theory of the price level (FTPL) and, 43– sudden stops and, 107–114
45, 48–50 uncertainty and, 308–312
foreign, 23 Institutions. See also Banks
Friedman model and, 3–4 building, 362–372
hyperinflation and, 50, 52–53, 308, 437, 440, 460 credibility and, 107–114
independent currency unions (ICUs) and, 147– defined, 360–361
148 development of, 74–79
literature on, 433–435 formal/informal, 360–361
nominal rigidities and, 454 government debt and, 260–267
output collapses and, 396–397, 403 inflation targeting and, 74–79
recovery and, 237–241 legal issues and, 264–267, 360–362
social welfare and, 308 moral hazard and, 75
switching effects and, 3–4 nongovernmental organizations (NGOs) and,
taxes and, 45 296, 316, 361
time inconsistency and, 95–96 organizations and, 361, 372
transition economies and, 355, 360 output collapses and, 404
very rapid, 147 public discussion and, 298–316
Inflation targeting, xv–xvi, 3 rapid privatization and, 349
accountability and, 71, 81 reform and, 362–372
advanced economies and, 72–74 regression analysis and, 372–373
Brazil and, 79–84 role of, 360–362
Calvo and, 71–74, 77, 79–80, 85, 87–89 strong fiscal, 74–77
central banks and, 71, 78, 82–84, 87 strong monetary, 77–79
Chile and, 79–80, 84, 86, 113 substitution and, 98, 107–114
constrained discretion and, 79 transition economies and, 360–374, 372–373
critics of, 79 Institutions Matter (World Bank), 360
currency substitution and, 72–73 Insurance, 96–97, 439
disinflation and, 80 Inter-American Development Bank, xix, 269, 445,
emerging market countries and, 71–90 459–460
exchange rate fluctuations and, 84–87 Interest rates, 80–81, 83, 85
fiscal imbalances and, 75–76 asset price continuity principle and, 60–61
five elements of, 71 asymmetric responses and, 16–17
flexibility and, 78–79 crises recovery and, 217
floating fears and, 73, 80, 85 Devereux-Engel model and, 4
information-exclusive strategy and, 71 dollarization and, 122
institutional development and, 74–79 globalization and, 176
International Monetary Fund (IMF) and, 87–89 independent currency unions (ICUs) and, 122
liability dollarization and, 73, 74 Lahiri-Singh-Végh model and, 26–27
prices and, 71 local-currency pricing (LCP) and, 4–10
reform and, 77 macroeconomic pessimism and, 160
shocks and, 78–79, 82–83, 86 nominal rule, 55–62
stability and, 71, 74–79 Obstfeld model and, 5–18
sudden stops and, 73–75 parity conditions and, 26–27, 62, 107
transparency and, 71, 79 pegging and, 41–42, 51
Information costs, 104–105 recovery and, 218, 231–234
asymmetric information and, 249–250 shock transmission and, 175
cheap talk and, 300–301, 309–310 sudden stops and, 101
Index 481

transitional credit growth and, 332–333 Ireland, 125, 151n10

verification costs and, 439 Isard, Peter, xviii, 295–324, 452
International Disaster Database, 391 Italy, 176
International financial institutions (IFIs), 349–350. Izquierdo, Alejandro, 121, 442
See also Institutions inflation targeting and, 73, 77
building of, 362–372 institution substitution and, 97, 99
role of, 360–362 output collapses and, 379–380, 392, 400
International Financial Statistics (IFS), 134, 272
International Monetary Fund (IMF), xvii, 437, Jaggers, K., 392
452 Jamaica, 317
Brazil and, 81, 84 Japan, xvii, 260
Calvo and, 437–438, 445, 453–456 financial intermediation and, 252
cheap talk and, 300–301, 309–310 globalization and, 172–174, 176, 178, 182–183,
common currency and, 123 186, 193, 198–200
deliberative democracy and, 302 Jazbec, Bostjan, xviii, 327–347
financial intermediation and, 249–250, 268 Johnson, S., 361
Guidelines on Conditionality and, 316–317 Jonas, Jiri, 85
highly indebted poor countries (HIPCs) and, Jones, C. I., 361
320 Journal of International Economics, 457
Independent Evaluation Office (IEO) and, 317– Journal of Monetary Economics, 453
318 Journal of Money, Credit and Banking, 450, 454
inflation targeting and, 87–89 Journal of Political Economy, 446, 454
institution building and, 363, 366–367 J. P. Morgan, 99, 251, 272
nongovernmental organizations (NGOs) and, J-shaped pattern, 108
296, 316
private bond markets and, 268–269 Kaminsky, Graciela L., xvi–xvii, 74, 99
program ownership and, 295–321 Calvo and, 438–439, 443
Public Information Notice and, 311 contract theory and, 439–440
Reinhart and, 456 financial intermediation and, 252
Research Department and, xiii, xviii globalization and, 171–215
social welfare maximization and, 303–308 recovery and, 217
special interest groups (SIGs) and, 298–299, Kehoe, Timothy J., 96, 458
313–315 Keynes, John Maynard, 18, 433–435, 446
stress test approach and, 249 King, Mervyn, 83
sudden stops and, 99 Kiser, S. L., 218
transition economies and, 456 Kisselev, Kate, 112
transparency and, 317 Kiyotaki, N., 337
International Standard Industrial Classification Klein, J. P., 408
(ISIC), 338 Klein, Michael W., 108
Investment Klingebiel, D., 81, 218
arbitrage condition and, 288–289 Klinger, B., 392, 424
capital markets and, 457–460 Kocherlakota, N., 41, 51, 53
capital mobility and, 279–293 Kolmogorov-Smirnov test, 178, 182, 187
crises recovery and, 217–241 Koo, J., 218
domestic production and, 160–161 Koopmans, T., 432, 447–449
labor and, 279–286 Korea, 74, 99, 106, 176, 200, 217, 260
macroeconomic fundamentals and, 225–231 Kraay, A., 361
macroeconomic pessimism and, 159–168 Krogstrup, S., 280
market clearing and, 162 Krugman, P., 217, 280
program ownership and, 295–321 Kuhn-Tucker condition, 34
rational herding and, 105 Kumhof, Michael, xvii, 114, 249–277, 454
systemic risk and, 175–176 Kydland, 449
482 Index

Labor Legal issues

arbitrage condition and, 288–289 government debt and, 264–267
capital mobility and, 279–286 institutions and, 360–362
contract theory and, 434 New York governing law and, 440–441
credit markets and, 332–333, 337 property rights and, 361
education and, 280–284 unanimous action clauses (UACs) and, 440–
efficiency and, 280–281 441
elasticity and, 161 Leiderman, Leo, 100–101, 175, 455–456
hazard rates and, 410 Leisure time, 282
IMF Guidelines and, 317 Lemons problem, 174
market clearing and, 162 Leone, Alfredo M., 88
maximum likelihood function and, 405 Lerrick, Adam, 97, 116
output collapses and, 404–405, 410 Levy-Yeyati, E., 123
productivity and, 160–161 (see also Production) Lewis, A., 171
skilled, 280–284 Li, S., 114
taxes and, 279–286 Liability dollarization, 73–74, 109
Laeven, L., 218 Liberalization, 327, 331
Lagrangians, 34 index of privatization (LIP) and, 354
Lahiri, Amartya, xv, 21–39 transition economies and, 354, 362–372
Lahiri-Singh-Végh model Liberia, 125
cash-in-advance constraints and, 24–25, 34–38 Lichtenstein, 125
equilibrium and, 27–29 Lin, D. Y., 408
fixed exchange rates and, 30–33 Liquidity, 97, 114–115
flexible exchange rates and, 29–33 credit markets and, 333–338
government and, 27 globalization and, 174–175
interest rates and, 26–27 output collapses and, 412 (see also Output
intuition and, 36 collapses)
Kuhn-Tucker condition and, 34 private bond markets and, 268–270
Lagrangians for, 34 transitional credit growth and, 333–334
output shocks and, 32–34, 37 Loans. See Debt
prices and, 23 London School of Economics, 432
traders/nontraders and, 23–27 Long-term capital management (LTCM), 171–
transversality and, 28 172, 175, 178, 200, 203
types of shocks and, 27–28 Loo-Kung, R., 379
utility maximization and, 25–27 Lopez-de-Silanes, R., 267
velocity shocks and, 31–32, 37 Love, I., 339–342
Lancaster, Kel, 432, 449 Lucas, Robert E., Jr., 23, 46, 116, 433–434, 460
Landerretche, Oscar, 80 Lula, 82–84
La Porta, R., 267 Lyons, R., 175, 187, 203
Latin America, xiv. See also Specific country
crises recovery and, 219–222 McCallum, B. T., 41, 51, 53
financial intermediation and, 266–267 McFadden, 396
fiscal health of, 217–222 McLiesh, C., 267, 330
peak growth levels and, 377 Maddala, G. S., 124, 131
Latin American and Caribbean Economic Magendzo, I. Igal, xvi, 121–156
Association (LACEA), 458 Magic Flute (Mozart), 435
Laubach, F., 77, 79 Malaysia, 174, 176, 187, 200, 217
Laxton, Douglas, 72, 77 Manakit, P., 218
Layton, A., 381 Manin, B., 312
Learner, E. E., 134 Mantel, Rolf, 447
Lee, J. W., 218, 392, 435n1 Marimon, R., 249
Leeper, E., 41 Marion, Nancy P., 108
Index 483

Markets transition economies and, 349–376 (see also

access issues and, 21–22 Transition economies)
asset price continuity principle and, 60–61 type of distortion and, 22
asymmetric information and, 249–250 weak-form, 178–191
capital mobility and, 279–293 Marshall, M. G., 392
chronological crisis data on, 204–213 Martinique, 126
common currency effects and, 123 Marty, Alvin, 450
contagion and, 98–106 Masson, P., 77
Council for Mutual Economic Assistance Masten, Igor, xviii, 327–347
(CMEA) and, 355 Mathematics, 446, 448, 452. See also Equations
crashes and, 176 Mathematics for Economists (Allen), 446
credibility and, 95–104 Meade, J. E., 122, 130
credit, 327–345 Mejı́a, L.-F., 379, 392, 400
crises recovery and, 217–241 Meltzer, Allan H., 97, 116
decentralized, 335 Mendoza, Enrique G., xvi, xix, 79
degree of integration and, 182–183 Calvo interview and, 445–461
derivative, 268–270 institutions and, 95–119
double mismatch and, 217 Menze, J., 266
emerging countries and, 71–90 Mercosur countries, 187
equilibrium and, 25–27 (see also Equilibrium) Merrill Lynch, 272
equity and, 99, 112–115, 172, 175, 182, 199, 266, Metzler, L. A., 41
285 Mexico, 75, 110, 113
export flexibility and, 379, 392–394, 400, 403– credibility and, 108
422 crisis of, xvi, 85, 96, 98–99, 103
financial intermediation and, 249–252, 260, 264, financial intermediation and, 252
266–272 globalization and, 173, 175–176, 203
Gini index and, 412 macroeconomic pessimism and, 159
globalization and, 171–215 (see also Globaliza- markets after 1994 crash and, 457–460
tion) Tequila Effect and, 99, 187, 441, 457–460
goods vs. assets, 22 wealth redistribution and, 96
hard currencies and, 97–117 Milesi-Ferretti, Gian Maria, 99
Herfindahl index and, 417 Mill, John Stuart, 121
imperfection in goods, 21–22 Mills, T. C., 381
incomplete insurance, 96–97 Minella, André, 78, 82
inflation targeting and, 71–90 Minoui, C., 380, 384
Lahiri-Singh-Végh model and, 23–38 Mishkin, Frederic S., xv, 71–93
liability dollarization and, 73–74 Mitra, Pritha, xvii, 217–246
liberalization and, 327, 331 Mody, A., 349–376
local-currency pricing (LCP) and, 4–7 Monaco, 122, 125
macroeconomic pessimism and, 159–168 Monetary policy. See Policy
Mundell-Fleming model and, 21–38 Money market, 224, 242–244
Obstfeld model and, 5–18 Moody’s, 200
output collapses and, 377–425 (see also Output Moore, J., 337
collapses) Mora, R., 381
pegging and, 55–62 Moral hazard
prices and, xvi (see also Prices) common currency and, 131–132
private bond, 268–270 institutional development and, 75
segmented asset, 22–38 output collapses and, 408–409, 424–425
stability mechanisms and, 55–62 Morandé, Felipe, 80
strong-form, 191–199 Morocco, 317
sudden stops and, 73–74 (see also Sudden Mozart, Wolfgang Amadeus, 435
stops) Mudambi, R., 381
484 Index

Mundell, Robert, 21, 133, 150, 449, 460 Organisation for Economic Co-operation and
Mundell-Fleming model, xiv–xv Development (OECD), 125, 289
Calvo and, 21 Ostry, J., 123
dollarization and, 21 Outcome equations, 134–135, 147–148
flexible exchange rates and, 22 Output collapses, 32–34, 37
imperfection in goods markets and, 21–22 big, 385–390
Lahiri-Singh-Végh model and, 23–38 causes of, 390–405, 424
open economy and, 21–28 comparative analysis and, 379
perfect mobility and, 21 convergence hypothesis and, 391
shocks and, 22 democracy and, 394, 410, 412
sticky prices and, 21 duration analysis and, 378–390, 394–422, 424
traders/nontraders and, 22 equilibrium and, 387–390
Mussa, Michael, 87 event definition and, 381–385
export flexibility and, 379, 392–394, 400, 403–
Nash equilibrium, 18n6 422
National Bank of Hungary, 85 fixed effects logit specification and, 403
National Monetary Council, 81 frailties and, 408
National Science Foundation, 454 Gini index and, 412
Natural disasters, 391 goodness-of-fit indicators and, 396
Nelson-Aalen cumulative hazard function, 387– hazard rates and, 387–390, 405, 408–425
388 Herfindahl index and, 417
Neumeyer, P., 112 human capital and, 404–405, 410
Neut, Alejandro, xvi, 159–169 inflation and, 396–397, 403
New Keynesian theory, 433 institutions and, 404
New York, 440–441, 453 literature on, 379–381
New York Times, 289, 457 long, 385–390
Nicolini, J. P., 249 maximum likelihood function and, 405
Niepelt, D., 41 natural disasters and, 391
Nobel Prize, xiii, xix, 116 Nelson-Aalen cumulative function and, 387–
Nongovernmental organizations (NGOs), 296, 388
316, 361 peak-trough ratios and, 377, 385
Nonparametric analysis, 145–147 political changes and, 400, 403
Non-Ricardian policy, 47, 52 postwar data and, 380
Norway, 176, 338 predicting, 392, 394–405
Null hypothesis, 178, 182, 187, 191 Prentice-Williams-Peterson model and, 413,
418–421
Oates, W. E., 279–280 pseudo R2 measure and, 396
Obstfeld, Maurice, xiv–xv, 279 recovery from, 405–422
Calvo and, 449, 452–453 regional effects and, 378–379, 385–390
exchange rates and, 3–19 regression analysis and, 390–405, 408–422
fiscal theory of the price level (FTPL) and, 51, 53 short, 385–390
Occam’s razor, 417 small, 385–390
Olivera, Julio H. G., 446–447 statistical approach and, 379, 390–405
Open economies sudden stops and, 378, 392, 394
crises recovery and, 217–241 time-specific effects and, 378–381, 394–405
government debt and, 251–270 unexploited product space and, 392–394
institutional development and, 74–79 unfinished episodes and, 380–381
Lahiri-Singh-Végh model and, 23–38 war and, 378, 380, 387, 390–408
macroeconomic pessimism and, 159–168 Weibull specification and, 410
Mundell-Fleming model and, 21–28 World Development Indicators (WDI) and, 377
ownership issues and, 295–321 Overlapping generations model, 446
Ordinary least squares (OLS) analysis, 148 Oviedo, P., 112
Index 485

Ownership Petersen, M., 340

cheap talk and, 300–301, 309–310 Peterson, A. V., 413
concept of, 295–299 Phelan, C., 41, 51, 53
creation of, 295–296 Phelps, Edmund S., xiii, xix, 460
defined, 295 on Calvo, 431–435
heterogeneous public and, 312–315 Columbia University and, 431–435, 447, 449–450
highly indebted poor countries (HIPCs) and, 320 growth theory and, 447
homogenous public and, 303–308 nominal rigidities and, 453
International Monetary Fund (IMF) and, 295– Phelps-Pollak paper, 434
321 Philippines, 176, 200, 260, 270, 317
nongovernmental organizations (NGOs) and, Poland, 176, 260, 270
296, 316 Policy, xiv, 455
public discussion and, 299–316 Brady Plan and, 439, 456
special interest groups (SIGs) and, 298–299, Calvo’s influence on, xiii–xix, 3, 433–435, 437–
313–315 443, 445
World Bank and, 296–297, 318 capital mobility and, 286–287
central bank independence and, 78, 82–84, 450–
Paasche, B., 112 451
Pakistan, 317 competition, 362–372
Panama, 122, 125, 130 contract theory and, 434, 438–440
Panizza, Ugo, 73 credibility and, 95–104, 107–114 (see also
Papell, D. H., 379–380 Credibility)
Pareto optimality, 285 crises recovery and, 217–241
Park, Y. C., 218 debt forgiveness and, 166
Parkin, M., 433 debt rescheduling and, 440–442
Parrado, E., 122 deposit rescheduling and, 438–439
Parsley, David, 99 emerging vs. advanced economies and, 72–74
Pegging, 63 exchange rates and, xiv (see also Exchange rates)
asset price continuity principle and, 60–61 expansionary, 166
Calvo and, 121 fiat money and, 41, 45
CFA franc and, 127 financial intermediation and, 249–252, 260, 264,
Chile and, 80 266–272
contagion and, 98–106 fiscal theory of the price level (FTPL) and, 41–64
credibility and, 97–104, 107–114 floating fears and, 73
dollarization and, 121–122 foreign aid and, 349–376
equilibrium and, 55–62 future and, 460–461
fiscal theory of the price level (FTPL) and, 41– hard currencies and, 97–117
42, 51, 55–62 IMF and, 87–89 (see also International Monetary
hard vs. soft, 121 Fund (IMF))
independent currency unions (ICUs) and, 121– inflation targeting, 71–90
129, 133, 139, 143, 146–150 institutional development and, 74–79
interest rates and, 41 Lahiri-Singh-Végh model and, 23–38
nominal rule, 55–62 legitimacy and, 305
nonparametric analysis and, 145–147 liability dollarization and, 73–74, 109 (see also
sudden stops and, 98–114 Dollarization)
transition economies and, 355, 360 liberalization, 327, 331, 354, 362–372
volatility and, 137, 139, 143, 146–147, 152n17 macroeconomic pessimism and, 159–168
Penn, 452–453 non-Ricardian, 47, 52
Perri, Fabrizio, 112 optimal rules for, 14–18
Perry, G., 260, 264 pegging and, 41, 55–63 (see also Pegging)
Peru, 176 privatization, 96, 349, 354, 362–372
Pesenti, P., 174 procyclicality and, 15
486 Index

Policy (cont.) macroeconomic pessimism and, 159–168

program ownership and, 295–321 Mundell-Fleming model and, 21–38
public discussion and, 288–316 Obstfeld model and, 5–18
recovery and, 217–241 sticky, 21, 249, 433, 445, 452–457
reform and, 166–167 (see also Reform) sudden stops and, 110, 112 (see also Sudden
Ricardian, 54–55, 65n11, 107 stops)
social welfare and, 303–308 switching effects and, 3–4
time inconsistency and, 95–96 Principles of Political Economy (Mill), 121
transition economies and, 349–376 (see also Pritchett, L., 379
Transition economies) Privatization
transparency and, 79, 317 reform and, 96
unanimous action clauses (UACs) and, 440–441 too-rapid, 349
verification costs and, 439 transition economies and, 354, 362–372
welfare analysis and, 14–16 Production
Political economy capital mobility and, 284–286
budget constraints and, 291–293 Cobb-Douglas function and, 160
equilibrium derivation and, 291–293 constant-returns-to-scale function and, 284–285
globalization and, 279–293 export flexibility and, 379, 392–394, 400, 403–412
model for, 280–293 financial intermediation and, 250
Politics, 451 Herfindahl index and, 417
capital mobility and, 279–289 local-currency pricing (LCP) and, 4–7
Cold War and, 279 macroeconomic pessimism and, 159–168
democracy and, 301–302, 315, 394, 410, 412 market clearing and, 162
globalization and, 279 Obstfeld model and, 5–18
output collapses and, 400, 403 output collapses and, 377–425 (see also Output
program design and, 302 collapses)
special interest groups (SIGs) and, 298–299, 313 Pareto optimality and, 285
voters and, 286–289, 302 sudden stops and, 110, 112 (see also Sudden
Ponzi games, 109 stops)
Portugal, 260 unexploited product space and, 392–394
Posen, A., 77, 79 Property rights, 361
Pou, Pedro, 76, 78 Przeworski, A., 305, 312, 315
Powell, Andrew, 75, 124 Pseudo R2 measure, 396
Prebisch, Raúl, 446 Public discussion
Prentice, R. L., 413 bias and, 300
Prescott, Edward, 449–450 cheap talk and, 300–301, 309–310
Pesenti, P., 217 common ground and, 300–301
Prices, xiv–xvi, 97, 115–116 constraints and, 299
asset price continuity principle and, 60–61 deliberative democracy and, 301–302, 315
credibility and, 452–457 drawbacks of, 315–316
crises recovery and, 223–225 education and, 312
Devereux-Engel model and, 4, 13, 17 functions of, 303–315
exchange rates and, 3 (see also Exchange rates) fuzzy thinking and, 296–297
fiat money and, 41, 45 heterogeneous public and, 312–315
fiscal theory of the price level (FTPL) and, 41–64 homogenous public and, 303–308
Friedman model and, 3–4 institutional capacity and, 299
globalization and, 174–175, 199 need for, 296–297
inflation targeting and, 71 nongovernmental organizations (NGOs) and,
instantaneous adjustments and, 59 296, 316
Lahiri-Singh-Végh model and, 23 program design and, 302–303
lemons problem and, 174 public uncertainty and, 308–312
local-currency (LCP), 4–10 reform and, 298–299
Index 487

signaling and, 305 Reform, 443

social welfare maximization and, 303–308 credibility and, 95–97, 107–114
special interest groups (SIGs) and, 298–299, credit markets and, 327, 335–338
313–315 inflation targeting and, 77
understanding role of, 299–302 institutions and, 362–372
voting and, 302–303 macroeconomic pessimism and, 166–167
Public Information Notice, 311 public discussion and, 298–299
Puerto Rico, 125 special interest groups (SIGs) and, 298–299
Purchasing power parity (PPP), 384 transition economies and, 362–372
Regression analysis
Rabanal, Pau, 88 credit markets and, 339–345
Race-to-the-bottom prediction, 280 exponential parameter and, 412–413, 416–417
Rajan, R., 338–343 fixed effects logit and, 403
Rank-sum test, 187 Gompertz parameter and, 412–413, 416–417
Rational expectations (RE), 460–461 hazard rates and, 408–422
Rawlsian maximin principle, 450 Herfindahl index and, 417
Razin, Assaf, xvii institutions and, 372–373
Calvo and, 449, 450, 454 log-logistic paramter and, 412–413, 416–417
globalization and, 279–294 log-normal parameter and, 412–413, 416–417
institutions substitution and, 99, 101, 107 output collapses and, 390–422
Real estate, 249–250 Prentice-Williams-Peterson model and, 413,
Rebelo, Sergio, 75, 107 418–421
Recession, xviii–xix, 108. See also Output transition economies and, 372–373
collapses Weibull specification and, 410
inflation targeting and, 75, 86 Reinhart, Carmen, xvi–xvii, 74, 454
Keynesian, 434 Calvo papers and, 455
macroeconomic pessimism, 159–168 crises recovery and, 217
Recovery financial intermediation and, 252, 264, 270
aggregate supply and, 224–227 globalization and, 171–215
asset market and, 223–224, 241–244 inflation targeting and, 73, 80, 85
corporate governance and, 218 International Monetary Fund (IMF) and, 456
datasets for, 226–229 private bond markets and, 260
depreciation rates and, 223 sudden stops and, 99–101
Dornbusch model and, 223 switching effects and, 3–4
double mismatch and, 217 Reinstein, A., 269
equilibrium and, 225–231, 241–244 Rhee, C., 269
exchange rate depreciation and, 234–236 Ricardian policy, 54–55, 65n11, 107
goods market and, 224, 241–244 Rigobon, Roberto, 99
growth rates and, 219–241 Risk. See also Hazard rates; Investment
inflation and, 237–238, 241 credibility and, 97–104, 107–114
interest rates and, 231–234 Mendoza-Uribe model and, 107–108
literature on, 217–218, 244n1 systemic, 175–176
macroeconomic fundamentals and, 219–231 Robinson, J., 361
model for, 223–244 Rodriguez, Carlos, 449, 450
money market and, 224, 241–244 Rodriguez, Francisco, xviii–xix, 377–428
output collapses and, 405–422 Rodrik, D., 217, 361, 393, 410, 412, 417
prices and, 223–225 Rogoff, K., 51, 53, 264
private sector and, 218 Roland, G., 330
rapid Asian, 217–218 Roldos, J., 218
simulations of, 225–226, 229–231 Romania, 338
slow Latin American, 218 Rose, A. K., 122–123, 125
Reddy, S. G., 380, 384 Rose, Susan, 432
488 Index

Rothschild, Michael, 445–446 macroeconomic pessimism and, 159–168

Roubini, Nouriel, 112, 174, 217 Mundell-Fleming model and, 22
Rudebusch, G. D., 381 Obstfeld model and, 9
Russia, 106, 114, 338 optimal policy rules and, 14–18
bond default of, 174 output, 32–34, 37 (see also Output collapses)
credit markets and, 331 permanent, 18n7
crisis of, 73–74, 80, 99, 101, 103, 174, 458 signaling and, 106
double mismatch and, 217 sudden stops and, 110, 112 (see also Sudden
financial intermediation and, 252 stops)
GKO government bills and, 178 technology, 9, 13, 15–16
globalization and, 172, 176, 178, 183, 186–187, types of, 22–23
192, 198, 200 velocity, 27–28, 31–32, 37
growth in, 349 world liquidity, 114
rapid privatization and, 349 Sichel, D. E., 381
sudden stops and, 101, 103 Sidrauski, Miguel, 447
transition economies and, 349–376 (see also Signaling, 106
Transition economies) Silk road, 203
Russian virus, 174 Simpson, J., 266
Sims, C., 41
Sachs, J., 134 Singapore, 174–176, 200
Sachs-Warner index, 134 Singh, Rajesh, xv, 21–39
Sack, B., 270 Siotis, G., 381
Sadka, Efraim, xvii, 101, 279–294 Siu, H. E., 249
Sahay, Ratna, xviii, 349–376 Slovakia, 176
Sala-i-Martin, X., 134 Smith, Gregor W., 7
Salomon Brothers, 175 Smith, Katherine A., 97, 109, 112
Samuelson, Paul, 446–447 Sobel, J., 309
Sargent, T., 41, 49, 74, 460 Social welfare, 303–308
Savastano, Miguel, 75, 80–81, 87, 264 South Africa, 73
Savastono, M., 77 Spain, 176, 200
Saxena, S. C., 379 Special interest groups (SIGs), 298–299, 313–315
Scarf, Herb, 447 Spillover, 172–173, 182, 186–187, 200
Schmidt-Hebbel, Klaus, 72, 80 Stability
Schmitt-Grohe, S., 249 Brazil and, 79–84
Schmukler, S., 175, 187, 203 Chile and, 79–80, 84
Schwartz, Gerd, 88 credit markets and, 327
Seignorage, 51, 107–108, 116, 121, 123, 139 emerging market economies and, 95
Senegal, 317–318 hard currency and, 98
Serven, L., 264 inflation targeting and, 71, 74–84
Sharma, S., 77 institutional development and, 74–79
Shleifer, A., 267, 330 Mendoza-Uribe model and, 108
Shocks, xv, xvii rational herding and, 105
asymmetric responses and, 16–17 recovery and, 217–241
credit markets and, 336–337 Tablita plan and, 453
crises recovery and, 217–241 time inconsistency and, 95–96
Dornbusch model and, 223 Standard & Poor’s, 200
globalization and, 171–175, 182–183, 192–194, Stanford University, 432
198–199 Stiglitz, Joseph, 447
inflation targeting and, 78–79, 82–83, 86 Stokes, S. C., 312, 315
information costs and, 105 Stress test approach, 249
local-currency pricing (LCP) and, 5 Sturzenegger, F., 123–124
macroeconomic aggregates and, 110 Subramanian, A., 361
Index 489

Sudden stops, 73–75, 443, 445 Tenev, S., 355

causes of, 100–101 Tequila crisis, 99, 187, 441, 457–460
contagion and, 98–106 Terrorism, 159
credibility and, 97–104 Thaicharoen, Y., 361
equilibrium and, 110, 112 Thailand, xvii
foreign direct investment and, 101 double mismatch and, 217
information costs and, 107–114 financial intermediation and, 252
institution substitution and, 107–114 globalization and, 172–176, 183, 186–187, 192–
interest rates and, 101 193, 198–199
International Monetary Fund (IMF) and, 99 recovery and, 218–219, 221, 226–241
liability dollarization and, 109 Theory of Value (Debreu), 447
output collapses and, 378, 392, 394 Three Essays on the State of Economic Science
rational herding and, 105 (Koopmans), 447
Tequila Effect and, 99 Tille, C., 174
wealth effects and, 107 Time inconsistency, 449–452
Summers, L. H., 121 Tobin, James, 41, 431–432, 435, 447
Svensson, Lars O., 79 Todd, P. E., 124
Swagel, 280 Total factor productivity (TFP) growth, 130
Sweden, 176 Touch of Evil, A (film), 431
Swift, Jonathan, 171 Townsend, R. M., 439
Switzerland, 260 Trade credit, 335–338
Transition economies, 456
Tablita stabilization plan, 453 Council for Mutual Economic Assistance
Talvi, Ernesto, 442 (CMEA) and, 355
hard currency pegs and, 121 credit markets and, 327–345
inflation targeting and, 73, 77 empirical analysis of, 338–345, 353–354
institution substitution and, 97, 99 European Bank for Reconstruction and
output collapses and, 379–380, 392 Development (EBRD) and, 355, 362, 372
Tamiroff, Akim, 435n1 financial sector development and, 328–332
Tanner, Evan, xvii, 249–277 foreign aid and, 349–376
Taxes, xiv, xvii geography and, 354–355
capital income taxation and, 279–293 inflation and, 355, 360
European Union and, 280, 290–291 initial condition effects and, 354–355
evasion of, 125–126 institutional roles and, 360–374
financial intermediation and, 249–250, 273n3 liberalization and, 354, 362–372
fiscal theory of the price level (FTPL) and, 44 output growth and, 332–338, 350–353
harmonization and, 290–291 policy performance and, 355–360
inflation, 45 privatization and, 354, 362–372
institution building and, 362–372 reform and, 362–372
international competition and, 279–280, 289– regression analysis of, 372–373
291 scale of, 349
labor and, 279–286 Soviet bloc and, 349
Mexico and, 110 Transparency
political-economy model for, 280–293 inflation targeting and, 71, 79, 82–83
race-to-the-bottom prediction and, 280 International Monetary Fund (IMF) and, 317
social welfare maximization and, 307–308 volatility and, 137
stochastic distortion and, 97 Transversality,
Taylor, A. M., 279 exchange rates and, 28
Taylor, John, 432–434, 449, 453 fiscal theory of the price level (FTPL) and, 45–
Taylor, L. W., 381 54, 58
Technology, 9, 13, 15–16 Treatment effects model, xvi, 133–134, 146
Teles, P., 249 Trebbi, F., 361
490 Index

Triple threshold model, 341–342 Velocity shocks, 27–28, 31–32, 37

Turkey, 75, 85, 99, 176, 260 Venezuela, 176, 182, 200, 385, 387
Verification costs, 439
Uganda, 307–308 Vietnam, 308, 448
Ukraine, 176, 338 Vishny, R., 267
Unanimous action clauses (UACs), 440–441 Vladkova-Hollar, I., 328
United Kingdom, 150, 176 Volatility, 124, 130
United States asset prices and, 159–168
central bank independence and, 78 common currency and, 137, 139
credit markets and, 339 credit markets and, 330, 337
globalization and, 172, 176, 178, 182–183, 186, openness and, 137
198, 203 pegging and, 137, 139, 143, 146–147, 152n17
government debt and, 251 strict dollarization and, 143
inflation and, 130 Voters, 286–289, 302
interest rate decline in, 101, 103
movable collateral and, 266 Wagner, Rodrigo, xviii–xix, 377–428
shock transmission and, 175 Wallace, N., 41, 49, 74, 460
United States Agency for International Wall Street, 73, 174
Development (USAID), 447 Wall Street Journal, 199
University of Buenos Aires, 447 Walrasian system, 446
University of California, 445 Wars, 448
University of CEMA, 437 output collapses and, 378, 380, 387, 390–408
University of Chicago, 437 World War I era and, 279
University of Maryland, 457, 459 World War II era and, 279
Uribe, Martin, 107, 109, 112, 249 Wealth effects, 107
U.S. Foreign Disaster Assistance (OFDA), 391 Weder, B., 183
Utilitarianism, 450 Wei, L. J., 408
Utility Weibull specification, 410
capital mobility and, 282 Welfare
credibility and, 97–104, 107–114 maximization of, 303–308
Lahiri-Singh-Végh model and, 25–27 optimal policy rules and, 14–16
political-economy model and, 280–293 public discussion and, 303–308
transitional credit growth and, 332–335 Welles, Orson, 431, 435n1
Uzawa, Hirofumi, 448–449 Wellisz, Stan, 449, 452
Werner, Alex, 457
Valderrama, D., 112 Wilcoxon test, 187
Vales, Rodrigo O., 80 Williams, B. J., 413
Value and Capital (Hicks), 446–447 Wolf, H., 123
van Rijckenghem, C., 183 Woodford, M., 41, 52–55, 74
van Rooden, R., 355 Wooldridge, J. M., 124, 396
Van Wincoop, E., 122 World Bank, xviii, 134, 266–267
‘‘Varieties of Capital Market Crises’’ (Calvo), credit markets and, 338
457 duration analysis and, 384
Végh, Carlos A., xv, 107 export data and, 392
Calvo and, 454–455 institutions’ role and, 360–361
exchange rates and, 21–39 ownership and, 296–297, 318
financial intermediation and, 250, 252 private bond markets and, 268–269
growth in transition economies and, 353–355 social welfare maximization and, 307–308
inflation targeting and, 72, 74 transition economies and, 372
pegging and, 41, 122 World Development Indicators (WDI), 134, 338,
recovery and, 217 377
Velasco, Andrés, 21, 96, 112, 159–169, 458 World Pension Association, 264
Index 491

World War I era, 279

World War II era, 279

Yale University, 431–432, 447–448, 460

Yan, I., 114
Yue, Z. V., 112
Yuen, Chi-Wa, 101

Zhang, M. J., 408

Zingales, L., 338–343