The Economic World View

Studies in the Ontology of Economics

The beliefs of economists are not solely determined by empirical

evidence in direct relation to the theories and models they hold.
Economists hold `ontological presuppositions', fundamental ideas
about the nature of being which direct their thinking about economic
behaviour. In this volume, leading philosophers and economists
examine these hidden presuppositions, searching for a `world view' of
economics. What properties are attributed to human individuals in
economic theories, and which are excluded? Does economic man exist?
Do markets have an essence? Do macroeconomic aggregates exist? Is
the economy a mechanism, the functioning of which is governed by a
limited set of distinct causes? What are the methodological implications
of different ontological starting points? This collection, which estab-
lishes economic ontology as a coordinated ®eld of study, will be of great
value to economists and philosophers of social sciences.

Uskali Mäki is Professor of Philosophy at the Erasmus University of

Rotterdam. He is an editor of the Journal of Economic Methodology and
member of the Executive Board of the International Network for
Economic Method. His recent publications include being an editor of
and contributor to The Handbook of Economic Methodology (1998),
Economics and Methodology: Crossing Boundaries (1998), and Fact and
Fiction in Economics (forthcoming).
The Economic World View
Studies in the Ontology of Economics

Edited by
È KI
USKALI MA
published by the press syndicate of the university of cambridge
The Pitt Building, Trumpington Street, Cambridge, United Kingdom

cambridge university press

The Edinburgh Building, Cambridge CB2 2RU, UK
40 West 20th Street, New York, NY 10011±4211, USA
10 Stamford Road, Oakleigh, Melbourne 3166, Australia
Ruiz de AlarcoÂn 13, 28014 Madrid, Spain
Dock House, The Waterfront, Cape Town 8001, South Africa
http://www.cambridge.org

# This collection Cambridge University Press 2001

The book is in copyright. Subject to statutory exception and to the provisions of relevant
collective licensing agreements, no reproduction of any part may take place without the
written permission of Cambridge University Press.

First published 2001

Printed in the United Kingdom at the University Press, Cambridge

Typeset in 10/12pt Times System 3b2 [c e ]

A catalogue record for this book is available from the British Library

Library of Congress Cataloguing in Publication data

The economic world view: studies in the ontology of economics / edited by Uskali MaÈki.
p. cm.
Includes bibliographical references and index.
ISBN 0 521 801761 ± ISBN 0 521 00020 3 (pb)
1. Economics ± Philosophy. 2. Economics ± Methodology. 3. Economic man. 4. Macro-
economics. 5. Microeconomics. I. MaÈki, Uskali.

HB72.E2664 2001
330.1 ± dc21 00±045499

isbn 0 521 80176 1 hardback

isbn 0 521 00020 3 paperback
For KEKLU
 Contents

Notes on the contributors page ix

Preface xv

I The what, why, and how of economic ontology

1 Economic ontology: what? why? how? 3
uskali mäki
2 The empirical presuppositions of metaphysical explanations
in economics 15
harold kincaid
3 Quality and quantity in economics: the metaphysical
construction of the economic realm 32
scott meikle

II Rationality and homo economicus

4 The normative core of rational choice theory 57
russell hardin
5 The virtual reality of homo economicus 75
philip pettit
6 Expressive rationality: is self-worth just another kind of
preference? 98
shaun hargraves heap
7 Agent identity in economics 114
john b. davis
8 Chances and choices: notes on probability and
belief in economic theory 132
jochen runde

vii
viii Contents

III Micro, macro, and markets

9 Essences and markets 157
john o'neill
10 The metaphysics of microeconomics 174
alex rosenberg
11 Ontological commitments of evolutionary economics 189
jack vromen
12 Is macroeconomics for real? 225
kevin d. hoover
13 The possibility of economic objectivity 246
don ross and fred bennett

IV The world of economic causes

14 Ceteris paribus laws and socio-economic machines 275
nancy cartwright
15 Tendencies, laws, and the composition of economic causes 293
daniel m. hausman
16 Economics without mechanism 308
john duprEÂ

V Methodological implications of economic ontology

17 Sargent's symmetry saga: ontological versus technical
constraints 335
esther-mirjam sent
18 Two models of idealization in economics 359
alan nelson
19 The way the world works (www): towards an ontology of
theory choice 369
uskali mäki

Name index 390

Subject index 395
 Contributors

Fred Bennett was, until his recent retirement, Director of Financial

Analysis for Industry Canada. He is now completing a doctorate in
philosophy at the University of Ottawa, and has published articles in
the area of political philosophy.
Nancy Cartwright is Professor in the History and Philosophy of Science
at the London School of Economics (since 1991), as well as Professor
of Philosophy at the University of California at San Diego (since
1998). She is the director of the Centre for the Philosophy of the
Natural and Social Sciences at the LSE and Fellow of the British
Academy. Her research interests include history and philosophy of
physics and economics, causal inference, and objectivity in science.
Cartwright's publications include Nature's Capacities and Their
Measurement (1989), How the Laws of Physics Lie (1983), and The
Dappled World: A Study of the Boundaries of Science (Cambridge
University Press, 1999). She co-authored the book Otto Neurath:
Philosophy between Science and Politics together with J. Cat, L. Fleck,
and T. Uebel (Cambridge University Press, 1995). Research projects in
which she is currently involved include `Measurement in Physics and
Economics'.
John B. Davis, Professor of Economics and International Business,
teaches International Trade and International Economics at Mar-
quette University. He is the author of Keynes's Philosophical Develop-
ment (Cambridge University Press, 1994), editor of New Economics
and Its History (1998), and co-editor of The Handbook of Economic
Methodology (1998). Among the journals he has published in are
Cambridge Journal of Economics, Economic Journal, Review of Poli-
tical Economy, History of Political Economy, Economics and Philos-
ophy, The European Journal of the History of Economic Thought, and
Journal of Economic Methodology. He has been a visiting scholar at
Cambridge University and Duke University. He is President-Elect of

ix
x Notes on the contributors

the History of Economics Society, and has been the editor of the
Review of Social Economy since 1987. Currently he is working on a
book to be published by Routledge on theories of the individual in
economics.
John DupreÂ, formerly Professor of Philosophy at Stanford University, is
currently Professor of Philosophy at Birkbeck College, University of
London, and Senior Research Fellow at the University of Exeter. He
works on the philosophy of science, with special interests in the
philosophy of biology and of economics. He is the author of The
Disorder of Things: Metaphysical Foundations of the Disunity of
Science (1993) and the editor of The Latest on the Best: Essays on
Evolution and Optimality (1987).
Russell Hardin, professor and former chair of politics at New York
University, is the author of Liberalism, Constitutionalism, and Democ-
racy (1999), One for All (1995), Morality within the Limits of Reason
(1988), and Collective Action (1982). He was for many years the editor
of Ethics. He is currently at work on a book on problems of the
rationality and epistemology of ordinary mortals.
Shaun P. Hargreaves Heap teaches at the University of East Anglia. He
has also taught at Concordia University in Montreal and the Sydney
University. His research is in macroeconomics and in philosophy and
economics. His publications include Rationality in Economics (1989),
The New Keynesian Macroeconomics (1992) and Game Theory: A
Critical Introduction (with Y. Varoufakis, 1995) and articles in the
Economic Journal, the Journal of Post-Keynesian Economics and
Kyklos. His current research is on the economics of the media.
Daniel M. Hausman is Professor of Philosophy at the University of
Wisconsin, Madison. His work has focused on the philosophy of
economics and the problem of causation. In addition to numerous
journal articles he has published Capital, Pro®ts and Prices: An Essay
in the Philosophy of Economics (1981), The Inexact and Separate
Science of Economics (Cambridge University Press, 1992), Essays on
Philosophy and Economic Methodology (Cambridge University Press,
1992), Economic Analysis and Moral Philosophy (Cambridge Univer-
sity Press 1996; with Michael S. McPherson), Causal Asymmetries
(Cambridge University Press, 1998), The Philosophy of Economics: An
Anthology (editor) (Cambridge University Press, 1984; 2nd ed. 1994)
and Economic Methodology: Crossing Disciplinary Boundaries (edited
with R. Backhouse, U. MaÈki, and A. Salanti) (1998). He is the former
co-editor of the journal Economics and Philosophy.
Kevin Hoover is Professor of Economics at the University of California,
Davis. He is Chairman of the International Network for Economic
 Notes on the contributors xi

Method and an editor of the Journal of Economic Methodology. He is

the author of numerous articles on topics in monetary economics,
macroeconomics, and economic methodology in journals such as
American Economic Review and the Journal of Economic Literature. He
is the author of The New Macroeconomics (Blackwell), and Causality
in Macroeconomics (Cambridge University Press).
Harold Kincaid is Professor of Philosophy at the University of Alabama
at Birmingham. He received his Ph.D in 1983 from Indiana University.
He is the author of Philosophical Foundations of the Social Sciences
(Cambridge University Press, 1996), Individualism and the Unity of
Science (1997), and numerous articles on topics in the philosophy of
the social sciences.
Uskali MaÈki is Professor of the philosophy of science and of economics
in particular at Erasmus University, Rotterdam. His research interests
cover topics in the philosophy of economics, such as realism, idealiza-
tion, explanation, rhetoric, the sociology and economics of eco-
nomics, and foundations of New Institutional and Austrian
economics. He has published in journals such as the Journal of
Economic Literature, Philosophy of the Social Sciences, Studies in the
History and Philosophy of Science, Economics and Philosophy, Cam-
bridge Journal of Economics, Perspectives on Science. He is a co-editor
of the Journal of Economic Methodology and of The Handbook of
Economic Methodology (1998), Economics and Methodology: Crossing
Boundaries (1998), and Rationality, Institutions and Economic Method-
ology (1993).
Scott Meikle is Reader in Philosophy at the University of Glasgow. He is
the author of Aristotle's Economic Thought (1995) and Essentialism in
the Thought of Karl Marx (1985), and is currently preparing a book on
Marx's relationship to the Anglo-Scottish Enlightenment and the
Aristotelian tradition in European philosophy.
Alan Nelson is Professor and Chair of the Department of Philosophy at
the University of California, Irvine. He has also taught at UCLA,
Pittsburgh, and Stanford, and has been a Fellow of the Zentrum fuÈr
InterdisziplinaÈre Forschung in Bielefeld. He has published widely in
the philosophy of science with special attention to the philosophy of
economics and to seventeenth-century science.
John O'Neill is currently Professor of Philosophy at Lancaster Univer-
sity. He has written widely on the philosophy of economics, ethics,
political philosophy and the philosophy of science. His publications
include The Market: Ethics, Information and Politics (1998) and
Ecology, Policy and Politics: Human Well-Being and the Natural World
(1993).
xii Notes on the contributors

Philip Pettit is Professor of Social and Political Theory at the Research

School of Social Sciences, ANU and also holds a regular Visiting
Professorship in Philosophy, Columbia University, New York. He is
the author of several books including, most recently, The Common
Mind: Essays in Psychology, Society and Politics (1993, 1996) and
Republicanism: A Theory of Freedom and Government (1997, 1999). He
is also the author of a large number of articles, including pieces in
Mind, Philosophical Review, Journal of Philosophy, Ethics, Philosophy
and Public Affairs, Economics and Philosophy and other major jour-
nals. Born in Ireland, he taught at universities there and in Britain
before moving to Australia in 1983.
Alexander Rosenberg is Professor of Philosophy and Director of the
Honors Program at the University of Georgia, Athens. His academic
speciality is the philosophy of science, especially the philosophy of
biology and of social and behavioural sciences. In addition to
numerous journal articles, he has published several books, including
Instrumental Biology or the Disunity of Science (1994), Economics:
Mathematical Politics or Science of Diminishing Returns? (1992),
Philosophy of Social Science (1988, 1995 enlarged edition), The
Structure of Biological Science (Cambridge University Press, 1985),
Hume and the Problem of Causation (1981), and Microeconomic Laws
(1976). He has taught at Syracuse University, the University of
California, Riverside, and has been a visiting professor at the
University of Minnesota and the University of California, Santa
Cruz. In 1994±5, he was a visiting lecturer in philosophy at Oxford
University.
Don Ross is Professor of Economics at the University of Cape Town, and
Convenor of UCT's Programme in Philosophy, Politics and Eco-
nomics. He is the author of over twenty articles, mainly on the
philosophical foundations of economics, of game theory and of
cognitive science, and has published three books. The most recent of
these is What People Want: The Concept of Utility from Bentham to
Game Theory. He is currently writing a book on equilibrium explana-
tion in history.
Jochen Runde is Lecturer in Economics at the Judge Institute of Manage-
ment Studies and Fellow of Girton College, Cambridge. Most of his
previous research has been in the methodology of economics, focusing
particularly on probability, uncertainty and decision theory, causality
and causal explanation, idealization and abstraction, and rational
choice theory. He is currently working on aspects of Austrian
Economics and the Information Theoretic Approach to Economics,
competing conceptions of social structure, tacit knowledge, credit-risk
 Notes on the contributors xiii

assessment in the UK banking industry, and case studies of the

development of large multinational corporations.
Esther-Mirjam Sent is Assistant Professor in the Department of Eco-
nomics and Faculty Fellow in the Reilly Center for Science, Tech-
nology, and Values at the University of Notre Dame. She has
published articles in the Cambridge Journal of Economics, Journal of
Economic Methodology, History of Political Economy, Journal of
Economic Behavior and Organization, and Philosophy of Science,
among other journals. Her book The Evolving Rationality of Rational
Expectations: An Assessment of Thomas Sargent's Achievements (Cam-
bridge University Press, 1998) was awarded the 1999 Gunnar Myrdal
Prize of the European Association for Evolutionary Political
Economy. She has recently started a research project on Herbert
Simon.
Jack J. Vromen is currently Assistant Professor of Philosophy of
Economics and member of EIPE (Erasmus Institute for Philosophy
and Economics) at Erasmus University, Rotterdam. His main research
interests are related to evolutionary and institutional economics. In
addition to articles on these topics, he has published a book Economic
Evolution: An Enquiry into the Foundations of New Institutional Eco-
nomics (1995) which was awarded the 1995 Gunnar Myrdal Prize of
the European Association for Evolutionary Political Economy.
 Preface

What is the world made of ? How does the world work? Is it governed by
causes? What is the nature of causality? Are some states of the world
necessary and some others impossible? Do universals exist? Do minds
exist? Do human beings have a free will? Are there social wholes
irreducible to their parts? These are examples of questions that philoso-
phers and scientists keep asking and answering, and have kept asking
and answering for centuries. They are concerned with the most funda-
mental components in our general world views. Some of the time, these
questions and answers have been explicit, while much of the time they are
implicit. Such questions and answers, at various levels of speci®city, are
ever-present also in economics: they are presupposed and they are
implied in the acts of economic reasoning, both theoretical and empirical.
The purpose of this book is to help start making them explicit. To make
such elements of economic reasoning explicit is to examine the ontology
of economics. It is to study the basic world views of economics.
The economic world view suggests an intellectual challenge from two
perspectives. It tends to be part of our everyday experience that economic
considerations (those related to money and the market, cost and bene®t
calculations, etc.) have been growing in importance in our social lives.
The academic world has also been witnessing, not only the growing
importance of economic constraints on academic life, but also the
growing importance of economics as a discipline, both as a resource for
studying social phenomena and as a target of study itself. In both cases,
questions arise as to what exactly it is that is growing in importance.
What exactly is the economic realm made of ?
I believe the study of economic ontology is a prerequisite for under-
standing economics as a scienti®c discipline. It is thus intended and
hoped that this volume will help de®ne a ®eld of study ± the ontology of
economics, or economic ontology ± so as to give it a relatively distinct
identity with a set of paradigmatic issues and ways of pursuing insights

xv
xvi Preface

into them; to help turn it into a collectively coordinated and cumulative

endeavour. Economic ontology is a ®eld to which both economists and
philosophers of economics and other social sciences can and should
contribute. The parallel recent development in general social theory
proceeds under the heading of `social ontology'; it is a project in which
social scientists and philosophers of society participate. It is hoped that,
in the context of economics and its philosophy, The Economic World
View will inspire more systematic efforts of the same sort in the future.
When Barry Smith, editor of the prestigious philosophy journal The
Monist, a few years ago asked me to guest edit a special issue and to
propose a topic, I did not hesitate long: it had to be on the ontological
foundations of economics. It seemed to me that there was a gap in the
otherwise rapidly growing literature on economic methodology and the
philosophy of economics: a systematic and collectively coordinated
project on economic ontology was missing. Scattered remarks and
contributions had been appearing here and there, but no self-consciously
collective and coordinated effort existed. I suggested the title, `The
Metaphysics of Economics' and started working. I worked quite hard,
and the harvest was good. Eight ®rst-rate papers were published. Those
eight papers, four of them rewritten, are included in the present volume.
It was clear that The Monist project had to be followed up. Editing a
larger volume of contributions was an obvious idea. I spread the
message, and once again the harvest was good. Several people proposed
contributions and started writing. I gave comments on the drafts that
were submitted, and many of the contributors generously served as
anonymous referees on other contributors' papers. In such respects it
was a genuinely collective endeavour. Under the pressure of the pro-
cedure, another ten papers survived, several of them having gone
through many rounds of revision.
There are several people whose help and support I want to acknowl-
edge. Barry Smith's invitation played a crucial role. Patrick McCartan at
Cambridge University Press was very encouraging from the beginning,
while Chris Harrison's role was invaluable at the end. Many of the
contributors have been extremely helpful in reviewing each other's
papers; also, several others acted as referees, assessing the submissions to
The Monist. At the ®nal stages of the project, the assistance of Frank
Hindriks and Judith de Putter has been indispensable. Many thanks to
them all. This project has been supported by the Foundation `Vereniging
Trustfonds Erasmus Universiteit Rotterdam' in the Netherlands and the
Academy of Finland research project `Economics as a Target and Tool
of Science Studies'.
Uskali MaÈki, Rotterdam, May 2000
Part I
The what, why, and how of ecomonic
ontology
1 Economic ontology: what? why?
how?

È KI
USKALI MA

So you're an economist? You study everything that can be gauged by the

measuring rod of money? You view human interaction in terms of supply
and demand in the market? You depict human action as seeking self-
interest in a calculative manner? Is this indeed your view of the world? If
you are an economist and somebody has attempted, or might attempt, to
embarrass you with such questions, you should read this book. If you are
a non-economist inclined to raise such questions, you too should study
the essays collected in this volume. If you are a philosopher interested in
the peculiar characteristics of the `dismal science', this is a book to read.
The book examines aspects of the economic world view from a variety of
perspectives by raising shamelessly deep questions. This is a book in the
ontology of economics.

1 What?

What is the economy made of ? What are its constituents and how do
they hang together? What kind of general principles govern its func-
tioning, and its change? Are they causal principles and, if so, what is the
nature of economic causation? What drives economic actors, and what
mental capacities do they possess? What is utility, or well-being? What is
uncertainty, or risk? Do aggregates exist? Do individual preferences and
social institutions exist, and in what sense? Are (any of ) these things
historically and culturally invariant universals, or are they relative to
context? These are some questions about the economy. The answers that
an economist will provide in response to such questions will include
concepts such as scarcity, rationality, preference, well-being, expectation,
choice, information, probability, strategy, convention, contract, wealth,
division of labour, market, exchange, money, equilibrium, coordination,

3
4 The what, why, and how of economic ontology

mechanism, and so on. These answers are seldom suf®ciently detailed

and re®ned to give us a precise idea of what exactly is presupposed about
the world in posing and answering the questions. The ontology of
economics takes as its task to offer detailed analyses and re®nements of
such basic concepts.
The study of economic ontology is concerned with what may be called
`the economic realm': the economic realm consists of those parts or
aspects of the universe which are set apart as constituting the subject
matter of economics. We may think of the economic realm as being
distinguished from other realms (the physical realm, the biotic realm, the
realm of aesthetics) by virtue of being composed of certain types of
entities or properties ± such as `that part of individual and social action
which is most closely connected with the attainment and with the use of
the material requisites of wellbeing' (Marshall 1920, p. 1) or `human
behaviour as a relationship between ends and scarce means which have
alternative uses' (Robbins 1935, p. 16). This suggests that the economic
realm has de®nite boundaries which separate it from other realms. On
the other hand, the history of the discipline suggests that such boundaries
are not completely sharp and ®xed and that there are rival and evolving
perceptions of how to draw them. The contents and con®nes of the
economic realm are `essentially contestable' as a popular phrase has it,
and they change along with the rest of the discipline of economics (see
Kirzner 1976). The question then arises whether the varying circumscrip-
tions of the economic realm are endeavours to discover something about
the objective constitution of the world, or whether they are subjective
impositions of structure on the social world.
The economic realm is connected to other realms in various ways. The
possibility of the existence of the economic realm presupposes the
existence of the physical, the biotic, and the psychic realms. How exactly
this dependence is to be characterized is itself an ontological issue; is it a
matter of reduction, or supervenience, or is some sort of emergence
involved? The functioning of the economic realm is shaped by the realms
of morality and politics. The exact character of this dependence ± and the
question whether the realms of morality and politics are made of stuff
that is different from that of the economic realm ± are likewise among
some of the major ontological issues related to economics. We may say
that the justi®cation of the discipline of economics requires a claim to
relative autonomy, but this does not yet give us a precise idea of what
such autonomy amounts to.
The contents and con®nes of the economic realm are suggested by the
ontological commitments that economists hold. A set of such commit-
ments has the nature of a Weltanschauung, a fundamental world view
 Economic ontology 5

with a focus on a selection of economically relevant aspects of the world.

Joseph Schumpeter (1954) used the word `vision' to capture very much
the same idea. There is a sense in which those commitments have a
foundational status: they are more fundamental than other elements in
one's belief system in that they are not as easily corrigible and that they
are presupposed rather than derived by the reasoning within the belief
system. The ontological commitments have the character of absolute
presuppositions as R. G. Collingwood called them. He viewed them as the
ultimate and unquestioned presuppositions in terms of which all the
relative suppositions are to be justi®ed (Collingwood 1940). Ludwig
Wittgenstein called them hinges on which our ordinary thoughts and
judgements turn, and the riverbed within which our thought ¯ows
(Wittgenstein 1969, §§97±9, 341±3). Adopting a metaphor from constitu-
tional economics, we may also view the ontological commitments as
providing part of the constitutional framework of a belief system. In this
spirit, James Buchanan refers to elements of a research programme
which are `rarely, if ever, challenged by those scholars who work inside
the intellectual tradition . . . These central elements are taken as pre-
suppositions, as relatively absolute absolutes, and, as such, they become,
themselves, the constraints (the constitution) within which the scienti®c
discourse is conducted' (Buchanan 1991, pp. 13±14). Thus there is
something basic or fundamental about the ontological presuppositions in
our belief systems.
It is in the nature of such a constitutional framework that it determines
the boundaries of a discipline or a line of research. Among the ontolo-
gical or metaphysical core principles of economics, including constitu-
tional economics, Buchanan lists individual autonomy, rationality of
choice, and spontaneous market coordination. `Unless those who would
be participants in the scienti®c dialogue are willing to locate the exercise
in the choice calculus of individuals, there can be no departure from the
starting gate . . . The principle of spontaneous coordination of the
market is the principle of our discipline' (Buchanan 1991, pp. 14, 22).
Buchanan thus takes such principles as de®ning the boundaries of a
research ®eld or programme. Such ontological principles have social
consequences: they effectively help identify the genuine participants in a
programme and exclude others as non-participants. `There is simply no
common basis for scienti®c argument, and ultimately agreement, with
those who choose to perceive social interaction either in purely con¯ic-
tual or purely idealistic visions. These visions are, indeed, alternative
``windows'' on the world' (p. 18). Legitimate membership in the conver-
sations of the economics profession thus requires that the world is being
viewed through shared ontological `windows'.
6 The what, why, and how of economic ontology

Another feature characteristically ascribed to the fundamental com-

mitments is implicitness. The constitutional world view they work with
or the ontological windows through which they view the world are
articulated by practising economists themselves only seldom and incom-
pletely, if at all; such explicit articulation is characteristically not
regarded as one their preoccupations qua economists. In contrast,
making explicit what tends to be implicit is exactly the preoccupation of
this book. Buchanan says that `the process through which individuals
choose among such windows remains mysterious' (p. 18). It is one
purpose of this book to start demystifying the process of choosing such
ontological windows. It is a prerequisite of this demysti®cation that we
understand what those windows are like. One has to start looking at the
ontological windows instead of merely looking through them. This is also
acknowledged by Buchanan in his remark about the insiders and out-
siders: `External intellectual challenges to the whole enterprise tend to be
directed at these elements in the core of the program. The ongoing
research within the constraints can, of course, proceed without concern
for these external criticisms, but practitioners need to be aware of the
core-imposed limits on the persuasive potential of the internalized
analytical exercise' (p. 14). It is a purpose of this volume to help
practitioners in economics be `aware of the core-imposed limits' of what
they can successfully do as such practitioners. A more ambitious goal
would be to try to help economists to make well-argued constitutional
choices, that is, reasoned choices of the constitutional framework within
which they do their work, or of the windows through which they view the
world ± or at least help them to be aware of that larger framework and to
design arguments for the constitution they have ended up working with,
whether as a matter of choice or perhaps as a matter of some path-
dependent academic process. The possibility of constitutional choice
implies that ontological constitutions are not ®xed, that in principle they
can be changed at will. Constitutional change may also ± and often does
± occur without much re¯ection; as Wittgenstein puts it, the riverbed
within which our thought ¯ows can itself change in response to the ¯ow
of thought (Wittgenstein 1969, §97).
In some cases and for some purposes, we had better distinguish
between two aspects of the ontology of economics: the ontological
convictions of an economist and the ontological presuppositions of an
economic theory. Questions about the economic world view can often be
transformed into questions about economic theories, taking on the
general form, `What does theory T presuppose concerning P?' For
example, `What exactly does theory T presuppose about the capacities
and dispositions of economic agents, or of the market mechanism?' A
 Economic ontology 7

re®ned account of the presuppositions of theory T does not necessarily

serve as an adequate account of the ontological convictions of an
economist using T: deep in her heart, the economist may not fully believe
in the world view presupposed by T. This means that the two are
contingently related: they may or may not coincide. In cases where they
do not coincide, there is a tension between them that may function as a
dynamic force driving theory development: the economist may be
motivated by an urge to resolve the tension by attempting to modify the
theory. Chapter 17 by Esther-Mirjam Sent and chapter 19 by Uskali
MaÈki contain discussions of the role that the ontological convictions of
economists play in their attitudes towards theories. Alex Rosenberg's
chapter 10 and Jack Vromen's chapter 11 stress the importance of
distinguishing between economic theories and economists' convictions.
The ambition of articulating and making explicit the ontological
underworld of economic beliefs tends to run counter to a strong reading
of Collingwood's belief that the absolute presuppositions of a belief
system are not and cannot be transparent to those whose thoughts are
based on them. He believed that it is only with hindsight that we can
clearly see what the basic commitments were, in other words that it is
only after the period in intellectual history building upon those elements
is over that we can make those building blocks fully explicit. I admit that
the task of explicating the presuppositions may be easier with hindsight,
but it is one of the (explicit) presuppositions of the endeavour exempli®ed
by the present volume itself that the explication of contemporary
commitments is not impossible. I should perhaps add that the explication
of the presuppositions of theories may be an easier task than the
explication of the ontological convictions of economists.
The attempt to understand economics as a scienti®c discipline requires
the examination of its ontology. Such an examination is also legitimate
from a more general point of view. Ontology is a branch of philosophy
that is once again enjoying a respectable status. It is generally understood
as the study of being as being, of the most general characteristics of all
things, whether actual, possible, or even impossible, whether physical,
mental, or social. The concerns of this book for the most part are not this
abstract or global. The focus is on ontological issues that are character-
istic of economics, whether or not they are shared by other disciplines
and other realms. This is a book devoted to studies in local or regional
ontology. Yet, it is obvious that ontological categories of various levels
of generality are relevant to understanding economics. The articulation
of the world view of economics involves various kinds of categories:
those which are predominantly and speci®cally economical; those that
are more general in their applicability but not universally applicable; and
8 The what, why, and how of economic ontology

those that apply to virtually all domains. This division is somewhat

arti®cial as the fact is that there is a continuum of categories rather than
sharply distinct sets. Thus we get a continuum of sets of categories in an
ascending order of generality, such as the following: money, market,
®rm, price; preference, belief, rationality, choice, rule, welfare; evolution,
aggregate, equilibrium; quality and quantity, probability, essence, causal
power, mechanism, existence, objectivity. The contributions to this
volume characteristically combine such levels. Kevin Hoover's chapter
12 examines the existence of aggregates; chapter 9 by John O'Neill
discusses the essence of markets; Russell Hardin's chapter 4 examines the
normative values at the core of rational choice theory; Jochen Runde
analyses probability and belief in chapter 8; Scott Meikle's chapter 3 sets
out to characterize the economic realm in terms of quality and quantity;
and chapter 13 by Don Ross and Fred Bennett argues for the objectivity
of both micro and macro economics in terms of real patterns.
A ®nal remark on the sense of `ontology' relevant to this book is in
order. `Ontology' and `metaphysics' are often used synonymously, but
the relevant concept of ontology is not given by the familiar pejorative
usages of `metaphysics'. This book is about local or special metaphysics
of an honourable and prestigious kind, of the kind that has recently
enjoyed a resurgence and rehabilitation both in philosophy and in the
special sciences. The present book is not about metaphysics in the logical
positivists' sense of unveri®able nonsense nor in Karl Popper's sense of
unfalsi®able non-science. In economic ontology, the relevant questions
are neither, `Is it meaningful?' nor, `Is it scienti®c?' Using the veri®ability
and falsi®ability criteria, most if not all of science would turn out to be
meaningless and non-scienti®c. The relevant questions in economic
ontology are, `What are the underlying presuppositions?' and, `How do
the presuppositions constrain and determine belief ?' as well as, `How
does and can one justify or criticize the presuppositions?' There is no
absolute break between ontological presuppositions and other claims of
science; there is rather a continuum of claims of various degrees of
generality and corrigibility.

2 Why?

There are several interrelated reasons for why the study of economic
ontology is desirable. One has to do with the boundaries of the discip-
line of economics and of the economic realm. Another has to do with
the grounds of the beliefs that economists hold; their ontological
 Economic ontology 9

commitments are among these grounds, along with conventional em-

pirical evidence. Yet another has to do with the reliability and appro-
priateness of the methods economists use; as a general rule, methods
should match the nature of the reality they are used to investigate. Let
me brie¯y elaborate on these three reasons.
Consider ®rst the issue of boundaries. Supposing one holds that the
economic realm is constituted by categories such as money, the market,
and calculative self-interest, or something weaker, such as instrumental
rationality, the question arises as to the boundaries of the realm so
circumscribed. For example, one may ask, as Shaun Hargreaves Heap
does in chapter 6, whether phenomena related to self-worth (or trust,
care, and esteem) belong to the economic realm by way of reduction to
preferences and self-seeking calculation. Likewise, does the notion of
market apply to science, or religion, and to which aspects of them? And
what exactly does one presuppose about the ultimate nature of the
constituents of certain regions of social life when proposing an economics
of politics and marriage, of crime and church-attendance? Are those
presuppositions defensible as some economists believe, or not, as many
non-economists suspect? How do ontological arguments about the way
the world is constituted bear on the ongoing dispute around `economics
imperialism' ± the expansionistic tendencies of economics in the social
sciences?
The second set of reasons is related to the idea that ontology plays a
role in justifying economists' beliefs and the lines of research they pursue.
The famous Duhem±Quine thesis suggests that empirical tests are unable
to fully determine the merits and demerits of any particular hypothesis or
theory so as to reliably discriminate it from its rivals. All theories or
theory-choices are underdetermined by evidence: there is slack between
theory and empirical evidence. This implies that there must be other
determinants to ®ll in the gap, so to speak. Various social factors have
been proposed to serve as such determinants of beliefs or theory choices,
such as the pursuit of fame and fortune by scientists, or the social
interests of the scienti®c community. The suggestion here, in no way a
novel one, is that ontological commitments are one determinant of
theory choice and theory development in science, economics included.
The general world views held serve as constraints that any successful
theory-candidate has to meet as a minimum condition of acceptability. If
economists view the world through metaphysical windows, this suggests
that there are limits to the power of conventional empirical testing.
Buchanan puts it rather strongly: `How can empirical evidence be made
convincing when such evidence must, itself, be perceived from only one
vantage point at a time? The naãÈvete of modern empirical economists in
10 The what, why, and how of economic ontology

this respect verges on absurdity' (1991, p. 18). If modern economists are

naãÈve in not recognizing the role of ontological windows and riverbeds in
theory choice, the study of economic ontology may help them in growing
up. The role of ontology in shaping the fate of theories is addressed in
chapters 17 and 19 by Sent and MaÈki as well as in Rosenberg's chapter
10.
There is a related thought that is worth exploring. It is a major task of
economic ontology to uncover the ontological presuppositions of
economic theories. The fact that economists hold those theories may then
be at least partly explained by the fact that they hold exactly those
presuppositional commitments. However, as suggested in the previous
section, the situation is not always that simple. It may often be the case
that the ultimate ontological convictions of an economist and the
ontological presuppositions of the theories he holds are not perfectly
tuned with one another. In a situation of such a mismatch between the
ontologies of an economist and of a theory, one way to rectify ontolo-
gical harmony in one's beliefs is to treat the theory instrumentalistically.
The very fact that economists are attracted by an instrumentalist
interpretation of theory may in some cases provide partial evidence for
the existence of such a mismatch. According to the intstrumentalist view,
a theory is not really to be believed to give us a truthful picture of what
the world is like, it is rather to be used as a useful tool for whatever
purposes there may be. Thus the study of economic ontology may help
explain why economists hold certain views of the nature of economic
theory ± and perhaps help them drop those views. Philip Pettit's chapter
5 seeks to show how economists might try to avoid an instrumentalist
view of the notion of homo economicus.
The third reason for examining economic ontology is that theories and
methods need justi®cation and that one way of attempting to justify
them is to argue that they somehow conform to the structure of the
world. The issue around a given method can often be phrased as one
concerning its appropriateness for addressing certain parts or aspects of
reality. Social scientists inspired by hermeneutics argue that quantitative,
formal, and causal modelling is not appropriate for studying the human
realm; some method of understanding the meaningfulness of human
behaviour is required to accommodate the speci®c nature of the human
realm. A weaker line of reasoning would be to admit that there are
causes out there and social phenomena are causally determined, but that
the method adopted by economists from mechanics, namely that of the
decomposition and composition of causes, employing idealizing assump-
tions and their relaxations, is not appropriate in social sciences. This is
because the interactions between the causes are signi®cant enough to
 Economic ontology 11

undermine the model of vector addition which underlies the method.

This is a venerable issue which is often discussed in terms of metaphors
such as those of mechanism, machine, and organism. Such metaphors
can be expressions of major ontological convictions, or, more weakly,
signals of entertaining a way of looking at the social world. Chapters 14,
15, 16, and 18 by Nancy Cartwright, Daniel Hausman, John DupreÂ, and
Alan Nelson address these fundamental issues.

3 How?

One can pursue economic ontology in a variety of ways. They differ from
one another in regard to the aims pursued, the theoretical resources
employed, and the relationships envisaged between the substance of
economics and general ontological doctrines. chapter 2 by Harold
Kincaid deals with some of these issues.
Economic ontology may be practised as an exercise in direct descrip-
tion of what is believed to be the fundamental nature of the economy, or
it may be viewed as an exercise in describing or prescribing the
ontological presuppositions underlying theoretical accounts of economic
phenomena or the ontological convictions held by economists. In the ®rst
guise, it is an attempt to directly represent economic realities, while in the
second mode, it is directly about economics and only indirectly about the
economy.
Another distinction is that between descriptive and revisionary on-
tology (Strawson 1959). We may pursue a description of the ontological
categories and presuppositions that we, or some others, work with.
Given that much of the ontological underworld in our belief systems is
implicit and hidden, the task of articulating it in explicit and systematic
form may not be easy. Descriptive ontology in this sense is a challenging
project. Revisionary ontology, on the other hand, is an attempt to
change, rather than just describe, the prevailing ontological categories
and presuppositions of a belief system: it is a matter of constitutional
revision. Examples of the latter include John DupreÂ's criticism of the
ontology of mechanism in chapter 16; the suggestions to revise the
ontology of economic agency in chapter 7 by John Davis; and Philip
Pettit's re-articulation of the self-seeking economic man in terms of
virtual reality in chapter 5.
Most of the contributions to the present volume belong to descriptive
rather than revisionary ontology. It is to be noted that these two
endeavours do not exclude one another. In particular, if one wants to do
12 The what, why, and how of economic ontology

revisionary ontology, say, in order to replace the prevailing ontological

windows of certain branches of economics, one had better have accom-
plished an analysis in descriptive ontology so as to be in possession of an
adequate account of the categories and presuppositions that underlie
those parts of economics. Some critical writing on economics that has
such a revisionary character is clearly not based on an adequate
descriptive ontology. It is dif®cult to do good descriptive ontology. It is
more dif®cult to do good revisionary ontology, based as it should be on
sound descriptive ontology.
These perspectives are connected with others, such as those of a priori
and a posteriori approaches to economic ontology. One can do
economic ontology in an a priori fashion by starting from some general
philosophical ontology, then try to apply it to parts of economics in
pursuit of description. If the ®t is poor, one may then switch to a
revisionary gear and propose that the ontological presuppositions of
economics had better be changed so as to establish a better ®t between
economics and general ontological doctrine. Economic ontology in this
mode may become an exercise in a priori imposition. Alternatively, one
may proceed by way of empirical case studies in descriptive economic
ontology so as to develop, in an a posteriori fashion, an accurate
account of the actual ontological presuppositions of particular economic
theories and the actual ontological commitments of particular econo-
mists. By comparing these to one another and to the ontologies of other
social sciences and scientists, one may then possibly end up with
identifying interesting mismatches and using them to develop arguments
of a revisionary kind.

4 The menu

Convinced as we are that we will be well nourished in a good company,

we are now asked to sit down to the meal and take a look at the menu.
The next two chapters in this starter section deal with the methodology
and substance of economic ontology. In chapter 2, Kincaid agreeably
argues for a continuity between metaphysical arguments and the sub-
stance of economics, while chapter 3 by Meikle re¯ects on the departure
of the world of economics from the ordinary commonsense world.
The ®rst main course will be served in part II, devoted to the concept
of the economic actor. Given the individualist leanings of much of
economics, the individual actor is a major component amongst the
furniture of the economic world. One's vision of the rest of the economy
 Economic ontology 13

very much depends on the properties attributed to economic actors ± and

so does one's belief in the reality of such actors. Hardin's chapter 4
examines the normative value of the actor's own welfare involved in
rational choice theory and contrasts this with other views. In chapter 5,
Pettit suggests a new way of thinking of the reality of homo economicus
driven by self-regarding desires, given that it runs counter to our
commonsense picture of human beings. Hargreaves Heap, in chapter 6,
argues that economists should not exclude self-worth and expressive
rationality in their representation of economic actors. Chapter 7 by
Davis explores the idea of conceiving of economic actors as genuine
agents, equipped with causal powers which have magnitude and dura-
tion. The ®nal chapter 8 in this section by Runde discusses attributions
of belief and concepts of risk and uncertainty in terms of the ontology of
probability.
Part III provides dining experiences in areas that lie beyond the sphere
of the individual actor, addressing ontological key issues in micro,
macro, and evolutionary economics. In chapter 9, O'Neill explains how
we can safely talk about the `essence' of markets ± and, one might want
to add, of many other things, just as economists do on a daily basis.
Rosenberg's chapter 10 seeks to identify the major components in the
ontology of micro economics, such as equilibrium, individualism, and the
intentional or folk psychological notions of preference and expectation.
In chapter 11, Vromen examines the ontological commitments of the
Nelson and Winter type evolutionary theory and argues that neoclassical
economists such as Alchian and Friedman may share them even if their
theories do not fully implement them. Moving on to the realm of macro,
in chapter 12 Hoover makes the case that aggregates can be viewed as
real once we are careful with the character of aggregates and what their
reality consists in. The ®nal chapter 13 of the section by Ross and
Bennett suggest that the objectivity of the economic realm can be
defended in terms of `real patterns', using examples from macro and
micro economics to make the case.
The chapters of part IV address the deep ontological issues related to
economic causation. They continue the long tradition of thinking of
society in terms of a mechanics of causes and effects. In chapter 14,
Cartwright imagines the economy made of machines equipped with
causal powers that run them. In the next chapter, Hausman raises doubts
about whether the causes will retain their identity and characteristic
tendencies as they act jointly ± which would be required for the
composition of causes. The ®nal chapter 16 of the section by DupreÂ
questions the suitability of the metaphor of mechanism in economics and
argues for a pluralistic ontology for economics.
14 The what, why, and how of economic ontology

The ®nal part V serves the dessert by exhibiting some of the methodo-
logical implications of ontological arguments in economics. Sent's
chapter 17 examines the ontological desideratum of treating economic
agents, economists, and econometricians symmetrically, and traces the
dif®culties met by Thomas Sargent in establishing this symmetry. In
chapter 18 by Nelson, the claim is made that the method of idealization
in economics should not be conceived after the model of Newtonian
mechanics, since economic systems, unlike mechanical systems, do not
satisfy the principle of continuity. The ®nal chapter 19 by MaÈki proposes
an ontology of theory choice where economists' convictions about the
way the world works sometimes function as constraints on what they
accept and reject.

References
Buchanan, James (1991) The Economics and the Ethics of Constitutional Order,
Ann Arbor: University of Michigan Press.
Collingwood, R. G. (1940) An Essay on Metaphysics, Oxford: Clarendon Press.
Kirzner, Israel (1976) The Economic Point of View, Kansas City: Sheed and
Ward.
Marshall, Alfred (1920) Principles of Economics, 8th edition, London: Macmillan.
Robbins, Lionel (1935) An Essay on the Nature and Signi®cance of Economic
Science, 2nd edition, London: Macmillan.
Schumpeter, J. A. (1954) History of Economic Analysis, New York: Oxford
University Press.
Strawson, Peter (1959) Individuals. An Essay in Descriptive Metaphysics, London:
Methuen.
Wittgenstein, Ludwig (1969) On Certainty, Oxford: Blackwell.
2 The empirical presuppositions of
metaphysical explanations in
economics

HAROLD KINCAID

Metaphysical explanations and analyses are common fare among com-

mentators on economics. Philosophers, economists and historians of
economic thought often evaluate, criticize, support or explain the prac-
tice of economics by invoking very general metaphysical facts. In what
follows I raise some skeptical questions about this practice on two
interrelated grounds. First and foremost I argue that analyses of
economics on allegedly general metaphysical grounds often rest on much
more substantive, empirical and contingent matters of fact ± on substan-
tive issues in economics itself ± than their advocates acknowledge.
Second, defending or criticizing economics on general metaphysical
grounds tends to ignore the complexity of economic research. Econo-
mists employ theories for different purposes in different contexts with
different interpretations. Metaphysical diagnoses of success or failure
generally must ignore this diversity and instead wrongly treat modern
economics as a homogeneous whole. The moral I draw from these two
points is not that metaphysical considerations play no role, but that to
understand their role we must pay much more attention to empirical,
local details of economic explanation.
I support these claims by looking at three cases where economics has
been evaluated by appeal to broad metaphysical considerations. Section I
below looks at a common argument among philosophers that goes like
this: Modern economics is a more or less dismal failure by the standards
of the natural sciences. Why is that? Neo-classical theory is essentially
committed to certain predicates or kinds. The best explanation for the
failure of neo-classical theory is that those essential predicates or kinds
do not cut nature at the joints ± in short, they are not natural kinds. This

``The Empirical Presuppositions of Metaphysical Explanations in Economics,'' Harold

Kincaid, The Monist, vol. 78, no. 3, pp. 368±85. Copyright # 1995, THE MONIST, La
Salle, Illinois 61301.

15
16 The what, why, and how of economic ontology

metaphysical argument is quite popular, and my concern shall be with

the argument form, not all its speci®c instantiations.
Section II takes up a related argument, this time one coming from an
economist and one more closely tied to fundamental issues in the ®eld. In
a very interesting history of the productive-nonproductive distinction in
economics, Helen Boss has implicitly endorsed the following argument:
Fundamental to any economic theory is a decision about what phe-
nomena to count as economic (or as production vs. consumption, etc.).
But any such distinction does not cut nature at the joints, for the
economic and social world is characterized by ``netness'' ± a thorough-
going interdependence that makes any such distinctions arbitrary. Thus
every economic theory rests essentially on subjective imposition, largely
determined by the normative views of the economists. I argue that
metaphysical facts such as ``netness'' by themselves do not warrant
Boss's conclusions; at issue, again, are substantive economics questions
undecided by general metaphysical facts. Once again the real issues are
substantive economic ones.
Finally, Section III turns to a different use of metaphysical principles
in economics, namely, to argue for the legitimacy and future prospects of
the neo-classical tradition. Here my target is appeal to individualist
metaphysical principles. A great many economists have taken there to be
(1) some obviously true facts about the relations of individuals to social
wholes that (2) warrant reductionist approaches to theory construction
in economics. I argue that neither is the case. The obvious metaphysical
principles about parts and wholes really rest on quite speci®c and
controversial economic assumptions. Moreover, the truth of those
assumptions by themselves do not support reductionist programs. Only
by the addition of further nontrivial economic assumptions can we get
any plausible implications for how economics can or must proceed.

I
A common view among philosophers of economics is that modern
economics has failed to meet the standards of the natural sciences. The
real puzzle is to explain that fact, particularly given the success with
which economics has applied advanced mathematics to its domain. One
standard explanation is a metaphysical one: the basic categories of
economics do not pick out natural kinds. Hence we hear Rosenberg
claiming that ``reasons and actions are not natural kinds''; that this claim
is an ``inference to the best explanation,''1 an explanation of the fact that
1
Alexander Rosenberg, Sociobiology and the Preemption of Social Science (Baltimore:
Johns Hopkins University Press, 1980), p. 115.
 The empirical presuppositions of metaphysical explanations 17

economic theory has ``not been improved, corrected, sharpened, speci-

®ed, or conditioned . . . None of these things have been done because
they cannot be done.''2 Nelson locates the problem most directly in the
basic notion of a ``commodity'' ± ``economics never gets it right because
commodity is not a natural kind'';3 he takes a similar conclusion to hold
for other key economic concepts as well. Both Rosenberg and Nelson
believe that at least part of the problem with economic kinds is their lack
of universality. Nelson: ``economics is not universal, for we are supposing
that there are societies to which it does not apply in virtue of their not
having any commodities''; hence ``economics is not universal and does
not express relations among natural kinds.''4 Rosenberg: the basic
categories of economics refer to ``states of members of the species Homo
sapiens, states which can only be de®ned . . . by components of a
spatiotemporally restricted particular individual,'' something that must
rule out the universality that scienti®c laws require.5 Similar conclusions
on broadly similar reasoning are reached by others, like DupreÂ, who view
natural science quite differently than does Rosenberg. So Dupre claims
that ``most of economic theory [is] an unpromising candidate for
epistemologically respectable science'' because of its ``indeterministic
metaphysical grounding'' ± in short, because economic categories do not
pick out stable kinds.6
Of course, if economics does not pick out natural kinds, we cannot
expect it to do what good science does: provide laws, predict with
precision and predict the unexpected, unify diverse phenomena, and so
on. We thus have an explanation for why economics fails. These
explanations are indeed metaphysical, for they appeal to facts about the
basic structure of reality to explain what economics can and cannot do.
Arguments of the kind sketched above rest on three basic claims:
(1) that there is a fact that needs to be explained, in this case, the ``fact''
that neo-classical economics has been a dismal failure, (2) that there are
speci®c essential categories of neo-classical theory, and (3) that the
hypothesis that those categories do not pick out natural kinds is a
better explanation for the alleged fact cited in (1) than any alternative

2
Alexander Rosenberg, Economics: Mathematical Politics or Science of Diminishing
Returns (Chicago: University of Chicago Press, 1992), p. 149.
3
Alan Nelson, ``Are Economic Kinds Natural Kinds?'' in Savage, W., ed., Scienti®c
Theories (Minneapolis: University of Minnesota Press, 1990), p. 130.
4
Cited in n. 3 above, p. 129, p. 130.
5
Cited in n. 1 above, p. 131.
6
John DupreÂ, ``Could There Be a Science of Economics?'' in French, P., Uehling, T., and
Wettstein, H., eds., Midwest Studies in Philosophy (Notre Dame, IN: University of Notre
Dame Press), vol. XVIII, pp. 374±5.
18 The what, why, and how of economic ontology

explanation. There are serious obstacles to establishing these claims,

some general and some speci®c. I shall begin with the former.
Any argument about the success or failure of modern economics or
about its essential presuppositions assumes that there is one theory at
issue. That assumption, however, seems questionable. As writers such as
Kuhn have pointed out, focusing on the theory of a given area overly
simpli®es scienti®c practice. For our purposes the most questionable
simpli®cation is the assumption that we can identify the essential
predicates of a theory. Real science seems to be much messier. Different
individuals work with different versions of a theory for different
purposes. Details are added and dropped, basic notions are given
different readings, and so on. Economics seems to be no exception in this
regard. Sometimes these differences come because a shared but very
general and abstract formulation of neo-classical theory has to be
interpreted to be applied; other times economists work with different
fundamental postulates, no matter how broadly construed. So basic
predicates like rationality get different readings depending on the appli-
cation and economist;7 allegedly fundamental axioms are dropped
depending on the context. For example, Hausman's careful attempt to
sort out the essential components in neo-classical theory identi®ed pro®t
maximization as central, yet an important body of recent work on the
theory of the ®rm is willing to give that up in favor of maximization by
individuals inside the ®rm.8 These dif®culties in identifying essential
predicates multiply when we note that neo-classical theory does not
exhaust modern economics. Much applied work, for example, arguably
proceeds largely independently of the postulates typical of abstract
general equilibrium theory.
We can see how these metamethodological worries come out in
practice by pointing out one implicit assumption at work in Rosenberg's
argument. Rosenberg argues that belief-desire psychology does not pick
out natural kinds and that this explains the failure of modern economics.
That diagnosis makes a crucial assumption: that economics is essentially
about individual behavior. Putting aside the question where macro-
economics ®ts in, for much of neo-classical microeconomics the basic or
primitive entities ± those whose internal structure is left unexplained ±
are aggregates or social entities such as ®rms, households, aggregate
market demand and so on. The fundamental laws of supply and demand

7
See Kenneth Arrow, ``Economic Theory and the Hypothesis of Rationality,'' in Eatwell,
J., Milgate, M., and Newman, P., eds., The New Palgrave (New York: W. W. Norton).
8
See Daniel Hausman, The Inexact and Separate Science of Economics (Cambridge:
Cambridge University Press, 1992).
 The empirical presuppositions of metaphysical explanations 19

arguably make claims about markets as a whole and have in large part
been tested by evidence about aggregate quantities. Moreover, at least
some of those tests arguably proceed independently of any very strong
assumptions about individual behavior.9 So even if the categories of
belief-desire psychology were not natural kinds, that would not by itself
explain the alleged failure of economics, at least when it came to the laws
of supply and demand as generally interpreted by economists.
If ``the'' neo-classical theory is a misnomer, then we should have
doubts both about steps (1) and (2) above. Any assessment of the neo-
classical tradition will have to be piecemeal, looking carefully at case-by-
case evidence. Such assessments are impossible without taking sides on
substantive issues in economic research, and it is unlikely that this
diverse work is a success or failure across the board. Thus there is
unlikely to be any general fact to explain or any single factor that
explains all.
Even if we were convinced that there was a single entity ``modern
economics'' that was largely a failure and which was committed to a
well-de®ned set of essential categories, I still do not think the metaphy-
sical diagnoses would be reasonable. The evidence for these diagnoses is
that they are supposedly the best explanations of the facts at issue.
However, the evidence for that claim is thin. There are at least the
following problems:
(a) To be the best explanation, a metaphysical diagnosis must rule out or
at least show implausible the obvious competing hypotheses explaining
the failure of neo-classical economics. There are numerous such competi-
tors ± for example, the dif®culties of testing with nonexperimental
evidence, the lack of incentives in the profession for data gathering, the
complexity of the phenomena at issue, the closeness of political and
moral values to the issues economists investigate, the role of ideology,
the relative scarcity of resources compared to the natural sciences, and so
on. These factors and complexes of them have to be ruled out before the
metaphysical account wins the day. That of course is a big task, but it is
one that cannot be avoided if we want to infer to the best explanation.
(b) Metaphysical explanations of scienti®c success and failure have their
own problems. In particular, we have good evidence from the natural
sciences that theories without natural kinds can be successful ± can
produce accurate predictions, explain new phenomena, open up fruitful
domains of inquiry, etc. ± for signi®cant periods of time and that theories

9
As I argue in Philosophical Foundations of the Social Sciences: Analyzing Controversies in
Social Research (Cambridge: Cambridge University Press, 1995), ch. 7.
20 The what, why, and how of economic ontology

with natural kinds can similarly fail. Biology and genetics more speci®c-
ally made enormous strides over a long period of time working with a
notion of the gene that ultimately proved not to be a natural kind; much
was learned about patterns of inheritance and the mechanism of gene
expression despite the fact that there was no single entity corresponding
to the notion of ``gene'' employed. Likewise, theories of continental drift
¯oundered for decades, despite the fact that their basic categories are
now counted as picking out natural kinds. These are just a few of the
many examples from the history of science apparently showing that
successful science does not entail that real kinds have been found and
that ®nding real kinds does not guarantee scienti®c success.10
(c) The notion of natural kinds on which these diagnoses depend is itself
troublesome. Both Rosenberg and Nelson see an important tie between
natural kinds and one criterion for lawfulness, namely, universality. The
idea is that natural kinds ground laws, that laws are separated from
accidental generalizations by the universality of the former, and that
good science is science that produces laws. These claims are philosophi-
cally controversial. Universality in the syntactical sense seems not to
matter at all, since any statement referring to particulars can be trans-
formed into one that does not. Important parts of the natural sciences do
refer to particulars ± geology, evolutionary biology, molecular biology,
ecology and the like refer to processes characterizing this planet. Laws,
moreover, seem neither suf®cient nor necessary for explanation, as an
exhaustive philosophical literature argues.11 No doubt there is some
important role for the distinction between laws and accidental general-
izations, between kinds and other groupings, and between universal and
particular statements. But the moral of the philosophical challenges to
these notions seems to be that we cannot evaluate entire areas of research
by pointing out that their statements do not have the right form ± it takes
a much more detailed look at actual empirical work.
Given this plethora of obstacles, it would be rash to conclude that
lacking natural kinds is the best explanation for the alleged failure of
economics.
Thus every step of the no-natural-kinds diagnosis relies on speci®c
empirical assumptions, ones that are far from necessary and certain.
Those assumptions are both economic and historical: they involve taking

10
See Larry Laudan, ``A Refutation of Convergent Realism,'' Philosophy of Science, vol.
48 (1981), 218±49.
11
See Bas van Fraassen, The Scienti®c Image (Oxford: Clarendon Press, 1980), for a survey
of the standard objections.
 The empirical presuppositions of metaphysical explanations 21

a stand on where and what neo-classical axioms are essential to explana-

tion, on the empirical success of those speci®c applications, on the role of
aggregate vs. individual-level variables, on the relative importance of
various obstacles to economic theorizing, on the relative scope of neo-
classical explanations, and so on. So the claim that economics has no
natural kinds turns out to rest crucially on substantive claims within
economics itself.
The proper moral to draw, however, is not that these no-natural-kinds
diagnoses are a dead end. It is rather that (1) the arguments for them will
have to appeal much more directly and convincingly to speci®c economic
considerations and (2) that any successful claim about a lack of natural
kinds is likely to be much less broad in scope. Thus while I am skeptical
about the claim that ``commodity'' is no natural kind or that economic
predicates do not pick out stable entities, it may well be that less global
claims are reasonable explanations for particular failures in economics.
In applied work on supply and demand there is the constant worry, for
example, about particular commodity groupings ± about when and
where commodities can be aggregated and treated as a single entity such
as ``food.''12 There is also the worry that speci®c empirical ®ndings are
confounded by changes in taste, income, etc. These are worries about
natural kinds and about stability, but when they are justi®ed, it is on the
basis of quite speci®c empirical considerations. Moreover, we should not
expect these debates to be as black and white as the philosophers have
pictured them ± we should expect them to ask whether speci®c economic
variables are stable enough for the purposes at hand, whether some
aggregation procedure allows suf®cient integration given the assumptions
of the study in question, and so on. What is stable or a kind for one set
of purposes in one context with one set of evidence may not be so in
another situation.

II
Philosophers are not, of course, the only ones to offer metaphysical
explanations. Economists offer them as well. In this section I want to
look at some arguments given by an economic historian tracing out the
role of the productive±nonproductive distinction in the history of
economic thought. The arguments again allegedly raise doubts about
economic natural kinds.

12
For an overview of the issues involved, see Angus Deaton and John Muellbauer,
Economics and Consumer Behavior (Cambridge: Cambridge University Press, 1980).
22 The what, why, and how of economic ontology

The productive±nonproductive distinction has a long, if troublesome,

history. In discussing that history up to the present, Helen Boss presents
an argument, metaphysical in nature, for a startling conclusion about
economics. The conclusion is that ``The productive/unproductive distinc-
tion turns out to be more than a trivial error,'' for ``the boundary
problem must be faced by all'';13 yet ``there are as many boundaries as
there are sovereign boundary drawers. There can be no `correct' bound-
ary.''14 So economic theories rest on an ultimate subjective imposition.
The apparent upshot is that the distinction between the economic and
noneconomic is value laden and with it the enterprise of economics
itself.15
Boss reaches this conclusion after convincingly pointing out again and
again the inconsistencies and arbitrariness of speci®c uses of the
productive±nonproductive distinction throughout the history of
economic thought. However, she is ultimately after bigger game and not
content to point out speci®c problems with speci®c economic theories.
There is, Boss implies, a larger moral. Attempts to draw the productive±
nonproductive distinction are doomed to fail, because they commit the
``input±output fallacy'': they treat certain activities as essential to a
productive process and yet deny that those activities are productive
themselves. However, according to Boss, the economic and social world
exhibits ``netness,'' a thoroughgoing interdependence. Attempts to sepa-
rate consumption and production, market and nonmarket, rent seeking
and pro®t seeking, and more generally the economic and noneconomic in
any sharp way fail time and time again, because they require a dichot-
omous distinction between what is continuous and interdependent.
Obviously Boss's argument has much in common with the arguments
of the previous section. Both raise doubts about the scienti®c quali®ca-
tions of economics on the grounds that it does not pick out natural
kinds. Similar reasoning is advanced by the many who describe economic
disputes as paradigm disputes in which order is ultimately imposed on
the economic world, not found.16 Whether they rest on philosophical
premises about intentionality or claims about interdependence in the
social world, all these arguments are of a piece in that they doubt the

13
Helen Boss, Theories of Surplus and Transfer: Pirates and Producers in Economic Thought
(Boston, MA: Unwin Hyman, 1990), p. 5.
14
Ibid., p. 277.
15
I qualify my interpretation here because Boss herself is not entirely clear on how far she
wants to push this claim, on exactly what value ladenness involves, on exactly what her
points say about the prospect of science in economics and so on.
16
See, for example, Richard Wolff and Stephen Resnick, Economics: Marxian versus Neo-
Classical (Baltimore, MD: Johns Hopkins University Press, 1987).
 The empirical presuppositions of metaphysical explanations 23

prospects of economic science because of some alleged very general facts

about the nature of social reality. Economics rests on a dubious
metaphysics.
Boss raises complex and dif®cult issues, ones that I cannot fully sort
out here. However, something less than complete clarity will suf®ce to
show that Boss's pessimistic conclusion is not inevitable and that the real
issues depend upon substantive empirical questions of social research,
not on the broad considerations about economic kinds she cites. Boss's
basic argument is that the social world is an interdependent world, so
that imposing any distinction between the economic and noneconomic or
between production and consumption is arbitrary and ultimately re¯ects
a theorist's value judgements. The premise of this argument is question-
able, and the inference drawn from it does not follow.
To evaluate Boss's claims we must ®rst get clearer on what netness
involves. Some distinctions will help. Economic and noneconomic pro-
cesses might be either causally or, for lack of a better word, existentially
dependent. Causal dependence is a relatively clear notion, at least to the
extent that causality itself is. One entity, event, kind, etc. is causally
dependent upon another just in case its properties are in part causally
produced by the latter. ``Existential'' dependence is a noncausal relation.
The left sides of objects, for example, cannot exist without the right sides
and are thus mutually dependent. Biological processes depend on
chemical processes, but the relation is apparently not one of causation,
for the processes do not involve two separate entities. Both kinds of
dependence can be categorized according to whether they are weak or
strong: strong dependence is a matter of necessity, weak dependence a
contingent rather than inevitable fact. Furthermore, causal dependence
varies along several dimensions. The causal factors involved may be
additive or nonadditive causes. Additive causes are ones where the joint
effect of economic and noneconomic causes are a simple function of their
separate effects. Noneconomic causes can also be minor or integral. In
the former case the economic causes produce most of the effects in
question; in the latter they do not. Obviously netness can take on
different forms depending which of these various categories are involved.
The idea of netness, construed causally, entails Boss's conclusions only
with the help of further assumptions that are controversial and that
involve taking a stand on fundamental empirical matters. If all social
phenomena are amenable to economic explanation, then netness does not
raise problems for drawing a line between the economic and the social,
for the social is economic. Of course, the idea that the economic approach
± as exempli®ed in rational-choice explanations ± can be applied to all
sorts of human behavior is one with many adherents. This economic
24 The what, why, and how of economic ontology

imperialism may ultimately be shown to be a dead end. But if it is, it will

be on empirical grounds. There is no metaphysical inevitability here.
Let us assume that the above empirical question was settled against the
economic imperialists. Nonetheless, netness alone does not entail that
any distinction between the economic and noneconomic is arbitrary, on
either the causal or existential interpretation of netness. The reason is
that neither causal nor existential dependence entails explanatory depen-
dence. To see how this can be, consider ®rst the case where every
economic process is causally dependent on noneconomic processes. If
those noneconomic causes are minor and/or additive, we can legitimately
leave them out of our economic explanations. We can do so because we
can either ignore their effects because they are suf®ciently small or
because we can subtract them. Natural science is predicated on such
practices, and even though ultimately all physical bodies are intercon-
nected by gravity, no one claims that this netness entails that geology or
biology, for example, is based on arbitrary divisions.
Of course, noneconomic causes might be neither minor nor additive.
Yet again there are substantive questions of economic theory at issue
here, questions that cannot be settled simply by pointing out that the
social and the economic are interconnected. In explaining the distribution
of income, for example, we can at least in principle make a distinction
between economic factors like investment in human capital and non-
economic factors like discrimination. Innumerable studies try to subtract
the effects of the latter from the former; there are prima facie plausible
reasons for thinking that in a competitive market the in¯uence of
discrimination will be driven to a minimum. So causal netness gets us to
Boss's subjectivist conclusion only by ruling out these possibilities. Again
substantive economic presuppositions are at work.
In this connection we should note that the ``input-out fallacy'' is no
fallacy at all. Boss's claim is that if a factor is a necessary input to a
productive process, then it must itself be considered productive of
economic value. No doubt many or most attempts to draw the produc-
tive±nonproductive distinction were badly motivated. But the failure was
not one of logic. The gravitational attraction of the Earth is a necessary
precondition for most economic processes, but treating it as non-
economic, nonproductive causal in¯uence is of course eminently reason-
able. One such counterexample is all it takes to show that excluding
necessary factors is no error in reasoning. It takes speci®c and nontrivial
economic arguments to determine which factors in¯uencing the pro-
duction process can be legitimately ignored.
What about existential netness? Does it fare any better in establishing
the arbitrariness of the economic±noneconomic distinction? Apparently
 The empirical presuppositions of metaphysical explanations 25

it does not, for such dependence arguably holds between all the special
sciences and physics. Biological processes, for example, depend upon
physical processes. If we followed Boss's reasoning, we would be led to
the conclusion that biological categories were arbitrary impositions. Of
course we should not swallow that conclusion but rather reject the
reasoning that led to it. Explanation is always qua description, and we
can thus explain one process in multiple and separate ways under
different descriptions. To take Boss's examples, one thing, a highway, or
one activity, work, may be both productive and nonproductive (con-
sumption) at the same time. Highways contribute to lower transportation
costs for corporations and are at the same time consumed by happy
campers on vacation. Work is not just toil and trouble but can also be
fun, a commodity to be consumed. However, there is no inconsistency or
arbitrariness in making such distinctions here, for we might well have
good grounds for believing that qua its contribution to productive
economic output the highway or work is productive and qua its contri-
bution to preference satisfaction it is not. Of course there are practical
obstacles to drawing these distinctions.Yet we have no a priori reason to
think any attempt to make these distinctions is doomed.
Behind Boss's general argument there perhaps lies another argument
that is also worth dispelling, for doing so will add more grist to my mill.
The argument is this:
Specifying any particular economic theory involves a decision about its domain
of application. Different economic theories pick out different domains, and often
there are equally good reasons for these commitments. Thus the sphere of the
economic rests on a fundamental and arbitrary imposition of economic categories
onto reality.

I see this argument in the background of Boss's discussion, because she

ultimately argues that the productive±nonproductive distinction is
closely tied to issues about where to draw the line between the economic
and noneconomic. She argues that every economic theory involves
commitments about the latter, and yet that this line can be drawn in
many different ways, making this fundamental decision about domain a
subjective imposition. This version of Boss's argument is akin to the
arguments of the last section, for it too traces the dif®culties of economic
science back to a lack of objective natural kinds.
A look at the natural sciences shows fairly convincingly that this
argument must be ¯awed. A ®rst hint that something is wrong comes
from the fact that there are cases in the natural sciences where important
categories can be applied in divergent incompatible ways. The concept of
a ``species,'' for example, plays an important role in biology and yet is
26 The what, why, and how of economic ontology

rendered differently by evolutionary biologists, ecologists, and paleontol-

ogists. Arguably this situation results not because two of these groups
are confused, but because the complex set of facts about individual
organisms can be divided up in multiple ways depending on the explana-
tory interests involved.17 This does not mean that biology is inherently
value laden but only that complex causal systems can be individuated in
various ways. That openness does not entail that these individuations are
inherently subjective, for not just any grouping counts as a species and
there are de®nite and motivated criteria for the speci®c species concepts
which biologists use.
A more fundamental objection to this argument is that decisions
about domains can be rationally motivated and need not be value laden
impositions. Any sophisticated look at the history of the natural sciences
will surely show that the object of explanation for most disciplines has
altered in sometimes subtle and sometimes drastic ways. That does not
make natural science categories inherently arbitrary, for domain change
and redescription can be motivated by a piecemeal process that is
rationally warranted at each step.18 So the mere fact that every
economic theory must de®ne its domain does not of itself entail that the
economic±noneconomic distinction is arbitrary and value laden.
Of course, what is not inevitable in theory may be common in
practice. Boss does a terri®c job of exposing many such speci®c miscues
± cases where economists distinguish the productive and nonproductive
or the economic and noneconomic on arbitrary, inconsistent, or other-
wise weakly warranted grounds. Here, however, her arguments are
really economic; they argue from within, rather than from without, and
on substantive economic grounds rather than on some very general
quasi-metaphysical facts about the nature of the social or economic
realm.
Once again, what looked like a metaphysical explanation depends for
its success ± if and when it is successful ± on concrete questions of
economic theory and investigation. It is not the very general metaphy-
sical facts that are doing the work.

III
I want to ®nish my argument by looking at methodological individu-
alism. Methodological individualism is often expressed and/or motivated
by appeal to very general ontological principles, principles that seem
17
See Philip Kitcher, ``Species,'' Philosophy of Science, vol. 51 (1984), 308±33.
18
For a discussion of how this goes, see Dudley Shapere, Reason and the Search for
Knowledge (Dordrecht, Holland: D. Reidel, 1984).
 The empirical presuppositions of metaphysical explanations 27

undeniable. In what follows I try to show that those principles are not
undeniable but rest on quite speci®c and contestable economic assump-
tions and that those principles support individualist methodology only
with the help of other substantive economic assumptions. Moreover,
much modern economics, even that with pristine neo-classical creden-
tials, does not exemplify individualism at all. Once again metaphysical
diagnoses tend to mistakenly treat modern economics as a homogeneous
whole.
Individualism has seemed so powerful a doctrine because it seems
supported by two ``metaphysical commonplaces'': that society is com-
posed solely of individuals and does not act independently of them.19
These metaphysical principles are generally thought to be undeniable and
to have important methodological implications ± for example, that any
macrolevel theory is reducible in principle to individualist accounts.
These two metaphysical commonplaces certainly seem reasonable at ®rst
glance. We can formulate them more precisely by borrowing work from
the philosophy of mind where physicalists claim that the physical
exhausts what there is ± every entity is either an entity described by
physics or some sum thereof ± and that the mental supervenes on the
physical ± in other words, once the physical facts are set, so too are the
mental facts. Transferred to the case of economics or the social sciences
more generally, the individualist metaphysical commonplaces are that
individuals exhaust the economic world and that once all the facts about
individuals are set, then so too are all other aggregate economic facts.
These metaphysical principles are not as trivial as they look. In fact,
they are open to multiple readings, and which reading is correct, if any, is
a matter of substantive economics, not grand metaphysics. As stated, the
exhaustion principle is arguably false for the obvious reason that a
society, an economy, or economic entity such as a corporation is
populated by more than individuals ± it includes material goods.
However, once we recognize that fact, much larger controversies lurk. If
these material goods are means of production, how are they to be
described? In physical-technological terms? As a particular quantity of
capital? As so many hours of embodied labor? How we answer the
question is of course taking sides on some fundamental economic issues,
namely, whether capital can coherently be considered a factor of
production, etc. Thus the apparently obvious principle that society is
made up of individuals turns out to raise deep and controversial
economic issues.

19
As asserted, for example, by J. N. Watkins, ``Methodological Individualism: A Reply,''
in O'Neill, J., ed., Modes of Individualism and Collectivism (London: Heinemann, 1973).
28 The what, why, and how of economic ontology

The supervenience principle is in the same boat. Here we have to ask

what exactly is the supervenience base ± on just what facts about
individuals do the macroeconomic facts depend? Answering that ques-
tion is again taking a position on important substantive economic issues.
Consider, for example, the overall pattern of income distribution, an
aggregate relationship. On what facts about individuals does that aggre-
gate relation supervene? In other words, which individual variables
suf®ce to determine the aggregate relationship? Are human capital
variables suf®cient? Or are factors such as race, sex, sector, and other
more social variables involved? Clearly there are fundamental issues in
economic theory involved here. Moreover, there is nothing special about
this particular example. When, for example, we ask what microeconomic
facts determine the macroeconomic facts we have to take a stance on the
role of expectations and their nature. So the commonplace metaphysical
truth of supervenience is no truism. It can be spelled out in different
ways; which version is correct is a controversial economic issue.
Even if the exhaustion and supervenience principles were obvious
metaphysical truths, the reductionist claims based upon them would
nonetheless require further economic assumptions. The reduction of one
theory to another requires a bridge law showing that the descriptions or
terms in the theory to be reduced are co-extensional with some term or
terms in the more fundamental theory. Moreover, these linkages must be
established in a way that allows the more fundamental theory to replace
the explanations of theory to be reduced. Exhaustion and supervenience
do not entail that such conditions will be met. To use a commonsensical
example, chairs are surely physical objects and the facts about chairs
supervene on the facts about molecules. Yet that does not mean we can
de®ne ``chair'' in physical terms, for there may be inde®nitely many
physical ways to make a chair. To use a philosopher's term, ``chair'' is
multiply realizable. Even when the requisite one-to-one linkage can be
found, it may not suf®ce for reduction. For example, de®ning an anti-
body as ``anything with such and such structure and eliciting an immune
response'' does not suf®ce to reduce immunology to biochemistry, for it
invokes unreduced facts about the immune system. The upshot is that
reducibility requires extra assumptions, ones empirical in nature.
In economics those assumptions again involve speci®c, contingent
assumptions about how the economic world works, assumptions that in
no way follow from some general metaphysical truth. Consider, for
example, what it would take to reduce explanations in terms of corpora-
tions to ones about individual behavior. We would need to show that
there is some ®xed set of ways that individual behavior can bring about
corporate behavior and that we could list and explain, at least in
 The empirical presuppositions of metaphysical explanations 29

principle, all those individual behaviors. It might well be that we will

eventually produce such an account. Yet surely that is an open issue, one
that depends on contingent and currently unknown facts about the
relation between corporate and individual behavior. If we think ®rms
maximize because economic selection has forced them to, then we might
wonder whether there is any ®xed set of individual behaviors that must
be involved, for economic selection of corporations might not ``care''
about individual-level detail, allowing an open-ended number of pro-
cesses at that level. The current diversity of individual-level mechanisms
offered by recent work on the ®rm might further support that prospect.20
At any rate, it should be clear that reducibility here rests on much more
than exhaustion and supervenience; it rests on quite speci®c empirical
matters to be decided by the course of economic inquiry.
Similar points can be made about reduction elsewhere in economics.
Reducing aggregate-level business cycle theory to some less aggregative
account assumes more than simply that lower-level facts ®x higher-level
ones. It also assumes that patterns at the most aggregate level are
brought about by some ®xed set of less aggregative variables. Whether
that holds might well depend on the level of aggregation involved: large-
scale macroeconomic processes might result from some ®xed set of
processes describing the labor market, credit markets, and so on while
those processes themselves might result at different times from inde®-
nitely many distributions of individual preferences, endowments, etc. I
point out these abstract possibilities for one reason: they again make it
clear the metaphysical commonplaces of exhaustion and supervenience
will not by themselves suf®ce to support reduction. It is the subsidiary
economic assumptions that decide the issue.
Mention of corporations and aggregate macroeconomic variables
leads naturally to the second point I want to make about individualist
metaphysics. Commentators inside economics and out identify methodo-
logical individualism as a central tenet of modern economics, sometimes
even economics from its inception. Yet that claim threatens grossly to
oversimplify modern economic practice. Modern economics is not ex-
hausted by neo-classical theory, and much work outside the neo-classical
tradition is explicitly anti-individualist. More important, work that does
fall into the neo-classical tradition is often not individualist in nature.
Microeconomics typically treats the ®rm and the household as funda-
mental entities, unexplained in individualist terms. Much supply-and-
demand analysis is about aggregate market behavior as is the data testing

20
For some off the possibilities, see Oliver Williamson and Sidney Winter, eds., The Nature
of the Firm (New York: Oxford University Press, 1991).
30 The what, why, and how of economic ontology

it; these results neither derive from nor directly support strictly individu-
alist neo-classical accounts of individual consumer behavior. Talk of the
metaphysics behind modern economics once again leads us astray. Thus
we have yet another reason for the conclusion reached above, namely,
that any evaluation of the metaphysical underpinnings of modern or
even neo-classical economics will have to be much more attuned to local
empirical detail.
So appeals to individualist metaphysics, like the metaphysical argu-
ments of the previous sections, carry little force on their own; if they are
compelling, it will be because of other substantive economic facts. Of
course, the three examples discussed in this paper are only a small batch
of evidence. But they should make us wary of the claim that economics
rests on metaphysical ``assumptions that are not themselves subject to
empirical con®rmation.''21 Metaphysical assumptions abound in eco-
nomics, but they are intimately tied to speci®c economic issues and must
be evaluated accordingly.22

References
Arrow, Kenneth (1987) ``Economic Theory and the Hypothesis of Rationality,''
in Eatwell, Milgate, and Newman, The New Palgrave, pp. 25±38.
Boss, Helen (1990) Theories of Surplus and Transfer: Pirates and Producers in
Economic Thought, Boston: Unwin Hyman.
Deaton, Angus and Muellbauer, John (1980) Economics and Consumer Behavior,
Cambridge: Cambridge University Press.
DupreÂ, John (1993) ``Could There Be a Science of Economics?'' in French,
Uehling, and Wettstein, Midwest Studies in Philosophy, Volume XVIII,
pp. 374±5.
(1993) The Disorder of Things: Metaphysical Foundations of the Disunity of
Science, Cambridge, MA: Harvard University Press.
Eatwell, J., Milgate, M., and Newman, P., eds. (1987) The New Palgrave: Utility
and Probability, New York: W. W. Norton.
Fine, Arthur (1986) The Shaky Game: Einstein, Realism and the Quantum Theory,
Chicago: University of Chicago Press.
French, P., Uehling, T., and Wettstein, H., eds. (1993) Midwest Studies in
Philosophy, Volume XVIII: Philosophy of Science, Notre Dame, IN: Uni-
versity of Notre Dame Press.
Hausman, Daniel (1992) The Inexact and Separate Science of Economics,
Cambridge: Cambridge University Press.

21
John DupreÂ, The Disorder of Things: Metaphysical Foundations of the Disunity of Science
(Cambridge, MA: Harvard University Press, 1993).
22
I would like to thank Uskali MaÈki and an anonymous referee for helpful comments on
this paper.
 The empirical presuppositions of metaphysical explanations 31

Kincaid, Harold (1996) Philosophical Foundations of the Social Sciences: Ana-

lyzing Controversies in Social Research, Cambridge: Cambridge University
Press.
Kitcher, Philip (1984) ``Species,'' Philosophy of Science, 51, 308±33.
Laudan, Larry (1981) ``A Refutation of Convergent Realism,'' Philosophy of
Science, 48, 218±49.
Nelson, Alan (1990) ``Are Economic Kinds Natural Kinds?'' in Savage, Scienti®c
Theories, pp. 102±36.
O'Neill, J., ed. (1973) Modes of Individualism and Collectivism, London: Heine-
mann.
Rosenberg, Alexander (1980) Sociobiology and the Preemption of Social Science,
Baltimore: Johns Hopkins University Press.
(1992) Economics: Mathematical Politics or Science of Diminishing Returns,
Chicago: University of Chicago Press.
Savage, Wade, ed. (1990) Scienti®c Theories, Minneapolis, MN: University of
Minnesota Press.
Shapere, Dudley (1984) Reason and the Search for Knowledge, Dordrecht,
Holland: D. Reidel.
van Fraassen, Bas (1980) The Scienti®c Image, Oxford: Clarendon Press.
Watkins, J. N. (1973) ``Methodological Individualism: A Reply,'' in O'Neill,
Modes of Individualism and Collectivism, pp. 166±78.
Williamson, Oliver and Winter, Sidney, eds. (1991) The Nature of the Firm, New
York: Oxford University Press.
Wolff, Richard and Resnick, Stephen (1987) Economics: Marxian versus Neo-
Classical, Baltimore, MD: Johns Hopkins University Press.
3 Quality and quantity in economics:
the metaphysical construction of the
economic realm

SCOTT MEIKLE

There is a feeling today, more or less vague but nonetheless real, that
accountants, economists, and others who deal mainly with money,
inhabit a different world from everyone else.1 This `economic world'
maps roughly onto the ordinary world of things, people, and the
activities of life, but it comes subtly adrift at important points in ways
that produce baf¯ement. Economists often insist that they live in the
`real world', with the heavy implication that their critics and detractors
live in a world of dreams, and though this insistence is sometimes enough
to silence opposition, it is seldom accepted as wholly convincing or
satisfactory.
My purpose here is to argue that this feeling that the economic world
is strange has substantial grounding of a kind not to be be brushed aside
as inexpert `folk economics'. The feeling presupposes a contrast between
a commonsense world and an economic world, and if there is such a
contrast to be drawn, it ought to be possible to say something about
these worlds and how and why they differ. I shall take it that the
commonsense world is the world as it is ordinarily thought of and
spoken about, and by `ordinary' thought and speech I mean those
conducted in terms of the concepts of the natural language we all use.
The economic world, accordingly, is the world as it is thought of and
spoken about using economic concepts.
In order to limit and de®ne the task I shall consider two ordinary
concepts which have fundamental places in our common conceptual
scheme for dealing with the world and its contents, that of a thing and
that of an activity. Ordinary English has been partly penetrated by
economic or market conceptions, but by considering old and funda-
mental concepts there may be some hope of navigating around this

1 I am grateful to Geoff Kay, Pat Shaw, Chris Martin, Uskali MaÈki, Jimmy Lenman, and
David Wiggins for commenting on earlier drafts of this paper.

32
 Quality and quantity in economics 33

obstacle. These ordinary or natural language concepts have logical

features which characterize them and connect them with other concepts
in the scheme, and this tissue of characteristics and connections I shall
refer to as the metaphysics of natural language. In order to ®nd forms of
expression suitable to serve the purposes of economic thought, econo-
mists have, since the eighteenth century when economics ®rst emerged as
a discrete science, reformed the metaphysics of the ordinary concepts of a
thing and an activity, and I shall chart those reforms in sections I and III,
and consider some of the consequences for understanding that they have
had in sections II, IV, and V.
The ordinary concepts of thing and activity, and their corresponding
economic ones of utility and labour, evidently contest the same ground
as elements of alternative or rival ways of handling certain portions and
aspects of the world of ordinary experience. In order to get some grip on
the confusions that arise in this contest, I shall draw contrasts between
modern market thinking and pre-modern thought which did not suffer
from this modern dif®culty. MacIntyre has illuminated aspects of
modernity by means of contrasts with pre-modernity in the philosophies
of knowledge, ethics, and politics, and it may be that the same is possible
in metaphysics and in thought about matters that are today called
`economic'.
Aristotle's philosophy sticks very closely to the metaphysics of natural
language. He saw it as the task of philosophy as he understood it to
discover the logical characteristics of the concepts in common use in the
natural language and the connections they have with other concepts in
the scheme. This is a descriptive conception of the philosophical task
rather than a reforming one, or `revisionary' one as Strawson put it.
Wittgenstein was alarmed at the relaxed habits which mainstream
anglophone philosophy has got into regarding attempts to reform the
metaphysics of natural language, which some in that tradition have
disdained as the `metaphysics of the Stone Age'. Whether or not one is
inclined to share Wittgenstein's general alarm, there is, I think, good
reason for caution about such reforms in the case of economics.
Aristotle's accounts of the ideas of a thing and an activity provide insight
into just what our ordinary ideas of these things are, both in classical
Greek and in English. That is one reason why Aristotle is a useful author
to consider here. Another is that pre-modern European thought was
conducted largely through the medium of the Greek and Latin inherit-
ance, principally the Aristotelian inheritance, and so the Aristotelian
tradition is rich in providing food for thought if one is looking for ways
of getting perspective on familiar modern ideas concerning which, where
they differ from traditional or ordinary ideas, there is a tendency
34 The what, why, and how of ecomonic ontology

unre¯ectively to suppose them to be the results of scienti®c improvement.

A third reason is that Aristotle's own thought in Nicomachean Ethics
(NE) 5.5 and Politics (Pol) I, 8±10, dealing with matters that today are
referred to as `economic', has had a profound in¯uence on the ideas of
modern critics of the market, especially Marx and Keynes.

I
Our ordinary idea of a thing includes the idea that things fall into kinds.
There are, for example, organic things and inorganic things, and each of
these kinds falls into other kinds, and each of them into yet others. This
ordinary idea has been philosophically explicated in the Aristotelian
manner by Wiggins in the following way: things are naturally occurring
entities, or artifacts (substances in Aristotelian jargon), which persist
through change, and whose identities are bound up with the continuity
of the path they trace through space and time, and with their membership
of natural kinds, into which they fall in virtue of their composition,
properties, structure, origin, typical or nomological behaviour, and so
forth. The notion of identity itself (in logical notation `='; the expression
`a = b' says that a and b are the same thing), as Wiggins has argued, is
primitive in the sense that it cannot be strictly analysed or dismantled
into more basic ideas, because there are no ideas more basic than it into
which it may be analysed. Furthermore, the notion of identity, primitive
as it is, presupposes the objects themselves that are the same or different,
together with the kinds into which these objects fall. Everything that we
can talk about or identify is a thing in this, that, or the other kind. By
reference to the kind it can be said what identity and difference criterially
amount to ± what it takes for a and b to coincide or be identical. In the
case of the particulars of ordinary experience the objects and kinds that
are presupposed are continuants or substances and the kinds into which
these fall. With these kinds come ways of coming into being, acting,
interacting, and passing away, in the light of which identities are to be
adjudicated.2
The attempt, beloved of philosophers of empiricist sympathies backed
by Humean anti-substance metaphysics, to analyse identity in terms of
properties (that is, `a = b' if they share all their properties) fails because it
gets things back to front. The fact that a and b share properties follows

2 See D. Wiggins, Sameness and Substance (Oxford 1980), 49±55. For an examination of
what is involved in the disputes over substance between empiricists like Hume and
Russell, and the Aristotelians or rationalists, see D. Wiggins, `Substance', in A. C.
Grayling (ed.), Philosophy: A Guide Through the Subject (Oxford 1995), 214±49.
 Quality and quantity in economics 35

from the identity of a and b (they are the same thing), and the identity is
not something constructed out of that community of properties. Identity
is a primitive notion transcending any philosophical reduction, so it is
not analysable at all, and the best we can do is to uncover and describe
the actual practices and criteria that are used in individuating and
reidentifying things of the different kinds that there are, in order to
enhance the understanding we already have of sameness or identity as
ascribed to things. These practices and criteria vary with the kind of
thing; those for one kind of living entity differ from those of another, and
from those for non-living entities, from those for the different kinds of
artifacts that there are, and from those for such a thing as a republic.
Thus the notion of identity itself involves the notions of the things there
are, and of their kinds, so that the notion of a thing, entity, or substance,
lies at the core of our language, and our equipment for thinking about
reality. It is not an idea that is very likely to be a suitable subject for
reform.
The notion of value in use found in the classical political economists,
Smith, Ricardo, and Marx, is coherent with the ordinary idea of a thing
through the notion of an artifact, or things that exist because we have
made them. The classical notion of value in use is that of a useful thing,
and in the overwhelming majority of cases the thing is designed and
made to have just those qualities which make it useful for a purpose, and
it is said to have value in use in virtue of that fact. `Useful' can be treated
as a two-place predicate, with the form `x is useful for y', where x is a
thing and y a purpose, and the classical notion of value in use ®ts that
form because it is tied to the notion of purpose.3 The purposes served by
useful things are of different kinds, and accordingly these things too fall
into different kinds according to the purposes they serve, for which they
are ®tted in virtue of their natural properties, which properties they were
deliberately constructed to have in order that they might serve those
purposes. So useful things are necessarily heterogeneous, and the kinds
into which they fall are incommensurable.
The classical notion of value in use remained part of the foundation of
thought about matters that today we call `economic' from Aristotle
through to Smith, Ricardo, and Marx. But it became recalcitrant as

3 The full statement of usefulness involves a three-place relation: x is useful to y in respect

of z. Here, however, we are concerned with the purpose which a certain sort of use value
is made in order to serve. A pen (x) is made to be useful for writing (z), and that is the
same for any person y, so as regards the use value of pens, y can be disregarded and the
predicate `useful' can be treated as a two-place predicate. This is not to say that a
particular y might not use a pen for getting coins out of washing-machines.
36 The what, why, and how of ecomonic ontology

economic thought developed in the nineteenth century, and the classical

notion of value in use was eliminated and replaced by the notion of
utility. Bailey makes no mention of useful things, or value in use, in the
early chapters of the Critical Dissertation in which he sets up his
conceptual apparatus.4 Mill, in his Principles of Political Economy,
retains the term, but smudges the distinction between value in use and
price in a way that causes value in use to lose some of its conceptual
independence. He de®nes wealth as `all useful or agreeable things, which
possess exchangeable value'.5 Jevons shifts the focus decisively away
from usefulness in consumption towards usefulness in buying and selling,
and as a consequence an independent notion of value in use is entirely
lost. Jevons was convinced that most of what was wrong with economics
in his time was due to the presence of qualitative notions, and he was
concerned to replace these with quantitative notions wherever it seemed
possible to him. Utility, or what had earlier been called `value in use',
was at the top of his list. He writes, in the preface to The Theory of
Political Economy, that concerning economics
I have long thought that it deals throughout with quantities . . . I have
endeavoured to arrive at accurate quantitative notions concerning Utility, Value,
Labour, Capital, &c . . . Mathematical readers may think that I have explained
some elementary notions, that of degree of utility, for instance, with unnecessary
prolixity. But it is to the neglect of economists to obtain clear and accurate
notions of quantity and degree of utility that I venture to attribute the present
dif®culties and imperfections of the science.6
The classical notion of usefulness, or value in use, was an obstacle to
these ambitions for quanti®cation, and Jevons introduces instead a
notion of usefulness, `utility', which is severed from the idea of things
that are useful for particular purposes, which is a heterogeneous notion
capable of only limited quanti®cation. His notion of utility is that of a
generic `usefulness' which is not discriminated into species, and which
may inhere indiscriminately in things of any kind at all. This is a
`usefulness' that is unconnected with the actual usefulness of a thing, or

4 S. Bailey, A Critical Dissertation on the Nature, Measures, and Causes of Value (London
1825).
5 J. S. Mill, Principles of Political Economy (New York 1969), 9. Mill's work was the
standard textbook until it was replaced by Marshall's Principles of Economics.
6 W. S. Jevons, The Theory of Political Economy (London 1879), vii±viii. Jevons begins the
preface to the second edition by explaining that `certain new sections have been added,
the most important of which are those treating of the dimensions of economic quantities.
The subject, of course, is one which lies at the basis of all clear thought about economic
science. It cannot be surprising that many debates end in logomachy, when it is still
uncertain . . . what kind of quantity utility itself is.' (Italics in original.)
 Quality and quantity in economics 37

with the purpose it was made to serve, and it is indifferent to the physical
constitution of things of that kind in virtue of which they are useful for a
purpose. This undifferentiated `usefulness' is an economic construct
which, unlike the ordinary concept of usefulness, is common to all
things. Usefulness, by this device, had been made into something
uniform, homogeneous, and measurable, just like money or exchange-
value, with which it could then be aligned or confused. The impression is
given that the dif®culty posed for economic thought by the recalcitrant
fact that things fall into incommensurable kinds has been overcome. The
differences of purpose which those things of different kinds are useful
for, have been put out of the picture and replaced by the single purpose
of exchanging them, that is, their usefulness in use has been subordinated
to their usefulness in exchange or buying and selling. Things are regarded
only or primarily as exchangeable items, and one differs from another
only or primarily in the magnitude of its value in exchange.
Jevons's concept of utility is not, as it might appear, a re®nement of
the concept of value in use, in the way that the concept of velocity is a
re®nement of the concept of speed. Before the development of mechanics,
the concept of speed allowed descriptions of the movement of a moving
body, and comparisons of one moving body with another in respect of
speed. But there was no precise way of saying how much speed a moving
thing had, or how much faster its movement was than that of another
moving thing, until the notion of speed was re®ned into the quantitative
notion of velocity, that is, so many units of length per unit of time.
Velocity is a re®nement of the ordinary notion of speed, because there is
a conceptual continuity between the two notions which makes it possible
to say that velocity is a quantitative version of speed. But that is not how
Jevons's notion of utility relates to the classical notion of value in use
and to the ordinary idea of a thing. These are simply different concepts,
and where value in use relates to the constitution of a thing in virtue of
which it is useful for the speci®c purpose which things of that kind are
meant to serve, Jevons's notion of utility does not.
By the time Marshall wrote, some twenty years after Jevons, it had
become possible to dismiss value in use without argument, and a thing's
usefulness came to be understood almost entirely as its usefulness in
buying and selling. Marshall writes:
`The word value' says Adam Smith `has two different meanings, and sometimes
expresses the utility of some particular object and sometimes the power of
purchasing other goods which the possession of that object conveys.' But
experience has shown that it is not well to use the word in the former sense. The
value, that is the exchange value, of one thing in terms of another at any place
and time, is the amount of that second thing which can be got there and then in
38 The what, why, and how of ecomonic ontology

exchange for the ®rst. Thus the term value is relative, and expresses the relation
between two things at a particular place and time.7
Marshall offers no argument, but merely the assertion that there is no
useful place in economics for the idea of value in use, though doubtless
he was right. Wicksteed was more forthcoming: `What we really have to
do is to put out of consideration the concrete and speci®c qualitative
utilities in which they [i.e. useful things] differ, leaving only the abstract
and general quantitative utility in which they are exchanged.'8 The
notion of the usefulness of things for purposes of life other than buying
and selling (Wicksteed's `concrete and speci®c qualitative utilities'), no
longer occupies a position of prominence in economics. In the volumi-
nous index of Schumpeter's History of Economic Analysis it does not rate
an independent entry, though there are two columns of entries under
`Value', all of which concern value in exchange or price.
The classical approach to the analysis of economic value, or price, was
based on the distinction between value in use and value in exchange, and
in turn the notion of value in use was based on the ordinary idea of a
thing. The artifacts that Smith, Ricardo, and Marx were interested in,
were conceived as a sub-class of natural things, whose differentiating
characteristic is that they occur, not by nature, but by virtue of human
efforts or `labour'.9 So products were seen as things continuous with the
natural world, and not as radically separated from it, even though their
chief interest was the money those things could be bought and sold for.
These established conceptual dispositions were redrawn in a revo-
lutionary way with the introduction of the notion of utility. The natural
constitution of things, their having a certain composition, certain proper-
ties, structure, their membership of natural kinds, and their being useful
for speci®c purposes, were now to be `put out of consideration' as
Wicksteed enjoined, and we are not to think of things, but of abstract
entities, utilities. It is, of course, precisely the differences being set aside
here that ordinarily concern people. The trouble with humans, as Evelyn
Waugh's Titoist quartermaster complained about wartime refugees, is
that `They all want sumpin' different'. These differences are the point, in
ordinary life. It is a serious matter to be told that we must think of things
in a way that is different from how we know them to be, and it is

7 A. Marshall, Principles of Economics (London 1898), 8. Bailey had followed the same
course much earlier. In his Critical Dissertation (see n. 4) of 1825, he omits altogether any
consideration of the usefulness of things for any other purpose than that of exchange.
8 Philip A. Wicksteed, The Common Sense of Political Economy (London 1933), ii, 714.
9 For a treatment of artifact identity see Wiggins, Sameness and Substance, ch. 3, sections 3
and 4.
 Quality and quantity in economics 39

something which, if we do it, might be expected to have consequences. In

this economic account of the nature of things, a major point of connec-
tion was severed between the `economic realm' and the natural realm of
things, people, and their activities, as we ordinarily think of them.10
The natural realm, as it will, keeps pushing through these metaphysical
revisions, as it does in the problem of aggregation. Leontief observes that
in trying to deal with the problem of determining the magnitude of
arti®cial aggregative objects such as `output of consumer goods' and
`average price level of agricultural goods', `the reduction in qualitative
variety is attained at the cost of ever increasing quantitative indetermi-
nacy', so that the economist `winds up either with a system of quantita-
tively well-de®ned relationships between qualitatively ill-de®ned
variables or with a set of quantitatively indeterminate ± or at least loosely
described ± relationships between sharply de®ned variables'.11 The
problem of aggregation is insoluble in principle, because the heteroge-
neity of things is not negotiable.
I have discussed nineteenth-century economists because it was they
who engineered the metaphysical reform. Economists have since moved
from cardinal ex post utility to ordinal ex ante utility expressed in
preferences. In doing so they have not reversed the reform, but merely
continued with it, while forgetting that there is a problem about hetero-
geneity and that there was a reform which was meant to solve it.

II
The loss of a clear grasp of the ordinary idea of a thing, which
accompanies this metaphysical reform, has serious consequences, two of
which will be considered in this section. Firstly, it makes dif®cult or
impossible the formulation of an adequate conception of wealth, that is,
a conception suf®ciently rich to allow making the pertinent and intelli-
gible distinction that is in fact made between wealth as money and wealth
as useful things. Secondly, it frustrates framing questions about the end
aimed at by market economy, considered as a system, which are

10 This severence and others like it have come with costs attached. For example, the
practice of calculating GDP by measuring movements of money in market transactions,
to the exclusion of considerations from the ordinary world such as well-being, and the
stock of physical and intellectual capacities, now has increasingly insupportable costs,
even in terms which economics is prepared to countenance. See, for example, Will
Hutton's discussion in The Guardian, 11 March 1996, p. 15.
11 W. Leontief, `The Problem of Quality and Quantity in Economics', in his Essays in
Economics: Theories and Theorizing (Oxford 1966), 55±6, and 46. See also his remarks
on the `production function', 46ff.
40 The what, why, and how of ecomonic ontology

intelligible and which are in fact asked.12 Marx and Keynes held that the
end of the market system is the accumulation of money rather than the
accumulation of useful things, and whether they were right or wrong in
this claim, the distinction between the alternatives is intelligible, perti-
nent, and worthy of consideration. However, if the end of `economic
activity', ambiguous as that expression is, is identi®ed as utility, then it
appears that the distinction cannot be drawn, or at any rate it is not clear
how or whether it can be drawn, and as long as that remains so the
distinction cannot be given consideration, and neither can theories like
those of Marx and Keynes which rest on it.
Keynes retained a grip on the ordinary idea of a thing, and on the
derivative idea of a useful thing. In his discussion of the choice of units in
the General Theory, he writes that the `national dividend', as de®ned by
Marshall and Pigou, `measures the volume of current output or real
income and not the value of output or money-income'. It depends on net
output, that is, `on the net addition . . . to the resources of the commun-
ity available for consumption or for retention as capital stock'. The
distinction he draws between volume of output or real income, and value
of output or money income, is a distinction between wealth as useful
things and wealth as money, and the distinction is the foundation of
those social and economic policies known as Keynesianism. He enters it
as `a grave objection . . . that the community's output of goods and
services is a non-homogeneous complex which cannot be measured,
strictly speaking, except in certain special cases', and he adds with
unconcealed sarcasm that `on this basis an attempt is made to erect a
quantitative science'. The heterogeneity and incommensurability of
things, and other conceptual dif®culties which must be faced by attempts
at making economics a quantitative or exact science, are, he says,
generally dismissed by economists as `conundrums' because businessmen
don't worry about them: `They are ``purely theoretical'' in the sense that
they never perplex, or indeed enter in any way into, business decisions
. . . which are clear-cut and determinate in spite of the quantitative
indeterminacy of these concepts'.13 This is so because business, in
Keynes's view, is concerned above all with wealth as money. He writes:
The distinction between a cooperative economy and an entrepreneur economy
bears some relation to a pregnant observation made by Karl Marx, ± though the
12 I shall resist here the not-very-strong temptation to enter into a discussion of preferences.
There is little reason to acquiesce in the strategy of circumventing discussion of the
possibility that market economy might have a systemic end, by the device of insisting
that ends be discussed only in relation to the preference schedules of individual choosers.
13 J. M. Keynes, The General Theory of Employment Interest and Money (London 1936),
38±9.
 Quality and quantity in economics 41

subsequent use to which he put this observation was highly illogical. He pointed
out that the nature of production in the actual world is not, as economists seem
to suppose, a case of C-M-C', i.e. of exchanging commodity (or effort) for money
in order to obtain another commodity (or effort). That may be the standpoint of
the private consumer. But it is not the attitude of business, which is a case of M-
C-M', i.e. of parting with money for commodity (or effort) in order to obtain
more money.14
Keynes is dealing here with the two types of economy, the `cooperative',
neutral monetary economy, or non-market economy which prevailed
throughout history until modernity, a version of which the Soviets tried
unsuccessfully to establish, and the market economy, or `disembedded'
economy as Polanyi called it.15 It is a distinction that absorbed Keynes.
For the present, however, I want to concentrate on the distinction
between C-M-C, in which the end aimed at is useful things (C), and M-C-
M, in which the end is money (M). Keynes says here that he found the
distinction in Marx, though he might also have got it from Aristotle,
whom he read, and who ®rst made the distinction in Politics book one.16
Marx himself took it from Aristotle and made it the cornerstone of his
analysis of market economy.17 The distinction has far-reaching impli-
cations, most of which Aristotle himself drew at least in outline.
Oikonomikeà (literally `running a household') was Aristotle's name for
the art of providing ourselves with the things we need.18 It `must either
®nd ready to hand, or itself provide, such things necessary to life, and
useful for the community of the family or the polis' (Politics, 1,
1257b27±30). Acquiring them is itself an art, the art of acquisition or
chreÃmatistikeÃ: `Of the art of acquisition then there is one kind which is by
nature a part of oikonomikeà ' (1256b27f.; 1256a10±13). In this art, money
is a means to getting useful things, C-M-C. But there is another kind,
chreÃmatistikeà in the bad sense, which aims, not at acquiring useful things,

14 The Collected Works of John Maynard Keynes (London 1971), xxix, 81

15 See Trade and Market in the Early Empires, ed. K. Polanyi, C. Arensberg, and
H. Pearson (Glencoe, Illinois 1957).
16 On Keynes's relation to Aristotle see A. Fitzgibbons, Keynes's Vision: A New Political
Economy (Oxford 1988), 91, 126, 131, and A. Carabelli, On Keynes's Method (London
1988). See also J. Coates, The Claims of Commonsense: Moore, Wittgenstein, Keynes and
the Social Sciences (Cambridge 1996).
17 See especially A Contribution to the Critique of Political Economy (London 1971), 27, 42,
50, 68, 117, 137, 155, and also Capital, i (London 1976), 65±6, 150±1, 162. The
distinction, and the notation which is due to Marx, are to be found passim in the
Contribution, in Capital, and throughout the three volumes of Theories of Surplus Value.
18 A fuller account of Aristotle's ideas can be found in my Aristotle's Economic Thought
(Oxford 1995). Needless to say, there are neo-classical `economic' interpretations of some
of Aristotle's ideas, and these are examined mainly in chapter 6 of that book.
42 The what, why, and how of ecomonic ontology

but at getting money, M-C-M. It too is concerned with acquisition, and

because of that people confuse it with the ®rst art, but it is really quite
different because its end is different: `The source of the confusion is the
near connection between the two kinds of wealth-getting; in either, the
instrument is the same, although the use is different, and so they pass
into one another; for each is a use of the same property, but with a
difference: accumulation is the end in the one case, but there is a further
end in the other' (1257b34ff.). According to Aristotle's theory of action,
actions are de®ned by their ends, and if two activities have different ends
they are different activities. Natural chreÃmatistikeà (C-M-C) aims at
getting useful things. Unnatural chreÃmatistikeà (M-C-M), which includes
not only trade (kapeÃlikeÃ) but any other activity when it is pursued for the
sake of money, aims at getting money.
Having distinguished these ends, Aristotle must give two de®nitions of
wealth, one as C and one as M. Wealth properly speaking (true wealth,
ho aleÃthinos ploutos, 1256b30f.; or natural wealth, ho ploutos ho kata
phusin, 1257b19f.), is `the stock of things that are useful in the community
of the household or the polis' (1256b30f., and 36±7). Wealth `of the
spurious kind' is money (it is perhaps possible here to glimpse one of the
more practical reasons why the early anglophone moderns, Hobbes,
Locke, & co., found Aristotle so objectionable). The distinction contrasts
sharply with the de®nitions of wealth to be found in the economics that
followed classical political economy, where it became usual in de®nitions
of wealth to elide the distinction between useful things and price. Mill, as
we saw earlier, elides it by de®ning wealth as `all useful or agreeable
things, which possess exchangeable value'. De®ning wealth in economics
nowadays means de®ning and measuring capital, and the issue at stake in
the capital controversy of the 1950s and 1960s was whether wealth
should be de®ned as a stock of heterogeneous goods or as sums of
money, and it may be regarded as a reprise within economics of the
problems Aristotle is tackling here.
Aristotle also distinguishes possession or the capacity to use wealth,
from ownership or the capacity to exchange it. `Wealth as a whole', he
writes in the Rhetoric, `consists in using things rather than owning them;
it is really the activity ± that is, the use ± of the property that constitutes
wealth', and he adds that the `de®nition of security is present possession
in such a way that the owner has the use of the goods, and that of
ownership is the right of alienation, whereby gift or sale is meant' (Rhet.
1361a19ff.). In contrast Marshall, taking no account of the distinction
between possession and ownership, de®nes wealth, not in terms of use,
but in terms of exchange or rights of transference: `a person's wealth'
consists in `those material goods to which he has (by law or custom)
 Quality and quantity in economics 43

private rights of property, and which are therefore transferable and

exchangeable', and `those immaterial goods which belong to him, are
external to him, and serve directly as the means of enabling him to
acquire material goods'.19 Cairncross sees no substantial distinction, but
only one of viewpoint, between wealth and capital, both of which he
regards as involving exchange; `Social capital . . . includes not only trade
capital, but also non-commercial assets that possess a money value . . .
The distinction between social capital and wealth is one of standpoint.
Capital is an agent in production . . . Wealth is a fund upon which we
can draw in consumption'.20 Samuelson gives no de®nition of wealth in
the eighth edition of his textbook Economics.
C-M-C' behaviour begins and ends with useful things. Its aim is to
acquire a useful thing that is needed; once it is acquired, exchange
reaches a natural terminus, and that thing leaves circulation and enters
consumption. M-C-M' behaviour has no natural terminus. In this form
of behaviour `money is the starting-point and the goal' (1257b22f.), and
since there is no difference of quality between one sum of money and
another, the only possible difference being one of quantity, this quanti-
tative growth of money is the only aim that M-C-M' can have. But if M
can be advanced to become M', so can M' be advanced to become M'',
and so on without limit: `there is no limit to the end it seeks; and the end
it seeks is wealth of the sort we have mentioned . . . the mere acquisition
of money' (1257b28f.); `it is concerned only with getting a fund of
money, and that only by the method of conducting the exchange of
commodities' (1257b21ff.); `all who are engaged in acquisition increase
their fund of money without any limit or pause' (1257b33f.). Roll
observes that `Aristotle's long discussion of the two arts of [chreÃmatistikeÃ]
. . . was not just an attempt to drive home an ethical distinction. It was
also a true analysis of the two different forms in which money acts in the
economic process: as a medium of exchange whose function is completed
by the acquisition of the good required for the satisfaction of a want; and
in the shape of money capital leading men to the desire for limitless
accumulation.' 21 In M-C-M' the particular natures of the useful things
falling under C are not especially important because C is not an end, but
merely a means to M, and here it may be possible to discern why the idea
of a useful thing should eventually have proved otiose in modern
economics.
True wealth has a limit because, being the stock of things that are

19 A. Marshall, Principles of Economics, 125.

20 A. Cairncross, Introduction to Economics (London 1960), 67±8.
21 E. Roll, A History of Economic Thought (London [1938] 1961), 35.
44 The what, why, and how of ecomonic ontology

useful in a community, a natural limit is reached when there are enough

of them for the purposes of the citizens in living well and having a
¯ourishing life, `for the amount of property which is needed for a good
life is not unlimited'. There is a limit in the art of politikeÃ, or the running
of a community or politeia, `just as there is in the other arts; for the
instruments of any art are never unlimited, either in number or size, and
wealth may be de®ned as the number of instruments to be used in a
household or in a polis' (1256b32ff.). Wealth is limited, as the means to
any end are limited, and wealth is a set of means for the ¯ourishing of the
citizens. `Limit' (peras) is an important idea in Aristotelian thought, and
it is a serious matter for Aristotle that in the pursuit of wealth as money
`there is no limit of the end it seeks'. True wealth consists in `those goods
. . . necessary for life and useful for the community of the city or
household', and they are limited to those needed to attain the ends of
those communities. The pursuit of wealth as money has no limit imposed
from without because it is not a means subordinate to an end, as the
pursuit of wealth as useful things has, and since it is a quantitative thing
it has no limit of its own, so it has no limit at all. C-M-C' is an institution
or form of behaviour with a limit built into its form. But it is in the
nature of M-C-M' that it has no limit built into its form (`there is no limit
to the end it seeks', 1257b28f.). For that reason those who pursue it are
engaged in a form of activity whose end is of such a kind that it has no
limit. Whether or not they are personally greedy is beside the point; the
point is that the end of the activity they are engaged in has no limit or
terminus.
Wealth, according to Aristotle, is a set of instruments or means for the
activities of life, not the end to which all the activities of life are directed.
If the members of a polis `associated in nothing more than military
alliance and the exchange of goods, this would not be a polis', because a
polis is a partnership in living well (Pol. 3, 1280b17±23, 29±35). The
speci®cation of what it is to live well, of the good life, is derived from an
analysis of human nature, given in De Anima, which provides a basis for
identifying what provisions and arrangements are best for creatures with
the needs and capacities humans have to ¯ourish as things of their kind.
Wealth, or true wealth, is a set of means for that ¯ourishing. Aristotle
means more by `living well' than is meant today by the `standard of
living'. Activities and relationships are the greater part of living well, and
means are subordinate to them as means are to ends.
The Aristotelian theoretical structure, its concepts, distinctions, sup-
porting analyses, and overall conceptual dispositions, has carried great
weight with modern critics of market economy, because it holds out the
possibility, which it itself partly realizes, of an integrated set of de®nitions
 Quality and quantity in economics 45

of wealth, human nature, and human well-being, that are drawn up, not
in economic terms, but in terms that are theoretically independent of
economics. This independence allows such a theory to act as a standard
against which it may be possible to test how well or badly market
economy, and its dedicated science of economics, are doing in serving
human good. Market economy will always pass the test if it is judged by
the accounts of human nature, human good, and wealth, offered by
economics and its associated utilitarian philosophy.

III
Activities are ordinarily thought to have a point or purpose, and as those
purposes differ so activities differ. In Aristotle's theory of action, which
sticks closely to the ordinary idea, this is expressed by saying that actions
are discriminated into kinds by their ends.22 Each kind of activity,
studying mathematics, fattening beef, practising medicine, going for a
walk, administering justice, or playing football, has an end for the sake
of which it is pursued. They are different kinds of activity because they
have different ends, and other differences between them, differences in
the sorts of movements called for, the instruments used, and the blends
of human capacities required, are what they are because of those
differences of end. Some have ends which are distinct from the activities
aimed at bringing them about, as the end of performing a surgical
operation is to make someone well again. The end of others is simply to
engage in the activity, like going for a walk. Activities of different kinds
cannot be added up because they are incommensurable. Although they
all take time, it is an irreducible fact that an hour of studying maths is
not an hour of the same thing as an hour of walking or performing
surgery. They share duration and perhaps some features that are
accidentally common, but that is all, and it isn't much because activities
are distinguished by their ends.
The heterogeneity and incommensurability of different kinds of
activity posed a dif®culty for economic thought in arriving at the kind of
conception of activity or `labour' it needed. Smith had spoken of labour
as a quantity, implying commensurability between kinds, and found the
idea dif®cult but failed to identify the dif®culty.23 Ricardo formulates it
clearly early in chapter 1 of the Principles: `In speaking, however, of

22 See NE 3, 1115b22, and Met. U 1050a22±4. This remains in the Aristotelian tradition.
When Aquinas asks if `human acts receive their species from their end?', he answers that
they do, Summa Theologiae, I±II q. 1 a. 3.
23 A. Smith, Wealth of Nations, ed. E. Cannan (London 1904), vol. i, 33.
46 The what, why, and how of ecomonic ontology

labour, as being the foundation of all value . . . I must not be supposed

to be inattentive to the different qualities of labour, and the dif®culty of
comparing an hour's or a day's labour, in one employment, with the
same duration of labour in another.' But his response is unconvincing:
`The estimation in which different qualities of labour are held, comes
soon to be adjusted in the market with suf®cient precision for practical
purposes.'24 This clearly will not do. The theory was that labour-value
underpins market-value, but arriving at a coherent idea of labour-value
was made dif®cult by the fact that there are different kinds of labour,
which are incommensurable because of the differences in kind. Ricardo,
like Smith before him, tries to surmount the obstacle by introducing
market estimations of different labours, and these are supposed to do the
commensurating. But this is to underpin labour-value with market-value,
and that is the wrong way round for the theory. As long as nothing was
done about this the classical theory of economic value was resting on a
circularity.
Economic thinkers at that time were not so ready to dismiss conceptual
dif®culties as `conundrums', and in seeking a solution the device they
adopted was to adjust the metaphysics of the idea of activity by
eliminating the cause of the problem, kinds. They didn't do things by
halves, and around the reform an entire theory of action was made up, or
adapted from one that had been made earlier, in which there were no
kinds of activities but only individual actions, and only one end to which
all actions were to be considered means. Activities differed, not by end,
but only in their ef®cacy in promoting the single end they were all
supposed to share, utility. Since there was only one end, there could be
only one kind of activity, not many as is ordinarily thought. For that
reason, the speci®c ends or purposes of distinct activities drop out of the
picture. The notion of kinds is removed by the reform, just as the notion
of kinds among useful things was removed by the metaphysical reforms
to the idea of a thing. (In deciding whether this was problem-avoidance
or problem-evasion, account must be taken of the fact that kinds still
thrive in ordinary thought and speech, where the reforms continue to be
entirely ignored.)
So the ground was cleared for the economic version of the idea of
activity. Activities count alike in economics, and they can be added up or
aggregated. `Labour' is regarded, not as a set of activities differentiated
into kinds by differences of end and quality, but as a quantity, indifferent
to ends, whose instances are commensurable and addable because they

24 D. Ricardo, The Principles of Political Economy and Taxation, ed. P. Sraffa (Cambridge
1986), 20. Smith had adopted the same solution in the Wealth of Nations.
 Quality and quantity in economics 47

form a single kind, and are all expenditures of a single undifferentiated

capacity, the capacity for labour. Labour is seen as the negative correlate
of utility, disutility, and as such it is not differentiated into kinds any
more than utility is.
The work of HaleÂvy and Macpherson, among others, shows convin-
cingly enough that modern anglophone utilitarianism was historically an
integral part of the development of economic thinking itself, and that it is
an expression in philosophical idiom of the economic view of the
world.25 Utilitarianism provided philosophical and ethical infrastructure
for the new economic view of the world which had been developing along
with market economy itself in the previous couple of centuries before
Bentham.26 Modern utilitarian philosophy is not neutral in respect of
economics, and its theory of action is inadmissible as a philosophical
witness for the defence in the case that can be brought against economics,
that the world of things and activities it portrays is not the world of
common understanding, that it is at odds with that world and with the
metaphysics of natural language.

IV
The suppression of kinds among activities entails the suppression of the
speci®c ends of activities, and many of the objections made to utilitarian
philosophy and economic thinking have been aimed at positions that are
consequences of this metaphysical reform. The weakness of the ®rst draft
of utilitarianism, subsequently known as act-utilitarianism, was that if
actions fall into only one kind, then actions must be ethically evaluated
individually, not by kinds since there are supposed to be none. This was
soon shown to be an unworkable basis for ethics, if by ethics we mean
something that is supposed to tell you how to live, and rule-utilitarianism
was devised to meet the dif®culty. HaleÂvy's `philosophic radicals' did not
adopt utilitarian ethics because it was consequentialist, but because they
wanted to hold that there was only one kind of action. Earlier versions of
utilitarian thought like that of Protagoras were inspired by other things,
but modern anglophone utilitarianism was inspired by economics.

25 The economic background to this revisionist movement in philosophy is the subject of a

classic treatment by Elie HaleÂvy, The Growth of Philosophic Radicalism (London 1928).
Much of C. B. Macphersons's work is directly or indirectly relevant here, but particularly
his paper `The Economic Penetration of Political Theory', collected in The Rise and Fall
of Economic Justice (Oxford 1987).
26 For an account see ch. xiii, `The Augustan Debate over Land, Trade and Credit', in
J. G. A. Pocock, The Machiavellian Moment: Florentine Political Thought and the
Atlantic Republican Tradition (Princeton 1975).
48 The what, why, and how of ecomonic ontology

More recent concern about utilitarian thinking has focused on its

insensitivity to the real point of activities, and to the ways in which
performances of activities are appraised. This insensitivity follows from
the utilitarian insistence on the commensurability of values, to which
Bernard Williams has drawn attention, and that in turn follows from the
elimination of kinds from among activities.27 The idea of an activity, as
distinct from an action, is an important element in the metaphysical
structure of the ordinary concepts we bring to bear when we think and
speak about the affairs and agency of ourselves and others. The
elimination of kinds tears the idea of an activity out of the scheme of
ordinary concepts and destructures ordinary thought and sensibility
about behaviour. Without it we would lose the idea of a calling or
vocation, and be unable to make discriminations, appraisals, and judge-
ments that are ordinarily thought important.
Performances are appraised as good or bad instances of a kind of
activity, and the criteria applied in such appraisals vary from one kind of
activity to another. The criteria by which a surgeon may be judged to
have done well will probably be unknown to any but medical people, and
they are different from those by which an advocate, architect, musician,
academic, or footballer may be judged to have done well, which in each
case involve subtleties of discrimination that are not quickly or easily
learned. When we don't know or understand the criteria by which
performances of an activity are appraised we are not in a position to
judge them, and we occupy the role of the ignoramus. The utilitarian
theory that there is a single standard for appraising performances in any
kind of activity rationalizes that sort of ignorance. All the world's
greatest hits sound alike on the didgeridoo.
This is a point of practical con¯ict between the ordinary world and the
economic world which has recently become acute as a result of the
introduction of market thinking into the management of non-commercial
activities in the public sector. The ends of some activities have come to be
protected by codes of ethics which preserve those ends and prescribe the
®tting conduct of the activity; for example, the Hippocratic oath,
collegiality and the disinterested pursuit of truth, and others that are less
determinate like the spirit of sport. According to the economic view,
however, kinds have been discontinued and with them the speci®c ends in
relation to which activities fall into kinds. So there is no reason in theory
to acknowledge the existence of codes of ethics which protect speci®c
ends, or to take them seriously, even if there might be practical reasons
for paying them lip-service when managing people who believe in them.

27 See for instance B. Williams, Morality (Cambridge 1972).

 Quality and quantity in economics 49

Practitioners of those arts, in trying to defend those codes of value, the

ends they embody, the kinds of relationships they need, and the supply of
means required for their pursuit, ®nd themselves forced onto the wrong
foot. They are put in the position of having to defend what they see as
the end of the activity they practise, as if it were either (i) a means to a
further end altogether which has no speci®c connection with the activity
or with the human good embodied in its end but only the generic
connection that everything is supposed to have with utility, or (ii) as if it
were part of an all-inclusive end, commensurable with all other parts of
that end, and tradeable with them.28 In this position they ®nd it dif®cult
to argue with the accountants and economists, who see in their case the
expression either of a dream world, or of the self-interest of a particular
interest group sometimes disparaged as `trade-union attitudes'.29 Since
their activities are supposed to be commensurable, the economist is apt
to suppose them to be quanti®able too. Consequently practitioners of
those activities are also often forced into the apologetic dilemma arising
from the enforced quanti®cation discussed by Bernard Williams: `Again
and again defenders of such values are faced with the dilemma, of either
refusing to quantify the value in question, in which case it disappears
from the sum altogether, or else of trying to attach some quantity to it, in
which case they misrepresent what they are about and also usually lose
the argument, since the quanti®ed value is not enough to tip the scale.' 30
Aristotle's chief concern about the introduction of money into Greek
society and culture was precisely its capacity to sow confusion in thought
about ends. Plato had worried about the effects of too much buying and
selling because `trade ®lls the land with wholesaling and retailing, breeds
shifty and deceitful habits in a man's souls and makes the citizens
distrustful and hostile' (Laws, 705a), and Aristotle probably shared those
sentiments. But Aristotle's main worry concerned the effects that M-C-M
has on all the other activities that make up the life of the polis. He
noticed that those not themselves involved in trade tend to use the arts
they practise as a way of getting money, and that this can have an
adulterating effect on the conduct of any kind of activity. His worry
arose from the fact that getting money is a distinct art with a distinct end,

28 J. S. Mill couldn't decide whether such activities were means to happiness or parts of it.
29 A standard professional response often provoked by considerations of this kind mentions
`corporatism', so it is apt to cite Marx: `The bourgeoisie has stripped of its halo every
occupation hitherto honoured . . . It has resolved personal worth into exchange value,
and in place of the numberless indefeasible chartered freedoms, has set up that single,
unconscionable freedom ± Free Trade', The Communist Manifesto, in Marx and Engels,
Collected Works (London 1975±), vol. vi, 487.
30 See B. Williams, Morality, 103.
50 The what, why, and how of ecomonic ontology

and when it becomes connected with the conduct of another art A (as it
always does since there is no distinct activity of making money other
than the coining or printing of legal tender) it does not leave the conduct
of art A unaffected. It introduces another end so that a con¯ict of ends
arises, as a result of which something must happen to the speci®c end of
art A; it can be subordinated and compromised by becoming merely a
means to the end of money-making.31
The general form of the problem can be illustrated with the example of
medicine, not because medicine is any more susceptible to the problem
than any other kind of activity, but just for the sake of an example. The
end of medicine is health. But if social arrangements are such that the
medical profession pursues it also for the sake of money, then health is
no longer its only aim. The practitioners will now be pursuing two ends
at the same time. Those ends can be combined in different proportions
by individual practitioners. In the best case the practitioner will give the
greatest priority to health and the least to money. Even in this case the
aim is still not simply health, but a minimum compromise between health
and the other end. In the worst case the practitioner gives the greatest
priority to money and the least to health. In this case he cannot disregard
health altogether, because the pursuit of money here is parasitic on the
pursuit of health, and there is a threshold in the pursuit of health below
which he cannot go and still effectively use the art for the pursuit of
money. He is using the medical art as a means to another end altogether.
In both the best and the worst cases, and at every point on the spectrum
between them, the practitioners are no longer pursuing health alone, and
they will not always do the same things they would have done if they had
been.
Aristotle criticizes the Sophists for pursuing money rather than
philosophy. What they do looks like philosophy because it `turns on the
same class of things as philosophy', but it differs in its end since the end
of philosophy is not making money, and activities that have different
ends are different however similar they may appear (Metaphysics G,
1004b17ff.). He concludes that what the Sophists do `is, as we said, a
kind of money-making', rather than a kind of philosophy (De Sophisticis
Elenchis, 165a23, 171b28). Many walks of life are intended as targets of
the criticism. Artisans like the maker of the Delphian knife produce
inferior things which cannot do what they are meant to do properly
because they were made for sale rather than use (Pol., 1252b1±5), and
the professional arts like medicine are targets too. Aristotle generalizes

31 For a more detailed account of the dif®culty and Aristotle's view of it, see my Aristotle's
Economic Thought, ch. 4 (1), 68±81.
 Quality and quantity in economics 51

the point: those not involved in commerce but who want to pursue
money `try to do so by some other means, employing each of the faculties
in an unnatural way [and] make all these faculties means for the business
of providing wealth [chreÃmatistikeÃ, that is, money-getting], in the belief
that wealth is the end and that everything must be directed to the end'
(Pol., 1258a8±14).32 Aristotle's worry is that commerce and its values
could penetrate into all the activities that make up the life of the polis or
society, corrupting them, causing a confusion of ends which would make
it dif®cult for the community to order its activities properly, and
obscuring any clear view of the end or point of social living, of what it is
to live well.
This kind of worry is rife today when almost all activity is regulated by
money, though many modern writers are less acutely aware of this
ambiguity in the conduct of activities than Aristotle was, even though
they live in a market economy where the situation is much worse, and he
didn't. The utilitarian reform of the metaphysics of the ordinary idea of
activity is well adapted to the modern market situation, because it gives
the impression of avoiding the kind of concern Aristotle had about the
ambiguous conduct of activities. The suppression of speci®c ends
removes the con¯ict of ends in favour of the end of pursuing money, for
which the single end of `utility' looks suspiciously like a euphemism.33
Aristotelian metaphysics, unlike utilitarian metaphysics, is not amen-
able to eliding category distinctions, distinctions between kinds, or
distinctions between ends, and because of this Aristotle is led to establish
a strict line of demarcation between the ordinary and the economic
worlds. He begins his thinking about matters that we would call
`economic' with a distinction between value in use and value in exchange

32 ChreÃmatistikeà is often translated nowadays in ways that blur the very distinction
Aristotle is drawing between the two arts of wealth-getting, getting wealth as useful
things and getting wealth as money. Rackham, in the Loeb edition of the Politics
(London 1926), at times translates it as `business', so that at 1257b35 Aristotle's
distinction becomes an opaque one between `the two arts of business'. Carnes Lord
makes it even more misleadingly a distinction between the two `forms of expertise in
business', Aristotle: The Politics (Chicago 1984).
33 Utilitarians might deny that they are committed to the view that the common currency of
utility is money, but as Bernard Williams observes, `they are committed to something
which in practice has those implications: that there are no ultimately incommensurable
values'. Williams adds that it is not an accidental feature of the utilitarian outlook that
the presumption is in favour of the monetarily quanti®able and against other values,
`because (for one thing) utilitarianism is the value system for a society in which economic
values are supreme; and also, at the theoretical level, because quanti®cation in money is
the only obvious form of what utilitarianism insists upon, the commensurability of
value', Williams, Morality, 103.
52 The what, why, and how of ecomonic ontology

(1257a6±13), the former falling into the category of quality and the latter
into that of quantity. When he subsequently comes to de®ne wealth and
acquisition (chreÃmatistikeÃ) he gives two de®nitions of each, one in terms
of value in use (C) and one in terms of value in exchange (M). He ®rst
de®nes and distinguishes wealth as useful things and wealth as money,
and then he transposes each of them into an end of action, distinguishing
two forms of behaviour C-M-C and M-C-M. So he is quite systematic in
separating the ordinary world from the economic world, and this gives
Aristotle's analysis a certain transparancy. Economic thought today is
expressed in terms which, in effect if not by design, merge the ordinary
world and the economic world rather than hold them apart, and this
gives it a certain opacity.
Aristotle's pre-modern clearheadedness in keeping the ordinary world
separate from the economic world has not endeared him to commercial
society, and anglophone moderns from Hobbes onwards have given him
an infamous reputation as an enemy of science and a defender of
aristocratic privilege which is entirely unfounded. Marx and Keynes were
sailing against a well set tide in taking the Aristotelian course rather than
the utilitarian one. They got from Aristotle a vantage point from which it
could be seen just how strange the economic world is: a world of
exchange value and M-C-M, in which wealth is money not useful things,
in which there is only one kind of activity not many, and in which there is
only one kind of useful thing.

V
In defence of these metaphysical reforms it might be said that they are
technicalities required for a special scienti®c purpose, and that this is
their justi®cation. Such a defence can work only on the condition that
the special scienti®c purpose (SSP) be de®ned in a way that clearly
demarcates the special terrain to which it applies in order that the use of
the technicalities may be restricted to that terrain, and prevented from
contesting with ordinary concepts for the role of favoured terms for
describing the world of ordinary experience. De®nitions of the SSP
generally fail to meet the condition. Robbins famously de®nes economics
as `the science which studies human behaviour as a relationship between
ends and scarce means which have alternative uses'.34 Such a de®nition
bridges the distinction between the ordinary world and the economic
world, and this makes it useless as part of a defence of the metaphysical
reforms as technicalities required for an SSP. Indeed, in de®ning

34 L. Robbins, The Nature and Signi®cance of Economic Science (London 1932), 15.
 Quality and quantity in economics 53

economics in terms that represent it as a branch of practical reason,

Robbins commissions the very confusion of the ordinary and economic
worlds that the defence is meant to defend against.
There is point in looking for ways of keeping economic discourse apart
from ordinary discourse if the aim is to prevent the reformed concepts
competing with ordinary language for the description of ordinary
experience. But if part of the aim is to enter into such a contest with a
view to winning it, then we are not dealing primarily with an SSP at all,
but with an attempt to change how things and activities are ordinarily
described and thought about. If that were the case, then a de®nition of
the sort offered by Robbins would be a contribution to the contest, and
its strength would be that it builds on the dif®culty everyone has in
keeping the ordinary world and the economic world apart by merging
them and making it even more dif®cult to unravel them. If this has been
the strategy it has been effective in creating confusion, and it has
produced some grotesque results in other areas of scholarship. In some
historical work on the ancient world, for instance, authors have stocked
antiquity with entrepreneurs, banks, interest-rates, money markets, and
other market phenomena that didn't exist, largely because of a failure to
make adequate safeguards for maintaining a distinction between the
language of the economic world and that of the ordinary world, so that
such authors have unre¯ectively resorted to economic or market lan-
guage for describing any form of social behaviour in any kind of society
and in any period of history.35
Not all schools of economic thought have muddied the waters in this
way. Marx and Keynes didn't, and Adam Smith himself maintained a
substantial separation between the ordinary world and the economic
world. Smith knew perfectly well that market economy was a recent
phenomenon. He identi®ed it as the fourth of his historical stages, the
`stage of commerce', and he was not led to read back into non-market
societies the concepts, behaviour, and institutions of a market economy.
He also held that `consumption' or useful things (C) `is the sole end and
purpose of production', and since he was aware that business pursues M
not C and that there are dangers in this, he thought that business needed
public regulation.36 Provided business was regulated properly, however,
he believed that the totality of business operations produced an outcome
for the society that was characterized by improving C. Smith could not
35 See M. I. Finley, The Ancient Economy (London 1985). The same kind of anachronism is
found in anthropology; see C. Meillassoux, `Essai d'interpretation du pheÂnomeÁne
eÂconomique dans les socieÂteÂs traditionelles d'auto-subsistence', Cahiers d'eÂtudes afri-
caines, 4 (1972), 38±67.
36 A. Smith, Wealth of Nations, vol. ii, 159.
54 The what, why, and how of ecomonic ontology

have formulated either position if he had merged the ordinary and the
economic worlds in the way that economics now does.
Little more has been done here than scratch the surface of what it is
that underlies the common feeling that the world of the accountants and
economists is odd or suspect, but perhaps it has been enough to show at
least that there is sense to be made of a distinction between the ordinary
world and the economic world. The temptation to run the two worlds
together is constantly present and I have tried to suggest that there is
intellectual and moral point in keeping them apart. Perhaps something
has been said also to sustain the belief that insisting on the priority of the
ordinary world is not to be dismissed as `folk economics', and that
economics does not do better and more scienti®cally something that is
ordinarily done less expertly, but does something else altogether. Exactly
what that is, is one of the threads that have been left dangling here.
Part II
Rationality and homo economicus
4 The normative core of rational
choice theory

RUSSELL HARDIN*

Introduction

Although it is commonly called a positive theory ± to imply that it is

purely descriptive and without value assumptions ± rational choice
theory is typically grounded in a powerful and simple value theory, from
which many of its most compelling results follow. In actual applications,
the word ``rational'' in rational choice theory is typically not a merely
formal term. It is also a substantive term that refers to particular values,
essentially welfare values. It may be true, as some rational choice
theorists insist, that the theory could be applied to actors whose values
are other than welfare and, especially, other than their own welfare. But
the remarkable success of rational choice explanations turns on the
surprising power they have when they are grounded almost exclusively in
the actors' own welfares.
To this extent, rational choice theory is, practically, a two-fold theory
that says what values govern individuals' choices and what the result of
those choices will be. Often, the result is collective rather than merely
individual, although rational choice explanations are strictly at the
individual level and can govern individual actions outside social inter-
actions as well as within such interactions. In the discussion here,
however, I will focus on rational choice explanations of social-political
interactions, including large group interactions.
A common complaint against rational choice theory is that it mistakes
concern with self-interest for rationality. The complaint is well grounded
but pointless. Rational choice theorists commonly do suppose that the
* An earlier version of this paper was presented to the political theory seminar at Yale
University. I wish to thank Stephen Elkin, Kristen Monroe, James Scott, Ian Shapiro,
and participants in that seminar for their comments. I also thank the Mellon and Russell
Sage Foundations for generous general support.

57
58 Rationality and homo economicus

agent's own welfare is the chief value of concern to the agent. But there is
no mistake involved in such an assumption. It is rather a hypothesis than
a direct ®nding. If it produces good predictions and explanations, it is a
good hypothesis; otherwise it is a bad hypothesis. In any case, it cannot
be judged a mistake merely by direct inspection of motivations. The critic
should want to show not that the assumption is wrong by inspection but
rather that it leads to wrong predictions and explanations.
It may be that such a theory will fail too often to be of explanatory
value. In actual fact, such a theory in the hands of Bernard Mandeville,
Anthony Downs, Gary Becker, and Mancur Olson seems to have swept
still less successful theories from the ®eld in various contexts. The vast
literature showing that Olson gets interest groups wrong is a literature
about the success of tiny fractions of people in various potential interest
groups who have taken some degree of collective action. A theory that
®ts the overwhelming bulk of people most of the time and only fails for
small fractions is so unlike any other theory in the dismal history of the
social sciences that, if it were Mendel and his theory of inheritance in
peas, we would think the data were phoney. That they are not phoney is
best attested by the work of the critics themselves who show very low
levels of contribution toward various collective provisions.
Oddly, the critics often seem to want to have their cake and eat it too.
They criticize moves by many rational choice theorists to include broader
concerns in individual utility functions. For example, William Riker and
Peter Ordeshook include a sense of ``citizen duty'' in their list of
``rational'' motivations to vote1 (Riker and Ordeshook 1962) and Gary
Becker (1962) says the utility function can be quite catholic in its
inclusions.2 (Critics say such moves reduce rational choice theory to
tautology, making it a matter of constant rede®nition to ®t what happens
rather than a matter of prediction or explanation. But they also complain
that a theory limited to self-interest is far too narrow to explain
individual behavior and social outcomes.)
In any case, the power of rational choice theory has come from its
clear assumption, most of the time, of a substantive value that drives
individual choices. That value is often captured in the rubric ``self-
interest,'' although that is the wrong term for a value. I have an interest
in having lots of money. Money is only an instrumental, not a consump-
tion or welfare value. It is a means, not an end. Most of my interests in

1 William H. Riker and Peter C. Ordeshook, ``A Theory of the Calculus of Voting,''
American Political Science Review 62 (1962): 25±42.
2 Gary S. Becker, The Economic Approach to Human Behavior (Chicago: University of
Chicago Press 1976).
 The normative core of rational choice theory 59

contemporary social science contexts are instrumental means, as money,

power, and ef®cacy in voting are. The underlying value of rational choice
theory is generally own welfare. This value theory gives rational choice
its bite. Without a substantive value theory, it is only conceptual, not
explanatory.

A quick survey of contemporary rational choice theory

There are two main classes of results in rational choice theory, one of
which is primarily conceptual, the other of which is primarily explana-
tory. First, Kenneth Arrow's impossibility theorem says, in brief, that
under well-speci®ed conditions a group's interests cannot even be de®ned
in the same terms as those that de®ne an individual's interests. It is a
fallacy of composition to speak of collective or group rationality as an
analog of individual rationality.
For the second result, Downs's economic theory of democracy ex-
plained why we might get tweedle-dum and tweedle-dee candidates in
competition, why voters might not know enough to vote their interests,
and why many might not vote at all. Olson, in a variant of the last of
Downs's points, showed why interest groups may not form at all to
represent many, especially widely popular, interests and why the political
interest group system has massive biases in favor of groups already
organized for other purposes. Hence, it is also a fallacy of composition to
suppose a group can act or be motivated to act as an individual can.
Note that Arrow's result requires no speci®c value theory. It follows
for virtually any value theory that is a preference theory, i.e., that orders
states of affairs. Downs's and Olson's results carry for any case in which
we might say there is a collective bene®t whose provision would bene®t
all the individuals in a relevant group, in which the costs of providing the
collective bene®t must be borne by the individuals, and in which the costs
and bene®ts can be aggregated into net bene®ts. Hence, these results
apply to cases in which there is a relatively clear substantive value at
stake. In what follows, the concern will generally be with explanatory
rational choice theory, which has generally been built on a relatively
well-de®ned value theory.
In the sections below, I will canvass contemporary rational choice
theory and then discuss a major confusion in what is at issue. I will then
discuss proposals for a rational choice theory without substantive values
and three non-welfarist value theories that could alternatively underlie
social choice and explanation. Finally, for the problem of group, indeed,
60 Rationality and homo economicus

revolutionary cooperation, I will consider the chief candidate theory of

politics that is not methodologically individualist: structuralist political
economy. I will conclude with brief remarks on the explanatory power of
the assumption of individualist welfarism.

The value theory of rational choice theory

One of the striking features of the individualist school of political

economy is that it starts from a value theory: individual welfare
somehow conceived, including, at its most nearly vacuous, simple
preferences. This value theory may be objectionable on various grounds,
sometimes including considerations of its internal consistency, but it is
relatively coherent and it seems to have strong and coherent implications.
When it is coupled with a modest variant of Hobbes's or any socio-
biologist's psychological theory ± that the ®rst or strongest interest of the
individual is the self ± it leads to a remarkable blending of positive and
normative conclusions. Indeed, this was the great invention of Hobbes.
By putting his positive theory of human psychology together with his
value theory of human welfare, he was able to devise a positive theory of
government that was at once seemingly a normative theory. That is a
trick many economists seem to think they turn every day.
Hobbes and then Locke insisted that politics centrally address welfare
issues. Hobbes thought religious differences should be subordinated to
political order, that individual citizens should, in their own interest, let
their sovereign decide their religious af®liation. Locke proposed leaving
religion out of politics in order to enable politics to focus on issues of
property, including property in oneself. Hobbes and Locke have prob-
ably been the two most in¯uential political philosophers in the modern
western tradition. As with the best of the rational choice theorists, a chief
reason for their in¯uence and the intellectual coherence of their argu-
ments is that they start from a particular value theory.
It may seem odd to modern positivists to say that an explanatory
theory gains coherence from a value theory. But that is genuinely the
story with individualist political economy. This makes sense if it is true
that people are motivated by values in their choices, especially if they are
motivated systematically by some set of values. Implicitly, at least, most
economists and most rational choice theorists are in the tradition of
Hobbes and Locke in assuming that people are so motivated. If their
assumption is largely true, then rational choice theory should yield good
explanatory results. Those who criticize the particular morality of the
 The normative core of rational choice theory 61

theory on the claim that people are more driven by other values should
not ®nd it objectionable per se to ground explanatory theory in value
commitments.
What difference has economic reasoning actually made in political
science? Obviously, it has contributed to major revisions in the areas that
Arrow, Downs, and Olson addressed, and more recently in the study of
Congressional organization, regulation, and campaign contributions. In
addition, I wish to assert what may be a controversial claim: Economic
reasoning has pushed us toward more closely relating normative and
positive concerns. For example, it has pushed us into giving normative
accounts of institutional arrangements. We assess the institutions for
how well they ®t our purposes. Earlier generations of social theorists
made normative judgments of institutions, but they often drew their
normative principles from elsewhere, not from the panoply of positive
principles they used. As in Hobbes, so in current work in individualist
political economy, the positive and the normative are driven by related
motors: individual incentives for individual bene®ts. This core concern is,
moreover, a powerful organizer of manifold problems and analyses
across a broad front. Theory grounded in it may often falter and give
seemingly wrong results, but it seldom fails to yield some kind of result.3
The argument here is not one of a necessary or conceptual relation
between the value theory assumptions and the rational choice structure
of the explanatory theory. It is, rather, a causal or explanatory
claim. Practitioners of the individualist approach most often assume,
with Hobbes, some version of self-interest as the goal the individual seeks
to serve. Downs and Olson produce their striking results from this
assumption.
But the assumption of self-interest is not necessary to the individualist
approach. For example, it does not lie behind Arrow's impossibility
theorem, which is arguably devastating for democratic theory.4 Arrow's
result does not depend on self-interest and could still apply even if
everyone chose from purely public-spirited motivations. Imposing rea-
sonable structures of self-interest to limit the range of plausible choices
actually helps to avoid Arrow's impossibility result in speci®c contexts.
Indeed, even a moral theorist concerned with action from non-self-
interested motives may wish to start from a framework of rational choice
in understanding individual moral choices. The moral theorist may even
3 Some may consider that a fault, not a virtue in the approach.
4 Kenneth J. Arrow, Social Choice and Individual Values (New Haven, Conn.: Yale
University Press 1951; second ed. 1963). See further, Russell Hardin, ``Public Choice vs.
Democracy,'' in John W. Chapman, ed., NOMOS 32: Majorities and Minorities (New
York: New York University Press 1990), pp. 184±203.
62 Rationality and homo economicus

wish to juxtapose self-interest against other motivations. And that

theorist may often be forced to conclude that self-interest is too powerful
in that contest ± it gives the better explanation of behavior.
This last observation suggests what is perhaps the greatest source of
controversy over the explanatory and moral propriety of individualist
political economy. It is not only the individualist focus, it is also the
nearly universal assumption of self-interest as the major explainer of
behavior that rankles. We seem to know from introspection that self-
interest does not rule our lives so extensively as it rules in individualist
political economy. Some academics can look within themselves and
seldom ®nd self-interest at work. Rational choice theorists seem to take
special pleasure in showing, or at least in making plausibility arguments,
that much of what we thought was moral or otherwise not self-interested
behavior is arguably self-interested after all. Sometimes the answer to
this debate is that both sides are right but in different ways. The self-
interest theorists may capture much of what people do politically even in
extreme contexts and the overwhelming bulk of what they do in more
ordinary contexts. But individuals acting from other commitments spear-
head many important activities, such as revolutionary, charitable, and
ideal-regarding actions. Therefore their commitments are a very impor-
tant part of explaining what happens even if they are statistically
uncommon commitments.

Interests, consumptions, welfare

It is common to say that people are rational in some contexts and not in
others. One might be relatively rational in considering mortgages for a
home but not in considering partners for marriage. It is merely an
extension of this view to suppose that in some general contexts, such as
different cultures, people may be not rational even though people in
other cultures might be. For example, James Scott has argued that the
very conservative peasants of southeast Asia are not rational. Rather,
they are profoundly risk averse and their choices are governed by a
``moral economy.''5 What he means is that they are not rational in the
sense of simply maximizing their production of rice. Instead of adopting
seed grains that would have much higher average annual yields, they

5 James C. Scott, The Moral Economy of the Peasant: Rebellion and Subsistence in Southeast
Asia (New Haven, Conn.: Yale University Press 1976).
 The normative core of rational choice theory 63

adopt very hardy, stable grains that will almost always produce at least
enough that they will not starve.
When Scott's thesis is summarized, it sounds like a rational choice
account.6 Peasants are not interested in maximizing resources only in the
sense of producing the most total food over the years or the most income
from it. They want such resources, after all, only as means, not as an end.
The end they want is a reasonably good life. The goodness of life depends
not only on resources but also on consumptions that the resources make
possible. Life with reasonable levels of consumption seems eminently
preferable to a life of ¯uctuation between great plenty and starvation.
Hence, the rational peasant might well choose Scott's risk averse policy.
Scott supposes peasants are smart but not rational in their risk
aversion. One might sooner conclude that rational choice ®ts the condi-
tions of scarcity, such as those of Scott's southeast Asian peasants,
particularly well. It is in open-ended contexts in which available resources
permit ful®llment of a wide range of values that rational choice might
seem dubious. Or, at least, rational choice grounded in the narrow value
of own welfare might seem dubious. I will return to this point below.
Scott's analysis is essentially Hobbesian in the following sense. Scott
supposes peasants will ®rst make sure of survival and only then go after
greater wealth. Hobbes supposed we must all put order uppermost and
therefore we accept government. In both cases, there is no tradeoff with
the rationality of amassing greater wealth; there is merely the prior need
for the sine qua non of survival, without which there can be no bene®t
from wealth. If the conditions faced by Scott's peasants are as he
describes them, their risk aversion is not a matter of ``moral'' economy
but only of rational choice.7
Scott speaks of ``preferences which do not make sense in terms of
income alone.''8 This is nothing unusual ± preferences are coherently
seen only over whole states of affairs, in which income is only part of
what is at issue. I do not have a preference for blue. I may have a
6 I will not discuss the whole of Scott's theory of his peasants, and one might suppose he
has claims for putting other behaviors under the rubric moral economy. I think that all
his important arguments, such as mutual insurance and reciprocity, are readily framed as
directly rational, as is seemingly extreme risk aversion in the choice of seed grains as
discussed here. The latter is, in any case, the most compelling part of Scott's empirical
analysis.
7 On the analogous risk aversion of the entrepreneur or manager in a market society, see
Richard H. Day, Dennis J. Aigner, and Kenneth R. Smith, ``Safety Margins and Pro®t
Maximization in the Theory of the Firm,'' Journal of Political Economy 79 (1971):
1293±1301. Would anyone wish to call this the ``moral economy'' of the entrepreneur or
manager?
8 Scott, Moral Economy, p. 35.
64 Rationality and homo economicus

preference for a blue shirt with certain other clothes to be worn in

particular circumstances. As is presumably true of Scott and virtually
everyone else as well, I similarly do not have an unrestricted preference
for higher income. If higher income entails giving up my academic life or
entails giving up more of my leisure time, I may not prefer it to my
present income with my present lifestyle and consumption pattern. There
are no preferences that ``make sense in terms of income alone.''
The issue here is a conceptually messy one at the very foundations of
rational choice theory in the welfarist school. We speak, sometimes
almost interchangeably, about interests, consumptions, and welfare. But
these are conceptually quite different. Our interests are what put us in a
position to consume and consumption typically brings welfare. I have an
interest in amassing resources, but resources are of no value per se ± I
want them only in order to be able to consume. Obviously, interests and
consumptions trade off with each other. If I consume some things, I must
expend some of my resources. (Some social theorists argue that we
should ground our normative theories in resources rather than in
welfare, that certain conceptual problems in welfare make it ®nally an
unworkable normative principle.9 Without resolving that issue, we may
all readily grant that resources are means without intrinsic value and that
what gives them instrumental value is the welfare they can bring us.)
When Scott speaks of a preference for income alone, he misuses the
concept of preference. I can prefer one thing to another all else equal or I
can prefer one completely de®ned state of affairs to another.10 I cannot
simply prefer higher income ± I must add what I prefer it to and I must
add some rough equivalent of the phrase ``all else equal.'' Similarly,
Scott's peasants can prefer the whole state of affairs that follows from
planting a highly productive but also highly fragile strain of rice to the
state that follows from planting a safer but less productive strain, or vice
versa. If I have either of these preferences and I act on it, then I, Scott's
peasant, am rational.
The beauty of Scott's account is that, after we correct the conceptual
confusions in it, the empirical case still stands its ground. Scott's analysis
translates into a rich rational choice theory of peasant behavior.11 The
principal errors are evidently not in the empirical account but only in the
interpretation of their economic signi®cance.
9 Amartya Sen, ``Equality of What?,'' pp. 353±369 in Sen, Choice, Welfare and
Measurement (Cambridge, Mass.: MIT Press 1982).
10 Russell Hardin, ``Rational Choice Theories,'' in Terence Ball, ed., Idioms of Inquiry:
Critique and Renewal in Political Science (Albany, N.Y.: SUNY Press 1987), pp. 67±91.
11 As Samuel L. Popkin argues in The Rational Peasant: The Political Economy of Rural
Society in Vietnam (Berkeley: University of California Press 1979).
 The normative core of rational choice theory 65

Incidentally, ``moral'' in this account of moral economy seems to mean

no more than that, given their established practice, which is eminently in
their interest, the peasants attach moral vocabulary to the practice. As do
people in many other contexts as well, they deduce an ``ought'' from an
``is.''12 Or perhaps they merely short-cut their reasoning and moralize
means to the ends they value.13
Finally, recall the point made by some critics (especially philosophers)
of rational choice that we should expect rational choice theory to fail
when resources become pro¯igate enough to allow the ful®llment of ever
more diverse consumptions. At that level of resources, resources or their
amassment may compete with consumption (they typically compete in
the sense of trading off resources for consumptions). I may be able to
gain more resources only by forgoing consumptions. Moreover, my
consumptions may become much more, not less competitive, just because
my indulgence of them is less constrained. Under the grim constraints of
subsistence, I cannot really choose much beyond limited consumption
with survival and more spectacular consumption brie¯y on the way to
not surviving. It is not that rational choice theory becomes increasingly
unstable with increasing resources, but that preference over the vast
range of possibilities becomes more idiosyncratic and perhaps even less
stable.

Rational choice without substantive values

There are several ways in which theorists avoid the assumption of

welfarism (or any other particular substantive value) as the basis of
rational choice. Two are especially instructive. In the ®rst, it is sometimes
assumed that each individual can have any values, welfarist or otherwise,
and that the theory will still work, as though the theory were somehow
neutral with respect to values. In the second, it is supposed that we get
12 Russell Hardin, One for All: The Logic of Group Con¯ict (Princeton, N.J.: Princeton
University Press 1995), pp. 60±5.
13 In his discussion of ``the moral economy of the English crowd,'' E. P. Thompson's
central concern is with the reaction of the general populace to high prices for grain in
years of short harvests. In those years, the people insist, often under threat of force, that
dealers and millers not charge substantially higher prices than they would in good years.
This is a ``moral'' concern insofar as it is an objection to pro®teering from adversity. E.
P. Thompson, ``The Moral Economy of the English Crowd in the Eighteenth Century,''
Past and Present 50 (1971): 76±136. Thompson sometimes frames his argument as a
derivation of an ought from an is ± as did, no doubt, the eighteenth-century English
crowd. The price in years of normal harvest becomes the morally right price.
66 Rationality and homo economicus

stable results without any assumption whatever about the internal

mental considerations of the actors as though the theory were purely
behavioral and not implicitly cognitive.
First, consider the neutral view of values. Arrow's theorem is based on
leaving it open to each individual to have any preference ordering over
various states of affairs. My preferences could be ordered by an altruistic,
sel®sh, spiteful, ideal-regarding, or any other principle. This, the most
abstract form of the utility function ± with any values plugged in ± is a
late, essentially theoretical development. Actual economic reasoning
about society typically continues the traditional presumption of welfare
values, the presumption that there are objective things that matter to
most people. In actual reasoning, for example, it is typically assumed
that people prefer more resources to fewer, more of some consumptions
within a given range to fewer, reasonable survival to death, and so forth.
Many applications of economic reasoning would make no sense without
such assumptions, which are so much a part of second nature that they
are not even explicitly stated.
Similarly, claims for the inherent neutrality of the liberal program are
prima facie false or at least exaggerated. Virtually all liberals and all
liberalisms share Hobbes's view of the substantive values of survival and
welfare. Hence, there is real, substantive content to the values defended
by liberalism, which is not an empty shell that can be ®lled with any
values willy-nilly. Indeed, Locke, an early liberal, was particularly
concerned to keep non-welfarist religious values out of politics in order
to secure welfarist values. The only neutrality we might insist on is the
neutrality of John Stuart Mill's skepticism in the determination of what
conduces to the welfare of other individuals. One person thinks parlor
games are the best way to enhance welfare at the end of the day, another
thinks relaxing mindlessly before the television is the best, another
thinks reading great classics is best. All may be right for themselves
alone. The fact that you think television is best for you is prima facie
evidence that it is best for you. I must be epistemologically neutral about
which is better for you in principle and must therefore defer prima facie
to your judgment unless I can ®nd causal relations that might sway your
views.
Millian skeptical neutrality is not neutrality with respect to welfare,
virtue, or whatever choice of fundamental value. It is epistemological,
not moral. It is neutrality with respect to judgments of what contributes
to welfare for a particular individual. Some things can be shown to have
systematic welfare tendencies, as health does or various resources and
talents might. But any such claim for systematic tendencies is empirical,
not a priori, and it can prove to be wrong in a particular momentary
 The normative core of rational choice theory 67

instance or even more generally for a particular person. Hence, these

claims may still be overridden by other considerations, such as par-
ticularly grim conditions that might make life seem worse than death, or
harmful causal effects of which the particular individual may be ignorant.
Enjoying wine this moment may seem to me to be a good thing, but it
may contribute to wrecking my life.
Although the value theory of most rational choice theory is not
neutral, the theory itself is neutral with respect to substantive social
problems it can address. It can address almost any problem, almost every
kind of problem. It is not a theory of class con¯ict, state formation,
revolution, or democracy. But it can be brought to bear on all of these.
Now turn to the purely behavioral view of choice ± choice that is no
longer rational in the usual sense, because it involves no mental supposi-
tions about agents. Gary Becker has argued that we do not need the
assumption of rational choice to get standard results in economic
analysis.14 He analyzes the demand curve on the assumption that
resources are limited rather than on the usual assumption that choice is
deliberately intended to maximize utility. He shows that the market
demand curve is negatively sloped even if individual households are
irrational in their consumption choices. Hence, Becker demonstrates that
we can get market regularity from individual irregularity.
There are three, perhaps too easy, responses to this move. First,
Becker himself does not follow up his claim thereafter but sensibly
returns to the more common assumption of rationality. He defends this
move on the compelling ground that the standard theory is so richly
developed that the gains from working with it in his areas of concern far
outweigh what gains might be had from attempting to create a new
theory.15
Second, rationality that is inherently grounded in substantive values of
welfare is closely related to resourcist accountings. The appeal of basing
theory in resources is that, ostensibly, the theory is then fully behavioral.
We need not posit mental principles for choosing. We merely lay out the
resource constraints. But this apparent escape from the supposedly
hidden world of the mental requires at its outset that we somehow intuit
or know how the resource constraints matter. We start from the mental,
we do not escape it. Becker's demonstration was not directed at the
philosopher's constant concern with the mental and the behavioral but
only at the understanding of large systemic patterns.

14 Gary S. Becker, ``Irrational Behavior and Economic Theory,'' pp. 153±68 in Becker, The
Economic Approach to Human Behavior (essay ®rst published 1962).
15 Becker, The Economic Approach to Human Behavior, p. 151.
68 Rationality and homo economicus

Third, in politics it is often not sensible to characterize collective

choices as choices by individuals in a market. The choices are very often
of collective provisions, not individual provisions. We might still expect
resource constraints to affect our collective survival, just as individual
resource constraints affect individual survival. But this will be a big
general issue, not an issue relevant to every collective choice. If we wish
to understand many particular collective choices or if we ®nd the results
of collective choice relatively stable or regular, we are likely to be driven
back to assuming the individuals involved in the choices are acting
rationally, not merely whimsically or incomprehensibly.
To a large extent, unless we can carry out a program such as Becker's,
the difference between assuming some value theory for the actors and not
assuming any is the difference between explanatory and conceptual
analysis in individualist political economy. If we impute certain substan-
tive values to the actors in our theory, we can say what behaviors or
choices follow from those value commitments. If we do not, we cannot
say much. Hence, we may treat rational choice theory as a strictly
positive theory in the following sense. The theorist takes the values of
individuals into account but the theorist need not take a value position.
If a theorist seems strongly committed to the descriptive claim that all or
most people share some value, one might suspect the theorist shares the
value. And some rational choice theorists may seem to hold the view that
self-interest is the ``right'' value for a person to have. But one could be
completely open as to whether there are any right values while still
thinking, descriptively, that some small set of values motivates a par-
ticular population in some context.

Non-welfarist value theories

Reliance on a substantive value theory is, of course, not unique to

rational choice theory. Theories, both normative and explanatory, have
been grounded in various non-welfarist value theories with speci®c
substantive content. Canvassing many of these or analyzing any one of
them in depth would take up far too much space for present purposes,
but it is useful to consider three major traditions: contractarianism,
Marxist labor theory of value, and communitarianism. All of these have
been applied both to normative justi®cation and to explanation of social
results.
 The normative core of rational choice theory 69

Contractarianism

In its explanatory mode, political contractarianism was ridiculed by

Hume and it has never recovered. Hume merely pointed out what
everyone knew: that no actual states seemed to have historical grounding
in an act of general popular agreement or contract. Still, contractar-
ianism, the political generalization of consent, is touted as a justi®catory
principle.16 In this, it trades on the appeal of consent in ordinary personal
contexts. But that appeal does not compose readily into a collective
analog. The apparent moral force of contractarianism is a fallacy of
composition. There might be a context in which composition from
individual to collective consent would work. But modern states are not
such contexts.
Hobbes clearly wanted justi®cation. Perhaps therefore many inter-
preters have tried to impose a moral contractarian vision on him, as
though he were arguing that because we have agreed to government we
are morally obligated to stick by our agreement. His actual justi®cation
is more nearly utilitarian because it is grounded in mutual advantage. He
says we are all better off to have a state and, once we have one, to avoid
dissension and revolution. Mutual advantage is, of course, part of what
is required for a standard contractarian argument to go through. But
Hobbes did not need to take the further step of claiming that prior
agreement makes obedience morally obligatory because mutual advan-
tage was suf®cient justi®cation for his argument, as it often would be for
a utilitarian justi®cation.

Labor theory of value

Turn to the labor theory of value. The chief candidate for Marx's value
theory is his labor theory of value, according to which the value of an
object is a function only of the amount of labor that has gone into
producing it. That value theory is conspicuously silly in general. For
example, if academic journals applied that theory to the acceptance of
articles for publication, the journals would deteriorate dramatically. But
if that value theory fails, the Marxist argument about exploitation
collapses with it. One might still think there is capitalist exploitation of

16 Russell Hardin, ``Contractarianism: Wistful Thinking,'' Constitutional Political

Economy, 1 (1990), 35±52.
70 Rationality and homo economicus

workers or ®rst world exploitation of the third world, but it will take a
new argument grounded in a different value theory to make the case.
Is the labor theory of value strictly normative or is it also explanatory?
It would be out of character for Marx not to have treated it as
explanatory. The account goes roughly as follows. Workers in some
circumstances, such as the nineteenth-century industrial factory, begin to
recognize their actual state through discussion of common interests. In
that state, they are exploited because they get only a fraction of the labor
value they contribute to the capitalist's product. As their understanding
grows, they achieve class consciousness, they fully recognize their
common interest as a class in opposition to another class. They are
therefore then in a position to act in common for their class interest when
relevant opportunities occur.
If we assert there is exploitation without grounding the claim in a
value theory, such as the labor theory of value, the claim is likely to
involve a naturalistic inference to a normative conclusion. We simply
look at the structure of payoffs in an interaction and infer exploitation
from nothing more than these facts. That would be an illegitimate and
uncompelling move. Therefore, the labor theory of value or some
alternative is necessary for Marxist argument, whether normative or
explanatory. That is why, if the labor theory of value proves to be
incoherent, then Marxist exploitation is incoherent as well ± unless,
again, it can be given grounding in some other value theory or normative
principle.
Incidentally, it should be clear that one can reject the labor theory of
value and claims of exploitation based on it and still argue for greater
equality or for redistribution of resources for welfare. Indeed, exploita-
tion per se is an odd notion that seems inherently to involve a naturalistic
inference of a normative claim from a factual assessment. We might
readily conclude that there is no exploitation but that there is grievous
inequality or deprivation that merits correction. As the demand for
unskilled labor falls in the United States, the permanently unemployed
underclass grows. They cannot sensibly be seen as a capitalist resource, a
reserve army of the proletariat available for controlling wages in the
most pro®table industries, although they may have that function in some
service industries such as some fast-food restaurants. It is increasingly
implausible to view their plight as one of exploitation. Most capitalists
might be better off if there were no underclass.
Oddly, Marx's own value theory seems likely to be regressive, not
progressive. For example, labor values cannot typically be expected to
trump exchange or market values in a free-wheeling market. Suppose
therefore that we leave it to central agencies to assign labor values in
 The normative core of rational choice theory 71

order to secure relevant incomes based in these values. These are likely to
be grounded in past ways of doing things. Hence, in a period of
technological transition (which is virtually always), we will tend to assign
values for the manufacture of, say, cloth that derive from earlier values
when there were no modern, mechanized looms. We will be Luddite. If
we are, then we perversely undercut the possibilities of economic welfare
for the large masses, whose poverty seemed to drive Marx's theorizing.

Communitarianism

Finally, consider an older view of what we might call cultural motivation,

a view that is now articulated in communitarian thought. Communi-
tarian argument is predominantly normative rather than explanatory.
Historically, cultural motivations have been used in explaining collective
action by diverse ethnic, religious, local, and national communities. In a
sense, the appeal to communitarian values allows us to escape the fallacy
of composition inherent in generalizations from individual to collective
interest by supposing that there is no problem of composition of
individual values.
Super®cially, there seems to be no fallacy because every individual is
supposed to want the same thing and to have the desire to contribute to
achieving it. On the contribution side of the claim, however, one may still
expect substantial con¯ict, just as in ordinary collective action contexts in
which we all want the same collective provision but it is in our interest
individually that others do the providing. Hence, communitarian values
do not resolve the problem of composition from individual motivation to
collective action. Their role is rather to admit non-welfarist values into the
individual's panoply, values that are determined somehow by the group.
Some community values are merely group-level aggregations of indi-
vidual-level values. For example, individuals may be expected to bene®t
from general economic growth in some contexts. Such welfarist values
seem unlikely to be problematic for rational choice theory unless they are
rei®ed at the individual level in the form of strictly group-regarding
values. Some community values, however, need not be welfarist ±
although it is too facile to conclude that they necessarily are not welfarist
merely on the testimony of those who hold them. For example, many
people may assert that the only reason to keep a promise is that it is
immoral on non-welfarist grounds to break one. And yet they may
generally only make promises for welfarist reasons and may typically
break them for welfarist reasons.
72 Rationality and homo economicus

Much of recent work on communitarian values is justi®catory. The

purpose of the work is to argue that communitarian values are good in
some sense and that they should be honored by individuals and polities.
Those who are troubled by news of atrocities and bloodshed in the name
of particular community values have solid ground for complaint against
too quick communitarian justi®cation. Much of the work has been
conceptual in trying to make sense of how values can get constituted at
all if there is no communal basis for at least some of them.17 So far, the
``theory'' is woefully underdeveloped. The relatively casual blending of
Aristotelian virtues with Hegelian community commitments that stands
in for communitarian value theory is too bland and unstructured to yield
systematic conclusions. Worse still, the value theories of contemporary
communitarians seem to share the regressive bias of the labor theory of
value. Those theories elevate a form of social Luddism to the level of the
good for a given community.18

The structuralist alternative to individual choice

Explanations grounded in these alternative value theories are still

methodologically individualist. In the face of major political phenomena
such as revolutions, the structuralist political economist may assert that
an individualist approach is inadequate even on the accounts of individu-
alist political economy. If the bulk of behavior would not be revo-
lutionary and only a relatively small number of individuals make a
revolution go, we cannot expect to know enough about relevant indi-
viduals to make competent predictions or even retrospective assessments
of revolutions. We cannot have a theory of revolution or great social
movements that depends on predicting whether a speci®c Lenin or
Gorbachev is available to lead, and we cannot believe Leo Tolstoy's
functionalist vision, at the end of War and Peace, that great times
necessarily produce great people. We will have to look to more
systematic structural considerations that ®nally enable the occasional
revolutionary to make a difference.
One answer that the individualist theorist can give to this criticism is
that the revolutions we see may be more or less accidental rather than
structurally determined. Hence, the structuralist approach does no better

17 Michael J. Sandel, Liberalism and the Limits of Justice (Cambridge: Cambridge

University Press 1982).
18 Hardin, One for All, chapter 7.
 The normative core of rational choice theory 73

than the individualist. Structuralists are often uninterested in individual

commitments and are willing simply to assume that relevant kinds will
manifest themselves under certain conditions. They can then focus on
explaining or understanding those conditions. The great strength of the
individualist school is in simple generalizations about individuals, gen-
eralizations that may transform understanding even while blurring
differences that seem to matter historically.
In the recent gentle revolutions of East Europe we may see seeming
vindication of both individualist and structuralist arguments. For many
commentators, the structural conditions of apparently declining econo-
mies seem to explain the revolutions. The accident of Gorbachev seems
to many others to be the chief explanation of at least the timing of the
revolutions. But individualists might further argue, as most western
economists do, that the structural conditions of declining economies are
eminently easily explained from individualist assumptions. The structur-
alist approach is therefore, at least in large part, merely shorthand for a
more precise and more extensive individualist account. Moreover, indivi-
dualist economists were perhaps more consistent than any other social
scientists in predicting the changing conditions of the political economies
of the East European nations. To trump them, structuralists would have
to argue that the major condition was not economic but rather political
and moral ± it was the urge for democracy and openness, for national
self-determination and the end of colonial domination, or whatever.
Ironically, that would be an odder move in the structuralist school than
in the individualist school.

Concluding remarks

If theorists of other persuasions are to take on individualist theorists on

an equal footing, they too will have to accommodate their theories to a
compelling value theory, which, at present levels of theoretical develop-
ment of value theories, means either to accommodate to something like
the individualist value theories of welfare and preference or to develop
some other theory far more extensively than has yet been done. If the
theory remains hollow at its core, it will be only variously of interest in
its occasional insights rather than generally of interest in its overall
program.
Perhaps rational choice theory has a natural tendency to develop a
value theory roughly like that which drives most of it. Because it is
individualist, it must be concerned with individual actions, hence, most
74 Rationality and homo economicus

likely, individual motivations. What motivates an individual? Typically,

the desire to achieve particular ends or values, perhaps especially self-
regarding ends and values, as rational choice theorists generally assume.
Rational choice theory is ®nally like sociobiology in the following
limited sense. The most obvious sociobiological inference about human
motivation is that humans seek their own interests. If they did not, they
would individually reduce their chances of survival. This inference leaves
open for investigation whether it could happen that group-oriented
behaviors get selected. Economic reasoning about human motivation
starts from the simplest self-interest assumption19 and then turns to the
same range of more complex aggregate or group-level issues. Still, in
both cases, the group-level results are derived from individual interests.
Even at the group-level, we would expect the individual-level result still
to hold ± if not always, at least very often, perhaps even predominantly
often.
Much of the power of the rational choice approach may be an
accident. It just happens that the welfarist value theory of economics
over the past two or three centuries grew in response to theoretical and
empirical attacks from vast numbers of economists and their critics.
Nothing even vaguely like that has happened to any other value theory.
As a result, the formal value theory of rational choice is richly articulated
and supple. Anyone who wishes to oppose some other value theory to it
is burdened with the virtually impossible task of ®rst making that other
theory articulate enough even to stay in debate. Worse still, at this stage
in development of various theories, all others are virtually forced to deal
with the formal and even the substantive value theory of rational choice
if they are to address similar problems. That is to say, other theorists are
likely to be forced to contribute to rational choice theory while pursuing
their own preferred alternative.20 The critics of rational choice may
®nally wish to weep at their complicity in its re®nement.
19 This was Hobbes's assumption long before it was a sociobiological inference.
20 See Russell Hardin, ``The Morality of Law and Economics,'' Law and Philosophy 11
(November 1992): 331±84, especially pp. 380±4.
5 The virtual reality of homo
economicus

PHILIP PETTIT

The economic explanation of individual behaviour, even behaviour

outside the traditional province of the market, projects a distinctively
economic image on the minds of the agents involved. It suggests that in
regard to motivation and rationality, they conform to the pro®le of homo
economicus. But this suggestion, by many lights, ¯ies in the face of
common sense; it con¯icts with our ordinary assumptions about how we
each feel and think in most situations, certainly most non-market
situations, and about how that feeling and thought manifest themselves
in action. What, then, to conclude? That common sense is deeply in error
on these matters? That, on the contrary, economics is in error ± at least
about non-market behaviour ± and common sense sound? Or that some
form of reconciliation is available between the two perspectives? This
paper is an attempt to defend a conciliationist position.
The paper is in ®ve sections. In the ®rst section I describe the economic
mind that is projected in economic explanation, whether explanation of
market or non-market behaviour. In the second section I argue that this
is not the mind that people manifest in most social settings and, in
particular, that it is not the mind that common sense articulates. In the
third section I show that nevertheless the economic mind may have a
guaranteed place in or around the springs of human action; it may have a
virtual presence in the generation of action, even action on which it does
not actually impact. In the fourth section I show that where the economic
mind has such a virtual presence, that is enough to license an important
variety of economic explanation: the explanation of the resilience or
robustness of certain patterns rather than the explanation of their
emergence or continuance. And then in a short ®fth section I show that
this sort of explanation ®ts with some established ideas about the
explananda of economic and social science.

75
76 Rationality and homo economicus

1 The thesis of the economic mind

There are two sorts of assumptions that economists make about the
minds of the agents with whom they are concerned. First, content-
centred assumptions about the sorts of things that the agents desire:
about which things they prefer and with what intensities. And, second,
process-centred assumptions about the way in which those desires, those
degrees of preference, issue in action.
The process-centred assumptions boil down to the assumption that
people's actions serve their desires well, given their beliefs about such
matters as the options available, the likely consequences of different
options, and so on. There are different theories as to what it is for an
action or choice to serve an agent's desires well, given the agent's beliefs:
about what it is for an agent to be rational. Many economists work with
relatively simple models but the family of theories available is usefully
exempli®ed by Bayesian theories of rationality (Eells 1982). According to
Bayesian theory, an action is rational just in case it maximizes the agent's
expected utility.
The Bayesian idea, roughly, is that every agent has a utility function
that identi®es a certain degree of utility, a certain intensity of preference,
for every way the world may be ± every prospect ± and a probability
function that determines, for each option and for each prospect, the
probability that the choice of that option would lead to the realization of
that prospect. An action will maximize the agent's expected utility just in
case it has a higher expected utility than alternative options, where we
determine the expected utility of an option as follows. We take the
prospects with non-zero probability associated with the option; we
multiply the utility of each prospect by the fraction representing the
probability of its being realized in the event of the option's being chosen;
and we add those products together.
So much for the assumptions that economists make about the way
desires or preferences lead to action. What now of the assumptions that
they make about the content of what human beings prefer or desire? The
main question here is how far economists cast human beings as ego-
centric in their desires. In order to discuss it, we need some distinctions
between different theses that each ascribe a certain egocentricity.

1. Self-centredness. This relatively weak claim says that people do what

they do as a result of their own desires or utility functions. They do
not act on the basis of moral belief alone; such belief issues in action,
only if accompanied by a suitable desire. And they do not act just on
 The virtual reality of homo economicus 77

the basis of perceiving what other people desire; the perception that
someone desires something can lead to action only in the presence of a
desire to satisfy that other person.
2. Weak non-tuism. This is a stronger claim, in the sense that it
presupposes the ®rst but represents people as intuitively more ego-
centric still. People's desires bear on how others behave and on what
happens to others, so the thesis goes, but such desires are not affected
by perceptions of what those agents desire, even for themselves;
people's utility functions, as it is often put, are independent of one
another (Gauthier 1986, 87).
3. Strong non-tuism. A stronger claim again: people's desires do not
extend, except instrumentally, to others. Not only do people take no
account of what others desire in forming their own desires in regard to
others; any desires they have for what others should do, or for what
should happen to others, are motivated ultimately by a desire for their
own satisfaction (Gauthier 1986, 311).
4. Self-regardingness. A thesis that presupposes 1 but represents an
alternative way of strengthening it to that represented by 2 and 3.
People's non-instrumental desires may extend to others, and they may
be responsive to the perceived desires of others ± 2 and 3 may be false
± but the more that the desires bear on their own advantage, the
stronger they are; in other words, people are relatively self-regarding
in their desires.

Economists almost universally accept the ®rst, self-centredness thesis.

Agents who are rational in any economically recognizable sense cannot
be led to action just by moral belief or the perception of what another
desires or anything of the sort; such belief or perception may affect what
they do but only through ®rst affecting what they desire. Some thinkers
toy with the possibility that agents may be capable of putting themselves
under the control of something other than their own desires: for example,
Mark Platts (1980) when he imagines that moral belief may motivate
without the presence of desire; Amartya Sen (1982, Essay 4) when he
speaks of the possibility of commitment; and Frederic Schick (1984)
when he canvases the notion of sociality. But economists are probably on
the side of common sense in urging that all action is mediated via the
desires of the agent (Pettit 1993, Chapter 1). In any case, that is what I
shall assume in what follows. There is a con¯ict between economics and
common sense, as I shall be arguing, but it does not arise in respect of
this ®rst thesis.
Do economists go beyond the rather uncontroversial form of self-
centredness articulated in thesis 1? They certainly do so to the extent that
78 Rationality and homo economicus

certain versions of the axioms of consumer choice theory go beyond

minimal requirements of rationality, self-centredness included, and imply
features like the downward-sloping demand curve. But that is not the
issue. The question is whether economists go beyond the postulate of
self-centredness in postulating any of the more egoistic theses, 2 to 4.
Many economic theories endorse weak and strong non-tuism. They do
so to the extent that various economic models assume that any good I do
you is, from my point of view, an externality for which ideally I would
want to extract payment: an external bene®t that I would ideally want to
appropriate for myself (or `internalize') (Gauthier 1986, 87). But this
seems to be a feature of particular models and not an assumption that is
essentially built into the economic way of thinking. And it is a feature
that affects only some of the standard results of the theories in question,
not all of them (Sen 1982, 93). I am not inclined to regard it as a deep
feature of economic thinking. It may have little or no presence, for
example, in the application of economic thought to social life outside the
market.1
While economics postulates self-centredness in the sense of thesis 1,
then, it does not necessarily suppose that people are non-tuistic in the
senses de®ned in theses 2 and 3. But, to come now to thesis 4, I do think
that the discipline is committed to the assumption that people's self-
regarding desires are generally stronger than their other-regarding ones:

1 Some may say that there is a deeper reason than the frequent use of non-tuistic
assumptions for thinking that economic thinking is strongly non-tuistic in nature. The
deeper reason, according to these theorists, is that in holding that agents act so as to
satisfy their desires, economists assume that agents act for the sake of achieving their own
desire-satisfaction: that is, for the sake of attaining a certain bene®t for themselves.
Anthony Downs gives countenance to this line of thought when, ironically enough, he
tries to explain how economists can make sense of altruism. `There can be no simple
identi®cation of acting for one's greatest bene®t with sel®shness in the narrow sense
because self-denying charity is often a great source of bene®ts to oneself. Thus our model
leaves room for altruism in spite of its basic reliance upon the self-interest axiom' (Downs
1957, 37).
The line of thought in Downs's remark is confused. Accepting the economic theory of
rationality may mean believing that people maximize expected utility but it does not mean
believing that they act for their own greatest bene®t. That a person maximizes expected
utility means that they act in the way that best serves their desires, according to their
beliefs, but not that they do so for the sake of maximum desire-satisfaction and, in that
sense, for the sake of their greatest bene®t. When I act on a desire to help an elderly
person across the road, I act so as to satisfy that desire but I do not act for the sake of
such satisfaction; I act for the sake of helping the elderly person. To think otherwise
would be to confuse the sense in which I seek desire-satisfaction in an ordinary case like
this and the sense in which I seek it when I relieve the longing for a cigarette by smoking
or the yearning for a drink by going to the pub.
 The virtual reality of homo economicus 79

that in this sense people are relatively self-regarding in their desires.

Whenever there is a con¯ict between what will satisfy me or mine and
what will satisfy others, the assumption is that in general I will look for
the more egocentric satisfaction. I may do so through neglecting your
interests in my own efforts at self-promotion, or through helping my
children at the expense of yours, or through jeopardizing a common
good for the sake of personal advantage, or through taking the side of
my country against that of others. The possibilities are endless. What
unites them is that in each case I display a strong preference for what
concerns me or mine, in particular a preference that is stronger than a
countervailing preference for what concerns others.2
The assumption that people are relatively self-regarding in their desires
shows up in the fact that economists tend only to invoke relatively self-
regarding desires in their explanations and predictions. They predict that
as it costs more to help others, there will be less help given to others, that
as it becomes personally more dif®cult to contribute to a common cause
± more dif®cult, say, to take litter to the bin ± there will be a lesser level
of contribution to that cause, and so on. They offer invisible-hand
explanations under which we are told how some collective good is
attained just on the basis of each pursuing their own advantage. And
they specialize in prisoner's dilemma accounts that reveal how people
come to be collectively worse off, through seeking each to get the best
possible outcome for themselves.3
The belief that people are relatively self-regarding shows up in other

2 Notice that this conception of self-interest is consistent with the recognition of a capacity
on the part of ordinary agents to identify with entities beyond themselves. See Pettit 1997,
Chapter 8.
3 It may be said against this that I am focusing on purely contingent aspects of economic
explanation: that there is no reason why economists should not develop their explanations
on the basis of other-regarding desires as well. Perhaps fewer people will put their litter in
a bin that becomes more dif®cult to access. But, equally, fewer people will put their litter
in a bin, if it comes to be generally believed that littering is not so bad after all: say if it
comes to be believed, however improbably, that littering has some good environmental
side-effects. Or so any economist should be prepared to admit.
This observation shows that economists can and should recognize the relevance of
relatively other-regarding desires. But it does not demonstrate that they must take those
desires to be potentially just as powerful as self-regarding preferences. And the
explanatory practice of economists manifests the contrary belief. The working assumption
behind economic explanation is that however much people may care for others, care for a
collective good, or care for some moral principle, their self-concern is likely to outweigh
the effects of such care, if it comes into con¯ict with it. That is why it must be a miracle in
the economics textbook if some aggregate or collective pattern emerges or continues when
the available self-regarding reasons argue against people's doing the things that the
pattern requires.
80 Rationality and homo economicus

aspects of economic thought too. It may be behind the assumption of

economic policy-makers and institutional designers that no proposal is
plausible unless it can be shown to be `incentive-compatible': that is,
unless it can be shown that people will have self-regarding reasons for
going along with what the proposal requires.4 And it may be at the root of
the Paretian or quasi-Paretian assumption of normative or welfare
economics that it is uncontroversially a social bene®t if things can be
changed so that all preferences currently satis®ed continue to be satis®ed
and if further preferences are satis®ed as well. This assumption is plausible
if the preferences envisaged are self-regarding, for only envy would seem
to provide a reason for denying that it is a good if some people can get
more of what they want for themselves without others getting less. But the
assumption is not at all plausible if the preferences also include other-
regarding preferences, as we shall see in a moment. And so the Paretian
assumption manifests a further, deeper belief: that the preferences with
which economics is concerned are self-regarding ones.
The Paretian assumption is not plausible ± certainly not as uncontro-
versial as economists generally think ± when other-regarding preferences
are involved, for reasons to which Amartya Sen (1982, Essay 2) has
directed our attention. Consider two boys, Nasty and Nice, and their
preferences in regard to the distribution of two apples, Big and Small.
Nasty prefers to get Big no matter who is in control of the distribution.
Nice prefers to get Small if he is in control ± this, because he is other-
regarding and feels he should give Big away if he is in charge ± but
prefers to get Big, if Nasty is in control: he is only human, after all. The
Paretian assumption suggests ± under the natural individuation of
options (Pettit 1991) ± that it is better to have Nice control the
distribution rather than Nasty. If we put Nice in control, then that
satis®es Nasty ± he gets Big ± and it satis®es Nice as well: Nice's
preference for having Big if Nasty is in control does not get engaged and
Nice's preference for having Small ± for giving Big away ± if he is in
control himself is satis®ed. But this is clearly crazy: it means that we are
punishing Nice for being nice, in particular for having other-regarding
preferences; and this, while apparently attempting just to increase
preference-satisfaction in an impartial manner. The lesson is that the
Paretian assumption is not plausible once other-regarding preferences
®gure on the scene and so, if economists think that it is plausible ± think

4 In fairness, however, I should note that this search for incentive-compatibility could be
motivated ± reasonably or not ± by the belief that however other-regarding most people
are, policies should always be designed to be proof against more self-regarding `knaves'.
See Brennan and Buchanan 1981.
 The virtual reality of homo economicus 81

indeed that it is uncontroversial ± that suggests that they only have self-
regarding preferences in view.
The upshot of all this, then, is that economists present human agents
as relatively self-regarding creatures who act with a view to doing as well
as possible by their predominantly self-regarding desires. These desires
are usually assumed to be desires for what is loosely described as
economic advantage or gain: that is, roughly, for advantage or gain in
the sorts of things that can be traded. But self-regarding desires, of
course, may extend to other goods too and there is nothing inimical to
economics in explaining patterns of behaviour by reference, say, to those
non-tradable goods that consist in being well loved or well regarded
(Pettit 1990; Brennan and Pettit 1993, 2000). The economic approach is
tied to an assumption of relative self-regard but not to any particular
view of the dimensions in which self-regard may operate.

2 The con¯ict with common sense

Does the picture ®t? Are human beings rational centres of predominantly
self-regarding concern? It would seem not. Were human agents centres of
this kind, then we would expect them to ®nd their reasons for doing
things predominantly in considerations that bear on their own advan-
tage.5 But this isn't our common experience, or so at least I shall argue.
Consider the sorts of considerations that weigh with us, or seem to
weigh with us, in a range of common-or-garden situations. We are
apparently moved in our dealings with others by considerations that bear
on their merits and their attractions, that highlight what is expected of us
and what fair play or friendship requires, that direct attention to the
good we can achieve together or the past that we share in common, and
so on through a complex variety of deliberative themes. And not only are
we apparently moved in this non-egocentric way. We clearly believe of
one another ± and take it, indeed, to be a matter of common belief ± that
we are generally and reliably responsive to claims that transcend and
occasionally confound the calls of self-regard. That is why we feel free to

5 Some might say that under the assumption that human beings are rational centres of
predominantly self-regarding concern ± this, in a Bayesian sense ± we ought to expect that
they would be, not only self-concerned, but also calculating: we ought to expect that they
would think in terms of the ledger of probabilities and utilities that ®gure in Bayesian
decision theory. I do not go along with this. Bayesian decision theory says nothing on
how agents manage to maximize expected utility; it makes no commitments on the style of
deliberation that agents follow. See Pettit 1991.
82 Rationality and homo economicus

ask each other for favours, to ground our projects in the expectation that
others will be faithful to their past commitments, and to seek counsel
from others in con®dence that they will present us with a more or less
impartial rendering of how things stand.
Suppose that people believed that they were each as self-regarding as
economists appear to assume; suppose that this was a matter of common
belief amongst them. In that case we would expect much of the discourse
that they carry on with one another to assume the shape of a bargaining
exchange. We would expect each of them to try to persuade others to act
in a certain way by convincing them that it is in their personal interest to
act in that way: this, in good part, by convincing them that they, the
persuaders, will match such action appropriately, having corresponding
reasons of personal advantage to do so. Under the economic supposition,
there would be little room for anyone to call on anyone else in the name
of any motive other than self-interest.
The economic supposition may be relevant in some areas of human
exchange, most saliently in areas of market behaviour. But it clearly does
not apply across the broad range of human interaction. The normal
mode under which people exchange with one another is closer to the
model of a debate than the model of a bargain. It involves them in each
presenting to the other considerations that, putatively, they both recog-
nize as relevant and potentially persuasive. I do not call on you in the
name of what is just to your personal advantage; did I do so, that could
be a serious insult. I call on you in the name of your commitment to
certain ideals, your membership of certain groups, your attachment to
certain people. I call on you, more generally, under the assumption that
like me you understand and endorse the language of loyalty and fair
play, kindess and politeness, honesty and straight talking. This language
often has a moral ring but the terminology and concepts involved are not
con®ned to the traditional limits of the moral; they extend to all the
terms in which our culture allows us to make sense of ourselves, to make
ourselves acceptably intelligible, to each other.
One way of underlining this observation is to consider how best an
ethnographer might seek to make sense of the ways in which people
conduct their lives and affairs. An ethnographer that came to the shores
of a society like ours ± a society like one of the developed democracies ±
would earn the ridicule of professional colleagues if they failed to take
notice of the rich moral and quasi-moral language in which we ordinary
folk explain ourselves to ourselves and ourselves to one another: the
language, indeed, in which we take our bearings as we launch ourselves
in action. But if it is essential for the understanding of how we ordinary
folk behave that account is taken of that language, then this strongly
 The virtual reality of homo economicus 83

suggests that economists must be mistaken ± at least they must be

overlooking some aspect of human life ± when they assume that we are a
relatively self-regarding lot.6
The claim that ordinary folk are oriented towards a non-egocentric
language of self-explanation and self-justi®cation does not establish
de®nitively, of course, that they are actually not self-regarding. We all
recognize the possibilities of rationalization and deception that such a
language leaves open. Still, it would surely be miraculous that that
language succeeds as well as it does in de®ning a stable and smooth
framework of expectation, if as a matter of fact people's sensibilities do
not conform to its contours: if, as a matter of fact, people fall system-
atically short ± systematically and not just occasionally short ± of what it
suggests may be taken for granted about them.
We are left, then, with a problem. The economic mind is that of a
relatively self-regarding creature. But the mind that people display
towards one another in most social settings, the mind that is articulated
in common conceptions of how ordinary folk are moved, is saturated
with concerns that dramatically transcend the boundaries of the self. So
how, if at all, can the economic mind be reconciled with the common-or-
garden mind?7

3 The economic mind as a virtual presence

The obvious answer for would-be conciliationists is to say that whereas

ordinary folk conform in most contexts to the picture of the common
6 We may note in passing that there is nothing surprising in the fact that our ordinary
encounters with one another are articulated and shaped by a non-egocentric language.
We are not just bargaining creatures who take one another's beliefs and desires as given
and seek out minimal terms of cooperation. We are creatures who also try to in¯uence
what we each believe and desire, under the assumption that when obstacles do not get in
the way ± when there is nothing we are disposed to fault about our circumstances ± then
we are susceptible to the same considerations in the formation of our beliefs and desires:
under the assumption, equivalently, that we are sensitive to the same norms of belief and
desire formation (Pettit 1993, Chapter 2). Given that we pursue this enterprise, it is only
to be expected that we should have evolved a language for framing culturally shared
expectations.
7 This problem may be dismissed by some thinkers on the ground that the literature on
conditional cooperation shows how economically rational individuals may cooperate out
of purely self-regarding motives (Axelrod 1984; Taylor 1987; Pettit and Sugden 1989). But
that would be a mistake. This literature shows that economically rational individuals may
come to behave cooperatively, not that they will come to think and talk in a cooperative
way.
84 Rationality and homo economicus

mind, the economic mind is still implicitly present in such contexts. But
how to interpret this? What does it mean to say that the economic mind
is implicitly present: that people are implicitly but not explicitly oriented
towards the self-regarding concerns that economists privilege?

The main model of the implicit±explicit distinction is drawn from a

visual analogy. It suggests that an explicit concern is something focal,
something directed to the centre of a subject's ®eld of vision, whereas an
implicit concern is a concern for what lies at the edge of that ®eld: a
concern for what is peripherally rather than focally tracked. If I explicitly
desire something, my desire is explicit in the sense in which I am explicitly
aware of the computer screen in front of me; if I implicitly desire
something, my desire is implicit in the sense in which I am ± or was a
moment ago ± only implicitly aware of the telephone at the edge of my
desk. Does this model help in explicating the idea that even if people are
not always explicitly of an economic turn of mind, they are at least
implicitly so?
The model certainly gives us a picture of what it might mean to say
that implicitly people are economically minded. It would mean that even
as people pay attention to the sorts of concerns engaged in ordinary
exchanges with others, even as they keep their eyes on the needs of a
friend, the job that has to be done, the requirements of fairness, they
invariably conduct some peripheral scanning of what their own advan-
tage dictates that they should do. The model does not deny the
appearance of more or less other-regarding deliberation but it does
debunk that appearance. It suggests that whether they are aware of it or
not, those who practise other-directed deliberation indulge a more self-
directed style of re¯ection in the shadows of the mind, on the boundaries
of their attention. Gary Becker (1976, 7) comes close to endorsing this
model when he writes: `the economic approach does not assume that
decision units are necessarily conscious of their own efforts to maximize
or can verbalize or otherwise describe in an informative way reasons for
the systematic patterns in their behaviour. Thus it is consistent with the
emphasis on the subconscious in modern psychology.'
But the focal±peripheral interpretation of the claim that people are
implicitly self-regarding does not make the claim seem particularly
compelling. We all admit that people profess standards from which they
often slip and that their slipping does usually relate to an awareness,
perhaps a deeply suppressed awareness, of the costs of complying with
the standards. We all admit, in other words, that weakness of will and
self-deception are pretty commonplace phenomena. But what the focal±
peripheral model would suggest is that the whole of human life is shot
 The virtual reality of homo economicus 85

through with this sort of failure: that what we take to be a more or less
occasional, more or less localized, sort of pathology actually represents
the normal, healthy state of the human organism. That is a fairly
outrageous claim. Most economists would probably be shocked to hear
that the view of the human subject which they systematically deploy is
about as novel, and about as implausible, as the picture projected in
classical Freudianism.
But if we reject the focal±peripheral way of reconciling the economic
and the common mind, are we forced to choose between the two pictures
of the human subject? Are we forced to choose between economic science
and common sense? Happily, I think not. There is a second, less familiar
model of the implicit±explicit distinction that is available in the literature
and it promises a different, more attractive mode of reconciliation.8
I call this the virtual±actual model. One area where it is sometimes
deployed ± though not in so many words ± is in explaining the sense in
which I may implicitly believe that 2 times 101 is 202, even when I have
never given a thought to that particular multiplication; or, to take
another example, the sense in which I may implicitly believe that Europe
has more than ten million inhabitants, when I have only ever thought
about the population of individual countries. I implicitly believe these
things in the sense that I am so disposed ± speci®cally, I am so familiar
with elementary arithmetic or with the population ®gures for European
countries ± that even the most casual re¯ection is suf®cient to trigger the
recognition that indeed 2 times 101 is 202, indeed the population of
Europe is more than ten million. I virtually believe the propositions in
question ± virtually, not actually ± but the virtuality or potentiality in
question is so close to realization that ordinary usage scarcely marks the
shortfall.9
I propose that if we are to follow the familiar conciliationist route of
describing people as economically minded, but not always in an explicit

8 I explore a third model of this distinction in Pettit (1998) but it doesn't seem to have any
application relevant to present concerns.
9 The implicitness of my belief that 2 times 101 is 202 should not be confused with the
implicitness of my belief, say, that for any number described in decimal notation, you get
double that number by following the sort of rule that you and I apply in computing 2
times 101: a rule, as it happens, that we would probably ®nd it hard to articulate. The
implicitness of the belief in the rule does not lend itself to modelling on the virtual±actual
pattern, but rather on some other analogy such as that provided by the focal±peripheral
picture, because it is clear that you and I do actually believe in the rule; we do actually
believe in it to the extent that we do actually rely on it. The implicitness of my belief that 2
times 101 is 202 is the implicitness of non-actuality, the implicitness of a belief that hovers
on the edge of realization, not the implicitness of a belief that is realized in some sub-
articulate fashion.
86 Rationality and homo economicus

fashion, we should try to spell out this claim by reference to the virtual±
actual model, not the focal±peripheral one. I think that it is not
implausible that people are virtually self-regarding in most contexts of
choice, even if they are not actually so. It is generally agreed that actual
self-regard plays a great part in market and related behaviour but that it
does not have the same sort of presence ± if it has a presence at all ± in
other contexts: for example, in contexts of ordinary family or friendly
interaction, in contexts of political decision, or in contexts of group
behaviour. What I suggest is that in such non-market contexts self-regard
may still have an important presence: it may be virtually if not actually
there; it may be waiting in the wings, even if it is not actually on stage.
Here is how self-regard might have a virtual presence in such contexts.
Suppose, ®rst of all, that people are generally content in non-market
contexts ± we can restrict our attention to these ± to let their actions be
dictated by what we might call the cultural framing of the situation in
which they ®nd themselves. A friend asks for a routine level of help and,
in the absence of urgent business, the agent naturally complies with the
request; it would be unthinkable for someone who understands what
friendship means to do anything else. There is an election in progress
and, the humdrum of everyday life being what it is, the agent sponta-
neously makes time for going to the polls; that is manifestly the thing to
do, under ordinary canons of understanding, and the thing to do without
thinking about it. Someone has left a telephone message asking for a
return call about some matter and the agent doesn't hesitate to ring back;
even if aware that there is nothing useful they can tell the original caller,
they shrink from the impoliteness, in their culture, of ignoring the call. In
the pedestrian patterns of day-to-day life, the cultural framing of any
situation will be absolutely salient to the ordinary agent and the ordinary
agent will more or less routinely respond. Or so at least I am prepared to
assume.
But that is only the ®rst part of my supposition. Suppose, in the
second place, that despite the hegemony of cultural framing in people's
everyday deliberations and decisions, there are certain alarm bells that
make them take thought to their own interests. People may proceed
under more or less automatic, cultural pilot in most cases but at any
point where a decision is liable to cost them dearly in self-regarding
terms, the alarm bells ring and prompt them to consider personal
advantage; and heeding considerations of personal advantage leads
people, generally if not invariably, to act so as to secure that advantage:
they are disposed to do the relatively more self-regarding thing.
Under these suppositions, self-regard will normally have no actual
presence in dictating what people do; it will not be present in deliberation
 The virtual reality of homo economicus 87

and will make no impact on decision. But it will always be virtually

present in deliberation, for there are alarms which are ready to ring at
any point where the agent's interests get to be possibly compromised and
those alarms will call up self-regard and give it a more or less controlling
deliberative presence. The agent will run under cultural pilot, provided
that that pilot does not carry them into terrain that is too dangerous
from a self-interested point of view. Let such terrain come into view, and
the agent will quickly return to manual; they will quickly begin to count
the more personal losses and bene®ts that are at stake in the decision on
hand. This re¯ection may not invariably lead to self-regarding action ±
there is such a thing as self-sacri®ce, after all ± but the assumption is that
it will do so fairly reliably.
If the suppositions I have described were realized, then it would be fair
to say that people are implicitly self-regarding: that they implicitly
conform to the image of the economic mind. The reason is that under the
model of virtual self-regard, no action is performed without self-
regarding consideration unless it fails to ring certain alarms: that is,
unless it promises to do suitably well in self-regard terms. What it is to do
suitably well may vary from individual to individual, of course, de-
pending on their expectations as to what is feasible and depending on
their self-regarding aspirations: depending on how much they want for
themselves, and with what intensity. But the point is that regardless of
such variations, the model of virtual self-regarding control does privilege
self-regard in a manner that conforms to the image of the economic
mind. Another way of putting this point is to say that under the model
described, an agent will generally be moved by certain considerations
only if they satisfy a certain negative, self-regarding condition: only if
they do not tend to lead the agent towards a certain level of self-sacri®ce.
Let the considerations push the agent below the relevant self-regarding
level of aspiration and the alarm bells will ring, causing the agent to
rethink and probably reshape the project on hand.
The position which self-regard is given under the model of virtual self-
regarding control is rather like that which it enjoys under Herbert
Simon's (1978) model of satis®cing as distinct from maximizing be-
haviour. People do pretty well in self-regarding terms, even if they do not
do as well as possible. And it may even be that virtual self-regarding
control enables them to do as well as possible in egocentric terms, for the
absence of self-regarding calculation in most decisions represents a
saving in time and trouble ± these are virtues emphasized by Simon ± and
it may also secure other bene®ts: it may earn a greater degree of
acceptance and affection, for example, than would a pattern of relentless
calculation.
88 Rationality and homo economicus

But is the model of virtual self-regarding control, in particular the

scenario of the alarm bells, a plausible one? The question divides in two.
First, is there any arrangement under which we can imagine that such
alarms are put in place? And second, if there is, can we plausibly
maintain that those alarms will reliably serve to usher self-regarding
deliberation into a controlling position in the generation of behaviour?
The alarms required will have to be informational; they will have to be
signals that this is the sort of situation where the agent's advantage may
be compromised, if cultural framing is given its head. So are there signals
available in ordinary contexts that might serve to communicate this
message? Clearly, there are. Consider the fact that a decision situation is
non-routine; or that it is of a kind where the agent's ®ngers were already
burned; or that it is a situation in which the agent's peers ± others who
might be expected to fare about as well ± do generally better than the
agent; or that some conventional or other assurances as to the responses
of others are lacking. Any such facts can serve as signals that the agent's
personal advantage may be in especial danger. Indeed it is hard to
imagine a situation where the agent's interests were likely to be compro-
mised in signi®cant measure by culturally framed demands ± compro-
mised in a measure that the agent would not generally tolerate ± without
such signals being present. Certainly it is reasonable to assume that
generally there will be signals available in such situations that the agent
should take care: signals to the effect that this is a situation where that
framing is liable to serve the agent less well than it ordinarily does.
The other question is whether it is plausible, given the availability of
signals of this kind, to postulate that the signals will generally tip agents
into a self-regarding sort of deliberation: a sort of deliberation that is
normally sidelined in favour of ®delity to the cultural frame. This issue is
wholly an empirical matter but it is an issue on which the weight of
received opinion speaks unambiguously. It has been common wisdom for
at least two thousand years of thinking about politics that few are proof
against temptation and few, therefore, are likely to ignore signals that
their self-interest may be endangered. Human beings may be capable of
reaching for the stars but, except for some romantic strands of thought,
all the streams in the western tradition of thinking suggest that if there is
opportunity for an individual to further their own interests, then they
can generally be relied upon, sooner or later, to exploit that opportunity:
all power corrupts. The main theme of the tradition is summed up in the
lesson that no one can be entrusted with the ring of Gyges that Plato
discusses: the ring that renders a person invisible and that makes it
possible for them to serve their own interests with impunity, at whatever
cost to the interests of others.
 The virtual reality of homo economicus 89

These lines of thought give support, therefore, to the picture described

above. They suggest that it is very plausible to think that even where
people pay no actual attention to relatively self-regarding considerations,
still those considerations have a certain presence and relevance to how
people behave. They are virtually present, in the sense that if the
behaviour rings the alarm bells of self-interest ± and there will be plenty
of such bells to ring ± the agent will give heed and will tend to let self-
regarding considerations play a role in shaping what is done.10
Under the emerging picture, then, there is a sense in which people are
always at least implicitly of the self-regarding cast of mind projected by
economists; if they are not actually self-regarding in their mode of
deliberation, they are virtually so: if self-regard does not actually occupy
the pilot's seat, it is always there in the co-pilot's, ready to assume
control. The picture is a rather non-idealistic representation of human
beings but it is not unnecessarily bleak. It emphasizes that in the normal
run, people are not calculatingly self-concerned: they articulate their lives
and relationships in the currency of received values and they generally
conform to the requirements of those values. Where it goes non-idealistic,
it does so only in the spirit of what we might call the Gyges axiom: the
principle that virtue ± ®delity to the demands of the cultural frame ± is
fragile and generally survives only under conditions where it is not
manifestly against the interests of the agent, only under conditions where
the alarm bells do not ring.
There are two further points to put to those who worry about the
alleged non-idealism of our picture. First, the picture leaves open the
possibility that in many cases some individuals will not heed the alarms
and will stick to what the culturally framed situation requires, by criteria
of common values, through the thick and the thin of self-sacri®ce. And,
second, the picture leaves room for the Aristotelian principle that people
become virtuous, become lovers of virtuous ways, through habituation in
those ways. It leaves room, not just for the possibility that some people
will be relatively heedless of the alarms described, but for the possibility
that such heedlessness may be facilitated in increasing measure by a
regime in which the alarms only rarely ring: a regime in which things are
well designed and people are free, in the silence of self-regard, to develop
an attachment to doing that which by the common values of the culture
is what the situation requires.
10 The picture of virtual self-regard may be modi®ed by being made subject to certain
boundary conditions. It might be held, for example, that the picture does not apply
universally, only under certain structural arrangements: say, that it does not apply in
family life, only in relations of a more public character. For related ideas see Satz and
Ferejohn 1994.
90 Rationality and homo economicus

4 The economic mind as an explanatory principle

We saw in the ®rst section that the economic mind is distinctively self-
regarding and in the second that it contrasts in this respect with the
common mind: the mind as articulated in common ways of thinking. The
last section gave us a picture under which it seems possible to reconcile
these two points of view: the points of view associated respectively with
economics and common sense. The common-sense viewpoint is valid to
the extent that ordinary folk manage their affairs most of the time
without adverting to their own interests; they are guided in their decisions
by what is required of them under the cultural framing of the situations
in which they ®nd themselves. The economic viewpoint is valid to the
extent that even when this is so, even when people are not explicitly self-
regarding in their deliberations, still self-regard has a virtual presence; it
is there, ready to affect what people do, in the event that any of the alarm
bells of self-interest ring.
The question which now arises, however, is how far the merely virtual
presence of self-regard is supposed to legitimate the economic explana-
tory enterprise: the enterprise of explaining various patterns in human
affairs by reference to rational self-regard.11 That self-regarding consid-
erations have a virtual as distinct from an actual presence in human
deliberation means that they are not actual causes of anything that the
agents do. They may be standby causes of certain patterns of behaviour:
they may be potential causes that would serve to sustain those patterns,
did the actual causes fail. But it is not clear how anything is to be
explained by reference to causes of such a would-be variety. After all,
explanation is normally taken to uncover the factors operative in the
production of the events and patterns to be explained; it is normally
taken to require a reference to actual causal history (Lewis 1986, Essay
22).
This dif®culty can be underlined by considering the explananda that
economic investigation is ordinarily taken to be concerned with in the
non-market area. These are, ®rst, the emergence of certain phenomena or
patterns in the past and, second, their continuation into the present and
future. The explanation of the emergence of any phenomenon ± say, the
emergence of a norm or institution ± clearly requires a reference to the

11 Apart from the problem that I go on to discuss, there is an issue as to how, non-
circularly, the economist is to tell the level of threat to self-interest at which an agent's
alarm bells ring. I cannot discuss this problem here but would just note that it is parallel
to the problem of determining an agent's aspiration level under Simon's (1978) satis®cing
model.
 The virtual reality of homo economicus 91

factors that were operative in bringing it into existence. And the explana-
tion of the continuation of any phenomenon, equally clearly, requires a
reference to the factors that keep it there.12 So how could a reference to
virtual self-regard serve to explain anything? In other words, how can
our model of the common-cum-economic mind serve to make sense of
the explanatory claims of economics, in particular of the economics of
non-market behaviour: of behaviour that is motored by the perception of
what situations demand, under relevant cultural frames, not by consid-
erations of self-regard?
The answer, I suggest, is that even if virtual self-regard is of no use in
explaining the emergence or continuation of any pattern of behaviour, it
can be of great utility in explaining a third explanandum: the resilience of
that pattern of behaviour under various shocks and disturbances.
Imagine a little set-up in which a ball rolls along a straight line ± this,
say, under Newton's laws of motion ± but where there are little posts on
either side that are designed to protect it from the in¯uence of various
possible but non-actualized forces that might cause it to change course;
they are able to damp incoming forces and if such forces still have an
effect ± or if the ball just drifts ± they are capable of restoring the ball to
its original path. The posts on either side are virtual or standby causes of
the ball's rolling on the straight line, not factors that have an actual
effect. So can they serve any explanatory purpose? Well, they cannot
explain the emergence or the continuation of the straight course of the
rolling ball. But they can explain the fact ± and, of course, it is a fact ±
that not only does the ball roll on a straight line in the actual set-up, it
sticks to more or less that straight line under the various possible
contingencies where perturbing forces appear and even have a temporary
effect. They explain the fact, in other words, that the straight rolling is
not something fragile, not something vulnerable to every turn of the
wind, but rather a resilient pattern: a pattern that is robust under various
contingencies and that can be relied upon to persist.
The resilience explained in this toy example may be a matter of
independent experience, as when I discover by induction ± and without
understanding why ± that the ball does keep returning to the straight
line. But equally the resilience may only become salient on recognizing
the explanatory power of the posts: this, in the way in which the laws
that a theory explains may only become salient in the light of the
12 I ignore the requirements of potential explanation ± fact-defective or law-defective
explanation ± as that enterprise is discussed by Robert Nozick (1974). It may be
interesting to know how something might have come about or might have continued to
exist under a different history, or under a different regime of laws, but the interest in
question is not that which motivates ordinary economic attempts at explanation.
92 Rationality and homo economicus

explanatory theory itself. It does not matter which scenario obtains. In

either case the simple fact is that despite their merely standby status, the
posts serve to resolve an important matter of explanation. They explain,
not why the pattern emerged at a certain time, nor why it continues
across a certain range of times, but why it continues across a certain
range of contingencies: why it is modally as distinct from temporally
persistent.
The lesson of our little analogy should be clear. As a reference to the
virtually ef®cacious posts explains the resilience with which the ball rolls
on a straight line, so a reference to a merely virtual form of self-regard
may explain the resilience with which people maintain certain patterns of
behaviour. Imagine a given pattern of human behaviour whose continua-
tion is actually explained by the cultural framing under which people
view the relevant situations or, more prosaically, by people's sheer
inertia. Suppose that that pattern of behaviour has the modal property of
being extremely robust under various contingencies: say, under the
contingency that some individuals peel away and offer an example of an
alternative pattern. The factors that explain its actually continuing may
not explain this robustness or resilience; there may be no reason why the
example of mutant individuals should not display a new way of viewing
the situation, for example, or should not undermine the effects of inertia.
So how to explain the resilience of the pattern? Well, one possible
explanation would be that as the contingencies envisaged produce a
different pattern of behaviour, the alarm bells of self-interest ring ± this,
because of the contrast between what different individuals are doing ±
and the self-regarding deliberation that they prompt leads most of the
mutants and would-be mutants back towards the original pattern.
The analogy with the rolling ball serves to show how in principle the
model of virtual self-regard may leave room for the economic explana-
tion of behaviour that is not actively generated by considerations of self-
regard. But it may be useful to illustrate the lesson more concretely.
David Lewis's (1969) work on convention is often taken as a ®rst-rate
example of how economic explanation can do well in making sense of a
phenomenon outside the traditional economic domain of the market. He
invokes the fact that conventions often serve to resolve certain problems
of coordination ± problems of a kind that can be nicely modelled with
game-theory techniques ± in explanation of such conventions. But what
is supposed to be explained by Lewis's narrative? Lewis is clearly not
offering a historical story about the emergence of conventions. And,
equally clearly, he is not telling a story about the factors that actually
keep the conventions in place; he freely admits that people may not be
aware of the coordination problem solved by conventional behaviour
 The virtual reality of homo economicus 93

and may stick to that behaviour for any of a variety of reasons: reasons
of inertia, perhaps, or reasons of principle or ideology that may have
grown up around the convention in question.
The best clue to Lewis's explanatory intentions comes in a remark
from a later article when he considers the signi®cance of the fact that
actually conventional behaviour is mostly produced by blind habit. `An
action may be rational, and may be explained by the agent's beliefs and
desires, even though that action was done by habit, and the agent gave
no thought to the beliefs or desires which were his reasons for action. If
that habit ever ceased to serve the agent's desire's according to his beliefs,
it would at once be overridden and corrected by conscious reasoning'
(Lewis 1983, 181; my emphasis). This remark gives support to the view
that what Lewis is explaining about convention, by his own lights, is not
emergence or continuance but resilience. He implies that the servicing of
the agent's ± as it happens, self-regarding ± desires is not the actual cause
of the conventional behaviour but a standby cause: a cause that would
take the place of a failing habit, so long as the behaviour remained
suitable; this, in the way that he says it would displace the remaining
habit at the point where the behaviour becomes unsuitable. And if the
servicing of self-regard is a standby cause of this kind, then what it is best
designed to explain is the resilience, where there is resilience, of the
conventional behaviour.
But it is not only the Lewis explanation of conventional behaviour that
lends itself to this gloss. Can we explain American slave-holding by
reference to economic interests (Fogel and Engerman 1974, 4), when
slave-holders articulated their duties, and conducted their business, in
terms of a more or less religious ideology? Yes, to the extent that we can
explain why slave-holding was a very resilient institution up to the time
of the civil war; we can explain why the various mutants and emancipa-
tionists never did more than cause a temporary crisis. Can we explain the
failure of people to oppose most oppressive states as a product of free-
rider reasoning (North 1981, 31±2), when it is granted that they generally
used other considerations to justify their acquiescence? Yes, so far as the
free-riding variety of self-regarding reasoning would have been there to
support non-action, to make non-action resilient, in any situation where
the other, actual reasons failed to do so and alarms bells rang. Can we
invoke considerations of social acceptance to explain people's abiding by
certain norms, as I have tried to do elsewhere (Pettit 1990), when I freely
grant that it is considerations of a much less prudential kind that keep
most people faithful to such norms? Yes, we certainly can. Self-regarding
considerations of social acceptance can ensure that normative ®delity is
robust or resilient if they come into play whenever someone begins to
94 Rationality and homo economicus

deviate, or contemplate deviation, and if they serve in such cases to

restore or reinforce compliance.
If it is granted that the resilience of phenomena like these is explicable
by reference to virtual self-regard, I should add, then it becomes
plausible that self-regard may explain something else as well. We are
assuming that the day-to-day continuation of the patterns in question is
explained in other terms: say, by reference to culturally established ways
of thinking. But the resilience-explanation, assuming it is sound, suggests
that there are likely to have been crises in the past where the virtual self-
regard invoked in the resilience-explanation was actually triggered and
where it had the effect of preserving the pattern under discussion. We
may or may not have independent evidence of such crises but it becomes
plausible to conjecture that there were some and that self-regard serves
to explain the continuation of the pattern, not in day-to-day situations,
but in the presence of those crises. I make this point, however, only in
passing. The main claim I wish to defend is that even if self-regard serves
in no way to explain emergence or continuation, still it can explain
resilience.
The upshot will be clear. We can make good sense of economic
explanation, even explanation of non-market behaviour, in terms of the
model of virtual self-regard whereby the economic mind is reconciled
with the common mind. That model recommends itself, then, on at least
two grounds. It shows that the assumptions which economists make
about the human mind, in particular about human motivation, can be
rendered consistent with the assumptions of commonplace, everyday
thinking. And it shows that so interpreted, the assumptions motivate a
promising and indeed developing programme for economic explanation:
and explanation, not just in the traditional areas of market behaviour,
but across the social world more generally.

5 A more general view

But not only does the story that we have told show how economics, even
an economics of self-regarding agents, can have something to say in
explanation of ordinary social behaviour: in particular, behaviour
outside the market. It does so in a way that ®ts the explanation to
broader patterns. While the idea of explaining the resilience of a
behavioural pattern ± its resilience as distinct from its presence or
emergence ± may look suspiciously novel, it connects closely with quite
familiar styles of theoretical explanation.
 The virtual reality of homo economicus 95

One is the explanation of the ®tness conferred by a certain trait: the

explanation that consists in showing why the trait is adaptive. That a
trait confers a certain degree of ®tness means that in the relevant
environment the bearer has a certain propensity to survive ± a certain
propensity to be replicated in ± a variety of more or less probable
contingences.13 Thus any explanation of ®tness by reference to the
adaptiveness of a trait is just like the explanations considered in the last
section. Where we spoke of explaining the resilience of conventional or
normative behaviour, of slave-holding or of prudence, we might just as
well have spoken of explaining the ®tness associated with those patterns
of behaviour.
This observation keys us to the fact that the explanations we men-
tioned are also akin to those so-called functional explanations in
sociology that attempt to explain the ®tness or survival potential of
certain institutions, presenting them as more or less ®xed features of the
society: as features ®t to survive a large range of contingencies (Pettit
1996, 2000).14 Take any institution such that whatever the reasons it
obtains now, its role in private or public life means that were it to come
under challenge, effects would materialize to keep it in place. The socio-
logical explanation that points up that role and that argues for the
survival potential of the institution by reference to that role parallels
quite nicely the explanations considered here. One such explanation
might argue for the survival potential of golf clubs by reference to the
fact that golf clubs enable members to establish important business
contacts; members may not be aware of this now but they would become
aware of it under those pressures that might otherwise drive them away.
Another such explanation might argue for the survival potential of a
harsh prison regime by reference to the fact that such a regime enables
politicians to satisfy the public outrage which reliably follows any
heinous crime: it enables them to display the right body-language,
presenting themselves as tough on crime.
But the most obvious pattern of explanation with which our story
connects is equilibrium explanation. This is the explanation of a fact or
pattern which does not show how it emerged or why it is present, but

13 Fitness is a probabilized version of resilience under which resilience can be increased, not
just through an increase in the number of contingencies guarded against, but also
through an increase in the probability of the contingencies against which guards are
provided.
14 What I sketch here, of course, is a revisionist account of functional explanation in
sociology: an account under which it might be better described as ®tness-explanation.
See Pettit 1996 for details and Pettit 2000 for a comparison between rational choice
theory and functionalist explanation.
96 Rationality and homo economicus

which demonstrates that the pattern is more or less inevitable, at least in

a certain context, by pointing out that any ways in which it is liable to be
disturbed would lead to correction. As an example Elliott Sober (1983)
offers us R. A. Fisher's explanation of the 1:1 sex ratio in many species.
The idea is that if a population ever departs from equal numbers of males
and females, then there will be a reproductive advantage favouring
parents who overproduce the minority sex and the 1:1 ratio will tend to
be restored. Such an equilibrium explanation can be seen, in our terms,
not as a distinctive way of explaining things ± not as a distinctive
explanans ± but rather as a way of explaining a distinctive explanandum.
That the sex ratio is in equilibrium, or that any pattern represents an
equilibrium ± strictly, a stable equilibrium ± is a way of saying that it
enjoys a particularly high degree of resilience. Being in stable equi-
librium, at least for a given context, is a limit case of being resilient.
It hardly needs saying that while the notion of a stable equilibrium is
variously construed, equilibrium explanation is a standard and staple
practice in ordinary economics. What appears at this point, then, is that
the story we told in vindication of economic explanation outside the
market area vindicates it in a manner that ought to appeal to most
economists. In pursuing equilibrium explanations economists show a
day-to-day concern with explaining the resilience of certain patterns and
not ± or at least not necessarily ± their emergence or persistence. Thus
they ought to have no dif®culty in recognizing the signi®cance of the
sorts of explanations discussed here.15

References
Axelrod, Robert (1984) The Evolution of Cooperation, New York: Basic Books.
Becker, Gary (1976) The Economic Approach to Human Behaviour, Chicago:
University of Chicago Press.
Brennan, H. G. and J. M. Buchanan (1981) `The Normative Purpose of
Economic ``Science'': Rediscovery of an Eighteenth Century Method',
International Review of Law and Economics, 1, 155±66.
Brennan, H. G. and Philip Pettit (1993) `Hands Invisible and Intangible',
Synthese, 94, 1993, 191±225.
(2000) `The Hidden Economy of Esteem', Economics and Philosophy, 16,
77±98.

15 This paper represents a further development of a theme in Pettit 1993. It overlaps in

some part with the text of three lectures given at the Ecole des Hautes Etudes en Sciences
Sociales, Paris and published as `Normes et Choix Rationnels', Reseaux, 62, 87±112. I
was greatly aided in preparing the ®nal draft by comments received from Uskali MaÈki,
Raimo Tuomela, and an anonymous referee.
 The virtual reality of homo economicus 97

Downs, Anthony (1957) An Economic Theory of Democracy New York: Harper.

Eells, Ellery (1982) Rational Decision and Causality, Cambridge: Cambridge
University Press.
Fogel, R. W. and S. L. Engermann (1974) Time on the Cross: The Economics of
American Negro Slavery, Boston: Little, Brown.
Gauthier, David (1986) Morals by Agreement, Oxford: Oxford University Press.
Lewis, David (1969) Convention, Cambridge, Mass.: MIT Press.
(1983) Philosophical Papers, Vol. 1, New York: Oxford University Press.
(1986) Philosophical Papers, Vol. 2, New York: Oxford University Press.
North, Douglas (1981) Structure and Change in Economic History, New York:
Norton.
Nozick, Robert (1974) Anarchy, State and Utopia, New York: Basic Books.
Pettit, Philip (1990) `Virtus Normativa: Rational Choice Perspectives', Ethics,
100, 725±55.
(1991) `Decision Theory and Folk Psychology', in Susan Hurley and Michael
Bacharach, eds., Essays in the Foundations of Decision Theory, Oxford:
Blackwell, 147±75.
(1993) The Common Mind: An Essay on Psychology, Society and Politics, New
York: Oxford University Press. Pb edition, with new postscript, 1996.
(1996) `Functional Explanation and Virtual Selection', British Journal for the
Philosophy of Science, 45, 1996.
(1997) Republicanism: A Theory of Freedom and Government, Oxford: Oxford
University Press.
(1998) `Practical Belief and Philosophical Theory', Australasian Journal of
Philosophy, 76, 15±33.
(2000) `Rational Choice, Functional Selection and Empty Black Boxes',
Journal of Economic Methodology, 7, 33±57.
Pettit, Philip and Robert Sugden (1989) `The Backward Induction Paradox',
Journal of Philosophy, 86, 169±82.
Platts, Mark (1980) Ways of Meaning, London: Routledge.
Satz, Debra and John Ferejohn (1994) `Rational Choice and Social Theory',
Journal of Philosophy, 91, 71±87.
Schick, Frederic (1984) Having Reasons: An Essay on Rationality and Sociality,
Princeton: Princeton University Press.
Sen, Amartya (1982) Choice, Welfare and Measurement, Oxford: Blackwell.
Simon, Herbert (1978) `Rationality as Process and as Product of Thought',
American Economic Review, 68, 1±16.
Sober, Elliott (1983) `Equilibrium Explanation', Philosophical Studies, 43,
201±10.
Taylor, Michael (1987) The Possibility of Cooperation, Cambridge: Cambridge
University Press.
6 Expressive rationality: is self-worth
just another kind of preference?

SHAUN HARGREAVES HEAP

1 Introduction

People commonly re¯ect on what they do and these re¯ections give rise
to feelings of self-worth, or sometimes the reverse when feelings of
shame, guilt and embarrassment are experienced. It seems plausible to
think that these feelings, or their anticipation, in¯uence action in some
way or another. Indeed, I take it for granted that, roughly speaking,
people act in ways which encourage a sense of self-worth. This paper is
concerned ®rst with whether the conventional account of action as
instrumentally rational offers a way of understanding such feelings and
their connection to action. Action in the instrumental model is under-
taken to satisfy preferences (or it conforms to a preference ordering) and
it might be natural to think of self-worth as just another kind of
preference which motivates people to act (or it is built in to a preference
ordering). However, I argue in the next section that action which
supports self-worth cannot be readily or easily subsumed within that
model because self-worth cannot be reduced to some kind of preference.
Does this matter? Since no simple account of action could hope to
capture exhaustively what makes people tick, it is not self-evident that
the failure in this regard necessarily tells against the mainstream reliance
on the instrumental model. However, in the third and fourth sections, I
argue that the failure to include the re¯ective capacities which are
responsible for the generation of self-worth in the ontology of economics
does important damage to mainstream economics. In particular, it leads
to an impoverished understanding of the three classic games of social and
economic life: the prisoner's dilemma, the coordination and the battle of
the sexes game.

My thanks go to Uskali MaÈki for comments on an earlier version of this paper.

98
 Expressive rationality 99

Throughout the argument, I shall be focusing on the type of public sense

of self-worth that comes from behaviour which conforms with (or
breaches) the shared standards or norms of a group. In these circum-
stances, the judgement about what is worthy in an action is shared by
others who subscribe to the norm. So when one's action conforms (or
breaches) a norm it is regarded as worthy (or shameful) for the same
reasons by a group. This endows action with a symbolic dimension: the
action becomes a voice to the shared reasons that ®nd it worthy, so
enabling people to say things about themselves to others who share that
norm. As a result, action that is taken in support of self-worth in such
conditions is often referred to as expressively rational (since the action
comes to express something about the individual, see Hargreaves Heap
1989). Thus the paper is concerned with whether expressive rationality is
reducible to instrumental rationality. Alternatively, as the shared stan-
dards which enable action to acquire this expressive dimension are the
key to much anthropological investigation, this paper might be under-
stood in yet a further way. It is concerned with whether the economists'
rational choice model can account for what anthropologists typically ®nd
distinctive about human activity: that is, the shared meanings people
attach to what they do; and if not what does economics lose through this
blindness to anthropology.1

2 Self-worth is not simply another kind of preference nor do people get a

sense of self-worth from preferences satisfaction

The conventional rational choice model of action often forgoes any

explicit account of the process behind actions. It is simply action which
conforms to the axioms of rational choice. This is a rather strange legacy
of behaviourism in economics (e.g. see Stewart 1995) and for the reasons
which make behaviourism generally dif®cult in social science (and
because this volume is concerned with ontology), I shall treat it as a
curiosity and focus on the motivational account of rational choice which
is supplied when the rational choice model is pressed to account for its
internal workings. This model depends on two elements: preferences over
outcomes and beliefs regarding the connection between actions and
1 Of course, Adam Smith in his Theory of Moral Sentiments made much of an individual's
capacity to consider how their actions would seem to some impartial spectator. So, this
paper might alternatively be viewed as study in what if anything has been lost to our
understanding of economic agency by forgetting the in¯uence of Adam Smith's impartial
spectator.
100 Rationality and homo economicus

outcomes. These elements together with an instrumentally rational

disposition supply the ontology of a person. Thus the rational agent of
rational choice theory acts having used their beliefs to calculate what
action will best satisfy their preferences. The question addressed in this
section is whether it is possible to make sense of action which seems to be
in¯uenced by a concern with a public sense of self-worth (or the
avoidance of guilt, shame, and embarrassment) within this model of how
people act. In other words, does the ontology of the rational choice
model allow for what anthropologists call expressive action?
I shall consider two possible ways of accounting for expressive actions
within the rational choice model. The ®rst and perhaps obvious way is to
treat self-worth as one among possibly many motives which lie behind a
particular person's preferences. Thus people can be acting in support of
their self-worth when they act on their preferences. The conventional
theory is usefully quiet for this purpose on the source of an agent's
preferences. The origin and nature of a person's preferences does not
matter for the theory, so it seems perfectly possible for a desire for self-
worth to be one of the motives which in¯uences a person's preferences
over outcomes. All that is required by the theory is that preferences
should be well-behaved in the sense that they supply a preference
ordering and that they can be taken as given. The ®rst requirement is
synonymous with rational choice theory and needs no further explana-
tion. The second is less frequently formally stated, perhaps because it is
too obvious to need saying. But it is nonetheless important because when
preferences change then, at the least, a complete account of action will
also require an account of preference formation. This is in part a point
about the scope of the instrumental model. It does not tell the whole
story in these circumstances, but this need not be especially worrying as it
is often rather dif®cult to tell the full story. If, however, the gap is ®lled
with an account of preference formation that makes preferences depend
endogenously on action, then the damage to the instrumental model is
potentially much deeper as the very sense that action ¯ows from
preferences will be undermined if the preferences themselves are formed
by action.
To assess this strategy for allowing self-worth into the instrumental
model of action, it is helpful to consider what is involved in the activity
of re¯ection which gives rise to feelings of public self-worth (or their
reverse). For this purpose, I distinguish two key elements in the self-
re¯ection on action that gives rise to self-worth.
First, a sense of self-worth comes from evaluating the action and this
entails comparing the action that was undertaken with others which were
available. It is the comparison that gives meaning to the action which
 Expressive rationality 101

then re¯ects well (or otherwise) on the agent. They could have done `x',
but they did `y' and that says something about the person. For example,
it will be plain that when I invite someone for dinner and I serve a formal
meal with several courses, I will be indicating to the guest that I honour
the occasion in a way which is very different from the meaning I would
convey by serving them baked beans on toast. My action here says
something about myself in relation to my guest. On the other hand, if I
am too poor to serve a formal meal, then the meaning that my guest will
attach to the receipt of beans on toast would be very different. Hence
what any particular action like the serving of baked beans on toast
means (or says about its author) is not ®xed: it depends on the set of
available actions from which it was selected.2
Secondly, the generation of self-worth from the meaning attached to
`doing ``x'' when ``y'' was available' sometimes depends on a public
standard or key for interpretation. For example, suppose I believe that
formalities are the bane of life and should be dispensed with whenever
possible. I express this commitment through my dinner menu by giving
my guest beans on toast as this is what happens to be the ®rst thing I see
in my cupboard. The guest knows that I could afford a formal meal with
several courses and so knows that beans are not to be interpreted as a
sign of poverty. However, I shall only obtain the sense of self-worth that
comes from showing, as I intend, that I am the kind of person who sets
no store in formalities, when my guest also interprets the receipt of beans
on toast in this way. If my guest interprets the receipt of beans on toast
in a different way, as meaning, for instance, that I take the guest to be
the kind of person who is not worth honouring with anything more than
beans on toast, then the action no longer serves my self-esteem. Indeed
quite the reverse is the case: I will have conveyed offence when I intended
something quite different.
In this example a shared standard or key for interpretation is crucial
for my action to generate public meanings that can then become a source
of my self-worth. I am not concerned with whether this is always the case
(as in anti-private language arguments), it is suf®cient for my purpose
that it is sometimes the case that a person's self-worth is experienced
through actions which have public meanings in which the agent ®nds a

2 In effect, this is a familiar point from linguistics and so perhaps needs no further
explanation when applied to the meanings associated with actions. See Culler 1976 on
how the meaning of any word depends in part in what it is not and so upon the whole set
of words in a vocabulary. For instance, the precise meaning of a referential word like
`river' will depend on the full set of words available for use in connection with moving
inland water: that is, whether there are words like `stream' or `brook' in the vocabulary.
102 Rationality and homo economicus

sense of self-worth.3 Thus I take it as a matter of fact that people's sense

of worth depends sometimes on how other people perceive them and I
draw the inference that people must share a language for interpreting
their actions for this to be possible, as otherwise no person can reason-
ably expect to convey reliable messages to others through their actions.4
If this brief analysis of how a public sense of self-worth is constituted is
accepted, then the ®rst strategy for subsuming self-worth into the
instrumental model of preference satisfaction looks shaky because the
`preferences' which re¯ect a sense of self-worth seem bound to become
context dependent in two senses. One type of dependence relates to the
set of options and the other relates to the norms or public standards for
interpreting selection from any given set. An example that is commonly
cited in the literature will help bring this out. Suppose a person chooses
an apple when offered a choice between an apple and a banana. Now
suppose they are offered the same choice again in the presence of third
option, a smaller apple, the person may well now choose the banana so
as to avoid the imputation of greediness, given the prevailing norm,
which might come from choosing what is seen to be a `large' apple. The
choice in these circumstances depends on both the set of options and the
prevailing norm (for example a different norm where the choice of a
large item was interpreted as a form of enthusiasm might plausibly have
generated the selection of the `large' apple). Such context dependence is
bound to be worrying for the instrumental theory of choice when the
primitives of choice are apples, bananas, and their like because this
dependence can be a source of apparent preference change, as in this
illustration when apples are sometimes preferred to bananas and some-
times bananas are preferred to apples.
Of course it is possible to avoid the inconsistency by making the
primitives more nuanced: the apple in the ®rst choice is not the same as
the `large' apple in the second choice even though it shares the same
physical characteristics. But this concedes that the preferences are

3 The distinction between shame and guilt is sometimes made to turn on whether the
standard is public or private. So, for those who follow this distinction, the remainder of
this paper should be understood as focusing on shame rather than guilt.
4 Although I wish to sidestep the private language argument here, it seems to me that the
public sharing requirement arises for essentially the same reason. It is a simple
consequence of avoiding judgements of self-worth becoming self-serving. If I am free to
judge the act potentially in any way that I please then the standard and the judgement are
prone to be partial in a way that will offer little psychic comfort. The sense of well-being
comes precisely from the fact that the standard is independent of oneself and the public
sharing of the standard is a way of achieving a measure of independence (see Taylor 1989,
and on a related distinction between internal and external reasons, see Williams 1981).
 Expressive rationality 103

not `given' in the sense that they can change for reasons which have
nothing to do with either the physical attributes of the good or the
rationality of the agents. Nevertheless at ®rst glance this need not seem
such a troubling concession since agents will still be acting in a broadly
maximizing fashion (the `utility' function which is maximized is now
de®ned not only over actions but over the set of available actions; see Sen
1994). However further re¯ection reveals three important problems.
Firstly both the Savage and von Neumann-Morgenstern approaches to
the axiomatization of this theory of rational choice under uncertainty
require that the primitives of choice are de®ned independently of the
choice set (and the states of nature; see Sugden 1991). So the existence of
context dependent preferences creates a signi®cant problem for those
who hold to the axiomatic interpretation of this form of rational action
(i.e. those who adopt a behaviourist approach to rational action and
sidestep any account of the internal workings of agency).
Secondly such context dependence creates problems for the empirical
testing of this theory of action. The trouble arises because there is always
a tricky question of interpreting whether such and such an event
constitutes a test in any theory and the scope of such interpretational
problems increases greatly when context is allowed to affect preference in
this way. The point is simple. There are rules (within a scienti®c
community) for interpreting whether, say, an apple really is an apple for
the purposes of testing the theory in question and once preferences apply
not just to apples but to `apples in their context', these rules have in effect
become much looser because the grounds for claiming that an apple is
not an apple have expanded from physical differences to include differ-
ences in context. Indeed, it is tempting to think that such a radical licence
to individuate the circumstances of an action might undermine com-
pletely the activity of empirical testing.
Thirdly, even if these empirical problems are not regarded as a decisive
objection to the instrumental theory of rational choice in these circum-
stances, it is nevertheless clear that the operationalization of that theory
will depend on developing a theory of context which seems likely to
draw, in part, on our understanding of how self-worth is publicly
generated. In other words, one can preserve the instrumental account
only by attaching to it the elements of a theory of expressive rationality.
Whether in these circumstances it still makes sense to talk of the
instrumental model may seem a matter of taste, but it may also turn on
precisely how context here interacts with preference (see the earlier
comments on the worry when preferences become endogenous). I do not
want to beg this question now, but in what follows, so as to avoid any
confusion, I shall keep the instrumental and expressive models distinct
104 Rationality and homo economicus

and use the instrumental model to refer to an instrumental model

unadorned by any expressive embellishment.
The second, and alternative, strategy for incorporating considerations
of public self-worth in the instrumental model would be to concede that
self-worth is indeed, in general, something different from preference
satisfaction. However, it so happens that we live in a society where
people get self-worth from satisfying preferences. The satisfaction of
preferences is the shared standard, so to speak. In this way the underlying
model of action is unaffected by the observation that people care about
self-worth because self-worth happens to come from behaving according
to the postulate of rational choice. So, happily, preference satisfaction
and self-worth amount to the same thing.
This strategy also suffers from problems because self-worth in this
sense could never be generated by the mere activity of satisfying your
preferences. That is, a person cannot, in general, value preference
satisfaction and generate self-worth by being good at this activity. The
reason is that while such a standard can be publicly shared, it cannot, in
general, be publicly veri®ed. It is impossible for a person to demonstrate
their competence in satisfying preferences unless their preferences are
publicly known (independently of action) and for genuinely personal
preferences this will not be possible. Of course, one could gain a sense of
self-worth from something like generating wealth (or consumption)
because this is a standard which, when shared, is publicly veri®able, but
wealth generation is not the same as preference satisfaction. Indeed
any move towards a model of norm-driven behaviour in the form of
wealth maximization (or some other publicly observable objective) rather
than preference satisfaction would seem in effect to concede that self-
worth rules absolutely (at least in so far as self-worth comes from
behaviour which conforms to norms). In other words, self-worth would
be incorporated but at the expense of the rational choice model of action
altogether.
If these arguments are granted then neither of the two obvious ways of
subsuming self-worth within the rational choice model seems likely to
work. Of course, we might perhaps reinterpret all action as norm driven
in support of self-worth instead and so solve the dilemma the other way
round. But this seems unsatisfactory in the sense that most people would
acknowledge that preference satisfaction captures some part of what
makes us tick. Indeed, if the model of norm-driven behaviour is general-
ized (that is it becomes the only account of action) then it seems liable to
suffer from the mirror image of the problem encountered by generalizing
the model of preference satisfaction to all forms of action. How, for
example, would the generalized norm-following model account for the
 Expressive rationality 105

agent's action when norms con¯ict? In such circumstances it is natural to

draw on some further model of individual agency to explain how choice
is made (e.g. perhaps the model of preference satis®ers). In short, it seems
that a concern with (a) self-worth which comes from following/breaching
norms and (b) individual preferences satisfaction need to be held as two
separate types of motivational in¯uence on behaviour. The one cannot
be sensibly reduced to the other whichever way the reduction is
attempted.
This conclusion naturally raises the question of what the relationship
is between the two types of motivation in practice. I have no general
answer to this question but I do have a contribution to offer. It comes in
the next two sections. I focus on two examples where an economics
which is based solely on preference satisfaction fails in important respects
and where the failures can be remedied through recognizing the agents'
concern for self-worth.

3 The role of self-worth in economic explanations: what is lost, part I

If agents are concerned to say something about themselves through their

actions and these messages projecting the agent's self-worth are to be
distinguishable from simple preference satisfaction, then it will be
important for the action not to be open to interpretation as possibly
mere preference satisfaction. This suggests that the concern with self-
worth will be at play in settings where either action is potentially under-
determined by the logic of preference satisfaction or where what is
entailed by preference satisfaction is quite clear and the agent chooses to
act in a manner which confounds this calculation. In other words,
expressive actions occur either by complementing preference satisfaction
decisions when the logic of preference satisfaction is indeterminate or by
overruling the clear logic of the calculating preference satis®er.
Good examples of the ®rst type of opportunity for expressive action
are provided by any interaction which when modelled solely in preference
satisfaction terms has the form either of a coordination game or a battle
of the sexes game or is a game which is repeated inde®nitely. The key
features of such games for the purpose of this argument is that they have
multiple Nash equilibria. This in turn means that the agent who is guided
solely by instrumental reason will not know what to do in these games.
The point is straightforward. It is commonly recognized by game
theorists that an instrumentally rational agent knowing only that all
agents are rational in this sense (and have common priors) and common
106 Rationality and homo economicus

knowledge of the game will only know that a Nash equilibrium strategy
combination should be played. They will not know, however, which of
the many Nash equilibria to select. In other words in such settings the
model of agents as preference satis®ers fails to supply a complete account
of how agents act. Since many social interactions fall into one of these
three types, this would appear to be both a signi®cant shortcoming for an
economics which restricts itself to the preference satisfaction model of
action and a signi®cant opportunity for agents to use their actions to say
something about their non-preference satisfying qualities.
How the concern for self-worth (in the presence of norms) might ®ll
the explanatory gap here is unmysterious. The norm can provide so to
speak a different kind of reason for action. It can single out one action
where instrumental reason has no way of distinguishing between the two.
Suppose the interaction is a two-person game and there are two Nash
equilibria (x, x) and (y, y). To choose `x' rather than `y' can mean
something when there is a norm in place and so agents who are
concerned for their self-worth can have a reason for choosing, say, `x'
rather than `y'. In effect the choice of `x' says something about the agent
with which they would like to identify and when they both choose `x' in
this way they reaf®rm the shared value encoded in the norm (and of
course they also solve the equilibrium selection problem).
It is important not to overstate this potential. This is not a general
claim that expressive action plugs all these explanatory gaps because
there are other ways of supplementing the rational choice model for this
purpose. For example, the Nash re®nement project in game theory has
introduced further principles of `rational' thought, like back and forward
induction and like pay-off and risk dominance, that can help reduce the
number of Nash equilibria in some games. Alternatively it is possible to
appeal to the concept of salience and focal points (see Schelling 1963).
Salience or focal points are coordinating schemas which just occur to
agents and it would be foolish to deny their importance. For example,
when Schelling asked one of his famous questions to students concerning
where they would go to meet a friend in New York who like themselves
did not know the meeting place, there was surprising agreement on
Grand Central Station. It is dif®cult to imagine that any re¯ective
process would have singled out this location. After all, the moment one
begins to think about places in New York, many seem to have equal
claims as a meeting spot. Instead, Grand Central must just stick out
when people think of New York (or thought of New York in the 1960s,
as I am reliably informed people are now more inclined to meet at the
Empire State Building!).
The claim here is more restricted. It is that where both agents re¯ect on
 Expressive rationality 107

their action in settings with multiple Nash equilibria, then the re¯ection
and the action take this expressive form. The re¯ection cannot rely on
instrumental reason alone because this points only to a Nash action. So
some other kind of reason (or, more loosely, re¯ective resource) must be
at work and I offer the model of expressive reason.5 People act in the
presence of norms to say things about themselves: the actions are, so as
to speak, expressive and not only instrumental.
To support this narrower claim there are examples where re¯ection
does seem to be at work in equilibrium selection and it is easy to see how
these re¯ections might ®t the model of expressive action which I have
sketched. For example when Schelling asked his students to play
anonymously the equivalent of a battle of the sexes division game, they
almost all opted for the 50±50 division. Salience here seems to have an
obvious potential re¯ective base in the shared values of a liberal
democracy concerning how different agents are to be treated (i.e. equally
in the absence of any further reason for distinguishing between them).
Likewise in disputes between people that have the structure of the battle
of the sexes game (where the different Nash equilibria divide the bene®ts
of agreement differently between the players), it is common to ®nd in
experimental settings that a norm encoding a shared theory of justice
enables one of the equilibria to be selected (see Roth 1988). Thus when
these players act in the presence of such a norm they express a commit-
ment to a set of values which is a source of self-worth. Each is saying to
the other that they believe in a theory of justice and in acting on this
theory they enjoy a sense of self-worth.
It may be tempting to imagine that re¯ection of this sort is more likely
in situations where the Nash equilibria differ signi®cantly, as for instance
in battle of the sexes games. Here the different Nash equilibria divide the
bene®ts differently and so people seem likely to seek reasons for the
difference whereas in a pure coordination game there is no difference
between the characteristics of the equilibria and so there is no apparent
need to explain why one is selected by reference to these characteristics.
Serendipidity ought to be enough. Indeed one might imagine that the
kind of unre¯ective biases which are built into a habitual or routine form
of decision-making select the equilibria in such circumstances without
provoking much of a re¯ective twitch. Interestingly, however, even some
important coordination games seem to involve re¯ection and expressive

5 It is worth noting that when one agent is known to act in some unre¯ective manner on a
bias of the Grand Central variety, then the other agent will be able to use instrumental
reason alone since the other agent's bias is enough to anchor the instrumental agent's
instrumental reasoning process.
108 Rationality and homo economicus

action. For example consider the fashion game where people like to dress
in ways which are similar to their peer group. This is a pure coordination
game and yet in practice precisely because the peer group itself is formed
by a web of shared beliefs those beliefs tend also to attach to the uniforms
of the group (see Douglas 1978, where much action acquires symbolic
content because of the interlocking webs of shared belief ).
An example of the second type of opportunity for expressive action
comes in interactions which when viewed from the perspective of
preference satisfaction have the form of a prisoner's dilemma. In such
settings action is uniquely determined by the calculation of preference
satisfaction as there is a dominant strategy for each agent. As a result,
there is scope for an agent to signal that they ®nd worth in some other set
of values by acting in a manner which confounds the expectation that
preference satis®ers play the dominant strategy. I choose this particular
example because while there are many situations where the instrumen-
tally rational course of action is uniquely determined and so there is
scope for acting in a manner which is clearly contrary to that of
instrumental reason, these other settings are less likely to offer the
opportunity for expressive action. The point here is that acting in a non-
instrumental way is always capable of interpretation as `irrationality'.
Since expressing `irrationality' scarcely seems a sound base for generating
self-worth, it will be important for the action to be construed as worthy
in some other sense even though it is irrational when judged by the lights
of instrumental reason. The great virtue of the prisoner's dilemma inter-
action in this respect is the way that a play of the dominated cooperative
strategy opens up the possibility of achieving the collectively superior
outcome in the game. Thus the play of the cooperative strategy may be
`irrational', but it is not so obviously stupid as `irrational' play is in other
settings.
Again, to support this claim, there is plenty of evidence of people
apparently not playing the dominated strategy in prisoner's dilemma
games and it is not implausible to interpret these actions expressively. Of
course, for the reason cited earlier, in many settings the action associated
with preference satisfaction will not be unambiguously known as a
person's preferences cannot be publicly veri®ed, independently of action.
Thus the evidence from experiments where the people are told that the
structure of the interaction has the form of the prisoner's dilemma are
particularly instructive; and there is considerable evidence here which
shows that agents do not always play the dominant strategy. Of course,
this evidence can always be interpreted as lapses from rationality rather
than examples of agents applying some other sense of reason. However,
it is interesting that play of the dominant strategy seems to be less likely
 Expressive rationality 109

in experiments where the partner is known and so might plausibly share

a set of beliefs which would enable each of the players to interpret the
message which each player is sending to the other about themselves.
Moving from the laboratory, perhaps the most telling examples come
from those studies where ®rms manage to solve the apparent prisoner's
dilemma associated with expending effort by appealing to sources of
`trust' within the culture of the company (see Aoki 1990, and Casson
1991).

4 The costs of ignoring expressive reason: part II

I have argued in the previous section that the explanation of behaviour in

an important class of economic interactions frequently requires that the
public sense of self-worth that is generated through expressive actions be
recognized as a source of motivation. Those economic theories which
ignore the expressive orientation of agency suffer corresponding gaps in
their explanatory canon. The damage, however, goes further than this
because theory is often used not only for explanation but also for
prescription.
A variety of prescriptive problems might arise when agencies are
guided only by the instrumental understanding of the situation. For
example the exclusively instrumental understanding may lead to a set of
objectives which are different from the objectives that would guide policy
when there is a more rounded view of what motivates individuals. I do
not consider this problem here. Instead, I accept the objectives which
come from an instrumental understanding and focus on a particular
problem which can arise when a change that is introduced on the basis of
an instrumental understanding of the situation also affects the expressive
orientation of agents and in addition the instrumental and expressive
aspects of the situation now pull in opposite directions. The result is a
policy change which is less effective than had been expected and may
even be counter-productive.
To make this concrete, suppose following the instrumental understand-
ing of why people behave, it is thought desirable to encourage people to
behave in some way by offering greater ®nancial rewards to that
behaviour. The problem which concerns me in this section arises when
the ®nancial encouragement also alters the expressive orientations which
are at play in such a way as to undermine or offset the effects of the
change in the ®nancial rewards on behaviour. In such circumstances, the
neglect of the expressive dimension actually means that the effects of
110 Rationality and homo economicus

the policy are not only imperfectly understood, they are liable to be
smaller than expected and may even be the reverse of what was intended.
Two examples will bring this out.
First consider a company that perceives its employees' effort decision
solely in instrumental terms and believes there is a one-sided prisoner's
dilemma which needs to be solved. The circumstances are that the
worker can expend high or low effort at work. There is no way of
monitoring effort, so workers are paid a ¯at rate. Thus the dominant
strategy for a worker is perceived by the company to be low effort
expenditure since in the absence of detection the worker is thought
obviously to prefer to work less rather than more as no extra reward
comes with effort. Perceiving this, the ®rm introduces a monitoring and
payment system which transforms the decision over effort. Effort is
monitored and a high level is ®nancially rewarded. The dominant
strategy for the worker now becomes `making an effort' (the equivalent
of the previously dominated cooperative move in the effort version of the
prisoner's dilemma game). Let us further suppose that before the
introduction of this new payment system there was a norm which
allowed workers to express a commitment to professionalism or crafts-
manship to fellow workers through working with high effort. The norm,
however, went unrecognized by the ®rm because it perceived the situation
solely through instrumental eyes and seeing only the logic of the
dominant strategy of defection, it decided to introduce the new payments
system.
The direct consequence of the new payment system is the undermining
of the capacity to express their professionalism through high effort since
this is now what is required by instrumental calculation. To `defect' with
low effort, the non-instrumental move in these new circumstances, does
not provide a substitute expressive action since it seems most likely to be
interpreted as `irrational'. After all, it would seem to serve no conceivable
other purpose. Thus if the norm of professionalism depends for its
continuing in¯uence on acts of instantiation, then the change in pay-
ments system will undermine the norm more generally and this will have
potential spillover effects for other arenas inside the company. Speci®c-
ally there may be other dilemmas which now can only be solved through
changes in the payment system and so on. The result is that resources
may have to be mobilized across a range of activities inside the company
to encourage cooperative play when this was actually hitherto largely
generated (and costlessly) by the expressive actions of the workforce. In
other words, the contractual encouragement destroys the source of trust
in the relationship which hitherto generated cooperation.
This is one sketch of a worrying interplay between the instrumental
 Expressive rationality 111

and the expressive. Is it far-fetched? Perhaps, but there is much evidence

to suggest that the introduction of market/contractual relations within
parts of the public sector have undermined the professional ethos which
had guided many employees in that sector in the UK. Of course the
existence of the ethos does not demonstrate that the `effort' problems
were thereby, in large measure, previously solved. But its existence and
sudden disappearance provides some support for the thought that the
expressive and instrumental cannot always coexist and so the encourage-
ment of one may come at the expense of the other.
The second possible example of this interplay at work, this time in the
reverse direction, comes from the worrying emergence of persistently
high levels of unemployment in many OECD countries. To illustrate this
possibility, suppose initially that unemployment rises in a country above
the equilibrium or natural rate as a result of a demand shock. The
resulting rise in unemployment is involuntary: that is, the unemployed
are willing to work at the prevailing wage or less but can ®nd no job
offers at that or a lower wage. As a result the unemployed cannot explain
their unemployment in instrumental terms. It is not that they have
chosen leisure. They are simply unable to ®nd a job. Suppose also that
this involuntary unemployment persists because the government takes no
remedial action with respect to demand. Instead the government attacks
the problem on the supply side by tightening the unemployment bene®t
rules (perhaps through lowering the real value of bene®ts or lowering the
duration, etc.). This is a policy which works with an instrumental
understanding of the unemployment problem because such changes
increase the incentives for the unemployed to look for work. The
rationale behind the policy is simple. If the incentive to look for work is
increased then the unemployed should expend more effort on ®nding a
job and increased effort should bring its own reward. The problem, of
course, is that increased looking does not help the unemployed ®nd a job
when there are no extra jobs being created due to demand de®ciency.
I argued before that the scope for expressive action depended on the
instantiation of norms which allowed action to acquire a symbolic
dimension. In this unemployment context, it is not implausible to argue
the same with respect to instrumental reason. If instrumental calculation
offers no remedy or is not successful in its application, then it seems
likely that agents will appeal less to instrumental reason in their own
psychological account of where they are and increasingly seek an
expressive explanation of their position. In other words the unemployed
will increasingly begin to see the state of unemployment as some form of
self-expression. Of course, the creation of an expressive motivation does
not just depend on this psychological desire to ®nd a form of reason
112 Rationality and homo economicus

which accounts for the person's position, it depends on the existence of

norms which would enable the activity of being unemployed to be
interpreted expressively by others. This in turn requires instantiation:
that is, other people who are unemployed. But this is, of course, precisely
why a high level of unemployment is worrying because it creates the
material for the instantiation of a new norm among the unemployed.
Thus a high level of involuntary unemployment when it persists for long
periods can be turned into a permanently high level of equilibrium
unemployment as the unemployed develop a culture of unemployment.
This is but one sketch of how the phenomenon of hysteresis with
respect to unemployment might arise. It turns on the competition
between instrumental and expressive reason which takes place once
involuntary unemployment persists and the bandwagon of an unemploy-
ment culture begins to roll. Perhaps there are better explanations or the
creation of a culture of unemployment plays only a small part. Neverthe-
less the key mechanism here, the competition between instrumental and
other forms of reason, has received considerable support in the experi-
mental psychological literature on cognitive dissonance and cognitive
evaluation theory (see Deci 1975). And in the presence of such competi-
tion there are general grounds for doubting the wisdom of prescriptions
which are based exclusively on the instrumental view of agency because
they will fail to take full account of the effects of a policy change (see
Frey 1992 and Pettit 1995, for related arguments).

5 Conclusion

This paper has argued that the re¯ective capacity that agents use to
generate a public sense of self-worth seems to have disappeared from the
ontology of mainstream economics as a result of its reliance on the
instrumental model of action. That re¯ective capacity is neither there
explicitly and nor can it be recovered implicitly. This disappearance
matters because the exercise of expressive reason, which I have associated
with that capacity, can account for some of the explanatory shortcomings
regarding equilibrium selection in games with multiple Nash equilibria
and the failure to explain the incidence of cooperative behaviour in
prisoner's dilemma games. These explanatory failings are signi®cant and
provide strong grounds for restoring the expressive orientation of agents
to economics. This argument is further reinforced by the prescriptive
weaknesses which may also come from overlooking the expressive
motives which are sometimes at play in action.
 Expressive rationality 113

In short, when Robert Burns wrote the following he was on to

something which mainstream economics has regrettably tended to over-
look:
O wad some Pow'r the giftie gie us
To see oursels as others see us!
It wad frae mony a blunder free us
And foolish notion.

References
Aoki, M. (1990) `Towards an Economic Model of the Japanese Firm', Journal of
Economic Literature, March, 1±27.
Casson, M. (1991) The Economics of Business Culture, Oxford: Clarendon Press.
Culler, J. (1976) Sausurre, Glasgow: Collins.
Deci, E. (1975) Intrinsic Motivation, New York: Plenum Press.
Douglas, M. (1978) The World of Goods, Harmondsworth: Penguin.
Frey, B. S. (1992). `Tertium Datur: Pricing, Regulating and Intrinsic Motivation'
Kyklos, 45, 161±84.
Hargreaves Heap, S. (1989) Rationality in Economics, Oxford: Basil Blackwell.
Pettit, P. (1995) `Institutional Design and Rational Choice', in R. Goodin (ed.),
The Theory of Institutional Design, Cambridge: Cambridge University Press.
Roth, A. (1988) `Laboratory Experimentation in Economics: A Methodological
Overview', Economic Journal, 98, 974±1031.
Schelling, T. (1963) Strategy of Con¯ict, Oxford: Oxford University Press.
Sen, A. (1994) `The Formulation of Rational Choice', American Economic
Review, Papers and Proceedings, 385±90.
Stewart, H. (1995) `A Critique of Instrumental Reason in Economics', Economics
and Philosophy, 11, 57±83.
Sugden, R. (1991) `Rational Choice: A Survey of the Contributions from
Economics and Philosophy', Economic Journal, July, 751±85.
Taylor, C. (1989) Sources of the Self, Cambridge: Cambridge University Press.
Williams, B. (1981) Moral Luck, Cambridge: Cambridge University Press.
7 Agent identity in economics

JOHN B. DAVIS*

1 Introduction

Many economists would agree that as a social science economics aims to

explain cause-and-effect relationships in economic life, and aims to do so
in close conjunction with an analysis of agency. Yet though economists
constantly refer to economic agents when they speak of individuals
making choices, the concept of agency itself is not sharply de®ned in
economics. I de®ne agency as the power to initiate and bring about
events. Economic agents, then, have the power to initiate economic
activity, and, together with other causal factors, such as capital,
institutional arrangements, natural resources, etc., help bring about
events seen as the causal effects of their actions. Choice, economists'
main concern, is thus an aspect of agency in that its analysis helps us
account for how economic agents act on the world. But choice needs to
be understood in a wider framework of cause-and-effect explanation if
economic agents are to be understood speci®cally as agents with powers
to initiate and bring about events when making choices.
Another way of seeing economic agents as agents in a causal sense is to
recognize that on this conception economic agents actually exist, and
that adopting a cause-and-effect view of economic agents commits one to
an ontological realism about economic agents in the sense distinguished
by MaÈki in his taxonomy of realisms (1992). Note that while most
economists would say they are interested in actually existing economic
agents, many would also allow that the logic of choice ± or how agents
ought to behave ± may be evaluated apart from the question of whether
there are agents whose choice behavior that logic describes (Sugden 1991,

* Thanks to Tony Lawson, Uskali MaÈki, Jochen Runde, and Alex Viskovatoff ± without
implication ± for thoughtful comments on an earlier version of this paper.

114
 Agent identity in economics 115

p. 752). On the view here, however, the investigation of economic agency

requires one to suppose that there are economic agents with real powers
to bring about events as the effects of their actions. This emphasis calls
for some re-orientation of contemporary economics, since with the
considerable development of axiomatic decision theory in the postwar
period pursued apart from a realism about economic agents, it seems fair
to say that the discipline has not focused on explaining how economic
agents actually bring about events.
In the discussion that follows I begin to set out an ontological frame-
work for an analysis of economic agents by developing stategies for
investigating the location and scope of economic agency, where each of
these considerations pertains to a relatively separate set of issues involved
in an ontology of agents. A related way of approaching the topic of
agency that has tangencies to philosophers' work on personal identity is
as the investigation of the identi®cation of economic agents or agent
identity in economics. This paper consequently might also be thought to
investigate how we identify or account for the identity of economic
agents in terms of the location and the scope of their agency. Or, using
philosophers' personal identity language, the paper investigates both
how we distinguish or individuate agents ± the business of locating them
± and how we re-identify or track agents through change in their
characteristics and surroundings ± the business of determining what
scope their agency possesses or sustains.
The rationale for using philosophers' thinking about personal identity
to investigate agent identity in economics needs to be highlighted at the
outset. In supposing economic agents actually exist, one in effect assigns
oneself the task of explaining how and for how long economic agents
exist. How they exist is a matter of accounting for where they are active.
I label this the question of agents' location, though it will be readily seen
that more is involved than simply picking out agents' site of operation,
since where they are active leads us immediately to claims about the
manner in which they are active and some ascription of powers to them.
The question of how long agents exist concerns the durability of their
powers. I label this the question of the scope of economic agency. Thus
described, it will be seen that the issues of location and scope of agency
are means of explaining agents' causal effectiveness. This focus is slightly
different than is involved in philosophers' investigation of personal
identity, since only some approaches to personal identity emphasize the
effects of individuals' actions on the world. But the agent identity
approach adopted in this paper shares the general realist concern
personal identity theorists have regarding the persistence and distinctness
of a type of real being in the world. For this reason, personal identity
116 Rationality and homo economicus

theory offers insights for economics that may help develop an onto-
logically realist view of economic agents as having causal powers.
MaÈki, in his taxonomy of realisms, distinguishes referential realism and
representational realism. Referential realism, or the semantical thesis that
the terms of language and science refer to real entities, connects in
important respects to the issue of how we locate economic agents. Our use
of language re¯ects our ontological commitments, and one such commit-
ment is that the things we refer to are real entities. But to successfully refer
to real entities implies that we can ontologically distinguish or individuate
those real entities in the world. Thus we may begin to investigate the
location of agents through an analysis of our modes of reference. In
contrast, representational realism, or the semantical thesis that attributive
statements ascribe properties to real entities, has important connections
to the issue of what scope or extent of power economic agents may be said
to have. Here our ontological commitments speci®cally concern the
magnitude and duration of economic agency, or how long and with what
force real agents exercise their powers as agents. Thus we may begin to
investigate the scope of agency through an analysis of how economic
agents are represented in the language of economic theory.
This paper consequently attempts an analysis of the location and scope
of economic agency in terms of these two forms of realism, so as to link
ontological and semantical thinking on the subject. It postpones a veristic
realist analysis of agents (also distinguished by MaÈki) on the grounds
that establishing truths about the forms of economic agency ®rst requires
that one identify those agents whose actions have real effects in economic
life. Roughly speaking, ontology precedes epistemology. As an illustra-
tive framework, the second section of the paper begins by critically
examining the standard neoclassical view of economic agents, and
characterizes it as unsuccessful in its ability to identify economic agents.
The third section of the paper then offers an alternative, realist approach
to locating economic agents, discusses routines as a form of phenomena
that exhibits agency relationships, and closes with a proposed criterion
for locating agents. The fourth section turns to how we may address the
dif®cult topic of agency's scope, considers the point-in-time and through-
time dimensions this involves, and introduces considerations pertaining
to multiple sources of agency and changes in its structure. The paper
concludes with three brief remarks to clarify principal themes of the
discussion.
 Agent identity in economics 117

2 The neoclassical approach to locating economic agents

As an ontological issue, the location of agency concerns how we

distinguish or pick out agents in economic life. To appreciate the nature
of the issue, it is worth noting at the outset that this aspect of the topic of
agent identity is less central to philosophers who investigate personal
identity. When philosophers analyze the conditions under which indi-
viduals maintain their personal identities, they generally suppose that
they know whose identity is at issue (namely, that of individual human
beings), and thus rarely stop to worry about where agency is located (and
whether individual human beings are indeed agents). For example, in
Par®t's (1984) famous teletransporter examples regarding personal sur-
vival, he imagines what might happen to himself were his own body
dematerialized and reconstituted elsewhere under various scenarios.
Most economists also tend to assume that locating agents is not an issue,
since they believe that individuals are automatically agents. Even when
economists concern themselves with situations in which people make
choices for others, mobilize others' resources, and act as others' agents,
that is, in principal±agent situations, they assume there is no dif®culty in
locating agency, since delegated authority stems from choices previously
made by autonomous individuals who are agents. But it seems fair to say
that without an account of what makes an individual an independent
agent in the ®rst place ± one in whom the power to affect the world exists
± it is question-begging to say that individuals are even in a position to
delegate authority and assign decision-making responsibility to others.
Unlike philosophers in their treatment of personal identity, then, econo-
mists cannot avoid ®rst confronting the problem of how we locate
causally ef®cacious agents preparatory to a full analysis of the scope and
extent of their powers as distinguishable agents.
Note ®rst, then, that when we use language to refer to things, we
suppose that we can locate and distinguish the things we refer to. If one
uses the term ``chair'' referentially, this implies one can pick out an object
thought to be a chair, and distinguish it from other pieces of furniture.
This suggests we ought to look at how economic theories make reference
to the things they name. But before doing so, we ought to ®rst ask
ourselves what conception of reference a theory involves, since it is well
possible a theory employs a view of reference that does not help us to
pick out the sorts of things we are interested in ontologically identifying.
For example, it might be my theory of reference that ``whatever my
friends call a `dog' does indeed pick out real dogs in the world,'' where
that they also refer to people they dislike as dogs tells us that my theory
118 Rationality and homo economicus

of reference does not enable me to consistently pick out real dogs. Of

course these friends do refer when they use the term ``dog,'' but because
their use of the term is often metaphorical, they do not thereby help us to
disinguish and locate real dogs. Here my theory of reference is inap-
propriate to the task at hand. And since our project here is not to
investigate the rhetorical use of language, but to pursue the ontological
project behind referential realism, we ought similarly to ask whether our
economic theories employ theories of reference that will allow us to
locate and distinguish real economic agents.
Let us distinguish two philosophical approaches to reference for
constative utterances, the early analytic Fregean theory dating from the
turn of the century (Frege 1892; Russell 1905), and the more recent
causal theory associated with scienti®c realism (Kripke 1971; Putnam
1970, 1973). The Fregean theory explains the meaning of a name
(common noun or proper name) in terms of its intention, or in terms of
that conjunction of properties analytically true of the name. As Putnam
characterizes it,

On the traditional view, the meaning of say, ``lemon,'' is given by specifying a

conjunction of properties. For each of these properties, the statement ``lemons
have the property P'' is an analytic truth; and if P1, P2, . . ., Pn are all of the
properties in the conjunction, then ``anything with all of the properties P1, . . .,
Pn is a lemon'' is likewise an analytic truth. (1970, p. 51)

The theory then supposes that the intension of a name determines its
extension, where a name's extension is that to which it refers ± or more
simply that meaning determines reference. The theory is neo-Kantian in
that the logical structure of language (rather than as for Kant the
psychological structure of the mind) determines what can be picked out
and said about things in the world. But this is problematic, since it
essentially renders the idea of a world independent of thought empty (as
Kant did the notion of things-in-themselves), creates ambiguities between
the act of judging and that which is judged, and generates paradoxes of
meaning variance across conceptual structures aÁ la Kuhn.
On the causal theory, in contrast, it is argued that names come to refer
in a quite different manner. Things are, as it were, ``baptized'' at some
historical point in time, and then generally retain their original names
through causal chains of events linking the inaugural naming to later
language-users, despite modi®cations in meaning and the way names are
understood. On this theory reference is accounted for in terms of our
causal interactions with things in the world and among us, and presup-
poses that we can ontologically distinguish real entities from one another.
Unlike the neo-Kantian approach, that is, things in the world are ®rst
 Agent identity in economics 119

supposed to be distinguishable and re-identi®able through change, and

this conditions the way in which we use language to refer to them. Put
differently, reference is not established in an a priori manner by picking
out things to which de®nitions should apply, but rather in an a posteriori
manner by af®xing names to the things we distinguish.
In Davis (1989), I argued that neoclassical economic theory in its
axiomatic form employs a Fregean approach to reference, and that,
because the Fregean approach does not provide an adequate account of
reference, axiomatic neoclassical economic theory thereby fails to distin-
guish and pick out real economic agents in the world. Neoclassical
economic theory relies on the Fregean theory of reference, essentially
because it distinguishes and de®nes distinct economic agents in terms
their different preference sets. Of course it is not analytically true that
particular individuals have the particular preferences they do, but it is
analytically true that any given individual is de®ned in terms of his or her
own preferences. Our language sometimes obscures this point, because it
allows us to say two individuals may ``have'' the same preferences.
However, the theory clearly holds that two individuals may not have the
same preferences in the sense of one doing the other's preferring.
Preferring, as a propositional attitude, always implicates the particular
individual doing the preferring, and is thus inseparable from that
individual. Understood in this way, the theory holds that distinguishing
two individuals is analytically equivalent to distinguishing two sets of
own preferences.
How, then, does reference actually fail on this view? In Davis (1995), I
argued that using an individual's own preferences to distinguish that
individual presupposes the very individual those preferences are meant to
distinguish. If a set of preferences picks out an individual, they must be
that individual's preferences and not someone else's preferences. But if
we have already picked out the individual to whom a set of preferences
belongs in order to call these preferences that individual's own prefer-
ences, we cannot then use those preferences to pick out that individual.1
Neoclassical theory thus involves itself in a circularity in its implicit
account of individual and agent identity, arguably causing it to employ
the ad hoc assumption that preferences are exogenous to foreclose critical
evaluation of its individualist project. For our purposes, however, the
conclusion that is of chief interest here is that the theory is unable to
1 An early antecedent to this argument is Bishop Butler's charge that Locke's view, that
individual identity depends on memory and states of consciousness, was circular. Butler
argued that memory presupposes rather than explains personal identity. Hume, whose
critique of the idea of the self is better known, essentially followed Butler in this argument
(see Noonan 1989).
120 Rationality and homo economicus

locate real economic agents in the world. Its claim that individuals are
agents, that is, is not supported by an understanding of reference that
gives us a means of actually distinguishing individuals as real economic
agents.
It thus follows that neoclassical theory lacks a way of systematically
distinguishing between agents who are single individuals and agents
made up of groups of individuals. Gary Becker's family decision-making
analysis is a good example (1976). Husbands' preferences are taken to
represent wives' and other family members' preferences, thus substituting
the preferences of a single individual for a group of individuals. But
because own preferences cannot pick out individuals, this account cannot
distinguish (as feminists have continually argued) the case of a husband
acting on his own preferences in the name of the family from the case of
a husband acting on some sort of truly shared family preferences.
Subsequent intra-family bargaining models have attempted to get around
Becker's dif®culty by assuming that husbands and wives are both
independent agents whose negotiations create shared preferences for the
family as a joint agent. But this similarly begs the question of whether
husbands and wives are independent agents in the ®rst place. In
comparison to Becker's analysis that cannot determine whether husbands
are single agents or representatives of group agents, the bargaining
approach cannot determine whether individual agents constitute a group
agent or a pre-existing group agent enfranchises individual agents.
But that neoclassical theory cannot credibly distinguish between
agents who are single individuals and agents made up of groups of
individuals, implies that it fails to locate agency in the world. And, failing
in this ®rst step, it cannot proceed to investigate the issues surrounding
determination of the scope of agency, where that concerns the duration
and magnitude with which distinguishable economic agents exercise their
powers. Indeed in Davis (1995), it was also argued that even were we to
suppose individuals are distinguishable economic agents (or locate
agency in individuals), neoclassicism's dynamical analysis of human
capital accumulation precludes its re-identifying them as distinct agents
through change. Speci®cally, since on time allocation models (Becker
1965) certain of an individual's preferences are enhanced in signi®cance
by past accumulations of human capital, human capital accumulations
indirectly contribute to the constitution of an individual's identity over
time. But that we accumulate particular human capital stocks is partly a
function of the price system and thus the decisions and preferences of
others. This indirect in¯uence of other individuals' preferences on any
given individual disrupts neoclassicism's attempt to distinguish indi-
viduals in terms of their own preferences, making their re-identi®cation
 Agent identity in economics 121

through change problematic. And lacking an ability to track individuals

through change, the theory has few resources to explain the scope of
agency in terms of its magnitude and duration.
What this discussion of neoclassical theory is meant to demonstrate is
that addressing the problems involved in explaining agency requires that
a theory be formulated so as to take ontological commitments seriously.
One cannot, then, simply assume individuals are always and everywhere
agents, because one's ideological predilections encourage one to think
this must be the case. Rather, given this intellectual predisposition, one
needs to establish when and how individuals are able to act as agents.
Neoclassical theory, to be fair, is hardly alone in its limited success in this
regard, since similar sorts of arguments could be developed to show how
methodologically collectivist theories fail to refer to distinguishable and
re-identi®able group economic agents. In the section that follows an
attempt is made to describe how one might go about locating economic
agents through an analysis that places ontological realism in the fore-
ground.

3 A realist approach to locating economic agents

How should we proceed if our ®rst task is to explain how economic

agents are distinguished from one another? The circularity problem
associated with the Fregean approach to reference arises when one
attempts to distinguish agents using an a priori, de®nitional criterion,
primarily motivated by theoretical concerns (as the neoclassical prefer-
ence criterion is motivated by the desire to apply the theory of instru-
mental rationality to all economic agents). Suppose, however, one were
rather to attempt distinguishing economic agents in an a posteriori
manner where this involved making explicit use of the causal theory of
reference. In this instance, theory is built up around an inherited set of
ontological commitments, namely, that such-and-such sorts of things
exist in the world, which we are able to refer to by developing our
theories around past namings (or ``baptisms'') meant to pick out and
distinguish these real entities. Theory from this perspective follows rather
than precedes reference, and accordingly allows different theories to
share objects of reference, and so avoid the problem of meaning
variance.
It will naturally be said in response to this suggestion that premising a
set of ontological commitments itself involves some theory, and that it is
illusory to think one can postpone theorization until one has agreed on
122 Rationality and homo economicus

what one is referring to. This is certainly correct in that there is indeed a
theorization implicit in one's ontological claims, but I will put this
dif®culty aside here by assuming that the sort of implicit theorizing we
imagine attaches to many of our ontological commitments is extra-
theoretical in the sense of not belonging to any particular theory
developed by an identi®able collection of professional theorists. Different
economic theories, for example, can refer to ``markets'' as real entities,
but in doing so they adopt a very basic conception not reducible to any
one group's theorization of markets. Speci®cally, markets involve ex-
change, the items exchanged are valued, traders in markets have different
interests, some form of property relationship is implied, etc. Loosely, we
might say that rather than presuppose implicit theorizations, the use of
terms such as ``markets'' signals widely held beliefs we may understand
as constituting collections of ``social facts.'' Following Gilbert (1989) we
may characterize social facts as facts in the eyes of reasonably informed
individuals in society, where their ordinary understanding of these facts
contributes a conceptual element separate from the further interpretative
contribution that comes of more systematic analysis in social science
theory. The prior element of understanding contributed by society both
underlies social facts' status and acceptance as facts in society, and serves
as a conceptual penumbra about these facts that assists their further
analysis in more formal social science investigation. An a posteriori social
science approach, then, attempts to anchor investigation in what the
social world takes as given, rather than in a series of deductions made
from theories constructed around views of a subject matter's essential
requirements.2
From this perspective, we may suppose, ®rst, that analysis of the set of
social facts relevant to agency begins with attention to evidence regarding
how economic agents themselves understand the distribution and scope
of agency in the economy. Presumably agents of any sort must rely on
some understanding of agency in order to act as agents. Here the causal
theory of reference comes into play in that an acknowledged set of social
facts would naturally include facts about how historically we have come
to pick out and refer to agents. On the causal theory, reference depends
on historical namings or ``baptisms'' and subsequent chains of connec-
tion as names evolve, change, and are modi®ed. Economic agents, then,
rely on this information in acting as agents just as they rely on various
other ordinary social facts.
Second, our investigation of those social facts relevant to agency

2 The approach here also distances itself from the traditional empiricist project of beginning
with scientists' collections of observation statements.
 Agent identity in economics 123

requires we also recognize that, because economics in its mission as

causal social science aims to characterize economic life in terms of
relatively enduring relationships, the evidence regarding agency relevant
to our interests in economics concerns what we may observe in persistent
patterns of economic activity. Clearly we do not expect to learn much
about agency from evidence about episodic phenomena. These two
points together imply that the social facts which underlie an analysis of
agency in economics are those recognized by the economy's agents in
connection with recurring patterns of activity investigated in science. It
seems uncontroversial to add, thirdly, that these recurring patterns of
activity involve intentional behavior in situations of social interaction.
Intentional behavior, of course, is our principal concern in investigating
economic agency, while social interaction allows our scope to include
relations between individual agents, between group agents, and between
individual and group agents.
Following the lead of Nelson and Winter (1982), then, I propose that
these relatively enduring patterns economists wish to investigate be
initially described as routines or as routinized behavior. I de®ne routines
in economic life and elsewhere as organized collections of rules and
procedures that guide behavior in regularly encountered circumstances.
Routines help individuals recognize domains of activity in which they
may act. Routines do not themselves have causal powers, but rather
create a framework in social settings that permits the assignment of
agency to individuals who act.3 Central in this regard is the normative
force possesed by the rules and conventions that make up routines. Rules
and conventions tell us what we ought to do when we participate in a
given routine. Thus an individual who abides by a routine becomes an
agent in accepting the normative force of the rules and conventions
characteristic of that routine.4
Other units of social analysis might be proposed, but there are
important advantages to focusing on routines in an analysis of agency.
First, since routines may be investigated for both individuals and groups,
an examination of routines per se is not likely to bias our analysis at the
outset toward either individualist or collectivist interpretations of the
location of agency. Indeed both individual and group responsibilities are
often recognizable in the duties that many routines imply. Second,
because routines tend to be relatively self-contained social practices, they
3 Obviously some routines are personal and carried out mostly apart from other
individuals. The discussion here focuses on those routines that involve social interaction.
4 I do not wish to suggest that agency does not exist outside of the context of routines.
Routines, however, provide a valuable framework in which agency may be explained in
economic life.
124 Rationality and homo economicus

appear more susceptible to analysis than many other more complex

forms of behavior. More complex behavioral settings can indeed plau-
sibly be argued to combine routines, so that these ``institutional''
environments are explained in terms of routines as their elements. Third,
routines are an ordinary feature of the social world, and as such their
common-sense description does not require immediate introduction of
elaborate theoretical constructs. In fact, almost anyone can both describe
a variety of routines in which people engage, and on prompting cite a
number of facts relevant to the scope and location of agency that these
routines imply.
For example, in economic life, say in the context of business ®rm
routines, many can describe routines employees regularly observe, and
analyze this evidence to produce an initial set of facts regarding the
location of agency in business ®rms. What sort of analysis we initially
attempt of course depends upon what cause-and-effect questions immedi-
ately concern us. Suppose we are interested in how a ®rm's organization
causes it to address new business, and we ®nd that a ®rm's routine is that
customer orders to purchase goods are sent sequentially to the billing
and shipping departments. That the group agent made up of the shipping
department does not act before the group agent made up of the billing
department locates agency at different points and different times in the
®rm with respect to new orders. If we also learn that departmental sub-
routines require that new business be processed by teams of individuals
in each department, we might further conclude that agency rests with
departments rather than individuals. Alternatively, if in each department
the routine adopted requires that the ®rst available individual take sole
responsibility for new business until it is concluded, we might say that
individuals act as agents, but that their agency is rotating and recurrent,
rather than continuous, where this is a matter of numbers of individuals
in a department, volume of business, time needed per order, etc. Or, a
department might include a routine for trouble-shooting unexpected
problems that is always delegated to the same individual. This sort of
case clearly would contrast with the revolving responsibility type of
routine in which individuals act in the name of their department.
Note that these simple examples treat routines as if they were isolated
from one another. However, routines clearly occupy places in larger
patterns of behavior and/or complex organizations, and how these larger
patterns of behavior and organizations are structured generally plays an
important role in determining which routines are exercised and which
ones go unexercised. Thus a trouble-shooter type agent in one depart-
ment of a business ®rm who has a particular set of duties as a
representative of that department might well ®nd that he or she has prior
 Agent identity in economics 125

duties according to a set of ®rm routines that transcend department

routines. Clearly then, analyzing the interconnections and hierarchical
relationships between routines in this fashion would be a step toward a
more general theory of action for business ®rms and other sorts of
organizations. But it still seems reasonable to suppose that routines
remain the building blocks for an analysis of economic agency, since
whether or not a given routine is exercised does not alter the power of
individuals to initiate sequences of events on the basis of that routine.
That is, if individuals acting in connection with routines are plausibly
described as the source of events, we may comfortably speak of the
action in which organizations engage in a derivative manner, prepared to
provide an analysis of how the organization structures action in terms of
its assignment of responsibilities to individuals and collections of indi-
viduals through their embedding in routines.
The shorthand claim that routines can be plausibly described as event-
generating may be reinforced by attention to what individuals involved
in routines tend to believe individuates particular routines as distinct
forms of activity. Note again, then, that an important dimension of
routines is that the rules and procedures they involve typically possess an
element of obligatoriness. Following a routine means one ought to follow
certain rules and adopt certain procedures in the circumstances in which
the routine applies. Putting aside whether we understand this as substan-
tively or procedurally rational, one way of capturing the normative force
routines possess is to think of routines as depending upon systems of
mutually reinforcing expectations held by the individuals who participate
in them. Routines may then be thought to be forms of conventional
behavior in which agents act interdependently in such a manner that each
recognizes rules and practices in routines have normative force. Thomas
Schelling, David Lewis, and Robert Sudgen have contributed to an
explanation of conventions as spontaneously emerging equilibrium solu-
tions to coordination problems in interdependent decision-making con-
texts. However, we cannot follow them too closely in their game-theory
formulation of conventional behavior if we want to explain agency by
analyzing the social facts routines imply, since their approach assumes
individuals are instrumentally rational or rational in Bayesian terms, and
we have seen that this presupposes that individuals are already character-
izable as independent agents. Since our a posteriori method is meant to
avoid the circularity and question-begging regarding agency inherent in
the neoclassical approach, our seeing the rule-governed routines as
systems of mutually reinforcing expectations must allow that, though
individuals may not always have the status of agents, they may nonethe-
less have mutually reinforcing beliefs about expected behavior.
126 Rationality and homo economicus

What, then, does treating routines as forms of conventional behavior

based on mutually reinforcing expectations contribute to our understand-
ing of agency? In a simple two-individual case, mutual expectations
regarding one individual's entitlement or responsibility to act as an agent
may be described as follows: individual x expects to do A, individual y
expects individual x to do A, individual x expects individual y to expect
individual x to do A, and so on potentially to higher order expectations.
What we see here is that in determining entitlement or responsibility to
act, expectations such as these confer agency status by recording the
judgments of the individuals involved. Since our method treats social
facts as widely recognized phenomena, treating routines as forms of
conventional behavior based on mutually reinforcing expectations im-
plicitly provides us with a criterion by which agents may be located and
distinguished from one another. Abstracting out this location criterion,
we may thus say that in the context of a given routine, an independent
agent is that individual or group of individuals that expects to act as an
agent, is expected by others to act as an agent, believes others have the
relevant expectation, is thought to have this belief by others, and so on.5

4 Addressing the scope of agency

The issue of the scope of economic agency concerns the magnitude and
duration of the powers that distinguishable economic agents may be
represented to realistically exercise as causal agents. Here there are
af®nities to the philosophers' topic of personal identity, since philoso-
phers often imagine situations in which an individual suffers an impair-
ment in his or her power to act, and then ask whether that individual's
identity is sustained. For example, an individual might lose consciousness
for an extended period of time. Does this person remain himself or
herself during this period? Or an individual might be transported into
some unfamiliar setting in which his or her chief skills and views fail to
apply. Must this individual become a new person? In economics,
however, the parallel topic of agent identity necessarily approaches the
magnitude and duration of agent powers from a somewhat different
perspective. Recall that philosophers interested in personal identity are
typically concerned with the fate of some already distinguished indi-
vidual. But economists cannot presuppose who or what an economy's
5 I put aside the question of what sort of regress is involved here. See Mongin and Walliser
(1988) for a discussion of when in®nite regressions are problematic and when not. In
general, not all in®nite regressions need be paradoxical.
 Agent identity in economics 127

agents are if they intend to actually explain agency. Nor for that matter
should economists follow philosophers of personal identity in supposing
that economic agents (when distinguished) are likely candidates for
survival through change in time and/or circumstances. Whereas philoso-
phers hope to demonstrate personal identity is sustained through a
variety of changes, or at least say under what conditions this may be the
case, in economic life organizational and behavioral developments often
erase the space occupied by particular routines together with the agency
relations they imply. Consequently, just as we cannot presuppose that in
economics we are automatically concerned with certain types of agents,
for example, individuals or classes in the methodological individualist
and collectivist traditions, so neither can we assume that certain types of
agents automatically retain the status of being agents across different
spheres of activity and through processes of change.
Explaining the magnitude and duration, or strength and permanence,
of particular forms of economic agency accordingly requires that we
place boundaries on distinguishable agency relations both at a point in
time and also through time. At a point in time, the magnitude or strength
of an agent's powers are a matter of how far-reaching the effects of an
agent's actions are. For example, if in a business ®rm a team of
individuals has responsibility for a set of cross-department activities, say,
with respect to ®rm internal audit procedures, the in¯uence of this team-
agent's actions may in certain circumstances outweigh the actions taken
by departmental agents. Conversely the scope of the latter's activities at
any point in time would be partly de®ned by scope the team-agent's
activities possessed. Regarding an agent's powers through time, that is,
their duration or permanence, we need to be able to say when a given
instance of agency ceases to operate, and when another emerges. Again
using the business ®rm as a type of organization, we might in this
instance look for changes that ended reliance on some routines while
creating the occasion for others. Technological change affecting infor-
mation storage surely would provide examples of this kind. Activities
and routines made obsolete by the appearance of new ones mark off the
boundaries of agency relations through time.
These remarks, however, only provide an initial outline of the dimen-
sions involved in explicating the scope of economic agency relations, since
they ignore the complexity of most real-world cause-and-effect relation-
ships. Following J. S. Mill, contemporary philosophers of economics
(Cartwright 1983, 1989; Lawson 1989; Hausman 1992; MaÈki 1994)6 have

6 I ignore here the differences between these individuals regarding interpretation of ceteris
paribus clauses.
128 Rationality and homo economicus

begun to sort out this complexity by distinguishing (i) between an agent's

having a power to act in principle and that agent's actions possessing
distinguishable effects in the presence or absence of overlapping and/or
countervailing sources of action, and (ii) between an agent's actually
exercising a capacity to act and not exercising that capacity, for whatever
reason. Thus in terms of our point-in-time example above regarding the
magnitude of an agent's powers, it could be found in terms of dimension
(i) that a team-agent responsible for ®rm audit procedures has routine
authority over individual departments, but that individuals within a
number of departments routinely delay providing the relevant data (a
case of countervailing power). Alternatively, rather than engage in
delaying actions, individuals within some departments could be found to
expedite ®rm audit routines (a case of overlapping power). Both of these
cases might further be complicated along dimension (ii) were individuals
in some departments able to engage in delaying actions, but did not do so
(a case of an unexercised countervailing power), or were individuals in
some departments able to engage in expediting actions, but did not do so
(a case of unexercised overlapping power). Parallel cases concerning
agency relationships through time might also be imagined.
These additional characteristics of real-world cause-and-effect rela-
tionships clearly make the analysis of the scope of agency relations more
dif®cult, but attention to them seems only to reinforce the importance of
®rst sorting out simple agency relationships before looking at the
interrelationships between them. Ontologically, that is, investigation of
the basic concepts of agent identity involving location and scope precedes
investigation of their manifestations in complex cause-and-effect phe-
nomena. This claim may seem at odds with the social facts approach
taken here, since it could be argued that ordinary individuals' appraisal
of routine behavior is recorded in terms of the manifestations or observed
effects of agency relations rather than in terms of their originating
sources. Against this, the position taken above is that ordinary indi-
viduals understand routines in terms of responsibilities to act, where
having responsibility is an element of being an agent. Thus ordinary
thinking regularly operates in simple ontological terms even though this
more philosophical language is typically unfamiliar to most people. In
contrast, what ordinary thinking is not particularly good at is moving
from recognition of basic relationships of responsibility and agency to an
analysis of complex agency patterns, both on account of the fact that
ontological matters are rarely enunciated in ordinary thinking, and
because of the inherent dif®culties involved in working out representa-
tionally adequate, ontologically uni®ed accounts of cross-cutting, multi-
dimensional agency interrelationships.
 Agent identity in economics 129

What the analysis here generally calls for, then, is the examination of
structures of routinized behavior in various domains of economic activity
within which it appears possible to distinguish decentralized responsibil-
ities for action on the part of different sorts of economic agents. The
business ®rm as a type of organization naturally lends itself to a variety
of examples, and the method illustrated in the examples above would
equally apply to the analysis of agency in the family, government, and
other distinguishable domains in which routines are commonly noted.
But our emphasis on routines as relatively enduring, and the relative
enduringness of these traditional domains of interest, should not be
thought to imply that the agency relations we may ascertain to lie behind
routines by this method constitute permanent causal elements for
economics as a social science. That is, in addition to the complexity we
encounter in terms of multiple sources of causality, complexity also
enters our analysis on account of the evolution of structures of routinized
behavior. This evolution, it seems fair to say, comes about because
economic agents' actions are both conditioned by the frameworks in
which they operate, and simultaneously condition or bring about change
in those frameworks themselves. Suppose we were to treat economic
evolution as a multi-mechanism process with both selection and adaptive
learning and also as a multi-level process with individuals and groups of
individuals acting as agents (Vromen 1995, p. 213). Then explaining the
scope and location of agency would require our considering the relative
success of different types of routines empowering different types of
agents in different types of environments. That evolution may well
transform the forms which agency assumes, however, only demonstrates
the importance of having a method for explaining its location and scope.

5 Three concluding remarks

(1) The discussion of agent identity here has proceeded as if one might
investigate the location and individuation of agents apart from treatment
of the scope of their powers as causal agents. There is modest justi®cation
for this in that the causal theory of reference adopted above allows us to
communicate about things in the world without agreeing upon theories
about these things. In practice, however, conventional location of
responsibility to act according to systems of mutually reinforcing expec-
tations quickly introduces scope of agency considerations into talk about
who or what an economy's agents are. Being able to say where
responsibility to act lies typically involves being able to say something
130 Rationality and homo economicus

about what that responsibility involves. At the same time, because action
often possesses considerable unintended effects, analysis of the scope of
agency goes beyond what our location criterion involves. Those party to
a convention cannot anticipate all that assignment of responsibility for
action implies, particularly where an agent's actions have transformative
effects upon the structure of routinized behavior itself.
(2) A theme emphasized in the discussion here is that there is an
important difference between the way philosophers approach the
problem of personal identity and the way economists need to approach
the problem of agent identity. My view is that if economists were to
begin to investigate agent identity they would tend to do so along the
lines of the problem of personal identity. The worst perpetrators of this
misconsception are neoclassical economists who generally assume that all
individuals are agents (and that households and ®rms, when acting as
agents, are resolvable into collections of individuals). But traditional
institutionalists and Marxists have made comparable mistakes by sub-
stituting their collective agents into the neoclassical equation. In both
cases, normative reasoning about preferred social subjects displaces
ontological reasoning about causal power and cause-and-effect relation-
ships. Economists, however, need to examine the way the world works if
they are to explain it. Indeed, policy goals regarding the well-being of
different types of agents may only be promoted with an adequate
understanding of the economic process.
(3) This paper began with reference to two methodological traditions
in economics regarding agency: methodological individualism and
methodological collectivism. The argument of this paper should have
made clear that these orientations are both ontologically naãÈve. Their
being so is in part a matter of historical forces alluded to in the remark
above. No one would deny that individuals and classes have been
defended as economic agents, because doing so has suited different social
thinkers' world views. Nonetheless, agency as a domain of analysis is
susceptible to careful examination. That it has not received it seems due
in part to a failure of philosophers and methodologists of economics to
focus on ontological issues. Recently, the topic of causality has begun to
receive some attention (e.g. Hoover 1990). The hope this paper concludes
with is that this interest will be further extended to investigation of the
agents that initiate cause-and-effect sequences.

References
Becker, G. (1965) ``A Theory of the Allocation of Time,'' Economic Journal, 65,
493±517.
 Agent identity in economics 131

(1976) The Economic Approach to Human Behavior, Chicago: University of

Chicago Press.
Cartwright, N. (1983) How the Laws of Physics Lie, Oxford: Clarendon Press.
(1989) Nature's Capacities and their Measurement, Oxford: Clarendon Press.
Davis, J. (1989) ``Axiomatic General Equilibrium Theory and Referentiality,''
Journal of Post Keynesian Economics, 11 (3), 424±38.
(1995). ``Personal Identity and Standard Economic Theory,'' Journal of
Economic Methodology, 2 (1), 35±52.
Frege, G. (1892) ``U È ber Sinn und Bedeutung,'' Zeitscrift fuÈr Philosophie und
philosophische Kritik, 100, 25±50. English translation, ``On Sense and
Reference,'' Philosophical Writings, ed. P. Geach and M. Black, Oxford:
Blackwell, 1952.
Gilbert, M. (1989) On Social Facts, London: Routledge.
Hausman, D. (1992) The Inexact and Separate Science of Economics, Cambridge:
Cambridge University Press.
Hoover, K. (1990) ``The Logic of Causal Inference,'' Economics and Philosophy, 6
(2), 207±34.
Kripke, S. (1971) ``Identity and Necessity,'' in Identity and Individuation, ed.
M. Munitz, New York: New York University Press, 135±64.
Lawson, T. (1989) ``Abstraction, Tendencies and Stylised Facts: A Realist
Approach to Economic Analysis,'' Cambridge Journal of Economics, 13,
55±78.
MaÈki, U. (1992). ``Friedman and Realism,'' Research in the History of Economic
Thought and Methodology, 10, 1±36.
(1994). ``Isolation, Idealization and Truth in Economics,'' in Idealization VI:
Idealization in Economics, ed. B. Hamminga and N. De Marchi, Amsterdam
and Atlanta: Rodolpi.
Mongin, P. and B. Walliser (1988) ``In®nite Regressions in the Optimizing
Theory of Decision,'' in Risk, Decision and Rationality, ed. R. Munier,
Dordrecht: Reidel.
Nelson, R. and S. Winter (1982) An Evolutionary Theory of Economic Change,
Cambridge, MA: Harvard University Press.
Noonan, H. (1989) Personal Identity, London: Routledge.
Par®t, Derek (1984) Reasons and Persons, Oxford: Clarendon Press.
Putnam, H. (1970) ``Is Semantics Possible?'', in Language, Belief, and Metaphy-
sics, ed. H. Kiefer and M. Munitz, Albany: State University of New York
Press, 50±63.
(1973) ``Meaning and Reference,'' Journal of Philosophy, pp. 699±711.
Russell, B. (1905) ``On Denoting,'' Mind, reprinted in Logic and Knowledge,
London: Allen and Unwin, 1958.
Sugden, Robert (1991) ``Rational Choice: A Survey of Contributions from
Economics and Philosophy,'' Economic Journal, 101 (407), 751±85.
Vromen, J. (1995) Economic Evolution. London: Routledge.
8 Chances and choices: notes on
probability and belief in economic
theory

JOCHEN RUNDE1

A recurring point of contention in rational choice theory is the legitimacy

or otherwise of assuming that decision-makers choose as if they were
guided by precise numerical probabilities of the consequences of their
actions. Economists, broadly speaking, fall into two groups on this issue.
One holds that beliefs can only be taken to correspond to point
probabilities in situations that approximate games of chance or where
numerical probabilities can be assigned to events on the basis of a
knowledge of relative frequencies. The other holds that beliefs should be
treated as point probabilities, entering economics via decision theory and
emerging as the parameters of consistent choice. I shall call these the
traditional and the Bayesian view respectively.
An important feature of the Bayesian view is that it makes it possible
to apply rational choice theory even in situations in which there are
no obvious `objective' probabilities. This feature has contributed
enormously to its popularity in economic theory and, until quite recently,
to the steady decline of the distinction between risk and uncertainty
that tends to accompany the traditional view (between situations in
which agents' beliefs about random outcomes are, or are based on, point
probabilities and situations in which they are not). In fact, even now
the dominance of the Bayesian view is such that some of its more
prominent proponents feel justi®ed in declaring the distinction a `sterile
one' (e.g. Hirshleifer and Riley 1992, p. 10). Yet there have always been
some who have continued to insist on the importance of uncertainty as
distinct from risk in economic analysis, particularly in the more
heterodox strands of the discipline. And, over the last ten years or so, a

1 I am grateful to Chuck McCann, Tony Lawson, Uskali MaÈki, members of Gay Meek's
M.Phil seminar on Philosophical Issues in Economics (Cambridge) and three anonymous
referees for comments on earlier versions of this paper.

132
 Chances and choices 133

similar view has begun to emerge in the mainstream literature in

economics and decision theory (Kelsey and Quiggin 1992). The purpose
of this paper is to examine this `dissident' view, focusing on some of the
ontological presuppositions that inform the traditional and the Bayesian
positions.
I begin in the ®rst section with a sketch of the traditional view. The
representatives I have chosen are Knight (1921) and Keynes (1973a),
partly because they are routinely cited in connection with the risk/
uncertainty distinction in economics and partly because they provide
philosophically sophisticated statements of two prominent versions of
the chance-based interpretation of numerical probability. One of my
aims in this section is to show why it is legitimate to speak about
Knightian or Keynesian uncertainty in the same breath,2 despite the
differences in the foundations of their respective theories. The second
section outlines the Bayesian model and attempts to account for its
popularity in modern economics. Finally, the third section assesses
the Bayesian model in its descriptive capacity as a representation of
the economic actor and examines the ontological shift in some
recent generalizations of the Bayesian model on Knightian or Keynesian
lines.

1 Chances

Situations of chance, in what follows, are de®ned as ones in which a

numerical probability may be assigned to some B on the basis of the
number of times it is instantiated in a class of otherwise homogeneous
A's. The central metaphysical presupposition of the chance-based
approach to probability, then, is the homogeneity or sameness of the
members of the classes on which it is based. I shall argue that the two
theories of numerical probability advanced by both Knight and Keynes
are at one on this point. This will allow us to derive a general criterion
for the traditional `Knightian' or `Keynesian' distinction between risk
and uncertainty.

2 For example, Anand (1993), Curley and Yates (1989), Epstein and Wang (1994), Fishburn
(1993), Hodgson (1997), Hogarth (1987), Kelsey and Quiggin (1992).
134 Rationality and homo economicus

Knight

Knight's (1921) Risk, Uncertainty and Pro®t is best remembered for

distinguishing between uncertainties that take the form of `a known
chance' (risks) and `true' uncertainties that do not. But his position
is in fact more subtle than this simple dichotomy might imply. For
as he presents it, there exists a gradation of uncertainties running
through the following three types of `probability situation': a priori
probability, statistical probability and `estimates' (Knight 1921,
pp. 224±5). As there are some problems with Knight's own presentation,
I shall use the following slightly reformulated version of his three
categories:3
1. A priori probability. The ideal case in which numerical probabilities
can be computed on general principles, namely where they are
assigned to equally probable, exhaustive and mutually exclusive
possible outcomes such as the six sides of a perfect die. Given a list of
such possibilities x1, x2,. . .,xn, the probability of any one of them
p(xi) = 1/n. If m of the x's are also y's, then p(y) = m/n.
2. Statistical probability. Situations in which frequencies are derived on
the basis of an `empirical classi®cation of instances [trials]' rather than
from abstract possibilities. The statistical probability r of some
outcome y is simply the proportion or relative frequency of a reference
class of empirically tabulated x's that are also y's. This might be
written px(y) = r.
3. Estimates. Situations in which a priori probabilities cannot be deter-
mined or in which there are not enough trials suf®ciently `like' to form
a reference class against which frequencies can be determined.
The ®rst two of these probability situations fall into the category of
risk, in Knight's terminology, the third into the category of uncertainty:
whereas it is possible to assign numerical probabilities to events on the
basis of a calculation of the `chances' in the ®rst two cases (risk), it is not
possible to do so in the third (uncertainty). The assignment of numerical
probabilities in situations of risk, according to Knight, is always based
on the assumption that `the unknown causes in a case will distribute
themselves according to the law of indifference among the different

3 The discussion of chance-based interpretations of probability in the previous version of

this paper published in the Monist was badly corrupted by the fact that I had originally
failed to notice these problems (now discussed in detail in Runde 1998). The present
version of this section supersedes the original entirely.
 Chances and choices 135

instances [trials]' (p. 219), where this `law' is an expression of the

metaphysical indeterminism of the world (p. 222). In the case of a priori
probability the trials are by hypothesis completely homogeneous except
for `really indeterminate factors' (although the key assumption in the
calculation of a priori probabilities is of course the equiprobability of
outcomes, rather than the homogeneity of trials). In the case of statistical
probability it is generally not possible to ensure that all but these factors
have been eliminated to arrive at a grouping of absolutely homogeneous
trials.4 Even so, where probabilities are assigned on the basis of statistical
frequencies it is necessary to assume that the differences between trials
`not subject to measurement or elimination are in fact indifferent' (p.
221). A judgement of the probability of some event x on the basis of a
knowledge of a statistical frequency, according to Knight, therefore
presupposes that there is no probabilistically relevant difference between
the trial that may produce x on the one hand, and each of the class of
trials that forms the basis on which the statistical frequency was
determined on the other.

Keynes

Keynes's Treatise on Probability is a huge work that touches on issues far

wider than can be summarized here. At its core, however, is the idea that
probability should be analysed as a relation of partial entailment
between the conclusion of an argument h and some set of evidential
premises e. These relations are presented as logical entities, the apprehen-
sion of which warrants some rational degree of belief in h intermediate
between certainty or maximum probability on the one hand (where h is a
logical consequence of e) and impossibility or minimum probability on
the other (where h stands in a contradictory relation with e). Relations of
partial implication are written h1 | e1 (read `h1 relative to e1') and, in
modern terms, the theory may be described as one that builds on binary
comparisons of the form h1/e1 * h2/e2 (the symbol * denotes the
relation `is at least as probable as' and the relations =* and >* are then
de®ned in the normal way for `is as probable as' and `is more probable
than'). It is important to recognize that Keynes's probability relations
form only a partially ordered set, that is, that in many cases it may not be

4 Knight acknowledges that the classi®cation of trials will involve the exercise of judgement:
`it is always possible to form classes if the bars are let down and a loose enough
interpretation of similarity is accepted' (p. 227).
136 Rationality and homo economicus

possible to say that one proposition, relative to the evidence, is more

probable than, less probable than, or as probable as, another (Kyburg
1995).
Keynes's ontology of logical probability relations is probably the most
characteristic and controversial feature of his theory.5 Yet he has
remarkably little to say about the probability relation itself, which he
presents as primitive not analysable in terms of simpler ideas. What he
does do is develop a formal logic of comparative or qualitative prob-
ability, much of which is devoted to conditions under which further
comparisons of probability relations can be derived from comparisons
already given. But he also shows how, under special circumstances,
numerically de®nite probabilities may be derived from qualitative com-
parisons of probability relations. The special circumstances are that the
relevant situation must be such as to allow the legitimate application of
the principle of indifference.
The principle of indifference, that `there must be no known reason for
preferring one of a set of alternatives to any other', provides a criterion
by which alternative hypotheses may be judged as equiprobable. Key-
nes's formalization of this criterion turns on the notion of the relevance
of differences in the evidence bearing on each of a pair of alternatives. As
the logic of relevance can be quite complex I shall mention only Keynes's
`simple' de®nition: e1 is irrelevant to h/e if h/e & e1 =* h/e; e1 is then
relevant to h/e if it is not irrelevant. Keynes uses this de®nition to spell
out the principle of indifference. The basic idea is that the evidence must
be symmetrical with respect to each of the alternatives, that there
must be no relevant evidence bearing on any one alternative that does
not bear equally on all of the others. The procedure Keynes has in
mind thus involves determining what parts of the evidence are relevant
to each alternative hypothesis by a series of judgements of relevance.
Then, if this evidence is of the same form for each of the alternatives, the
principle authorizes a judgement of equiprobability, that is, that hi/e =*
hj/e, for all i, j.
Keynes goes on to elaborate a second requirement for the legitimate
application of the principle of indifference, namely that the relevant
alternatives be `ultimate' or indivisible. This is to ensure that the
instances to which probabilities are to be assigned are not capable of
being split into sub-alternatives of the same form. This condition is
5 The major thrust of Ramsey's (1926) famous critique of the Treatise is to question the
existence of the logical probability relations posited by Keynes. I have argued elsewhere
that Keynes's subsequent concession on this point does not disturb his logic of
comparative probability, which does not depend on any particular theory of partial
beliefs and how they may arise (Runde, 1994a).
 Chances and choices 137

necessary to avoid the classic paradoxes to which naãÈve applications of

the principle of indifference fall prey.6 Then, if the list of equiprobable
and indivisible alternatives h1/e, h2/e, . . . hn/e is exhaustive (and they are
mutually exclusive) the conditions necessary to arrive at numerically
de®nite probabilities within the framework of comparative probability
are met.7 Let H = h1/e & h2/e & . . . & hm/e (0 < m < n). The (Keynesian
version of classical or a priori) probability of H is then p(H/e) = m/n.
Clearly the conditions necessary to assign probabilities in this way are
likely to be restricted largely to games of chance. Like Knight, however,
Keynes also suggests that it may sometimes be possible to assign
numerical probabilities on the basis of relative frequencies. The version
of the frequency theory sketched in chapter 8 of the Treatise, presented
as a special case of his own theory in terms of relations between sets of
propositions, also makes heavy use of judgements of relevance and
indifference. As relative frequencies are truth-frequencies relative to
some reference class, the problem for the frequency theorist is to ®nd the
appropriate reference class. Keynes (1973a, p. 113) de®nes the reference
class as the class of propositions of which everything that is relevant in
the circumstances is known to be true of the proposition to which the
probability is to be assigned. The determination of a numerical prob-
ability of some proposition on the basis of a relative frequency may then
be sketched as follows. Suppose we want to determine the probability of
proposition y knowing that of a class of n x's, m are also y's. The x's
satisfy the requirements of a reference class, on Keynes's theory, if (i) our
relevant knowledge about each member of that class is symmetrical with
our relevant knowledge about the next member, and (ii) our relevant
knowledge about each member of that class is symmetrical with our
relevant knowledge about y. If such a class can be found, according to
Keynes, a numerical probability may be assigned to a proposition on the

6 One famous example concerns the probability of some book being red in the absence of
any evidence to the contrary. A naãÈve application of the principle of indifference would
lead to equal probabilities being assigned to the probability of a book being red and being
not-red. But the same rule could equally well be applied to the hypothesis that the book is
black or that it is blue, leading to the contradictory situation in which there are three
mutually exclusive hypotheses each as probable as not.
7 It is possible to distinguish between two categories of `Keynesian' uncertainty within the
framework of comparative probability. The ®rst consists of the cases in which it is not
possible to determine numerical probabilities but it remains possible to make qualitative
comparisons of relations of partial implication. The second category consists of cases in
which it is not possible even to compare relations of partial implication, either because
they are not comparable, or simply because they do not exist or, if they do, are not
known.
138 Rationality and homo economicus

basis of its truth-frequency within that reference class. In the present case
px(y) = m/n.

Comparison

The two types of theory of probability advanced by Knight and Keynes

are usually classi®ed as opposites. Knightian statistical probability is
often given as a variant of what Hacking (1975) calls aleatory prob-
ability, for example, whereas Keynes's `logical' approach is generally
regarded as a variant of epistemic probability. The difference is that
whereas aleatory probabilities are regarded as a (measurable) property of
the external world, epistemic probabilities are regarded as a property of
the way in which we think about the external world. This dichotomy is
re¯ected in the way that Keynes writes of the probabilities of proposi-
tions (about outcomes or events) rather than of outcomes or events
themselves, and in the different ways in which Knight and Keynes think
of the concept of indifference. Whereas Knight's references to indiffer-
ence concern gaps in the causal determination of reality, Keynes's
concern gaps in our knowledge of the causal determination of reality.8
Yet as will already have become clear, Knight and Keynes have
remarkably similar views about the situations in which it is possible to
determine numerically de®nite probabilities. Both of them tend to think
of numerical probabilities as deriving from chance-based setups and both
of their accounts therefore presuppose the existence of classes of equi-
probable, exhaustive and mutually exclusive outcomes or alternatives (in
the case of a priori probability) or reference classes of more or less
homogeneous `instances' or trials on the basis of which statistical
frequencies can be determined. There are also similarities at the episte-
mological level, not least in that their claims about the `objectivity' of
8 For example, Knight writes: `We do not merely feel that we know no reason why the coin
shall fall heads or tails; we know in a positive sense that there is no reason, and only under
this condition do we make the probability judgement with any con®dence. And
furthermore, as already argued, it appears that only on condition that there is no reason
would the results of experience con®rm the judgement, as they do. The entire science of
probability in the mathematical sense is based on the dogmatic assumption that the
ultimate alternatives are really equally probable, which seems to the writer to mean real
indeterminateness' (p. 222). Keynes, in contrast, is concerned with the absence of
knowledge of reasons for regarding one alternative as more probable than another.
Whereas Knight is quite explicit about his belief in the real indeterminacy of the universe,
Keynes is more circumspect, leaning towards a determinist view, but ultimately declaring
agnosticism on the matter (Keynes 1973a, pp. 311±38).
 Chances and choices 139

probabilities (or judgements of probability) are far weaker than is

commonly supposed. We have seen that Knight is fully alive to the fact
that judgements regarding the homogeneity of instances are agent-
relative, and Keynes describes his formulation of the principle of
indifference as `displaying its necessary dependence upon judgements of
relevance and so bringing out the hidden element of direct judgement or
intuition, which it has always involved' (Keynes 1973a, p. 69).9 The main
difference here is that Keynes attempts to formalize judgements of
`likeness' by giving explicit epistemic criteria for judgements of (ir)rele-
vance and, accordingly, indifference.
Modern proponents of the traditional chance-based interpretation of
probability in economics tend to be somewhat less explicit on the issues
considered in this section than are Knight and Keynes. But the ontolo-
gical presuppositions are much the same, namely that there exist identi®-
able classes of more or less homogeneous trials that are in reasonable
agreement with the mathematical concept of independently repeated
events (e.g. Bewley 1986; Davidson 1988; Hicks 1979; Lawson 1988;
Lucas 1981).10 Situations of risk, and leaving aside the special case in
which it is possible to determine numerical probabilities on an a priori
basis, are ones in which there exist identi®able classes of more or less
homogeneous trials on the basis of which frequencies can be determined.
Situations of uncertainty are ones in which such classes do not exist or, if
they do, cannot be identi®ed.

9 Keynes in fact sometimes goes out of his way to ensure that his theory is not interpreted
as placing too much emphasis on the subjectivity of individual judgements (see 1973a,
p. 56). His famous 1937 QJE passage on uncertainty, moreover, suggests that, like
Knight, he too sometimes thinks in terms of a gradation of uncertainties depending on
the degree of homogeneity of the members of the class of `instances' relative to which
frequencies are determined: `By ``uncertain'' knowledge, let me explain. I do not mean
merely to distinguish what is known for certain from what is only probable. The game of
roulette is not subject, in this sense, to uncertainty; nor is the prospect of a Victory bond
being drawn. Or again, the expectation of life is only slightly uncertain. Even the weather
is only moderately uncertain. The sense in which I am using the term is that in which the
prospect of a European war is uncertain, or the price of copper and rate of interest
twenty years hence, or the obsolescence of a new invention, or the position of private
wealth owners in the social system in 1970. About these matters there is no scienti®c
basis to form any calculable probability whatever' (Keynes 1973b, pp. 113±14).
10 Commenting on the rational expectations hypothesis, for example, Lucas (1981,
pp. 223±4) remarks that it will not `be applicable in situations in which one cannot guess
which, if any, observable frequencies are relevant: situations which Knight called
``uncertainty.'' It will most likely be useful in situations in which the probabilities of
interest concern a fairly well de®ned recurrent event, situations of ``risk'' in Knight's
terminology.'
140 Rationality and homo economicus

2 Choices

On the Bayesian view the probability of a proposition or event is

interpreted simply as the strength of the actor's personal belief in that
proposition or event.11 It is often argued that the strength of personal
beliefs is a causal property of those beliefs, re¯ected in the extent to
which actors are prepared to act on them (Ramsey 1926, p. 71), and
which may be measured by examining choice behaviour in betting
situations. Suppose we wish to ®nd the actor's subjective probability of
some proposition h, p(h). Bayesians propose that this can be done by
®nding the lowest value P* that the actor would be prepared to exchange
for a gamble that pays S if h is true and nothing if h is false.12 Two
assumptions are crucial here. First, the value P* is assumed to exist.
Second, and as the second formulation below suggests, P* must leave the
actor exactly indifferent between accepting the gamble and refusing it.
Given that the actor is an expected utility maximizer, his or her
declaration of indifference between P* on the one hand and S if h, 0 if : h
on the other, is then often written as:
P* = p(h)S + (17p(h))(0)
or,
p(h)(S7P*) + (17p(h))(7P*) = 0
This gives a real-valued probability p(h) = P*/S.
It is not hard to explain the appeal of this approach in economic
theory. First, it is extremely simple and based on what is widely
regarded as no more than a systematic and analytically tractable
expression of the common-sense idea that people's actions issue from
their desires and their beliefs (in conjunction with the old and equally
venerable idea that actions should be evaluated in terms of the expected
11 Ramsey (1926), de Finetti (1964), and Savage (1954) (readers interested in the differences
between the different versions of the Bayesian model should consult Fishburn 1981 or
Kreps 1988).
12 P and S are assumed to be (numerically de®nite) utility values rather than amounts of
money. I have adopted the (Ramseyan) approach of taking utility as primitive to keep
things as simple as possible. Savage (1954), the author economists most frequently
appeal to in justifying their use of subjective probabilities, de®nes probabilities in a more
general way in terms of preferences over `acts' (the concept of a consequence is primitive
in Savage's theory and every mapping from the set of possible `states of the world' into
the set of possible consequences constitutes an act). Ramsey's and Savage's systems are
nevertheless quite similar in their implications, and the noted difference will not affect
the argument at the level of abstraction pursued here.
 Chances and choices 141

utility of their consequences). Second, since it is possible to bet on just

about any decidable proposition or event, the Bayesian interpretation of
probability brings a large class of cases within the compass of the
probability calculus that could not be on chance-based interpretations.
This then makes it possible to argue that the traditional distinction
between risk and uncertainty becomes redundant. And third, the theory
appears quite modest in ontology insofar as it avoids making claims
about observer-independent probabilities (hence de Finetti's famous
remark that `Probability does not exist'). In particular, it does not
presuppose the existence of classes of repeated events and is often
praised for making it possible to model beliefs without getting immersed
in deep questions about the connection between actors' beliefs and
`reality' (Lucas 1981, p. 223).
But the feature of the Bayesian interpretation of probability that might
appeal most to many economists is the contention that it is a condition of
rationality that an actor's subjective degrees of belief be related so as to
conform to the axioms of the probability calculus. This contention is
supported by the so-called Dutch Book argument, that an actor whose
beliefs do not meet this requirement could be induced to accept a
sequence of bets that would result in a sure loss whatever the outcome of
the events bet upon. It is at this point, however, that the theory's
modesty about mind-independent probabilities begins to be offset by the
rather stringent prerequisites of its own. For in order for the Dutch Book
argument to go through it is necessary, not only that actors must evaluate
options in terms of the mathematical expectation of the value of their
consequences and that they must always be prepared to act (bet) on their
beliefs, but also that they be prepared to bet against the hypothesis in
question at the least odds that characterize their degrees of belief
(Kyburg 1978). In the above example this would mean that the actor
would also have to agree to accept a payment of P* in exchange for a
gamble that involves a loss of S if h is true and nothing otherwise, such
that P*/S = 7P*/ 7S = p(h).13 I shall assume that this condition is met
for the moment and return to when it is not later on.
An example may be useful to illustrate how a Dutch Book may
occur.14 Suppose that there are two mutually exclusive hypotheses h and
k, and that the actor is prepared to exchange up to the following values
to take the following gambles:

13 This condition is implied by the expression P* = p(h)S + (17p(h))(0), but not by a

declaration of indifference between P* on the one hand and S if h, 0 if h on the other (see
below).
14 Adapted from Schick (1986).
142 Rationality and homo economicus

Fig. 1

(1) p(h)S to receive S if h, 0 if : h

(2) p(k)S to receive S if k, 0 if : k
and to demand at least
(3) p(h or k)S in exchange for having to give up S if (h or k), 0 if
: (h or k).
Suppose also that someone accepts these gambles at the stated odds
and that the probabilities implicit in (1), (2) and (3) above are such that
p(h) + p(k) 7 p(h or k) < 0. These `probabilities' violate the probability
axioms which require that p(h) + p(k) = p(h or k) for mutually exclusive
events. It can be shown that the actor has entered into a Dutch Book.
To see this consider the possible outcomes. There are three: h is true, k
is true, or neither, leaving the three scenarios in Fig. 1. The actor's
degrees of belief are `incoherent' because he or she will incur a sure loss
whichever of the three possible eventualities obtains. For example, if h
turns out true the actor will gain S + p(h or k)S and pay or lose
p(h)S + p(k)S + S. Since p(h)S + p(k)S < p(h or k)S, the net gain will be
less than zero. The same will be the case if h is true or (h or k) is true.
Such losses will be avoided only if probabilities are assigned such that
p(h) + p(k) = p(h or k). Economists count the idea that actors whose
beliefs violate the axioms of probability are open to economically self-
destructive behaviour as a strong argument in favour of the Bayesian
view. The argument is analogous to the money pump argument used to
justify transitive preferences.
I have emphasized the Dutch Book argument to highlight some of
what the claim that actors' choices reveal their subjective probabilities
entails. In particular, the existence of numerically de®nite subjective
probabilities seems to presuppose both that actors have real-valued
degrees of belief at the back of their minds and that they are prepared to
bet for and against every uncertain hypothesis at the same least odds that
correspond to these degrees of belief. This is equivalent, in terms of the
 Chances and choices 143

Bayesian framework, to their being able to identify sure consequences P*

that are exactly equal in value to mathematical expectation of a pair or
random consequences (S and 0), where the value of the mathematical
expectation of the random consequences depends on beliefs that the
whole procedure is aimed at quantifying.

3 Choices and chances

Introspection and experience suggest that these requirements, that actors

have real-valued degrees of belief and that they are always willing to take
bets at all the odds that correspond to these degrees of belief, are seldom
met. This may be no objection to the Bayesian view in its normative
capacity, when it is used as a means of checking and improving the
consistency of one's decisions. Indeed, it may even be a possible
justi®cation for it.15 But this is not the way it is typically used in
economic theory, where it tends to be thought of as a representation of
behaviour that is approximated in the behaviour that economists study
(e.g. Kreps 1988, p. 6). And in this descriptive capacity, the generality the
theory achieves in virtue of its ontological modesty about `objective'
probabilities is only achieved at the price of having to live with rather
immodest metaphysical claims about the extent to which actors have, are
in touch with, and are prepared to act on, what are assumed to be
quanti®able values and beliefs. Of course, I am taking a very literal
reading of the theory here, and the noted immodesty may be no
particular cause for embarrassment on instrumentalist or other non-
realist interpretations of economic theory.16 Nevertheless, and notwith-
standing the argument that it is suf®cient justi®cation for the theory that
only those who behave `as-if ' on the basis of economically sound
reasoning will survive in a competitive environment (Alchian 1950;
Friedman 1953), many economists do seem to believe that any predictive
power that formal decision theory may have has to come from a real
tendency human actors have to deliberate and act in accordance with its
prescriptions (e.g. Sugden 1991).
That said, there are two well-known challenges to the Bayesian
commitment to real-valued beliefs which, as they involve what seem to be
systematic violations of the theory, are no less problematic on its
15 Of course there are also those who reject the normative interpretation of the Bayesian
model (e.g. Anand 1993).
16 Indeed, as Uskali MaÈki (private correspondence) notes, ontological immodesty may
encourage instrumentalist interpretations.
144 Rationality and homo economicus

Fig. 2
instrumentalist than on its realist interpretations. The ®rst of these
challenges is that if people were in fact rational Bayesians, and if it were
true they often have very different beliefs, then we should observe
considerably more betting behaviour than we in fact do. For it is an
implication of the Bayesian view that rational actors whose beliefs differ
will always enter into bets with one another. This can shown diagramati-
cally. Suppose there are two individuals 1 and 2 who are uncertain about
two mutually exclusive events h and k. Each of them is endowed with an
initial amount of money c that is independent of which of the two events
obtains. This situation may be depicted in terms of a square Edgeworth
box, where the amount of money received in event h is measured on the
horizontal axis and the amount of money received in event k is measured
on the vertical axis. Let E be the initial endowment and suppose that the
two individuals are risk averse, that is, that their indifference curves are
convex to the origin.17 Then consider the following scenarios in Fig. 2.
The absolute slope of the indifference curves is equal to p(h)/p(k) at all
points on the 458 line.18 In the ®rst ®gure the two actors share the same
subjective probabilities p1(h)/p1(k) = p2(h)/p2(k). Indifference curve tan-
17 As expected utility maximizers, the actors' objective functions have the form
U = p(h)u(ch) + p(k)u(ck), where p(h) and p(k) are the probabilities of h and k and u(ch)
and u(ck) are the utilities of the amounts of money received in events h and k respectively.
The objective function describes a locus of combinations of ch and ck which are equally
preferred, and there is one such `indifference curve' for every value of U (with expected
utility increasing away from the origin). Indifference curves that are convex to the origin
are termed risk averse because they re¯ect a preference for less rather more variation in
the amounts of money contingent on each event.
18 The slope of the indifference curves (the marginal rate of substitution of money
contingent on each event) is obtained by total differentiation:
dU = p(h)u'(ch)dch + p(k)u'(ck)dck = 0. Rearranged, this expression gives:
dck p…h†u0 …ch †
ˆÿ
dch p…k†u0 …ck †
As ch = ck on the 458 line, it follows that the slope of the indifference curves on that line
must be equal to p(h)/p(k).
 Chances and choices 145

Fig. 3

gency (equilibrium) accordingly occurs at E on the diagonal and they will

not take opposite sides of a bet. In the second ®gure, however, the actors'
probabilities differ (p1(h)/p1(k) > p2(h)/p2(k)) and their indifference curves
are not tangential at E on the diagonal. Here the actors will increase their
utility by betting, that is, by trading in contingent claims to a point off the
diagonal. The question which proponents of the Bayesian view must
answer, then, is why there seems to be so much less betting behaviour
than there seem to be differences in opinion. Why do people tend to shun
bets even though they disagree?
The second challenge to the Bayesian interpretation of probability is
provided by the well-known counter-examples given by Ellsberg (1961).
One of these involves two urns, Urn I known to contain 100 balls each of
which may be either black or red, and Urn II known to contain exactly
50 black balls and 50 red balls. The actor wins a prize of $100 in the
event of correctly predicting the colour of the next ball drawn. I shall
write RI for a bet on a red ball from Urn I, BII for a bet on a black ball
from Urn II, etc. Most people profess indifference between RI and BI
and RII and BII respectively, implying that p(RI) = p(BI) = p(RII) = p(BII).
When offered a choice between RI and RII (BI and BII), however, most
display a preference for betting on Urn II. The trouble is that this
implies that p(R1I) > p(RI) and P(BII) > p(BI), contradicting the previous
choices.
A second example involves an urn known to contain 30 red balls and
60 balls, each of which may be either black or yellow. The actor is given a
choice between options I and II, and then between options III and IV
below, the prize depending on the colour drawn, as in Fig. 3.
It turns out that most people prefer I over II and IV over III. Again
there is a contradiction as these preferences imply that p(R) > p(B) and
p(B) > p(R) respectively. The conclusion, in both this and the previous
example, is that there is no unique additive probability distribution
146 Rationality and homo economicus

consistent with the observed pattern of choices maximizing the expected

value of some utility function.
The absence of widespread betting and the aversion to `vague'
probabilities illustrated by the Ellsberg examples suggest that most
people do not or do not act `as if ' they assign point probabilities to the
outcome of all events. Proponents of the Bayesian view tend to respond
in two ways. One is to admit that there may be plausible psychological
reasons for the noted behaviour, but that these ultimately constitute
failures of rationality. In other words, the strategy is to fall back on the
normative justi®cation of the theory. For example, Hirshleifer and Riley
(1992, p. 34) write:
The dissident literature claims that the discrepancies revealed by these results [such
as the Ellsberg experiments] refute the economist's standard assumption of
rationality, or at least the expected utility hypothesis as a speci®c implication of
that assumption. We reject this interpretation. A much more parsimonious
explanation, in our opinion, is that this evidence merely illustrates certain
limitations of the human mind as a computer. It is possible to fool the brain by the
way a question is posed, just as optical illusions may be arranged to fool the eye.
Discovering and classifying such mental illusions are fruitful activities for psy-
chologists, but these paradoxes are of relatively little signi®cance for economists.
The second response is to interpret a reluctance to bet and the Ellsberg
choices as rational responses to the possibility of asymmetric infor-
mation, that is, that actors may shun bets or discount `vague' prob-
abilities because of a suspicion that the person offering the bet knows
something they do not (Hirshleifer and Riley 1992, p. 36, n. 12; Morris
1993). Clearly this may be a possibility in some cases, although, as
Morris (1993, p. 27) admits, decision-makers tend to display an aversion
to `vague' probabilities even when strategic considerations are ruled out.
An alternative possibility is to accept the consequences of the apparent
fact that the central prediction of the Bayesian model in its descriptive
capacity, that people's choices are or are `as if ' they are informed by real-
valued subjective probabilities, is, in general, false. Before moving on to
this possibility, however, I should like to return brie¯y to Knight (1921,
p. 219) and Keynes (1973a, p. 82), who would both recommend assigning
equal probabilities to the contents of the ®rst of the `ambiguous' urns
described above. But this should not be construed as an argument in
favour of the Bayesian view. According to Keynes's decision theory it is
rational to prefer to be guided by probabilities determined on the basis of
greater evidential `weight', the amount or completeness, in some sense, of
the relevant evidence on which a judgement of probability is based. The
evidential weight in favour of assigning probability of 1/2 each to red and
black is greater in the case of urn I than urn II, for example, thus
 Chances and choices 147

providing a rationale for the observed choices. Keynes (1973c, p. 148)

later suggests a link between weight and con®dence, distinguishing
between the `best estimates we can make of probabilities and the
con®dence which we make them' (p. 240). Knight (1921, p. 227) makes a
similar point, arguing that the action that issues from an opinion
depends as much on the con®dence in, as on the favourableness of, that
opinion. The distinction between a judgement of probability and the
con®dence with which it is made has no place in the world of a
committed Bayesian because it drives a wedge into the link between
choice and degrees of belief on which it is founded.19

Choice under uncertainty and economic theory

The foregoing arguments should not be taken to suggest that our choices
are not informed by the strength of our beliefs, merely that our beliefs
tend to be much more vague and partial than the Bayesian model
implies. We often seem to be able to at least grade possible choice-
outcomes in terms of probability, if only roughly, and these gradations
often seem to be useful in guiding our choices. A Bayesian might express
this in the following terms. Suppose the actor is not prepared to commit
to de®nite values of P* and P* such that p(h) = P*/S = 7P*/ 7S. He or
she may nevertheless be prepared to accept a bet on h at odds suf®ciently
favourable and to accept a bet against h at odds suf®ciently favourable
(shunning bets either way at any odds in between). In other words, the
actor may be prepared to give up any value below P* in order to receive
S if h is true (and nothing otherwise) and be prepared to accept any value
above P* in return for having to give up S if h is true (and nothing
otherwise), where P* < P*. Beliefs would thus take the form of intervals,

19 As Knight emphasizes this distinction, I would resist recent interpretations that suggest
that he `implicitly accepted the modern view that by modelling individuals as able to
choose consistently among unknown outcomes, we in effect represent them as always
having subjective probabilities' (LeRoy and Singell 1987, p. 398; see also Langlois and
Cosgel 1993, p. 460; Lawson 1998, p. 51). It is true that Bayesians sometimes argue that
they can accommodate this distinction (e.g. Hirshleifer and Riley 1992, pp. 10±11,
175±7), the idea being that actors' con®dence in their probabilities is re¯ected in the
incentive they have to postpone choice until they can update their `vague' (but
numerically de®nite) priors in the light of new information. The problem, however, is
that this proposal undermines the Bayesian account of where priors are supposed to
come from. If actors' probabilities are supposed to be re¯ected in their propensity to
choose and act on them, how then are priors to be established when they prefer not to do
so?
148 Rationality and homo economicus

Fig. 4

the lower and upper bounds of which would be given by p*(h) = P*/S and
p*(h) = 7P*/ 7S respectively. These intervals may overlap, giving rise
to situations that might look like those in Fig. 4.
As B3 dominates B2 it is possible to argue that B3 is unambiguously
more probable than B2. But it is not possible to arrive at an unambiguous
ranking of B2 and B1 in terms of probability, or, for that matter, of B3
and B1 (the possibility of non-comparability that Keynes is so insistent
on). The above schema provides a possible expression of the situation in
which people are reluctant to bet both ways at the same odds on some
event, is immune to Dutch Books (see Kelsey 1994, pp. 436±7) and
provides a way of thinking about the Ellsberg examples. In the case of
the urn containing 30 red balls and 60 balls in an indeterminate
proportion of black and yellow, for instance, the statistical probability of
red is 1/3 and the probability of black (yellow) is anywhere in the interval
[0, 2/3]. On this `multiple prior' approach to the problem, no colour is
unambiguously more or less probable than, or as probable as, any of the
others, and it is hard to see how those who resist committing themselves
to a unique probability distribution, as the Bayesian view requires,
should be regarded as irrational. It follows, on this line of reasoning, that
no option unambiguously dominates (or is indifferent to) any of the
others in expected utility terms. Determinacy can only be ensured by
introducing additional decision rules, the most popular of which seems to
be a maximin criterion.20
But what if indeterminacy were retained? This possibility, almost
20 See GaÈrdenfors and Sahlin (1982), Gilboa and Schmeidler (1989, 1993), Kyburg (1990,
pp. 237±8), Levi (1986, pp. 128±40) and, for some recent applications and discussions of
the multiple prior approach in economic theory, Bewley (1986), Epstein and Wang
(1994), Kelsey (1993, 1994), Kelsey and Quiggin (1992), and Gilboa and Schmeidler
(1993). An alternative approach to `Knightian' uncertainty replaces the Bayesian prior
with a non-additive measure or capacity (Gilboa 1987; Schmeidler 1989). The underlying
idea here is that people discount `ambiguous' probabilities: by assigning probabilities of
1/3 to red and 1/4 each to black and yellow in the second of the Ellsberg examples, for
instance, the paradox is resolved. But as this approach remains committed to real-valued
degrees of belief (although these will no longer conform to the standard probability
axioms), it is to this extent open to the same objections that I have been raising against
 Chances and choices 149

SQ A3
z‚‚‚‚‚‚‚‚‚‚‚}|‚‚‚‚‚‚‚‚‚‚‚{
j expected utility
|‚‚‚‚‚‚‚‚{z‚‚‚‚‚‚‚‚} |‚‚‚‚‚‚‚‚{z‚‚‚‚‚‚‚‚}
A1 A2
Fig. 5.

unheard of in economic theory, is explored in an imaginative paper by

Bewley (1986). Bewley generalizes the Bayesian model by dropping
completeness and introduces a privileged `status quo' (SQ) position
amongst the set of available options.21 Beliefs are represented as convex
sets of classical probability distributions, generating intervals of prob-
abilities similar to those depicted above. The decision rule is that an
option is preferred to another if and only if it yields a higher expected
utility with respect to every distribution in that interval. (If beliefs
correspond to intervals, of course, then so will the expected utility ± the
`non-numerical mathematical expectation' (Keynes 1973a, p. 344) ±
associated with each option.) The `status quo' is retained unless there is
an option which dominates it in the sense just described. To illustrate,
consider the expected utility intervals associated with three possible
actions A1, A2 and A3 (Fig. 5). SQ and A2 (A2 and A3) are not
comparable, but SQ dominates A1, and A2 and A3 both dominate A1. A3
will be the preferred option as it is the only one that unambiguously
dominates SQ (note that the fact that A3 dominates the status quo
whereas A2 does not restores comparability between A2 and A3).
Now I would not want to go to great lengths to defend Bewley's
particular formalism. His decision rule is extremely conservative, for
example, and the implicit claim that people have sets of probability
distributions at the back of their minds, while perhaps reasonable in
cases such as Ellsberg's urns, seems no less dubious as a general claim
than claiming that their beliefs are point-valued. Bewley's approach, and
the multiple prior approach more generally, nevertheless marks two
signi®cant shifts from the ontology of orthodox Bayesianism. In the ®rst
place, it drops the metaphysics of real-valued degrees of belief and
thereby rehabilitates the distinction between risk and uncertainty.
Second, it goes some distance towards loosening the connection between

the Bayesian model. For some recent economic applications, see Dow and Werlang
(1992a, 1992b, 1994).
21 Completeness is the requirement that for every pair of options x and y, the decision-
maker either prefers x to y, prefers y to x, or is indifferent between them. Completeness
rules out non-comparability of options.
150 Rationality and homo economicus

degrees of belief and dispositions to act on which the Bayesian approach

is founded. This makes it possible to express the situation in which
people respond to uncertainty by refraining from making a choice and
taking an action (in Bewley's framework an option is not taken unless
acceptance, which always involves a departure from one's existing
circumstances, is preferred to rejection). Not only does all of this seem to
accord with what we know about ourselves, it also appears to be
consistent with a range of economic phenomena that seem resistant to
standard Bayesian analysis: the tendency for bid (buyer) prices for
insurance against uncertain contingencies to be systematically less than
asking (selling) prices, even if the insured event results in losses to the
bidder and not to seller; the adoption of conventions in the face of
uncertainty, such as the practice of `going on' as before except in so far
as there are de®nite reasons to expect a change (Keynes 1973c); rule-
following, habits and behavioural routines (Cyert and March 1992;
Hodgson 1997; Vanberg 1994); the incompleteness, rigidity and long-
term nature of wage contracts (Mukerji 1998; Williamson 1985); and that
more liquid assets are often preferred to less liquid assets that dominate
them in return (Runde 1994b).

4 Conclusion

Proponents of what I have called the traditional view tend to resist

assuming that actors are guided by sharp numerical probabilities save for
the situations in which they have knowledge of (the relevant) chances.
The ontological basis for the traditional `Knightian' or `Keynesian'
distinction between risk and uncertainty, on this view, turns on the
absence or otherwise of classes of equiprobable (exclusive and exhaus-
tive) outcomes in the case of a priori probabilities, or classes of
homogeneous trials or suitably de®ned reference classes in the case
statistical frequencies. Situations of uncertainty arise where such classes
do not exist or, at least, cannot be identi®ed.
The `betting' interpretation of probability associated with the Bayesian
view provides an ingenious means of transforming decision-making
under uncertainty, in the traditional sense, to decision-making under
risk. For this reason, and because denying the existence of subjective
probabilities is widely perceived as coming uncomfortably close to
denying that actors are able to make consistent choices over gambles, it
continues to serve as the dominant theory of probability and belief in
modern economic theory. Yet the generality and simplicity the Bayesian
 Chances and choices 151

approach achieves in virtue of avoiding claims about the existence of

`objective' probabilities (or probability relations) should not be allowed
to distract from the strength of its own ontological presuppositions. I
have focused on two of these, that actors have real-valued degrees of
belief and that they are always willing to take bets at all odds that
correspond to these degrees of belief. These presuppositions are highly
implausible on descriptive grounds and, I have suggested, overly de-
manding requirements even on the theory's normative interpretations.
Economic choices are generally not informed by a knowledge of chances
or their Bayesian surrogates. The epistemological concerns that moti-
vated so much of the economic theorizing of Knight and of Keynes are
as relevant as they ever were.

References
Alchian, A. A. (1950) `Uncertainty, Evolution and Economic Theory', Journal of
Political Economy, 58, 211±21.
Anand, P. (1993) Foundations of Rational Choice Under Risk, Oxford University
Press.
Bewley, T. (1986) `Knightian Decision Theory: Part 1', Cowles Foundation
Discussion Paper, no. 807.
Curley, S. P. and J. F. Yates (1989) `An Empirical Evaluation of Descriptive
Models of Ambiguity Reactions in Choice Situations', Journal of Mathema-
tical Psychology, 33, 397±427.
Cyert, R. M. and J. G. March (1992) A Behavioural Theory of the Firm, 2nd
edition, Oxford: Blackwell.
Davidson, P. (1988) `A Technical De®nition of Uncertainty and the Long-Run
Non-neutrality of Money', Cambridge Journal of Economics, 12, 329±37.
de Finetti, B. (1964) `Foresight: Its Logical Laws, Its Subjective Sources', in
H. E. Kyburg and H. E. Smokler (eds.), Studies in Subjective Probability,
New York: John Wiley, 93±158.
Dow, J. and S. R. Werlang (1992a) `Excess Volatility of Stock Prices and
Knightian Uncertainty', European Economic Review, 36, 631±8.
(1992b) `Uncertainty Aversion, Risk Aversion and the Optimal Choice of
Portfolio', Econometrica, 60, 197±204.
(1994) `Nash Equilibrium under Knightian Uncertainty: Breaking Down Back-
ward Induction', Journal of Economic Theory, 64, 305±24.
Ellsberg, D. (1961) `Risk, Ambiguity and the Savage Axioms', Quarterly Journal
of Economics, 75, 643±69.
Epstein, L. G. and T. Wang (1994) `Intertemporal Asset Pricing Under Knightian
Uncertainty', Econometrica, 62, 283±322.
Fishburn, P. C. (1981) `Subjective Expected Utility: A Review of Normative
Theories', Theory and Decision, 13, 139±99.
152 Rationality and homo economicus

(1993) `The Axioms and Algebra of Ambiguity', Theory and Decision, 34, 119±37.
Friedman, M. (1953) `The Methodology of Positive Economics', in Essays in
Positive Economics, Chicago: Chicago University Press, 3±43.
GaÈrdenfors, P. and N. E. Sahlin (1982) `Unreliable Probabilities, Risk Taking
and Decision Making', Synthese, 53, 361±86.
Gilboa, I. (1987) `Expected Utility Theory with Purely Subjective Non-Additive
Probabilities', Journal of Mathematical Economics, 16, 65±88.
Gilboa, I. and D. Schmeidler (1989) `Maximin Expected Utility with a Non-
Unique Prior', Journal of Mathematical Economics, 18, 141±53.
(1989) `Updating Ambiguous Beliefs', Journal of Economic Theory, 59, 33±49.
Hacking, I. (1975) The Emergence of Probability, Cambridge: Cambridge Uni-
versity Press.
Hicks, J. R. (1979) Causality in Economics, Oxford: Basil Blackwell.
Hirshleifer, J. and J. G. Riley (1992) The Analytics of Uncertainty and Infor-
mation, Cambridge: Cambridge University Press.
Hodgson, G. (1997) `The Ubiquity of Habits and Rules', Cambridge Journal of
Economics, 21, 663±84.
Hogarth, R. M. (1987) `Decision Making Under Ambiguity', in R. M. Hogarth
and M. W. Reder (eds.), Rational Choice: The Contrast Between Economics
and Psychology, Chicago: Chicago University Press, 41±67.
Kelsey, D. (1993) `Choice Under Partial Uncertainty', International Economic
Review, 34, 297±308.
(1994) `Maxmin Expected Utility and Weight of Evidence', Oxford Economic
Papers, 46, 425±44.
Kelsey, D. and J. Quiggin (1992) `Theories of Choice Under Ignorance and
Uncertainty', Journal of Economic Surveys, 6, 133±53.
Keynes, J. M. (1973a) A Treatise on Probability. The Collected Writings of John
Maynard Keynes, vol. 8, London: Macmillan.
(1973b) `The General Theory and After: Part II Defence and Development',
The Collected Writings of John Maynard Keynes, vol. 14, London: Mac-
millan.
(1973c) `The General Theory', The Collected Writings of John Maynard
Keynes, vol. 7. London: Macmillan.
Knight, F. H. (1921) Risk, Uncertainty and Pro®t, Chicago: University of
Chicago Press.
Kreps, D. M. (1988) Notes on the Theory of Choice, Boulder: Westview.
Kyburg, H. E. (1978) `Subjective Probability: Criticisms, Re¯ections, and Prob-
lems', Journal of Philosophical Logic, 7, 157±80.
(1990) Science and Reason, New York: Oxford University Press.
(1995) `Keynes as a Philosopher', History of Political Economy, 27, 7±32.
Langlois, R. N. and M. M. Cosgel (1993) `Frank Knight on Risk, Uncertainty
and the Firm: A New Interpretation', Economic Enquiry, 31, 456±65.
Lawson, T. (1988) `Probability and Uncertainty in Economic Analysis', Journal
of Post Keynesian Economics, 11, 38±65.
 Chances and choices 153

LeRoy, S. F. and L. D. Singell (1987) `Knight on Risk and Uncertainty', Journal

of Political Economy, 95, 394±406.
Levi, I. (1986) `The Paradoxes of Allais and Ellsberg', Economics and Philosophy,
2, 23±53.
Lucas, R. E. (1981) Studies in Business Cycle Theory, Cambridge, MA: MIT
Press.
Morris, S. (1993) `Risk, Uncertainty, and Hidden Information', CARESS
Working Paper, 93±10.
Mukerji, S. (1998) `Ambiguity Aversion and Incompleteness of Contractual
Form', American Economic Review, 88, 1207±31.
Ramsey, F. P. (1926) `Truth and Probability', in D. H. Mellor (1978) Founda-
tions: Essays in Philosophy, Logic, Mathematics and Economics, London:
Routledge & Kegan Paul, 60±100.
Runde, J. H. (1994a) `Keynes After Ramsey: In Defence of A Treatise on
Probability', Studies in History and Philosophy of Science, 25, 97±121.
(1994b) `Keynesian Uncertainty and Liquidity Preference', Cambridge Journal
of Economics, 18, 129±44.
(1998) `Clarifying Frank Knight's Discussion of the Meaning of Risk and
Uncertainty', Cambridge Journal of Economics, 22, 539±46.
Savage, L. J. (1954) The Foundations of Statistics, New York: John Wiley.
Schick, F. (1986) `Dutch Bookies and Money Pumps', Journal of Philosophy, 83,
12±119.
Schmeidler, D. (1989) `Subjective Probability and Expected Utility Without
Additivity', Econometrica, 57, 571±87.
Sugden, R. (1991) `Rational Choice: A Survey of Contributions from Economics
and Philosophy', Economic Journal, 101, 751±85.
Vanberg, V. J. (1994) Rules and Choice in Economics, London: Routledge.
Williamson, O. E. (1985) The Economic Institutions of Capitalism, New York:
Free Press.
Part III
Micro, macro, and markets
9 Essences and markets

JOHN O'NEILL

Socialists and liberals have engaged in a long-standing debate in political

philosophy about the desirability of markets. Those debates have focused
on a series of questions about the market: the kind of moral character it
fosters, its tendency to enhance or diminish human welfare, the distri-
bution of goods it promotes, its relationship to political democracy and
freedom, its compatibility with socialist goals, and so on. Recently, the
very possibility of this debate has been questioned. The whole tradition
of argument about the market is rejected on the grounds that it assumes
an ``essentialist'' view of the market.1 Both defenders of the market and
its traditional socialist critics assume that it is possible to talk of ``the
market.'' They assume that different markets share some essential nature
such that one can engage in a general discussion of the relation of the
market to moral character, welfare, justice, freedom, democracy, and so
on. However, the argument goes, that essentialist assumption should be
rejected. The standard arguments between defenders and critics of the
market rest on a mistake.

``Essences and Markets'' by John O'Neill, The Monist, vol. 78. no. 3, pp. 258±275.
Copyright # 1995. THE MONIST, La Salle, Illinois 61301.
1 For typical examples of such attacks on market essentialism see B. Hindess, Freedom,
Equality, and the Market (London: Tavistock, 1987), pp. 147±58, T. Cutler, ``Social
Theory and Social Policy,'' Economy and Society, 17 (1988), 251±71; J. Tomlinson, The
Unequal Struggle? British Socialism and Capitalist Enterprise (London: Methuen, 1982),
ch. 7, and ``Market Socialism,'' in B. Hindess (ed.), Reactions to the Right (London;
Routledge, 1990). The anti-essentialism about markets of Hindess and Cutler has its roots
in a general anti-essentialism that appears in their earlier work: see for example the
criticism of Hilferding for his essentialism in A. Cutler, B. Hindess, P. Hirst, and
A. Hussain, Marx's ``Capital'' and Capitalism Today (London: Routledge and Kegan
Paul, 1977). p. 69. Another in¯uential anti-essentialist ``post-Marxist'' text which includes
speci®c criticism of essentialism about the economy is E. Laclau and C. Moufe, Hegemony
and Socialist Strategy: Towards a Radical Democratic Politics (London: Verso, 1985),
pp. 75±85 and passim.

157
158 Micro, macro, and markets

This rejection of the existence of any essence to ``the market'' has clear
appeal in an intellectual world dominated by a variety of positions that
employ that most empty of pre®xes ``post.'' Thus, the rejection of market
essentialism is attractive to those ``post-Marxists'' and ``post-liberals''
who assume that the market is an achievement which the recent demise of
``actually existing socialism'' shows that we can now look beyond: the
debate over the relation between socialism and the market is one that we
can put behind us since it depended on the assumption that the market
has certain essential properties which render it incompatible with soci-
alism. In the ``post-Marxist'' and ``post-liberal'' world, the debate moves
onto new territory concerning the proper institutional framework in
which markets can operate.2 Amongst ``post-modernists'' the term ``es-
sentialist'' is used as a term of abuse: we live in a world without essences,
one in which Heraclitus is proclaimed a hero. Essentialism is rejected
variously for being incompatible with the recognition of ``difference,'' for
entailing a reduction of social categories to non-social natural categories;
it is taken also to be philosophically naãÈve, having been effectively
demolished by Wittgenstein. The post-modern moves against essentialism
are most frequently employed in discussions of gender. However, the
same moves can and have been made in debates in political economy and
appear in particular in discussions of ``market essentialism.''
My aim in this paper is to defend the questions about the market that
have been traditional to political philosophy against these recent anti-
essentialist arguments. I argue that if essentialism is presupposed by these
questions, the debate is none the worse for it.3 Neither liberal defenders
of the market nor its socialist critics are in error in making essentialist
assumptions. Much of the recent criticism of essentialism is founded on a
caricature which is defended by nobody. In the ®rst section I draw a
more accurate picture of the essentialist position by looking back at the
work of the founder of essentialism, Aristotle, and outlining something
of his in¯uence. The second section shows that, properly characterized,
the essentialist position is immune to the recent anti-essentialist wave of
arguments that the various positions that are ``post'' this or that have
offered. However, to reject those arguments is not to render essentialist
claims about the market immune from criticism. I conclude by brie¯y
outlining what a defensible and empirically grounded criticism of essenti-
alist assumptions about the market would need to look like.
2 I criticize this line of argument in J. O'Neill ``Essentialism and the Market,'' Philosophical
Forum, 26 (1994).
3 I leave aside here the issue as to whether essentialism is necessarily presupposed in the
traditional debate. As I note in section 2.D, many of the questions that the traditional
debate assumes may survive the failure of an essentialist approach to markets.
 Essences and markets 159

1 Aristotle, essences, and markets

The distinction between an entity's essential and accidental properties,

the ascription of essences to social institutions, and the use of essentialist
assumptions in discussion of the market, can all be traced back to the
work of Aristotle. His work provides a paradigmatic example of an
essentialist position.

A Essence and accident 4

The essential properties of an entity of a particular kind are those

properties of the object that it must have if it is to be an object of that
kind. Accidental properties of an entity of a particular kind are those
properties it has, but could lack and still be an entity of that kind. To
take a standard example from the modern natural sciences, consider
what it is for a substance to be copper. There are properties an entity
must have if it is to be a specimen of copper, e.g., ductility, malleability,
fusibility, electrical conductivity, atomic number 29, and so on. There are
other properties possessed by some or all samples of copper that they
could lack and still be copper. Thus, for example, it may be this lump of
copper is a doorstop. It is a property it could lack and still be copper. It
may be that all copper comes in lumps smaller than that of the Taj
Mahal. However, that property, if it is true of all lumps of copper, is not
one that they must have. It is possible, given the nature of copper, for
them to come in sizes greater than that.
A few standard observations about essences thus de®ned:
1. Essence precedes investigation and requires investigation. The essen-
tial properties of objects are properties that we discover that they have by
investigation, not by simply looking at them or by looking in a diction-
ary. The electrical conductivity of copper and its speci®c atomic number
are properties that we once did not know copper possessed and that we
now know that it does. The essence of an object precedes its discovery.
The appearance of an object does not reveal to us directly its essence.
2. Many essential properties of objects are dispositional properties that
are only actualized in certain circumstances. Thus ductility, malleability,
fusibility, and electrical conductivity are all clearly dispositional properties

4 Aristotle, Topics, Book I.

160 Micro, macro, and markets

± capacities and powers that particular samples of copper have, which

they exhibit in certain conditions. To discover those properties requires
that those conditions be realized. A sample of copper may never exhibit
the powers and capacities that it possesses. That this is the case is one
reason why the discovery of the essences of natural objects involves
experimental investigation: powers can be discovered by setting up those
conditions in which they are exhibited.
3. Some of the essential properties of objects are dependent upon others
and part of the purpose of scienti®c investigation is to discover those
dependencies: it is the atomic structure of copper that explains its
ductility, malleability, fusibility, and electrical conductivity.5 However,
the dependent-essential properties of objects are no less real than those
that are explanatorily prior to them. To explain is not to explain away.

B Essence and development

Aristotle's biological and social writings contain a second and distinct set
of teleological claims about essences that need to be distinguished from
the basic claims just made. Aristotle asserts that an entity's essence can
be revealed in the pattern of its development, that is, that there is a
normal developmental sequence to an entity of some kind which repre-
sents an unfolding of the essential nature of that entity, such that the
®nal ``mature'' stage of the normal developmental sequence exhibits most
clearly the essential nature of the entity. Entities that exhibit such
developmental sequences include both biological beings and social
institutions. Consider, for example, his account of the development of
the polis from the earlier associations of household and village:
[The polis] is the end or consummation to which those associations move, and the
``nature'' of things consists in their end or consummation; for what a thing is
when its growth is completed we call the nature of that thing, whether it be a man
or a horse or a family . . .6
Aristotle here treats the development of both individual organisms ±
humans and horses ± and of social institutions ± families and political
communities ± as exhibiting in the same way a development from
5 In the language of Locke, the atomic structure of copper is its real essence, the
explanatorily dependent properties, its nominal essence (J. Locke, Essay Concerning
Human Understanding, III, 3). See also Aristotle's distinction between ``real'' and
``nominal'' de®nitions in Posterior Analytics, Book II.
6 Aristotle, Politics, I.ii, trans., E. Barker (Oxford: Clarendon Press, 1950).
 Essences and markets 161

immature to mature states. The developmental pattern exhibits the

unfolding of the essential nature of a being of that species, that develop-
ment that would occur if not interrupted by some external accident. For
both organisms and institutions the essential nature of an object is
exhibited in its mature stage since at that stage all the potential powers of
that kind of being are fully developed.

C Essences and markets

Aristotle uses both the general essence/accident distinction and the more
particular teleological account of essences to describe the market as a
social institution. To characterize those properties of the market which it
possesses in virtue of the kind of economic institution it is, as against
accidental features of this or that market. Aristotle starts by considering
the features of the market that distinguish it from non-market economic
institutions, speci®cally the household. He proceeds thus in virtue of his
taxonomic approach to the characterization of an essence which he
inherits from his biology: a sentence characterizing an essence of an entity
of a particular kind standardly has the form of a noun phrase indicating a
genus modi®ed by an adjectival phrase that expresses the differentiae,7
the properties that are peculiar to that kind.8 The market is de®ned as a
mode of acquisition (genus) which considers objects for exchange and
which knows no limits to the accumulation of wealth (differentiae).
In the Politics Aristotle distinguishes between two forms of acquisition,
the economic and the chrematistic: the former is characteristic of the
household,9 the latter is characteristic of the market. The two modes of
acquisition are correlative with two possible uses that articles of property

7 In the Metaphysics the de®nition of the essence of a species is reduced to just one
differentiating property (Metaphysics, VII 12 and VIII 6). However, in the De Partibus
Animalium he claims that ``no single differentia . . . can possibly express the essence of a
species,'' and defends, rather, de®nition through clusters of differentiae (De Partibus
Animalium I. 2±3).
8 See Topics, Book I, 4±6 and Book VI. The procedure of de®ning the nature of a being of
a particular kind through consideration of what is peculiar to it is not always felicitous ±
what is peculiar to entities of a certain kind is not co-extensive with that which is
characteristic of their nature.
9 The household here needs to be understood as a unit of both production and
consumption, and not, as it is in the modern world, only as a unit of consumption: in the
classical world commerce operated at the margins of the household. On the central place
of household in the ancient economy and the marginal role of commerce see G. St. Croix,
The Class Struggle in the Ancient World (London: Duckworth, 1981), ch. III, iv.
162 Micro, macro, and markets

can have: the ®rst a use which is ``peculiar and proper'' to the object ±
say of a sandal, to be worn ± the second a use as an item to be exchanged
with others.10 Economic acquisition, that of the household, considers
acquisition only with respect to the object's primary use, as an object
that satis®es a need. Objects thus employed constitute, for Aristotle, the
real components of wealth and, in opposition to Solon, Aristotle asserts
that this wealth has bounds: ``the amount of household property which
suf®ces for a good life is not unlimited.''11 The second form of acquisi-
tion, the chrematistic,12 is concerned with the accumulation of the means
of exchange in the market: ``It is the characteristics of this second form
which lead to the opinion that there is no limit to wealth and property.''13
This form of wealth-making is often confused with the ®rst, but the two
are distinct, and constitutive of quite different institutions. While there is
a limit to the accumulation of natural goods in the household, namely
the needs they satisfy, there is no limit to the acquisition of the means of
exchange: ``There is no limit to the end it seeks; and the end it seeks is
wealth of the sort we have mentioned [i.e., wealth in the form of currency]
and the mere acquisition of money.''14
The development of the second art of acquisition, the chrematistic, is
presented by Aristotle as the outcome of a natural developmental
sequence leading to market institutions. The market begins in simple
exchange, barter, in which ``things which are useful are exchanged
themselves, and go directly, for similar useful things,'' which in turn
develops into exchange through the medium of money, which in turn
gives rise to the new and distinctive art of acquisition:
When, in this way, a currency had once been instituted there next arose, from the
necessary process of exchange [i.e., exchange between commodities with money
serving merely as a measure], the other form of the art of acquisition which
consists in retail trade [conducted for pro®t] . . . The result has been the
emergence of the view that the art of acquisition is specially concerned with
currency. . .15
The chrematistic art of acquisition, because it involves acquisition of
objects not for their ``proper and peculiar'' properties, is, for Aristotle,
``unnatural'' for humans, and hence to be held in check. It needs to be

10 Aristotle, Politics, Book I, ch. 8, 1256b 27ff, p. 26.

11 Aristotle, Politics, 1236b 31.
12 The use of the term ``chrematistic'' I outline here is that which Aristotle employs most
widely in the Politics. Aristotle does however sometimes use it in a more neutral sense:
for an account of the different uses see Barker in Aristotle, Politics, pp. 226±7.
13 Aristotle, Politics, 1256b 40ff, p. 27.
14 Aristotle, Politics, I. ix 1257b 28ff, p. 31.
15 Aristotle, Politics, I. ix.
 Essences and markets 163

held in check, however, because it is the normal developmental outcome

of exchange.
The essential nature of market exchange is revealed for Aristotle in its
developed ``unnatural'' form. In this ®nal state the dispositional proper-
ties of markets are exhibited. The most signi®cant of these for Aristotle is
the particular moral character it tends to produce: Aristotle's political
theory is by and large concerned with the tendency of different social and
political institutions to issue in different virtues and vices. The market,
where its development is unchecked, tends to issue not in the virtues
constitutive of a ¯ourishing human life, but in the vice of pleonexia, the
disposition to want more than is proper.

D Aristotelian essentialism: distinctions and in¯uences

Aristotle's essentialism is a complex of claims that are logically indepen-

dent of one another. One might accept his basic distinction between the
essential and accidental properties of an object without accepting the
further teleological speci®cation of essences in terms of normal patterns
of development. Indeed it is worth noting that the example of essentialist
descriptions of copper I used in section A does not conform to the
Aristotelian teleological model. Likewise one might accept essentialism in
its basic form without accepting his particular taxonomic approach to the
speci®cation of essences. Again, it is possible to adopt an essentialist
position in either its basic form or its teleological elaboration for entities
in the natural world, but to reject it for the social world. One might accept
essentialism about social institutions generally and markets in particular,
but reject the teleological elaboration of essentialism that Aristotle offers.
The options open to the ``essentialist'' are much more varied and complex
than recent anti-essentialist caricatures of the position allow.
The picture of Aristotle's in¯uence on subsequent essentialist thought
about social institutions in general and the market in particular is
similarly complex. Both Aristotle's essentialism and his discussion of the
market have had a large in¯uence on philosophical and economic
thought. Aristotelian essentialism, as far as description of the natural
world is concerned, has had a long-standing in¯uence and indeed has
recently undergone something of a revival.16 In the social realm it is

16 See, for example, D. Wiggins, Sameness and Substance (Oxford: Blackwell, 1980),
H. Putnam, Philosophical Papers, vol. 1 (Cambridge: Cambridge University Press, 1975),
and S. Kripke, Naming and Necessity (Oxford: Blackwell, 1972). In the philosophy of
164 Micro, macro, and markets

possible to ®nd theorists, most notably Hegel and Marx, who self-
consciously defend Aristotle's essentialism in its full teleological form.17
Aristotle's discussion of the market has also exercised a large in¯uence
on subsequent economic thought.18 In some theorists both philosophical
and economic in¯uences converge. The most notable example of such a
thinker is Marx, whose work exhibits the in¯uence of Aristotle's basic
distinction between essence and accident, his teleological speci®cation of
essences, and his more particular discussion of the market and its speci®c
mode of acquisition.19
However, Aristotle's in¯uence in economics has often been indepen-
dent of any explicit or self-conscious appeal to his essentialist premises.
Moreover, where essentialist reconstructions of later economics is pos-
sible, they often require the rejection of Aristotle's teleological speci®ca-
tion of essences. Consider for example the work of Karl Polanyi whose
work owes an explicit debt to Aristotle's discussion of the household and
market. Aristotle's contrast between household and market is taken by
Polanyi to provide the key to the analysis of the development of modern
market society: Aristotle's ``famous distinction of householding proper
and money-making . . . was probably the most prophetic pointer ever
made in the realm of the social sciences; it is certainly still the best
analysis we possess.''20 The development of modern market society, the
story of Polanyi's The Great Transformation, is a story of the escape of
the economy from the social and ethical limits that constrained it in pre-
modern societies: Aristotle's greatness lies in his anticipating the con-
sequences of this disembedding of the economy from constraining
institutions.21 However, Polanyi employs Aristotle's account of modes of
acquisition without self-conscious reference to his essentialist presupposi-

science the work of Harre has been particularly in¯uential in reviving essentialism,
although Locke's essentialism is often the immediate in¯uence. See R. Harre and
E. Madden,Causal Powers (Oxford: Blackwell, 1977).
17 For an excellent discussion see S. Meikle, Essentialism in the Thought of Karl Marx
(London: Duckworth, 1985).
18 On Aristotle's in¯uence in ecological economics see J. Martinez-Alier, Ecological
Economics (Oxford: Blackwell, 1987). See also J. O'Neill, Ecology, Policy and Politics:
Human Well-Being and the Natural World (London, Routledge, 1993), ch. 10 and
``Polity, Economy, Neutrality,'' Political Studies, forthcoming.
19 For a further elaboration of Marx's essentialism and its debt to Aristotle see S. Meikle,
Essentialism in the Thought of Karl Marx (London: Duckworth, 1985), especially ch. 3.
See also A. Wood, Karl Marx (London: Routledge and Kegan Paul, 1981), part 5, and
J. O'Neill, ``Essentialism and the Market,'' Philosophical Forum, 26 (1994).
20 Polanyi, The Great Transformation, pp. 53±4; cf. Polanyi ``Aristotle Discovers the
Economy'', in Primitive, Archaic and Modern Economies (Boston, MA: Beacon Press,
1957).
21 K. Polanyi, The Great Transformation (Boston, MA: Beacon Press, 1957), p. 54.
 Essences and markets 165

tions. Moreover, Polanyi's arguments are clearly inconsistent with Aris-

totle's speci®c teleological account of the development of market ex-
change: a central thesis of The Great Transformation is that there is no
natural developmental sequence in the development of the modern
market economy.22 While, as I show below, Polanyi's work is best
interpreted in essentialist terms, it is in terms of Aristotle's basic
essentialism without its teleological turn.
Aristotelian essentialism in a non-teleological form also provides the
best philosophical foundation of much in the Austrian tradition of
economics.23 However, some of the theorists in that tradition, most
notably Hayek, have a complex relation to the Aristotelian tradition.
Hayek is strongly opposed to the tradition of criticism of the market
represented by Aristotle and his socialist successors such as Marx and
Polanyi. He is aware of the Aristotelian heritage of modern socialism and
deeply critical of it.24 However, Hayek does take over and rework the
Aristotelian distinction between household and market. The distinction
between them is reinterpreted not as that between economic and chrema-
tistic arts of acquisition, but as that between economic and catallactic
social orders. The account of the economic order is built on the house-
hold. However, its distinguishing feature is now understood to be the
deliberate subordination of means to a shared set of ends.25 The market
is not in the proper sense of the term an economy, since ``its activities are
not governed by a single scale or hierarchy of ends.''26 The market rather
consists of a network of economies, of households and enterprises, each
with its own ends, each having to adjust and coordinate its activities with
others through exchange, but together ungoverned by any order of ends.
The market order is governed only by abstract rules of property and
contract. An order of that kind Hayek refers to as a catallaxy.27
At the level of his philosophical presuppositions the relation is still

22 See especially Polanyi, The Great Transformation, ch. 5.

23 On the relation between Austrian economics and Aristotelian essentialism see B. Smith,
``Austrian Economics and Austrian Philosophy'', pp. 2±5 in W. Grassl and B. Smith
(eds.), Austrian Economics (London: Croom Helm, 1986) and ``Aristotle, Menger and
Mises: An Essay in the Metaphysics of Economics'', in B. Caldwell, Carl Menger and His
Legacy in Economics (Durham, NC: Duke University Press, 1990). For an essentialist
reading of Austrian economics see U. MaÈki, ``Mengerian Economics in Realist
Perspective'', in B. Caldwell, Carl Menger and His Legacy in Economics (Durham, NC:
Duke University Press, 1990) and ``Scienti®c Realism and Austrian Explanation,''
Review of Political Economy, 2.3 (1990), 310±44.
24 See, for example, Hayek, The Fatal Conceit (London: Routledge, 1988), pp. 45±8.
25 F. Hayek, Law, Legislation and Liberty, vol. 2 (London: Routledge and Kegan Paul,
1976) p. 107.
26 Ibid., p. 108. 27 Ibid., p. 109.
166 Micro, macro, and markets

more complex. Hayek's economic thought has been presented as essenti-

alist by some of his recent anti-essentialist critics, for example Hindess.28
Others, notably Gray, have given a strong anti-essentialist reading to
Hayek's own methodological claims about his work.29 My own view is
that, whatever Hayek's own philosophical conception of his work, his
writing is best understood in realist and essentialist terms.30 For example,
Hayek's claim that the market is a catallactic order is best interpreted as
a putative discovery about the nature of the market order and likewise
his rejection of the claim that the market order is an economy is best
understood as a criticism of a misconception about its nature.31 In the
following I will assume that the charge of essentialism made against
Hayek by his critics is one that can be made out. However, against these
critics, I argue this is a virtue of his work, not a vice.
In the rest of the paper I defend essentialism about markets against
recent criticism. In doing so I consider only the claim that the market has
an essential nature. I will bracket any critical discussion of the claims
made by Aristotle and Marx that the essential nature of the market is
revealed in a developmental sequence. I do so because they are not
relevant to the question of whether or not the traditional debates of
political philosophy about the market rest on a mistake. While the
debates may assume essentialism about the market, they do not normally
depend on any assumption about the normal developmental sequence of
the market.

2 In defence of essentialism

A How not to talk about essences

Criticism of essentialism often begins with quite misleading characteriza-

tions of what is meant by ``essence.'' Consider for example the criticism
of market essentialism offered by Hindess against both defenders of free
markets such as Hayek, and socialist critics of the market such as Marx:

28 B. Hindess, Freedom, Equality, and the Market (London: Tavistock, 1987), ch. 9.
29 Gray offers a Kantian reading of Hayek's position which is set against the ``Aristotelian
method of seeking essences or natures of things''; J. Gray, Hayek on Liberty (Oxford:
Blackwell, 1984), p. 6. For criticism of Gray see M. Peacock, ``Hayek, Realism and
Spontaneous Order,'' Journal of the Theory of Social Behaviour, 23 (1993), 249±64.
30 Gray himself acknowledges that this is possible. J Gray, Hayek on Liberty, p. 117.
31 Ibid., pp. 112ff.
 Essences and markets 167

These various positions arrive at their assessments of the market in rather

different ways, but they nevertheless share an essentialization of the market and
the problems that this generates for social analysis. To write of essentialism in
this context is to say that the market is analyzed in terms of an essence or inner
principle which produces necessary effects by the mere fact of its presence.32
The criticism relies on a caricature of what it is to say that an entity has
an essence, and it results in an account of essentialism about the market
that neither liberal proponents of markets nor socialist critics hold. In
using the language of essences one is not, as Hindess claims, describing
inner principles that produce ``necessary effects'' by the mere fact of their
presence. Rather, one is concerned with the nature and explanation of
the capacities and powers of objects to produce certain effects. The
liberal and socialist theorists that Hindess criticizes claim that the market
has certain dispositional properties ± to facilitate the accumulation of
capital without limit, to foster vices, say that of pleonexia, or virtues,
such as those of the autonomous character, to determine a particular
price for a commodity, and so on ± and are concerned to explain these. It
does not follow that these dispositions are always exhibited. That
Hindess wrongly characterizes essentialism undermines his more substan-
tial criticisms of market essentialism. The criticisms are aimed at posi-
tions that nobody holds.

B The institutional contexts of the market

At the centre of Hindess's substantial criticism of market essentialism is

the claim that markets do not and could not exist in an institutional
vacuum. Markets appear in a variety of institutional contexts. In
elaborating this claim Hindess makes two points ± that markets presup-
pose other institutions, and that markets operate in different institutional
contexts. Both points can and should be accepted. Neither, however,
entails the falsity of essentialism.
In criticizing Hayek's essentialism, Hindess accuses him of inconsist-
ency in also holding that market institutions presuppose non-market
institutions.
[I]n parts of The Constitution of Liberty (especially Ch. 4, ``Freedom, Reason and
Tradition'') he suggests that the effective workings of the market depend on the
presence of the appropriate traditions and moral codes. But in spite of that

32 B. Hindess, Freedom, Equality, and the Market, p. 149.

168 Micro, macro, and markets

recognition his anti-planning polemic achieves its super®cial appearance of

plausibility by essentializing ``the market.''33
Hayek's recognition that market transactions presuppose a set of prior
non-market institutions sets him against pure contractarian versions of
liberalism which hold that all relations could be modelled on contract.
However, Hayek is guilty of no logical inconsistency in combining non-
contractarianism about markets with essentialism. That a particular kind
of object or institution can exist only given particular conditions is quite
irrelevant as to whether the object or institution has essential properties.
That there are institutional preconditions of markets does not entail that
markets lack essential powers. Hayek's claim that the market has
institutional prerequisites is quite consistent, for example, with the claim
that it is essentially a catallactic order that reduces coercion and thereby
cultivates human freedom.
Hindess's second point is that markets operate in different institutional
contexts:
what is shared by all markets is little more than the fact that something is
marketed in them. Otherwise they are highly differentiated. Markets always
operate under speci®c institutional conditions, which can vary considerably from
one case to another. What is meant by institutional conditions in this context are:
the market actors (large corporations, government departments, small businesses,
producer and retail co-operatives, private individuals, etc.) and the resources
made available to them; legislative regulation and other forms of administrative
and political controls; customary and other informal constraints on acceptable
behaviour; linkages with and spillovers into other markets engaging different
actors and controls.34
That markets operate under a number of distinct institutional conditions
and have different effects given these conditions is uncontentious.
However, it does not follow that essentialism about markets is false or
that the only general statement one can make about markets is that
``something is marketed in them.''
Two initial general points are in order here. First there is nothing in
essentialism that disallows variation in the properties of different
instances of some kind of entity: the distinction between accidental and
essential properties serves in part to distinguish what is of the nature of a
thing and what varies. Second, that the essential properties of markets
might be exhibited only in certain institutional conditions does not entail
that they are not present nor that reference to them might not feature in
the explanations of the behaviour of markets.
An essentialist account of the market can allow that there exist a

33 Ibid., p. 159. 34 Ibid., p. 150.

 Essences and markets 169

variety of accidental features of speci®c market relations in particular

institutional contexts. These conditions will often entail that the powers
of markets are not exhibited. There may exist constraints that ``inhibit''
the exercise of those powers. Consider Polanyi's use of the concept of
chrematistic acquisition to specify the nature of market economies. It is
only in modern conditions, in which the market is ``disembedded,'' that
the potentialities of a market economy based on the principle of the
unrestrained acquisition of wealth without limit are fully exhibited. In
saying that the market is disembedded in modern capitalism Polanyi is
not claiming that markets here exist without any institutional context.
He recognizes that it was the product of and still relies upon a particular
political and legal framework.35 When Polanyi refers to the disembed-
ding of the market he is referring to the disappearance of legal and
customary regulatory constraints on the working of the market. In these
conditions the potentialities of markets become visible. However, those
potentialities existed within markets in their embedded state. The relation
between markets and surrounding institutional contexts was one of
tension. It was this tension to which Aristotle was responding in making
his distinction between household and money-making. Reference to
those potentialities appears in Polanyi's explanation of the collapse of
the Speenhamland Law which prevented a free market in labor.36
The co-existence of non-market relations alongside market relations
between agents might also entail that the market does not exhibit its
essential properties. For example, it might be the case that, as Hegel
claims, market relations are essentially impersonal, and this might be
taken to be a virtue or vice of markets. However, the impersonality of
market relations might often not be exhibited in virtue of other personal
ties between particular actors. Ties of kin, particular historical loyalties
between members of a community, and so on, might mean that market
transactions within a small village community do not exhibit the imper-
sonality of those of the city. The defender of the claim that markets are
essentially impersonal might still quite properly argue that the personal
bonds are accidental features of those speci®c market transactions: as the
surrounding accidental ties are eroded, markets tend to impersonality.
That essentialism about the market is compatible with acceptance of
the existence of variation in markets in different institutional conditions
also undermines a criticism of essentialism, popular in postmodern
literature. The term ``essentialist'' has, in postmodernist circles, become
a term of abuse. The major reason for this anti-essentialism is the

35 Polanyi, The Great Transformation, p. 139.

36 Ibid., ch. 7.
170 Micro, macro, and markets

assumption that essentialism is incompatible with difference. Postmoder-

nists claim to celebrate difference and diversity and to say that a certain
class of entities has an essential nature that they all share is taken to entail
a denial of difference. In the postmodern thesaurus the term ``essence''
appears in the same list as ``uniformity'' and ``homogeneity.'' Similarly,
the assumption that essentialism is incompatible with difference underlies
the rejection of essentialism about markets. It is assumed that to talk of
there being an essential nature of markets involves a denial of the
possibility of differences between distinct markets. That assumption is
false for the reasons outlined. Essentialism is quite compatible with the
recognition of variation. That a number of entities share some common
nature is quite consistent with the existence of differences between them.
There is, however, a partial concession that needs to be made to the
anti-essentialist about markets. There is certainly a danger associated
with essentialism of taking properties of one particular variant of a
species of being to be essential properties of all. Thus, for example, there
feminists criticize essentialism, it is often the way that culturally speci®c
attributes and relations are taken to be shared by all women that is the
object of criticism. Likewise it is a mistake to assume speci®c accidental
properties of a particular market order, say that in Western Europe or
the United States, to be essential properties of all markets. However, the
defensibility of speci®c essentialist claims needs to be distinguished from
the defensibility of essentialism. That some essentialist theorists make
false claims does not entail the falsity of essentialism.

C Social meaning and essence

Thus far I have outlined the essentialist position by way of parallels

between the use of the language of essences to describe natural objects
and its use in social theory. There are dangers in parallels between
essentialist claims about human beings and social institutions, and
essentialist claims about natural objects, which invite criticism of essenti-
alism in social theory. Thus another common reason why some feminists
are critical of any essentialist account of the nature of women is that it is
assumed that it entails biological reductionism and a commitment to a
purely biological explanation of gender differences and power relations.
If this is the case then the anti-essentialist criticism would be well-
founded. However, essentialism need not involve any such failure to
recognize the differences in the nature of the objects of the human
sciences and those of the natural sciences. More speci®cally, the essenti-
 Essences and markets 171

alist can accept that institutions like markets are not natural objects: in
particular, essentialism is consistent with the claim that social objects are
constituted by relationships and acts which have social meanings whereas
natural objects are not. Consider, for example, the account offered by
Hegel, a strong essentialist in the Aristotelian tradition, of the essential
differences between the contractual market sphere of civil society on the
one hand and the family on the other. Hegel claims that ``marriage, so
far as its essential basis is concerned is not a contractual relation.''37 The
relationship of marriage partners is such that one's identity is partially
constituted by that relationship. In contrast, contracts are essentially
between persons considered as ``self-subsistent persons,'' whose identity
is independent of the contractual relation. Whatever might be said for or
against Hegel's position, his arguments are based on differences in the
social meanings of different relationships and acts. Contractual relations
and personal relations are essentially different in virtue of the meanings
constitutive of them. In this respect at least, the essential natures of social
objects are different from those of natural objects. However, the essenti-
alist need have no dif®culty in recognizing that the objects of the human
sciences do have such distinct properties.

D Wittgenstein, games, and family resemblance

Another major source of inspiration for the rejection of essentialism,

including that about the market, is Wittgenstein's well known criticism of
the ``craving for generality'':
There is . . . the tendency to look for something in common to all the entities
which we commonly subsume under a general term. ± We are inclined to think
that there must be something in common to all games, say, and that this
common property is the justi®cation for applying the general term ``game'' to the
various games; whereas games form a family the members of which have family
likenesses.38
I have no argument at all with Wittgenstein's argument here. However, I
do not believe that it undermines essentialism. It is quite true that there is
no reason to assume that there must be something in common to all
entities that fall under a general term which justi®es the application of

37 Hegel, Philosophy of Right (T. Knox trans.) (Oxford: Oxford University Press, 1967),
163R.
38 L. Wittgenstein, Blue Book (Oxford: Blackwell, 1960), p. 17; see also Philosophical
Investigations (Oxford: Blackwell, 1958), paragraphs 66 and 67.
172 Micro, macro, and markets

the term. However, no general conclusion follows from that about the
truth of essentialism. The scienti®c endeavour to specify and explain the
nature of copper or certain species of animal or plant is not ruled out by
Wittgenstein's argument. The proper conclusion that follows from
Wittgenstein's discussion is that one cannot assume in advance that there
is some set of essential properties shared by all entities that fall under
some concept, not that no essential properties exist. Turning to markets,
it is quite possible that different markets, like games, may turn out to
share only some family resemblances. It does not follow that the attempt
to discover essential properties is a mistake. It does mean that one may
be unsuccessful. In the end it is a matter of empirical investigation.
Two further points need to be added. First, Wittgenstein's conclusion
is quite independent of the nominalist thesis that the only thing in
common between all entities that fall under a general term X is that they
are called X, such that, in principle, we could extend the reference of
term to whatever we like. The entities that fall under a term do share
family resemblances ± ``a complicated network of similarities overlapping
and crisscrossing: sometimes overall similarities, sometimes similarities in
detail.''39 Those similarities are real ± there does exist a network of
properties that thread together entities that fall under a term. If it is the
case that markets are, like games, united only by a set of family
likenesses, those likenesses are real.
Secondly, this existence of real family resemblances is suf®cient for the
intelligibility of many of the traditional questions asked about markets in
political philosophy. For example questions like ``What effects do markets
have on the moral character?'' like the question ``What effects does
engagement in games have on the moral character?'' make perfect sense
even in the absence of any single essential property shared by all markets
or all games. Likewise it makes sense to investigate the reasons for
blocking markets in certain goods ± votes, bodily parts, blood and so on ±
just as it makes sense to ask of games whether it is appropriate to play
them at a funeral. The existence of a family network of real resemblances
will suf®ce for such traditional questions raised in political philosophy.

E How to criticize market essentialism

Wittgenstein's argument does highlight the openness of essentialist

claims to empirical criticism. It is important to distinguish between

39 Wittgenstein, Philosophical Investigations, paragraph 66.

 Essences and markets 173

arguments aimed quite generally against essentialism and speci®c em-

pirical arguments against an essentialist claim about some set of entities
that fall under a general term. My criticism in this paper is entirely of
those arguments aimed quite generally against essentialism. That general
anti-essentialist arguments fail does not show that there must be some
form of market essentialism that is true. There may be good empirical
evidence speci®cally aimed only against essentialism about markets. It
may be that markets do not share any essential characteristics, and hence
that no essentialist theorist will get it right.
This last point can be developed in essentialist terms. It may be that in
exploring the differences between markets, what looked like one species
is in fact many. Dore opens a paper on ``What Makes the Japanese
Different?'' as follows:
Sheep come in all shapes and sizes. So do goats. In fact some sheep look like
goats, and vice versa. But as biological systems they are distinct; they won't
interbreed. Shoats and geep don't exist. Are capitalisms like that? Is it true that
there are different types of capitalism, and that the differences between them are
systematic. . . . ?40
Super®cially similar beings can turn out to be essentially different. It
might be the case that, as Dore suggests, super®cially similar economic
orders turn out to have quite different natures. The differences are
systematic. Those claims are not only consistent with essentialism about
social orders, they assume an essentialist programme. What is denied is
the more speci®c essentialist claims about the market. If it is right, some
of the essentialist claims made about the market would turn out to be
false, and theorists like Marx, Polanyi, and Hayek would have failed to
distinguish essentially different social orders. Whether this is the case is
in the end a question of empirical investigation. However, such empirical
criticism of market essentialism is not to be confused with those founded
on a general rejection of the language of essences. My main purpose in
this paper has been to clear the ground for proper discussion of
essentialist claims about the market by weeding out recent growths of
anti-essentialist scrub.41
40 R. Dore, ``What Makes the Japanese Different?'' in C. Crouch and D. Marquand (eds.),
Ethics and Markets (Oxford: Blackwell, 1993) , p. 66.
41 My thanks to Russell Keat, Uskali MaÈki, Mark Peacock, Andrew Sayer, and Barry
Smith for their comments on earlier versions of this chapter. I owe a special debt of
gratitude to the late Mary Farmer with whom I had several conversations on the issues
discussed here.
10 The metaphysics of microeconomics

ALEX ROSENBERG

The study of economics has been a going concern among philosophers

for the better part of twenty years without very many people even
noticing that economics has a metaphysics. Indeed, among economists
the term ``metaphysical'' is probably an epithet of opprobrium, employed
to suggest that a claim is untestable or otherwise without cognitive
signi®cance. Philosophers of economics will admit to the existence of an
epistemology of economics ± the study of the nature, extent, and
justi®cation of economic knowledge. But even this is controversial
among some economists.1 The notion that economics might have a
metaphysics as well as an epistemology will be scoffed at by those
economists doubtful about the enterprise of the philosophy of economics
altogether. Nevertheless, philosophers of economics should have no
trouble identifying a metaphysics in some of the theoretical commitments
of economics.
Or at least philosophers who have adopted a Quinean view of
metaphysics as continuous with physics ± and science generally, will have
no trouble ®nding metaphysical compartments of economic theory. To
see why, recall the Quinean teaching that there is no well-founded
distinction between necessarily true propositions, true in virtue of
meaning alone ± analytic truths, and contingent statements, true in virtue
of content ± synthetic truths. Nor do scienti®c claims meet experience
for tests one at a time; rather they meet experience in units no smaller
than entire theories. Accordingly, the difference between sets of metaphy-
sical statements and sets of empirical ones cannot be a matter of kind,
say testable versus non-testable, or necessary versus contingent; instead

``The Metaphysics of Microeconomics'' by Alex Rosenberg, The Monist, vol. 78, no. 3,
pp. 352±67. Copyright # 1995. THE MONIST, La Salle, Illinois 61301.
1 Cf. Donald McCloskey, The Rhetoric of Economics (Madison, WI: University of
Wisconsin Press, 1985), who rejects epistemology as a standard on which to assess
economics.

174
 The metaphysics of microeconomics 175

the difference is at most one of degree. The difference between metaphy-

sics and theory is one of generality and abstractness, and where metaphy-
sics begins and theory leaves off turns out often to be arbitrary.2 For this
reason there is no really adequate and illuminating de®nition of meta-
physics to be found in philosophy.
It is not surprising that sometimes research programs in the sciences
are driven by commitments which might as well be deemed metaphysical
as theoretical: determinism, reductionism, atomism. The role of meta-
physics in directing research programs will be even more prominent in
the empirically underdeveloped and non-experimental sciences. In the
social and behavioral sciences there are fewer hard experimental or
observational ®ndings to guide theorizing; the need for some direction in
theorizing can only be ®lled by appeal to very abstract and general
constraints often derived from other disciplines via philosophy ± episte-
mology, metaphysics, or their amalgam in the philosophy of science.
Thus, many research programs in psychology and sociology have been
inspired by metaphysical dicta variously inspired by understandings and
misunderstandings of everything from Darwinian evolution to Heisen-
bergian uncertainty. So it is with a good deal of economic theory.
Important features of the theory are determined by metaphysical com-
mitments both borrowed and invented.3 Among empirically well-devel-
oped disciplines ± the physical sciences, in particular ± metaphysical
disputes are rare because they have already been settled by the success of
some basic theories and the failure of others: controversies surround the
signi®cance of experimental data, and the competition among theories to
account for these data.
Interestingly, in all the other social sciences, debates about method-
ology and epistemology bulk larger, and it is the absence of much
agreement on what the data are that leads social scientists often to be
guided by metaphysics. Economics, especially neoclassical microeco-
nomics, like physical science, is almost entirely free of such debates. But
this does not show that there is among economists agreement about what
constitutes the data and what the data show. Rather, it betokens a
substantially greater agreement about metaphysics among economists
than in other social disciplines.
Metaphysics sans phrase is sometimes de®ned as the study of the
2 See W. V. O. Quine, From a Logical Point of View (Cambridge, MA: Harvard University
Press, 1953).
3 I argue for this view at some length in Alex Rosenberg, ``Methodology, Theory and the
Philosophy of Science,'' Paci®c Philosophical Quarterly, 66 (1985), 377±93, and Alex
Rosenberg, ``Normative Naturalism and the Role of Philosophy,'' Philosophy of Science,
57 (1990), 34±43.
176 Micro, macro, and markets

ultimate constituents of reality ± what fundamental kinds of things there

are, and how the differing things in this world and out of it can be shown
to be examples of these basic kinds. Mutatis mutandis the metaphysics of
a discipline is therefore the basic set of kinds or types of entities
countenanced by the established theory or theories of the discipline.
Microeconomics is the traditional theoretical core of the discipline of
economics, a widely known body of concepts, generalizations, and
analytical strategies shared by almost all academic economists. It is the
similarity across textbooks of economics, compared to the vast diversity
among the textbooks of other social sciences that reveals the meta-
physical homogeneity of the discipline.
The metaphysics of economics is in part the study of the basic kinds
over which the theory quanti®es ± that is what sorts of things the truth of
the theory commits us to. But, there is potentially something more to the
metaphysics of economics. For, it may be that economists are not fully
committed to the existence of entities of the sort the theory's truth
requires. If so, the metaphysics of economists seems to come into con¯ict
with the metaphysics of economics, for we need to distinguish what
economic theories state from what economists who use them believe.
This inconsistency between economists' beliefs and their favorite theories
may be tolerable to some economists, as inconsistencies seem tolerable in
other disciplines, like physics.4 But inconsistency will not sit well with
philosophers, and for that matter with others interested in whether to
believe economists or their theories. Of course, to some economists,
metaphysical inconsistency is just as unacceptable as any other kind, and
for the same reason: at least one of the inconsistent propositions must be
false, and false beliefs are to be expunged.
In this paper I attempt to identify some components of the metaphysics
of microeconomics, and to explore the interpretations of the theory
economists often provide because of their discomfort with the metaphy-
sical commitment of their theory. Even independent of the Quinean
conception of metaphysics, there appear to be signi®cant metaphysical
assumptions behind well-known methodological strictures of economic
theory. For example, conventional welfare economics subscribe to a
clear-cut distinction between facts and values, between normative claims
which cannot be empirically adjudicated and factual claims which can be.
If pressed to defend this orthodoxy the defender will begin with episte-
mological considerations which suggest that we cannot test moral
prescriptions, but eventually the defender needs to seek a metaphysical

4 Consider for example wave/particle duality, renormalization in quantum electro-

magnetism, or attitudes towards Bell's inequality.
 The metaphysics of microeconomics 177

basis for this claim, one which explains why there are no moral facts, or
what it is about their character that unsuits them for accommodation by
economic theory. There are, of course, philosophical theories that
economics could help itself to for such a task,5 but the task is clearly a
metaphysical one.
Relatedly, economic theory needs a metaphysical argument for its
rejection of interpersonal comparability of preferences. It is easy, of
course, to maintain a digni®ed silence when faced with the demand to
justify the claim that normative matters cannot be adjudicated beyond
the satisfaction of a criterion of Pareto optimality among incomparable
personal preferences. But without a fundamental argument ± a metaphy-
sical one ± the rejection of interpersonal comparisons of preference will
face repeated rejection, and microeconomic theory will continue to be
threatened with rejection for ideological motives. Or again, consider the
resolute refusal of microeconomic theory to accommodate a distinction
between wants and needs. Such a distinction is obvious outside eco-
nomics, and seems to have a reasonable foundation in biology, psychol-
ogy, and common sense. But economics is blind to the wants/needs
distinction largely because there seems no way to accommodate the
distinction while preserving the character of the theory of consumer
choice. Here too philosophers have entered where economists have
declined to tread.6
As with other cases, the economist's defense of these claims may be
stigmatized as metaphysical to the extent that it is not empirical.
Economists will be criticized for continuing to embrace a fact/value
distinction, or the interpersonal incomparability of preferences, or the
assimilation of wants to needs as matters of metaphysical commitment.
Some defenders of economic theory's legitimacy in the absence of
empirical con®rmation have implicitly embraced this metaphysical
defense, especially writers on economic methodology who have adopted
Lakatos's notion of a scienti®c research program.7 They have identi®ed a
``theoretical hard core'' of claims beyond the reach of empirical test
which are in effect metaphysical commitments beyond empirical
challenge.8

5 For example, J. L. Mackie, Ethics: Inventing Right and Wrong (New York: Penguin
Books, 1977).
6 See David Braybrooke, Meeting Needs (Princeton, NJ: Princeton University Press, 1987).
7 Imre Lakatos, ``The Methodology of Scienti®c Research Programs,'' in Philosophical
Papers, vol. 1, pp. 1±74 (Cambridge: Cambridge University Press, 1978).
8 For the best example of this strategy see E. Roy Weintraub, General Equilibrium Analysis:
Studies in Appraisal (Cambridge: Cambridge University Press, 1985).
178 Micro, macro, and markets

Change, stability, and equilibrium

The explanatory role of equilibrium ± general and partial ± in economic

theory must rest primarily on its metaphysical attractions. General
equilibrium theorizing derives the existence of a unique stable equi-
librium of supply and demand in every market for every good from the
present into the in®nite future from the preferences of individual agents
and their current endowment of goods (given some further assumptions
such as certainty, the absence of increasing returns to scale, in®nite
divisibilities, an in®nite number of agents, etc.). Much economic theori-
zing involves attempts to vary the assumptions of a general equilibrium
model and determine whether the equilibrium is still derivable. For
example, economists seek to know whether a general equilibrium exists if
we surrender assumptions about returns to scale, certainty about the
future, the number of consumers, the privacy of goods, the imposition of
various governmental policies, etc. Less demanding in its assumptions is
the approach of partial equilibrium, which examines how changes in one
market can move its equilibria in supply and demand from one point to
another, on the assumption that the change will not signi®cantly
in¯uence prices on any other market.
It is pretty clear to most people that general equilibrium does not
characterize any real market, not even as a ®rst approximation, and that
partial equilibrium is at best the exception to the rule of economic
processes. So, why the attachment of economic theory to such
equilibrium-approaches. A sociological and historical approach like
Mirowski's will ®nd the sources of this attachment to equilibrium in the
nineteenth-century envy of physical theory, and the conscious aping of
its theoretical structure by Walras and others.9 But this analysis cum
critique fails to do intellectual justice to economic theory, and fails to
recognize the important metaphysical commitment that explanations in
physics and economics share.
In both physics and economics, and for that matter evolutionary
biology, the object of theory is to explain changes: displacement or
motion in physics, diversi®cation and evolution in biology, exchange of
commodities in economics. But, how are such changes best explained?
Plato famously held that no sort of change can be explained. Therefore
Plato forswore natural science and causal inquiry.10 Newtonian me-

9 See Philip Mirowski, More Heat than Light (Cambridge: Cambridge University Press,
1989).
10 See, for instance, The Phaedo.
 The metaphysics of microeconomics 179

chanics provides one of the most powerful approaches to the explanation

of changes. For Newtonian mechanics shows us that what we view as
change is really rest ± the absence of change. What we take to be change
± the motion of the planets around the sun for example ± is actually an
equilibrium in a closed system conserving total energy. The solar system's
behavior re¯ects the absence of interfering forces, stasis, changelessness.
By showing that what we identity as change is in reality the result of the
constancy of more profound forces, and by showing how the system
returns to a new equilibrium when new forces intervene, Newtonian
mechanics provides a deeply satisfying, perhaps the most deeply satis-
fying explanation for apparent change. Similarly, contemporary evolu-
tionary theory explains diversity of ¯ora and fauna as the result of
changes in gene ratios caused by interfering forces ± mutation, migration,
drift, and environmental change, all of which disturb an equilibrium. The
result is evolutionary adaptation as organisms interact with one another
and with the environment over generations in the diirection of a new
adaptational equilibrium. Nature selects among random variants those
more well adapted to local environments ± where the local environment
means not just a place in space and time, but also competing organisms,
and for that matter other adaptations of the organisms evolving. The
ceaseless variation of the biosphere is in fact either merely cyclical
variation that masks the constancy of gene ratios or the result of
persistent movement in the direction of local optimal equilibrium solu-
tions to adaptational problems. Here too, change is explained by
showing that beneath it is changelessness or else movement back towards
a changeless state.
The commitment to equilibrium explanations in these disciplines is
deep enough to be metaphysical. Even in the succession of theories in
physics, from Newtonian mechanics to the special and general theories of
relativity, and to quantum electro- and chromodynamics, equilibrium
strategies of explanation have been preserved.
So too in economics. There seems little that is constant in economic
activity as we observe it. How to bring explanatory illumination to this
buzzing, blooming confusion? The apparently ceaseless change that
characterizes economic activity would be most powerfully and most
satisfyingly explained if we could show that beneath the appearance of
change was the reality of changelessness, or at least that the actuality of
change was but a ceaseless readjustment back towards states that would
be changeless once attained. General equilibrium theory shows that a
constant total quantity of goods is exchanged in trade in such a way as
more nearly to satisfy a set of constant preference functions, and that
once the trades are made, in the absence of interfering forces, nothing
180 Micro, macro, and markets

more of economic interest will happen. Moreover, economists hope to

show that the continual interference of exogenous shocks that charac-
terize a real market simply produces decentralized responses the net
effect of which is always to move the market back towards unique,
stable, and changeless equilibria. If economists ®nd it dif®cult to free
themselves from the commitment to equilibrium explanations, the reason
is not hard really to understand. When it comes to explaining change
there are few more appealing stratagems than reducing change to stasis.
The commitment to equilibrium explanation is metaphysically if not
empirically as well grounded in economics as it is in physics or biology.
The trouble is that the appeal to equilibrium is not well-founded in
economic data the way it is in evolutionary or mechanical dynamics.
Whence the decidedly metaphysical ± that is, non-empirical ± commit-
ment to its theoretical centrality.

Methodological individualism

Perhaps the most widely recognized metaphysical commitment of eco-

nomics is to some version of a doctrine of methodological individualism:
roughly the proposition that the behavior of economic systems, econo-
mies, markets, industries, ®rms, unions, and other collective entities are
to be explained by appeal only to the behaviors of their parts. Some
version of this thesis stands behind the methodological dictum that
economic phenomena of whatever level ± the individual trade, the ®rm,
the industry, the whole economy ± are ultimately to be explained
adequately by showing that they are the exclusive effects of the sum of
the behaviors of individual economic agents ± consumers, producers,
®rms, etc. The methodological dictum is evident outside of microeco-
nomics in the widespread demand that macro-theories be provided at
least eventually with complete microfoundations. Within microeco-
nomics, the very existence of a theory without microfoundations is
impossible: all microeconomic analysis begins with individual agents
responding to boundary conditions and then aggregates these responses
to determine a market effect.
The individualism of microeconomic theory is resolute, and has had
remarkable explanatory pay-off. Perhaps most striking is the unwilling-
ness of economists to surrender their commitment to individualism in the
face of apparently recalcitrant phenomena dealt with in game theory and
the theory of public goods. Public goods ± like police protection ± are
ones which cannot be provided to some consumers without being
 The metaphysics of microeconomics 181

provided to others. Thus individuals have an incentive to decline to pay

for them in the hope that others will do so, and thus provide a
commodity to non-players for nothing. Since all individuals are capable
of thus reasoning, in many cases we should not expect public goods to be
provided. However, it is obvious that many such goods are provided. It
would be easy to explain the provision of public goods simply by
appealing to group-preferences and the coercion of individuals by the
group to pay for the good.
But such explanations are ad hoc, and many of them are undermined
by the insight that rational agents all have an incentive to free-ride on
group preferences, thus leading to undersupply or serious enforcement
problems we do not in fact observe. Much of the advance in our
understanding of public goods and of the social and political institutions
which they often constitute is the result of a commitment to individu-
alism which refuses to countenance such irreducible group entities, and
seeks instead conditions that made the provision of public goods really
individually rational.
The resolute reductionism of microeconomics has been extolled as one
of its important strengths, akin to the atomism of successful physical
theories like the kinetic theory of gases. Elsewhere in social science, and
especially in some parts of sociology, under Durkheim's in¯uence, holism
continues to hold sway.
There have been endless discussions of what exactly a commitment to
individualism consists in. As with other disputes in the philosophy of
science which began as linguistic, this one started out as the question
whether social or other sorts of ``wholes'' and their properties could be
de®ned exhaustively by appeal to individuals and their properties. But
like the other disputes, the problem of holism versus individualism has
been transformed over time from a de®nitional problem into an explana-
tory one.11 It is now the question of whether adequate fundamental
explanation of all phenomena adverts only to individuals and their
properties. While the original dispute might have been settled by the
provision of suitable de®nitions, such de®nitions were not forthcoming.
The current version of the dispute may eventually be settled by the
advance of theory in the social sciences. But since theory has not
advanced far enough to settle the matter one way or another, the
continued dispute about holism versus individualism has become meta-
physical.

11 See, for instance, the work of Jon Elster, especially Nuts and Bolts for Social Science
(Cambridge: Cambridge University Press, 1988), and Explaining Technical Change
(Cambridge: Cambridge University Press, 1983).
182 Micro, macro, and markets

Philosophers, economists and others are reduced to providing a priori

considerations ± mainly in the form of rhetorical questions and tu quoque
arguments for and against individualism. Perhaps the best argument for
individualism is the observation that if there were no individuals there
would be no institutions, coupled with the rhetorical question of where
does the remainder come from, if institutions are more than the sum of
the individuals who participate in them? Holists will reply that on the one
hand they are not committed to the existence of a mysterious substance
above and beyond individuals that together with them constitutes social
whole. Rather, the interaction of individuals brings into existence higher-
level properties of social institutions which explain other features of
institutions without themselves being explainable by simple aggregation
from the properties of their component individuals.
Though philosophy may keep this metaphysical dispute alive, it can
never settle it. Only advances in empirical theory or decisions about the
scope of economics can do so. Should a theory whose axioms quantify
only over individuals, their expectations and preferences come to be
deemed adequate in its explanation of some body of aggregate economic
phenomena then of course individualism will be vindicated. Alternatively,
should economists decide to restrict their study only to such institutions as
can be explained by appeal to the behavior of individuals only, individu-
alism in economics will have been vindicated by the device of a question-
begging restriction on economics' scope. Until such time, individualism
will remain a metaphysical commitment widely shared by economists.

Intentionality

Meanwhile, not only does the development of microeconomics re¯ect a

commitment to individual human behavior as its fundamental explana-
tory factor. It also endorses a conception of the causes of human
behavior which involves taking sides on issues of the gravest metaphy-
sical import in western philosophy: the nature of mind. For economic
theory treats human behavior as choice and identi®es the causal variables
in determining behavior as expectations and preferences.
Expectations and preferences are cognates for beliefs and desires.
Thus, microeconomic theory is committed to explaining individual
behavior ± choice ± as the effect of the joint interaction of beliefs and
desires. Microeconomic theory makes very strong claims about prefer-
ences, and traditionally has operated under very unrealistic assumptions
about the accuracy and completeness of expectations.
 The metaphysics of microeconomics 183

The standard assumptions of the theory of consumer choice include

the claim that the agent can rank all commodities or other alternatives in
a weak transitive order, that the agent can identify all those alternatives
that are attainable for choices given the agent's resources, and that the
agent will always choose the attainable alternative that ranks highest. In
other words, the theory is committed to treating economic agents as
goal-directed systems that instantiate intentional states. But unlike
ordinary humans, economic agents are omniscient both about their own
preference orders, their budget constraints, and the attainable set of
outcomes. It is this omniscience that obscures the intentional character of
the expectations and preferences which drive economic choice. Inten-
tional states ± beliefs and desires ± are referentially opaque and not truth-
functional: the substitution of co-referential terms and co-extensive
predicates in statements describing them can convert such statements
from true descriptions of what people want and believe to false ones
describing neither their wants nor beliefs. Thus, for example, Lois Lane
may believe that Clark Kent is cowardly, but disbelieve that Superman
is pusillanimous, even though the names ``Clark Kent'' and ``Superman''
are co-referring, and the predicates ``cowardly'' and ``pusillanimous'' are
co-extensive. So much is well-known.
But when agents are omniscient, and perfectly rational in the econo-
mist's sense ± their preference orders are complete, continuous, and
transitive ± their set of beliefs will be closed under substitution of
coextensive terms, as will their wants. Complete information and transi-
tivity of preferences make it dif®cult to see immediately that expectations
and preferences are the intentional variables to which common sense
adverts under the names ``belief '' and ``desire.''
To the extent that economists view their discipline as methodologically
allied with the natural sciences, it is committed to a naturalistic meta-
physics for intentional states ± that is to the view that intentional states
determine choice in accordance with causal laws we can discover. Thus,
on their most obvious interpretation, as generalizations about the causes
of individual behavior, the fundamental assumptions of microeconomic
theory attribute the same psychological processes to all economic agents,
and explain differences among agents as the effects of different prefer-
ences and expectations. What is metaphysical here ± in the classical,
philosophical sense of the term ± is the commitment to causal interaction
between the mind and the body on the basis of causal laws. Assuming the
falsity of dualism, this commitment is tantamount to a strong type-type
materialist identity theory of the relation between brain states and
mental states.
There is another reason to identify these psychological commitments
184 Micro, macro, and markets

of microeconomic theory as metaphysical ± in a potentially derogatory

sense. It is well known that there is a long-term tradition of indifference
by economists to ®ndings and theories in psychology ± even favorable
ones ± which might bear on the truth, falsity, or tenability of these
commitments about the causes of individual human action. To the extent
a substantive commitment is recalcitrant to countervailing empirical
evidence, it would be a pardonable positivist criticism to call the
commitment ``merely metaphysical.''
There are reasons well known in the philosophy of psychology to be
pessimistic about prospects for uncovering any such psychological laws
of the sort economists need in any case. And this makes for what is
perhaps the deepest metaphysical mystery of microeconomic theory: the
conjunction of explanatory individualism and indifference to the truth
of its claims about individual psychology makes for a metaphysical
problem of major proportions for microeconomics. If the basic explana-
tory categories of economics are individual expectations and preferences
and yet economists have little interest in the further question of
whether their claims about preferences and expectations are reasonable,
approximately correct empirical hypotheses, then questions must arise
about the cognitive status of economics. To the extent that method-
ology has metaphysical rami®cations, answers to these questions must
have importance for anyone interested in the metaphysics of micro-
economics.
One traditional way in which economists have responded to this
problem is simply to adopt a crude behaviorism about choice, an
interpretation of the axioms of the theory known as ``revealed prefer-
ence.'' Another, more radical approach is to reject any interest in or
commitment to individual choice behavior as a subject for economics.
This stratagem completely deprives economics of commitment to
methodological individualism. But this is a small price to pay for the
freedom from any need to defend the truth of the theory's assumptions.
Still a third, more traditional approach is to treat the theory as an
idealization without ontological commitments as to what there is.

Behaviorist relief from metaphysics?

Ever since Paul Samuelson discovered operationalism (or was it B. F.

Skinner he discovered?) many economists have argued that their theory
makes no claims about the psychological processes that determine
choice. This at any rate seems the simplest account of how ``revealed
 The metaphysics of microeconomics 185

preference theory'' got started. Samuelson (and Houthekker)12 realized

that if preferences and expectations could between them determine
choice, then holding expectations constant, we could read preferences off
of choices. By offering an agent many pairs of commodities and allowing
the agent to select the object preferred in each case, a map of the agent's
preferences could be constructed, provided only the agent's choices
honored the assumption of weak transitivity: for every pair of commod-
ities a, b, either a is preferred to b or vice versa or the agent is perfectly
indifferent between a and b. From a map so constructed additional
intermediate choices can be predicted. If all we need for the theory of
consumer choice is that actual choices be transitive, then we can remain
agnostic about the causes of these choices; we need not embrace
extravagant claims about the psychological structure of tastes and
preferences. And, of course, by suitable employment of the von
Neumann-Morgenstern theory of uncertainty we can extend our beha-
viorism about preferences to a behaviorism about expectations as well:
they will turn out to be choices made as probabilities vary while
preference-behavior remains constant.
For a time this sort of behaviorism was popular among economists
who thought twice about the apparent falsehood of their theory of
consumer choice as a psychological theory. Revealed preference theory
also comports with the commitment to methodological individualism.
Both may be justi®ed on the epistemic ground that all we ever see are
individual economic agents; we can directly observe neither aggregates of
them nor the ``insides'' of any one of them.
The trouble with revealed preference theory is, of course, the trouble
with behaviorism in psychology and positivism in epistemology.
Behavior can only reveal preference if preference causes behavior. If
preference is just a synonym for behavior that satis®es intertemporal
transivity then not only is behavior often not preference ± say, whenever
what the non-behaviorist calls tastes change ± but the economist must
forswear the explanation of individual choice behavior. Moreover, by
declining to commit ittself to the causes of behavior ± ``internal''
expectations and preferences ± the economist surrenders the aim of
explaining why consumers and producers make the choices they do.
Additionally, they are committed to interpreting their own theory in a
way that fails to take its metaphysics seriously.
De®ned as the science which studies how scarce resources are distribu-

12 Paul Samuelson, Foundations of Economic Analysis (Cambridge, MA: M.I.T. Press,

1947); Hendrick S. Houthakker, ``Revaled Preference in the Utility Function,''
Economica, n.s., 17 (1950): 159±74.
186 Micro, macro, and markets

ted and ought ef®ciently to be distributed, economics has long been

committed to explaining individual choice, and of course, to explainin
aggregate economic data by showing that they are the consequences of
such choice. The only reason to surrender this goal is the realization that
it is unattainable. Here the distinction between the commitment of
economics and of economists becomes crucial. The theory makes a claim
which the theorists either ignore, or reinterpret. Economists have not so
much admitted that explaining individual choice is unattainable as they
have pretended that it never was a goal of theirs. Instead, many
economists have claimed that the goal of theory is simply to systematize
the data of markets, industries, and whole economies. For these purposes
it is convenient to couch their theory in the form of assumptions
ostensibly about the preferences and expectations of individuals, but
such claims are not to be taken seriously. Rather, following Friedman,13
they are to be understood as merely heuristic devices, convenient ®ctions
whose justi®cation consists only in their success in systematizing aggre-
gate data. Assume for the nonce that the theory of consumer behavior
and the theory of the ®rm actually do systematize aggregate data about
supply and demand in the market for consumer goods and factors of
production. Then the question immediately arises why these two theories
successfully systematize observations. It would be quite convenient for
economists if they could help themselves to an instrumentalist approach
to theories, which eschews metaphysical issues ± like whether the theory
is true or false. According to such an approach, the question of why a
theory is empirically adequate is an epistemically impertinent question
which no scientist is obliged to answer. Any attempt to answer it is an
excursus into metaphysics most pejoratively construed. On this view the
most plausible answer to the question of why a theory is empirically
adequate ± the claim that it is approximately true ± is a groundless
excursion into metaphysics since no empirical data are available to
warrant the truth of its claims about unobserved underlying mechanisms.
Regardless of the plausibility of this doctrine in the interpretation of, say,
quantum mechanics, it seems remarkably unavailing in any defense of
microeconomic theory. Quarks are observationally inaccessible, and for
this reason, instrumentalists have some epistemic basis for skepticism
about the truth ± approximate or otherwise ± of the ontological claims of
quantum physics. However, economic agents ± consumers, producers,
®rms, etc. are hardly beyond our observational powers. Accordingly, the
question remains: assuming a set of claims about consumers and

13 Milton Friedman, ``The Methodology of Positive Economics,'' in Essays in Positive

Economics (Chicago: Chicago University Press, 1953).
 The metaphysics of microeconomics 187

producers does systematize aggregate data, what explains this fact? If the
answer is the approximate truth of the claims about consumers and
producers, then there seems little reason beyond a commitment to an
overstrict empiricism to deny that the theory makes claims about
individual action as well as claims about the market.
But of course, we know that the claims of the theory of consumer
behavior and the theory of the ®rm are factually false and not even
approximately correct for the vast range of choices people actually make
over time in the real world. What is even more important, the claim that
the theories of consumer choice and the theory of the ®rm actually do
systematize aggregate data is, to say the least, highly controversial. And
between them these two facts about the empirical bearing of economic
theory make for the gravest of metaphysical problems: for they raise the
question of what economic theory is about, and what sort of an
enterprise it constitutes.
A theory which systematizes observations on the basis of assumptions
we know to be factually false is not by itself problematical. Newtonian
mechanics provides a ®ne example of such a theory. The theory embodies
a large number of idealizations and abstractions ± from frictionless
planes to point masses ± about whose non-existence there seems to be no
question. But the approximation of physical objects and processes to
these abstractions is close enough to make the theory which employs
them predictively unproblematical. The theory raises some metaphysical
problems just because it is on the one hand empirically adequate and on
the other metaphysically bootless. But while the problems the theory
raises are metaphysical, it raises fewer epistemological problems, just
because of its empirical adequacy.

A separate science?

In wrestling with what is essentially the same question of reconciling

economics' appeal to a set of generalizations about individual choice
with the predictive weakness of economic theory, Daniel Hausman has
argued that economists are committed to the existence of a ``separate
science.''14 Though he does not say so, this is I think the fundamental
metaphysical commitment of economists.
Every attempt to establish a science re¯ects a commitment to ``sepa-

14 Daniel Hausman, The Inexact and Separate Science of Economics (Cambridge:

Cambridge University Press, 1992), p. 222.
188 Micro, macro, and markets

rateness,'' that is, to the existence of a small number of tractable

independent variables that explain a signi®cant range of the data which is
the discipline's explananda. Successful sciences are those in which the
commitment to separateness has been vindicated, not by metaphysical
considerations, but by empirical successes. Economics still awaits the
unambiguous vindication of its separateness by this success. The long
period in which it has remained committed to separateness in the absence
of empirical success is a testimony to the strength of its metaphysical
commitments.
11 Ontological commitments of
evolutionary economics

JACK VROMEN

1 Introduction

Any attempt at offering a discussion of `an ontology of evolutionary

economics' is immediately fraught with dif®culties. First, evolutionary
economics is proliferating so rapidly that any attempt runs the danger of
being obsolete before it is published (see, for example, Hodgson 1997).
Second, at present evolutionary economics does not constitute an
unequivocally de®ned school or movement. It rather covers many
different approaches and standpoints (see, for example, Langlois and
Everett 1994 and Nelson 1995). And, ®nally, even if we con®ne our
attention to a few clearly circumscribed branches within evolutionary
economics and forget about the rest, it is not at all clear what exactly we
are looking for if we want to focus on their ontologies (or ontological
commitments).1 Given these dif®culties some preliminary remarks are in
order.
`Evolutionary economics' in this paper comprises two distinct
branches: Nelson and Winter's evolutionary theory and evolutionary
game theory. This already involves a choice that will affront many self-
confessed proponents of evolutionary economics. To many Nelson and
Winter's evolutionary theory (let alone evolutionary game theory) is
already too close to neoclassical economics to be called `evolutionary
economics' at all. Such proponents of evolutionary economics typically
plea for a radical break with equilibrium theorizing and stress phe-
nomena like nonlinear and self-organizing systems (see, for example,
Day and Chen 1993) and endogenously created novelty and ± conse-
quently ± the open-endedness of economic evolutionary processes (see,

1 For a study of ontological commitments of evolutionary economics with a focus quite

different from mine, see, e.g., Herrmann-Pillath (1996).

189
190 Micro, macro, and markets

for example, Saviotti and Metcalfe 1991). If this is what evolutionary

economics is all about, then this paper does not address evolutionary
economics at all. My justi®cation for concentrating on the two branches
mentioned above is that they both clearly associate themselves with
evolutionary theorizing and that they both attract much attention in the
economics profession.
What then could the ontology of these two branches of evolutionary
economics possibly be? As a ®rst attempt we could loosely follow a
Quinean strategy of identifying the ontological commitments of the
theories that are formulated within these branches: what entities must
exist, and what properties must they have for the theories in question to
be true? But following this strategy is not unproblematic. It has been
observed by some that it may be one of the peculiarities of economics
that what is stated explicitly in economic theories need not coincide with
what their advocates ± the economists ± believe to be true of the economy
(see, for example, Caldwell 1992, 1994; Rosenberg 2001). Rosenberg
argues that if economists as a matter of fact do entertain beliefs about
the actual economy different from what their theories explicitly state,
then we should distinguish between the metaphysical commitments of
economists and the metaphysical commitments of economics (Rosenberg
2001, pp. 176, 186). Apparently Rosenberg believes it is possible to
identify the metaphysical commitments of economics without looking at
beliefs of the economists using its theories, simply by taking the theories
at face value. By contrast, I shall argue that in identifying the metaphy-
sical (or rather ontological) commitments of economics the beliefs of
economists using its theories should be taken into account. If the beliefs
of economists do not match with what their theories express (or rather
appear to express), then we should investigate why they nevertheless
cling to their theories. Only after this investigation we can tell what
ontological commitments are involved in their theories.
The paper is organized as follows. I start with a discussion of
Rosenberg's views on the ontological presuppositions of neoclassical
economics. I then turn to the so-called selection arguments put forward
by Alchian and Friedman. The selection arguments, it will be argued, do
not only foreshadow subsequent theoretical developments in the two
branches of evolutionary economics mentioned above, but also reveal
ontological commitments of at least some neoclassical economists that
give rise to a reinterpretation of neoclassical economics. Next the basic
tenets of Nelson and Winter's evolutionary theory and evolutionary
game theory will be discussed. The paper concludes with some ®nal
re¯ections on the differences in ontology between the various versions of
neoclassical economics and of evolutionary economics.
 Ontological commitments of evolutionary economics 191

2 Prelude

My main reasons for taking Rosenberg's work as a starting point are the
following. First, Rosenberg has done some pioneering and provocative
work on the metaphysics and ontology of neoclassical microeconomics
(for a critical reaction, see, for example, Ross 1995). The second reason is
that Rosenberg has recently published several papers on the very same
exemplars of evolutionary arguments and theorizing that I address in this
paper (see Rosenberg 1992b, 2001). Since Rosenberg is an acknowledged
expert in the philosophy of biology, the views that he expresses in these
papers cannot be left unaddressed. Let us have a closer look then at what
Rosenberg has to say on both subjects ± ®rst, the metaphysics and
ontology of neoclassical microeconomics and, second, the exemplars of
evolutionary arguments and theorizing in economics.
As has been observed also by other commentators, Rosenberg's
criticism of neoclassical microeconomics proceeds from the presupposi-
tion that neoclassical microeconomics tries to explain individual be-
haviour as the effect of the joint interaction of beliefs and desires (see, for
example, Rosenberg 2001, p. 182). To be more precise, Rosenberg
identi®es expectations and preferences as `. . . cognates for beliefs and
desires' (ibid.) in neoclassical economics as the causal variables deter-
mining human behaviour. It will turn out to be expedient to split up
Rosenberg's presupposition about neoclassical economics (NE) in two
parts:
(NE1R) among other things, neoclassical microeconomics pur-
ports to explain the behaviour of individual human
agents;
(NE2R) neoclassical microeconomics pinpoints preferences and
expectations as the two (interacting) causal determi-
nants of the behaviour of individual human agents.2
Rosenberg's criticism of neoclassical economics mainly focuses on
(NE2R). According to Rosenberg, it is (NE2R) that makes neoclassical
microeconomics of a piece with folk psychology. And the problem with
folk psychology, Rosenberg argues, is that it does not allow for improve-
ments in prediction beyond the predictive powers that are re¯ected in our

2 Henceforth propositions with subscript 1 refer to the tasks or goals that some type (or
school) of economic theory is said to pursue. Propositions with subscript 2 denote what
some type (or school) of economic theory is said to take to be the actual causal
determinants of individual behaviour (i.e., behaviour of individual agents).
192 Micro, macro, and markets

everyday projections of people's actions. In Rosenberg's view the key

concepts of folk psychology ± desire and belief ± do not denote natural
kinds; they do not cut the world at its joints.
What is curious, however, is that Rosenberg himself expresses some
doubts as to whether (NE2R) is really representative of neoclassical
microeconomics. Rosenberg refers to revealed preference theory and von
Neumann-Morgenstern cardinal utility theory as self-conscious attempts
to purge neoclassical microeconomics from its af®liation with folk
psychology (Rosenberg 2001, pp. 184±6). As Rosenberg himself ob-
serves, adherence to these theories allows neoclassical economists to
remain agnostic about the causes or determinants of individual human
behaviour. Yet Rosenberg does not take the acceptance of these theories
by neoclassical economists to be indicating that there may be something
wrong with (NE2R). He does not seriously consider the possibility that
the popularity of these theories among neoclassical economists calls for a
reinterpretation of neoclassical microeconomics, or at least for a quali®-
cation of (NE2R).3 No, for Rosenberg (NE2R) captures an invariant
essential of the neoclassical programme. Given this ®xation of the
neoclassical programme, revealed preference theory and von Neumann-
Morgenstern cardinal utility theory cannot appear to Rosenberg other
than as ingenious attempts of neoclassical economists to sidestep the
vexing problems inherent in the neoclassical programme.
Rosenberg argues that these attempts fail. An instance of individual
human behaviour can only be explained in terms of the preferences and
expectations that are identi®ed in revealed preference theory and von
Neumann-Morgenstern cardinal utility theory, he argues, if this instance
is really caused by these preferences and expectations (see Rosenberg
2001, pp. 184±6). Here (NE1R) comes in. Rosenberg's argument here
clearly presupposes that individual human behaviour is an explanandum
of neoclassical microeconomics. But, again, it is noteworthy that Rosen-
berg himself casts doubt on the truth of this presupposition.4 Rosenberg
appears to be aware of remarks of neoclassical economists such as
Friedman, stating that the behavioural assumptions are not meant to
explain individual behaviour, but to systematize aggregate data (of
industries, markets and whole economies; Rosenberg 2001, p. 186). As in
the case of (NE2R), this acknowledgement does not seem to affect
Rosenberg's understanding of neoclassical economics. Rosenberg does

3 If we are to believe Binmore (1994), then nowadays all neoclassical economists embrace
revealed preference theory and von Neumann-Morgenstern cardinal utility theory (and
have thus left the old `Benthamite', folk psychological interpretation behind).
4 And so do Kincaid (2001) and Ross (2001), for example.
 Ontological commitments of evolutionary economics 193

not call (NE1R) into question. Again, what he does instead is to argue
that this escape route does not bring neoclassical economists relief either.
For, Rosenberg asks himself, how could theories of individual behaviour
systematize aggregate data without being at least approximately true
(Rosenberg 2001, pp. 186±7)?5
At this point Rosenberg brings in a new element. The problems with
explaining individual behaviour in terms of preferences and expectations
are not just of a metaphysical kind (the problem that preferences and
expectations are not natural kinds), Rosenberg argues, but also of an
empirical kind: `. . . we know that the claims of the theory of consumer
behaviour and the theory of the ®rm are factually false and not even
approximately correct for the vast range of choices people actually make
over time in the real world' (Rosenberg 2001, p. 187). Now this seems to
be in need of some further clari®cation. On the basis of the above
discussion, one could expect that Rosenberg believes that preferences
and expectations (as cognates for desires and beliefs) should play no role
in explanations of individual behaviour. After all, in Rosenberg's view its
dedication to these folk psychological derivates prevents neoclassical
economics from achieving major improvements in predictive power.
Given that in Rosenberg's mind neoclassical economics is bound to
explain individual behaviour (even if it is mainly interested in system-
atising aggregate data), it seems to follow that economists are well-
advised to dispense with preferences and expectations altogether (and
thus to abandon the neoclassical programme).
This is not the conclusion that Rosenberg draws, however. Rosenberg
seems to hold the view that consumers and producers do have desires
and beliefs and that these mental states are consequential for the actions
that consumers and producers undertake. What is more, he argues that
desires and beliefs are indispensable in explaining individual behaviour
(see Rosenberg 1992a, p. 151). What Rosenberg takes to be false in the
theories of the ®rm and of consumer behaviour are rather the additional
claims made in the theories that consumers and producers have the best
possible beliefs (rational expectations) and that they choose the best
possible action given their desires (preferences) and their beliefs. In other
words, Rosenberg takes it to be beyond doubt that real consumers and
producers are less rational than the ones that are portrayed in the
neoclassical theories of consumer behaviour and of the ®rm (see, for
example, Rosenberg 1992a, p. 116).

5 See also Rosenberg (1992a), p. 129, where he argues that it is improbable that predictive
improvements at the aggregate level can exceed predictive improvements at the individual
level.
194 Micro, macro, and markets

Thus it must be concluded, I think, that in Rosenberg's eyes the

situation for economics is quite hopeless. According to Rosenberg,
economics is wedded to explain individual behaviour. As we have seen,
he thinks that economics is committed to this even if its sole purpose
were to be to systematize aggregate data. Furthermore, Rosenberg holds
that explaining individual behaviour necessarily invokes desires and
beliefs. Hence, the af®nity between economics and folk psychology is not
incidental but fundamental and inescapable. Maybe economic theories
could be made more realistic by relaxing their rationality assumptions.
But the hope of achieving real predictive improvements beyond our
everyday capacity to predict is forlorn, Rosenberg believes, because
preferences and beliefs do not denote natural kinds. Is Rosenberg
committed to the view then that economics is destined never to attain
scienti®c respectability?

3 Neoclassical economics and the selection arguments

Rosenberg (1992b) regards the selection arguments put forward by

Friedman (1953) and Alchian (1950) as just another desperate attempt to
defend neoclassical economics.6 Rosenberg contends that the selection
arguments are attractive to neoclassical economists because they appear
to offer a way out of the acknowledged empirical inadequacy of the
theories of consumer and ®rm behaviour. Friedman's and Alchian's
arguments hold out the promise, Rosenberg argues, that the neoclassical
claim about ®rm behaviour can be vindicated even though it is granted
that not all ®rms are rationally informed pro®t-maximizers. But, again,
Rosenberg argues that the hope is forlorn; this line of defence is also
bound to fail.
There certainly are several things to be said in favour of Rosenberg's
treatment of the selection arguments. Rosenberg is right that selection
arguments did pop up in economics in order to defend neoclassical
economics. And they were indeed devised to get around the problem of
the apparent falsity of the neoclassical theory of the ®rm. These merits
notwithstanding, my main problem with Rosenberg's treatment of the
selection arguments is again that he views them as mere ad hoc escape
manoeuvres. Rosenberg simply refuses to consider the possibility ser-
iously that the selection arguments reveal more than just the wish of
neoclassical economists to evade dif®cult problems with their theories.

6 See also Rosenberg (1992a), pp. 183±93, and Rosenberg (1994).

 Ontological commitments of evolutionary economics 195

By contrast, I do not want to rule out the possibility in advance that by

studying the selection arguments carefully we can learn something about
neoclassical economics itself. To be more precise, I shall examine the
possibility that the arguments reveal beliefs of (at least some) neoclassical
economists about the workings of competitive markets; beliefs that cast a
different light on the nature of their theoretical enterprise than Rosen-
berg does. But before I can come to that I ®rst have to put the selection
arguments in their historical context.
In the 1930s and 1940s the empirical adequacy of the neoclassical
assumption of pro®t maximization was challenged from within the
economics profession, notably by the Oxford Research Group in the UK
and by Lester in the USA.7 In the ensuing marginalism controversy these
so-called antimarginalists held that the neoclassical (marginalist) theory
of the ®rm had to be discarded. Questionnaires and subsequent inter-
views clearly pointed out, the antimarginalists claimed, that entrepre-
neurs do not base their decisions on the type of marginalist deliberations
that neoclassical theory ascribes to them. What they found instead was
that entrepreneurs typically use rules of thumb, such as the `full cost'
pricing rule in setting their prices. The antimarginalists took these results
to be ample evidence that the neoclassical theory of the ®rm was
unrealistic. They concluded from this that this neoclassical theory was to
be rejected.
The neoclassical theory of the ®rm was defended by Machlup (1946).
Machlup's main strategy was to show that the answers to the question-
naires given by the entrepreneurs can be rephrased in terms of the
marginalist principles underlying the neoclassical theory of the ®rm.
What Machlup pointed out in particular was that behaviour that is
routinized rule-following from the (®rst person's) point of view of the
entrepreneurs themselves, can be interpreted as maximizing behaviour
from a (third person's) theorist's point of view. For this purpose
Machlup drew an analogy with an experienced automobile driver. When
making the decision whether or not to overtake a truck that is proceeding
ahead of her at slower speed, Machlup argues, an experienced automo-
bile driver typically `sizes-up' the relevant factors in a routine way. Yet if
we want to develop a `theory of overtaking', Machlup contends, we
would have to break the decision down into its factors and to state how
changes in any of the factors would affect the driver's actions. The
message is clear: the antimarginalists may well be right that just as with
the experienced driver's behaviour most of the behaviour of entrepre-
neurs is routinized. But they are wrong in concluding from this that a

7 See Vromen (1995) for a detailed discussion.

196 Micro, macro, and markets

theoretical account of entrepreneurial behaviour in terms of conscious

deliberations and calculations is incorrect. Machlup seems to be con-
vinced even that an `explanation of an action must often include steps of
reasoning which the acting individual himself does not consciously
perform (because the action has become routine)' (Machlup 1946,
p. 535).
Alchian's classic (1950) paper was prompted by Alchian's feeling that
both the antimarginalist criticisms and Machlup's defence were mis-
guided.8 In Alchian's view, both the antimarginalists and Machlup
proceed from the mistaken assumption that the neoclassical theory of the
®rm is meant to be a theory of the behaviour of individual ®rms. They
both confuse `axioms' with `theorems', Alchian argues: the axioms of the
neoclassical theory of the ®rm pertain to the behaviour of ®rms, but its
theorems are about aggregate industry (or market) behaviour. To be
more precise, the theorems of neoclassical economics are generic or
pattern predictions about changes in aggregate industry variables. A
paradigmatic example is the pattern prediction that, ceteris paribus, the
aggregate capital/labour ratio in an industry increases if (real) wage rates
rise (see Alchian 1950, p. 17; see also Lester 1946, p. 65).
Now what Alchian wants to point out is that even if the antimarginal-
ists were right that individual ®rms do not behave as stated in the
neoclassical axioms, neoclassical pattern predictions still hold.9 Or, to
use Alchian's own terms, even if the axioms are false, they can neverthe-
less be used to derive the neoclassical theorems. The axioms still are
useful tools for analysing industry behaviour, Alchian argues, because
`market selection' is going on: `impersonal market forces' see to it that
®rms in the industry that happen to make positive pro®ts grow at the
expense of ®rms that happen to suffer losses. In the paradigmatic
example mentioned above, after (real) wage rates have risen those ®rms
in the industry will have a higher probability of survival that have
relatively lower costs because of their relatively higher capital/labour
ratio. As a consequence, the average capital/labour ratio in the industry
increases.
Roughly the same idea can be found in Friedman (1953). Friedman
also argues that even if the antimarginalists' empirical ®ndings were to be
taken seriously, the neoclassical theory of the ®rm can be maintained
because there are processes of `natural selection' going on in markets that
8 See Alchian in Zerbe (1982), p. 149.
9 Alchian does not argue that we can a priori know that the neoclassical theorems hold in
all cases. In each and every case the theorems have to be tested empirically. But he seems
to be quite con®dent that, whatever the results of these tests, the neoclassical axioms can
be retained as a useful starting point in analysing industry behaviour.
 Ontological commitments of evolutionary economics 197

lead to the same results as the ones predicted by the neoclassical theory.
At ®rst the relevant passages in Friedman (1953) might suggest that
Friedman defends the neoclassical theory as a theory about the be-
haviour of individual ®rms. For he argues that unless the behaviour of
®rms approximates behaviour that is consistent with the assumption of
pro®t maximization (or the `maximization-of-returns hypothesis' as
Friedman prefers to call it), the ®rms will not remain in business for long
(p. 22). Friedman claims that the assumption of pro®t maximization
appropriately summarizes the conditions for survival.10 But from the
remainder of the paper (and from other papers written by Friedman) it
becomes clear that Friedman just like Alchian takes the neoclassical
theory of the ®rm to be a theory of aggregate industry behaviour (see
also Hammond 1991).
In the same passage, Friedman remarks that the argumentative
support thus given to the maximization-of-assumptions hypothesis is in
part similar to the argumentative support that Savage and he have given
for the expert billiard player hypothesis.11 The expert billiard player
hypothesis states that before making his shots, an expert billiard player
makes lightning calculations and solves a set of mathematical equations,
resulting in the identi®cation of the uniquely best shot. Friedman and
Savage (1948) bet that this hypothesis yields excellent predictions about
the expert billiard player's behaviour even though presumably no expert
billiard player really goes through these calculations. Expert billiard
players are likely to rely on their skill and experience when playing the
game instead of executing complicated calculations.
Friedman and Savage's discussion of the expert billiard player hypoth-
esis is reminiscent of Machlup's experienced automobile driver analogy.
Like Machlup, Friedman and Savage want to illustrate the point that the
question whether a hypothesis is descriptively unrealistic of the decision
processes individuals go through is irrelevant when assessing the useful-
ness of the hypothesis. The assessment of the usefulness of the hypothesis
should be based on other grounds.12 And acceptance of a hypothesis
does not mean commitment to the belief that the hypothesis is descrip-
tively realistic. The hypothesis that Friedman and Savage (1948, p. 281)
develop and defend in the paper is based on von Neumann-Morgenstern

10 In Vromen (1995), I call this Friedman's intermediate claim, stressing that his ultimate
claim is much more akin to Alchian's claim. Related to this, I also argue there that
`conditions' are to be understood as suf®cient and not as necessary conditions.
11 Friedman himself speaks of `evidence' instead of `argumentative support'. See Vromen
(1995), pp. 29±34, for a detailed discussion.
12 Of course, the ground that Friedman and Savage think is to be decisive ± whether the
hypothesis yields accurate predictions ± is different from Machlup's.
198 Micro, macro, and markets

cardinal utility theory and is meant to rationalize observable behaviour

by attributing utility functions to individuals. No claim is made that
individuals really have these functions in their minds nor that they
actually are engaged in attempts at maximization. The claim is only that
they behave as if they actually maximized the functions.
What characterization could we give of the type of neoclassical
economics that Alchian and Friedman are defending with their selection
arguments? First, it seems that Rosenberg's presupposition (NE1R) has to
be replaced by something like:
(NE1A&F) neoclassical economics purports to predict industry
behaviour.13
Furthermore, it seems that especially Friedman prefers to remain
agnostic about the `immediate determinants' of consumer and producer
behaviour. This suggests that Rosenberg's (NE2R) has to give way to
something like:
(NE2A&F) neoclassical economics does not pinpoint any speci®c
mental states of individual agents (and, indeed, does
not pinpoint any causes) as the actual determinants
of their behaviour.14
Thus what makes Alchian and Friedman feel con®dent that neoclas-
sical economics is on the right track in predicting industry behaviour is
not their belief that individuals are actually conducting the calculations
that neoclassical economists are ascribing to them. What is behind their
con®dence is rather their belief that `impersonal market forces' lead to
the very same tendencies in industry behaviour that would accrue if
individuals were to do literally what neoclassical economists ascribe to
them. We can unpack this belief as follows:15
(NE3A&F) `market selection' is a dominant force in shaping
industry behaviour
and

13 Later on, Machlup (1967) also subscribes to this view.

14 Clark (1997) and Satz and Ferejohn (1994) make the similar point that neoclassical
economics (and rational choice theory in general) need not, and indeed should not, be
understood as entailing speci®c claims about individual psychology.
15 Henceforth, propositions with subscript 3 refer to the forces (or mechanisms or causes)
that some type (or school) of economic theory is said to believe to be the forces shaping
industry (or market) behaviour. Propositions with subscript 4 relate to the implications
at the industry or market level that some type (or school) of economic theory is said to
associate with the workings of these forces.
 Ontological commitments of evolutionary economics 199

(NE4A&F) `market selection' produces the same tendencies in

industry behaviour as fully informed and fully ra-
tional ®rms would do.16
In his comments on the selection arguments, it does not seem to dawn
on Rosenberg that (NE1R) and (NE2R) are called into question. Rosen-
berg is only concerned with showing that selection arguments do not give
comfort to neoclassical economics as he interprets it: as a theory about
individual behaviour. In particular, Rosenberg does not show any
awareness that if the selection arguments are taken seriously, they imply
both (NE1A&F), that the neoclassical theory of the ®rm is about industry
behaviour, and (NE2A&F) (NE3A&F) and (NE4A&F), that the tendencies
that the neoclassical theory predicts at this aggregate level are not
supposed to be brought about by rationally informed maximizing ®rms,
but by `market selection'.17 He fails to see that (NE3A&F) and (NE4A&F)
together provide an answer to his question: `. . . assuming a set of claims
about consumers and producers does systematize aggregate data, what
explains this fact?' (Rosenberg 2001, pp. 186±7).

4 Nelson and Winter's evolutionary theory

No other book is so much referred to in `evolutionary economics' as

Nelson and Winter's An Evolutionary Theory of Economic Change (1982).
It has become the classic in evolutionary economics. Although not all
proponents of evolutionary economics agree with everything in Nelson
and Winter's book, it still is the main reference point and a major source
of inspiration. In this section I want to point out that in turn Nelson and
Winter have been inspired by Alchian's and Friedman's selection argu-
ments. Much of what Nelson and Winter are doing in their book can be
understood as a further elaboration and examination of what was
already advanced in the selection arguments.

16 To repeat, Alchian and Friedman do not hold that we can know a priori that the
neoclassical tendencies hold. Whereas testing of the neoclassical assumptions (or
`axioms' as Alchian calls them) is misguided, Alchian and Friedman contend, testing of
its implications (or `theorems' as Alchian calls them) remains necessary.
17 Rosenberg himself does not seem to reject the idea that there is something like `natural
selection' going on in competitive markets out of hand. But what he does reject out of
hand is that such an idea underlies the acceptance of neoclassical theory by at least some
economists.
200 Micro, macro, and markets

Rosenberg on Nelson and Winter

But let us ®rst have a look on what Rosenberg (1994) has to say on
Nelson and Winter's evolutionary theory. According to Rosenberg,
Nelson and Winter are inspired by Darwinian evolutionary theory in
biology. Indeed, his comments on their theoretical endeavour is based on
the presupposition that Nelson and Winter aim `to show how economic
processes instantiate Darwinian natural selection' (Rosenberg 1994,
p. 404). Rosenberg argues that Nelson and Winter's attempt fails to
develop an economic theory that is strictly analogous to Darwinian
evolutionary theory. In his view Nelson and Winter's attempt is likely to
surrender to Lamarckism. As in Lamarckism almost everything is
possible (`anything can evolve into anything by any means', p. 405),
Rosenberg argues, this would shed no special light on economic
processes. Rosenberg then develops the outlines of what could be a fully-
¯edged Darwinian theory in economics. He winds up with something like
a Simon and Cyert and March type of organization theory that, as he
contends, only few economists are prepared to accept.
It is hard to avoid the impression that Rosenberg's comments are
based on a cursory reading of Nelson and Winter's book. Let it suf®ce to
quote the following passage from Nelson and Winter's book at length:
[Relatedly,] our theory is unabashedly Lamarckian: it contemplates both the
`inheritance' of acquired characteristics and the timely appearance of variation
under the stimulus of adversity.
We emphatically disavow any intention to pursue biological analogies for their
own sake, or even for the sake of progress toward an abstract, higher-level
evolutionary theory that would incorporate a range of existing theories. We are
pleased to exploit any idea from biology that seems helpful in the understanding
of economic problems, but we are equally prepared to pass over anything that
seems awkward, or to modify accepted biological theories radically in the interest
of getting better economic theory (witness our espousal of Lamarckianism).
(Nelson and Winter 1982, p. 11)
Two things are made immediately clear in this passage. First, in contrast
to what Rosenberg asserts, Nelson and Winter are not looking for an
economic evolutionary theory that is strictly analogous to Darwinian
evolutionary theory in biology. They are looking for `better economic
theory', using only those insights and parts of biological theories that are
conducive to this theoretical purpose. Second, and again contrary to
what Rosenberg claims, Nelson and Winter do not surrender to La-
marckism. They wholeheartedly embrace Lamarckism right from the
start, because they think it provides a better starting point for `better
 Ontological commitments of evolutionary economics 201

economic theory' than Darwinism. The only valuable idea in Rosenberg's

comments is, I think, that as such the notion of Lamarckism does not tell
us very much about the `nuts and bolts' of economic processes. The
obvious conclusion to draw from this idea, I venture, would be to study
Nelson and Winter's `Lamarckian' economic theory in a more detailed
and careful way. This is not what Rosenberg does, however. In this
section I shall redress this omission. Ironically, the upshot of my
discussion will be that a Simon and Cyert and March type of organiza-
tion theory that Rosenberg thinks would be a promising candidate for an
unadulterated Darwinian economic theory precisely provides the key to
the understanding of the alleged `Lamarckian' element in Nelson and
Winter's economic theory.

Nelson and Winter's evolutionary theory as an explication of

`appreciative' orthodox theory

Nelson and Winter call the selection arguments of Friedman and

especially Alchian `direct intellectual antecedents' of their own book
(p. 41). In particular, what they take over from these arguments is the ex
post viewpoint: when there is uncertainty, the success or failure of an
action cannot be determined ex ante. Success or failure can only be
ascertained ex post then, after the action has passed the `market test'. But
whereas the modus operandi of market selection remained somewhat
implicit in the selection arguments, Nelson and Winter endeavour to
make it explicit. Market selection favours ®rms that make positive pro®ts
over those that make negative pro®ts (suffer losses), they argue, because
making positive pro®ts means acquiring additional means that the ®rms
then can use to expand, whereas suffering losses means diminishing
means and, hence, compulsory contracting. In their models, the shifts in
market shares that may follow from this are called selection effects.18
Nelson and Winter believe that this idea of market selection reveals a
genuine insight into how markets actually function. They subscribe to
the view that `natural selection' is an important force in moulding market
and industry behaviour. Nelson and Winter also agree with Alchian cum
suis that this view, though it is most of the time not made explicit
anymore, has always been with economic theory. Indeed, they argue that

18 With this interpretation Nelson and Winter elaborate upon Becker's (1962) interpret-
ation of (and adherence to) the selection arguments. In Becker's view, the selection
arguments show that just like households, ®rms face `budget constraints'.
202 Micro, macro, and markets

the central place given to the force of market selection in their evolu-
tionary theory is consonant with the emphasis put on notions like
competition, competitive pressure and competitive forces in orthodox
economic theory (p. 32). And like orthodox economic theory, Nelson
and Winter contend, the focus of evolutionary economic theory is on
analysis of the larger systems (like industries and markets) and not on the
individual actors (like ®rms or the individual stakeholders of ®rms;
p. 51): `our theory is a theory about market processes' (p. 41).19
It is here that Nelson and Winter's distinction between formal and
appreciative theory becomes relevant (pp. 9, 32±3, and 46±8; see also
Nelson 1995, pp. 50±3).20 Formal theory is characterized by heroic
abstractions. It selects a few variables, mechanisms or relationships,
formalizes them and traces their consequences. Tractability and elegance
of models (mostly, but not necessarily mathematical models) are valued
highly in formal theory. Formal theory is developed, re®ned and
modi®ed in theoretical and foundational research. Appreciative theory,
on the other hand, is used mostly in applied and policy-oriented research.
Appreciative theory refers to a way of looking at phenomena, or a
framework of appreciation that a group of theorists share with one
another (p. 46). The intuitions (p. 33) and beliefs (see Nelson 1995, p. 50)
of theorists about the phenomena they are studying are re¯ected in their
appreciative theory.
With its characteristic commitments to the assumptions of individual
maximizing behaviour and aggregate equilibrium behaviour formal
orthodox economic theory lives up to the demands of delivering rigorous,
tractable and elegant models. These models are used to conduct static
and comparative static analysis. According to Nelson and Winter
appreciative orthodox economic theory entertains the very same complex
and `messy' notions and ideas that are put centre stage in their own
evolutionary theory: evolution, change, dynamic processes, adaptation,
learning and search (see also Nelson 1995, p. 49). It is clear that these
biology-oriented notions do not ®t readily with formal orthodox theory's
assumptions.
Orthodox economists do not seem to be very bothered about this
tension. Why not? One reason is provided by the selection arguments.

19 See for similar views, Mayr (1982) ± in the realm of biology ± and Metcalfe (1989) ± in
the realm of economics.
20 Although I believe that this distinction is suggestive and (at least potentially) contains
valuable insights, I can only agree with Nelson and Winter (1982, p. 9) that their
treatment leaves a lot wanting. Many different features seem to be lumped together in
these notions. In what follows I shall concentrate on those features that seem to be
relevant for the purposes at hand.
 Ontological commitments of evolutionary economics 203

The selection arguments attempt to reconcile formal and appreciative

orthodox theory. They do so by arguing that the central claims of formal
orthodox theory are backed up by appreciative orthodox theory. This
attempt at a reconciliation presupposes that formal orthodox theory is
not taken at face value. Indeed, the proponents of the selection argu-
ments urge us to reinterpret the claims of formal orthodox theory
considerably. The neoclassical theory of the ®rm, they tell us, is not
about individual ®rm behaviour but about aggregate industry behaviour.
The neoclassical theory does not claim that individual ®rms actually
solve the maximization problems that it ascribes to them. It is said to
claim at most that the behaviour of ®rms surviving market selection
coincides with that of hypothesized ®rms solving the maximization
problems. Neither does the theory claim that industries are in equilibrium
all of the time. What is claimed is that industries tend to move in the
direction of equilibria.
This has consequences for our attempts to identify the ontological
commitments of economic theory. For if we reinterpret economic theory
along the lines outlined above, economic theory is not ontologically
committed to grant the existence of mental states being the `immediate
causal determinants' of individual behaviour. Or, to put it in the terms
proposed by Nelson and Winter, for formal orthodox theory to be true
the behaviour of individual agents need not be caused by their mental
states. What orthodox theory is ontologically committed to is captured
in its appreciative theory: the existence of market selection which is held
to be causally responsible for industry behaviour. Thus interpreted, it is
not so much certain properties attributed to individual agents as certain
properties attributed to (competitive) markets that are believed to be
crucial in orthodox theory.
This still leaves us with a question. If orthodox economists really
believe that what is actually going on in economies is quite something
different from what is expressed in their formal theories, why then do
they cling to their formal theory? Why do not they formalize their beliefs
and intuitions as they are revealed in their appreciative theory and
present that theory as their theory? According to Nelson and Winter,
orthodox economists' reasons for refusing to go this way have changed
over time. Marshall's primary reason for developing formal static
orthodox theory further, instead of trying to provide the ground work
for a truly dynamic evolutionary theory (even though he famously saw
the `Mecca' of economics lying in `economic biology'), seems to have
been that at that time no techniques were available to build tractable
dynamic models. But now, Nelson and Winter contend, with all kinds of
techniques at their disposal to build such models, the resistance among
204 Micro, macro, and markets

orthodox economists must be explained in terms of their predilection for

simple and elegant models.
If Nelson and Winter are right in this (and I think there is much to say
in favour of it), then it is primarily aesthetic and not ontological reasons
that prevent orthodox economists from leaving their formal theory. This
all the more supports the idea that the ontological commitments of
orthodox economic theory cannot be read off directly from its formal
theory. Its appreciative theory should also be taken into account.
In Nelson and Winter's view, appreciative orthodox theory coincides
to a large extent with their own appreciative evolutionary theory. They
see the main difference between orthodox theory and their own evolu-
tionary theory lying in their respective formal theories. Orthodox econo-
mists apparently take their belief that their formal theory and their
appreciative theory are reconcilable with one another ( joined with their
aesthetic predilection) to be a suf®cient reason to cling to their formal
theory. By contrast, Nelson and Winter insist on it that the core ideas in
appreciative theory should be re¯ected in their formal theory by explicitly
modelling them. Borrowing MaÈki's (1994) terminology, orthodox appre-
ciative theorists are methodological instrumentalists in this respect ± they
accept a (formal) theory despite the fact that they do not take the causal
mechanisms that are described in the theory to be literally true ± while
Nelson and Winter are methodological realists ± they believe that a
theory is only acceptable if it speci®es the causal mechanisms that are
thought to be actually operating in reality. In this, Nelson and Winter
follow a suggestion made by Koopmans about the role that `natural
selection' plays in Alchian's and Friedman's arguments: `if this is the
basis for our belief in pro®t maximization, then we should postulate that
basis itself and not the pro®t maximization which it implies in certain
circumstances' (Koopmans 1957, pp. 140±1).

Further ontological commitments of Nelson and Winter's evolutionary

theory

`Market selection' is modelled explicitly by Nelson and Winter. One of

the objectives they pursue with this is to examine systematically under
what conditions evolutionary processes governed by `market selection'
produce neoclassical results.21 In other words, they do not believe
21 In fact, many different models appear in Nelson and Winter's book that serve many
different purposes. One of these purposes, that has been taken up later also by Dosi et al.
(1988), is to account for technical innovation and change.
 Ontological commitments of evolutionary economics 205

(NE4A&F) offhand. In fact, from Winter (1964) onwards especially Winter

has gone at great length to demonstrate in a rigorous formal way that
`market selection' produces neoclassical results only under quite restric-
tive conditions.22 From Winter (1964) also stems the assumption that
®rms follow ®xed, rigid behavioural rules. Winter seems to have intro-
duced this assumption merely to be able to derive de®nite results from his
model. But in Nelson and Winter's (1982) evolutionary theory this
assumption is assigned the same grounding status as the assumption that
there is `market selection' going on.
In Nelson and Winter's view, understanding ®rm behaviour means
identifying the routines that ®rms follow. Most of the informal discus-
sions in their book are devoted to bring home this view. Nelson and
Winter develop their notion of routine at the ®rm level in analogy with
the notion of skills at the level of individual persons. They argue that
`routine' closely resembles Polanyi's notion of tacit knowledge and Ryle's
notion of `know how'. Routines are enabling and restricting at one and
the same time ± they both determine what ®rms can and cannot do.
Nelson and Winter contrast their view of ®rms displaying (quasi-)
automatic routine behaviour with what is stated in formal orthodox
theory: ®rms are engaged in conscious and deliberate choice. They stress
in particular that in their view ®rms are not able to respond in the
prompt and ¯exible way to environmental changes that formal orthodox
theory envisages. A ®rm's extant routines severely restrict the ®rm's
opportunities to change its behaviour in response to environmental
changes. If changes in a ®rm's behaviour are possible at all, they at the
very least take time.
Nelson and Winter argue that they are much closer to orthodox
appreciative theory also in this respect. They take Machlup's experienced
automobile driver and Friedman's (and Savage's) expert billiard player
to illustrate that Machlup and Friedman also hold that much of actual
behaviour is routine behaviour.23 Nelson and Winter's catch phrase
`routines as genes' clearly indicate that Nelson and Winter believe that
routines instruct ®rm behaviour in much the same way as the genes of an
individual organism instruct its behaviour.
Still another important element of Nelson and Winter's evolutionary
theory has to be mentioned. As already alluded to above, Nelson and

22 In some cases (when dealing with comparative statics and growth theory), however,
Nelson and Winter demand from their own models that they are able to reproduce
neoclassical results.
23 In the previous section we have seen, however, that Friedman does not seem to be
wedded to any particular view on the `immediate determinants' of ®rm behaviour.
206 Micro, macro, and markets

Winter do not hold that routine ®rm behaviour necessarily implies ®rm
behaviour that is completely unresponsive to changing circumstances. In
the ®rst place, a routine may be conditional in the sense that the routine
speci®es different lines of behaviour under different conditions. Changes
in ®rm behaviour that result from such conditional routines when
conditions change are placed under the rubric of `along-the-rule effects'
in some of Nelson and Winter's models. And, second, Nelson and Winter
assume that if, for whatever reason, the prevailing routines of ®rms cease
to yield satisfactory results, ®rms engage in search activities for better
routines. Changes in ®rm behaviour resulting from such search efforts
are called `search effects' by Nelson and Winter.

`Lamarckianism' in Nelson and Winter's theory?

Here we encounter a supposedly Lamarckian element in Nelson and

Winter's evolutionary theory: the timely appearance of variation under
the stimulus of adversity. New routines are likely to enter the industry if
the old ones do not work well any more. And if the new routines turn out
to do reasonably well, Nelson and Winter do not only assume that the
®rms who have ®rst found the routines stick to them, but also that they
will be imitated by other ®rms in the industry. The new routines can thus
spread over the population. This supposedly stands for the second
Lamarckian element in Nelson and Winter's theory: the `inheritance' (or
better transmission) of acquired characteristics. Given the ambiguities
that surround the use of the notion Lamarckism, however, what does
putting this label of Lamarckism on Nelson and Winter's theory add to
our understanding of their theory?24 Nothing, I venture.25 It is more
instructive to follow Nelson and Winter's own modelling efforts in
treating `search effects' separate from selection effects and along-the-rule
effects.
Nelson and Winter assume that Simon's notion of satis®cing provides

24 There seem to be more strict and more loose interpretations of Lamarckism around. Hull
(1982), for example, adopts the strict interpretation that Lamarckism is committed to the
claim that changes in phenotype cause changes in genotypes.
25 Note that my problem with the label Lamarckism is quite different from Rosenberg's.
Rosenberg's problem seems to be that because Nelson and Winter revert to Lamarckism,
their evolutionary theory lacks de®nite economic meaning. In contrast, my problem is
that without putting the label of Lamarckism on it Nelson and Winter's theory already
has de®nite economic meaning (if you take the trouble of looking carefully at it), and
that using the label is more likely to create confusion than clari®cation.
 Ontological commitments of evolutionary economics 207

the key to the understanding of the processes in which search effects are
produced. Firms are assumed to engage in search only if their extant
routines yield pro®ts that lie below some aspiration level. Search
terminates as soon as some new routine is found that leads to pro®ts
exceeding the aspiration level.26 At ®rst it may seem that Nelson and
Winter run into trouble here: how is the assumption that ®rms are
engaged in satis®cing behaviour ± that seems to be clearly conscious and
deliberate goal-directed behaviour ± to be reconciled with Nelson and
Winter's basic contention that ®rm behaviour is (quasi-) automatic
routine behaviour? Nelson and Winter's way out of this is to assume ®rst
that engaging in search behaviour need not imply that conscious
decisions are made. Whether ®rms engage in search behaviour or not
may itself be routine-guided: dissatisfactory results may `trigger' search
behaviour in an (quasi-) automatic way. And, second, Nelson and
Winter argue that search behaviour itself is also routine behaviour;
search behaviour is assumed to be guided by second-order search
routines. In this they follow Cyert and March (1963) and, again, Simon
with his notions of procedural rationality and selective search.27 These
are precisely the authors who and the key ideas which Rosenberg
recommends to Nelson and Winter for making their theory a genuinely
Darwinian evolutionary theory!
We are now in the position to complete our picture of the basic
explanatory structure of Nelson and Winter's economic evolutionary
theory. The main purpose that Nelson and Winter ascribe to evolu-
tionary economics (EE) can be formulated as follows:
(EE1N&W) Economic evolutionary theory purports to explain
changes in aggregate industry behaviour.
As the actual determinants of ®rm behaviour are routines in Nelson and
Winter's view, we can state for a start:
(EE2N&W) Economic evolutionary theory takes ®rm behaviour
to be (quasi-) automatically instructed by their rigid
routines.
To repeat, it is important to see that (EE2N&W) can account for
changes in ®rm behaviour. Rigid routine following by ®rms can neverthe-
less lead to changes in their behaviour. It can do so in two different ways.
26 To be fair, it must be added that Simon's notion of satis®cing also allows for adjusting
aspiration levels in an upward or downward direction.
27 Indeed, it has been noted by Simon himself that the kind of learning that his notions of
procedural rationality and selective search imply closely resemble Darwinian natural
selection (see Simon 1983, p. 40).
208 Micro, macro, and markets

First ®rms may have relatively sophisticated conditional routines that

may lead them to change their behaviour when conditions change
(Nelson and Winter's `along-the-rule effects'). And, in the second place,
among the routines that ®rms have are routines that may trigger search
behaviour (depending on the actual results obtained) and search routines
that guide their search behaviour. Thus Nelson and Winter's `search
effects' may come about. Needless to say, both along-the-rule effects and
search effects result in changes in ®rm behaviour that in turn can bring
about changes in industry behaviour.
Finally, Nelson and Winter believe that `market selection' can produce
`selection effects' that also bring about changes in industry behaviour.
Thus we have:

(EE3N&W) Economic evolutionary theory takes `market selec-

tion' to be a dominant force in bringing about
changes in industry behaviour.

Note the similarity between (NE3A&F) and (EE3N&W). Whether `market

selection' produces changes in industry behaviour depends on the
`materials' it works on. In Nelson and Winter's theory the `materials' are
provided by the ®rms' routines. `Selection effects' will obtain only if there
is variety (or diversity) in routines. Now Nelson and Winter assume that
routines of ®rms are ®rm-speci®c. Routines are assumed to differ from
®rm to ®rm. This assumption is not only instigated by their belief that
real-to-life ®rms actually are different in their operating procedures, but
also by their belief that it is wrong, methodologically speaking, to assume
from the outset that one and only one type of ®rm exists (or can exist). In
particular, they think that it is wrong to premise a standard of perform-
ance that is independent of the characteristics of ®rms (p. 94). The
`representative ®rm' should not be assumed to exist from the outset. If it
exists at all it cannot be anything but the end product of evolutionary
processes.
We can capture the facts, ®rst, that Nelson and Winter take routines to
be ®rm-speci®c, and, second, that routines do not only comprise (®rst-
order) operating procedures, but also (second-order) search routines, by
replacing (EE2N&W) by:
(EE2N&W') Economic evolutionary theory takes the behaviour
of each and every ®rm in the industry to be
instructed by its own speci®c set of routines.
Among these ®rm-speci®c routines are routines that
determine under what conditions the ®rm in ques-
tion engages in search behaviour and (second-
 Ontological commitments of evolutionary economics 209

order) search routines that guide the ®rm's search

activities.28
We can complete our picture of Nelson and Winter's evolutionary
theory by adding a fourth characteristic stating that orthodox (neoclas-
sical) industry results are not reproduced by Nelson and Winter's
evolutionary theory under all conceivable conditions:
(EE4N&W) `market selection', working on routine-following
®rms, produces the same tendencies in industry
behaviour as fully informed and fully rational ®rms
would do only under particular conditions.

5 Evolutionary game theory

In his discussion of the relation between economics and evolutionary

biology, Rosenberg asserts that `the terms of the trade are always in the
direction from economics to biology and not vice versa' (Rosenberg
1994, p. 385). Rosenberg apparently believes that biology never had a
signi®cant impact on economic theorizing. This is in line with Rosen-
berg's general point that economics is quite different from evolutionary
biology.
What Rosenberg seems to overlook here is that his own assertion
about economics exporting ideas to biology implies that ideas stemming
from economics later on have been taken on board in biology. This
seems to run counter to his general point that economics and evolu-
tionary biology share next to nothing with each other. Indeed, Rosen-
berg himself refers to Malthus and to the Scottish moral philosophers
(such as Hume and Smith) who paved the way for Darwin to develop
his notion of natural selection. Apparently, then, Rosenberg acknowl-
edges that there have been rudimentary ideas of `natural selection'
around in `classical economics'. The only way for Rosenberg to
reconcile his assertion and his general point seems to be to hold that
these ideas that once were present in classical economics disappeared
from economics in its transition to neoclassical economics. In section 3
I have argued that these ideas did not vanish completely from
economics. In this section I argue that Rosenberg's assertion by itself is
questionable as well. For it overlooks that recently a signi®cant transfer

28 This does not mean that Nelson and Winter take search behaviour to be an altogether
deterministic affair. Their models of search behaviour also contain stochastic elements.
210 Micro, macro, and markets

of ideas from evolutionary biology to economics has been taking place:

the adoption (and transformation, as we shall see) of evolutionary game
theory by economists.

Evolutionary game theory: some preliminaries

Evolutionary game theory has been developed by biologists. Path-

breaking work in this ®eld has been done especially by John Maynard
Smith. Maynard Smith and his collaborators tried to accommodate
traditional game theory, as pioneered by von Neumann and Morgenstern
(1944) and made more accessible by Luce and Raiffa (1957), in such a
way that it would be suited to analyse animal contests. It is important to
note that the type of game theory Maynard Smith developed is radically
different from the traditional type of game theory. The traditional type
of game theory focuses on the question what ideally rational players
would (or should) do in situations where the payoffs of the behaviour of
players do not only depend on their own behaviour, but also on the
behaviour of others. The players are assumed to have common know-
ledge of the preferences of all players. Under this assumption of complete
transparency of the game to all players, in the traditional type of
(noncooperative) game theory fully rational players are typically
assumed to be able to reason their ways immediately to a Nash
equilibrium.
Evolutionary game theory differs from this by removing the assump-
tions of ideal (or unbounded, or omniscient) rationality and common
knowledge. It is assumed instead that if individual players converge on a
Nash equilibrium at all, they can do so only after having gone through
some laborious and time-consuming evolutionary process. In the crudest
biological version of evolutionary game theory, individual players are
taken to be organisms that are genetically programmed to follow certain
behavioural strategies. The individual `players' are assumed to have no
choice at all. Whether they like it or not, they are predisposed to play a
certain strategy. As the individuals are assumed to pass on their genes to
their offspring (in the simplest models it is assumed that reproduction is
asexual), their descendants are assumed to be programmed to play the
same strategy.
Constancy in the behavioural predispositions of individual organisms,
however, does not necessarily imply that change is absent also at the level
of the population. For the crucial assumption is that the force of natural
selection impinges on the individuals. Natural selection can produce
 Ontological commitments of evolutionary economics 211

differential reproduction. The ®tter individuals are assumed to produce

more offspring than the less ®t ones, so that the frequencies (proportions)
of the genes of the former in the population's gene pool increase at the
expense of the frequencies of the latter. Thus we can say that in this
crudest form of evolutionary game theory changes at the population level
are explained by natural selection working on a variety of individuals
that are assumed to be unchanging in the relevant (genetic) respect.
Maynard Smith introduced the notion of an Evolutionarily Stable
Strategy (ESS) as the appropriate concept of a stable equilibrium in this
setting (see Maynard Smith 1982). The basic idea is that if all individuals
in a population have converged on playing an ESS (they all play ESS),
then no mutant strategy (a strategy that is slightly different from the
ESS) can have a higher ®tness than the ESS. Thus, if a mutant strategy
happens to enter a population in which all other individuals play ESS,29
it will be wiped out by the force of natural selection. Hence the situation
of all individuals playing the ESS will be restored. Evolutionary game
theoretic analyses in terms of ESS's are static analyses, for such analyses
do not account for the processes of change that might lead populations
to converge on ESS's. Such dynamic processes were ®rst explicitly
modelled in Taylor and Jonker (1978). The replicator dynamics (also
sometimes called `Malthusian dynamics' or the `replicator equation') that
Taylor and Jonker formulated captures the simple idea that the popula-
tion's average ®tness is the threshold value that determines whether the
population share of some strategy increases or decreases. If the ®tness of
playing some strategy exceeds the population's average ®tness, then the
proportion of players in the population playing the strategy increases; if
it is below the population's average ®tness, then the proportion de-
creases.
In most of the applications of evolutionary game theory in economic
contexts, the crudest biological interpretation of evolutionary game
theory is repudiated. In particular, the idea that the behaviour of
individuals is fully determined by pre-established programs is aban-
doned. Economic agents, it is maintained, are less dumb and more
sophisticated than the animals that ®gure in the crudest biological
interpretations. It is typically assumed that individuals are capable of
learning and imitating other successful individuals. What these less crude

29 An important assumption here is that the population is very large, so that the chance of
two mutants meeting each other is negligible (compared to the chance of a mutant
meeting an individual playing the ESS). Another important assumption is that pairs of
individuals are drawn randomly from the population to play the game (there is no
`selective play').
212 Micro, macro, and markets

interpretations have in common with the cruder ones, however, is that

the reasoning and calculating skills of individuals are less developed and
less sophisticated than the ones that rational choice theory ascribes to
them. In the more re®ned interpretations, individuals are not assumed to
be farsighted and omniscient, or `unboundedly' rational, but myopic and
boundedly rational only (see, for example, Mailath 1992). Furthermore,
although it is not assumed in the more re®ned interpretations that the
behaviour of individuals is completely preprogrammed genetically, it is
still assumed that the learning and imitation behaviour is programmed in
some sense (see, for example, Weibull 1995, chapters 4 and 5). Learning
and imitation behaviour, it is assumed, are governed by ®xed decision
rules. It is sometimes argued that in social and cultural evolution, memes,
the term that Dawkins (1976) coined to designate the units of `replica-
tion' in imitation, take the place of genes (see, for example, Binmore
1994).

Whence the attractiveness of evolutionary game theory?

The crucial behavioural assumption in evolutionary game theory that the

behaviour of individuals is programmed clearly parallels Nelson and
Winter's contention that routines `as genes' instruct the behaviour of
®rms. Of course, a difference is that the elementary behavioural units in
Nelson and Winter's evolutionary theory are ®rms, and not individuals
(their constituent parts) as in evolutionary game theory. But the basic
explanatory structure seems to be the same: change at an aggregate level
of analysis is explained by evolutionary forces working on programmed
elementary behavioural units.
It is questionable, however, whether evolutionary game theorists are in
general as committed to the belief that human persons really display
(pre)programmed behaviour as Nelson and Winter are to the belief that
®rms really follow routines (quasi-) automatically. Most evolutionary
game theorists do not seem to be committed to the view that human
persons are computing machines or ®nite automata of some sort or
another.30 What they are primarily interested in, it seems, is in the
properties or the `behaviour' of some hypothesized theoretical system.
The fact that many evolutionary game theorists typically investigate
properties of different systems with differently programmed behavioural

30 See, e.g., Binmore and Samuelson (1992), p. 279, for two different interpretations of the
analysis of computing machines and ®nite automata in game theory.
 Ontological commitments of evolutionary economics 213

units already suggests that they are not claiming that any of their models
adequately (or truly) represents the `real world' that we inhabit. They are
`¯eshing out' the system-level implications of certain behavioural as-
sumptions, leaving aside the issue what behavioural assumption is most
realistic. Their theoretical endeavours are directed at ®nding answers to
questions like: what would happen at the system level if the individuals
involved were programmed in such and such ways?
Sometimes it seems that game theorists in economics resort to
evolutionary game theory in order to overcome dif®culties with conven-
tional noncooperative game theory (see Binmore 1995 and Rizvi 1997).
The key solution concept in conventional noncooperative game theory
is, as we have seen, that of a Nash equilibrium. One major dif®culty,
however, is that it is not at all clear why ideally rational players would
or should converge on a Nash equilibrium. A second dif®culty is that
many games typically have several Nash equilibria. This gives rise to the
`equilibrium selection problem': in the case of multiple equilibria, which
equilibrium is to be selected on what grounds? So many different
re®nements of the Nash equilibrium concept have been proposed to
solve this problem in conventional game theory, that in the end almost
any Nash equilibrium could be justi®ed in terms of one re®nement or
another.
In this problem setting, evolutionary game theory provides an attrac-
tive way out. Evolutionary game theory holds out the promise to solve
both dif®culties. So far, this hope does not seem to be forlorn. Systems
that behave according to the replicator dynamics turn out to have only
(symmetric) Nash equilibria as stable stationary points. Evolutionary
game theory can also solve the equilibrium selection problem. In evolu-
tionary game theory, the equilibrium selected is a function of the
equilibrating process by means of which it is arrived at. And the
equilibrating process in turn may not only depend on the replicator
dynamics by which it is governed, but also on the initial situation from
which the process starts. This means that `details' of equilibrating
processes, their historical `contingencies' and `accidents', must be taken
into account.31 Much of what is currently going on in evolutionary game
theory seems to concentrate on the issue whether the results that are
derived from replicator dynamics are invariant under perturbations of
the details in the speci®cation of the dynamics.
On the other hand, at least to some the attractiveness of evolutionary
game theory does not stem entirely from its potential to solve problems

31 `History matters' one could say; evolutionary game theory can be used to illustrate `path
dependent' processes and `lock-in' effects.
214 Micro, macro, and markets

`internal' to game theory. What is appealing in evolutionary game theory

to some is that it dispenses with the `grotesque' exaggerations of the
rationality of players that characterize conventional game theory. Evolu-
tionary game theory, it is thought here, can tell us more about the ways
people actually play games than conventional game theory (see, for
example, Sugden 1986, p. 16). This feature of evolutionary game theory
seems to be important especially for those who want to make use of game
theory to explain real-to-life phenomena. A fascinating example of the
latter is provided by attempts to use game theory in order to shed more
light on `the spontaneous evolution' of conventions, institutions and
norms. These evolutionary game theorists are not interested in the
properties of hypothesized theoretical systems for their own sake, but in
how the insights gained in investigating these properties can enhance our
understanding of how prevailing conventions, institutions and norms in
society have (or, to put it more carefully could have) evolved.

`Theoretical' and `applied' evolutionary game theory

Binmore, Kirman and Tani (1993) make some useful distinctions that
nicely bring out the differences in theoretical interests within the ®eld of
evolutionary game theory. Binmore et al. do not only distinguish
between theoretical work and applied work in game theory, but also
between ®ve general purposes for which game-theoretic models may be
designed: prediction, explanation, investigation, description, and pre-
scription. The ®rst distinction relates to types of issues game theorists
address. When game theorists analyse properties of game-theoretic
models (which is also called `gamesters looking in' by Binmore et al.,
p. 3), they are engaged in theoretical work. Applied work is done when
game theorists use their analyses to address `phenomena out there'
(`gamesters looking out'). Binmore et al. argue that in principle each of
the ®ve general purposes mentioned above can be aimed at both in
theoretical and in applied work. This leaves us with ten different types of
game-theoretic analyses.
In the present context of discussing the ontological commitment of
evolutionary game theory our attention can be con®ned, I think, to only
two of them: theoretical work (also sometimes called `pure' or `founda-
tional' work) devoted to the investigation (or `exploration') of the proper-
ties of theoretical systems, and applied work directed to the explanation
(or `explication') of the spontaneous evolution of conventions, institu-
tions and norms. Examples of work falling within the former category
 Ontological commitments of evolutionary economics 215

are Hansen and Samuelson (1988), Schaffer (1989), and Weibull (1994).
They all examine the conditions that are to be met for evolutionary
processes to result in a situation in which all agents behave as if they
were fully rational. Within the latter category fall for example Sugden
(1986) and Binmore and Samuelson (1994). Of course, there are also
borderline cases such as Young (1993) containing both elements of
theoretical investigation and applications to study the spontaneous
evolution of conventions.
We can summarize the work done in the ®rst category of theoretical
evolutionary game theory (tegt) as follows:
(EE1tegt) Theoretical evolutionary game theory investigates
properties of hypothesized systems (it explores models
of population behaviour in particular);
(EE2tegt) Theoretical evolutionary game theory represents indi-
viduals as programmed computing machines (and as
®nite automata in particular).
Theoretical evolutionary game theory typically investigates whether
different models, populated by computing machines with different
degrees of sophistication, have different properties. The fact that `indi-
viduals' with different degrees of sophistication are often considered by
the same game theorist indicates that theoretical evolutionary game
theory as such is not committed to any particular view on the essential
properties of real-to-life individuals. In particular, evolutionary game
theorists engaged in theoretical work do not seem to claim that indi-
viduals are ®nite automata.
Applied evolutionary game theory (aegt), on the other hand, seems to
be committed to a particular view on the essential properties of indi-
viduals. In its attempts to explain how conventions and institutions
evolved spontaneously applied game theory does seem to involve a
particular view on the causal role that individuals play in such processes
of spontaneous evolution. After all, how can such game-theoretic
accounts claim to provide an explanation of such processes if they get the
role that individuals play wrong? Things turn out to be a little bit more
complicated, however. As Binmore et al. point out, applied game theory
does not straightforwardly claim to explain how conventions and institu-
tions de facto evolve spontaneously. It rather gives `possible stylized
explanations of how things might have come about' (Binmore et al. 1993,
p. 5). Applied evolutionary game theory does not claim that the actual
ways in which conventions, institutions and norms came about corre-
spond precisely to the dynamic processes that it speci®es. It only shows
that it is possible or logically coherent to entertain the view that they
216 Micro, macro, and markets

emerged spontaneously out of the interactions of boundedly rational

individuals (p. 6).32
I take it that this last statement does not imply that applied evolu-
tionary game theorists leave open the possibility that real individuals are
`unboundedly' (perfectly) rational after all. What they leave open is the
possibility that the essential properties of individuals do not exactly
match the properties that are ascribed to them in their models. Real
individuals may be different from how their models portray them. The
rationality of real individuals may be either more or less bounded than
that of the individuals populating their models. What applied evolu-
tionary game theorists do not question, however, is that the rationality of
real individuals falls short of being perfect. Furthermore, real individuals
may be quite unlike the mechanical (machine-like) rule followers in their
evolutionary game-theoretic models. Like theoretical evolutionary game
theory, applied evolutionary game theory is not committed to the view
that there is no volition, consciousness, deliberation, and ratiocination
involved on the part of the individuals.
Evolutionary game theory as such is committed to the assumption that
individuals are boundedly rational (in some sense, and to some degree).
This assumption, together with the claim that the behaviour of bound-
edly rational individuals can be represented theoretically in terms of
mechanical rule following, can be found in theoretical and applied
evolutionary game theory. But game theorists engaged in theoretical
evolutionary game theory do not seem to be wedded to the view that
individuals really are boundedly rational. They simply want to explore
what follows from the assumption that individuals are boundedly
rational. By contrast, although game theorists engaged in applied game
theory do not claim that real individuals are boundedly rational exactly
to the degree as they are depicted in their models, they do seem to hold
that real individuals are boundedly rational.
Thus we have:
(EE1aegt) Applied evolutionary game theory gives possible
stylized explanations of the spontaneous evolution of
conventions, institutions and norms;
(EE2aegt) Applied evolutionary game theory holds that indi-
viduals are boundedly rational (their imitating and ±

32 Binmore et al. argue that the main intellectual achievement of such applications of game
theory is one of liberation: it provides a ®rst step to break out from ®xed preconceptions
that imprison our thoughts (one such preconception being that conventions and
institutions can only come about by conscious design).
 Ontological commitments of evolutionary economics 217

trial-and-error type of ± learning behaviour can be

described in terms of rule following).

6 Further re¯ections on evolutionary economics' ontological commitments

Now that we have had a closer look at Nelson and Winter's evolutionary
theory and evolutionary game theory, we can delineate general tenets in
their ontologies and compare these with those in neoclassical economics.
In the previous section we already started with delineating shared tenets
in the ontologies of Nelson and Winter's evolutionary theory and
evolutionary game theory by noting that in both theories the elementary
behavioural units (®rms and individual agents respectively) are taken to
display programmed behaviour. We quali®ed this observation by arguing
that whereas Nelson and Winter really seem to believe that routines
instruct ®rm behaviour in much the same way as genes programme the
behaviour of organisms (EE2N&W), evolutionary game theorists only
want to go as far as to assume that individuals behave as if they were
programmed (EE2tegt, EE2aegt).
Indeed, I argued that evolutionary game theorists can coherently
entertain the view that individuals act purposefully and choose con-
sciously. Nelson and Winter put quite some effort in challenging this
view. They contrast their notion of quasi-automatic routine behaviour
with that of ¯exible, deliberate choice. At the same time they also
acknowledge, however, that ¯exible, deliberate choice can be described in
terms of rule following. It is demonstrated by Nelson and Winter
themselves that optimizing behaviour can be translated into rule-gov-
erned behaviour (Nelson and Winter 1982, p. 150). This is the reverse
manoeuvre of the one that Machlup (1946) made to defend neoclassical
economics: rule-following behaviour can be redescribed and thus theore-
tically accounted for in terms of optimal behaviour.
This once again shows that the ontological commitments of economic
theories cannot be read off directly from how individual agents and the
system in which they are operating are depicted in theory. Some
economic theory may either suggest that individual agents are taken to
act consciously and deliberately, whereas its defenders believe that much
individual behaviour is displayed in a mindless, automatic routine way,
as in Machlup's case. Or, conversely, an economic theory may suggest
that individual agents are believed to behave as a machine, whereas its
practitioners may believe that individual agents act consciously and
deliberately, as in the case of evolutionary game theory.
218 Micro, macro, and markets

Thus evolutionary game theorists do not necessarily subscribe to

Nelson and Winter's view that individual behaviour is routine most of
the time. Ontologically speaking, however, they do have several beliefs
about individual agents and their essential properties in common with
Nelson and Winter. There is ®rst of all the recognition that individual
agents are only boundedly rational. The capabilities of individual agents
to foresee future contingencies, to gather and process information, and
to calculate best responses are severely restricted. The possibility that the
behaviour of boundedly rational agents equals that of hypothesized
unboundedly, perfectly rational agents cannot be foreclosed a priori. But
Nelson and Winter and evolutionary game theorists concur in the view
that (pure luck aside) boundedly rational agents can at most be expected
to display such behaviour after having gone through laborious and time-
consuming evolutionary process. And indeed both types of evolutionary
economists agree that it is one of the tasks of evolutionary economics to
®nd out under what conditions this can be expected to happen.
This brings us to a second point. Individual agents and their essential
properties are seen as products of the past. Another way of putting this is
to say that the operating characteristics of individual agents are believed
to exhibit some degree of inertia. The past casts a lasting shadow over the
present and future in that past occurrences and processes severely restrict
the opportunities of individual agents to adapt to new environmental
circumstances. Individual agents are unable to shape their operating
characteristics ex nihilo (or de novo). It is likewise considered to be
impossible that individual agents revise their operating characteristics
overnight.
Besides this feature of inertia there is a third related point on which both
proponents of evolutionary economics seem to concur. The operating
characteristics of individual agents are considered to be agent-speci®c.
Individual agents are believed to differ in their operating characteristics.
This can be called the feature of heterogeneity.33 What characterizes a
particular agent is his or her set of routines or programmes. Indeed, one
could say that agents are identi®ed by their routines. As Binmore (1992,
p. 414) puts it, in evolutionary game theory programmes must be seen as
substitutes for players. This feature marks a clear break with economic
theorizing in terms of representative agents. Again there is no presumption
that after having gone through an evolutionary process different agents
having different characteristics cannot wind up having the same represen-
tative characteristics. But there is no presumption to the opposite effect

33 This element of heterogeneity is also stressed in present-day computational economics

(see, e.g., Vriendiggs).
 Ontological commitments of evolutionary economics 219

either. Whether homogenized representative agents will be ultimately

produced in evolutionary processes is precisely one of the questions that
both types of evolutionary economics investigate.34
So far we have been discussing mainly beliefs of evolutionary econo-
mists about essential properties of individual agents (be it ®rms or
individual human beings). But in Nelson and Winter's evolutionary
theory a crucial role is played also by beliefs about essential properties of
the system (or regime) in which agents are operating. Competitive
markets are believed to exert selection pressure on individual agents. The
root idea behind this is that individual agents have to compete for scarce
resources. Scarcity here does not so much refer to a predicament
necessitating (intra-)individual choice as to an inevitable inter-individual
con¯ict of interests. All individuals are dependent for their scale of
operation (if not for their survival) on the amount of resources that they
manage to acquire. And the resources that the one is able to acquire are
subtracted from the remaining stock of resources that are left to others to
compete for.
This idea of a putative ubiquitous `brute fact of life' should not be
confused with that of a competitive market, however. If competition for
scarce resources is seen as belonging to the `natural order', the competi-
tive market is best viewed as one `social order' among others that can be
chosen to organize it.35 There are some (following Hayek, for example)
who prefer to view the competitive market as a social order that evolved
spontaneously rather than by ®at or by concerted action. But even they
do not deny that a choice can be made to replace the competitive market
by other market forms or by various forms of planning.36 Moreover, it
has been observed by many that safeguards have to be installed and
enforced in order to prevent the competitive market from `naturally'
transforming into a monopoly.37

34 To some (e.g. Witt 1993, Foss 1994, and Hodgson 1997) it is one of the de®ning
characteristics of evolutionary economics that processes of economic evolution are not
only assumed to start with already existing variety, but also to constantly engender
novelty and, hence, new variety. In my account of Nelson and Winter's evolutionary
theory and evolutionary game theory this characteristic is not put centre stage, although
it is to some extent captured in (EE2N&W'), (EE2tegt), and (EE2aegt).
35 Cf. Hirshleifer's distinction between `natural economy' and `political economy' (Hirsh-
leifer 1987). This also explains why `market selection' is more properly called a social
than a natural mechanism. See, e.g., MaÈki (1990).
36 This is not so say that society is malleable to our liking. Attempts to replace competitive
markets may fail (or may not lead to the goal intended). And, of course, in the eyes of
Hayekians any attempt to do so is regrettable.
37 If some ®rm consistently outcompetes other ®rms in the market (and its emerging
monopoly position is not endangered by new entrants), then of course the market ceases
220 Micro, macro, and markets

The fact that market form can be altered by a decree issued by some
collective agent or by concerted action of individual agents does not imply
that market participants can subvert the regime of the competitive market.
Indeed it can be said to belong to the essence of competitive markets that
market participants lack the power to do so. In competitive markets ®rms
are prohibited to act in concerted or orchestrated ways and, hence, are for
example unable to inaugurate barriers to new entrants.
Nelson and Winter's idea that ®rms in competitive markets are
exposed to selection pressure ®nds its counterpart in evolutionary game
theory in the idea of external environments imposing selection pressure
on individuals. Types that are relatively well adapted to the environment
spread in the population at the expense of types that are relatively poorly
adapted. Here the properties of the prevailing environment determine to
a great extent what types come to dominate the population. One may be
intuitively led to think of a natural environment here. But if there is
frequency-dependent selection, the properties of the social environment
can be said to have an impact on the selection process as well. In
frequency-dependent selection, as in the paradigmatic Hawk-Dove game,
the adaptedness of the types depends on the prevailing frequency (or
proportion) of the types in the population. Instead of saying that the
natural and social environment jointly (co-)determine the evolutionary
process here, one could also say that the evolutionary process does not
only depend on properties of the prevailing natural environment but also
on properties of the prevailing population.
Once again individuals are believed to be subject to pressure that they
cannot control. Individuals are believed to be unable to withstand the
pressure, let alone to overrule or nullify it. The rules of the game are
believed to be as `inert' (or ®xed) as the operating characteristics of the
players. Unlike the natural environment the social environment can
change in processes of frequency-dependent selection. But this does not
mean that the rules of the game are alterable as well. On the contrary, if
there are changes in the social environment, they unfold according to
®xed, pre-established rules of the game.
In sum, then, the ontological commitments of Nelson and Winter's
evolutionary theory and those of evolutionary game theory overlap to a
considerable degree. Both species of evolutionary economics can be said
to refer crucially to a natural and social environment exerting selection
pressure on individual agents. Although the social environment is not
taken to be immutable, it is believed to be beyond the reach of individual

to be competitive. Among others, late-nineteenth-century American economists have

stressed this possibility of `self-annihilation' of competition.
 Ontological commitments of evolutionary economics 221

agents to change it. Furthermore, as to the properties of the individual

agents themselves, both species of evolutionary economics concur on the
view that individual agents are boundedly rational, that their essential
operating characteristics exhibit inertia and that they differ (or, at the
very least, can differ) among different individual agents. Where the two
species differ is in their view on the immediate determinants of individual
behaviour. While Nelson and Winter hold that individual agents follow
rules in an unthinking machine-like way most of the time, evolutionary
game theory only goes so far as to claim that the behaviour of individual
agents can be represented theoretically in terms of rule-following be-
haviour.

References
Alchian, Armen A. (1950) `Uncertainty, Evolution, and Economic Theory',
Journal of Political Economy, 58: 211±21.
Becker, Gary (1962) `Irrational Behavior and Economic Theory', Journal of
Political Economy, 70: 1±13.
Binmore, Ken (1992) Fun and Games. A Text on Game Theory, Lexington, Mass.:
D. C. Heath and Company.
(1994) Playing Fair. Game Theory and the Social Contract (Part I). Cambridge,
Mass.: MIT Press.
(1995) `Foreword' to Weibull (1995), pp. ix±xi.
Binmore, Ken, Alan Kirman and Piero Tani (1993) `Introduction: Famous
Gamesters', pp. 1±25 in Ken Binmore, Alan Kirman and Piero Tani (eds.),
Frontiers of Game Theory, Cambridge, Mass.: MIT Press.
Binmore, Ken and Larry Samuelson (1992) `Evolutionary Stability in Repeated
Games Played by Finite Automata', Journal of Economic Theory, 57:
278±305.
(1994) `An Economist's Perspective on the Evolution of Norms', Journal of
Institutional and Theoretical Economics, 150: 45±63.
Caldwell, Bruce J. (1992) `Commentary' on Alan Nelson's `Human Molecules',
pp. 135±49 in Neil de Marchi (ed.), Post-Popperian Methodology of Eco-
nomics, Boston, Dordrecht, etc.: Kluwer Academic Publishers.
(1994) `Two Proposals for the Recovery of Economic Practice', pp. 137±53 in
Roger E. Backhouse (ed.), New Directions in Economic Methodology,
London: Routledge.
Clark, Andy (1997) `Economic Reason: the Interplay of Individual Learning and
External Structure', pp. 269±90 in J. D. Drobak and J. V. C. Nye (eds.), The
Frontiers of the New Institutional Economics, San Diego: Academic Press.
Cyert, Richard M. and James G. March (1963) A Behavioral Theory of the Firm,
Englewood Cliffs, New Jersey: Prentice Hall.
Dawkins, Richard (1976) The Sel®sh Gene, Oxford: Oxford University Press.
222 Micro, macro, and markets

Day, Richard H. and Ping Chen (1993) Nonlinear Dynamics and Evolutionary
Economics, New York, Oxford: Oxford University Press.
Dosi, G., C. Freeman, R. Nelson, G. Silverberg, and L. Soete (eds.) (1988)
Technical Change and Economic Theory, London: Pinter.
Foss, Nicolai (1994) `Realism and Evolutionary Economics', Journal of Social
and Evolutionary Systems, 17: 21±40.
Friedman, Milton (1953) `The Methodology of Positive Economics' in Essays in
Positive Economics, Chicago: University of Chicago Press.
Friedman, Milton and Leonard Savage (1948) `The Utility Analysis of Choices
Involving Risk', Journal of Political Economy, 56: 279±304.
Hammond, J. Daniel (1991) `Early Drafts of Friedman's Methodology Essay',
paper presented at the Annual Meeting of the History of Economics Society,
June.
Hansen, R. G. and W. F. Samuelson (1988) `Evolution in Economic Games',
Journal of Economic Behavior and Organization, 10: 107±38.
Hayek, Friedrich von (1948) `The Meaning of Competition' in Individualism and
Economic Order, Chicago: University of Chicago Press.
(1968) `Competition as a Discovery Procedure', reprinted as pp. 254±65 in
New Studies in Philosophy, Politics and Economics and the History of Ideas,
London: Routledge & Kegan Paul.
Herrmann-Pillath, Carsten (1996) `On the Ontological Foundations of Evolu-
tionary Economics' (mimeo, paper presented at the Second Workshop on
Evolutionary Economics, University of Witten-Herdecke, 25±27 October
1996).
Hirshleifer, Jack (1987) `Natural Economy Versus Political Economy', in
Economic Behaviour in Adversity, Sussex: Wheatsheaf Books, pp. 169±93.
Hodgson, Geoffrey M. (1997) `Economics and Evolution and the Evolution of
Economics', pp. 9±40 in Jan Reijnders (ed.), Economics and Evolution,
Aldershot: Edward Elgar.
Hull, David L. (1982) `The Naked Meme', pp. 273±327 in H. C. Plotkin (ed.),
Learning, Development and Culture, New York: John Wiley & Sons.
Kincaid, Harold (2001) `The Empirical Presuppositions of Metaphysical Expla-
nations in Economics', this volume, pp. 15±31.
Koopmans, Tjalling (1957) Three Essays on the State of Economic Science, New
York: McGraw-Hill.
Langlois, Richard N. and Michael J. Everett (1994) `What is Evolutionary
Economics?' pp. 11±47 in Lars Magnusson (ed.), Evolutionary and Neo-
Schumpeterian Approaches to Economics, Boston, Dordrecht, etc.: Kluwer
Academic Publishers.
Lester, R. A. (1946) `Shortcomings of Marginal Analysis for Wage-employment
Problems', American Economic Review, 36: 63±82.
Luce, R. D. and Howard Raiffa (1957) Games and Decisions, New York:
Dover.
Machlup, Fritz (1946) `Marginal Analysis and Empirical Research', American
Economic Review, 36: 519±54.
 Ontological commitments of evolutionary economics 223

(1967) `Theories of the Firm: Marginalist, Behavioral, Managerial', American

Economic Review, 57: 1±33.
Mailath, George J. (1992) `Introduction: Symposium on Evolutionary Game
Theory', Journal of Economic Theory, 57: 259±77.
MaÈki, Uskali (1990) `Scienti®c Realism and Austrian Explanation', Review of
Political Economy, 2: 310±44.
(1994) `Isolation, Idealisation and Truth in Economics', in Bert Hamminga
and Neil de Marchi (eds.), Idealisation IV: Idealisation in Economics,
Amsterdam, Atlanta, Georgia: Rodopi.
Maynard Smith, John (1982) Evolution and the Theory of Games, Cambridge:
Cambridge University Press.
Mayr, Ernst (1982) The Growth of Biological Thought, Cambridge, Mass.:
Harvard University Press.
Metcalfe, Stanley (1989) `Evolution and Economic Change', in A. Silberston
(ed.), Technology and Economic Progress, Basingstoke, Hampshire: Mac-
millan Press.
Morgan, Mary S. (1995) `Evolutionary Metaphors in Explanations of American
Industrial Competition', pp. 311±37 in Sabine Maasen, Everett Mendelsohn
and Peter Weingart (eds.), Biology as Society, Society as Biology: Metaphors,
Dordrecht/Boston: Kluwer Academic Publishers.
Nelson, Richard (1995) `Recent Evolutionary Theorizing about Economic
Change', Journal of Economic Literature, 39: 48±90.
Nelson, Richard and Sidney Winter (1982) An Evolutionary Theory of Economic
Change, Cambridge, Mass.: The Belknap Press of Harvard University Press.
Rizvi, A. (1997) `The Evolution of Game Theory' (paper presented at Erasmus
EIPE Seminar, 17 December 1997).
Rosenberg, Alex (1992a) Economics ± Mathematical Politics or Science of
Diminishing Returns?, Chicago: University of Chicago Press.
(1992b) `Neo-classical Economics and Evolutionary Theory: Strange Bedfel-
lows?', PSA: 174±83.
(1994) `Does Evolutionary Theory Give Comfort or Inspiration to Economics?'
pp. 384±407 in Ph. Mirowski (ed.), Natural Images in Economics, Cam-
bridge: Cambridge University Press.
(2001) `The Metaphysics of Microeconomics', this volume, pp. 174±88.
Ross, Don (1995) `Real Patterns and the Ontological Foundations of Micro-
Economics', Economics and Philosophy, 11: 113±36.
Satz, Debra and John Ferejohn (1994) `Rational Choice and Social Theory',
Journal of Philosophy, 91: 71±87.
Saviotti, P. Paolo and J. Stanley Metcalfe (1991) `Present Developments and
Trends in Evolutionary Economics', pp. 1±30 in P. Paolo Saviotti and
J. Stanley Metcalfe (eds.), Evolutionary Theories of Economic and Technolo-
gical Change, Chur, Reading, etc.: Harwood Academic Publishers.
Schaffer, Marc E. (1989) `Are Pro®t-maximizers the Best Survivors?', Journal of
Economic Behavior, 11: 29±45.
Simon, Herbert A. (1983) Reason in Human Affairs, Oxford: Basil Blackwell.
224 Micro, macro, and markets

Sugden, Robert (1986) The Economics of Rights, Co-operation and Welfare,

Oxford: Basil Blackwell.
Taylor, P. and L. Jonker (1978) `Evolutionarily Stable Strategies and Game
Dynamics', Mathematical Biosciences, 40: 145±56.
Van Damme, Eric (1994) `Evolutionary Game Theory', European Economic
Review, 38: 847±58.
Von Neumann, John and Oskar Morgenstern (1944) Theories of Games and
Economic Behavior, Princeton: Princeton University Press.
Vriend, Nicholaas (1995) `Self-organisation of Markets: An Example of a
Computational Approach', Computational Economics, 8: 205±31.
Vromen, Jack J. (1995) Economic Evolution. An Inquiry into the Foundations of
New Institutional Economics, London: Routledge.
Weibull, JuÈrgen W. (1994) `The ``As If '' Approach to Game Theory: Three
Positive Results and Four Obstacles', European Economic Review, 38:
868±81.
(1995) Evolutionary Game Theory, Cambridge, Mass.: MIT Press.
Winter, Sidney (1964) `Economic ``Natural Selection'' and the Theory of the
Firm', Yale Economic Essays, 4: 225±72.
Witt, Ulrich (1993) `Evolutionary Economics: Some Principles' pp. 1±16 in
Ulrich Witt (ed.), Evolution in Markets and Institutions, Berlin: Springer
Verlag.
Young, Peyton (1993) `The Evolution of Conventions', Econometrica, 61: 57±84.
Zerbe, R. O. Jr. (ed.), (1982) Research in Law and Economics, 4, Greenwich,
Conn.: JAI Press Inc.
12 Is macroeconomics for real?

KEVIN D. HOOVER*

All are keeping a sharp look-out in front, but none suspects that
the danger may be creeping up from behind. This shows how
real the island was.
J. M. Barrie, Peter Pan

Children are often thought to be peculiarly honest ± witness the story of

``The Emperor's New Clothes.'' My title comes from a group of my
academic children: ®rst-year graduate students. I teach a mandatory class
in macroeconomic theory to graduate students in both an economics
department and an agricultural economics department. The students in
agricultural economics are typically more interested in crop patterns or
natural resources ± relentlessly microeconomic topics ± than in unem-
ployment, GDP growth, or interest rates. Each year at least one student,
who I assume comes from the agricultural economics department, writes
on the anonymous class evaluation something like this: ``If macro-
economics were real economics ± which it is not! ± this would have been a
good class.'' What is one to say to the honest and piercing doubts of an
academic child?
The idea that macroeconomics stands in need of a microfoundational
base is a commonplace among economists. I shall argue that what
motivates this belief are principally ontological concerns, naãÈvely, but
pointedly expressed, in my students' questions about the reality of
macroeconomics. I shall argue that ontological reduction of macro-
economics to microeconomics is untenable. Thus, while the program of
microfoundations may illuminate macroeconomics in various ways, it
cannot succeed in its goal of replacing macroeconomics.
* I am grateful to Uskali MaÈki and Thomas Mayer for valuable comments on the ®rst
version of this paper, and to D. Wade Hands for some pointers on the literature on
supervenience. This research was supported under National Science Foundation grant
No. 9311930.

225
226 Micro, macro, and markets

To begin at the beginning, it might help to de®ne the key terms.

``Macroeconomics'' is sometimes thought to be the economics of broad
aggregates, and ``microeconomics'' the economics of individual economic
actions. Although he did not use the terms ``microeconomics'' and
``macroeconomics,'' John Maynard Keynes (1936, pp. 292±3) drew a
related distinction: microeconomics is the theory of the individual
industry or ®rm; macroeconomics is the theory of output and employ-
ment as a whole. As Maarten Janssen (1993, ch. 1) shows, these
alternative de®nitions cut in somewhat different ways. They are,
however, similar enough for present purposes, since any quanti®cation of
output and employment as a whole, is bound to involve broad aggregates.
Macroeconomics is thus that area of economics that treats of GDP,
unemployment, interest rates, the ¯ow of ®nancial resources, exchange
rates, and so forth.
Uskali MaÈki (1996) offers a careful discussion of realism in relation to
economics that might help to de®ne the term ``real'' in the title. MaÈki
distinguishes between ontological realism, which raises questions about
what there is, and semantic realism, which raises questions about the
connection between language and what there is. Semantic realism can be
analyzed further, but the central claim of this essay is ontological:
macroeconomic aggregates exist. More precisely, and again using MaÈki's
terminology, I will argue that macroeconomic aggregates exist externally
(i.e., independently of any individual human mind) and objectively (i.e.,
unconstituted by the representations of macroeconomic theory). This
claim, if it can be sustained, undermines one of the central rationales for
the program of microfoundations for macroeconomics that since at least
the mid-1940s has dominated thinking among economists.

1 The program of microfoundations

Before this century, the most common de®nition of economics was

epitomized by Alfred Marshall (1920, p. 1): ``[A] study of mankind in the
ordinary business of life; [economics] examines that part of individual
and social action which is most closely connected with the attainment
and with the use of the material requisites of well-being.'' This de®nition,
which is reasonably hospitable to macroeconomics, has now largely been
supplanted by that of Lionel Robbins (1935, p. 16): ``Economics is the
science which studies human behaviour as a relationship between ends
and scarce means which have alternative uses.'' On Robbins's de®nition,
economics must be fundamentally about the individual.
 Is macroeconomics for real? 227

Modern macroeconomics developed in the wake of Keynes's General

Theory of Employment Interest and Money (1936). Typical elements of
Keynes's analysis were the consumption function, which related aggre-
gate consumption to aggregate national income, the investment function,
which related aggregate investment to the general rate of interest, and the
liquidity preference function, which related the aggregate stock of money
to aggregate national income and the general rate of interest. It is easy to
understand that in a profession committed to Robbins's de®nition of
economics, such aggregate relationships were at best rather unappealing
way-stations on the path to an individualist economics. The program of
microfoundations, as it has developed over the past ®fty years, aims to
explain all macroeconomic properties of the economy ± in principle, at
least ± by reference to the behavior of rational economic actors such as
postulated by microeconomics.1
Approval of the program of microfoundations is almost universal
among economists. Those economists who have re¯ected on the matter
at all deeply typically associate microfoundations with methodological
individualism (e.g., Janssen 1993, pp. 26±9 passim; Boland 1979, chs. 2
and 5).2 Blaug (1992, p. 44) de®nes methodological individualism as the
principle that ``. . . asserts that explanations of social, political, or
economic phenomena can only be regarded as adequate if they run in
terms of the beliefs, attitudes, and decisions of individuals.''
Methodological individualism is a doctrine about explanation. Despite
lip-service to it, it is not widely practiced by economists. The reason is
what I have elsewhere labeled the ``Cournot problem'' after its lucid,
early formulation by Augustine Cournot (1838/1927, p. 127), the nine-
teenth-century mathematician and economist: there are too many indi-
viduals (®rms and consumers) and too many goods to be handled by
direct modeling.3 Mark Blaug (1992, p. 46) observes that few explana-
tions of macroeconomic phenomena have been successfully reduced to
their microfoundations, so that insistence on microfoundations would
eliminate explanations of macroeconomic phenomena tout court. Even
Lucas (1987, pp. 107±8), an important advocate of the program of
microfoundations, holds up only the hope of the elimination of distinc-
tion between microeconomics and macroeconomics.
The commitment of economists to methodological individualism is
1 See Weintraub (1979) and Janssen (1993) for general discussions of microfoundations for
macroeconomics.
2 I must be careful not to leave a false impression: both Janssen and Boland are critics of
the program of microfoundations.
3 Hoover (1988, p. 135). Also see Hoover (1988, ch. 9, esp. section 2; ch. 10, pp. 241±4);
Friedman (1955).
228 Micro, macro, and markets

thus not grounded in successful applications. Rather it appears to be

based on an instinctive belief in ontological individualism: the doctrine
that all that exists fundamentally for the economy are individual
economic actors. Lucas and his fellow new classical economists have
promoted representative-agent models, a class of models in which the
mathematical methods of microeconomic optimal choice are applied to a
single individual who takes national income as his budget constraint and
whose choices are taken to represent the aggregate choices of the
economy, because they appear to achieve the reduction of macro-
economics to microeconomics as required by the program of microfoun-
dations for macroeconomics. A. P. Kirman (1992) severely criticizes the
representative-agent model, not because it aspires to methodological
individualism, but because it fails to ful®ll the necessary conditions for
perfect aggregation, so that the representative agent in the models fails to
represent actual individuals successfully. Methodological individualism
remains the goal. Similarly, David Levy (1985) argues that complete
methodological individualism is impossible, because, given imperfect
information, individual economic actors must make reference to collec-
tive entitites as part of their own decision-making processes. Neverthe-
less, Levy (1985, p. 107) writes: ``These collectives have no real existence
but are simply the product of theories.'' While defending macro-
economics against the strong claims of the program of microfoundations,
Blaug (1992, p. 45) nevertheless writes: ``ontological individualism is
trivially true.''
It is important to understand that there are some senses in which
neither the methodological nor the ontological individualist denies the
existences of aggregates, collectives, or wholes. No one denies that GDP
calculations can be made and reported, or even that GDP may have
some locally stable relationship to unemployment or average interest
rates or some other aggregate. Similarly, no one denies the existence of
social organizations such as governments or ®rms (in the sense that talk
of governments and ®rms conveys meaning). What is typically denied is
that such aggregates or organizations are among the fundamental units
from which economic reality is constructed.
Hayek (1979, ch. 4) argues that such entities are secondary, and that
the role of a social science is ``compositive'' ± that is, that it must explain
these entities as arising from the fundamental individual components.4
Hayek (1979, ch. 6) denies that the wholes that social science explains

4 While not denying that aggregates such as GDP or the general price level can be
calculated, Mises (1949/1966, p. 217) goes further than Hayek in arguing that they are
quite devoid of meaning (also see Lachmann 1976, p. 96).
 Is macroeconomics for real? 229

through compositive methods are subject to scienti®c laws. He holds up

the attempt to connect them through laws as an example of Whitehead's
``fallacy of misplaced concreteness.'' He writes: ``the wholes about which
we speak exist only if, and to the extent to which, the theory is correct
which we have formed about the connection of the parts which they
imply, and which we can explicitly state only in the form of a model built
from those relationships'' (Hayek 1979, p. 98). Hayek thus argues that
aggregates exist, but derivatively rather than fundamentally, and that in
MaÈki's terminology they do not exist objectively (i.e., unconstituted by
the representations of theory).5 Still, even Hayek does not endorse
practicable methodological individualism, stressing the importance of a
reduction to microfoundations in principle and himself citing the
Cournot problem (Hayek 1979, pp. 74±5, esp. fn. 8).

2 Is macroeconomics ontologically problematic?

One might concede the main point of the last section ± namely, that the
drive for microfoundations is driven by ontological individualism ± yet
not believe that any interesting metaphysical issue is involved, because
the ontology of economics is too well understood by common sense to
pose any serious puzzle. MaÈki (1996, p. 434), for example, contrasts folk
economics with scienti®c economics, arguing that scienti®c economics
merely presents modi®cations of the `` `ontic furniture' of the general folk
views of `man' and society.'' He lists some types of possible modi®ca-
tions: selection, abstraction, idealization, exaggeration, projection, aggre-
gation. But he maintains that none of these modi®cations or
combinations thereof ``accomplishes a major departure from the ontic
furniture of the ordinary realm. No new kinds of entities or properties are
introduced'' (MaÈki 1996, p. 435). MaÈki illustrates his point with a
discussion of the business ®rm in standard neoclassical analysis, and
concludes ``that folk economics and neoclassical economics have real
business ®rms as their shared referent even though they represent these
®rms differently.''
MaÈki's case for the ``ontological commonsense realism'' of economics
in which the ontic furniture poses no special challenges to the under-
5 The terminology of ``fundamental'' or ``derivative'' existence is fraught with dif®culties. It
is beyond my purpose to try to sort such matters out here. It is enough for the point at
hand, however, to note that Hayek does not believe that economic aggregates can be
causes in their own rights. They might serve as some sort of shorthand, but he argues that
there is always an adequate causal mechanism independent of that shorthand.
230 Micro, macro, and markets

standing is more persuasive for some parts of economics than others. I

want to argue that for some important macroeconomic aggregates, it is
not particularly cogent, and that some such aggregates do not share a
referent with folk notions. The case can be illustrated with reference to
the related notions of ``real GDP'' (or ``real income'' or ``real output,''
these terms being used almost interchangeably) and the ``general price
level.''
On any interpretation, macroeconomics takes a larger view of the
economy and deals with aggregates, which are, in turn, constructed from
characteristics of individual economic actors. It is helpful to distinguish
two types of aggregates.
First are what we might call natural aggregates: simple sums or
averages. Examples of natural aggregates are the level of total employ-
ment or the average rate of interest on six-month commercial paper. I
call these natural aggregates because they are measured in the same units
(i.e., they have the same dimensionality) as the individual components
that they comprise and, therefore, preserve a close analogy with those
individual components. Employment, for instance, is measured by the
number of workers or the number of man/hours at both the level of a
single individual and in aggregate. The rate of interest on a bond and the
average rate of interest on a group of similar maturity bonds are both
expressed as a percentage yield per unit time.
A second type of aggregate, what we might call synthetic aggregates,
are important for macroeconomics. I call these synthetic because they
are fabricated out of components in a way that alters the structure of
the components, so that they are dimensionally distinct from the
components and so that there is no close analogy (despite their some-
times sharing a common name) with the components. The nature of the
synthesis is well illustrated by the general price level. The notion of a
general price level aims to capture a pre-scienti®c insight: ``a dollar just
ain't what it used to be!''; ``when I was a lad a penny would buy what a
quarter does these days.'' To capture this insight, one would like to
have some notion of the average level of prices. A simple average will
not of course work: (10¢/orange + 20¢/apple + $27,948/Volvo station-
wagon)/3 does not convey any useful information. One cannot add
apples and Volvos, as they say.
It might be argued that any sort of an average is altogether the wrong
way to start. What is really wanted is an estimate of the price of money
itself, and not the average price of goods. The price of money would, like
the price of oranges, be a single number. The general price level could be
de®ned to be its inverse (1/pm). Since relative prices of goods change
because of changes in the conditions of demand and supply, there would
 Is macroeconomics for real? 231

be, at best, a rough proportionality between individual prices and the

general price level. Indeed, it would permit one to isolate which changes
in individual prices were the result of ``real'' factors and which of
monetary factors.
This approach, however, does not do justice to the pre-scienti®c
insight, for it does not provide us with a concept or a measurement of
prices that is independent of highly particular and highly inadequate
theoretical models. To see this consider, how one would actually
determine the value of pm. One might, for instance, write down a
complete Walrasian general equilibrium model in which commodities
were expressed in natural quantities, prices in terms of money, and all
assets denominated in money were valued using pm. Aside from the
impracticality of formulating and solving such a model for an actual
economy (the Cournot problem again), it is well-known that pm might
not be determinable in such a model, or if it is, might not be unique
(Hahn 1965; Samuelson and Sato 1984; see Hoover 1988, ch. 5, section 1
for a simple exposition). The essence of the problem is that the real
quantity of money (i.e., the useful services that it provides), unlike the
real quantities of apples or Volvos, depends fundamentally on the price
of other goods. In adjusting the prices of various goods, including
money, unique convergence may not be possible, because each time
prices adjust to remove an excess demand or supply, the quantity of
money changes ± possibly in a discontinuous manner ± which can
increase rather than diminish some of the excess demands or supplies.
This problem in the foundations of monetary theory has yet to be
satisfactorily resolved. But what if it had been? It would tie the notion of
the general price level extremely closely to a particular theoretical
analysis. The measurement of pm would be ``derived'' rather than
fundamental (see Ellis 1966, ch. 8). The pre-scienti®c notion is not tied to
such a derivation. That would not pose any special problem if the general
price level derived in this way correlated closely with numerous other
theoretical methods of deriving it, which in turn correlated reasonably
with the pre-scienti®c sense of a general rise in prices.6 Temperature
provides an example of what is wanted (Ellis 1966, ch. 6). The notion of
hotter and colder is pre-scienti®c. The ®rst attempts to provide quanti®-
cation relied on some presuppositions ± e.g., the linearity of the expan-
sion of the various ¯uids used in thermometers ± but were not tied to
particular theories. Temperature measures can now be derived from
particular theoretical understandings ± e.g., from the kinetic theory of

6 Avogadro's number, for example, can be computed to take the same value from numerous
theoretically independent methods (see the discussion in Hacking 1983, pp. 54±5).
232 Micro, macro, and markets

gases. Such derived measures show considerable consilience with the pre-
scienti®c notions of hotter and colder and with the atheoretical measure-
ment systems. They permit the extension of temperature scales beyond
ordinary experience ± e.g., to the measurement of the temperature of the
sun ± but retain their independence from particular theories because of
the consilience of measurements derived from different theoretical
starting points. In contrast, the measurement of the general price level
remains at the atheoretical stage in which the makers of price index
numbers, Laspeyres, Paasche, and Fisher, are the economists' Fahrenheit
and Celsius.
The point of raising these dif®culties in measuring the general price
level is not that the existence of aggregates is tied to their measurement.
Rather it is that the dif®culties in measuring them help to expose what
problematic entities they are, and undermine the appeal of seeing them as
close analogues of their components (particular prices, particular goods,
and so forth). The disanalogies can be made clearer through a more
detailed examination of the general price level.
The fundamental insight of the index number is that one can avoid
some of the dimensional nonsense of averaging disparate prices by
averaging percentage rates of change instead. A simple average, however,
does not capture the commonsense feeling for the degree of price change.
A change in the price of gasoline should count for more than, say, a
change in the price of caviar in measuring the change in the general price
level. How to weight various price changes turns out to have an
irreducible degree of arbitrariness.
In general the percentage change in the general price level, indicated as
p (where p is the logarithm of the general price level), is related to the
individual underlying prices as
Dp ˆ f …Dp1 ; Dp2 ; . . . ; Dpn †; (1)
where pj is the price of good j for j = 1, 2, . . . n. Now, the properties that
theoretically restrict the functional form of f(.) are very weak:
1. if Vj Dpj  
< 0; then Dp < 0.
2. Dp  maxfDp1 ; Dp2 ; . . . ; Dpn ; g:
3. Dp  minfDp1 ; Dp2 ; . . . ; Dpn ; g.
Conditions 2 and 3 together imply an obvious corrollary:
4. if Dp1 ˆ Dp2 ˆ . . . ˆ Dpn ; then Dp ˆ Dpj for any j.
Condition 1 says that if each price increases (decreases) the general
price level must itself increase (decrease), and that the general price level
cannot change if no individual price changes. Conditions 2 and 3 say that
 Is macroeconomics for real? 233

the general price level cannot increase by more than the largest nor
decrease by less than the smallest individual price change. Condition 2
says that if every individual price changes equiproportionally, so must
the general price level. An in®nite number of functions ful®ll these
conditions, and the range of consistent changes in the general price level,
given a ®xed set of underlying price changes is wide. In practice, price
indices are generally linear
P
m P
m
Dp ˆ wj D pj ; where wj ˆ 1: (2)
jˆ1 jˆ1

For m < n, this formulation recognizes the practical fact that price
indices are based on samples of selected goods. The weights wj in these
indices are chosen in practice to capture the pre-scienti®c sense of the
amount of a price rise. Two common weighting schemes with rationales
in microeconomic consumer theory are the Laspeyres index, which
chooses the weights to re¯ect the share of each good in base period
consumption, so that the general price level effectively measures the
changing cost of a ®xed bundle of goods, and the Paasche index, which
chooses weights to re¯ect the share of each good in current period
consumption, so that some compensation is made for substitution from
relatively more expensive to relatively cheaper goods in the face of
changes in relative prices.
Neither index is ``correct''; there are an in®nite number of indices lying
between the two; and economists have from time to time argued the case
for other indices with different weighting schemes.7 The non-uniqueness
of the price index is important for the point of this essay. It is a
fundamental property. A price index is an attempt to quantify the pre-
scienti®c insight that the value of money changes. The different admis-
sible price indices are not, however, approximations to some true under-
lying general price level. The general price level is in some fundamental
sense non-scalar, although there is no currently acceptable scienti®c
re®nement that captures that fact.8 No similar inde®niteness attaches to
any of the prices of the underlying individual goods.
The change in the general priceR level, p, may be integrated over time to
t‡n
generate a price level …pt‡n ˆ t Dpdt ˆ p0t‡n ‡ c†. The constant of
integration (c) permits us to choose an arbitrary base usually Pt = 1 or
7 Generally one expects the Laspeyres index to be greater than the Paasche index, but this
can be guaranteed only if certain regularity conditions are imposed on preferences that
may not always hold for individual agents.
8 This may be an area in which the theory of fuzzy sets would be helpful. The use of scalar
indices may account for some portion of the apparently irreducible randomness in
estimated macroeconomic relations.
234 Micro, macro, and markets

Pt = 100, where P = exp( p), for some desired base period t (other base
values are less common, but not unknown). P differs from the particular
price of, say, a Volvo, not just in its intrinsic inde®niteness, but also in its
dimensions. The dimensions of the price of a Volvo are dollars/Volvo; the
dimensions of P are period-t dollars/base-period dollars. The dimensions
of P are not the dimensions of the price of any good. They appear to be
the inverse dimensions of the price of money, taking base-period money
to be the numeraire. Given the inde®niteness of the price index, however,
it is evident that the price of money is unlike the price of other goods, and
represents a substantial departure from pre-theoretic notions of price.
The price index is used to normalize the price of particular goods, thereby
to decompose individual price changes into a common or general element
and a ``real'' or relative (to the index) element. That this operation is not
obvious to common sense will be evident to anyone who has taught
elementary economics or read policy analysis by non-economists.
Real GDP is another important example of a synthetic aggregate.
Considered as national income, nominal GDP adds up the incomes of
each individual in the economy and is an obvious extension of the
accounting framework for business or personal income. In a major
innovation in economic analysis, the national accounting framework
since the 1930s establishes the three-way identity between the sum of all
incomes, the value-added in production, and the value of all ®nal goods
and services. That these other methods of computing GDP have obvious
commonsense analogues is less clear. If ®nal goods (i.e., goods that are
not inputs into other production processes) are indicated by Qj, then
nominal GDP is
P
n
YN ˆ P j Qj j ˆ 1; 2; . . . n: (3)
jˆ1

The dimension of income is dollars/unit time. Money provides the

common unit that is essential if disparate goods are to be added.
If some or all of the prices of individual goods increase, it is obvious
that nominal GDP could increase without any of the individual quan-
tities changing. In the utilitarian framework that underlies economics
this is an undesirable characteristic, because the measure of income has
changed without the underlying utility, which is assumed to be generated
by the quantities of the goods themselves, changing. It is clearly desirable
to correct for changing prices. The usual way to do this is to compute
real GDP as
Y R ˆ Y N =P: (4)
Real GDP is often treated as the analogue of an individual good. It
 Is macroeconomics for real? 235

does not, however, have the dimensions of a real good. Rather its
dimension is base-period dollars (not dollars/unit good). Real GDP is a
derivative measurement. One gets a different measurement for it for
every different admissible price index. It inherits the fundamental fuzzi-
ness of the general price level.
The analogy of real GDP to an individual good is suggested to some
by the possibility of perfect aggregation. If, for example, relative
Pn prices
are constant (i.e., pj /pk is constant for all j and k), then jˆ1 Pj;t Qj;t
(where the t in the subscript indicates the base time period t) can be
normalized by choosing the units for the Qj,ts so that each Pj,t = 1. Then
nominal GDP at time n can be written
P
n P
n
Pj;t‡n Qj;t‡n ˆ Pt‡n Qj;t‡n . (5)
jˆ1 jˆ1

In this case, conditions 1 to 4 above insure that P is unique. Some

conclude therefore that in this limited case, one can treat the summation
in the right-hand side of equation (5) as a natural aggregate quantity
analogous to an individual quantity. The conditions for constant relative
prices are almost certainly never ful®lled, but even if they were the
summation is not analogous to an individual quantity. The general price
level P in (5) still has the dimension period-n dollars/period-t (i.e., base
period) dollars. To sum heterogeneous goods, they must still be con-
verted to a common denominator, and in this case, the summation still
has the dimensions of period-t dollars. This would be more perspicacious
if (5) were written as
P
n P
n
Pj;t‡n Qj;t‡n ˆ Pt‡n 1j;t‡n Qj;t‡n ; (6)
jˆ1 jˆ1

where the subscripted numeral 1 is a place holder for the dimensional

conversion.
The general price level and real GDP are the most important aggre-
gates in macroeconomics. There are many others. Each mixes the
characteristics of simple and synthetic aggregates to different degrees.
Average interest rates were cited above as an example of a simple
aggregate, but when averaging is across nonhomogeneous maturities and
risk classes, interest rates too are complicated by the fundamental
problems of index numbers. Aggregation of employment across skill or
quality levels faces similar considerations. There are other derivative
quantities as well. The real rate of interest is de®ned to be the market
interest rate less the percentage change in the general level of prices (Dp).
Like real GDP, the real rate of interest inherits the fundamental fuzziness
of the general price level.
236 Micro, macro, and markets

The history of quantitative economics demonstrates that even the use

of simple averages represented a dif®cult conceptual leap. On the best
interpretation what is accepted to common sense is relative. To treat
synthetic aggregates as mere extensions of commonsense notions appears
in comparison to make a category mistake. Despite their deceptively
related names, there is no simple analogy between the general price level
and individual prices or between quantities of individual goods and real
GDP.

3 The supervenience of macroeconomics on microeconomics

Synthetic aggregates, at least, are not direct extensions of folk ontology.

It is clear, however, that, if their independent reality is to be demon-
strated in a sense more fundamental than that one can always calculate
them according to some algorithm, we must ®rst show that such
aggregates cannot be reduced to properties of individual economic
actors. Aggregates are in fact calculated; they clearly do not exist in a
separate Platonic realm; and ontological individualism has immediate
appeal, because we all have ®rst-hand experience as economic actors.
Any account of the autonomy or nonreducibility of macroeconomic
aggregates must account, therefore, for the relationship of the individual
to the aggregate.
Macroeconomic aggregates I believe supervene upon microeconomic
reality. What this means is that even though macroeconomics cannot be
reduced to microeconomics, if two parallel worlds possessed exactly the
same con®guration of microeconomic or individual economic elements,
they would also possess exactly the same con®guration of macro-
economic elements. It is not the case, however, that the same con®gur-
ation of macroeconomic elements implies the same con®guration of
microeconomic elements.
Biology provides analogies and disanalogies for economics. Alexander
Rosenberg (1985, ch. 4, section 8, ch. 6, section 3, passim) applies the
notion of supervenience to the relationship of functional biology (macro)
to molecular biology (micro). To take one example, hemoglobin is an
element in functional explanations of the operations of the cardio-
pulmonary and circulatory systems of higher animals (Rosenberg 1985,
ch. 4, section 2). At the molecular level, hemoglobin is not a single
chemical, but a family of chemicals. To be hemoglobin at the macro
level, a molecule must possess nine particular proteins at critical
junctures in the molecular structure. Across different species the approxi-
 Is macroeconomics for real? 237

mately 140 remaining proteins that the hemoglobin molecule comprises

vary considerably. Similarly, Rosenberg argues that Mendelian genetics
supervenes on a molecular base. Mendelian genetics uses a conceptual
scheme that is not easily mapped onto molecular features, but neverthe-
less identical molecular con®gurations produce identical genetic
behavior.
The notion of supervenience was initially suggested in the phil-
osophy of mind as a method of retaining the dependence of the mental
on the physical, while at the same time denying psychophysical laws
(see Kim 1978, p. 153). Rosenberg draws on the analysis of Jaegwon
Kim (1978). For Kim, supervenience is a relationship between two
distinct realms of properties (or relations). Consider Ij, which is a
conjuction of properties in the micro realm in which every one of the
properties or its complement form one of the conjuncts.9 Each Ij is
then a complete characterization of a possible micro state, and the
disjunction of every Ij de®nes every possible micro state. Consider the
Aj, constructed mutatis mutandis for the macro state. A family of
macro properties is supervenient on a family of micro properties when
any objects which share the same micro properties necessarily share
the same macro properties. Kim (1978, pp. 152±3) shows that one can
derive the following relationship:
I1 _ I2 _ . . . _ In ? Ak, for any Ak, (7)
where the Ih, h = 1, 2, . . ., n, are a subset of the Ij.10 Of relation (7), Kim
(1978, p. 153) says: ``I don't see how such generalizations could fail to be
lawlike.''
Using Kim's analysis, Rosenberg argues against the autonomy of
Mendelian genetics. The conceptual scheme of Mendelian genetics (the
macro level) does not map easily into the conceptual scheme of molecular
biology (the micro level). Mendelian genetics permits explanation of
phenomena not easily explainable directly from the molecular level.
However, Mendelian genetics fails to account for some phenomena
within its scope. Rosenberg argues that Mendelian genetics supervenes
on molecular biology, and that molecular biology is the more scienti®-
cally advanced, more fundamental and autonomous theory. Mendelian
genetics is reducible in principle (that is the upshot of Kim's analysis in
9 The account of Kim's analysis here omits most of the technical details (these are also
reproduced in Rosenberg 1985, pp. 113±16), and changes his notation. The identi®cation
of the distinct realms of properties as ``micro'' and ``macro'' is my addition ± literally ad
hoc ± and does not signi®cantly affect Kim's analysis.
10 Kim goes on to show that in some cases the implication in (7) can be strengthened to a
biconditional.
238 Micro, macro, and markets

(7) above), but it retains heuristic power, because something like the
Cournot problem prevents the practicable application of molecular
biology to some phenomena in which Mendelian genetics is relatively
successful.
Can a similar argument be applied to economics? I do not think so.
Rosenberg (1992, p. 129) himself has argued that the intentional char-
acter of microeconomics limits its scienti®c development: microeconomic
``theory's prediction and explanation of the choices of individuals
[cannot] exceed the precision and accuracy of commonsense explanations
and predictions with which we have all been familiar since prehistory.''
In fact, macroeconomic explanation and prediction is not only often
better, but may have more scope for improvability. An electric supplier
could not say when Mary Smith will switch on her oven, but it may know
pretty precisely how many kilowatts it must supply at a given time, based
on an aggregate analysis of past behavior. Insurance companies know
that whether an individual is, say, a smoker or obese matters probabil-
itistically to his chances of dying. But the company would go broke
trying to predict individuals' precise dates of death.11
It is important to remember that it is not macroeconomic theory that
supervenes on microeconomic theory, but macroeconomic reality that
supervenes on microeconomic reality. The disabilities of microeconomic
theory thus prove, at most, that there can be no automatic presumption
that microeconomics is more basic, because more successful, and that
macroeconomics is merely heuristic. The critical relationship is the
reducibility in principle suggested by (7) above. To begin to undermine
reducibility in the case of macroeconomics, it helps to note a crucial
disanalogy with biology. Reduction appeals to biologists because it
removes scienti®cally suspect teleological explanation common in evolu-
tionary biology and other functional accounts. The aim of reduction in
economics, however, is precisely the opposite: macroeconomics appears
mechanical and dehumanized, and the point of the program of micro-
foundations is reintroduce human decision-making as an explanatory
element. The point is to recover intentionality.
Kim's analysis posits two levels of properties that are (semantically at
least) distinct and then investigates how they must be related if one set is
supervenient on the other. Intentionality at the microeconomic level
undermines the distinctness of microeconomic properties from macro-
economic properties. Levy's argument (see section 1 above) that indi-
vidual economic actors will invariably make reference to social wholes
and aggregates is even more fundamental than he imagines. In evaluating

11 Both these examples are repeated verbatim from Hoover (1995a).

 Is macroeconomics for real? 239

the future individuals must form expectations about real prices and real
quantities. Independently of the uncertainty of the future, the Cournot
problem implies that it is impracticable to solve good-by-good, price-by-
price, period-by-period planning problems in all their ®ne detail. The
best that one is practically able to do is to work with aggregates. The
information on which these are based is fundamentally monetary.
Economic actors must use estimates and expectations of the general price
level and real interest rates to form any practical assessment of their
situations.
Hayek (1979, p. 62) writes:
in the social sciences it is necessary to draw a distinction between those ideas
which are constitutive of the phenomena we want to explain and the ideas which
either we ourselves or the very people whose actions we have to explain may have
formed about these phenomena and which are not the cause of, but theories
about, the social structures.
What Levy's argument demonstrates is that Hayek is mistaken, that how
people theorize about the economy is constitutive of macroeconomic
phenomena.12 Since people cannot theorize about certain sorts of
phenomena without appealing to macroeconomic categories ± that are
not themselves reducible to microeconomic categories ± the Cournot
problem introduces analytical constraints, not only in practice, but in
principle as well. The distinctiveness of the properties at the micro-
economic and macroeconomic levels is breached, undermining Kim's
analysis, because complete characterizations of the microeconomic must
include characterizations of the macroeconomic on the part of individual
agents.
To challenge the applicability to economics of the reductionism in
principle, implicit in Kim's analysis and in Rosenberg's application of
it to biology, does not challenge the notion that macroeconomics
supervenes on microeconomics. Kim's analysis is epistemological: it
argues that there must be laws that would permit us to draw
connections between the micro and macro levels. The point here is
ontological: even though macroeconomics cannot be reduced to
microeconomics as the program of microfoundations suggests, the
elements of macroeconomics could not exist without the substrate of
microeconomic individuals.

12 In contrast to Hayek, his fellow Austrian-school economist Mises (1943, p. 252) argues
that knowledge of economic theory can prevent the mistaken investments that fuel the
business cycle.
240 Micro, macro, and markets

4 Two arguments for the reality of macroeconomics

So far we have argued that the ontological status of macroeconomic

entities is problematic in the sense that, like other entities posited by
scienti®c theories, they are not part of our commonsense ontic furniture.
Furthermore, the nature of the relationship through which the elements
of macroeconomics supervene on the elements of microeconomics pre-
cludes direct reduction of the macroeconomic to the microeconomic,
even in principle. If macroeconomic entities exist, they cannot be said
therefore to exist only derivatively, despite their supervenience on micro-
economic entities. It remains to argue directly for the existence of macro-
economic entities.
The ®rst argument is based on the argument from manipulability
championed by Ian Hacking (1983, esp. pp. 22±4): ``If you can spray
them, then they are real.'' Hacking argues that convincing evidence of
the reality of the electron is found in experiments aimed at detecting the
existence of free quarks, in which niobium balls are charged by
``spraying'' them with electrons. The general point is that an entity
de®ned by a scienti®c theory has real existence when procedures that are
best understood as using the entity referred to by the theory as a tool to
manipulate other parts of the world, such as in laboratory experiments. I
take this argument to be related to the ``no-miracles'' argument for the
reality of scienti®c entities.13 The best explanation of why theories are
predictively successful, including successfully predicting the consequences
of using them to design experimental or engineering manipulations of the
world is that the entities posited by them in fact exist ± anything else
would be an inexplicable miracle.
It is common to denigrate the empirical success of economics (see e.g.,
Rosenberg 1992, pp. 18, 56, 112, 238, passim). It is true that economics
does not have the precision of physics or chemistry, although it is less
clearly inferior to meteorology, geology, climatology, and parts of
biology ± to name just a few of the less exact, but nevertheless scienti®c
disciplines. The reputation of economics for predicting poorly arises
partly because people seek unconditional forecasts (``what will happen
tomorrow?'') while economic theories typically predict only conditionally
13 MaÈki (1996, section 6) argues that the no-miracles argument and other arguments from
manipulability cannot be successfully applied to economics, even if they apply to physical
sciences. (There may, of course, be other arguments ± and MaÈki supplies some ± for
existential beliefs about economic entities.) The essence of my position is that
macroeconomics shares characteristics with physical sciences that microeconomics may
not.
 Is macroeconomics for real? 241

(``tomorrow X will happen if Y happens''). Quanti®ed economic relations

are at best locally stable: the precise estimate of the price elasticity of
demand for Volvos changes with changes in the range of alternative
brands and models, with changes in the proportion of academics to the
total population, and with changes in other background conditions.
Nevertheless, qualitatively stable relations are well established: e.g.,
demand curves slope down (i.e., when the price of Volvos rises, sales of
Volvos fall). And there is often enough local stability that useful
quantitative assessments are possible. Can irreducible macroeconomic
aggregates be manipulated as well?
The answer seems to be clearly yes. Consider the following example.
Almost every macroeconomic theory predicts that suf®ciently large
expansions of government expenditure will change (probably increase)
nominal GDP and the general level of prices.14 Different theories differ
in their precise understanding of the mechanisms. Similarly, no macro-
economic theory disputes the ability of the Federal Reserve to use its
ability to supply or remove reserves from the banking system to set the
level of the Federal funds rate (the rate at which one commerical bank
borrows from another overnight). The empirical evidence in support of
these propositions is also overwhelming. Now consider two irreducibly
macroeconomic aggregate entities: the real rate of interest (i.e., the
market rate of interest less the percentage change in the aggregate price
level (p)) and the yield curve (an aggregate relation portrayed as a graph
of market interest rates against time to maturity of the associated bonds).
Both the real rate of interest and the yield curve are synthetic aggregates
and both are entities with causal powers in some economic theories.
Every macroeconomic theory that I know predicts that actions that
increase the general price level or the Federal funds rate will shift the
yield curve upwards in the short-run. And, at least if the changes are
unanticipated, increases in the general price level will reduce the level of
the real interest rate. The empirical evidence for these effects is over-
whelming, and indeed are easily con®rmed by anyone willing to read the
Wall Street Journal regularly for a month. Just like the electron, some
macroeconomic aggregates not only can be controlled, but can be used to
manipulate other macroeconomic aggregates.
The second argument is related to the ®rst. Nowak (1980) and others

14 The caveat ``almost'' and the ambiguity over the direction both hinge on the ®nancial
market. If the interest elasticity of money demand were zero (empirically a false
supposition), there would be no change in prices; if the substitution effect on money
demand induced by an increase in government bonds ®nancing an increase in
government expenditure were large enough (again unlikely), the price level could fall.
242 Micro, macro, and markets

have argued that the principal method of constructing scienti®c theories

is idealization. Nowak's (1980, p. 29) paradigm idealization statement is:
If G(x) and p1(x) = 0 and . . . pk-1 = 0 and pk = 0 then
F(x) = fk(H1(x), . . . , Hn(x)) (8)
where Hi (i = 1, . . . , n) denote primary factors and the pj ( j = 1, . . . , k)
denote secondary factors. An idealized theory is one which picks out the
primary factors by setting the secondary factors to extreme values: zero
or, represented here, without loss of generality, as pj = 0.
Were G(x) a known and exhaustively complete theory of the phenom-
enon within its explanatory range such that one could accurately specify
each of the secondary factors that were set aside, then the distinction
between primary and secondary factors would in fact be be unclear,
because our complete knowledge of G(x) would allow us for example to
replace (8) with
If G(x) and H1(x) = 0 and . . . Hn71 = 0 and Hn = 0 then
F(x) = fn( p1(x), . . . , pk(x)). (9)
In the case of either (8) or (9), releasing the idealizing conditions (pk = 0
or Hi = 0) allows us to recover the complete theory, G(x). Idealization has
been reduced to a fancy name for an arbitrary selection of ceteris paribus
conditions or to a formal nesting relationship for theories.
Hoover (1994) argues that in an empirical context, the method of
idealization has power only if we recognize that not all of the idealizing
conditions can be explicitly stated. The claim to distinguish between
primary and secondary factors is then a claim that the primary factors
are the essence of the matter. Idealized theories thus aim to identify,
isolate and relate the real essences or causally effective capacities of
economic reality.15 The success of such an idealized theory then amounts
to an ontological claim for its primary factors.
That Keynesian macroecononomics could be cast as an idealization
that employs macroeconomic aggregates essentially is beyond doubt. The

15 Cartwright (1983, 1989) argues for realism with respect to causal capacities, but for an
instrumentalist interpretation of scienti®c laws. Laws are either literally false (``the laws
of physics lie'' ± to quote the title of Cartwright's (1983) earlier book) or are merely
phenomenal ± i.e., atheoretic regularities. Hoover (1994a) argues that if idealized models
represent essences, then phenomenal laws are necessary bridges to take the place of those
secondary factors that cannot in fact be identi®ed explicitly. MaÈki (1992) argues that
Nowak con¯ates idealization with isolation, which comprises idealization, omission, and
other techniques. To apply MaÈki's account we would have to say that the omission of
secondary factors amounts to a claim that retained primary factors are the essence of the
matter.
 Is macroeconomics for real? 243

major competitor to Keynesian macroeconomics today, new classical

macroeconomics, trades on an explicitly microfoundational approach.
There is, however, less here than meets the eye. The currently most
popular new classical macroeconomic theory is embodied in the real-
business-cycle model (see Hoover 1995b and forthcoming). The propo-
nents of this representative-agent model would like it to be interpreted as
an idealization from a complete Walrasian general equilibrium model of
the economy in which distributional issues are idealized out of the model
so that what remains is a one-agent, one-good, one-price representation
of the economy. This would work if the Walrasian model (the analogue
for G(x) in Nowak's schema) were both true and known in detail. At
least the second condition is false, which undermines the evidential basis
for the ®rst condition.
Empirically, far from isolating a microeconomic core, real-business-
cycle models, as with other representative-agent models, use macro-
economic aggregates for their testing and estimation. Thus, to the degree
that such models are successful in explaining empirical phenomena, they
point to the ontological centrality of macroeconomic and not micro-
economic entities. The appeal to the methods of microeconomics does
not in this case amount to the successful implementation of the program
of microfoundations, for they are but the simulacrum of micro-
economics. The relationship between models that are microeconomic in
form and their macroeconomic empirical implementation is metapho-
rical. The nature of metaphorical connection deserves futher exploration.
It is enough for the present purpose to understand that, at the empirical
level, even the new classical representative-agent models are fundamen-
tally macroeconomic in content.

5 The contingency of economic reality

Unless one is committed to a certain kind of apriorism, then what we

judge to be real depends on experience, on the nature of our theorizing,
and the success of our scienti®c and everyday practices. The best guess of
a scientist in 1789 would have been that phlogiston was real, although
most scientists would today say that it is not. The point is not that reality
is constituted by our theorizing: for a realist it is not only the case that
phlogiston is not real today, but that it was not real in 1789. The point is
rather that our knowing whether or not something is real is a scienti®c
fact like other scienti®c facts, which are established by argument and
evidence, and about which we may be mistaken. There is, therefore, no
244 Micro, macro, and markets

timeless, certain answer to the question in the title of this paper. What I
have sought to show in this essay is that the nature of microeconomics
and macroeconomics ± as they are currently practiced ± undermines the
prospects for a reduction of macroeconomics to microeconomics. Both
microeconomics and macroeconomics must refer to irreducible macro-
economic entities. These macroeconomic entities occupy ontologically
independent places in economic theory. To the degree that such theories
are empirically successful, the best account of these macroeconomic
entities is that they are real.

References
Blaug, Mark (1992) The Methodology of Economics: Or How Economists Explain,
2nd edition, Cambridge: Cambridge University Press.
Boland, Lawrence (1979) The Foundations of Economic Method, London: George
Allen and Unwin.
Cartwright, N. (1983) How the Laws of Physics Lie, Oxford: Clarendon Press.
(1989) Nature's Capacities and Their Measurement, Oxford: Clarendon
Press.
Cournot, Augustin (1838/1927) Researches into the Mathematical Principles of the
Theory of Wealth, Nathaniel T. Bacon (trans.), New York: Macmillan.
Ellis, Brian (1966) Basic Concepts of Measurement, Cambridge: Cambridge
University Press.
Friedman, Milton (1955) ``Leon Walras and His Economic System: A Review
Article,'' American Economic Review, 45(5): 900±9.
Hacking, Ian (1983) Representing and Intervening: Introductory Topics in the
Philosophy of the Natural Sciences, Cambridge: Cambridge University Press.
Hahn, Frank H (1965) ``On Some Problems of Proving the Existence of
Equilibrium in a Monetary Economy,'' in Frank H. Hahn and F. P. R.
Brechling (eds.), The Theory of Interest Rates, London: Macmillan.
Hayek, Friedrich von (1979) The Counter-revolution of Science: Studies in the
Abuse of Reason, Indianapolis: Liberty Press.
Hoover, Kevin D. (1988) The New-classical Macroeconomics: A Skeptical
Inquiry, Oxford: Blackwell.
(1994) ``Six Queries About Idealization in an Empirical Context,'' Poznan
Studies in the Philosophy of Science and the Humanities, 38: 43±53.
(1995a) ``Why Does Methodology Matter for Economics? A Review Article,''
Economic Journal, 105: 715±34.
(1995b) ``Facts and Artifacts: Calibration and the Empirical Assessment of
Real Business Cycle Models,'' Oxford Economic Papers, 47: 24±44.
(forthcoming) ``Quantitative Evaluation of Idealized Models in the New
Classical Macroeconomics,'' in Nancy Cartwright and Martin Jones (eds.),
Correcting the Model. Poznan Studies in the Philosophy of Science and the
Humanities, Amsterdam: Rodopi.
 Is macroeconomics for real? 245

Janssen, Maarten (1993) Microfoundations: A Critical Inquiry, London: Rout-

ledge.
Keynes, John Maynard (1936) The General Theory of Employment Interest and
Money, London: Macmillan.
Kim, Jaegwon (1978) ``Supervenience and Nomological Incommensurables,''
American Philosophical Quarterly, 15(2): 149±56.
Kirman, A. P. (1992) ``Whom or What Does the Representative Agent Repre-
sent?'' Journal of Economic Perspectives, 6(2): 117±36.
Lachmann, Ludwig (1976) ``On the Central Concept of Austrian Economics:
Market Process,'' in Edwin G. Dolan (ed.), The Foundations of Austrian
Economics, Kansas City, MO: Sheed and Ward.
Levy, David M. (1985) ``The Impossibility of a Complete Methodological
Individualist: Reduction When Knowledge is Imperfect,'' Economics and
Philosophy, 1(1): 101±8.
Lucas, Robert E., Jr (1987) Models of Business Cycles, Oxford: Blackwell.
MaÈki, Uskali (1992) ``On the Method of Isolation in Economics,'' Poznan Studies
in the Philosophy of Science and the Humanities, 25: 289±310.
(1996) ``Scienti®c Realism and Some Peculiarities of Economics,'' in R. S.
Cohen et al. (eds.), Realism and Anti-Realism in the Philosophy of Science,
Boston Studies in the Philosophy of Science, Vol. 169, Dordrecht: Kluwer,
pp. 425±45.
Marshall, A. (1920) Principles of Economics: An Introductory Volume, 8th edn,
London: Macmillan.
Mises, Ludwig von (1943) `` `Elastic Expectations' and the Austrian Theory of the
Trade Cycle,'' Economica, ns 10(3): 251±2.
(1949/1966) Human Action: A Treatise on Economics, 3rd edn, Chicago: Henry
Regnery.
Nowak, L. (1980) The Structure of Idealization: Towards a Systematic Interpret-
ation of the Marxian Idea of Science, Dordrecht: Reidel.
Robbins, L. (1935) An Essay on the Nature and Signi®cance of Economic Science,
London: Macmillan.
Rosenberg, Alexander (1985) The Structure of Biology, Cambridge: Cambridge
University Press.
(1992). Economics: Mathematical Politics or Science of Diminishing Returns?
Chicago and London: Chicago University Press.
Samuelson, Paul A. and Ryuzo Sato (1984) ``Unattainability of Integrability and
De®niteness Conditions in the General Case of Demand for Money and
Goods,'' American Economic Review, 74(4): 588±604.
Weintraub, E. Roy (1979) Microfoundations. The Compatability of Microeco-
nomics and Macroeconomics, Cambridge: Cambridge University Press.
13 The possibility of economic
objectivity

DON ROSS AND FRED BENNETT

Among philosophers of science, the phrase `positive economics' is apt,

these days, to produce snickers. Economists tend to be less embarrassed
by the term ± Lipsey and Crystal, after all, use the phrase as the title of
their bestselling introductory text ± but only slightly. In a 1994 paper,
Samuel Weston defends the utility of a positive/normative distinction in
economics, but in a strikingly tepid way. He concedes the impossibility of
`a purely positive, that is, value-free or ethically neutral economics' (p.
5), but argues, following Machlup (1969), that economists should
preserve the concept of relative normativity in order to encourage self-
consciousness about the more overtly ideological in¯uences on their
theories and policy recommendations. For Weston, the positive/norma-
tive distinction is based on purely pragmatic considerations, and is to be
drawn despite the fact that it has no sound epistemological or ontological
basis. This sort of defence of the distinction has polemical import only
against the most extreme denial of the possibility of objectivity. Such
extreme denial, like most forms of philosophical scepticism, must be
motivated on the basis of highly general epistemological theses. Engaging
it, therefore, is unlikely to turn on any issues that arise from within the
practice of economics, or of any other special science. This is illustrated
by Weston's own argument, much of which could be applied with only
minor amendments to any science that suggests practical advice ± which
is to say, to any science at all.
In the following discussion, we wish to provide a much stronger
defence of the possibility of positive economics, from a perspective that
sees the challenges to economic objectivity as arising from, and needing
to be answered with regard to, particular problems that arise within
economics. This does not mean that broad philosophical issues are to be
bypassed. Indeed, some quite general problems, properly called `meta-
physical,' will be the focus of much of our attention. Furthermore, we
expect that our defence of positive economics, if successful, could serve

246
 The possibility of economic objectivity 247

as a model for a defence of the possibility of objectivity in a number of

other sciences. However, our argument will not be a reply to general
epistemological scepticism, because we take that position to be unanswer-
able ± and uninteresting, just because neither the thesis itself nor any
attempted refutations of it depend on knowing anything speci®c about
the sciences to which it is meant to be applied. If general epistemological
scepticism is true, then inquiry should simply stop; but since we could
never be sure that scepticism is true, we are best off continuing to pursue
scienti®c inquiry on the assumption that it is not. This answer will not
suf®ce, however, against less sweeping sceptical theses advanced on the
basis of problems arising within speci®c sciences. Among recent critics,
Rosenberg (1991) and Dupre (1993) have advanced precisely such theses
concerning economics.
Weston's paper will serve as a useful foil for getting the discussion
started. He considers four grounds for supposing that value-free eco-
nomics is impossible. First, there is the fact that all scienti®c activity
presupposes that truth-seeking is valuable activity in the ®rst place, even
in circumstances where this activity threatens to undermine other valued
ends and pursuits. Second, because economics arises from recognition of
the fact of scarcity, economic theory is almost always directly relevant to
questions of distribution, and these questions are, in turn, among the
most fundamental issues of political and moral debate. Third, certain
assumptions of value are built into sciences by virtue of the fact that their
central terms are irremediably infected with value-leaden connotations.
This is particularly true in economics, which inherited much of its basic
vocabulary from utilitarian and proto-utilitarian political philosophy.1
Fourth, economists justify much of their activity by reference to the
policy advice that they are expected to give; and since there is no such
thing as a distributively neutral economic policy, any research directed
towards producing policy includes a normative element.
Let us consider the extent to which these four problems genuinely
undermine the possibility of positive economics. The ®rst and the third
are of least interest, since they are not based on distinctive aspects of
economics, and can be applied with equal justi®cation to any science.

1 `Utility', which is arguably the single most fundamental concept of contemporary

economics, is the best example of a term which has passed into the discipline through
political philosophy, undergoing several substantial changes in meaning over the course
of this history (see LaCasse and Ross 1991; Ross 1999). Weston, citing Myrdal (1954,
1958), gives as examples of value-leaden terms `economic integration', `productivity',
`equilibrium', `balance', and `adjustment'. The point of calling such terms `value-leaden'
should be evident: `adjustments' are naturally assumed to be for the better, and to call
something `productive' or `balanced' is implicity to offer approval of it.
248 Micro, macro, and markets

This may not seem immediately to be true of the third complaint, which
directs our attention to the fact that economics, unlike most sciences,
takes as fundamental many terms which have overtly normative uses
outside of their technical context. Notice, however, that given Weston's
conception of the pragmatic point of the positive/normative distinction,
this should make relative objectivity easier to achieve in economics than
in other sciences. If the terminology used in, say, biology, is infected with
assumptions of value to which its users are socially conditioned to be
blind (as feminist critics such as Harding (1986) have charged), then, on
Weston's reading, this should leave them epistemically worse off than
economists, who can scarcely be similarly unaware of the normative
relevance of concepts derived from moral philosophy. Among those who
attack scienti®c objectivity on the basis of some version or near relative
of the Sapir-Whorf hypothesis, almost none imagines that any science
could be rendered objective by being purged of normative terms, since
they hold such purging to be impossible. The best one can do, according
to such views, is maximize self-awareness of the normative assumptions
built into and reinforced by one's language. Therefore, it is hard to
imagine why we should think that the use in economics of overtly
normative terms constitutes a special challenge to its objectivity.
Weston's second and fourth problems do seem to direct our attention
to distinctive aspects of economics. The two problems are closely related.
Indeed, they may be seen as expressions of a single, slightly more general
problem: to what extent does the individuation of an economic domain
in the ®rst place result from concerns with distributional fairness (or the
lack of it) and the need for policies that regulate and in¯uence distri-
bution? If one believes that there would be no domain for economics in
the absence of certain normative interests, then one is likely to suppose
that economists' selection of problems as meriting their attention will
re¯ect those interests, especially where it is expected that study of a
problem will eventuate in policy recommendations. Here, indeed, is a
fundamental challenge to objectivity. To say that a particular aspect of
the world is imposed on it from an interest-driven perspective, rather than
discovered, seems conceptually equivalent to saying that the aspect in
question is not an objective feature of the world.
In preparing this issue for examination, we must proceed carefully with
the distinction between imposition and discovery. Saying that something
would not be an objective feature of the world if it were accessible only
from an interest-driven perspective is not equivalent to claiming that in
order to be objective, a feature would have to be evident to investigators
who cared about nothing in particular. Such investigators appear to be
impossible, since investigation must be motivated and organized by some
 The possibility of economic objectivity 249

set of interests or other. This point warns us against taking the question
before us as an historical one. It is quite plausible to think that had our
environment never presented people with scarcity, it would never have
occurred to anyone that there might be such things as economic general-
izations. But if this were the sceptic's point, then it would fail, after all, to
tell in any special way against economics. There would likely be no
science of geology if we were intelligent clouds of dust particles living in
open space.2 If the sceptic's concern is to offer an interesting basis for
suspicion about the possibility of positive economics, then she must
identify objectivity with perspective-independence in principle. What does
`perspective-independence in principle' mean? Suppose that we identify C
as the set of cognizers whose perceptual capacities and neural complexity
(biological or arti®cial) permits recognition of the patterns in data that
would be generalized over by the set of scienti®c theories `in the limit'
(i.e., the set of scienti®c theories that would be accepted if science were
®nished. The idea goes back at least to Pearce; for a clear re®nement of
the notion, see Friedman (1979).) A science fails to describe a set of
objective features in the world, the sceptic might then say, if its domain
could not be consistently individuated from outside the perspective of
some set of cognizers smaller than C. Interpreting the sceptic this way
does not, at least prima facie, saddle her with an unreasonably heavy
burden of argument, to judge from the sorts of positions on questions of
fundamental ontology that have been popular among philosophers.
Reductionists, in the venerable sense of Oppenheim and Putnam (1958),
should be comfortable with this conception of objectivity, since such a
conception seems to be a necessary part of the motivation for reduc-
tionism in the ®rst place. As for those ± the majority these days ± who are
non-reductionists about the special sciences, most are functionalists of
one sort or another. Though a functionalist about the objects of
quanti®cation of a given special science S denies the reducibility of S on
epistemic grounds, most functionalists seek to preserve the variety of
objectivity de®ned above for S by endorsing one sort or another of
supervenience hypothesis (see Kim 1989; Marras 1993) about the proper-
ties featuring in S's generalizations. The discussions in Savellos and
YalcËin (1995) reveal an array of competing notions about what super-
venience precisely amounts to, but we need not concern ourselves with
their differences here; the point, for now, is that all supervenience
hypotheses are motivated by the concern that a given putatively autono-
mous special science inherits in-principle objectivity from a speci®able

2 The counterfactual here seems to us a bit less implausible than the idea that intelligence
could arise in the absence of the selection pressures produced by scarcity.
250 Micro, macro, and markets

relation of either causal (though accidental) or nomic co-variance

between the properties it recognizes and those recognized at a base level
with respect to which in-principle objectivity is already presumed. It
might be objected that appeals to supervenience are in fact motivated, as
was Oppenheim and Putnam's sketch of reduction, by reference to
concern with the unity of science, rather than with objectivity. However,
the physicalism presumed by both projects is itself best motivated by the
conviction that if anything exists objectively, it is physical properties and
objects. Furthermore, functionalists of the sort who are most interested
in supervenience generally believe that uni®cation of the sciences is an
unlikely or impossible ambition in any case. In general, then, most
attempts to show that the objects featured in the domain of some special
science are not part of objective reality generally appeal, at least
implicity, to the idea that they neither reduce to nor supervene upon any
types of objects which the sceptic does take to be real. (The objects with
respect to which this sort of debate has been loudest among philosophers
are, over the past couple of decades, the propositional attitude concepts
used in the generalizations of psychology.)
The foregoing remarks might suggest the following conclusion where
economics is concerned. To show that economics tracks objectively
existing objects or processes, one would have to show that the properties
in terms of which these objects or processes are individuated either
reduce to or supervene upon other properties whose objective existence is
less contentious. Note that the base properties need not be physical; if
microeconomic properties ultimately reduce to or supervene upon
physical properties, they presumably do so by way of psychological
properties. Now, of these two possibilities, reduction or supervenience,
only one actually seems to be a live option. If one presumes that
economics tracks an objective domain, than few sciences offer more
compelling instances of the problems with reductionism. To borrow an
example from Fodor (1974), it seems inconceivable that the economic
type `monetary exchange' could turn out to be coextensive ± let alone
nomically coextensive, as required by reductionism ± with any physical,
or, for that matter, psychological type. The same can be said of virtually
all of the types to which macroeconomic generalizations, and axioms in
microeconomic theories, make reference. The objectivity of economics
thus appears to be straightforwardly incompatible with reductionism.
This, in turn, leaves us with a nicely sharp central question: does the
objectivity of economics depend on the claim that economic properties
supervene on physical or (more plausibly) psychological properties?
Many philosophers hold views that would lead them to answer `yes' to
this question. This seems, for example, to be an essential part of the basis
 The possibility of economic objectivity 251

for the scepticism about the possibility of economic objectivity expressed

by Dupre (1993). Economics, according to Dupre (1993, p. 260), is
contaminated from the beginning by normative assumptions because the
types to which it seeks to assign properties ± unemployment being his
example ± are invented rather than discovered, in the sense that the
typology is imposed on the world by economists, rather than found in it,
as a result of ideological or more narrowly political motivations. DupreÂ's
second claim does not derive from a review of case studies in macro-
economics. Rather, it is an application of a book-length metaphysical
argument to the effect that nothing like unemployment could be dis-
covered. A crucial part of this argument is a case against the claim that
special sciences such as economics are (as it were) `metaphysically
legitimated' by appeals to supervenience hypotheses (pp. 96±8). I will
not review DupreÂ's particular criticism of the plausibility of establishing
supervenience claims, because I ®nd it unpersuasive. However, I am in
agreement with him, and with Kim (1989), that supervenience is less an
alternative to reduction than a re®nement of it. More importantly,
philosophers have not yet provided an adequate formulation of the
claimed asymmetry of the supervenience relation (i.e., the assumption
that mental properties should depend on physical properties in some
way, but that physical properties should not, in general, depend on
mental properties) (Marras 1993). The fact that we can therefore not yet
be said to fully understand the concept is good grounds for hoping that
the possibility of economic objectivity does not depend on our being able
to show that economic properties are supervenient. Finally, and most
decisively, attempting to ground economic objectivity in the claim that
economic properties supervene on psychological ones would commit us
to a cure worse than the disease. Since the fundamental axioms of
microeconomic theory are, famously, utterly unrealistic as generaliza-
tions about individual behaviour and motivational structures, commit-
ment to seeking a supervenience base for microeconomics in psychology
would entail destroying microeconomics in order to save it. And where
else could one even imagine looking for a supervenience base for micro-
economics?
The view that objectivity in special sciences requires supervenience
rests, we contend, on a failure to appreciate that reduction and super-
venience do not constitute an exclusive disjunction of routes to objec-
tivity. In earlier work (Ross 1995; Ross and LaCasse 1995), one of us has
argued, with direct reference to economic examples, that economics
should be viewed as studying what Dennett (1991) has called `real
patterns'. We do not have space to reiterate that argument here. What we
wish mainly to do in the present discussion is to provide a fuller account
252 Micro, macro, and markets

than Dennett has done of the sense in which real patterns exist objec-
tively.3 This will enable us to articulate the most reasonable sense in
which economic objectivity is possible and to explain why philosophers
have so frequently supposed that it is not possible. We will then offer
some examples of economic patterns which seem to us to be undeniably
real in the sense that we will defend.
Dennett's `Real Patterns' is his mature response to the charge, levelled
from many quarters over the years, that his view on the ontology of
intentional states such as beliefs is a form of instrumentalism. Speci®c-
ally, Dennett holds that intentional states are ascribed to certain sorts of
systems ± intentional systems ± in the course of explaining and predicting
their behaviour. Since intentional states are thus attributed properties of
whole systems according to Dennett, his view contradicts the most
common sort of realism about beliefs and desires, according to which
they are either nomically coextensive with, or supervene upon, particular
neural states or disjunctions of neural states. To many, this has appeared
to involve him in a denial of realism about intentional states altogether;
they seem to emerge as at best idealizations, like frictionless planes, or at
worst mere ®ctions whose invention serves predictive purposes, rather as
epicycles do in Ptolemaic astronomy. Dennett's early denials of this
charge (e.g., Dennett 1979) usually seemed unconvincing, and the reason
for this, we can now see, is that he in fact cannot accommodate his view
of intentional states to any of the traditional varieties of realism. What is
needed is a non-traditional variety of realism, and this Dennett duly sets
out to provide in `Real Patterns'.
Before beginning the account of real patterns, it is worth drawing
attention to the similarity between the polemical situation that called
forth Dennett's paper and the controversy over the status of economics
with which we are concerned here. From the time of his earliest work,
Dennett has talked of intentional properties as emerging when the
intentional stance is assumed towards a system whose behaviour exhibits
too much differential sensitivity to its environment to permit it be
tracked at the level of physical causes and effects. It is this reference to
stances, more than anything else, that has made the charge of instrument-
alism against Dennett appear inescapable. He seems to be admitting,
from the outset, that intentional states exist only under a certain
perspective that can be taken towards ethological phenomena. Now, the
charge against economic objectivity advanced by Dupre and others, and

3 The full presentation of the relationship between our extension of Dennett's view and
Dennett's own self-interpretation will appear in Ross (2000) and in Dennett's (2000)
reply.
 The possibility of economic objectivity 253

accepted by Weston, is essentially similar: that the objects of economics

also emerge only under a certain perspective, in this case one that is
motivated by a set of socio-political values. In effect, the entire discipline
of economics is accused of resting on instrumentalism. This accusation
gains credence from the fact that many economists, following Friedman
(1953), have explicitly endorsed instrumentalism as an account of their
methodology, failing, apparently, to recognize that this move entails a
surrender of their ambition to be engaged in positive science. In light of
this, it is clearly worthwhile asking whether Dennett's way out of his
dilemma can also serve as the basis for a defence of positive economics.
We contend that it can.
We will now reconstruct Dennett's idea of a real pattern. The word
`reconstruct' should be emphasized here. Dennett provides a host of
examples to demonstrate the utility of granting existence to patterns of a
certain sort, but he does not provide explicit criteria for the reality of a
pattern. These must be arrived at by induction on his examples, and on
the basis of his remarks about the conditions under which reference to a
pattern is necessary in order to track some data. We will work from the
idea of a pattern in general, and then inquire as to what properties a
pattern must have if it is not to be shaved away by Occam's razor.
Patterns, let us say, are organizations of data from perspectives, where
the structural relations among the data that a pattern recognizes are
suf®ciently stable over time to permit some process or processes to be
tracked from the relevant perspective. Patterns so de®ned come cheap;
the world teems with them. This is not suf®cient reason in itself, however,
for denying reality to them. In deciding whether we should regard a given
pattern as real, the crucial question, as noted above, is this: is it
something that is imposed, on the data, or is it more accurately viewed as
being discovered in the data? In the case of some patterns ± for example,
the grand `cycles' in human affairs that certain sorts of historians are
disposed to construct, or macroeconomic cycles of the kind exempli®ed
by `Kondratiev waves' ± scepticism (which is maintained by the majority
of historians in the ®rst case, and by the majority of economists in the
second) rests largely on intuitions to the effect that the patterns are
imposed on the data rather than discovered in it.4 Is this intuition
reliable? Consider the second example. Kondratiev waves are ®shy
scienti®c objects because they are a function of a periodicity choice that
is arbitrary. We know, as matters of mathematical fact, both that an
in®nite number of periodicities are available for the organization of data

4 As noted above, this is the position maintained by Dupre (1993) with respect to all of the
objects of economics.
254 Micro, macro, and markets

in a series, and that, given this in®nity, regularities will appear in some of
them even if the series is generated by a chaotic physical process. To
predict a regularity given a previously untested periodicity and then ®nd
it there is epistemically impressive; but to merely report discovered
regularities is not, and does not in itself justify the construction of
theories to explain them. Notice, however, how natural it is here to speak
of `mere discoveries'. Our intuitions are correct in their judgement that
something is `imposed' on the data by long-wave theory, but it is not the
waves themselves that are imposed. After all, once one has chosen the
periodicity and the variables of measurement, the question of the shape
of the curve through the data is a factual one. Kondratiev waves are `in'
the data and are not imposed on them; what is imposed is the assumption
that tracking these waves is of epistemic signi®cance.
If these re¯ections lead us to admit the reality of patterns, however,
then we appear to be stuck with an in®nite number of real objects. This
does not represent a direct violation of Occam's principle. That principle
± or, at least, the version of it that is a justi®ed part of scienti®c
rationality ± does not concern the number of entities that we hold there
to be; rather, it concerns the need for limits on recognizing the existence
of particular posited entities. (In its most modest version, it is simply the
acknowledgement that any ontological principle that excludes nothing is
worthless.) The Kondratiev wave through the data to the present time is
not posited but found. However, part of what is implied in saying that
the discovery is without epistemic signi®cance is that we have no good
reason to suppose that the pattern will continue through the next period
in the series. In this respect, the wave violates projectibility requirements
on the objects of scienti®c generalizations.5 This was always true of it,
despite the fact someone using it to predict the future before the most
recently completed period would have been fortuitously right. Here,
®nally, is a reason for denying reality to the pattern, and holding it to be
a mere artifact: there is not, and never has been, a perspective from
which the pattern is stable.
At least two notions employed in the above analysis require clari®ca-
tion: we must say what we mean by `projectibility', and explain our
appeal to `perspectives'. Let us begin with projectibility. We are per-
suaded by the arguments of Cartwright (1983) that sciences do not
produce true nomic generalizations; we thus cannot de®ne projectibility

5 Of course, we do not wish to have to say that only patterns that continue to be
instantiated were ever real. Projectibility to a next unanalysed case in some logically
structured sequence of cases is the relevant necessary (though not, as we shall see,
suf®cient) condition here.
 The possibility of economic objectivity 255

in terms of laws. Cartwright (1989) has also argued, drawing on a rich

analysis of the history of econometrics, that the generalizations implicit
in privileging some curves through data over others are parasitic on the
identi®cation of causal relations holding among particulars; generaliza-
tions are then summary approximations of the quantitative causal
propensities of particulars that have been arranged into types. This helps
us to re®ne our diagnosis of what is wrong with `long-wave' theories in
macroeconomics: they report mere correlations, unjusti®ed by any
suggestions, let alone evidence, as to what causal relations among
particulars might explain the observed correlations in question. Note,
however, that following Cartwright in denying that the generalizations
produced by sciences are laws does not entail the conclusion that arriving
at generalizations is not an important goal of sciences. Furthermore, it
does not eliminate the signi®cance of projectibility. We will discuss the
appropriate range of the projectibility requirement in the course of
explaining our reference to perspectives. As far as the notion itself is
concerned, we shall simply say this: a generalization is projectible just in
case our discoveries of causal relations ground con®dence in, and help to
explain, the prediction that it will apply approximately to a (not
necessarily in®nite) run of future or counterfactual cases in which the
relevant causal relations obtain.
We now expand on our reference to the role of perspectives in
identifying real patterns. To `track a real pattern' is simply to recover, in
some encoding, the information instantiated in the pattern. Information
must ¯ow through channels between sources and receivers; hence, a
perspective is simply the receiving end of an informational channel, that
is, any point along the channel where noise (i.e., increased entropy) has
not so degraded the information at the source that it has ceased to be
recoverable. Where any scienti®c generalizations are concerned, the
scope of existence claims must be wider than the actual world, since we
do not want to restrict existence to patterns that contingently occupied
perspectives have tracked or could track. On the other hand, the set of
logically possible perspectives is too wide, since this will lead to a reality
criterion that will exclude almost nothing. Let us say instead that what is
relevant is the set of physically possible perspectives; put more precisely,
the set of perspectives available in the possible worlds that are nearby
according to physics. This is a very large set, to be sure, but it is not
in®nite, since it is restricted by physical limitations on information ¯ow.
There is, for example, no perspective from which an occupant can track
events occurring outside its light-cone, nor a perspective from which an
occupant can track patterns whose processing at speeds near that of light
would require computational resources of unbounded size. Nor can
256 Micro, macro, and markets

information ¯ow across singularities on the time-like dimension (e.g., the

Big Bang) or a space-like dimension (e.g., black holes).
We are now ready to state a necessary condition on the reality of a
pattern: if a pattern is real, then it must be recoverable from at least one
physically possible perspective. This condition is clearly not suf®cient,
however. It fails to exclude Hegelian historical patterns or Kondratiev
waves,6 or such artifacts as the arbitrary conjunction of two stable
patterns. Occam's principle does forbid our granting reality to patterns
that are epistemically redundant. Therefore, we add a second necessary
condition (which also incorporates our de¯ationary understanding of
projectibility), this one partly based on a direct suggestion of Dennett's:
if a pattern is real, then it must encode information about at least one
other type of structure S where that encoding is more ef®cient, in
information-theoretic terms, than the bit-map encoding of S, and where
for at least one of the physically possible perspectives under which the
pattern is projectible, there exists an aspect of S which cannot be tracked
unless the encoding is recovered from the perspective in question. The
two necessary conditions just stated are, we maintain, jointly suf®cient
for the reality of a pattern.
It should be clear that patterns de®ned according to the above criteria
need not supervene on any `lower-level' objects or sets of objects. Indeed,
the very idea of `levels' is undermined by identifying existence with
pattern-reality. Since any informational encoding more ef®cient than a
bit-map encoding involves compression at the receiving end of the
channel, it is possible to recover various aspects ± that is, a multiplicity of
patterns ± from any set of brute events that are reliably connected by
causal relations. Thus a given set of events may be simultaneously
tracked, to equally essential but different epistemic purposes, from a
perspective that picks out (e.g.) psychological types as real patterns and a
perspective that picks out economic types as real patterns. We may thus
speak of `an economic perspective' without implying that that perspective
could be voluntarily discarded without loss of objective knowledge once
an actual perceiver comes to occupy it. Nor does reference to `sets of
events' presume ultimate ontological bedrock; sets of events are them-
selves individuated as real patterns from some other perspective, perhaps

6 We should make clear that, as naturalists, we do not presume that the reality of these
patterns, or any others that are suggested, should be rejected on a priori grounds. We
have already discussed our basis for suspicion about Kondratiev waves above. As for
Hegelian historical patterns, the grounds for insisting that an ontological principle must
sanction their exclusion are empirical: Hegel's data were notoriously selective, interpreted
through the lens of a highly idiosyncratic metaphysics, and cannot be said to have stood
the test of predictive success.
 The possibility of economic objectivity 257

physics. The Dennettian view, as we have enlarged and interpreted it, is

consistent with the primacy of physics in just the sense, we would argue,
that the actual practice of science incorporates materialism: all sciences
save mathematics and logic con®ne their attention to the class of
physically possible worlds, and physics is the science that de®nes this
class. This is why special sciences may not violate the generalizations of
the physics contemporary with them, but no symmetrical limitation
holds in the other direction.
We hope that the motivation for referring to the patterns satisfying the
Dennettian criteria as `real' is now clear: nothing subjective enters into
determining which patterns meet the criteria. What else could possibly
motivate a denial of reality to such patterns other than intuitions about
physicalistic reductionism? But these intuitions are just what is at stake.
They are called into question precisely by the fact that they render it
deucedly hard to account for the patterns of explanation and theory
development in special sciences such as psychology and economics.
Naturalists should not require sciences to respect a priori philosophical
preconceptions such as reductionism; the burden of argument must lie
with the philosophical critics of scienti®c practice. Thus against both
reductionists and those, like DupreÂ, who push particular sciences into
defensive corners by assuming that only reduction or supervenience can
legitimate them, it is suf®cient to merely produce a viable alternative.
The ontological view that we have derived from Dennett's suggestions,
we contend, constitutes such an alternative.
While the above considerations seem to us to establish a very strong
presumption in favour of the possibility of positive economics, one
naturally wants to add something that suggests the actuality of positive
economics. Only very timid claims can rest everything on burden-of-
proof arguments. So what is needed are some motivations for believing
that at least many objects of economics are real patterns. In other work
(Ross 1994; Ross and LaCasse 1995), one of us has provided detailed
examples of real patterns identi®ed by microeconomics. Here, we will
present what seem to us to be some real patterns which require organiza-
tion of data in terms of macroeconomic patterns.
Could someone be said to be in good epistemic touch with the world if
they could not track the type `currency'? Perhaps it will be objected that
currency is plausibly supervenient. But on what disjunction of physical
tokens might currency futures supervene? What about the patterns that
relate unemployment levels, in¯ation rates and the currency supply?
Which tokens could even be proposed as possibly ¯eshing out their
supervenience base? Of course, we do not suppose that we can get far in
persuasion by means of rhetorical questions. We will thus revisit the
258 Micro, macro, and markets

problems with supervenience accounts after reviewing our examples.

Their point is to suggest patterns that seem to have a deeply robust
reality, based on the fact that near-certain disappointment awaits any
economic manager who fails for long to track them, and that indepen-
dent worldly pressures will force them to the attention of anyone whose
personal values incline him to discount them, as Mr Chernomyrdin in
Russia discovered during his time as Prime Minister, and as the
tempermentally Marxist-inclined ANC government of South Africa has
recognized since assuming power. Appeal to such cases will likely strike
many readers as being suspiciously reminiscent of the sorts of `no free
lunch' aphorisms tossed glibly about by ®scally conservative politicians
and in corporate boardrooms. The illustrations are not taken to establish
our philosophical point by themselves, however. We should not coyly
pretend that we are not sympathetic with many of the common aphor-
isms. But the aphorisms are typically uttered in a context of philosophical
naãÈveteÂ. If someone says that `there is no free lunch', do they mean that
there are no free lunches as a function of someone's or some instituional
complex's policy preferences, or that there could not be free lunches as a
matter of objective fact? Clearly, the ahporisms are often used in both
senses. Our illustrations below are of patterns which plausibly exist
independently of anyone's policy decisions. Having already presented, in
the discussion above, a philosophical perspective under which the
existence of such types of patterns can be maintained non-mysteriously,
we now defend the existence of some particular patterns by applying a
strategy familiar in the literature on scienti®c realism. It is excellent
evidence that something is real if attempts to pretend that it isn't
consistently lead to frustration, and if this `kicking back' by the world
occurs through a small set of causal mechanisms of which a comfortable
level of theoretical understanding is available. It is highly relevant to our
case for their independent reality that these patterns are not artifacts of a
world whose distributional structures were designed according to a
particular ideology; that is, macroeconomic patterns were not brought
into being by economists. Finally, these are cases of patterns which, if not
tracked, would leave us epistemically impoverished with respect to our
disinterested understanding of the world. To reiterate an example one of
us (Ross 1994) has discussed elsewhere, and with respect to which little
or nothing of normative policy-relevance is at stake, one cannot under-
stand important features of civilizations, such as ancient Greece and
Rome, that themselves lacked the economic perspective unless one brings
those economic patterns to bear on one's analysis of them. Without the
economic perspective, for example, we would miss the contribution to
 The possibility of economic objectivity 259

the collapse of the Roman Empire made by the ruinous in¯ation that was
caused by the continuous wage-extortion of the legions.
We shall, then, consider some examples of macroeconomic patterns
whose reality is suggested by the fact that policy-makers, in pursuit of
normatively based goals with respect to distributional fairness or other
political considerations, must work around them, being unable ± by
virtue of their objective reality ± to wish or legislate them away. We have
deliberately chosen examples whose reality is controversial, in the sense
that governments trying to acknowledge them must overcome pressures
from publics or sectors that do not acknowledge their reality. Frequently,
of course, governments have either failed to overcome these pressures,
have themselves been victims of subjectivist fallacies, or perhaps have
been fully aware of the reality of the patterns, but have felt obliged by
social concerns to pursue policies which, to those with different norma-
tive agendas, make governments appear to be ignorant of economic
reality. Clearly, however, some governments have simply denied
economic reality. When Victor Geraschenko became Chair of Russia's
Central Bank in 1992, he declared on television that `It's impossible to
apply economic theory to Russia' (Rosett 1993). As Hoover (1995,
p. 732) says of this remark:
it rightly strikes most economists of widely differing political persuasions as
absurd: whatever theory of the relationship of central bank behavior to hyper-
in¯ation that they subscribe to, they think it applies as surely to Russia as to the
United States, Germany or Papua New Guinea. Even Marx, who would place his
economic analysis at the service of a particular political philosophy, does not see
the truth of his analysis as depending on the truth of that philosophy.7
For reasons to be explained at the end of the paper, which result from
the current state of macroeconomic theory, we will not be able to name
the patterns whose existence is suggested by the non-malleability of
macroeconomic forces for which we will be arguing. Our point, however,
is that the goal of macroeconomic science should be to identify the
patterns responsible for this non-malleability.
The failure of political leaders either to accept, or, in other cases, to act
in the light of, the reality of macroeconomic patterns or changes in
existing patterns has not been uncommon in diverse sorts of political
systems. A commonly cited example is the response (or lack thereof ) of
®scal policies throughout the world to changes which occurred in the
international economy subsequent to the rapid rise in energy prices in
1973. N. Roulini and J. D. Sachs (1989) reviewed the ®scal performance
of the major OECD countries in the ®fteen years following the energy
7 We are grateful to an anonymous referee for bringing this example to our attention.
260 Micro, macro, and markets

shock and concluded that political considerations had been allowed to

determine the economic response: `We suggest that in several countries
the slow rate at which the post '73 ®scal de®cits were reduced resulted
from dif®culties of political management in coalition governments'
(p. 903). But political reasons for ignoring real economic patterns are not
limited to situations involving coalition governments; they are, perhaps,
endemic to most normal democratic situations. Alesena and Tabellini
(1987) argue that `when political power alternates randomly between
competing parties, each government will be tempted to leave a legacy of
high debt for its successor, whose spending priorities it is not likely to
share' (p. 909). With regard to the speci®c circumstances of the 1980s,
Roulini and Sachs attribute `the onset of large de®cits to the growth
slowdown and rise in unemployment after 1973, as well as the sharp rise
in real interest rates after 1979' (1989, pp. 909±10). But it was not the
change in the circumstances itself that caused the ®scal problem: it was
the actual ®scal policies pursued in the face of the changed circumstances.
Irrespective of the reasons for the policies chosen ± failure to recognize
the reality of the pattern, or, alternatively, the lack of the political
capacity or will to choose different policies ± the result was the same:
the growth of real government spending in the years just after 1973 was largely
`uncontrollable' in the sense that previous spending commitments based on pre-
1973 economic conditions were politically dif®cult to adjust for several years . . .
Another reason for induced de®cits was the intentional application by some
countries of Keynesian aggregate demand policies in the face of the growth
slowdown . . . Right or wrong many governments are loath to raise taxes or
lower government spending during a recession. (Alesena and Tabellini 1987,
pp. 919±20)
Some countries ± notably Germany, Japan and the UK among the major
economies ± did recognize the new reality and organized their ®scal
affairs to stabilize the ratio of public debt to GDP; others did not.
A salient example of a country that did not adopt new policies in the
face of the reality of changes in economic circumstances is Ireland. Even
prior to the oil shock of 1973, Ireland had relied heavily on borrowed
funds to support its development plans. As the OECD noted in 1971:
The key role of public investment in the development process has had a marked
impact on public ®nance. With capital expenditure rising rapidly . . . a public
sector de®cit equivalent to some 10 to 15 per cent of GNP has had to be ®nanced
. . . In consequence, there has been a steady upward drift in outstanding national
debt. (OECD 1971, p. 31)
But it was the failure to change policies in the face of the real change in
the global economy that occurred in the 1970s which caused the problem
 The possibility of economic objectivity 261

to escalate. By 1975, the OECD indicated that de®cits were being

incurred to ®nance current consumption as well as capital expenditures:
`Re¯ecting the efforts to cushion the rise in unemployment last year and
this, the public sector de®cit has risen sharply to about 20 per cent of
GNP' (OECD 1971, p. 5). By the early 1980s the problem had taken on
suf®cient magnitude that the Irish government acknowledged the need to
address the ®scal imbalance; but action did not follow words. As the
OECD wrote in 1982:
A principal objective of [Irish] governments since 1978 has been to tighten the
setting of economic policy. But against the background of a depressed inter-
national economy the scale of retrenchment originally sought has not been
achieved . . . the main explanation lies in overruns on current expenditure plans.
(OECD 1982, p. 26)
The inevitable logic of the pattern took hold as debt service payments
as a percentage of current expenditure rose from 17.9 in 1975 to almost
24 in 1982. The effect of compounding continued so that by the end of
1984 outstanding government debt amounted to 128 per cent of GNP
with the foreign component comprising about 54 per cent of GNP
(OECD 1985, p. 30). Aside from the effect on budgetary policy of the
requirement to devote approximately one-third of government revenues
to service the debt, the need to service the foreign component had
gradually come to erase the original bene®t of the foreign borrowing,
namely, increased capital availability:
The initial impact of external government borrowing was to increase the
resources available to the country to consume or to invest. As a result, high
public sector de®cits did not deprive the private sector of investible funds . . .
From 1979 to 1981 new net external borrowing of the private sector minus
relatively modest interest payments on earlier borrowing amounted to a cumula-
tive 25 per cent of annual GNP (i.e. about 8 per cent each year), a hefty addition
to the resources available to the country. (OECD 1985, p. 30)
Clearly, then, there were powerful incentives, shared by private and
public interests alike, in accumulating capital through borrowing.
However, wishes here were not determinants of economic reality, as
Dupre might want us to suppose; they were thwarted by it. A policy
which involved denial of the principle that an increase in a capital stock
must be driven, at least at some point in the causal nexus, by increased
ef®ciency or volume of production, ultimately ran aground because of
that denial:
But maintaining a high level of borrowing for a number of years eventually
means that debt service charges `use up' all new borrowing . . . public sector net
external borrowing in 1983 and 1984 did little more than ®nance interest charges
262 Micro, macro, and markets

on past debt . . . by 1984 interest payments exceeded net new external borrowing
by 1 per cent of GNP. (OECD 1985, p. 31)

By this point, total government debt stood at approximately 128 per

cent of GNP with the external portion representing 70 per cent of GNP;
service of the external portion consumed 9 per cent of total export
earnings. Driven by escalating interest costs, debt as a percentage of
GNP reached 140 in 1986. By the early 1990s the debt situation limited
the ability of the government to manage the economy, and was itself a
destabilizing factor:
Heavy government indebtedness has the dual effect of making the public ®nances
highly sensitive to movements in interest rates while also contributing to such
movements, because it makes real interest rates more susceptible to attacks on the
currency . . . Ireland remains rather vulnerable to the costs and risks of interest-
rate cycles, and will probably have to continue to have to pay and risk premium
on borrowing while the debt/GNP ratio remains substantially above the inter-
national amount. (OECD 1990/91, pp. 48±50)

The rest of the 1990s was dominated by the need to bring Ireland's
debt to GNP ratio in line with the requirements of the new European
Union Common Currency. This goal was, of course, achieved, and,
happily, an Ireland that is no longer attempting to operate ®scal policy in
de®ance of economic reality now enjoys one of the highest rates of
growth and of rising real incomes in the world.
The Irish example is a classic, and by no means unique, case of a
government which, for whatever reason, acted as if the patterns of
economics are not objective and are malleable in light of other political
priorities. In the early stages, infrastructure development and economic
growth probably were stimulated by the high level of borrowing ±
foreign borrowing in particular, which increased the country's ability to
invest above the level its domestic savings would have supported. But
ultimately, the basic pattern of compounding interest on debt limited the
extent to which this policy could be pursued and eventually came to
dictate much of the rest of the country's economic policy. Again, we
should emphasize that our point is not that the policy choices of the Irish
government were necessarily `wrong' in any normative sense. Rather, we
are simply using the Irish example to illustrate the point that certain real
patterns exist at the macroeconomic level, and that these patterns will
cause particular policies to have particular outcomes, regardless of the
motivations of policy-makers.
As noted above, Ireland, at the time of the 1973 energy crisis, was
particularly vulnerable to a debt shock by reason of its heavy state
borrowing to drive publicly funded programmes of economic develop-
 The possibility of economic objectivity 263

ment. Such a policy has problems of its own if it ignores the economic
reality that all resources, and capital in particular, have opportunity
costs; the fact that the capital is supplied by the government does not
negate this fact. This second objective pattern observable in the
economic realm grows out of the fact that all economic patterns arise
from the fact of scarcity. As discussed above, critics of the possibility of
economic objectivity have often argued that because economics is a
reaction to the fact of scarcity in the world, with the consequence that
economic choices are often (if not always) about questions of distri-
bution, objectivity is inherently unobtainable; all economics is, regard-
less of appearances, about values and normative choices. We claim that
far from eliminating the possibility of objectivity, the fact of scarcity
gives rise to a major objective feature of the economic domain: all
resources have alternative uses and capital in particular, regardless of its
source, is never `free'. When it is employed in one use part of the cost of
producing its output in that use is the forgone production or output
that would have been obtained had it been employed in its next best
alternative use.
The objective existence of opportunity cost is often implicitly (at
least) denied by politicians and members of the public who argue that
certain goods and services are `public' services and should be produced
and distributed free regardless of the cost of the capital required to do
this. Failure to recognize or accept the reality of the pattern is often
related to what The Economist has called `a suspicion of pro®t itself ' (9
December 1995, p. 66). There is a general feeling that `social' ends and
needs do not merely override the question of economic cost ± i.e.,
make the cost worth paying ± but, in fact, mean that there is no real
cost when the capital is publicly provided. This is particularly true
when the subsidized activity appears to be self-supporting, in that once
the government provides the initial capital, the service generates
suf®cient funds to operate without requiring further infusions of
money, although not earning pro®t. But, as The Economist says, `When
the state provides capital, or guarantees suppliers against ruin, as in the
production of many public services, the risks involved are greatly
reduced. But capital is still not free. Taxpayers' money has plenty of
alternative uses' (ibid., p. 66). This is not to say that pro®t as
calculated by accountants according to Generally Accepted Accounting
Principles (GAAP) is necessarily an accurate re¯ection of the true
economic return or lack thereof to capital. GAAP are designed to
present the picture from the point of view of the shareholder alone and
may differ from a full economic accounting. According to Jones (1981,
p. 15):
264 Micro, macro, and markets

Publicly relevant pro®t is quite different from privately relevant pro®t for two
sets of reasons: ®rst, publicly relevant categories are different from privately
relevant categories; second, publicly relevant prices differ from privately relevant
prices.
An example of the difference between pro®t from the perspective of a
private shareholder and pro®t from the perspective of the economy as a
whole is provided by consideration of corporate income taxes. For the
corporation, these represent a cost and hence a deduction from pro®t; for
the economy as a whole, however, they represent part of the surplus
produced by capital after the costs of all inputs are paid and are,
therefore, counted as part of the return on that capital. An example of
the difference between privately and publicly relevant prices is the price
of foreign exchange. A private entrepreneur faces the market price for
foreign exchange while in reality the foreign exchange may be worth
more to the economy as a result of subsidies, tariffs etc., which distort
market equilibria. Similar discrepancies may obtain in labour and other
factor markets.
Regardless of the need to adjust market prices to re¯ect underlying
economic reality, the fact remains that capital has a price even when
provided by governments.8 Consequently, when it is employed for
`policy' reasons in uses where it does not earn its opportunity cost, the
economy pays a price in terms of less output elsewhere in order to obtain
the desired policy outcome. A good example of this is the situation that
prevails in Canada with publicly owned electric utilities. In most of
Canada, electricity is sold at prices below those that would generate
allocative economic ef®ciency but at a level suf®cient to allow the utility
to at least break even in a ®nancial sense (Economic Council of Canada
1986): `As a rule, prices are set as low as possible, consistent with
®nancial soundness. Thus revenue earned is suf®cient to meet the
®nancial requirements of the corporations' (p. 34). But ®nancial suf®-
ciency does not equate to economic ef®ciency:
In several respects, though, these requirements understate the opportunity cost of
the resources used. This is particularly true of returns to capital. For the most
part, little or no return on equity is earned, and the interest paid on debt is low
because of government guarantees . . . In addition, the rents paid to provincial
governments on water resources are minimal, and the pro®ts earned on exports
subsidize the domestic price of electricity. The consequence of this underpinning
are sales of electricity for which production costs exceed returns. (p. 34)
One might object that this opportunity cost should be balanced against

8 For a discussion of the calculation of the opportunity cost of capital in a Canadian

context, see Jenkins (1981) and Burgess (1981).
 The possibility of economic objectivity 265

a so-called `social surplus' in terms of a broader welfare measure. After

all, the utilities are ®nancially self-suf®cient and there are desirable policy
reasons for keeping the price of electricity low. It would seem that one
can achieve a socially desirable policy objective at zero cost to the
taxpayer. This, however, is not the case; the opportunity cost of capital is
an objective reality and the cost to society as a whole of the sub-normal
rates of return can be calculated. One study (Berkowitz and Halpern
1981, table 10±9, p. 236) using 1980 ®gures estimated that Canadian
electrical utilities earned only one-half the opportunity cost of the capital
invested in them while another estimated forgone rents on water rights in
1979 at $4 billion in Ontario, Quebec, Manitoba, and British Columbia
(Zucker and Jenkins 1984, table 4.1, p. 30). A third study examined the
economic loss to Canada arising from the overproduction of electricity
as a consequence of underpricing and the cost of investment distortion
that resulted from appraising projects using interest rates that did not
fully re¯ect the social opportunity cost of capital; the cost for the single
year 1978 was estimated to be between $1.4 and $2.4 billion (Jenkins,
unpublished).
It should be stressed that the above noted losses do not come about
because publicly owned utilities are in any necessary sense less ef®cient at
the production of electricity than private providers. The cost arises from
a failure of allocative ef®ciency due to a failure to recognize and recoup
the opportunity cost of the resources used in producing the electricity. It
is incurred because of a desire by governments to ignore one of the
objective realities of economics: scarcity is a fact, given demand schedules
and budget constraints that are also facts, and this implies the objective
reality of opportunity costs. As the Economic Council of Canada (1986,
p. 35) notes:

the pricing and investment decisions of public ®rms are a re¯ection of government
policy objectives. The distortions that have occurred under public ownership are
not inherent features of this organizational mode. If provincial governments
could accept pricing behaviour that was close to that required for economic
ef®ciency, a change to private ownership would not offer any advantages in this
respect.

The conclusion that can be drawn from the above discussion is that the
foundation of economics in the fact of resource scarcity, far from limiting
the possibility of objectivity, provides the very basis of that objectivity, a
point stressed by Robbins (1932) in his classic essay on the philosophy of
positive economics. Decisions about whether or not to price electricity
below the rate necessary to recoup the opportunity cost of capital are
indeed normative and subjective ± there may be social or political
266 Micro, macro, and markets

reasons why such a policy may be deemed desirable. It is obviously not

possible to enter directly into the minds of policy-makers and ascertain
whether or not the reality of opportunity cost enters into the policy
calculus. What observation surely does suggest, however, is that political
discussion of such matters rarely, if ever, includes explicit recognition of
the full cost of a choice to price government services below the full
opportunity cost of the resource consumed. However, the fact that this
choice imposes a cost on the national economy is an objective fact
grounded in the reality that resources are scarce.
We will conclude our set of examples of objectively existing patterns
with a brief discussion of what economists call the `lump of labour'
fallacy. This can, of course, only constitute a fallacy if there is a fact of
the matter about which one can hold a fallacious view; thus its very
identi®cation depends on the idea that there are real patterns in the
macroeconomic realm. The fallacy ± that the output of an economy, and
hence the amount of work available, is ®xed ± has a long history. It was
Ricardo who, in 1821, was perhaps the ®rst major economist to give
credence to the idea:
the opinion entertained by the labouring class, that the employment of machinery
is frequently detrimental to their interests is not founded on prejudice and error,
but is conformable to the correct principles of political economy.9 (Ricardo 1821,
p. 321)
Concerns that technological innovation in the form of new products or
process innovations will lead to massive unemployment are widely held
at the popular level and often ®nd support in political discourse.
Opponents of this view, on the other hand, argue that the reverse is true
± technological innovation creates employment. With regard to product
innovation, it can certainly be argued that the invention and entry into
mass use of hitherto unknown products such as home computers and
microwave ovens ultimately creates new jobs, in the manufacture and
service of such goods, that did not previously exist. Similarly, process
innovations and the consequent improvements in productivity which
they generate can be expected, ultimately, to lead to increased rather
than decreased employment. While process innovation may enable a
given amount of output to be produced with less labour, and hence, in
the short run, lead to increased unemployment, such process innovations,
like product innovations, ultimately generate increased demand. In-
creased productivity reduces costs, which in turn leads to lower prices,
higher wages or increased pro®ts. The ®rst two results increase real wages
9 We are grateful to an anonymous referee for pointing out that, in falling into this fallacy,
Ricardo was in fact changing his previously expressed view.
 The possibility of economic objectivity 267

and consequently raise aggregate demand for both goods and workers.
On the other hand, if higher productivity serves only to increase pro®ts,
these will ultimately drive increased investments, which will also increase
output and employment.10 This is not to deny that switching effects may
occur within the ambit of a business cycle, where as the marginal
productivity of capital increases during an upswing, the opportunity cost
of wages rises. In this case, if wage rates are suf®ciently sticky, unemploy-
ment levels may actually be increased by higher rates of capital invest-
ment. However, to whatever extent we should expect switching, we
should equally well predict re-switching on the opposite slope of the
cycle. So long as investment increases capacity faster than the growth of
the labour-supply, and so long as employed labour is fully employed,
investment-driven growth between the beginning of one cycle and the
end of the next will reduce unemployment. We are unaware of any
historical case where capacity growth over the full course of a cycle has
been unaccompanied by a fall in unemployment except in conditions
where one of the two ceteris paribus conditions above has been violated.
Of course, this observation is very hard to test, given the basic dif®culties
in measuring cycles and in unambiguously demonstrating that the ceteris
paribus conditions are satis®ed. However, we claim that, given the very
simple logic supporting the fundamental pattern, the burden of argument
concerning its robustness should lie with the sceptic.
The literature on switching effects generally assumes a constant rate of
productivity per marginal unit of capital stock. Concerns over so-called
`technological unemployment' are thus a distinct issue, though they often
compound the `lump of labour' fallacy by applying it to an analysis in
which it has already been committed once before increasing returns to
scale (either driven by technology or by ef®ciencies of volume) are taken
into account. There are then two viewpoints as to the ultimate impact of
technological innovation. The ®rst ± that of Ricardo and much popular
opinion ± is that technological improvements have a negative impact on
employment because fewer workers can produce the same level of
10 It should be noted that these remarks pertain only to the rate of involuntary
unemployment, and not to the structure of unemployment. Technological change may
indeed contribute, as has often been pointed out, to a situation in which a stratum of
permanently unemployable people is created. If the business cycle in a sector is such that
labour markets do not periodically clear, the employability of those who have been out
of work longest may diminish to vanishing point, as their increasing lack of familiarity
with new technology holds them constantly at the back of every job queue. If the basic
insight of endogenous growth theory is correct, the existence of such a stratum may act
as a brake on growth as average value of human capital in the sector diminishes.
However, this fact does not somehow imply that the `lump-of-labour' fallacy is not a
fallacy.
268 Micro, macro, and markets

output. The second view holds that innovation will, through increasing
incomes and generating new products, lead to increasing (or at least not
decreasing) levels of employment through the mechanism of consumer
demand for new products and/or rising incomes being used for increased
consumption. Which result will obtain in which circumstances is not a
normative issue or a question of values. Rather it depends on the
objective pattern prevailing in a given economy at a given time and the
changes underway in that economy. What is at issue ± or, at least, what
ought to be at issue ± are not aspects of the relative costs of supply-side
factors, but rather the shape of the pattern economists call declining
marginal utility of income. If the collective utility to an economy of
increased income is low ± i.e., most individuals don't want to consume
more ± then the ®rst view may be correct. In other words, the ultimate
net impact on employment levels will depend on the demand responsive-
ness of prices to lower costs and the elasticity of demand. In a world of
fully satiated consumers with consequent total inelasticity of demand,
demand would not rise to match an increase in productive capacity, and
the result could be increased unemployment. But what is the evidence for
one view as opposed to the other? It seems to argue overwhelmingly in
favour of the conclusion that rapid technological innovation goes hand
in hand with increasing rather than decreasing employment. In a study
prepared for the OECD, Pascal Pettit (1995, p. 21) concludes:
Only in ceteris paribus universes is a faster increase in productivity synonymous
with lower employment growth and more unemployment. The whole experience
of the golden years of capitalism stresses the contrary: unprecedentedly high
levels of productivity gains were accompanied by full employment.

The Economist (25 November 1995, p. 67) cites ®gures showing that
for the last 100 years in the United States rising productivity has gone
hand in hand with rising employment (and rising incomes). One can
probably safely conclude that, for most economies, the marginal utility
of income is not sharply declining; a ¯at-sloped collective marginal utility
of income curve is a very plausible real pattern. This is not to say that a
shift in the pattern is not possible or that some other pattern may not
apply in certain economies. Again, however, real patterns are not to be
identi®ed with nomic generalizations. Nor is it to say that technological
change will inevitably increase employment of the type for which many
of those displaced by technological innovation will be suited, nor that the
bene®ts and costs of technological innovation are in any sense `fairly' or
equitably distributed. It is just to say that there are objective patterns
which exist and which cannot be ignored, dismissed or overridden by the
normative aspects of economic policy-making.
 The possibility of economic objectivity 269

Striking as these examples seem to us to be, it is problematic from the

point of view of the philosophy of science to rest the case for economic
objectivity mainly on instances from macroeconomics. The dif®culty is
that one cannot claim without controversy that macroeconomics has
arrived at a theoretical consensus that would permit us to produce a
stable typology of the real patterns it studies, or to measure the causal
propensities of its crucial variables. (For elaboration of this point, see
Ross and LaCasse 1995, pp. 474±8.) Thus the defender of objectivity in
macroeconomics can only proceed, as we have done here, by suggestive
enumeration of cases. We cannot seek a generic account of real macro-
economic patterns through elucidation of macroeconomic foundations
for the simple reason that macroeconomics has no settled foundations. In
light of this observation, it is not surprising that reductionists and
supervenience-theorists have been doubtful about the possibility of
macroeconomic objectivity. The metaphysic of patterns, however,
removes the need to build macroeconomic reality from foundations. Our
defence of that reality may instead proceed directly: the best evidence we
could have that there is an objective domain of macroeconomic patterns
is simply the fact that the patterns kick back if ignored. This was the
point our short stock of examples above is intended to illustrate. We
doubt that an economically literate reader will have serious dif®culty in
adding to it.
Where microeconomics is concerned, the dependence relation between
foundations and a plausible empirical reference-class for theoretical
terms we ®nd in defending a realist understanding of macroeconomics is
reversed. In microeconomics, we ®nd a consensus as strong as in any
science over theoretical foundations, and a large body of impressive
applications of the theory; however, it is surprisingly dif®cult to identify
the theory's objects. The view for which one of us has argued elsewhere is
that the fundamental core of microeconomics is best explicated by means
of the concepts of game theory, and that, under this interpretation,
games ± the set of patterns to which strategic interactions converge under
selection pressures of various sorts ± are the basic microeconomic
objects. The defence of this claim is the main subject of Ross and
LaCasse (1995). Here, we will just point out that if the claim is accepted,
then something along the lines of the real-patterns metaphysic is essential
to the ambition that microeconomics be an empirical science, since
possible objects that might constitute a supervenience base for games are
even harder to imagine than for the macroeconomic patterns exempli®ed
above. This is partly because game-players, that is, loci of ®xed pre-
ference-orderings over outcomes, can generally be token-identi®ed only
with aspects of objects identi®ed from outside of the economic stance,
270 Micro, macro, and markets

such as people or ®rms, since people and ®rms, over the course of their
biological or institutional biographies, demonstrably and frequently
reverse their tastes. Worse for the defender of a token-supervenience
account of microeconomics, the relation is typically many-to-many: an
economic agent may be mapped with least strain on to a set of aspects of
several people or ®rms. Worse yet, there are no individuation conditions
on such putative tokens independent of the economic perspective; so the
imagined token-identity would amount to nothing more than a gratui-
tous metaphysical ¯ourish. Fortunately, if the metaphysic of real patterns
elucidated earlier in this paper is accepted, then there is no need for
resort to such ad hoc devices in support of the claim that micro-
economics, like macroeconomics, tracks aspects of objective reality.
The claim that both macroeconomics and microeconomics have
objectively existing domains of investigation obviously does not imply
that most, or even much, work in economics is normatively disinterested.
That policy problems have an overwhelming in¯uence on which
economic phenomena are chosen for investigation is surely true; but the
same thing can be said of the medical disciplines and of large parts of
physics and chemistry, without this impugning their claims to be positive
sciences. It may also be true, as has sometimes been charged, that due to
its lack of secure theoretical foundations a good deal of work in macro-
economics resembles arbitrary curve-®tting. But in this case, what we
should say is that macroeconomics has not yet settled on an effective
method of tracking its domain; and this is an entirely different claim
from the statement that positive macroeconomics is impossible. The ideal
of objective science is of immense importance as a regulating principle in
research, even if in certain sciences it is rarely met. Researchers will not,
however, go on respecting an ideal that they have come to believe cannot
even be approximated. For this reason, if not also for the sake of truth,
we should be very cautious concerning arguments for the impossibility of
objectivity. It is possible ± though, we think, highly unlikely ± that
societies would be better off, in some sense which welfare economics has
tried, without notable success, to articulate, if economic policy choice
were overtly driven by ideology rather than facts. But we would surely
not be better off for mistakenly supposing that we have no other
choice.11
11 We would like to thank the following people for their useful comments and criticisms on
earlier drafts of this paper: Chantale LaCasse, Maurice Lagueux, Uskali MaÈki, Robert
Nadeau, two anonymous referees, and audiences before whom versions of the paper
were presented at l'Universite du QueÂbec aÁ MontreÂal and at the World Congress of
Logic, Methodology and Philosophy of Science in Florence, Italy. Ross is also grateful
to the Social Sciences and Humanities Research Council of Canada for ®nancial support.
 The possibility of economic objectivity 271

References
Alesena, A. and G. Tabellini (1987) `A Positive Theory of Fiscal De®cits and
Government Debt in a Democracy', National Bureau of Economic
Research, Washington, D.C., United States, working paper no. 2308.
Berkowitz, M. K., and P. J. Halpern (1981) The Role of Crown Corporations in
the Ef®cient Production and Use of Electricity, v. 1: An Economic Analysis of
the Production, Investment and Financing Behaviour of the Canadian Elec-
trical Utility Industry, Toronto: Institute for Policy Analysis, University of
Toronto.
Burgess, D. (1981) `The Social Discount Rate for Canada: Theory and Evidence',
Canadian Public Policy, 7: 344±83.
Cartwright, N. (1983) How the Laws of Physics Lie, Oxford: Oxford University
Press.
(1989) Nature's Capacities and their Measurement, Oxford: Oxford University
Press.
Dennett, D. (1979) `True Believers', in A. Heath (ed.), Scienti®c Explanation,
Oxford: Oxford University Press.
(1991) `Real Patterns', Journal of Philosophy 88: 27±51.
(2000) `With a Little Help From My Friends', in D. Ross, A. Brook and
D. Thompson (eds.), Dennett's Philosophy: A Comprehensive Assessment,
Cambridge, MA: MIT Press / Bradford.
DupreÂ, J. (1993) The Disorder of Things, Cambridge, MA: Harvard University
Press.
Economic Council of Canada (1986) Minding the Public's Business, Ottawa:
Supply and Services Canada.
Fodor, J. (1974) `Special Sciences', Synthese, 28: 77±115.
Friedman, Michael (1979) `Truth and Con®rmation', Journal of Philosophy, 76:
361±82.
Friedman, Milton (1953) Essays in Positive Economics, Chicago: University of
Chicago Press.
Harding, S. (1986) The Science Question in Feminism, Ithaca: Cornell University
Press.
Hoover, K. (1995) `Why Does Methodology Matter for Economics?', Economic
Journal, 105: 715±34.
Jenkins, G. (1981) `The Public Sector Discount Rate for Canada: Further
Implications', Canadian Public Policy, 7: 399±407.
(unpublished) `Public Utility Finance and Economic Waste', Harvard Institute
for International Development, Harvard University, ms.
Jones, L. (1981) Performance Evaluation of Public Enterprises, Washington, D.C.:
World Bank Discussion Paper 122.
Kim, J. (1984) `Concepts of Supervenience', Philosophy and Phenomenological
Research, 45: 153±76.
(1989) `The Myth of Nonreductive Materialism', Proceedings and Addresses of
the American Philosophical Association, 63: 31±47.
272 Micro, macro, and markets

LaCasse, C. and D. Ross (1991) `Reply to Norman, Has Rational Economic

Man a Heart', Eidos, 8: 235±46.
Machlup, F. (1969) `Positive and Normative Economics', in R. Heilbroner (ed.),
Economic Means and Social Ends, Englewood Cliffs: Prentice-Hall.
Marras, A. (1993) `Psychophysical Supervenience and Nonreductive Materi-
alism', Synthese, 95: 275±304.
Myrdal, G. (1954) The Political Element in the Development of Economic Thought,
Cambridge, MA: Harvard University Press.
(1958) `International Integration', in P. Streeten (ed.), Value in Social Theory:
A Selection of Essays on Methodology by Gunnar Myrdal, New York: Harper
and Brothers.
Oppenheim, P. and H. Putnam (1958) `Unity of Science as a Working Hypoth-
esis', in H. Feigl, G. Scriven and G. Maxwell, eds., Minnesota Studies in the
Philosophy of Science, vol. II, Minneapolis: University of Minnesota Press.
Organisation of Economic Cooperation and Development (1971, 1975, 1982,
1985, 1990/91, 1993) Economic Surveys: Ireland, Paris.
Pettit, P. (1995) `Technology and Employment: Key Questions in a Context of
High Unemployment', STI Review, 15: 13±47. Paris: OECD.
Ricardo, D. (1821) Principles of Political Economy and Taxation, republished in
P. Sraffa (ed.), The Works and Correspondence of David Ricardo, vol. I,
Cambridge: Cambridge University Press, 1962.
Robbins, L. (1932) An Essay on the Nature and Signi®cance of Economic Science,
London: Macmillan.
Rosenberg, A. (1991) Economics: Mathematical Politics or Science of Diminishing
Returns? Chicago: University of Chicago Press.
Rosett, C. (1993) `Rooted in Soviet Past, Russia's Central Bank Lacks Grasp of
Basics', The Wall Street Journal, 222: 59: 3±7.
Ross, D. (1995) `Real Patterns and the Ontological Foundations of Microeco-
nomics', Economics and Philosophy, 11: 113±36.
(1999) What People Want: The Concept of Utility From Bentham to Game
Theory, Cape Town: University of Cape Town Press.
(2000) `Rainforest Realism: A Dennettian Theory of Existence', in D. Ross,
A. Brook and D. Thompson (eds.), Dennett's Philosophy: A Comprehensive
Assessment, Cambridge, MA: MIT Press / Bradford.
Ross, D. and C. LaCasse (1995) `Toward a New Philosophy of Positive
Economics', Dialogue, 34: 467±93.
Roulini, N. and J. Sachs (1989) `Determinants of Budget De®cits', European
Economic Review, 33: 903±33.
Savellos, E. and U È . YalcËin (1995) Supervenience: New Essays, New York:
Cambridge University Press.
Weston, S. (1994) `Toward a Better Understanding of the Positive/Normative
Distinction in Economics', Economics and Philosophy, 10: 1±17.
Zucker, R. and G. Jenkins (1984) Blue-Gold: A Study of Hydro-Electric Rents in
Canada: A Study Prepared for the Economic Council of Canada, Ottawa:
Supply and Services Canada.
Part IV
The world of economic causes
14 Ceteris paribus laws and
socio-economic machines

NANCY CARTWRIGHT*

1 Why economics is not allowed ceteris paribus laws

Economics differs from physics, we are told, in that the laws economics
studies hold only ceteris paribus whereas those of physics are supposed to
obtain universally and without condition.1 Does this point to a metaphy-
sical difference between the laws the two disciplines study or does it
re¯ect merely a de®ciency in the level of accomplishment of economics as
compared to physics?
The conventional regularity account of laws tells us it must be the
latter. On this account a theoretical law is a statement of some kind of
regular association,2 usually supposed to hold ``by necessity.'' The idea
of necessity is notoriously problematic. Within the kind of empiricist
philosophy that motivates the regularity account it is dif®cult to explain
what constitutes the difference between law-like regularities and those
that hold only by accident, ``nonsense'' correlations that cannot be relied
on. I shall not be concerned with necessity here; I want to focus on the
associations themselves. These can be either universal, in which case the

* ``Ceteris Paribus Laws and Socio-Economic Machines'', Nancy Cartwright, The Monist,
vol. 78, no. 3, pp. 276±94, Copyright # 1995, The Monist, La Salle, Illinois 61301.
Research for this paper was sponsored by the ``Modeling in Physics and Economics''
project at the Centre for the Philosophy of the Natural and Social Sciences, London
School of Economics. I would like to thank the members of that research group for help
with this paper, as well as Mary Morgan and Max Steuer. The paper was read at a
symposium on realism in the Tinbergen Institute; I would also like to thank the students
and faculty who participated in that symposium for their very helpful comments.
1 See, for instance, J. J. C. Smart, Philosophy and Scienti®c Realism (London: Routledge

and Kegan Paul, 1963), who does not think even biology has proper laws.
2 C. G. Hempel and P. Oppenheim, ``Problems in the Concept of General Law,'' in

A. Danto and S. Morgenbesser (eds.), Philosophy of Science (New York: Meridian

Books, 1960).

275
276 The world of economic causes

law is deterministic, or they may be merely probabilistic. The regularity

account is grounded in a version of empiricism that traces back to the
philosophy of David Hume. Empiricism puts severe restrictions on the
kinds of properties that appear in Nature's laws, or at least on the kinds
of properties that can be referred to in the law-statements we write down
in our theories. These must be observable or measurable.3 It also restricts
the kinds of facts we can learn: the only claims about these quantities
that are admissible into the domain of science are facts about patterns of
their co-occurrence. Hence the speci®cation of either an equation (in the
case of determinism) or of a probability distribution over a set of
measurable quantities becomes the paradigm for a Law of Nature.
These two assumptions work together to ban ceteris paribus laws from
the Nature that we study. Together they tell us that all the quantities we
study are qualitatively alike. There are no differences between them not
®xed either by the patterns of their associations or by what is observable
or measurable about them. As a consequence it becomes impossible to
®nd any appropriate category that would single out conditioning factors
for laws from the factors that fall within them. What then could it mean
to include as a law in one's theory that (to use an example I will discuss
later) ``In conditions C, all pro®table projects will be carried out''? All
that is law-like on the Humean picture are associations between measur-
able quantities ± and that's it. The only way a condition could restrict the
range of an association in a principled or nomological way would be via
a more complex law involving a general association among all the
quantities in question. The effect of this is to move the conditioning
factor C inside the scope of the law: ``All projects that are both pro®table
and satisfy conditions C will be carried out.'' That's what the Laws of
Nature that we are aiming for are like. If theories in economics regularly
leave out some factors like C (whatever C stands for) from the ante-
cedents of their laws, and hence write down law claims that turn out to
be, read literally, false, then economics is regularly getting it wrong.
These theories need to keep working on their law-statements till they get
ones that express true regularities.
I defend a very different understanding of the concept of Natural Law
in modern science from the ``laws = universal regularities'' account I
have been describing.4 We aim in science, I urge, to discover the natures
of things; we try to ®nd out what powers or capacities they have and in
3 Sometimes philosophers substitute for ``measurable'' or ``observable'' the requirement
that the features studied be ``extensional'' or ``occurrent'' as opposed to ``dispositional.''
To my mind the latter categories make no sense.
4 N. Cartwright, Nature's Capacities and Their Measurement (Oxford: Oxford University
Press, 1989); and ``Aristotelian Natures and the Modern Experimental Method,'' in
 Ceteris paribus laws and socio-economic machines 277

what circumstances and in what ways these capacities can be harnessed

to produce predictable behaviors. I call this the study of natures because
I want to recall the Aristotelian idea that science aims to understand
what things are, and a large part of understanding what they are is to
understand what they can do, regularly and as a matter of course.5
Regularities are secondary. Fixed patterns of association among measur-
able quantities are a consequence of the repeated operation of factors
that have stable capacities (factors of this kind are sometimes called
``mechanisms'') arranged in the ``right'' way in the ``right kind'' of stable
environment. The image is that of a machine with set components that
must be assembled and shielded and set running before any regular
associations between input and output can be expected. In the case of
economics we can summarize this way: regularities are a consequence of
the repeated successful operation of a socio-economic machine.
I take it that the basic information we need in economic theory is
information about the characteristic behavior of the components and
about how to set them running together. This information is not itself a
report about regular associations between measurable quantities,
although it can be used to predict regular associations. The Humean view
assumes the contrary ± all the information there is to learn in economic
theory is immediately about the regular associations that obtain in the
economy. The Humean account of laws goes naturally with a covering
law theory of prediction and explanation. One set of regularities ± the
more concrete or phenomenological ± is explained by deducing them
from another set of regularities ± the more general and fundamental. The
distinction is like that in econometrics between structural equations and
reduced-form equations.6 The alternative theory of explanation in terms
of natures rejects the covering-law account: you can't have regularities
``all the way down.''
The example I will present in Section 4 is one in which we cross levels
of scale, or aggregation, in moving from explanandum to explanans. The
model I will describe is one in which a (highly idealized) macro-economic
regularity is shown to arise from the rational behavior of individual

J. Earman (ed.), Inference, Explanation and Other Philosophical Frustrations (Berkeley,

CA: University of California Press, 1992).
5 But I do not think that modern science follows the Aristotelian tradition in identifying the
nature of something (what it can do) with what it is (its de®nition); rather it insists that
the features and structures we study be identi®able independently of any speci®c claim
about what they can do.
6 Only note: it follows from what I shall argue in section 2 that only reduced-form
equations ®t the regularity account of laws; the structural equations cannot be read in the
same way as genuine reports of regularities.
278 The world of economic causes

agents. This crossing of levels is not important to the point. We could

attempt to explain macroeconomic regularities by reference to capacities
and relations that can only sensibly be attributed to institutions or to the
economy as a whole, with no promise of reduction to features of
individuals.7 On the other hand we could undertake ± as psychologists
often do ± to explain behavioral or psychological regularities by reference
to behavioral or psychological characteristics of the individuals involved.
The point is that in all these cases it is never laws, in the sense of
regularities, that are fundamental. No economic regularity, no matter
how basic or important or central or universal we think it is, can stand
on its own. If there are laws in the regularity sense to record in economic
theories, that is because they have been created.
Ceteris paribus conditions play a special role when explanation
depends on natures and not on covering laws. On the natures' picture
there is a general and principled distinction between the descriptions that
belong inside a law statement and those that should remain outside as a
condition for the regularity described in the law to obtain. The regula-
rities to be explained hold only ceteris paribus; they hold relative to the
implementation and operation of a machine of an appropriate kind to
give rise to them. The hypothesiss I propose in this paper is that the
covering-law account is inappropriate for much of the work in modern
economic theory. Economists usually talk in terms of models rather than
theory. In part they do so to suggest a more tentative attitude to the
explanations in question than would be warranted in a fully-¯edged well-
con®rmed theory; and in part they do so to mark a difference in the
degree of articulation: theory is a large-scale (and not necessarily
formalized) outline whereas a model gives a more speci®ed (and forma-
lized) depiction. But I think the terminology also marks a difference of
importance to the hypothesis of this paper. ``Theory'' suggests a set of
covering laws like Maxwell's equations, laws of the kind physics'
explanations are supposed to begin from. Models in economics do not
usually begin from a set of fundamental regularities from which some
further regularity to be explained can be deduced as a special case.
Rather they are more appropriately represented as a design for a socio-
economic machine which, if implemented, should give rise to the
behavior to be explained. I illustrate in Section 4 with a game theory
example.

7 For an account of how to do non-reductionistic social science, see David Ruben, The
Metaphysics of the Social World (London: Routledge, 1985).
 Ceteris paribus laws and socio-economic machines 279

2 What's wrong with regularities?

The most immediate problem with identifying natural laws with regula-
rities is that, as John Stuart Mill observed,8 regularities are few and far
between. That is why economics cannot be an inductive science. What
happens in the economy is a consequence of a mix of factors with
different capacities operating in a particular environment. The mix is
continually changing; so too is the background environment. Little is in
place long enough for a regular pattern of associations to emerge that we
could use as a basis for induction. Even if the situation were stable
enough for a regularity to emerge, ®nding it out and recording it would
be of limited use. This is the point of Trygve Haavelmo's example of the
relation between the height of the throttle and the speed of the auto.9
This is a useful relation to know if we want to make the car go faster. But
if we want to build a better car, we should aim instead at understanding
the capacities of the engine's components and how they will behave in
various different arrangements.
The second problem with regularities is that most of the ones there are
do not re¯ect the kind of fundamental knowledge we want, and indeed
sometimes have. We want, as Mill and Haavelmo point out, to under-
stand the functioning of certain fundamental rearrangeable com-
ponents.10 Most of what happens in the economy is a consequence of the
interaction of large numbers of factors. Even if the arrangement should
be appropriate and last long enough for a regular association to arise
(that is, they make up a socio-economic machine) that association would
not immediately teach us what we want to know about how the parts
function separately. Hence the laments about the near impossibility in
economics of doing controlled experiments designed especially to do this
job.
But are not the laws for describing how the parts function themselves
regularity statements? One central point of this paper is to urge that the
answer to this question is No. Consider what kinds of regularity
statements these would have to be. They are usually thought to record

8 J. S. Mill, ``On the De®nition of Political Economy and On the Method of Philosophical
Investigation in that Science'' [1850] in Collected Works (Toronto: University of Toronto
Press, 1967). See also ``On the Logic of the Moral Sciences,'' A System of Logic [1872],
also in the Collected Works.
9 T. Haavelmo, ``The Probability Approach in Econometrics,'' Econometrica, 12 (1944),
Supplement.
10 Parts are fundamental to the extent that their capacities stay relatively ®xed over a wide
(or wide enough) range of circumstances.
280 The world of economic causes

claims about what regularly follows if the component in question were to

operate on its own. The ®rst thing we should note is that if this is what a
law is, there are very few true laws to be had, for regularities of this kind
are rare indeed and the majority of these are hard won in a physics
laboratory. Secondly, even if we had such regularities they could not
perform the explanatory and predictive jobs that we expect of laws. The
reasons for wanting a regularity account of laws quite naturally motivate
a corresponding covering-law account of explanation and prediction. As
noted in section 1, on the covering-law account to explain a phenomenon
is to show it to be an instance of a general law, which is generally taken
to mean ``a general regularity.'' Collections of laws about what happens
when mechanisms operate on their own are not then going to be able to
explain the bulk of the phenomena we want them for, phenomena that
result from the joint operation of a mix of mechanisms. We can of course
begin to give an account here invoking ``laws'' or recipes for interaction.
The question is, can we give an account within a purely regularity theory
of laws?
Let us turn next to the idea of a mechanism operating on its own. We
may conceive of the demand mechanism in terms of individual prefer-
ences, goals and constraints or as irreducibly institutional or structural.
In either case on the regularity account of laws the law of demand
records the regular behavior that results when the demand mechanism is
set running alone. This paradigmatic case (which Haavelmo himself uses
in talking about laws for fundamental mechanisms) shows up the
absurdity of trying to describe the capacities of mechanisms in terms of
regularities. No behavior results from either the supply or the demand
mechanism operating on its own, and that is nothing special about this
case. In general it will not make sense to talk about a mechanism
operating on its own. That's because in this respect economic mechan-
isms really are like machine parts ± they need to be assembled and set
running before any behavior at all results. This is true in even the most
stripped-down cases. A rigid rod, for example, must not only be af®xed
to a fulcrum before it becomes the Simple Machine called a ``lever''; it
must also be set into a stable environment in which it is not jiggled about.
The same is true in economics. Consider an analogous case, a kind of
economic lever that multiplies money just as the lever multiplies force. In
this example banking behavior will play the role of the rigid rod with
``high-powered money'' as the force on one end. The reserve ratio will
correspond to where the fulcrum is located. Here is how the money-
multiplying mechanism works: the ``central bank'' has a monopoly on
making money. Following convention, let us call this high-powered
money. The commercial banking system blows up, or multiplies, the
 Ceteris paribus laws and socio-economic machines 281

high-powered money into the total money supply. The banks do this by
lending a fraction of the money deposited with them. The larger the
proportion they can lend (i.e. the smaller the reserve ratio) the more they
can expand money. Two factors are at work: the proportion of high-
powered money held as currency, which is one minus the proportion
deposited; and the proportion lent, which is one minus the reserve ratio.
Suppose, for example, that high-powered money = £100. All of it could
be deposited in a bank. That bank could lend £80. All £80 could be
deposited in another bank. That bank could lend £64, and so on. The
total of all deposits, 100 + 80 + 64 + . . . = 500.
Assume that banks lend all they can, which as pro®t makers they are
disposed to do. Then we can derive
M = (1+cu/re+cu) H (1)
where H = high powered money; re = the reserve ratio; 17 re = li, the
lending ratio; cu = the rate of currency to deposits; M = the money stock.
Equation (1) is like the law of the lever. Regularity theorists would like to
read it as a statement of regular association. But that is too quick. We do
not have a description of some law-like association that regularly occurs.
Rather we have a description of a socio-economic machine that would
give rise to a regular association if it were set running repeatedly.
This simple example illustrates both of the central points I want to
argue in this paper. First, regularity theorists have the story just upside
down. They mistake a sometime-consequence for the source. In certain
very felicitous circumstances a regularity may occur (or, to look to the
counterfactual account I will discuss brie¯y in the next section, it may be
set ready to occur), but the occurrence ± or even the possibility of the
occurrence ± of these regularities is not what is nomologically basic. We
want to learn how to construct the felicitous circumstances where output
is predictable. Our theories must tell us what the fundamental mechan-
isms available to us are, how they function, and how to construct a
machine with them that can predictably give rise to a regular association.
And the information that does this for us is not itself a report of any
regular association ± either a real regular association (one that occurs) or
a counterfactual one (that might occur). Second, the example shows the
special place of ceteris paribus conditions. Equation (1) holds ceteris
paribus, that is under very special circumstances we may designate C. C
does not mark yet another variable like H, M, cu and re that has
mistakenly been omitted from equation (1), either by ignorance or sloth.
Rather it marks the entire characterization of the socio-economic
machine that equation (1) describes and that if properly shielded and run
repeatedly will produce a regular rise and fall of money with deposits.
282 The world of economic causes

The example also reminds us that, like the rigid rod, the socio-
economic lever as well must be carefully shielded to keep it intact and to
ensure that it alone affects the outcome. Even in very developed models
the need for shielding is often left implicit. I think that has to do with the
very high level of abstraction at which theoretical models operate. Think
about the rigid rod attached to the fulcrum. What counts as shielding if
the rod is to function properly as a lever will be very different depending
on the material of the rod and the fulcrum and the method of ®xing them
together. In general, if the level of abstraction is too high, about all that
can be said is ``ensure nothing interferes.'' In this case the model has not
¯eshed in the full set of ceteris paribus conditions lying behind the law
derived in it but only reminded us that more still needs to be done for
any real-life application.
I use the word ``mechanism'' in describing the money multiplier.
Regularity theorists also talk about mechanisms. But for them a mech-
anism can only be another regularity, albeit a more ``fundamental'' one.
``More fundamental'' here usually means either more encompassing, as
in the relation of Newton's to Kepler's laws, or having to do with smaller
parts as in the relation between individuals with their separate behaviors
and expectations on the one hand and macro-economic regularities on
the other. Economics does not have any fundamental laws in the ®rst
sense. But the second sense will not help the regularity theorist either
since explanations citing regularities about individuals have all the same
problems as any other. When we say ``mechanisms,'' I think we mean
``mechanism'' in the literal sense ± ``the structure or adaptation of parts
of a machine.''11 The little banking model is a good case. In our banking
example the rigid relationship between high-powered money that the
central bank creates and the ultimate size of the money stock depends on
the currency-to-deposit ratio not changing, and on every commercial
bank being fully lent up, that is, being tight on, but not beyond, the legal
reserve ratio. It is not immediately obvious that a group of commercial
banks can ``multiply'' the amount of currency issued by the central bank
when no one bank can lend more than is deposited with it. The banking
system can lend more even though no one bank can. How much more?
That is what the ``law'' tells us. But it is not derived by referring to a
regularity, but rather by deducing the consequences of a mechanism. I
illustrate the relations between economic models and regularities with
another example in section 4. But ®rst I should discuss brie¯y one
philosophical loose-end.

11 De®nition 1, The Concise Oxford Dictionary, 8th edn.

 Ceteris paribus laws and socio-economic machines 283

3 Philosophical aside

When confronted with the fact that there seem to be a lot more laws than
there are regularities, regularity theorists are apt to defend their view by
resorting to counterfactual regularities. Laws are identi®ed not with
regularities that obtain but with regularities that would obtain if the
circumstances were right. It is fairly standard by now to go on to analyze
counterfactuals by introducing an array of other ``possible'' worlds.
Laws then turn out to be regularities that occur ``elsewhere'' even though
they do not occur here ± that is, even though they do not occur at all. As
a view about what constitutes a Law of Nature this revised (or ``mod-
alized''12 ) version of the regularity account seems even more implausible
than the original. A Law of Nature of this world ± that is, a Law of
Nature full stop since this is the only world there is ± consists in a
regularity that obtains nowhere at all.
There are two strategies, both I think due to David Lewis, that offer
the beginnings of an account of how possible regularities can both be
taken seriously as genuine regularities and also be seen to bear on what
happens in the actual world. The ®rst takes possible worlds not as
®ctional constructions but as real.13 The second takes them as book-
keeping devices for encoding very complicated patterns of occurrences in
the ``real'' world. In this case the truth of a counterfactual (about the
``real'' world) will be entirely ®xed by the complicated pattern of events
that actually occur in the ``real'' world. Rather than employing this very
complicated semantics directly we instead devise a set of rules for
constructing a kind of chart from which, with the aid of a second set of
rules, we can read off whether a counterfactual is true or false. The chart
has the form of a description of a set of possible worlds with a ``nearness''
relation de®ned on them. But the description need not be of anything like
a world, the ``worlds'' need not be ``possible'' in any familiar sense, and
the special relation on them can be anything at all so long as a recipe is
provided for how to go from genuine ``occurrent,'' ``measurable'' or
``observable'' facts about our world to the ordering on the possible
worlds. The trick in both cases, whether the possible worlds are real or
only function as account books, is to ensure that we have good reasons
for making the inferences we need; that is, that we are able to infer from
the truth of a counterfactual as thus determined to the conclusion that if

12 ``Modal'' because it employs concepts of necessity and possibility.

13 This view is explicitly defended in D. Lewis, On the Plurality of Worlds (Oxford:
Blackwell, 1986).
284 The world of economic causes

we really were repeatedly to instantiate the antecedent of the counter-

factual (in the ``actual'' world), the consequent would regularly follow.
This of course is immediately guaranteed by the interpretation that reads
claims about counterfactual regularities just as claims about what
regularities would occur if the requisite antecedents were to obtain. But
then we are back to the original question: What sense does it make to
claim that laws consist in these nowhere existent regularities?
The point of this question is to challenge the regularity theorist to
explain what advantage these non-existent regularities have over a more
natural ontology of natures which talks about causes, preventions,
contributing factors, triggering factors, retardants and the like. We may
grant an empiricist point of view in so far as we require that the claims of
our theories be testable. But that will not help the regularity theorist
since causal claims or ascriptions of capacities are no harder to test than
claims about counterfactual regularities and indeed, I would argue, in
many cases you can't do one without the other.14 So far as I can see the
advantage is supposed to be that the regularity account, even if it relies
on regularities that are not real, does not employ any funny concepts. We
use only concepts describing measurable (or ``observable'' or ``occur-
rent'') quantities and their occurrence at given places and times. The
propositional connectives used (``if X were to happen, then Y would
happen'') do not have a truth-functional semantics (by contrast with ``if
X then Y'' which is equivalent to ``not-X or Y'') as they should from a
strict Humean point of view. But, it seems, this is a problem that many
regularity empiricists are prepared to live with (though not purists like
Bas van Fraasen).
But what is funny about concepts like causality by contrast with
concepts of so-called measurable or occurrent properties? David Hume
had an associationist theory of concept formation that he combined with
a copy theory of impressions and ideas. Our ideas are either copies of our
impressions or are constructed from copies of impressions by a method
much like truth-functional logic. Impressions in turn are copies of the
sensory qualities of things; they are the ``way-they-look'' of purely
sensible qualities. On this story predicates like ``. . . is red'' are supposed
to stand for sensible qualities;15 ``. . . is lapping'' is not sensible. So claims
like ``The ball is red'' are supposed to make sense (to the extent that we
can construct a concept of a ball from our impressions) whereas ``The cat
14 I have defended the ®rst of these claims in Nature's Capacities and their Measurement
(see n. 4, above); the second is argued in ``Aristotelian Natures and the Modern
Experimental Method'' (also cited in n. 4, above).
15 The other so-called ``primary and secondary qualities'' are supposed to be sensible as
well.
 Ceteris paribus laws and socio-economic machines 285

is lapping up the milk'' does not, except as a claim more or less about the
locations of a sequence of color patches. Since there are no impressions
derived from Nature from which we could copy causal ideas, we do not
on this story really have any concept of causality. We only think we do.16
I do not think that any part of this story is plausible 200 years later.
There is no evidence for thinking that the idea of a ``sensible quality'' is a
reasonable one to use in our account of Nature.17 I suspect that our
concept of sensible qualities is a back formation from the Cartesian
concept of an ``impression,'' which itself is a philosophical construct with
no scienti®c backing. Philosophers sometimes try to substitute ``occur-
rent'' instead of ``sensible'' but this is equally problematic. In the ®rst
place, the distinction ``occurrent/dispositional'' is itself questionable. But
more importantly, even if it could be drawn, there is no account of why
singular causings are not occurrent nor of why the occurrent is either
more real than the dispositional or more epistemically accessible. The
associationist theory of concept formation provided an account of
epistemic access that did this job for Hume. We rightly do not hold with
associationism nowadays. But there is no good alternative that shows
that causings are not among what can be known.18 The contemporary
Humean attack on causation is, so far as I can see, groundless. It is an
attempt to keep Hume's conclusions even though we do not accept his
premises nor have any reasonable substitute for them. We may have no
adequate theory of concept formation, yet it is clear that associationism
cannot be maintained, and its consequent constraints on what we can
sensibly think and say should not dictate our methodology.

4 An example of an economic machine

The game-theoretic model proposed by Oliver Hart and John Moore in

their ``Theory of Debt Based on the Inalienability of Human Capital''19
16 Hume hypothesized that the idea of causality could be a copy of our impression of our
own internal state of expectation that one event will follow another. In this case we have
a concept, but it is not the concept we usually think it is.
17 Hugh Mellor's attacks on the distinction between occurrent and dispositional properties
support this point of view (H. Mellor, The Matter of Chance [Cambridge: Cambridge
University Press, 1971]), as does Sydney Shoemaker's account of properties as nexes of
causal laws (S. Shoemaker, ``Identity, Cause and Mind'', in Ontological Essays
[Cambridge: Cambridge University Press, 1984]).
18 There are, of course, a number of psychological experiments that show that we can be
misled about causality, as about anything else.
19 O. Hart and J. Moore in Theoretical Economics Discussion Papers, London:
286 The world of economic causes

provides a good example of a blueprint for a socio-economic machine. I

pick this example not because it is especially representative of recent
work in economic theory but rather because the analogy with machine
design is very transparent in this case and the contrast with a covering-
law account is easy to see. The central idea behind the model is that
crucial aspects of debt contracts are determined by the fact that
entrepreneurs cannot be locked into contracts but may withdraw with
only small (in their model no) penalties other than loss of the project's
assets. That means that some debt contracts may be unenforceable and
hence inef®ciency may result, i.e. some pro®table projects may not be
undertaken. Hart and Moore derive a number of results. I shall discuss
only the very ®rst (Corollary 1) as an illustration of how socio-economic
machines give rise to regularities. As Hart and Moore describe, ``Cor-
ollary 1 tells us that inef®ciency arisesP only if either (a) there is an initial
sunk cost of investment K > L0 ( = pi i = 1, . . . , n) and/or (b) the
project's initial returns are smaller than the returns from the assets'
alternative use (in particular, r1 < 11).''20
The model presents a toy machine that if set running repeatedly
generates an economically interesting regularity described in Corollary 1.
The inputs can vary across the identity of individual players, sunk costs
(K), income streams (r1, . . ., rn), liquidation-value streams (11, . . ., 1n),
and the initial wealth of the debtor (w0). The output we are considering is
the regularity described in Corollary 1:
R: All pro®table projects which have no initial sunk costs and whose initial
returns are at least as large as the returns from the alternative use of the assets
will be undertaken.
The model lays out a number of features necessary for the design of a
machine: it tells us (i) the parts that make up the machine, their proper-
ties and their separate capacities; (ii) how the parts are to be assembled;
and (iii) the rules for calculating what should result from their joint
operation once assembled. The components are two game players
(``debtor'' or ``entrepreneur,'' and ``creditor'') who (a) have the same
discount rates; (b) are motivated only by greed; (c) have perfect certainty;
and (d) are perfect and costless calculators. The arrangements of the
players is claustrophobic: two players set against each other with no

STICERD TE (1991), 1233, London School of Economics. Also in the Quarterly

Journal of Economics, 109 (1994), pp. 841±879.
20 Hart and Moore, 1991, p. 19. The result need mention only initial returns because of the
simplifying assumption Hart and Moore make that if the return at any period is greater
then the liquidation income at that period, this will continue to be the case in all
subsequent periods.
 Ceteris paribus laws and socio-economic machines 287

possible interaction outside. Later extensions consider what happens

when either the debtor or the creditor has pro®table reinvestment
opportunities elsewhere, but these are ®xed opportunities not involving
negotiations with new players. Another extension tries in a very ``indirect
and rudimentary fashion'' (Hart and Moore 1991, p. 40) to mimic with
just the two players what would happen if the debtor could negotiate
with a new creditor. Other assumptions about the arrangement are
described as well; for instance, as part of the proof of R it is assumed
``for simplicity'' that the debtor ``has all the bargaining power'' in the
original negotiation (p. 17). The central features of the arrangement are
given by the rules laid out for the renegotiation game, plus a number of
further ``simplifying'' assumptions about the relations among capital
costs, project returns and liquidation returns that contribute to the proof
of R.
Rules for calculating how parts function together vary widely across
domains.21 Component forces, which have the capacity to produce
motions, add vectorially. Simultaneous-equation models in econometrics
suppose that when a number of mechanisms act together, the equations
associated with each of them separately must all be satis®ed at once. In
electronics there are well-known theorems that reduce any complicated
arrangements of components to a simple circuit with known behavior.
And in game theory various concepts of equilibria tell us how to calculate
the outcomes of the joint action of players with speci®ed capacities (e.g.
perfectly or imperfectly informed, costless calculators, etc.). For the Hart
and Moore model it is supposed that the moves of the players together
constitute a subgame-perfect equilibrium.
The model is less informative about the two remaining features
necessary for describing the operation of a machine: (iv) what counts as
shielding and (v) how the machine is set running. As we have seen, Hart
and Moore describe a very closed world. There are just two players and
they are locked into the very special game they are playing. The shielding
conditions are implicit: nothing must occur that distorts the rules of the
game and nothing must affect the choice of debt contracts other than the
requirement that they constitute a subgame-perfect equilibrium. Re-
peated running simply means playing the game again and again. This is
consistent with my suggestion earlier that models like this are described
at a very high level of abstraction. The model speci®es abstract functional
relations between the parts that can be instantiated in various different
institutional arrangements; what counts as shielding will depend heavily

21 This is part of the reason it is dif®cult to formulate a general informative philosophical

account of mechanisms and their operation.
288 The world of economic causes

on what the speci®c material instantiation is. This is especially true in

game-theoretical models, where few clues are given about what real
institutional arrangements can be taken to constitute any speci®c game.
To derive the regularity R, Hart and Moore show that in their model
all subgame-perfect equilibria for their date-zero renegotiation game
satisfy a set of conditions (Proposition 1) that in turn imply Corollary 1.
The derivation proceeds by unpacking the concept of equilibrium in an
exogenously given extensive form game. In particular the derivation
employs no general laws that might be mistaken for claims about
regularities. There are no empirical generalizations nor any equations on
the analog of SchroÈdinger's or Maxwell's; there are not even any
theorems of game theory written down. Rather the argument employs
concepts and techniques of game theory (some of which are of course
validated by reference to theorems) plus the general characterization of
equilibrium to derive constraints on what the pay-offs will be from any
equilibrium in the game described.
Turn now to the questions about ceteris paribus conditions and the
regularity theory of laws. R holds (at best) only ceteris paribus: if
conditions like those of the model were to occur repeatedly, then in those
situations all pro®table ventures with no sunk costs and good enough
initial returns would be undertaken. On the picture of Natural Law that I
have been advocating, something like the converse is true as well. If a
regularity like R is to obtain as a matter of law, there must be a machine
like the one modeled by Hart and Moore (or some other, with an
appropriate structure) to give rise to it. There are no law-like regularities
without a machine to generate them. Thus ceteris paribus conditions have
a very special role to play in economic laws like R. They describe the
structure of the machine that makes the laws true.
The relation of laws to models I describe is familiar in economics,
where a central part of the theoretical enterprise consists in devising
models in which socio-economic regularities can be derived. But it is
important to realize how different this is from the regularity theory.
Look back to the regularity theory. R is not a universal association that
can be relied on outside various special arrangements. On the regularity
theory law-likeness consists in true universality. So there must be some
universal association in the of®ng or else R cannot be relied on at all,
even in these special circumstances. If an association like R appears to
hold in some data set, that cannot be a matter of law but must be viewed
as merely a chance accident of a too-small sample unless there is some
kind of true universal association to back it up.
The difference between an account of Natural Law in terms of
nature's capacities and machines and the regularity account is no mere
 Ceteris paribus laws and socio-economic machines 289

matter of metaphysics. It matters to method as well, both the methods

used in the construction of theory and those used in its testing. R tells
us that ceteris paribus all pro®table ventures will be taken up ± except in
conditions (a) and (b). The regularity theory invites us to eliminate the
ceteris paribus clause by extending this list to include further factors ±
(c) (d) (e), . . . (x) ± until ®nally a true universal is achieved: ``All
pro®table ventures that satisfy (a) (b), . . . (x) will be taken up.'' This
way of looking at it points the investigation in an entirely wrong
direction. It focuses the study on more and more factors like (a) and (b)
themselves rather than on the structural features and arrangements like
those modeled by Hart and Moore that we need to put in place if we
want to ensure ef®ciency.
The regularity theory also carries with it an entourage of methods for
testing that have no place here. When is a model like that of Hart and
Moore a good one? There are a large number of different kinds of
problems involved. Some are due to the fact that theirs is a game-
theoretic model; these are to some extent independent of the issue raised
by the differences between a regularity and a capacity view of laws. The
advantage to game theory is that it makes the relationship between the
assumptions of the explanatory model and laws like R that are to be
explained in the model very tight. The results that are ``derived'' in the
model are literally deduced. The cost is that the rules of the games that
allow these strict deductions may seem to be very unrealistic as represen-
tations of real-life situations in which the derived regularities occur. As
Hart and Moore say about their own model, ``The game may seem ad
hoc, but we believe that almost any extensive form bargaining game is
subject to this criticism'' (1991, p. 12).
Let us lay aside these special problems involving games and think
about the relations between models and the results derived within them
more generally. The point is that the results derived may be rendered as
regularity claims, but the relationship between the structures described
by the model and the regularities it gives rise to is not again itself one of
regular association. So our whole package of sophisticated techniques ±
mostly statistical ± for testing regularity claims are of no help in the
decisions about choice among models. How do we decide? As far as I
can see we have no articulated methodology, neither among philoso-
phers nor among economists (though we may well have a variety of
unarticulated methods). My reasons for attending to the metaphysics of
economic laws stem primarily from these methodological concerns. So
long as we are in the grip of the regularity view ± which happens to
economists as well as to philosophers ± we are likely to restrict our
efforts at methodology to re®ning ever more precisely our techniques at
290 The world of economic causes

statistics and to leave unconsidered and unimproved our methods for

model choice.

5 Three aspects of socio-economic machines

``Socio-economic laws are created by socio-economic machines.'' The

slogan points to three distinct theses, which are separable and can be
argued independently. The ®rst, the one that I am most prepared to
defend, follows Aristotle in seeing natures as primary and behaviors,
even very regular behaviors, as derivative. Regular behavior derives from
the repeated triggering of determinate systems whose nature stays ®xed
long enough to manifest itself in the resulting regularity. This feature
does not point particularly to a machine analogy, though, in opposition
to an organic one, as is apparent from the work of Aristotle himself.
Organic analogies usually suggest a kind of irreducible holism: the
behavior of the components when separated has little to teach about how
they work together in the functioning organism. By contrast, the
machine analogy stresses the analytic nature of much economic thought.
This is the second thesis: much of economic work is based on the hope
that we can understand different aspects of the economy separately and
then piece the lessons together at a second stage to give a more complete
account. This idea is central to the use of idealization and correction that
is common throughout economics. Bounded rationality is expected to be
like unbounded but with modi®cations; international trade at best tends
to move prices towards each other but they are often modeled as being
equal; in planning models the planner is often assumed to have no other
goal than social welfare just as the ®rm manager is assumed to maximize
shareholder value of the ®rm; and so forth. I do not want to urge this
second thesis suggested by talking of socio-economic machines as
vigorously as the ®rst. There is no guarantee that the analytic method is
the right method for all the problems that economics wants to treat. But
it is, I think, by far and away the one we best know how to use.
The third thesis is one about which evidence is divided. Ordinary
machines do not evolve. They have to be assembled, and the assembly
has to be carefully engineered. Is the same true of socio-economic
machines? I do not intend in using the description to urge that the answer
must be ``yes.''22 One of the most clear-cut examples of a ``designed''
institution in economics is the International Monetary Fund (IMF),

22 I owe the following discussion to Mary Morgan.

 Ceteris paribus laws and socio-economic machines 291

which resulted from negotiations over two radically different designs put
forward for an international body to run post-Second World War
international monetary arrangements. In large part the design adopted at
Bretton Woods in 1944 was for a ``fund'' from which member countries
could draw previously deposited reserves, rather than a ``bank'' which
could create credit on deposits. Economic historians have tended to rate
the institution a failure because of its design faults: (1) it was not devised
to effect the transition from the war years to the peace years; (2) it ran
out of reserves (as predicted by Keynes) in the 1960s and so had to create
SDR (special drawing rights) to service the growth in the international
economy; and (3) it was unable to bring pressure to bear on surplus
countries to revalue, compromising its ability to avoid damaging cur-
rency crises and oversee orderly changes in exchange rates, leading
ultimately to the collapse of the arrangements in 1971. By contrast,
economic historians have tended to rate the gold standard of the pre-
1914 economy as a much more effective institution ± an institution which
was not designed but which evolved gradually over the nineteenth
century. They suggest that this institution worked well because its
evolutionary character allowed it to be operated with considerable
¯exibility and tacit knowledge, rather than because it stuck to some
supposed ``rules of the game'' or was operated according to some agreed-
upon economic theory.23

6 Concluding remarks

I began with the conventional claim that the laws of economics hold only
ceteris paribus. This is supposed to contrast with the laws of physics. On
the regularity account of laws this can only re¯ect an epistemological
difference between the two: economists simply do not know enough to
®ll in their law claims suf®ciently. I have proposed that there is a
metaphysical difference as well. Laws in the conventional regularity sense
are secondary in economics. They must be constructed, and the know-
ledge that aids in this construction is not itself again a report of some
actual or possible regularities. It is rather knowledge about the capacities
of institutions and individuals and what these capacities can do if
assembled and regulated in appropriate ways.

23 A standard good source is J. Foreman-Peck, A History of the World Economy:

International Economic Relations since 1850 (1993).
292 The world of economic causes

Does this really constitute a difference between economics on the one

hand and physics on the other? I think not. It is sometimes argued that
quantum ®eld theory and the general theory of relativity function like
true covering-law theories. They begin with regularities that are both
genuinely universal and true (or, true enough); the phenomena to be
explained are just special cases of these very general regularities. Perhaps.
Most of physics works differently. Like economics, physics uses the
analytic method: we come to understand the operation of the parts ± for
example, Coulomb's force, the force of gravity, weak and strong nuclear
forces, or the behavior of resistors, inductors and capacitors ± and we
piece them together to predict the behavior of the whole. Even physics, I
would argue,24 needs ``machines'' to generate regularities ± machines in
the sense of stable con®gurations of components with determinate
capacities properly shielded and repeatedly set running. If this is correct
then differences in the metaphysics of Natural Laws that I have been
describing are not differences between economics and physics but rather
between domains in which the covering-law model obtains and those in
which the analytic method prevails. Economics and physics equally
employ ceteris paribus laws, and that is a matter of the systems they
study, not a de®ciency in what they have to say about them.
24 For a defense of this claim see N. Cartwright, ``Fundamentalism vs the Patchwork of
Laws,'' Proceedings of the Aristotelian Society (1994); and ``Aristotelian Natures and the
Modern Experimental Method'' (cited in n. 4, above).
15 Tendencies, laws, and the
composition of economic causes

DANIEL M. HAUSMAN1

John Stuart Mill is an empiricist and an inductivist. He believes that the

grounds for beliefs concerning matters of fact are ultimately perceptual
experiences and that generalizations are justi®ed inductively by observa-
tion of their instances and implications. Yet Mill believes that inductive
methods are not directly applicable to most subject matters. His methods
of induction are suitable only to domains in which few causal factors are
at work, while most subject matters involve the simultaneous action of
many causal factors. The possibilities of experimental intervention
increase the range of direct inductive inquiry, but that range is still
limited.
Mill believes that scienti®c knowledge of complex subject matters can
nevertheless be attained. If one is able to determine inductively the laws
governing single causal factors, then one can deduce what the conse-
quences of multiple causes acting simultaneously will be. Mill calls this
procedure ``the deductive method'' or ``the method a priori,'' but both
names are misleading. The deductive method is, in fact, an indirect
inductive method, in which the laws of individual causes are separately
determined by inductive methods. The role of deduction is to determine
what follows from these laws in complex circumstances. The evidence
that inductively supports the premises of a deductive argument is
supposed to be the inductive basis for one's belief in the argument's
conclusions (1843, bk. 2, ch. 3, §3). In Mill's words:
When an effect depends on a concurrence of causes, these causes must be studied
one at a time, and their laws separately investigated, if we wish, through the

1 This essay derives from Hausman (1995). Sections 2 and 3 are largely new. Nancy
Cartwright, Wade Hands, Uskali MaÈki, and Elliott Sober provided helpful criticisms of
earlier drafts. This work was carried out while I was the Ludwig Lachmann Fellow at the
London School of Economics, and I would like to thank the Charlottenburg Trust for its
support.

293
294 The world of economic causes

causes, to obtain the power of either predicting or controlling the effect; since the
law of the effect is compounded of the laws of all the causes which determine it.
(1843, bk. 6, ch. 9, §3)

If one wants to ``obtain the power of predicting or controlling'' an effect

such as projectile motion through understanding its causes, one needs to
investigate separately the separate causal factors (gravity, momentum,
friction) and their laws.
This notion of ``compounding of the laws of all the causes,'' of
deducing the consequences of the concurrence of a plurality of causes, is
problematic. Mill takes explanations in dynamics as paradigmatic and
does not pause over the notion of adding up the effects of different
causes. But the notion of ``compounding'' is not simple or straightfor-
ward even in the case of classical mechanics. I shall disentangle some of
the principles Mill invokes in attempting to explain how scienti®c
knowledge of complex subject matters is possible, and I shall argue that
his account fails. Some of the problems in acquiring knowledge of
complex subject matters arise from details of Mill's philosophy of
science, but some remain for non-Millians to grapple with, too.

1 The deductive method in economics

As Mill recognized, one can ®nd regularities in complicated phenomena.

These regularities, which Mill called ``empirical laws,'' are a valuable part
of science, because they constitute data which theories should explain
and because they may be of use. But empirical laws are not explanatory,
and they are a precarious basis for prediction. Until they are linked to
underlying causal laws, one does not know when they can be relied upon
and when they are likely to break down. So scientists should not rest
content with empirical laws. They should seek to uncover what Mill
called ``causal laws,'' and they cannot succeed in doing so by applying
direct inductive methods to complex subject matters. When I speak of
``laws,'' I shall be referring to causal laws.
Social phenomena are particularly inappropriate candidates for the
method of ``speci®c experience'' or ``the method a posteriori,'' because
they are complex, available in limited varieties, and not subject to
appreciable experimental manipulation. Physicists might stop there with
expressions of pity for the plight of the social theorist. But social
phenomena and especially economics are Mill's main interest, and he
wants to explain how one can acquire scienti®c knowledge of them. The
 Tendencies, laws, and the composition of economic causes 295

explanation lies in Mill's ``direct'' deductive method.2 This method is not

applicable to the general science of society, because there are too many
causes ever to ascertain all their laws and to determine their combined
effects, but it will work in subject matters such as economics in which, in
Mill's opinion, there are few really signi®cant causes.
When Mill ®rst introduces the deductive method in Book III of A
System of Logic, he envisions scientists deriving the laws governing
complex phenomena from the laws of all the relevant causes. Suppose,
for example, Wilson is sick, and one would like to know whether
penicillin will help cure Wilson.3 The method a posteriori demands an
inquiry into whether others with symptoms resembling Wilson's recov-
ered more often or more rapidly when given penicillin. The method a
priori in contrast draws upon knowledge of the causes of Wilson's
symptoms and of the operation of penicillin to decide whether penicillin
will help cure Wilson. Both methods are ``empirical'' and involve testing.
The difference is that the former attempts to use experiment or observa-
tion to learn about the complex phenomenon directly, while the latter
employs observation to study the relevant component causal factors.
In an example such as this one, the deductive method seems unobjec-
tionable, but in economics, causal factors that are known to be signi®cant
are left out of the story. As Mill was well aware, economic agents may be
motivated by all sorts of passions ± whether they be patriotic, malevolent,
benevolent, or neurotic ± which are left out of economics. Mill seems to
be of two minds about whether the omission of relevant causal factors is
scienti®cally acceptable. On the one hand, he criticizes members of the
``school of Bentham'' (including his father, James Mill) for theorizing
about government without incorporating all the causes, whether signi®-
cant or not (1843, bk. 6, ch. 8, §3). But when it comes to economics, Mill
apparently recommends just the methodological practice that he con-
demns in his father. For the correct method of including all the causes

2 The so-called inverse deductive method seems a strained attempt to defend large-scale
historical speculation of the sort that Mill admired in Comte. In principle there is nothing
strained or questionable about the inverse deductive method. Like the direct deductive
method, one ``deduces'' a social regularity from the underlying causal laws and a
description of the particular circumstances. What makes Mill call it the ``inverse deductive
method'' is just that the social regularity is ®rst derived from observation, rather than
obtained from the deduction, so that the deduction rather than the observation serves as a
veri®cation that the regularity is indeed a derivative causal law. But in his discussion of
the inverse deductive or historical method, Mill greatly relaxes the demand that the social
regularity be deduced from underlying causal laws. Mill argues that merely showing that
empirical social regularities are not ruled out by the causal laws is enough to lend some
weight to them (1843, bk. 6, ch. 10).
3 Compare this to Mill's own example (1843, bk. 3, ch. 10, §6).
296 The world of economic causes

``within the pale of the science'' is not feasible. Economists must set their
sights lower and aim only at a hypothetical science of tendencies which is,
in Mill's view, generally ``insuf®cient for prediction'' yet ``most valuable
for guidance'' (1843, bk. 6, ch. 9, §2).
Let us call this sort of deductive method, that so closely resembles the
method of Mill's father, ``the inexact deductive method,'' because it
incorporates only some of the causes. Mill defends this method as follows:
The motive which suggests the separation of this portion of the social phenomena
from the rest, and the creation of a distinct branch of science relating to them, is,
that they do mainly depend, at least in the ®rst resort, on one class of
circumstances only; and that even when other circumstances interfere, the
ascertainment of the effect due to the one class of circumstances alone is a
suf®ciently intricate and dif®cult business to make it expedient to perform it once
for all, and then allow for the effect of the modifying circumstances; especially as
certain ®xed combinations of the former are apt to recur often, in conjunction
with ever-varying circumstances of the latter class. (1843, bk. 6, ch. 9, §3)
The defences Mill offers here for employing this inexact deductive
method seem to be (1) practical ± there is no alternative (2) metaphysical
± although the results are only hypothetical, the tendencies persist even
when there are other disturbing causes, and (3) pragmatic ± this is an
ef®cient way of theorizing, and more order can be found this way than in
any other.4 I shall be questioning these defences.
In the case of economics, theorists following the deductive method ®rst
borrow basic ``laws'' from the natural sciences or from psychology,
which Mill regards as an introspective experimental science. Then
theorists deduce what follows from them in various circumstances.
Finally, veri®cation is essential (though not in order to test the basic
laws, which are already established and could not be cast in doubt by the
empirical vicissitudes of a conclusion deduced from a partial set of
causes). It is unclear whether veri®cation is necessary in order to regard
the deductively derived generalizations as economic laws, or whether
veri®cation merely determines the applicability of these laws.5
4 Surely Mill's father might have given the same argument in his own defense. There is an irony
here in the fact that recent extensions of neoclassical economic models to politics largely
recapitulate James Mill's account of political behavior. See for example Buchanan (1975).
5 Compare Mill (1836), pp. 325±6 and Mill (1843), bk. 3, ch. 9, §3 and bk. 6, ch. 9, §1, and
see De Marchi (1986) and Hutchison (1998). Mill writes, ``To verify the hypothesis itself a
posteriori, that is, to examine whether the facts of any actual case are in accordance with
it, is no part of the business of science at all, but of the application of science'' (1836,
p. 325). It is not clear from the text of Mill's writings whether Mill regarded the deductive
method as a distinctive method of theory appraisal or whether he regarded it as the
implementation of standard inductive methods when theorizing about complex phe-
nomena. I am indebted to Abraham Hirsch for my understanding of these problems.
 Tendencies, laws, and the composition of economic causes 297

Applying the deductive method in economics is even messier than the

discussion so far suggests. The laws that one derives are not only inexact;
they are sometimes drastically in con¯ict with the phenomena. These
empirical inadequacies are only to be expected, since many causal factors
are left out of the derivation. Furthermore the premises in the deduction
do not consist exclusively of established laws and true descriptions of the
relevant circumstances. Frequently included among the non-law premises
are extreme simpli®cations, such as claims that commodities are in®nitely
divisible or that knowledge is perfect. Since some of the implications of
such premises are bound to be wildly off the mark, one wonders what
evidential weight such deduction could have. Presumably the fact that
one can derive a purported causal law L for a complex system from such
polyglot premises give one reason to accept L only if the simpli®cations
are dispensable or if they are in some sense reasonable approximations in
the particular domain.

2 Inexact laws

The supposed ``laws'' of the component causes are also problematic,

since it is questionable, at least in the case of political economy, whether
they are really laws at all. For example, Mill believes that the most basic
law of political economy is that people seek more wealth, yet he asserts
that it is absurd to maintain that people do in fact always seek more
wealth (1843, bk. 6, ch. 9, §3)! What's going on?
``Laws'' such as ``People seek more wealth'' are supposed to identify
relevant causal factors. These may be counteracted by other factors and
prevented from operating, and so, for example, people will not always
seek more wealth. But a desire for more wealth remains a signi®cant
causal factor. When economists state that people desire more wealth they
are stating this truth.
Exactly what truth is this? From the premise, ``Desire for wealth is a
causal factor,'' how is any conclusion about what people will do
supposed to follow? One possibility is that the claim that people seek
more wealth is a counterfactual claim. If other causal factors were
absent, or if they were to balance, or if they were not to push too strongly
in another direction, then people would always seek more wealth. One
can use such a subjunctive in explaining or predicting things, if, in the
circumstances involving the phenomena to be explained or predicted,
other ``disturbing'' causal factors are absent, or they balance or they do
not push too strongly in another direction. Another possibility, which
298 The world of economic causes

differs only in its metaphysics, is that the claim that people seek more
wealth is true only when quali®ed with a ceteris paribus clause: if other
causal factors are absent, or they balance, or they do not push too
strongly in another direction, then people always prefer a larger com-
modity bundle to a smaller one. When the ceteris paribus condition is
met, such a quali®ed law can be used to explain and to predict
phenomena.6
According to both proposals if some antecedent (ceteris paribus)
condition is (or were) met, then everyone does (or would) seek more
wealth. This antecedent condition is vague. One can specify some of it,
but one cannot specify all of it. For example, if an agent is gagged and
chained to a wall, then the antecedent condition is not met. But if an
agent is competent, free, and knowledgeable about the options, the
antecedent condition may still fail to be met. The set of possible
interferences or ``disturbing causes'' is heterogeneous and impossible to
specify completely. I argued that quali®ed laws can nevertheless have
de®nite meanings and truth conditions and that it is possible to gather
evidence that con®rms or discon®rms them. In effect, I argued that
inexact laws are exact laws inexactly formulated and known, and that by
examining the performance of the generalization without its vague
quali®cations one can ®nd reason to af®rm or to deny that one has a
genuine (but inexact) law (Hausman 1992, ch. 8; 1981, ch. 7).
Although I have been wedded to this picture for a long time, I have
not been blind to its unsatisfactory features:
1. It is questionable whether any of the ``inexact'' claims of economics or
of any other discipline are inexact laws. Is there any precise quali®ed
or counterfactual claim that is inexactly expressed as ``People prefer
more wealth''?
2. Why should one believe that there are any such claims? Satisfying all
the justi®cation conditions might make the hope that there are such
claims less unreasonable, but it is hard to see what would constitute
good evidence that the claims of economics are truly inexact laws.
3. This framework sits uncomfortably with the practice of economists.
When do they explore whether these generalizations meet conditions
such as the ones I have formulated elsewhere? Why don't they address
the question of what are the main factors that interfere or disturb the
fundamental ``laws'' of economics? Is it only a failure of nerve and a

6 Because of metaphysical qualms about counterfactuals, I preferred the non-counterfactual

ceteris paribus construal, but the differences between regarding such claims as subjunctive
or as quali®ed are not germane to the discussion here.
 Tendencies, laws, and the composition of economic causes 299

desire to dodge uncomfortable questions that leads economists so

consistently to avoid the terminology of laws?

Although economists have unreasonable philosophical scruples about

the explicit use of causal language, it is more natural to claim that the
desire for more wealth is one of the causal factors in¯uencing choices
than to claim that there is some complicated law relating desire for
wealth to choices. Since it is independently plausible to take the explana-
tion of an event to consist in citing its causes, why not jettison the whole
complicated story told above and say simply that economists explain by
citing causes and predict by inferring effects from causes?

3 Tendencies and mechanisms

It is natural to describe the component causes as giving rise to ``ten-

dencies.'' This is the interpretation of Mill defended by Nancy Cartwright
(1989, esp. ch. 4), Uskali MaÈki (1992, and esp. 1993) and Geert Reuten
(1996). Drawing on the work of Roy Bhasker, Tony Lawson defends a
related view that science should formulate ``transfactuals'' that describe
the ``non-empirical activity'' of mechanisms that are operating whether
or not their operation manifests itself (1997, pp. 23, passim; see also
Bhaskar 1975). On this interpretation, claims such as ``People seek more
wealth'' express tendencies, capacities, or mechanisms rather than laws.
Such tendencies sometimes give rise to regularities, and it is by virtue of
the connections between tendencies and regularities that claims about
tendencies can be tested. The tendencies are not, however, themselves
regularities. For example, people's tendency to seek more wealth might
give rise to nearly universal wealth-seeking behavior in circumstances
such as those that characterize stock markets, while the tendency gives
rise to little wealth-seeking behavior in Buddhist monasteries. Yet it is
equally ``there'' in both contexts. Unlike claims concerning regularities,
the tendency claim is not restricted in scope, and the tendency remains
intact even when its manifestation is masked or counteracted.
If one models the principles of economics as claims about causal
mechanisms, capacities, or tendencies rather than directly as inexact
laws, then the deductive method would be more directly a matter of
compounding causes rather than a matter of deducing (derivative laws).
Although framing the problem of theorizing about complex systems this
way may ®t the language of economists more naturally, has one made
any real progress? Has one escaped the task of explaining how apparently
300 The world of economic causes

false or inexact principles can be both explanatory and a reliable basis

for prediction or has one merely postponed it?
It might be argued that an advantage of jettisoning talk of laws is that
one no longer faces any imperative actually to ®nd an exact account.
What counts on a causal model is identifying deep discriminating causes
and discovering the mechanisms according to which they operate. To do
this does not require that one produce an account that is freed from all of
the inessential falsehoods of current accounts. On a view that takes the
principles of economics to be inexact laws, in contrast, one does not
know precisely what the principles mean until that distant ideal future
when their ceteris paribus clauses have been ®lled in. Although explana-
tions in economics can, no doubt, be improved, it seems absurd to
maintain that they remain inadequate and ill-understood until their
ceteris paribus conditions can be fully speci®ed.
Yet, without knowing all the ``disturbing causes,'' how con®dent can
one be that the mechanism one identi®es is really operating and
responsible for the phenomena to be explained? If there is no inexact law,
then on many analyses of causation, one has not correctly identi®ed a
causal factor. If one has no sense of what the inexact laws are, to what
extent can one be said to understand the mechanism?
An apparently much more forceful argument can be made for the
tendency construal. According to the inexact-law model, a claim such as
``people seek more wealth'' tells one about choice only when the ceteris
paribus condition is met. ``But that will not do . . . Even if these
regularities did hold ceteris paribus ± or, other things being equal ± that
would have no bearing on the far more common case whether other
things are not equal'' (Cartwright 1989, p. 177; see also Lawson 1996,
esp. pp. 408±9).
Consider a more transparent example from simple physics. On the
inexact-law view, Galileo's law tells one how bodies fall (or would fall)
when there are (or if there were) no non-gravitational forces acting on
them. So, by itself, Galileo's law says nothing about how bodies fall in the
real world where other forces (such as air resistance) are always acting on
them. Lawson would conclude that one should interpret Galileo's law as
a transfactual describing ``something that is going on, that is having an
effect, even if the actual (possibly observable) outcome is jointly co-
determined by (possibly numerous) other in¯uences'' (1996, p. 408).
Cartwright rejects transfactuals, and she emphasizes instead that ``part of
the point of taking capacities seriously as things in the world, and not
just particularly strong modalities, [is] that they should remain intact
from one kind of situation to another'' (1989, p. 163).
One can thus argue in favor of theorizing in terms of tendencies on the
 Tendencies, laws, and the composition of economic causes 301

ground that, in contrast to the inexact-law model, one can talk about
factors that may be intact or even ``operating'' even when they are
canceled out or outweighed by other factors. But I think that this
argument is a mistake. Rather than an objection to a quali®ed or
subjunctive universal generalization construal, the fact that inexact laws
apply only when there are no ``disturbing causes'' seems to me a virtue of
talking in terms of laws rather than tendencies. Galileo's law does not
suf®ce to make inferences about what happens when other forces are
acting. One also needs principles concerning how to combine different
factors. And it is worth separating the laws of the individual causes from
the laws of their composition. In the case of mechanics, the principles of
composition are very simple ± it is just a matter of vector addition. In
economics, on the other hand, principles of composition are complicated
and controversial. These rules of composition should be distinguished
from the individual principles of economics themselves. There are no
grounds here for rejecting an inexact-law account.
In a similar vein, Nancy Cartwright has argued that ``Causal inter-
actions are interactions of causal capacities, and they cannot be picked
out unless capacities themselves can be recognized'' (1989, p. 164). If a
causal law is simply a universal generalization, then the only thing that
one can say about causal interaction is that the consequents in inexact
laws sometimes fail to hold. There is no way to distinguish between
failures due to causal interaction and failures due to errors in one's
purported laws or resulting from the in¯uence of some further disturbing
cause.
Like the last objection, this one is, I believe, mistaken. As soon as one
knows some of the principles of composition, one can make predictions
about what will happen when causal factors are combined. When these
predictions are disappointed, one can investigate whether some disturb-
ing cause (some violation of the ceteris paribus condition) was respon-
sible, or whether the factors fail to combine in accordance with the
principles of composition. Only in the latter case will there be a causal
interaction.
I am not sure whether one understands methodology and explanation
better if one focuses on laws or if one focuses on tendencies or even
whether it turns out to matter. The decision ultimately depends on what
theories of scienti®c explanation and theory construction are most
satisfactory. At a super®cial level both accounts appear to be capable of
explaining how people's pursuit of more wealth can be a fundamental
cause, even though there is no universal law saying that people seek more
wealth. I do not think that Mill's writings clearly commit him to either a
tendency or a ceteris paribus law view of component causes, and indeed
302 The world of economic causes

Mill does not distinguish clearly between these interpretations of com-

ponent causes. In the past I have preferred the ceteris paribus law view of
component causes, because it is more modest metaphysically, and
because I have been leery of taking the revolutionary step of demoting
laws from their central role in the understanding of science. But a
decision does not need to be made here, since both the law and tendency
interpretations permit one to describe the dif®culties that arise in
``compounding'' causes.

4 The composition of causes

To make possible the ``deduction'' of the net force in mechanics, one needs
to know more than just how the causes act separately ± that is, the separate
force laws. One also needs to know the law governing their combination,
and this law cannot be derived from knowledge of how forces act
separately. To speak, as Mill does, of a deductive method, is misleading
because the law governing the conjoint operation of causes cannot be
deduced from the laws governing the component causes separately.
How could Mill have thought otherwise? In Book 3 of his Logic Mill
writes:
Now, if we happen to know what would be the effect of each cause when acting
separately from the other, we are often able to arrive deductively, or a priori, at a
correct prediction of what will arise from their conjunct agency. To render this
possible, it is only necessary that the same law which expresses the effect of each
cause acting by itself shall also correctly express the part due to that cause of the
effect which follows from the two together. This condition is realized in the
extensive and important class of phenomena, commonly called mechanical,
namely, the phenomena of the communication of motion (or of pressure, which is
the tendency to motion) from one body to another. In this important class of
cases of causation, one cause never, properly speaking, defeat or frustrate
another; both have their full effect. . . This law of nature is called, in dynamics,
the principle of the Composition of Forces: and, in imitation of that well-chosen
expression, I shall give the name of the Composition of Causes to the principle
which is exempli®ed in all cases in which the joint effect of several causes is
identical with the sum of their separate effects. (1843, bk. 3, ch. 6, §1)

Mill is claiming that one can deduce what will happen when causes are
combined from the laws of their separate action plus the principle of the
Composition of Causes, which says that ``the joint effect of several causes
is identical with the sum of their separate effects.'' I shall call this
principle, ``the additivity assumption.'' In mechanical phenomena, one
 Tendencies, laws, and the composition of economic causes 303

``adds up'' the forces (or the accelerations they cause) and gets the right
answer. It would be more informative to call the method Mill is
espousing ``compositional'' rather than ``deductive'' for the derivation of
the combined effect is a process of adding and subtracting.
If one assumes additivity, then the deductive method can work, but it
is not clear what additivity means outside of the special case of
mechanics. In some instances one can take it literally as, for example,
when one interprets a change in quantity demanded as a sum of an
income and a substitution effect.7 But how is one supposed to ``add up''
the consequences for behavior of, for example, uncertainty, time pre-
ference, diminishing marginal rates of substitution and diminishing
returns? What goes on in much of economics is more like deducing than
adding, and it is unlike what Mill envisioned.
Even if one could make clear what was involved in ``adding'' causal
factors in economics, it is hard to see what justi®cation there might be for
assuming that the causes of some economic phenomenon are additive. In
mechanics, it seems easy to defend the assumption that causes are
additive. One can alter the air-resistance and measure the change in the
acceleration of a falling body. The predictions of a formula derived with
the help of vector addition from Galileo's law and laws of friction are in
good agreement with the data. But the same possibilities of experimenta-
tion that show that one has the correct combined law also make the
deductive method unnecessary: Mill's methods of induction can be
applied directly to the combined law. Proceeding by studying the
component causes separately might have many advantages, but it is not
needed to justify the claim that the formula combining gravitational and
frictional forces is indeed a causal law.
What justi®es the compositional assumption when one cannot directly
establish the combined law? The effect of multiple causal factors acting
together might be completely different from ``sum'' of their separate
effects. Causal factors may interact, and Mill has provided little reason
to believe that tendencies, such as the tendency to seek more wealth, are
still ``acting'' in the presence of other causes. One might think that one
could answer this skepticism about the possibility of developing eco-
nomics deductively merely by successfully carrying out the steps of the
deductive method. And I think that in fact Mill is con®dent of the
possibility of developing economics deductively because of how neatly
his and Ricardo's economics follows from simple and plausible premises.

7 But note that demand curves are decomposed into income and substitution effects rather
that built up from them. As I discuss below, much of economics does not in fact involve
``adding'' of the sort that Mill has in mind.
304 The world of economic causes

For example, given the capitalist's drive for pro®ts, the worker's inclina-
tion to breed, and diminishing returns in agriculture, Ricardo and Mill
held that rates of pro®t should decline, and population and rents should
increase. But the classic argument for diminishing rates of pro®ts and
increasing population and rents is not aptly described as a procedure of
adding component causes, because none of the three factors violates any
of the ceteris paribus conditions attached to the other factors. For
example, diminishing returns is not a force that needs to be added or
subtracted from the drive to maximize pro®ts, and the latter is not added
to or subtracted from the propensity to propagate. One needs deductive
logic rather than an additivity assumption to derive the economic
conclusion concerning pro®ts and rents, and so the success (such as it
was) of an economics based on such a derivation does not support the
additivity assumption.8

5 Can the deductive method work?

Even if Mill's economics followed his compositional method, its apparent

success would not vindicate the method. Suppose that the derivation of
diminishing rates of pro®t and increasing rents did exemplify Mill's
deductive method and suppose (contrary to nineteenth century data),
one found that the rate of pro®t did diminish and rents and population
did increase. Would one then be justi®ed in regarding this trend as a
causal law? The demands of the deductive method would have been met,
but what reason is there to believe that the factors mentioned cause the
diminishing rates of pro®t and increasing rents? Empirical studies only
establish the existence of an empirical law. To believe that one has a
causal law, one has to believe that the deduction correctly displays how
the individual causes act together. But this is what needs showing. Mill
has no answer for those who doubt whether causal laws of complex
phenomena such as economies can be deduced from the laws of the
separate causes.
If the derivation of economic laws were a matter of strict logical
deduction from a set of true premises, these qualms would be unfounded.
But, as we have seen, the derivation of economic laws depends on
8 Deductive logic is not enough. One also needs some sort of persistence or non-interaction
assumption to justify the claim that the separate causal factors continue to operate in the
presence of other factors. Additivity is assumed in the treatment of technological
improvements, but that part of the theory is so unsuccessful that no support for the
deductive method can be found in it.
 Tendencies, laws, and the composition of economic causes 305

incomplete premises, simpli®cations concerning initial conditions, and a

vague additivity assumption. So there are ample grounds for doubt. A
messy derivation of an economic generalization from simpli®cations,
plausible ceteris paribus laws, and an additivity assumption can increase
one's con®dence in the correctness of the generalization, but Mill believes
that science should provide inductive proof. Unless there are grounds to
take the component causes as additive, the deductive method does not
provide the sort of decisive supporting argument that Mill thought
sciences ought to provide for their conclusions. Mill might appear to
close this gap in his argument for a deductive economics in the following
famous passage:
The laws of the phenomena of society are, and can be, nothing but the laws of the
actions and passions of human beings united together in the social state. Men,
however, in a state of society, are still men; their actions and passions are
obedient to the laws of individual human nature. Men are not, when brought
together, converted into another kind of substance, with different properties; as
hydrogen and oxygen are different from water. . .Human beings in society have
no properties but those which are derived from, and may be resolved into, the
laws of the nature of individual man. In social phenomena the Composition of
Causes is the universal law. (1843, bk. 6, ch. 7, §1)
There are two serious problems with this quotation. First, since one is
not, in Mill's view, able to acquire appreciable knowledge of social
phenomena without employing the deductive method, one cannot know
whether Mill's claims about the relations between social phenomena
and human nature are true without taking for granted what he is trying
to show. Second, this passage is stronger on assertion than on argu-
ment. Mill argues only that the psychological laws governing human
beings in society are the psychological laws governing individual human
beings, but nothing follows from this concerning the character of social
laws in general. To reach the conclusion that ``the laws of the
phenomena of society are, and can be, nothing but the laws of the
actions and passions of human beings united together in the social
state,'' Mill needs to show that all laws of social phenomena derive
from the laws of psychology and of the natural sciences. Mill has not
provided such an argument, and by addressing the different question of
the relations between the properties of compounds and constituents, he
has muddied the waters. For many properties of societies (consider, for
example, social mobility or rate of increase of the money supply)
resemble properties of individual human beings as little as the properties
of water resemble those of hydrogen.
Mill does not believe that the additivity of effects always obtains, and
he goes on immediately after the passage just quoted to write,
306 The world of economic causes

This principle [of the Composition of Causes], however, by no means prevails in

all departments of the ®eld of nature. The chemical combination of two
substances produces, as is well known, a third substance with properties different
from those of either of the two substances separately, or of both of them taken
together. Not a trace of the properties of hydrogen or of oxygen is observable in
those of their compound water. (1843, bk. 3, ch. 6, §1)

The composition of causes is here identi®ed with a third principle, which

one might call ``the summation of properties.'' Mill ®nds that the
phenomena of chemistry are not mechanical, because many of the
properties of compounds are not ``sums'' of the properties of their
constituents, even though they might be literally deducible from general-
izations governing their constituents and how they combine. Mill slides
back and forth between questions concerning (1) what can be deduced
about complex subject matters from a knowledge of the ``laws'' govern-
ing individual causal factors and principles of combination (2) whether
effects of causes can be ``added,'' and (3) whether the properties of
compounds are similar to those of their constituents.
Although the composition of causes is not true of all phenomena,
nevertheless
The former case, that of the Composition of Causes, is the general one; the other
is always special and exceptional. There are no objects which do not, as to some
of their phenomena, obey the principle of the Composition of Causes; none that
have not some laws which are rigidly ful®lled in every combination into which the
objects enter. The weight of a body, for instance, is a property which it retains in
all the combinations in which it is placed. (1843, bk. 3, ch. 6, §2)

Mill's grounds for making the general metaphysical claim of the pre-
ponderance of additivity are that all objects ``obey the principle of the
Composition of Causes'' with regard to some of ``their phenomena.''
This is what one would expect if physical properties are mechanical and
all objects have some physical properties. But it does not follow that
``The former case, that of the Composition of Causes, is the general
one.'' And even if this conclusion were justi®ed, it would not in turn
provide a strong reason to regard the messy deductions from simpli®ca-
tions and psychological laws as social laws. There is nothing here to ®ll
the gap for Mill and to justify employing the deductive method.
Someone who follows Mill's prescriptions for economics identi®es a
set of potentially signi®cant causes of the phenomena of interest,
ascertains their separate laws, and, assuming additivity, ``adds up'' their
separate effects to determine their joint effects. If the results ®t the data,
then one concludes that one has causal knowledge of the phenomena of
interest. Even in the best of cases, this method is precarious because of
 Tendencies, laws, and the composition of economic causes 307

the additivity assumption. Deriving generalizations in this way surely

gives one some further reason to believe that they are true and explana-
tory, but the process only provides secure justi®cation for the derived
laws when the deductive method turns out not to be necessary after all.

References
Bhaskar, Roy (1975) A Realist Theory of Science, Hemel Hempstead: Harvester
Press.
Buchanan, James (1975) The Limits of Liberty: Between Anarchy and the
Leviathan, Chicago: University of Chicago Press.
Cartwright, Nancy (1989) Nature's Capacities and their Measurement, Oxford:
Oxford University Press.
De Marchi, Neil (1986) ``Discussion: Mill's Unrevised Philosophy of Economics:
A Comment on Hausman,'' Philosophy of Science, 53: 89±100.
Hausman, Daniel (1981) Capital, Pro®ts, and Prices: An Essay in the Philosophy
of Economics, New York: Columbia University Press.
(1992) The Inexact and Separate Science of Economics, Cambridge: Cambridge
University Press.
(1995) ``The Composition of Economic Causes,'' The Monist, 78: 295±307.
Hutchison, Terence (1998) ``Ultra-Deductivism from Nassau Senior to Lionel
Robbins and Daniel Hausman,'' Journal of Economic Methodology, 5:
43±91.
Lawson, Tony (1996) ``Developments in Economics as Realist Social Theory,''
Review of Social Economy, 14: 405±22.
(1997) Economics and Reality, London: Routledge.
MaÈki, Uskali (1992) ``The Market as an Isolated Causal Process: A Metaphysical
Ground for Realism,'' in Bruce Caldwell and Stephan Boehm (eds.),
Austrian Economics: Tensions and New Developments, Dordrect: Kluwer, pp.
35±59.
(1993) ``Isolation, Idealization and Truth in Economics,'' in Bert Hamminga
and Neil de Marchi (eds.), Poznan Studies in the Philosophy of the Sciences
and the Humanities 38: 147±68.
Mill, J. S. (1836) ``On the De®nition of Political Economy and the Method of
Investigation Proper to It,'' repr. in Collected Works of John Stuart Mill, vol.
4 (Toronto: University of Toronto Press, 1967).
(1843) A System of Logic, London: Longmans, Green & Co., 1949.
Reuten, Geert (1996) ``A Revision of the Neoclassical Economics Methodology ±
Appraising Hausman's Mill-Twist, Robbins-Gist, Popper-Whist,'' Journal of
Economic Method, 3: 39±68.
16 Economics without mechanism

Â
JOHN DUPRE

1 Mechanism

A standard and natural approach to the metaphysics of economics is to

start with a careful scrutiny of economic theory and practice, or some
suitable part thereof, and decide what metaphysical picture is implied by,
or presupposed by, or most consistent with this theory and practice. My
approach in this paper will be rather different. I shall start by presenting
and motivating some broad metaphysical ideas, and then discuss the
question how economics would look in the light of those ideas. The
rationale for this methodology is as follows. Economics has developed in
the context of a broad set of metaphysical assumptions that have existed
in parallel with the last four hundred years of science. Although the
metaphysical picture in question has primarily been developed in relation
to the physical sciences, it has been of fundamental importance in the
development of economics.1 In a research program carried out over a
number of years and culminating in my book, The Disorder of Things
(1993), I have developed a broad critique of this metaphysical picture.
An obvious sequel to this project is an examination of the implications of
the rejection of this picture for various areas of science. Economics is an
especially appropriate candidate for such an examination because its
high status among the social sciences derives, arguably, precisely from
the extent to which its practice reveals a commitment to this classical
metaphysics.2
In The Disorder of Things I referred most generally to this classical
1 Historians of science have recently traced the intimate connections between the rise of
physical science and the development of neoclassical economics in the late nineteenth
century (see e.g. Mirowski 1989).
2 Since writing this paper I have come across Tony Lawson's important book Economics
and Reality (1997), which applies some quite similar metaphysical ideas in some detail to

308
 Economics without mechanism 309

metaphysics as mechanism. Mechanism, of course, is a doctrine that has

something to do with machines ± in fact I have come to believe that the
connections with machines are deeper and more interesting than I had
appreciated when I wrote the book.3 In the book I discuss at length three
different doctrines that I take to be characteristic ingredients of mech-
anism: essentialism, reductionism, and (something like) determinism.
One might very crudely relate these doctrines to machines in the
following way: determinism says that machines are predictable and
reliable: the more reliable the better the machine. The world is a perfect
machine (originally it was taken to have been built by God, after all) and
so is perfectly reliable. This reliable production of more or less complex
behavior is achieved by the way a machine exploits the simpler reliable
behavior of its parts. Thus the way to understand a machine is always in
terms of a decomposition into parts, and perhaps parts of parts, and
parts of parts of parts, etc. This is a form of reductionism or mereological
determinism, the determination of the behavior of the whole by the
behavior of the parts. Finally, the preceding doctrines presuppose that it
is possible to distinguish unequivocally what the parts of a machine are.
Thus, to take a familiar example, the piston is a genuine part of a car in a
sense in which the dashboard, containing parts that relate to a variety of
quite distinct systems, is not. This connects with a central aspect of
scienti®c essentialism, the idea that science aims to discover an objec-
tively real classi®cation of things.
The debate over micro-reductionism in science concerns the extent to
which science is committed to explaining the behavior of objects in terms
of the behavior of their parts. An extreme and classical kind of
reductionism holds that all laws governing the behavior of complex
objects should be deducible from the laws of a lower-level science and
thus, ultimately, the laws of all sciences should be deducible from those
of particle physics.4 Hardly anyone holds this kind of strict reductionism
any more, and at least among philosophers of biology and of the social
sciences, reductionism even in more moderate versions is distinctly
unfashionable. While it is clear that the investigation of structural factors
plays a central role in the development of science, it is generally conceded
that the laws or causal knowledge to be discerned in these sciences is in
some quite strong sense autonomous from any accessible facts about
underlying structure. One might imagine that the decline of reductionism

economics. I discuss this book, and some minor disagreements with my own position, in
Dupre (1999).
3 The connections with Nancy Cartwright's work generally, and her contribution to this
volume speci®cally, are obvious and not coincidental.
4 A classic statement is Oppenheim and Putnam (1958).
310 The world of economic causes

would signal the decline of mechanism. But surprisingly, nothing could

be further from the truth.
A fundamental issue here is the relation between reductionism and
what I called ``(something like) determinism.'' I avoid referring to
determinism tout court for the obvious reason that much of science, and
most signi®cantly fundamental physics, is not now thought to be
deterministic. Contemporary positions remain close to determinism,
however, in one of two ways. First, it is often said that the indeterminism
disappears at a very low level of material aggregation, so that indeter-
minism occurs only at a level far beneath our concern. It does seem to me
obvious that it will be impossible reliably to shield macrodeterminism
from the indeterminism at the microlevel, but I won't pursue this. A
more important point is that an indeterministic microlevel can be just as
comprehensive as a deterministic one. Even if it couldn't specify the
behavior of each particle, it might specify the probability distribution of
all possible behaviors, and do so in a way fully determined by facts at the
microlevel. This is the doctrine of the causal completeness of the
microlevel, of which microlevel determinism is a special limiting case,
and it is a doctrine that is still widely believed.
The causal completeness of the microlevel is, of course, a necessary
condition for reductionism. Reductionism hypothesizes that in the end
everything will be explained by microphysics; so microphysics cannot
leave anything out. It is, however, far from suf®cient, as revealed by the
greater prevalence of subscription to the former than to the latter. An
excellent source for seeing how these positions are separated is in the
recent work of Alexander Rosenberg, notably Economics: Mathematical
Politics or Science of Diminishing Returns (1992), and Instrumental
Biology or the Disunity of Science (1994). Rosenberg harbors no doubts
as to the completeness of the microphysical, but he is also committed to
the view that biology and the social sciences are irreducible to physics.
According to Rosenberg, the trouble with these sciences is that their
categories are physically heterogeneous. In biology this is due to natural
selection. Nature selects for function not structure, and many different
structures may serve the same function. In economics the problem is
compounded. For not only are humans biological objects, but their
properties of relevance to economics are intentional properties; and
Rosenberg argues that many different physiological properties might
ground the same intentional property. Thus these properties are twice
removed from connectibility with the physical. But since the causal
properties of an object are, nonetheless, determined by its physical
structure, objects that share the same biological or economic properties
will be causally heterogeneous. These sciences are, therefore, imperfect.
 Economics without mechanism 311

They are defensible as (sometimes) instrumentally useful for ®nite beings

such as us, but more computationally and cognitively talented creatures
would look directly at the physical structure of things and hence get
straight to the true causal powers. It is doubtful, in fact, whether
Rosenberg even sees much instrumental utility in economics.
All this is part of a familiar picture. It is the picture of mental
epiphenomenalism, of the mental realm as a causally inert patina on the
underlying physical ± perhaps neurological ± reality, and of voluntary
action as mere illusion. And indeed, it is a view that makes everything
macroscopic epiphenomenal, causally inert. But, ®nally, it is a view
which philosophers driven on the one hand by an increasing realization
of the hopelessness of reductionism, and on the other by admiration for
physics, are frequently found to embrace.
In criticizing this view, I want to begin by attempting to allocate the
onus of violating common sense. On my view things at many different
levels of organization and complexity ± animals, planets, plants, elec-
trons, corporations, etc. ± have autonomous causal powers.5 ``Autono-
mous'' here means not reducible to a consequence of the causal powers
of the physical particles of which they are composed. An immediate
consequence of this view is that there can be no level of organization,
including that of microphysics, at which the behavior of objects is
completely determined by some set of laws describing only objects at that
level. Levels of organization are too intimately interconnected for there
to be any possible insulation of one level from the in¯uence of other
levels. But many philosophers, Rosenberg for example, ®nd it extra-
ordinary and implausible to deny the completeness of microphysical law
for describing microphysical phenomena. My question, then, is whether
we should ®nd it more plausible to deny the completeness of physics, or
to deny the ultimate causal ef®cacy of everything above the microphy-
sical level. Of course I prefer the former course.
Let me elaborate a little what I see as the strangeness of Rosenberg's
position. He believes that everything that happens happens in accordance
with, and as determined by, the universal laws that govern the behavior
of the physical particles of which everything is composed. The buzzing,
blooming confusion of the phenomenal world, on the other hand, is a
mere epiphenomenal froth on this underlying microphysical reality. The
macrosocopic, qua macroscopic, is, for Rosenberg, causally inert. I

5 This set of views, which I there describe as ``promiscuous realism,'' is defended in detail in
The Disorder of Things. The thesis encompasses both the autonomy of objects at different
levels of structural complexity, and the necessity of diverse and independent classi®cations
of objects at the same level (about which, see further below).
312 The world of economic causes

might naãÈvely suppose that it was the sight of the bear and my belief that
bears are dangerous that causes me (perhaps foolishly) to scamper up a
tree. But on Rosenberg's view it is misleading to refer here to the sight of
the bear, which suggests that it is by virtue of being a bear that I react to
the object as I do. On Rosenberg's picture it must be by virtue of being a
certain aggregation of physical particles that the object has the causal
consequences it does. Perhaps most of the aggregations of physical
particles that would succeed in constituting a bear would have the
relevant causal consequence, and perhaps there is a selective explanation
for this happy coincidence ± an explanation that might even explain why
an ontology of bears is one that serves my reproductive interests well
enough. But for all this, the ontology, or perhaps mythology, of bears,
lions, snakes, and so on is one forced on us only by our cognitive
limitations. If we were smart enough we would identify things as just the
precise aggregation of physical particles that they were and predict their
behavior from the laws of physics; we would certainly not hamper our
thinking with the physically heterogeneous categories of biology. It is
hardly necessary to spell out how central economic concepts ± price,
®rm, market, etc. ± would fare on this picture, and Rosenberg is as
skeptical about economics as his metaphysics would suggest. As I have
noted, he sees economics as dependent on an epiphenomenon (the
intentional) of an epiphenomenon (the biological) (see Rosenberg 1992).
Here it is implausible to expect even the rough correlations that we might
hope for in biology between macroscopic kinds and microstructural
make-up. Like Bishop Berkeley, I am unwilling to sacri®ce the familiar
world to a scientistic philosophy, though unlike Berkeley I am greedy
enough to want to endorse the unobservable realities of physics as well.
It is useful to recall that however implausible Berkeley's immaterialism
may seem, he did brilliantly demonstrate that Locke's version of materi-
alism wasn't exactly plain old common sense.
Rosenberg sometimes associates his position with empiricism (e.g.
1994, p. 10). This strikes me as remarkable. He sometimes writes as if
empiricism were just whatever method best accounts for the success of
physical science as, for example, when he attributes to empiricism a
reliance on theoretical explanation by uni®cation (1994, p. 11). But
empiricism, which for all its admitted dif®culties and necessary quali®ca-
tions, I want to embrace as a cardinal empirical virtue, has something to
do with grounding in experience. And surely the privileging of the
unobservable over the observable is the antithesis of empiricism. I do not
want to endorse the extreme empiricism of those, like Bas van Fraassen
(1981), who deny that we have any reason to believe in the ontology of
unobservables. But I do think the spurious adoption of the mantle of
 Economics without mechanism 313

empiricism obscures the counter-intuitive nature of the kind of physic-

alism Rosenberg supports. And I also think that a proper recognition of
the distance such a position takes us from a grounding in experience is an
important part of the way in which such a physicalism violates common
sense.
Rosenberg, then, defends a ``materialist and mereological determinist
approach to biological and, a fortiori, social systems: they are . . .
`nothing but' physical ones'' (1994, p. 55). The determinist aspect of this
position is obviously subject to question in view of the indeterministic
aspects of quantum mechanics, but as I have said I don't think this is the
crux of the matter and I shall pass over it here.6 He concedes that the
mereological determinism, the determination of the behavior of wholes
by the behavior of their parts, is not an empirically supported position
since we cannot in fact ``systematically derive the biological from the
physical'' (p. 55). I suspect, however, that he thinks that mereological
determinism is a condition on any materialism worthy of the name. I
disagree. I consider myself a materialist but, being also an empiricist, I
see no reason to believe in mereological determinism.
Before going any further, I do want to applaud one aspect of
Rosenberg's position. Rosenberg thinks that microphysics describes a
domain of objects that is exhaustive, in the sense at least that everything
is physically composed of them and of nothing else, and that conform to
universal and exceptionless laws. Rosenberg correctly concludes that the
behavior of the larger things composed of these microphysical entities
must be fully determined by the latter, and thus that they are in a sense
mere epiphenomena of the microphysical. Acknowledging the failure of
the reductionism that would substantiate this view Rosenberg, again
correctly in my view, resorts to the supervenience and mereological
determinism which he treats, again rightly in my view, as reductionism
though only for much more computationally talented beings than we.
These are ambitious claims and, as he and I agree, claims with no direct
empirical support. But they are indeed consequences of his views of the
microphysical and Rosenberg should be applauded for forthrightly
embracing them. I emphasize this point because it often seems that
philosophers want to have their cake ± totalizing physics ± and eat it too
± higher level autonomy. I think here of a certain kind of fuzzy
compatibilism about the free will question and certain appeals to both
supervenience and autonomy in the philosophy of mind.

6 Rosenberg takes the line that nature asymptotically approaches determinism at higher
levels of aggregation (1994, p. 61). For some reason he also (erroneously) attributes this
view to me.
314 The world of economic causes

Perhaps I also want to have my cake and eat it, because despite
believing in higher-level autonomy I do consider myself a materialist,
perhaps even a physicalist. I agree with Rosenberg that the microphysical
realm is exhaustive: the existence of immaterial minds, souls, ectoplasm,
deities, etc. strikes me as, to say the least, unlikely. Hence I believe that if
somehow all the physical particles in the universe were annihilated there
would be nothing left ± perhaps not even empty space. Where I differ is
in the commitment to universal physical laws. Thus the sense in which I
agree that the biological or even economic realms are ``nothing but'' the
physical is an extremely weak one having to do only with substantial
composition. I endorse a very weak physicalism because I endorse, in a
sense, only a very weak physics.
Why do so many philosophers, and presumably physicists, believe that
the laws of physics are exceptionless and universal? Of course physical
laws are generally expressed as universally quanti®ed, but surely we
should not be prisoners of our notation? As Nancy Cartwright has been
telling us for a number of years, we know that the laws of physics are not
universally true, at least as in any form in which we know how to express
them (Cartwright 1983). They are true only under quite stringent condi-
tions of the absence of any of an inde®nite range of interfering factors. If,
improbably, we were able to specify every such possibly interfering
factor, the most we would have achieved would be a universal law with
extremely restricted application. No doubt it is widely assumed that the
laws of physics can always be modi®ed in ways that specify explicitly the
consequences of interaction with any possible interfering factor. But then
even at the physical level we have gone far beyond what is empirically
supported and are expressing mere faith in hypothetical laws that we
neither possess nor have any reason to anticipate possessing.
I am aware that this line of argument is liable to produce a certain
amount of exasperation. Surely, it will be said, I must admit the
extraordinary successes of the physical sciences, and surely such successes
would be inexplicable if the world were not in broad outline as physical
scientists have supposed it to be: subject to underlying laws of complete
universality and generality. It is common to emphasize here the remark-
able improvements in predictive accuracy achieved by physicists over the
last four centuries (see Rosenberg 1994, p. 36). But of course the fact that
physicists are doing something very well doesn't imply that they have the
most sophisticated grasp of what they are doing so well. (If it did there
would surely be no use for philosophers of science ± a consequence for
which, I must admit, some enthusiasts for the physical sciences would be
willing to bite the bullet.)
Arguments of the sort just mentioned begin to look rather different in
 Economics without mechanism 315

the light of much recent work in the history and philosophy of physics.
Nature does not pronounce positively or negatively on our theories as
soon as we care to ask her. Without denying that experiments can often
impressively con®rm or refute theories, it has become clear that getting
experiments to produce illuminating results is extremely dif®cult. Re-
turning once again to the literal sense of mechanism, contemporary
large-scale physics experiments most obviously, but also older simpler
ones, are complex bits of machinery. Like other machines, it generally
takes a great deal of work and expertise to set them up and make them
work in the ways they are supposed to.7
This leads us to what is perhaps the most widespread basis for
mereological determinism, if perhaps a less sophisticated one from a
philosophical point of view than one grounded more directly in physical
law. A bluff commonsensical kind of realist is liable to insist that it
would be inexplicable or even miraculous that our machines could work
with as much reliability as they do if the theories about the world held by
the designers of those machines were not somewhere close to the truth.
Would you really get on an airplane if you didn't believe that physics
dealt in universal laws? The connection between mechanism and the
unity of science, I believe, is intimately related to the unity of purpose
and coherence of parts characteristic of a good ± indeed a perfect ±
machine. So it seems to me appropriate that the crux of this debate
should turn on a correct appreciation of the nature and signi®cance of
machines.

2 Machines and models

Most machines are not, of course, intended to demonstrate scienti®c

theories. Indeed one of the more interesting suggestions that has come
out of various parts of the history of science and technology is that the
relation between these two is anything but unidirectional. It may be that
much scienti®c theory ± thermodynamics and the steam engine provide
the locus classicus ± has developed to explain the successes of engineers
rather than the successes of engineers being made possible by ante-
cedently established scienti®c theories. But in view of the point just
mentioned about the machine-like nature of most experiments in physics,
we can see that this point is not crucial. If physics experiments are a kind

7 This point has been documented in detail by a variety of students of physics in history
(Galison 1987), sociology (Pickering 1984), and even anthropology (Traweek 1988).
316 The world of economic causes

of machine then the laws of physics are constructed to explain the

workings of machines either way. Of course there is more to it. Machines
that do what we expect lead us to theories that suggest new kinds of
machines, and so on. I am not denying the role of the theories, just
insisting on the closure of the circle which relates physical theories to the
machines that con®rm and implement them, and consequently on the
dif®culty of arguing from this kind of support for physical laws to their
universal range of application.
What I want to say, then, is that machines are very special parts of the
world. For machines mereological determinism is almost true (almost,
since no machines are perfect) but machines are a poor model for the
world in general. Rhetorical support for this position might be gained by
reconsidering William Paley's famous thought experiment of the watch
found on an apparently deserted island. Paley, of course, wants us to
think of the watch as similar in many respects to naturally occurring
organisms. But the force of the thought experiment comes, paradoxically,
from the anomalous character of the watch: it is quite unlike anything we
might expect to ®nd lying on the beach on a desert island. Unlike Paley,
we are not led to think of a similarity because committed to the view that
organisms are artifacts. According to contemporary science, watches are
artifacts, plants and animals aren't. The dissimilarity between the two,
apart from con®rming what many believe, that the world was not created
by an intelligent being, should also lead us to question the appropriate-
ness of machines as a model for naturally occurring objects. Indeed, if
the world were really a huge piece of machinery, one might wonder why
useful, if more modest, machines are so hard to make.
It remains incumbent upon me to say something about what is special
about machines. The ®rst point I want to suggest is that it helps to see
the behavior of machines not so much as determined but as heavily
constrained. Suf®cient constraint, of course, can amount to determina-
tion. Machines work when they have the capacity to do what they are
intended to do, and when they are constrained from doing anything else.
The simplest example might be a machine such as a lever. The rigidity of
the lever and the solidity of the pivot leave the end no option but to rise
when the other end is lowered. More interesting would be a relatively
complex machine such as a car. A series of effectively rigid connections
between the piston and the drive wheels give a car the capacity to move
forward when gasoline is ignited in the cylinders. The dif®culty in
making a reliable car has mostly to do with blocking all other capacities
± for the cylinder to burst or melt, for the piston to seize or ¯y out of the
end of the cylinder, for various connections to fail, etc. The car runs
because it has been rendered incapable of doing anything else.
 Economics without mechanism 317

Of course I do not want to deny that there must be some natural

regularities to make all this possible. Many of these concern rather
simple capacities of materials. Steel rods seldom break except under
extreme forces; mixtures of gasoline and air almost always explode when
ignited. Note that these can typically be seen as the deterministic upshots
of very large numbers of indeterministic capacities exercised at the
microlevel. No deterministic tendencies of hydrocarbon molecules to
oxidize are required to guarantee the explosion of a cylinder full of a
suitable air and gasoline mixture, nor is any deterministic tendency of
any molecule to escape the gas tank required to get about the right
amount into the cylinder. This provides a speci®c explanation of reliable
regularities that offers no promise of generalization to all causal
sequences. I take another similar example (though again Rosenberg
(1994) has a quite different account) to be those cases of natural selection
that approximate to deterministic processes.
Economists, on the whole, do not build machines but models. (I shall
discuss the extent to which they may sometimes build something like
machines at the end of the paper.) There is, however, an interesting
parallel between model-building and machine-building. As I have noted,
much of the work of building a good machine involves shielding it from
the possible in¯uence of interfering or disrupting forces. The problem
with model-building, whether in economics, population genetics, so-
ciology, or many other areas of science, is also provided by the possibility
of interfering in¯uences not considered in the model. Models, on the
other hand, have a huge advantage over machines: whereas machines
must ®nd a way of actually blocking interfering forces, models simply
abstract from such forces or, in other words, ignore them. Here are all
the advantages of theft over honest toil.
Obviously enough, however, this advantage of models is also their
great weakness. Whether they are good models depends on how impor-
tant the factors the models ignore are in determining the behavior of the
system being modeled. Where models fail to correspond closely with the
empirical reality they aim to model ± as notoriously occurs with many
economic models ± this may be because the elements included in the
model do not have the properties attributed to them, or because other
factors overwhelm the effect of those included. Popperians may note
with distress that this means a model can never be refuted: we may
always suppose that we have abstracted away from important factors.
But this is not my main point here. Let us assume that the model is
correct though perhaps incomplete; its failings, that is to say, are all of
the second kind. My point, then, is that inside every model is a machine
trying to get out. As long as we stick to modeling and passive observa-
318 The world of economic causes

tion, we can only hope to improve the model by incorporating ever more
factors into it. On the other hand there is also the quite different
possibility of trying to build a machine in which only the factors included
in the model are allowed to operate. Others are excluded in the ways
characteristic of machine-building. Thus a correct model, in the sense
that the causal powers of its elements are correctly described, may be
thought of as a possible machine. And a model may be correct in this
sense even if it is empirically quite inaccurate. This sense in which
economics is a mechanistic science, and the extent to which economists
may actually aspire to be machine-builders, will be taken up in more
detail later in this paper.

3 Parts and wholes

In this section I shall try to discharge one further onus clearly incumbent
upon my position, an explanation of what I take to be the actual
signi®cance of structural explanation, explanation of the behavior of the
whole in terms of properties of the parts. Here I want to reiterate a thesis
that I suggested at several points in The Disorder of Things. Structural
explanation, I suggest, is required to explain how a complex object has
the capacities it has, but does not generally tell us which capacities will be
exercised when. Thus physiology and biochemistry have provided won-
derful insights into the abilities of organisms to metabolize, move,
reproduce, and so on. But more is needed to determine when these
capacities will be displayed. Thus even for a capacity as ``mechanistic'' as
metabolism of food no amount of biochemistry will tell us when food to
be metabolized will be introduced into the system. Here we might note an
important characteristic of machines: machines typically have both
relevant capacities and controls.8
It is true that for relatively simple organisms a behavior may be more
or less deterministically elicited by features of the environment. But even
in this case we cannot in principle predict the production of behavior,
nor even its probability, merely from examination of the structure of the
organism. The behavior of the individual can, at best, only be predicted
from an extensive knowledge of the organism and its environment.
Typically this will give us no more than an estimate of the probability

8 I have discussed elsewhere some differences between organisms, especially humans, and
machines in the context of the free will debate (Dupre 1996). One important point is that
humans do not have controls.
 Economics without mechanism 319

that the organism will encounter the triggering environment. In terms of

real possibilities for reductive science, it seems to me that here we have a
quite suf®cient obstacle to any practical reduction quite apart from
Rosenberg's concerns about selection for function. In my book (1993) I
elaborate such an argument in terms of the necessity of appealing to
abstractions in real mereological explanations.
All explanation is partial: not everything can be explained at once. For
the purposes of explanation one takes certain matters for granted, and
others as in need of illumination. This elementary observation points to
an important quali®cation of my concession that capacities are amenable
to structural explanation. No doubt there are some very complex
physiological facts concerning my brain and my ®ngers that explain my
ability to write this paper. On the other hand, for the electronic traces I
am producing to be this paper, a comparable diversity of social facts
must also obtain. Starting with the social facts that give the traces, when
electronically processed into words, public meanings we could move with
increasing speci®city towards the social practices that make possible the
philosophy of science. Thus my capacity to write this paper is contingent
on many social facts in addition, no doubt, to structural facts about me. I
take the general dependence of language on social convention to show
the same to be true for a vast range of mental facts. This joint
determination of capacities by internal structure and external context
makes the possession of autonomous capacities by complex objects quite
unproblematic and innocent of the metaphysical mystery-mongering
with which it seems sometimes to be associated. Moreover, such autono-
mous capacities make so-called ``downward causation'' the determina-
tion of parts by wholes, unproblematic.9 It happens all the time: when,
for instance, I decide to raise my arm I cause the motion of billions of
physical particles. Although the case of human behavior seems to me the
most compelling, I take it that the same account could be applied to
much humbler systems. Indeed, in view of the indeterminism of the
microphysical and the determinism of macroscopic mechanical phe-
nomena, it seems natural to say that macroscopic mechanical movements
of an object cause the movements of its microphysical parts. Needless to
add, if I am right about this, there can be no possibility of the causal
closure of physics.
Most philosophers, as I have said, are aware of the impossibility of

9 An earlier advocate of the importance of downward causation was Campbell (1974). Kim
(1993, essay 17) is another philosopher who has recognized the commitment of non-
reductive physicalism to downward causation, but who thinks the idea is highly
problematic, perhaps incoherent.
320 The world of economic causes

actual reductive explanation in many cases, and their physicalist intui-

tions are often defended by appeal to so-called supervenience (see again
Rosenberg 1992, 1994). Indeed supervenientism has reached near epi-
demic proportions in recent years. Supervenience is a kind of reduc-
tionism, but not for mere mortals. To say that one domain supervenes on
another is to say that the former is wholly determined by the latter such
that if we had suf®ciently powerful ± perhaps in®nite ± minds, we would
see everything there was to see by looking at the supervened-upon, that is
microstructural, level. As supervenience buffs like to put it, once the
physical is ®xed, so is everything else. As indicated by the preceding
discussion of context dependence, however, even in relatively simple
cases, let alone cases from the human sciences, the relevant part of the
supervened upon domain may be extensive. Indeed it is arguable that
much of the universe would be required to provide the full basis on
which many macroscopic properties supervene. Whether or not this is
objectionable from a strictly ontological perspective, it is surely more
than enough to substantiate my claim that the whole reductionist story is
without empirical credentials. As we continue to grapple with the three
body problem, a unity of science accessible only to a being capable of
grasping and analyzing the microphysical state of the entire universe is
not something within reach of any imaginable empirical evidence.

4 Economics

So much for the mechanical universe, the universe in which everything

that happens is purely the consequence of the playing out of the laws of
microphysics harnessed by the ingenious structuring of countless micro-
physical components. I shall now make some suggestions as to how the
rejection of this metaphysical picture might bear on our understanding of
the science of economics. I begin by noting that the situation described in
the ®rst part of the paper presents a very troubling dilemma for the
metaphysical foundations of economics. On the one hand, it should be
clear that traditional monistic physicalism problematizes the very possi-
bility of economic phenomena. No doubt it used to be possible to
suppose that economics might be more or less smoothly reduced to
individual psychology, and thence through neurophysiology to chemistry
and physics. But few would stake much on such a future today. The
kinds of entities that ®gure fundamentally in economic theory ± ®rms,
consumers, physical capital, etc. ± are, from a purely physical point of
view, massively heterogeneous. Worse still, perhaps, many ± interest
 Economics without mechanism 321

rates, money supplies, etc. ± do not look physical at all. Thus a plausible
post-reductionist physicalist perspective on economics will be one like
that of Rosenberg (1992), who sees economics as an epiphenomenon on
an epiphenomenon and as, therefore, of even highly questionable instru-
mental utility. If the ultimate viability of an area of science depends on
its connectability with fundamental physics, then surely economics is in
deep trouble. I have no such expectations for science and thus my
pluralistic ontology, complete with downward causation by higher-level
phenomena, is in principle much friendlier to the possible reality of a
domain of economic phenomena. However, I fear that not all economists
will welcome my philosophical support, for several reasons.
The ®rst such reason is the threat to objectivity, in the sense of value
neutrality. It is undoubtedly a part of the self-image of economists that
much of what they do is pure ``positive'' science; and notoriously
``normative'' or ``welfare'' economics has been perceived as something of
a disciplinary backwater ± not the place to seek a Nobel prize, anyhow.
Economists consider themselves to be investigating an independent
reality of economic phenomena, and whether their mathematical models
correctly describe those phenomena depends solely on how things are out
there. The metaphysical picture, of course, that would best motivate that
practice would be one that applied the mechanistic metaphysics of
physical science directly to economics. But it has become clear that that
is impossible. Although my metaphysical position restores the possibility
of genuinely causally ef®cacious economic entities,10 entities that may
even be linked together in quasi-mechanical ways, it does so only by
severing the connection with physical science. And by doing so, I shall
argue, it in turn threatens the picture of economics as an independent
causal order open to fully disinterested enquiry.
The problem is that once one rejects the assumption that economic
entities are physically homogeneous parts ± even very complex parts ± of
the fabric of the cosmic machine, one is forced to look in more concrete
detail at the way economic kinds are constructed. We need to examine,
that is to say, the economic principles according to which economic
entities are classi®ed into economic kinds. And once we do this it is
impossible to escape the conclusion that value judgments are deeply
embedded in that constructive process. I suppose that some economists
still think that the amount of work done last week in the US is as

10 For more speci®c examples, MaÈki (1990) argues for treating money and entrepreneurship
as causally ef®cacious economic entities. General defences of the relevant kinds of causal
capacities against traditional Humean objections can be found in Dupre and Cartwright
(1988) and, in most detail, in Cartwright (1990).
322 The world of economic causes

objective a matter of fact as the amount of work done in lifting a

container onto a truck ± just a lot harder to measure. And as I have
noted, as a historical matter it appears that the use of the same term is far
from coincidental. But it is easy to show how untenable this view is.
Traditionally work has been de®ned as activity for which some remu-
neration is received. But though it is easy to see the attraction of this
de®nition from a particular interest in publicly visible processes of
production, from a broader perspective the de®nition is entirely arbi-
trary. Why should my labor in raising potatoes in my back yard not
count as work if the same labor provided in exchange for wages for a
commercial market gardener does count as work? More generally, why
should almost all domestic work ± cleaning, cooking, caring for children
± not count as fully as work as, say, building nuclear warheads or writing
jingles for beer commercials. Some inadequacies in the simple traditional
de®nition have been recognized in recent years, and various alterations
have been made. My point here is not to criticize a particular de®nition
nor to advocate any particular alternative. It is rather to insist that there
is no unequivocal answer to such questions. Rational progress with such
an issue can only be made by asking, ®rst of all, what the point is of
making such a de®nition in the ®rst place, and then considering what
aspects of the de®nition will best serve the relevant ends. It is not hard to
discover such goals in the case of work, though unfortunately some of
these normative views are hard to reconcile. Sometimes work is thought
of as a thing to be minimized (Adam Smith's ``toil and trouble,''
something to be contrasted with ease, leisure, relaxation). But it is also
thought of as something fundamental to a person's sense of identity, and
something that should be a major feature of any normal adult life.
Unemployment ± lack of work ± is more unequivocally considered a bad
thing, to be minimized subject, perhaps, to con¯icting economic desi-
derata. Such broad normative considerations will unavoidably affect the
desirability of de®nitions of work for economic purposes.11
Many, perhaps most, central economic concepts have such familiar
evaluative content. GDP is generally assumed to be good, and its
increase, economic growth, still better. In¯ation, for reasons that are
somewhat obscure,12 is generally assumed to be very bad. And so on.
Since economic policies are directed, admittedly with questionable
11 The interrelations and con¯icts between various conceptions of work, and the historical
evolution of such concepts are discussed in greater detail in Gagnier and Dupre (1995)
and Dupre and Gagnier (1996).
12 Given the amount of personal debt, especially mortgage debt, in a country such as the
US, it is clear that in¯ation would be a boon to a very large segment of the population.
This point is curiously absent for most public discussion of in¯ation. Cynics might
 Economics without mechanism 323

success, at promoting some of these things and avoiding others, the

decisions that go into de®ning such concepts will affect what activities
public policies aim to promote. On current criteria, for instance, a major
oil spill will qualify, by virtue of the economic activities involved in the
attempts to clean it up, as an economic windfall. While I do not suggest
that governments encourage oil spills, the typical failure to include
degradation of natural resources as an item in GDP accounting is only
one of several respects in which such procedures are open to debates with
profound quantitative implications.13 I am not suggesting that there is
some objectively correct de®nition of GDP that is thereby violated, only,
again, that whether a particular item is included will have large effects on
how particular activities or events show up in economic outcomes, and
thus, very probably, will affect which actions are carried out. Thus how
economics is practiced, even at the stage of concept construction, can
have important effects on the world. Since I have no reason to deny the
obvious here, I take this to be an illustration of the causal autonomy of
economic agents, though I do not claim that an epiphenomenalist could
not tell an adequate story to take account of what I have said so far.
Unwelcome though all this may be to economists there is now direct
empirical evidence of the way in which the very business of doing
economics can affect the nature of economic phenomena. Some intri-
guing recent research has suggested that the study of economics actually
makes students conform more closely to the axioms of economic man
(Frank, Gilovich, and Regan 1993). At the end of an introductory
economics class students were found more likely to default in prisoner's
dilemma type games, and generally appeared more inclined to act
according to canons of economic rationality. Economics professors were
found considerably less likely to donate to charity or subscribe to public
television than professors in other disciplines (though here it is hard to
sort out cause and effect). Given the general currency in our present
culture of economic ideas emphasizing and often applauding the pursuit
of individual self-interest it is hardly improbable that such ideas could
have a considerable in¯uence on behavior far beyond the narrow con®nes
of professional economics; and if so one cannot adequately evaluate the
empirical successes, such as they are, of contemporary economic theory
without considering the possibility that the behavior that generates the
predicted outcomes is, to some degree, generated by the very theory used

wonder whether the of®cial horror of in¯ation evident in much government policy
re¯ects a greater concern with the interests of bankers than of the general population.
13 For instance Waring (1988) provides a compelling account of the role of assumptions
about gender in the de®nitions underlying national income accounting.
324 The world of economic causes

to predict it. Thus it is arguably impossible to separate purely empirical

evaluations of the success of an economic research program from
normative concerns about the desirability of the model of human nature
that the program presupposes. I do not take this to be a problem peculiar
to economics, though I do think the behavioral consequences of neoclas-
sical economic ideology, to the extent that they occur, are, to say the
least, unfortunate. Indeed it seems to me quite plausible that any theory
of human behavior with wide currency and application will have some
effect on its own subject matter. Adding the present point to my previous
remarks about the unavoidable normative elements in the construction
of economic concepts and the subsequent use of these concepts in
guiding public policy, it seems to me clear that there is no sensible
prospect of a value-free economics. The fact/value distinction, in short,
should go the way of the Berlin wall.
These observations about the value-ladenness of economics relate to
mechanism and the unity of science only in so far as reductive mechanism
might have offered the most plausible line of resistance to them. I now
come to a second respect in which I expect my ontological reassurances
to meet with less than unquali®ed enthusiasm from economists. Although
I do not believe that the preceding remarks show that economic concepts
fail to refer to aspects of reality, I do believe that they show that reality
does not determine a uniquely correct set of economic concepts. Norma-
tive considerations are able to gain a grip on economic concepts because
the causal order alone does not determine how these concepts should be
de®ned. I continue to insist that these concepts may re¯ect aspects of
reality, because relative to a set of normative or other desiderata there
might perfectly well be better or worse ways, even a best way, of de®ning
a concept, such being determined by the real causal capacities of real
economic agents. In The Disorder of Things I argued for such a pluralistic
view of biology, speci®cally that there was no unique taxonomy of
biological organisms determined by nature.14 Here I am suggesting that
the same must be true for economics, though as I hope to have made
clear, this underdetermination may be of even greater practical con-
sequence in economics. For the case of biology I argued that different
taxonomic schemes would be most appropriate for different investigative
purposes. For the case of economics not only will different investigative
purposes call for different conceptualizations of the phenomena, but in

14 Critics of this earlier view have often been unwilling to accept the compatibility of
realism with the failure of uniqueness. No doubt this is related to the curiously prevalent
view that the belief that scienti®c claims are often true is somehow incompatible with the
truism that science is socially constructed.
 Economics without mechanism 325

addition the systematic adoption of a particular way of doing economics

can be expected to affect, perhaps profoundly over time, the nature of
the phenomena being investigated. This point makes clear the necessity
of recognizing the normative dimension in the choice of a set of economic
concepts. The only possible foundations for economics, I suggest, are in
political philosophy.
Having worked my way down to foundations, let me return to some-
thing more super®cial, ontology and the limitations of mechanism. The
standard methodology of economics is, broadly speaking, mechanistic.
Economic models analyze a phenomenon into a number of basic
economic constituents and attempt to show how the behavior of these
constituents produces a certain global behavior of the whole. One should
remember, of course, that these are models, and there is some debate
over how accurately models should be required to correspond to the
reality they represent. On the other hand, scientists are generally realists;
and, Milton Friedman15 notwithstanding, economists had better not be
instrumentalists given the rather limited empirical success that their
science has enjoyed. So I shall assume that economic models, while
certainly not required to mirror a complete reality (whatever that is), are
surely successful only to the extent that they re¯ect with some ®delity
some part of social reality.16 Such a view is expressed in the best-known
introduction to economics, by Samuelson (1983). ``Most economic
treatises,'' Samuelson writes, ``are concerned with either the description
of some part of the world of reality or with the elaboration of particular
elements abstracted from reality.'' In the next paragraph he writes, ``In
every problem of economic theory certain variables (quantities, prices,
etc.) are designated as unknowns, in whose determination we are
interested. Their values emerge as a solution of a speci®ed set of relation-
ships imposed upon the unknowns by assumption or hypothesis. These
functional relationships hold as of a given environment and milieu.''
These functional relationships, given the ®rst quotation, describe the
relationships between parts of an underlying mechanism that is assumed
to generate the variables of interest. So it appears that economic practice
typically assumes that the world it models is somewhat machine-like.
Samuelson's remarks ®t well, I take it, with my suggestion above that
``inside every model is a machine trying to get out.''
Here two questions immediately arise. A familiar question throughout

15 Friedman (1953) provides a much discussed instrumentalist account of economics. I have

added to the voluminous critical literature elsewhere (Dupre 1994).
16 For a much more detailed analysis of the extent to which economics can be treated as a
realistic science, see MaÈki (1989).
326 The world of economic causes

the history of economics is the following: given the necessity for

abstraction (admitted by Samuelson and everyone else, as far as I know)
in laying bare economic machinery, how useful will the identi®cation of
such machinery be? Will the output of the machinery be recognizable
against the noise of innumerable interfering phenomena? But I am
interested here in a more fundamental ontological issue: Do we have any
reason to believe in such economic machines at all? Here the dilemma
mentioned above, that of ®nding a place for economics either in the
monistic world of traditional mechanism or in my alternative pluralistic
world, surfaces even more sharply. The economist supposes that the
world contains economic machines. But the traditional mechanist be-
lieves only in physics machines. And with the failure of reductionism it
seems impossible that a world full of physics machines can have any
room for economic machines. My world, on the other hand, has room
for all kinds of machines, even though, constantly getting in each other's
way, they cannot be expected to display the absolute reliability of the
mechanists' One Big Machine. I say that my world has room for all these
different kinds of machines, but ± and this is the other horn of the
dilemma ± that doesn't mean that they are really there. In a largely
disordered universe there is no a priori guarantee that economics will be
possible at all, at least in anything like the form currently expected of a
mathematical science. How likely is it that there really are such
machines?
Here two different possibilities must be distinguished. First, there is
the question whether there might be naturally occurring economic
machines out there in the world waiting to be discovered, and second
there is the question whether people ± governments, corporations,
experimental economists, or whoever ± might be able to build such
machines. The orthodox faith among economists in the ef®cacy of
markets might be seen as the somewhat paradoxical idea that the only
good economics machines are the naturally occurring ones. The ones we
try to build are seen, by contrast, as massively and chronically unreliable.
Indeed one not uncommon view is that only if we resolutely refuse to
tamper with naturally occurring economic machinery will it work at all. I
suppose a conceivable rationale for such a view might be the analogy to
biological evolution. Certainly there is little prospect of our building
biological organisms better than nature, even if we can tinker with nature
a bit these days. On the other hand the analogy is hardly compelling. The
evidence for the exquisite adaptation of economic entities to their
environment comparable to that of organisms is, to say the least,
unimpressive. It is important to note, returning once again to the
dilemma I have just been emphasizing, that whether one adopts a
 Economics without mechanism 327

monistic physicalism like that of Rosenberg, or the kind of pluralism that

I prefer, such a position is as lacking in a priori philosophical rationale as
it is in empiricial support. The assumption that economic phenomena are
organized throughout in a machine-like structure is plausible only in a
world that is both mechanistic at the basic level, and susceptible of
reductive explanation all the way down to that level. There is little reason
to think we live in such a world. One wonders, even, whether the view of
economics in question may not have a signi®cant ideological component.
I certainly do not mean to deny that there is any naturally occurring
economic machinery. (``Naturally occurring'' here means only not delib-
erately contrived.) It does seem plausible that in certain geographically
localized markets and in large well-organized markets for homogeneous
commodities there may be quite ef®cient supply and demand equalizing
machines. The capacity for this to happen has been well understood for
over two centuries, and it is not implausible that in suitable cases no
signi®cant in¯uences beyond economic self-interest occur to interfere
with the exercise of the capacity. Such competitive markets, at any rate,
are the most plausible candidates for moderately reliable naturally
occurring economic machines. On the other hand with many naturally
occurring markets there is a wide range of well-understood interfering
factors such as varying degrees of monopoly and product differentiation
and limits on the ¯ow of information as well, of course, as a whole range
of externalities that make the naturally occurring outcome, even if an
equilibrium, far from optimal. Moreover, it is doubtful whether some of
the social phenomena that have been theorized in this way (an example I
shall mention brie¯y below is that of labor markets) have even the
rudimentary characteristics of such machinery. A fortiori, the aggre-
gation of all these more or less reliable more or less machines, as
envisaged for example by general equilibrium theory, is unlikely to be
anything at all like a machine.
What of our ability to build economic machines? Certainly there are
plausible cases in which we have built quite reliable such machines. One
example, which I borrow from Nancy Cartwright (this volume), is the
mechanism by which central banks operate through commercial banks to
attempt to control the interest rate and money supply. This is a simple
mechanism, and in many cases there is good reason to believe that the
intended causal connections will be strong enough to outweigh unantici-
pated interferences. The standard model of this process, in other words,
may re¯ect a set of causal relations that dominate any likely interfering
factors. It is also worth recalling here the familiar idea that the intention
to build such a machine may well be an important causal factor in
bringing about the desired result, a factor that might provide a major
328 The world of economic causes

advantage for constructed over natural machines. On the other hand,

there are grounds for skepticism about our general ability to build
economic machines that will reliably serve the purposes for which they
are intended. It may well be dif®cult or even impossible to construct
economic machines that work reliably to keep a whole economy under
some sort of control. Depending on the variety and prevalence of
interfering factors, unpredictable or even random from an economic
viewpoint, there is no reason to assume any convergence on a determi-
nate economic outcome as we incorporate more factors into our models
and attempt to control more variables.
The two models I have had in mind to this point, of discovered and
constructed economic machines, might be seen as the extremes of a
familiar political-economic spectrum: untrammeled competition and
totalitarian central planning. There is no doubt a good deal of consensus
that some point between these extremes is most desirable from a practical
point of view. I now want to suggest that a pluralistic metaphysical
foundation provides a plausible basis for sketching the elements of such a
position. Rather than the homogeneous smoothly interconnected market
mechanisms envisioned by neoclassical economics, we might rather
imagine a diverse array of mechanisms, including more or less con-
strained market mechanisms, operating more or less predictably and
reliably, and embodying more or less explicit goals. To some degree these
mechanisms have been consciously engineered and to some degree they
have evolved from efforts to coordinate more primitive individualistic
motivations.
It is noteworthy that this picture meshes very closely with a certain
conception of the agenda of institutionalist economics:
The institutionalist . . . does not have an all-encompassing, ®xed analytical
framework from which to analyze the world. He has no black box. The
institutionalist economist is ®rst and foremost eclectic, recognizing the import-
ance of market forces, but at the same time concerned with the historical,
institutional, and social forces that limit the role and in¯uence of markets . . .
Lacking any con®ning ideology the institutionalist is a tinkerer . . . He spurns as
unrealistic and as inducing blindness attempts to seek one universal explanation
of all events. (Mangum 1988, p. 202)
The contrast between this picture and that typically suggested by
neoclassical economists is well illustrated by a consideration of the labor
market. The neoclassical economist sees one dominant phenomenon,
individuals selling their labor to the highest bidder, the highest bidder
being the employer who can extract the largest marginal product from
the individual's labor. Apparently anomalous phenomena are treated in
a strictly Panglossian manner. Those who ``choose'' to eschew a college
 Economics without mechanism 329

education do so because further education would only reduce the present

value of their human capital. That others maximize this value by
attending college re¯ects an exogenous difference in abilities which
ultimately can only be explained genetically. The unemployed choose not
to work because they prefer leisure (or welfare payments) to the highest
offer available for their services. And so on. The labor market turns out
to be a well-oiled machine. But apart from any empirical implausibility
of this picture, I have argued that whether one is a monistic physicalist or
a pluralist there is no a priori likelihood that such an integrated economic
machine exists. If one is a monist it is unlikely that there are any
economic machines; if one is a pluralist there will be too many, all liable
to interact and interfere with one another.
An institutionalist picture, on the other hand, might see the labor
market in close accord with the ontological picture I sketched above.
There are many different job-allocating machines, loosely interconnected,
partially motivated by explicit aims, and resulting in more or less
predictable and more or less desirable ®nal outcomes. Some of these
machines will be such things as hiring committees and the personnel
departments of ®rms or institutions. Others will be local or national
governmental institutions providing constraints on machines of the
former kind. Others again will be cultural phenomena determining the
employment priorities or expectations of people from different social
strata. I call all these machines because they embody certain more or less
reliable causal relations and, more importantly, because mechanism is
the mode in which they are assimilated into scienti®c understanding. But
it will be obvious that such eclectic variably related bits of machinery will
not be susceptible to the kind of homogenizing scienti®c theory that has
acquired such prestige in ®elds from physics to economics. It will rather
require local and detailed investigation appealing, as necessary, to
history, cultural anthropology, sociology, and so on as well as economic
theory.17
The crucial point, whether one is a physicalist or a pluralist, is to insist
that there is no reason to expect anything like causal completeness at the
economic level. What are the implications of all this for the practice of
economics? I can do little more than speculate and summarize here. First,
one might note the perspective on economic history. Perhaps this is the
17 As I suggested in passing above, it is also possible that there is nothing much like a labor
market anywhere. The historian William Reddy (1984) argues in detail that there was
nothing like a market for labor in the eighteenth-century French textiles industry from
1750 to 1900. Reddy suggests that market discourse in this context was entirely
rhetorical. It seems entirely possible that these conclusions might apply much more
widely.
330 The world of economic causes

least controversial conclusion: economic history is history. Perhaps more

coherent than just one damn thing after another, but certainly not the
observable upshot of the action of universal laws. Moving from there to
systematic economics, the obvious corollary is that there is little point in
looking for such laws. Abstractions purporting to give a mechanistic
model of an entire economy probably serve little useful purpose beyond
winning Nobel prizes for their authors. Perhaps the more schematic
formulations of standard macroeconomics may gesture towards some
real causal capacities, though the notorious disappearing Phillips curve
suggests that these may be rather transient properties of particular
historical economies.
But perhaps we can still build economic machines. This is not intended
as a brief for Stalinism. For one thing, I have no wish to deny that the
causal interactions characteristic of a market are among the most
effective causal levers we have with which to build economic machinery.
Where they are usable without undesirable side-effects, markets are
surely a better way of distributing goods than, say, a committee in the
Ministry of Commodities. On the other hand, it strikes me as perfectly
obvious that markets are tools not ends in themselves. In the many cases
in which, for various reasons many well-known, they fail, we should
surely be prepared to look for more appropriate tools. Not only do we
have much more chance of ®guring out what the likely or possible
consequences of our economic actions are than we have of discovering,
by abstract mathematical analysis, what will happen if we do nothing,
but knowledge of the former kind is likely to be enormously more useful.
By analogy with machine-making, we can hope to develop a wide range
of knowledge of what can interfere with the intended consequences of
our economic actions and develop ways of blocking such possible
interferences. Empirical investigation of the causal capacities ± and here I
include, of course, historical investigation ± would be essential as a
foundation for such a practical, engineering-like economics. And at a
policy level, none of this makes much sense without prior or at least
concurrent discussion of the economic aims that we aim to promote.
There are certainly parts of economics that are practical, empirical,
and value-driven. But equally certainly these are not the aspects of
economics that carry the most prestige in the contemporary practice of
the discipline. Nonetheless, if my philosophical analysis is correct these
are the directions in which economics has most chance of making
progress.18

18 I would like to thank Uskali MaÈki for a number of very helpful suggestions on a
penultimate draft of this paper.
 Economics without mechanism 331

References
Campbell, Donald T. (1974) `` `Downward Causation' in Hierarchically Orga-
nized Biological Systems,'' in F. J. Ayala and T. Dobzhansky (eds.), Studies
in the Philosophy of Biology, London: Macmillan Press Ltd.
Cartwright, Nancy (1983) How the Laws of Physics Lie, Oxford: Oxford
University Press.
(1990) Nature's Capacities and their Measurement, Oxford: Oxford University
Press.
DupreÂ, John (1993) The Disorder of Things, Cambridge, MA: Harvard University
Press.
(1994) ``Could There Be a Science of Economics?'', Midwest Studies in
Philosophy, XVIII: 363±78.
(1996) ``The Solution to the Problem of the Freedom of the Will,'' Philosophical
Perspectives, 10: 385±402.
(1999) Review of Tony Lawson, Economics and Reality, Feminist Economics, 5:
121±6.
DupreÂ, John and Nancy Cartwright (1988) ``Probability and Causality: Why
Hume and Indeterminism Don't Mix,'' Nous, 22: 521±36.
DupreÂ, John and Regenia Gagnier (1996) ``A Brief History of Work,'' Journal of
Economic Issues, 30 (2): 553±60.
Frank, Robert H., Thomas Gilovich and Dennis T. Regan (1993) ``Does Studying
Economics Inhibit Cooperation?'', Journal of Economic Perspectives, 7 (2):
159±71.
Friedman, Milton (1953) ``The Methodology of Positive Economics,'' in Essays
in Positive Economics, Chicago: University of Chicago Press.
Gagnier, Regenia and John Dupre (1995) ``On Work and Idleness,'' Feminist
Economics, 1 (3): 1±14.
Galison, Peter (1987) How Experiments End, Chicago: University of Chicago
Press.
Kim, Jaegwon (1993) Supervenience and Mind, Cambridge: Cambridge University
Press.
Lawson, Tony (1997) Economics and Reality, London and New York: Routledge.
MaÈki, Uskali (1989) ``On the Problem of Realism in Economics,'' Richerche
Economiche, 43 (1±2): 176±98.
(1990) ``Practical Syllogism, Entrepreneurship, and the Invisible Hand,'' in
Don Lavoie (ed.), Economics and Hermeneutics, London: Routledge.
Mangum, Stephen L. (1988) ``Comparable Worth: The Institutional Economist's
Approach,'' in Garth Mangum and Peter Philips (eds.), Three Worlds of
Labor Economics, Armonk, NY: M. E. Sharpe, Inc.
Mirowski, Philip (1989) More Heat than Light, Cambridge: Cambridge Univer-
sity Press.
Oppenheim, Paul and Hilary Putnam (1958) ``The Unity of Science as a Working
Hypothesis,'' in Minnesota Studies in the Philosophy of Science, vol. 2, ed.
H. Feigl et al., Minneapolis: University of Minnesota Press.
332 The world of economic causes

Pickering, Andrew (1984) Constructing Quarks, Chicago: University of Chicago

Press.
Reddy, William (1984) The Rise of Market Culture, Cambridge: Cambridge
University Press.
Rosenberg, Alexander (1992) Economics: Mathematical Politics or Science of
Diminishing Returns, Chicago: University of Chicago Press.
(1994) Instrumental Biology or the Disunity of Science, Chicago: University of
Chicago Press.
Samuelson, Paul A. (1983) Foundations of Economic Analysis (enlarged edition),
Cambridge MA: Harvard University Press.
Traweek, Sharon (1988) Beamtimes and Lifetimes, Cambridge, MA: Harvard
University Press.
van Fraassen, Bas C. (1981) The Scienti®c Image, Oxford: Oxford University
Press.
Waring, Marilyn (1988) If Women Counted, San Francisco, CA: Harper and
Row.
Part V
Methodological implications of
economic ontology
17 Sargent's symmetry saga:
ontological versus technical
constraints

ESTHER-MIRJAM SENT

Ever since the rise of rational expectations economics in the 1960s, the
underlying assumption of rational expectations has both been defended
for being realistic and been attacked for being unrealistic. The concept of
rational expectations starts from the idea that individuals should not
make systematic mistakes. Agents are not stupid, they learn from their
mistakes, and draw intelligent inferences about the future from what is
happening around them. While the adaptive expectations hypothesis had
the disturbing implication that it allowed individuals to make systematic
forecasting errors period after period, the rational expectations hypoth-
esis asserted that people learned from their mistakes. It was based on the
idea that guesses about the future must be correct on average if
individuals are to remain satis®ed with their mechanism of expectations
formation. Rational people would take all available information into
account and then discount that information into the future. If errors
followed a pattern, they held information that could be used to make
more accurate forecasts. The resulting predictions might still be wrong,
but what mattered was that the errors would be random. People with
rational expectations did still make mistakes, but not the same ones each
time. Individuals could differ from one another in their expectations and
still be rational if they were using different information. But when all
these individual expectations were added together, errors tended to
cancel out ± producing an aggregate view of the future that re¯ected all
the available information.
While some economists embraced rational expectations because it
allowed a more realistic interpretation of expectations formation, others
have argued that, despite the rational expectations school's insistence
that expectations should be endogenous, it has dodged the crucial
question of how expectations are actually formed (Arrow 1978; DeCanio
1979; Friedman 1979; Pesaran and Smith 1992). What some of these
critics ignore, however, is that rational expectations economists have

335
336 Methodological implications of economic ontology

given different answers to the question of how we are to understand

rational expectations as statements of economic reality. Hoover (1988,
pp. 14±16) identi®es at least two interpretations of the hypothesis. A
weak form is that people do the best they can with the information they
have. Some rational expectations economists, however, have called this
interpretation vacuous. Moreover, it is not clear that this weak form
actually implies the mathematical properties of rational expectations
models. A strong form is that people actually know the structure of the
model that truly describes the world and use it to form their expectations.
Some rational expectations economists, though, have described this form
as silly. What remains is the usual fallback position that the truth of a
hypothesis does not matter as long as it generates true predictions.
Unfortunately, this argument undercuts the belief of rational expecta-
tions economists that any acceptable model must incorporate rational
expectations, since models with a good predictive record may meet this
requirement irrespective of whether they include the rational expectations
hypothesis. Complicating matters even more is the fact there are many
interpretations of rational expectations. For what does it mean for
individuals not to make systematic mistakes? Do they play games? Do
they solve general equilibrium models? Do they ®t time-series models?
According to MaÈki (1994, p. 236), ``[t]he most important methodo-
logical issue in economics has been and persists to be over what is called
the `realism' of theories and their `assumptions'.'' MaÈki argues that there
is a need for reorienting the assumptions issue towards a discussion of
the nature of assumptions in each speci®c case. This paper will analyze
the nature of the rational expectations assumption. Rather than lumping
all the rational expectations work together, this paper starts from the
realization that there is no universal interpretation of the idea of rational
expectations. Instead, it will evaluate Thomas Sargent's analysis of the
rational expectations assumption. Rather than looking at his accomplish-
ments from the perspective of orthodox philosophy of economics or
conventional history of economic thought, this paper examines the
stories Sargent is likely to have told when he was adopting different
interpretations of rational expectations, in an attempt to understand the
alternatives available to him, the choices he made, and the consequences
of those decisions.
This paper will show that Sargent entertained different interpretations
of rational expectations during different phases. It will further illustrate
that these phases are connected through Sargent's continuous attempts
to establish symmetry among agents, economists, and econometricians.
Whereas Keynesian models typically posited the government as an agent
with rational expectations and private agents with adaptive expectations,
 Sargent's symmetry saga 337

Sargent felt that ``[t]he people in Washington aren't all that much
smarter than anybody else'' (Sargent quoted by Guzzardi 1978, p. 74).
Whereas the hypothesis of adaptive expectations typically posited agents
who were making systematic forecasting errors period after period and
economists and econometricians who were fully knowledgeable, Sargent
posited agents who were inspecting and altering their own forecasting
records just like economists and econometricians in their attempts to
eliminate systematic forecasting errors. According to Sargent (1993,
p. 21), ``[t]he idea of rational expectations is . . . said to embody the idea
that economists and the agents they are modeling should be placed on an
equal footing: the agents in the model should be able to forecast and
pro®t-maximize and utility-maximize as well as the economist ± or
should we say econometrician ± who constructed the model.''1 Or: ``The
concept of a rational expectations competitive equilibrium . . . has the
attractive property that . . . [the agents] in the model forecast . . . as well
as the economist who is modeling them'' (Sargent 1987a, p. 411). Hence,
Sargent saw no reason for superiority of one category of individuals over
another group of people. Therefore, he sought to establish symmetry
among agents, economists, and econometricians in terms of the infor-
mation they possess, the techniques, theories, and models they employ,
and the forecasts they develop. Rather than defending this symmetry,
Sargent criticized the ``realism'' of asymmetry and, instead, took sym-
metry for granted.
Since Sargent never explained the justi®cation of symmetry among
agents, economists, and econometricians, I can only speculate on this
issue. First, Sargent may defend symmetry as a quality inherent in things.
This has an ontic interpretation, meaning that it entails an intrinsic
invariance among agents, economists, and econometricians. Second,
Sargent may argue that symmetry follows from a power of recognition
inherent in the mind. This interpretation is partly ontic and partly
epistemic, which can be clari®ed by an appeal to Kantian regulative
principles, implying that symmetry among economists, econometricians,
and agents is an ideal of reason (see Falkenburg 1988).2 Third, Sargent

1 Also see Sargent (1987c, p. 76): ``[R]ational expectations possesses the de®ning property
that the forecasts made by agents within the model are no worse than the forecasts that
can be made by the economist who has the model.'' Further see Sargent (1987a, p. 440):
``We implement the hypothesis of rational expectations by assuming that agents'
expectations about unknown random variables equal the linear least squares projections
on certain information sets to be speci®ed.''
2 Kant held that the concept of a systematic unity of nature was an ideal of reason directed
towards a completeness of empirical knowledge unattainable through concepts of the
understanding. Ideals of reason in Kant's sense bring about a generalization or
338 Methodological implications of economic ontology

may claim that symmetry is a matter of metaphor. According to this

epistemic interpretation, Sargent created a metaphor of the agent as an
economist and econometrician (see Mirowski 1989a).3 In order to accom-
modate these different justi®cations, I will interpret Sargent's search for
symmetry as an ontological constraint on theory formation. Since this
interpretation can be explained both in constructivist and realist fashions,
it is neutral with respect to the different justi®cations of symmetry.
Furthermore, the ontological constraint will gain meaning as the follow-
ing sections illustrate how Sargent gave the rough notion of symmetry a
more precise formulation by embedding it in different frameworks.
As we will witness in the following sections, symmetry as an ontolo-
gical constraint on theory formation has guided Sargent's work on
rational expectations. Each of the sections is a brief case study of how
Sargent's search for symmetry resulted in different interpretations of
rational expectations and kept getting obstructed by his encounters with
technical constraints (for more, see Sent 1998b). The next section
discusses how Sargent came to the idea of rational expectations as an
econometric concept. The second one analyzes Sargent's attempts to
interpret rational expectations as both an econometric and a theoretic
construct. The third section evaluates his efforts to incorporate general
equilibrium theory into the symmetry structure. The fourth section
discusses Sargent's eventual interpretation of rational expectations as the
®nal outcome in a learning process.

1 Phase one: rational expectations through distributed lags

This section analyzes Sargent's interpretation of rational expectations

economics in the late 1960s and early 1970s. At that time, the concept of
totalization of concepts of the understanding. They do not possess objective reality, since
the completeness of empirical knowledge of particulars can never be the object of
objective knowledge. However, they lead to regulative principles that can guide our
acquisition of knowledge to the establishment of systematic unity. Since symmetry
principles enable us to discover parts of this speci®c systematic structure, they have a
regulative character in Kant's sense.
3 The recognition underpinning this epistemic interpretation is that we live in a world of
broken symmetries and partial invariance. Foundational conservation principles are
factually false and no posited invariance holds without exceptions or quali®cations.
However imperfect the world, human reason operates by means of assigning sameness
and differences through symmetries in an attempt to force a reconciliation of constancy to
change. Our very livelihoods, in the broadest possible sense, are predicated upon
invariants whose existence cannot be proven but whose instrumentality renders our
actions coherent.
 Sargent's symmetry saga 339

adaptive expectations was under severe attack for ®tting models that
forecast better than agents, as it allows individuals to make systematic
forecasting errors period after period. This was an obstacle in Sargent's
search for symmetry. The hypothesis of adaptive expectations postulates
that individuals use information on past forecasting errors to revise
current expectations. Objections to the hypothesis included, ®rst, that it
is entirely backward-looking, and that all mechanistic backward-looking
extrapolative rules allow the possibility of systematic forecasting errors
for many periods in succession. Critics argued that the suboptimal use of
available information is hard to reconcile with the idea of optimization
that is the foundation of most microeconomic analysis. Second, no
widely accepted economic theory was offered to explain the magnitude of
the adjustment parameter. Some economists sought to meet these objec-
tions by using the concept of rational expectations, by taking the idea
that individuals should not make systematic errors as their point of
departure (Begg 1982).
Inspired by these changes in his environment, Sargent attempted to
satisfy the ontological constraint of symmetry by connecting the deter-
minism employed by neoclassical economists and the randomness advo-
cated by econometricians through the use of rational expectations. He
was troubled by the fact that there was an unclear link between time-
series econometrics, in particular distributed lags, and the models that
were then being used. Sargent said to Klamer (1983, pp. 63±4): ``What I
mean by the links not being clear is that often the models that we used
had no randomness in them. They analyze individual behavior in a
context in which there is no uncertainty, but they treated the data
probabilistically, thus adding randomness. That procedure is not a tight
one, not even an understandable one. The statistical model you're using
implies that there is an environment in which there's uncertainty,
whereas the economic model that you're using assumes that away. The
hunch is, and it's a hunch that turned out right, that it's not just a matter
of adding a random term. If there really is uncertainty, it ought to change
the way you think about individual behavior.'' Sargent noticed that in
econometrics, empirical prediction requires the use of probabilistic ideas.
Once these are introduced in econometric method, connecting the work
of econometricians and economists required them to be introduced in
economic theory as well.
Sargent chose to approach symmetry by starting from the perspective
of time-series econometrics, in particular distributed lags, and the term
structure of interest rates (Sargent 1968, 1969). Through his analysis of
distributed lags for interest rates, Sargent became aware of the role of
expectations, because orthodox neoclassical theory stated that they
340 Methodological implications of economic ontology

in¯uence the relationship between spot and forward rates, nominal and
real rates, and short and long-term rates. Furthermore, expectations, or
more speci®cally rational expectations, provided Sargent with an answer
to how symmetry might be achieved, as they allowed him to introduce
probabilistic ideas in economic theory as well. He considered rational
expectations a more elegant way to resolve the separation between the
randomness of distributed lags in econometrics and the determinism of
neoclassical models. Since Sargent initially started from the viewpoint of
econometrics, he was led to choose an econometrically motivated inter-
pretation of the concept of rational expectations, which involved treating
the econometrician and the agents in the model in a symmetric fashion.
Since he focused on interest rates, Sargent also encountered the
importance of LeÂvy stable distributions with in®nite variance and the
associated problem of constructing statistical estimators. For example,
Richard Roll (1970) showed that some interest rates follow a member of
the class of LeÂvy stable distributions. Furthermore, during the late 1960s
and early 1970s, LeÂvy stable distributions began to attract the attention
of many scholars working in economics (Mirowski 1989b, 1990). Investi-
gations showed how certain variates conformed to LeÂvy stable distribu-
tions, Monte Carlo studies were mounted to place some bounds on
sampling behavior, and attempts were made to link estimation to the
formalisms of linear programming. Roll (1970) suggested extending his
work by using Sargent's spectral-method analysis of interest-rate se-
quences and discussed these issues with his colleague Sargent and his
student Robert Blattberg at Carnegie-Mellon University. Furthermore,
LeÂvy stable distributions promised to explain why MSAE outperformed
least squares in a study of investment decisions by Sargent's thesis
adviser, John Meyer (Glauber and Meyer 1964). Inspired by these
developments, Sargent published a paper in 1971, together with Blattberg
(Blattberg and Sargent 1971), studying the performance of various
estimators where the disturbances follow distributions that have fatter
tails than does the normal distribution.
The fact that distributions with in®nite variance tend to have thick or
heavy tails implies that large values or outliers will be relatively frequent.
It means that sample variances grow unpredictably and without bound
with increases in the sample size. Because the least squares technique
minimizes squared deviations, it places relatively heavy weight on out-
liers, and their presence can lead to estimates that are extremely sensitive.
If the variance does not exist, it is obviously impossible to obtain a
meaningful variance estimator and the least squares estimator will not
possess its usual minimum variance property, for the Gauss-Markov
theorem will not hold. This in turn implies that the conventional F and t
 Sargent's symmetry saga 341

tests on the coef®cients could be very misleading, for the classical central
limit theorem will not hold. If the error distribution is so thick-tailed that
the mean as well as the variance does not exist, then the least squares
estimator cannot be unbiased because its mean will also not exist. In the
best case, stable distributions have only one moment of integral order
and, therefore, estimators cannot depend on any moments higher than
the ®rst. Only in a few cases are there explicit expressions for their
density or distribution functions that would enable us to concretize the
algorithms for estimating the parameters and, therefore, precise state-
ments about the sampling behavior of estimators could generally not be
written down.
The upshot is that with LeÂvy stable distributions almost every tech-
nique of modern econometrics is useless and would have to be discarded.
As a result, almost all references to stable LeÂvy distributions in economic
variates disappeared by the mid-1970s and many of the earlier enthusiasts
recanted with regard to stable LeÂvy distributions (Mirowski 1989b,
1990). The threat of LeÂvy stable distributions, which would require
econometrics to search for algorithms for estimating the parameters, was
averted by ignoring them, without a direct critique of the earlier ®ndings
of in®nite variance. The strict deterministic stance of neoclassical theory
is incompatible with economic variates following distributions that do
not guarantee the existence of algorithms for estimating parameters.
Hence, neoclassical econometrics is a set of techniques forged to satisfy
the ontological constraint of reconciling the determinism embodied in
neoclassical theory and the apparent stochastic nature of economic
variates. It ignores the problems with constructing statistical estimators
under stable laws, primarily by privileging the method of least squares to
test or estimate its basic regression model. This requires the distribution
of the random variables to be Gaussian and hence assures the existence
of an algorithm for estimating the parameters. Randomness, therefore, is
tamed by assuming that variances are ®nite.
Sargent was especially troubled by these technical constraints, for he
wanted to satisfy the ontological constraint of establishing symmetry
between techniques used by agents and the models developed by econo-
metricians. In particular, his econometrically motivated interpretation of
rational expectations required the availability of statistical estimators.
Whereas LeÂvy stable distributions previously only threatened neoclas-
sical econometrics, they could now also compromise economic theory
based on rational expectations. When stable laws enter the stage,
econometricians and agents would run into problems with the construc-
tion of statistical estimators. Rather than relinquishing the econometri-
cally motivated interpretation of rational expectations through
342 Methodological implications of economic ontology

distributed lags, Sargent somewhat silently gave up LeÂvy stable distribu-

tions with in®nite variance. Only a few explicit discussions of LeÂvy stable
distributions can be found in his later publications. While Roll had
illustrated the importance of in®nite variance for his data, Sargent used
the same data and assumed that the random term follows a distribution
with ®nite variance (Sargent 1971). He subsequently extended this study
and noted that ``Roll argues that the evolution of bill rates is more
adequately described by assuming that they are drawn from one of the
stable distributions with in®nite variance. While that speci®cation is
certainly an interesting one, abandoning the assumption of covariance
stationarity has its costs'' (Sargent 1972, pp. 75±6). Imposing covariance
stationarity and ignoring stable distributions with in®nite variance
assured that the algorithms for estimating the parameters can be
concretized, a prerequisite for the introduction of rational expectations.
In this ®rst phase of his quest for symmetry among agents, economists,
and econometricians, therefore, Sargent encountered the technical con-
straint that LeÂvy stable distributions lack an algorithm for estimating the
parameters. This obstructed his attempts to connect the randomness in
the models used by econometricians and agents with the determinism in
the models developed by economists. Therefore, relinquishing LeÂvy
stable distributions and their technical constraints served Sargent well in
his attempts to satisfy the ontological constraint of symmetry among
agents, economists, and econometricians. Yet, this led him to adopt the
``unrealistic'' assumption that data previously shown to have exhibited
in®nite variance now followed a distribution with ®nite variance. Hence,
this phase illustrates the ®rst instance of the intricate interplay of
ontological and technical constraints in Sargent's rational expectations
economics.

2 Phase two: rational expectations through vector autoregressions

While econometricians were the ®rst promoters of rational expectations,

their initial focus on methods for restricting the parameters of lag
distributions subsequently changed to restricting vector autoregressions.
Though the same ontological constraint of establishing symmetry among
agents, economists, and econometricians guided Sargent, the change in
his econometric environment led him to advocate a different connection.
Whereas we saw in the previous section how Sargent tried to use rational
expectations to restrict distributed lags, we will witness in this phase how
Sargent, in¯uenced by this change in his econometric environment, tried
 Sargent's symmetry saga 343

to use rational expectations to restrict vector autoregressions. Whereas

Sargent had earlier tried to turn agents in economic models into
econometricians using distributed lags, we will see in this section how
Sargent tried to establish symmetry between agents in economic models
and econometricians using vector autoregressions. The times are the late
1970s and early 1980s, the places are the University of Chicago and the
University of Minnesota at Minneapolis, and the supporting roles are
performed by Christopher Sims and Lars Hansen.
In his attempts to satisfy the ontological constraint of symmetry,
Sargent continued to start from the viewpoint of econometrics during
this phase of his work. However, rather than focusing on distributed
lags, he started employing vector autoregressions. For Sargent (1987a,
p. 241), ``[s]tochastic processes provide a natural context in which to
formulate the problem of prediction . . . When the econometric model
occurs in the form of a vector version . . . it is said to be a vector
autoregression.'' The vector autoregressive model was designed especially
to forecast. It tried to overcome many of the defects of the structural
approach by relying on statistical regularities only. Whereas the struc-
tural approach attempted to use economic theory and historical data to
simulate the structure of the economy as a system of equations, vector
autoregressions were not based on economic theories at all. Whereas
large national econometric models were successful in the 1950s and
1960s, their performance hit rock bottom in the 1970s. They did not
successfully predict and could not explain the simultaneous high in¯ation
and unemployment rates. Vector autoregressions, on the other hand,
seemed capable of producing forecasts that were, compared to the
standard kind, more accurate, more frequent, and cheaper. It is a
straightforward, powerful, statistical forecasting technique that can be
applied to any set of historical data.
Some econometricians, like Christopher Sims and Sargent, responded
to the evidence that naãÈve time-series models frequently appeared to offer
better forecasting performance than other econometric models by
focusing on representing the data relative to the theory (Sargent and
Sims 1977). Sargent and Sims were fellow graduate students at Harvard
University in the 1960s and fellow professors at the University of
Minnesota in the 1970s. According to Sargent, ``[l]earning from Chris
Sims about time series and about Granger-Sims causality and how
that ®ts in was fun . . . Very early on I had a hunch that Chris' stuff
would ®t in with rational expectations'' (Klamer 1983, p. 74). Sims, one
of the pioneers of vector autoregressions, believed that theoretical
restrictions in statistical inference should be kept to a minimum (Sims
1980). In¯uenced by Sims and the change of focus in the econometrics
344 Methodological implications of economic ontology

community, Sargent concentrated on restricting vector autoregressions

on the econometric side of symmetry. As before, Sargent ended up with
an econometrically motivated interpretation of rational expectations.
Rather than handling distributed lags, symmetry between agents and
econometricians now required that agents with rational expectations ®t
vector autoregressions. Rather than solving structural models, agents
model ``without pretending to have too much a priori economic theory''
(Sargent and Sims 1977) or ``using methods not based on explicit
economic theories'' (Sargent 1979). While the theoretical background for
time series was developed in statistics, the main applications arose in
communications engineering. Sargent (1987b, p. xxii) claimed that in the
sense that ``the language that macroeconomists speak has changed . . .
there has been a rational expectations revolution.'' But this language of
applied macroeconometrics was borrowed from engineering (Sent
1998a).
In order to satisfy the ontological constraint of symmetry, Sargent
sought to incorporate economic theory in the structure of vector
autoregressions and rational expectations. Hence, when he adopted the
method of vector autoregressions, he wanted to use the acquired
statistical information to construct a theoretical model. According to
Sargent, the ``atheoretical'' approach of macroeconomics should be used
to motivate theoretical assumptions. However, a major technical con-
straint, in the form of observational equivalence (Sargent 1976), hindered
Sargent in fully meeting the ontological end of symmetry. He discovered
that models compatible with the natural rate hypothesis and models that
are incompatible with it could both generate the very same time-series
relations. Within any policy regime, there were an in®nite number of
equally accurate representations of the data. If one of these forms was
invariant to changes in the policy regime, then the other forms would in
general not also be invariant. The natural and non-natural rate hypothe-
ses both generated identical observable consequences and both formed
the basis for equally good forecasts, as long as the policy regime
remained constant. To incorporate economic theory in his symmetry
structure, Sargent felt he needed to overcome this technical constraint of
observational equivalence by establishing a stronger connection between
vector autoregressions and economic theory.4
This is where Lars Hansen enters the picture. He graduated from the
University of Minnesota, Sargent's employer, in 1978 and subsequently
collaborated with Sargent on many papers. Hansen and Sargent (1981a,

4 I should note that while these problems led Sargent to change his approach, they never
bothered Sims one bit, because Sims could care less about symmetry.
 Sargent's symmetry saga 345

1981b) responded to the problem of observational equivalence by

synthesizing structural estimation and time-series analysis and showing
that time-series models were not necessarily atheoretical. Hansen and
Sargent (1991b, p. 1) argued that their ``goal has been to create a class of
models that makes contact with good dynamic economic theory and with
good dynamic econometric theory.'' But what is good dynamic economic
theory? What is good dynamic econometric theory? Having grounded
``good dynamic econometric theory'' in the engineering tools of vector
autoregressions, Hansen and Sargent (1990, 1991b) searched for ``good
dynamic economic theory'' in the engineering theory of recursive
dynamics and linear optimal control. The trouble was that this combin-
ation was technically not terribly successful, dif®cult to implement, and
based on controversial assumptions. Sargent (1987b, p. 7) acknowledged
that ``in order to make . . . models tractable enough for macroeconomic
work, their preferences, technology, and endowments have typically been
so simpli®ed, and so much has been abstracted, that it is often dif®cult to
take their predictions seriously.'' Because ``internal consistency [or
symmetry] is always purchased with simpli®cation and abstraction.'' In
particular, Sargent's set-up relied on the assumptions of linear-quadratic
models and time invariance on the economic-theory side and covariance
stationary, linearly indeterministic models on the econometric-method
side.
Though the combination of vector autoregressions, recursive dynamics
or linear optimal control, and rational expectations helped Sargent to
satisfy the ontological constraint of symmetry in this phase, the technical
constraints he encountered led him to represent the world as being either
linear quadratic and time invariant, or covariance stationary and linearly
deterministic. These technical constraints illustrated that his models were
technically not terribly successful and dif®cult to implement. Just like in
the previous phase, the technical constraints led Sargent to adopt
``unrealistic'' representations that were based on controversial assump-
tions. It is important to stress that Sargent himself acknowledged that, as
a result of the complex interplay between ontological and technical
constraints, it is dif®cult to take the predictions of his models in this
phase seriously.
In addition, Sargent became aware of the fact that his analysis relied
on outdated engineering techniques. In particular, techniques were being
developed to analyze nonstationary and nonlinear systems. At the same
time, some economists started arguing that nonstationary behavior was
an important aspect of the economy and that major economic variables
have nonlinear relationships. Even Sargent's hero Granger began advo-
cating the new techniques for nonstationarity and nonlinearity (Granger
346 Methodological implications of economic ontology

1994; Granger and TeraÈsvirta 1993). Sargent had largely avoided ques-
tions about the way in which economic agents make choices when
confronted by a perpetually novel and evolving world. This was so,
despite the ontological importance of the questions, because of the
technical constraint that his tools and formal models were ill-tuned for
answering such questions. Changes in his environment and the appear-
ance of a few extra technical constraints were necessary, though, to
convince Sargent to move to a complexity approach to prediction, using
insights from the economics workshop he attended at the Santa Fe
Institute. Before discussing this ®nal phase, the next section will outline
the additional technical constraints that Sargent encountered.

3 Phase three: theoretically motivated rational expectations

This phase is centered on Sargent's eventual interpretation of rational

expectations as individual rationality and mutual consistency of percep-
tions. From roughly the early to mid-1980s, Sargent focused on incorpor-
ating general equilibrium theory in his framework of rational
expectations and vector autoregressions.5 He believed that the true
system was composed of such state variables as taste, technology, and
policy, so that a model was structural if its equations were strictly
linked to the deep parameters of state variables. In this interpretation, a
vector autoregression was a reduced form of a system. The general
equilibrium framework imposed full theoretical restrictions on the coef®-
cients in the vector autoregression. Whereas the previous two phases in
Sargent's work started with the conception of agents as little econome-
tricians while economists were added as somewhat of an afterthought,
the phase discussed in this section started with the conception of agents
as little economists while econometricians were added as somewhat of an
afterthought.
While Lucas had used general equilibrium theory from the start, it
took Sargent until the late 1970s to move in this direction. During that
time, he spent a year as a visiting professor at the University of Chicago
and took two courses from Lucas. He said in his interview with Klamer
(1983, p. 62): ``I don't really work much with Lucas. I spent a year at
Chicago. I took two courses from him. He's a very good teacher. I learn
from him. I read his papers. He's been a big in¯uence on me.'' His

5 Sargent did not make use of Walrasian general equilibrium analysis, but he employed
representative agent analysis instead.
 Sargent's symmetry saga 347

encounters with Lucas led Sargent (1981a, p. 214) to explore the impli-
cations ``of a single principle from economic theory. This principle is that
people's observed behavior will change when their constraints change.''
He restricted ``things so that the dynamic economic theory is of the
equilibrium variety, with optimizing agents and cleared markets.''
Sargent sought to satisfy the ontological constraint of symmetry by
linking the vector autoregressions employed by econometricians and the
general equilibrium theory developed by economists through the concept
rational expectations, because ``[r]ational expectations modeling pro-
mised to tighten the link between theory and estimation, because the
objects produced by the theorizing are exactly the objects in which
econometrics is cast'' (Hansen and Sargent 1991a, p. 3). Besides, ``Lucas
and Prescott [had done] much to clarify the nature of rational expecta-
tions as an equilibrium concept, and also pointed the way to connecting
the theory with observations'' (Sargent 1987c, p. 76). Rational expecta-
tions modeling resulted in vector autoregressions: ``This is an attractive
assumption because the solutions of such problems are known to imply
that the chosen variables . . . can exhibit serial correlation and cross-
serial correlation'' (Sargent 1981a, p. 215). Hence, in this interpretation,
agents have expectations that are rational when these depend, in the
proper way, on the same things that economic theory says actually
determine that variable. A collection of agents is solving the same
optimum problems by using the relevant economic theory and the
solution of each agent is consistent with the solution of other agents.
Econometric methods can then be used to estimate the vector autoregres-
sions that result from this economic model.
For Sargent, satisfying the ontological constraint of symmetry among
agents, economists, and econometricians with this set-up was facilitated
by the fact that general equilibrium theory involved an a priori bias
towards symmetry among agents. In particular, there are three pieces of
evidence in making the case that general equilibrium theory has always
had trouble with distinctly differentiated actors. First, Edgeworth's
analysis of exchange relied on keeping the number of types of agents
constant through cloning. Second, in the analysis of uniqueness and
stability, there are aggregation problems if agents are very different as
shown by the Sonnenschein±Debreu±Mantel result. Third, the existence
of general equilibrium cannot be proved in an economy in which agents
are so different that they are fully specialized and in which the possibility
of self-suf®ciency is the exception rather than the rule. Unfortunately,
space limitations do not allow me to discuss these technical issues in
detail (for more, see Sent 1998b). In addition, it is dif®cult to explain
them in non-technical terms. Regardless, for Sargent the technical
348 Methodological implications of economic ontology

constraints associated with general equilibrium theory led him towards

symmetry not only among categories of people but also within groups of
individuals.
Though Sargent had ®nally met the ontological end of symmetry, our
narrative does not have a happy ending here. Instead, Sargent en-
countered new technical constraints following from the combination of
rational expectations, general equilibrium theory, and vector autoregres-
sions. First, if there is symmetry among the agents, then there is no reason
for them to trade with each other, even if they possess different infor-
mation. Instead of there being a hive of activity and exchange, Tirole
(1982) proved that a sharp no-trade theorem characterizes rational
expectations equilibria (Sargent 1993, p. 113). Second, agents and econo-
metricians have to be different in order to justify the error term. When
implemented numerically or econometrically, rational expectations
models need to impute more knowledge to the agents within the model,
who use the equilibrium probability distributions in evaluating their Euler
equations, than is possessed by an econometrician, who faces estimation
and inference problems that the agents in the model have somehow solved
(Sargent 1987c, p. 79). Third, there is a need for asymmetric actors in
rational expectations economics for the concept of policy recommenda-
tions to make sense. In particular, making recommendations for improv-
ing policy amounts to assuming that in the historical period the system
was not really in a rational expectations equilibrium, having attributed to
agents' expectations about government policy that did not properly take
into account the policy advice (Sargent 1984, p. 413). A fourth problem
deals with the issue of conceptualizing learning if agents are thought to be
little econometricians. In particular, econometric metaphors of reasoning
possess a blind spot for the process of information search and errors made
in information collecting, because econometric theories of inference and
hypothesis testing are applied after the data have been collected; they do
not start until the variables and numbers needed for the formulas are
available (Sargent 1993, p. 23). These problems, combined with the ones
outlined in the previous section, eventually jointly transformed Sargent's
entire program. As we will discover in the following section, Sargent tried
to reimpose symmetry among agents, economists, and econometricians
by making them all boundedly rational.
In this phase, therefore, technical constraints associated with general
equilibrium theory led Sargent to embrace symmetry not only among but
also within categories of individuals. However, due to the technical
constraints he subsequently encountered in his attempt to connect the
techniques used by agents, the theories constructed by economists, and
the models developed by econometricians, Sargent was unable to main-
 Sargent's symmetry saga 349

tain symmetry within the set-up he had developed. A ``realistic'' account

required heterogeneous agents, an asymmetric government, and differ-
ence between agents and econometricians. Again it needs to be empha-
sized that Sargent himself recognized these outcomes of the involved
interaction between ontological and technical constraints. In response, he
sought to reestablish symmetry through the use of bounded rationality.
This is the ®nal phase in his work to which we will now turn.

4 Phase four: convergence to rational expectations

This section discusses how Sargent eventually changed his attitude

towards rational expectations in response to developments in the late
1980s. During this period, Sargent became involved with the Santa Fe
Institute. Complexity, intractable unpredictability, spontaneous self-
organization, adaptation, nonlinear dynamics, computational theory,
upheavals at the edge of chaos, inductive strategies, new developments in
computer and cognitive science ± these were some of the themes taken up
by researchers at the Santa Fe Institute. Begun by a number of
distinguished physicists at the Los Alamos National Laboratories, the
Santa Fe Institute originally had nothing to do with economics. This
changed with a workshop on ``Evolutionary Paths of the Global
Economy'' from 8±18 September 1987 at the Institute campus in Santa
Fe (Anderson, Arrow, and Pines 1988). The gathering was successful
enough to continue the economics program at the Institute, which has
anywhere from eight to ®fteen researchers in residence at any given time,
evenly divided between economists and physical scientists. Brian Arthur,
who served as a director of the Institute's economics program ``sees a
`Santa Fe approach' emerging that views the economy as a complex,
constantly evolving system in which learning and adaptation play a
major role. Eventually . . . this could expand the current view, much
in¯uenced by classical physics, that depicts the world as relatively simple,
predictable, and tending to equilibrium solutions'' (Pool 1989, p. 703).
One area that received a great deal of attention during the workshop
was the speci®c question of how economic agents take the future into
account when making decisions. The axiom of rational expectations
seemed patently untrue to the physical scientists, who were acutely aware
of the dif®culties inherent in predicting the future.6 The problem in
6 Note that the earlier defense of rational expectations by Sargent and the current criticism
of rational expectations by physicists relied on different notions of ``realism.'' Spelling
these out would take us too far from the main narrative in this paper.
350 Methodological implications of economic ontology

developing a more ``realistic'' model was that if economic agents were

assumed to be able to anticipate the future, but not perfectly, then it is
hard to know just how imperfect rationality should be. One suggestion
was to develop theoretical economic agents that learned in the way actual
economic agents did. These suggestions for incorporating learning were
music to Sargent's ears, because his engineering metaphors turned out to
be not as convenient as he had previously thought and he had developed
a keen interest in including learning in the context of rational expecta-
tions models to restore symmetry among agents, economists, and econo-
metricians. Before analyzing Sargent's embracement of a Santa Fe-type
approach, we need to ®rst analyze prior attempts by Sargent to deal with
the problems outlined in the previous two sections.
The asymmetry among agents, economists, and econometricians that
followed from the technical constraints within the setting of rational
expectations, general equilibrium theory, and vector autoregressions did
not sit well with Sargent. In response, he was led to revise part of his
framework in the mid-1980s in an attempt to satisfy the ontological
constraint of symmetry. We witness Sargent (1993) relinquishing rational
expectations, when he made a ``call to retreat from . . . rational expecta-
tions . . . by expelling rational agents from our model environments''
(p. 3) and ``to create theories with behavioral foundations by eliminating
the asymmetry that rational expectations builds in between the agents in
the model and the econometrician who is estimating it'' (pp. 21±2).
Instead of starting from rational expectations, Sargent focused on agents
with adaptive expectations in work mostly co-authored with Albert
Marcet, who was a graduate student at the University of Minnesota
during Sargent's tenure there and who subsequently followed in Sargent's
footsteps by accepting an assistant-professor position at Carnegie Mellon
University (Marcet and Sargent 1986, 1988, 1989a, b, c, 1992). The
models they developed were adaptive in the sense in which that term is
used in the control literature (but not in the macroeconomics literature).
In particular, the agents were assumed to behave as if they know with
certainty that the true law of motion is time invariant. Because the agents
operate under the continually falsi®ed assumption that the law of motion
is time invariant and known for sure, the models do not incorporate fully
optimal behavior or rational expectations.
Unwilling to relinquish rational expectations entirely, Sargent did not
see learning as anything really new in economics. He saw it as a way of
strengthening the standard ideas and dealing with their problems ± as a
way of understanding how economic agents will grope their way toward
neoclassical behavior even when they are not perfectly rational (Sargent
1993, p. 23). He tried to reinforce rational expectations by focusing on
 Sargent's symmetry saga 351

convergence to this equilibrium (Marcet and Sargent 1992, p. 140). He

also tried to use learning with adaptive expectations to deal with some of
the problems associated with rational expectations (Sargent 1993, p. 25).
Finally, incorporating learning could assist in the computation of
equilibria (Marcet and Sargent 1992, p. 161).
Once he moved to restore symmetry among agents, economists, and
econometricians by incorporating learning, Sargent had to ®gure out
what version of the learning assumption he wanted to use. Since he was
after symmetry in the setting of adaptive agents in economic theories and
vector autoregressions in econometric models, he could only satisfy his
ontological constraint if the adaptive agents were to use vector autore-
gressions (Marcet and Sargent 1992, p. 140). Under this rendition, then,
agents used the same kind of vector autoregressions and least-squares
estimation techniques as econometricians. This new framework,
however, did not fully allow Sargent to satisfy the ontological constraint
he had established due to encounters with new technical constraints.
Agents still had to be quite smart, because ``although sequential appli-
cation of linear least squares to vector autoregression is not rational for
these environments, it is still a pretty sophisticated method'' (Marcet and
Sargent 1992, pp. 161±2). In particular, agents were assumed to be
almost fully rational and knowledgeable about the system within which
they operate, lacking knowledge only about particular parameters in
laws of motion of a set of variables exogenous to their own decisions,
which they learn about through the recursive application of least squares.
Thus, the representation resulting from the technical constraints was
``unrealistic'' in the sense that agents were assumed to have already
formed a more-or-less correct model of the situation in which they were,
and learning was just a matter of sharpening up the model a bit by
adjusting a few knobs. Since Sargent pictured economists and econome-
tricians as being far from rational and knowledgeable about the system
they analyze, this ``unrealistic'' picture still left him with a rather weak
attempt at satisfying his ontological constraint on theory formation.
Unhappy with the ``unrealistic'' interpretation of learning under
adaptive expectations, Sargent wanted something closer to the way
economists and econometricians learn. How could he circumvent the
technical constraints that resulted in the ``unrealistic'' representation?
Sargent (1993) thought he could restore symmetry ``by expelling rational
agents . . . and replacing them with `arti®cially intelligent' agents who
behave like econometricians. These `econometricians' theorize, estimate,
and adapt in attempting to learn about probability distributions which,
under rational expectations they already know'' (p. 3). Sargent had
become an enthusiast for the arti®cial intelligence approach to learning
352 Methodological implications of economic ontology

after the ®rst economics workshop at the Santa Fe Institute: ``My interest
in studying economies with `arti®cially intelligent' agents was spurred by
attending a meeting . . . at the Santa Fe Institute in September 1987'' (p.
vi). Instead of assuming that agents were perfectly rational, they could be
modeled as being arti®cially intelligent and learning from experience like
real economic agents. Instead of modeling the economy as a general
equilibrium, societies of interacting arti®cially intelligent agents could be
organized into an economy. Reluctant to give up ideas like representative
agents or completed arbitrage and to renounce general equilibrium
analysis, Sargent did not go all the way with Santa Fe. Rather than using
arti®cially intelligent systems to think about populations, he saw them as
models of an individual's brain. Rather than relinquishing the neoclas-
sical notion of an equilibrium, he focused on convergence to equilibrium
(Marimon, McGrattan, and Sargent 1990).
Sargent saw what he called his bounded rationality program as an
effort to restore symmetry among agents, economists, and econometri-
cians. Whereas technical constraints had earlier frustrated his attempts to
satisfy this ontological constraint of symmetry through the use of
rational expectations, Sargent now pictured agents, economists, and
econometricians alike as being boundedly rational but converging to
rational expectations. Ironically, however, the move to arti®cial intelli-
gence came along with technical constraints that left Sargent with a new
asymmetry between him and the agents in his models. When Sargent
made the agents more bounded in their rationality, he had to be smarter
because his models became larger and more demanding econometrically.
According to Sargent (1993), ``an econometric consequence of replacing
rational agents with boundedly rational ones is to add a number of
parameters'' (p. 168) because there ``are many choices to be made in
endowing our arti®cial agents with adaptive algorithms'' (p. 134) and we
``face innumerable decisions about how to represent decision making
processes and the way that they are updated'' (p. 165). Furthermore,
arti®cial intelligence did not allow Sargent to fully establish symmetry,
because the proliferation of free parameters in the bounded rationality
program left him with an asymmetry between economists and econome-
tricians: ``Bounded rationality is a movement to make model agents
behave more like econometricians. Despite the compliment thereby made
to their kind, macroeconometricians have shown very little interest in
applying models of bounded rationality to data. Within the economics
profession, the impulse to build models populated by econometricians
has come primarily from theorists with different things on their minds
than most econometricians'' (pp. 167±8).
This ®nal phase illustrates how Sargent's attempts at satisfying the
 Sargent's symmetry saga 353

ontological constraint of symmetry continued to be frustrated as a result

of his encounters with technical constraints. Sargent himself acknowl-
edged that neither learning through adaptive expectations nor learning
through arti®cial intelligence established the symmetry he sought.
Whereas the technical constraints associated with adaptive expectations
excluded agents from the symmetry structure, the technical constraints
associated with arti®cial intelligence continued to exclude agents from the
symmetry structure and further left Sargent with an asymmetry between
economists and econometricians. Will Sargent ever be able to satisfy the
ontological constraint of symmetry among economists, econometricians,
and agents? Will he continue to encounter technical constraints that keep
him from establishing symmetry? How will the ontological and technical
constraints interact? We will have to wait and see.

5 Conclusion

An evaluation of the ``realism'' of the rational expectations assumption

must be seen in context. One of the individual contributors, Thomas
Sargent, entertained different interpretations of rational expectations in
different periods. This paper has outlined four case studies of Sargent
trying to establish symmetry among agents, economists, and econome-
tricians. The ®rst one was staged in the late 1960s through early 1970s.
The events explored in the second case study took place in the late 1970s
through early 1980s. The third one was set in the early to mid-1980s. The
events discussed in the ®nal case study took place in the late 1980s
through early 1990s.
Sargent's community in the late 1960s through early 1970s consisted of
his thesis adviser Meyer, his colleague Roll, and his student Blattberg. In
this setting, Sargent made the initial decision to focus on the randomness
of time-series econometrics, the determinism of neoclassical economic
theory, connecting economic theory and econometric method through
rational expectations, and the term structure of interest rates. As a result
of those decisions, Sargent was led to adopt an econometrically moti-
vated interpretation of rational expectations and to acknowledge the
importance of LeÂvy stable distributions. The technical constraints he ran
into were that for LeÂvy stable distributions there was no general estima-
tion method and the properties of estimators could only be investigated
in an indirect way. He accommodated these problems by giving up LeÂvy
stable distributions despite the evidence in favor of their ``realisticness.''
Sims and Hansen were Sargent's collaborators in the late 1970s through
354 Methodological implications of economic ontology

early 1980s. In¯uenced by his colleague Sims and a change in his

economic environment, Sargent tried to employ rational expectations for
restricting vector autoregressions and to use the acquired statistical
information to construct a theoretical model. The trouble was that the
technical constraint of observational equivalence implied that the natural
and non-natural rate hypotheses both generated identical observable
consequences and both formed the basis of equally good forecasts. When
Sargent tried to get rid of this technical constraint by collaborating with
Hansen on recursive dynamics and linear optimal control models, he
found new technical constraints in that his models were technically not
terribly successful and dif®cult to implement. Furthermore, these con-
straints led Sargent to develop models based on ``unrealistic'' assump-
tions. In the early to mid-1980s, Sargent hung out with Lucas. In this
environment, he initially decided to focus on general equilibrium theory,
vector autoregressions, and rational expectations. This promised to allow
him to establish symmetry not only among but also within categories of
individuals. However, he found that asymmetry appeared as a con-
sequence of the no-trade theorems, incorporating information gathering,
error term justi®cation, and making policy recommendations. Sargent
tried to accommodate these technical constraints and restore symmetry
by adopting adaptive expectations and arti®cial intelligence. Marcet and
the Santa Fe Institute were part of Sargent's environment in the late
1980s through early 1990s. In these surroundings, Sargent ®rst focused
on adaptive rather than rational expectations. This reduced the asym-
metry among agents, economists, and econometricians, but technical
constraints prohibited Sargent from fully satisfying the ontological
constraint of symmetry. Sargent tried to incorporate a more ``realistic
interpretation of learning by ®nally adopting a version of arti®cial
intelligence that was limited to convergence with representative agents.
Yet, technical constraints continued to frustrate Sargent's search for
symmetry since he was left with asymmetry between himself and agents
and between economists and econometricians.
Sargent entertained different interpretations of rational expectations in
different periods. In the late 1960s through early 1970s, he used an
econometrically motivated interpretation of rational expectations with a
focus on restricting distributed lags. In the late 1970s through early
1980s, this emphasis changed to restricting vector autoregressions.
During both these phases, Sargent started with the conception of agents
as little econometricians while economists were added to the symmetry
picture as somewhat of an afterthought. In the early to mid-1980s,
Sargent focused on how rational expectations in a general equilibrium
framework could lead to vector autoregressions. During this phase, he
 Sargent's symmetry saga 355

started with the conception of agents as little economists while econome-

tricians were added as somewhat of an afterthought. In the late 1980s
through early 1990s, Sargent tried to show convergence to rational
expectations through learning by agents, economists, and econometri-
cians alike through the use adaptive expectations or arti®cial intelligence.
Furthermore, Sargent's choices were partly inspired by his social environ-
ment; Meyer, Roll, and Blattberg in the late 1960s through early 1970s,
Sims and Hansen in the late 1970s through early 1980s, Lucas in the
early to mid-1980s, and Santa Fe and Marcet in the late 1980s through
early 1990s. It was further shown that as a result of each interpretation of
rational expectations, Sargent had to deal with different technical
constraints. He quietly relinquished LeÂvy stable distributions when they
turned out to be too threatening. He resorted to convenient engineering
metaphors in his quests for rationality and a more scienti®c macro-
economics. He was unable to retain symmetry within the rational
expectations framework. Finally, he took the convenient way out by
wreaking all kinds of travesties upon the standard Santa Fe approach.
Rather than analyzing rational expectations economics in general, this
paper has followed one rational expectations economist, Sargent,
around. Rather than criticizing or defending his work based on evalua-
tions in terms of realism, this paper has shown how Sargent's own
attempts to satisfy the ontological constraint of symmetry resulted in
different interpretations of rational expectations and kept getting ob-
structed by his encounters with technical constraints. Rather than
imposing outside standards, this paper has illustrated that Sargent was
unable to meet his own standards as a result of the intricate interplay of
ontological and technical constraints. This kind of analysis provides an
alternative to the many different, simple, and equally (un)compelling
stories about the rational expectations revolution that have been circu-
lating in the methodology and history of thought communities up until
now.

References
Anderson, Philip W., Kenneth J. Arrow, and David Pines (eds.) (1988) The
Economy as an Evolving Complex System, Redwood City: Addison-Wesley.
Arrow, Kenneth J. (1978) ``The Future and the Present in Economic Life,''
Economic Inquiry, 16(2): 157±69.
Begg, David K. H. (1982) The Rational Expectations Revolution in Macroeco-
nomics, Baltimore: Johns Hopkins University Press.
Blattberg, Robert and Thomas J. Sargent (1971) ``Regression with Non-Gaussian
Stable Disturbances: Some Sampling Results,'' Econometrica, 39(3): 501±10.
356 Methodological implications of economic ontology

DeCanio, Stephen (1979) ``Rational Expectations and Learning from Experi-

ence,'' Quarterly Journal of Economics, 93(1): 47±57.
Falkenburg, Brigitte (1988) ``The Unifying Role of Symmetry Principles in
Particle Physics,'' Ratio, 1(2): 113±34.
Friedman, Benjamin M. (1979) ``Optimal Expectations and the Extreme Infor-
mation Assumptions of `Rational Expectations' Macromodels,'' Journal of
Monetary Economics, 5(1): 23±41.
Glauber, Robert R. and John R. Meyer (1964) Investment Decisions, Economic
Forecasting, and Public Policy, Boston: Harvard Business School.
Granger, Clive W. J. (1994) ``Forecasting in Economics,'' in Neil A. Gershenfeld
and Andreas S. Weigend (eds.), Time Series Prediction: Forecasting the
Future and Understanding the Past, Reading, MA: Addison-Wesley.
Granger, Clive W. J. and Timo TeraÈsvirta (1993) Modeling Nonlinear Economic
Relationships, Oxford: Oxford University Press.
Guzzardi, Walter (1978) ``The New Down-to-Earth Economics,'' Fortune, 21
December: 72±9.
Hansen, Lars P. and Thomas J. Sargent (1981a) ``Formulating and Estimating
Dynamic Linear Expectations Models,'' in Robert E. Lucas and Thomas J.
Sargent (eds.), Rational Expectations and Econometric Practice, Minneapolis:
University of Minnesota Press.
(1981b) ``Linear Rational Expectations Models for Dynamically Interrelated
Variables,'' in Robert E. Lucas and Thomas J. Sargent (eds.), Rational
Expectations and Econometric Practice, Minneapolis: University of Minne-
sota Press.
(1990) ``Recursive Linear Models of Dynamic Economies,'' National Bureau
of Economic Research Working Paper No. 3479.
(1991a) ``Introduction,'' in Lars P. Hansen and Thomas J. Sargent (eds.),
Rational Expectations Econometrics, Boulder: Westview Press.
(1991b) ``Recursive Linear Models of Dynamic Economies,'' unpublished
manuscript. Stanford University, Hoover Institution.
Hoover, Kevin D. (1988) The New Classical Macroeconomics, New York: Basil
Blackwell.
(2001) ``Is Macroeconomics for Real?'', this volume, pp. 225±45.
Kim, Kyun (1988) Equilibrium Business Cycle Theory in Historical Perspective,
Cambridge: Cambridge University Press.
Klamer, Arjo (1983) Conversations with Economists, Savage: Rowman and
Little®eld.
Latour, Bruno (1987) Science in Action, Cambridge, MA: Harvard University
Press.
Lawson, Tony (1994) ``A Realist Theory for Economics,'' in Roger E. Backhouse
(ed.), New Directions in Economic Methodology, London: Routledge.
Lucas, Robert E. (1987) Models of Business Cycles, Oxford: Basil Blackwell.
Lucas, Robert E. and Thomas J. Sargent (1979) ``After Keynesian Macroeco-
nomics,'' Federal Reserve Bank of Minneapolis Quarterly Review, 3 (Spring):
1±16.
 Sargent's symmetry saga 357

(1981) ``Introduction,'' in Robert E. Lucas and Thomas J. Sargent (eds.),

Rational Expectations and Econometric Practice, Minneapolis: University of
Minnesota Press.
MaÈki, Uskali (1994) ``Reorienting the Assumptions Issue,'' in Roger E. Back-
house (ed.), New Directions in Economic Methodology, London: Routledge.
Marcet, Albert and Thomas J. Sargent (1986) ``Convergence of Least Squares
Learning Mechanisms in Self-Referential Linear Stochastic Models,''
Hoover Institution Working Papers in Economics E-86±33.
(1988) ``The Fate of Systems With `Adaptive' Expectations,'' American
Economic Review, 78(2): 168±72.
(1989a) ``Convergence of Least Squares Learning Mechanisms in Self-Refer-
ential Linear Stochastic Models,'' Journal of Economic Theory, 48(2):
337±68.
(1989b) ``Convergence of Least Squares Learning in Environments with
Hidden State Variables and Private Information,'' Journal of Political
Economy, 97(6): 1306±22.
(1989c) ``Least Squares Learning and the Dynamics of Hyperin¯ation,'' in
William Barnett, John Geweke, and Karl Shell (eds.), Economic Complexity:
Chaos, Sunspots, and Nonlinearity, Cambridge: Cambridge University Press.
(1992) ``The Convergence of Vector Autoregressions to Rational Expectations
Equilibrium,'' in Alessandro Vercelli and Nicola Dimitri (eds.), Macroeco-
nomics: A Strategic Survey, Oxford: Oxford University Press.
Marimon, Ramon, Ellen McGrattan, and Thomas J. Sargent (1990) ``Money as a
Medium of Exchange in an Economy with Arti®cially Intelligent Agents,''
Journal of Economic Dynamics and Control, 14(2): 329±74.
Mirowski, Philip E (1989a) More Heat than Light, Cambridge: Cambridge
University Press.
(1989b) ``'Tis a Pity Econometrics isn't an Empirical Endeavor: Mandelbrot,
Chaos, and the Noah and Joseph Effects,'' Richerche Economiche, 43(1±2):
76±99.
(1990) ``From Mandelbrot to Chaos in Economic Theory,'' Southern Economic
Journal, 57(2): 289±307.
Muth, John F. (1960) ``Optimal Properties of Exponentially Weighted Fore-
casts,'' Journal of the American Statistical Association, 55: 299±306.
(1961) ``Rational Expectations and the Theory of Price Movements,'' Econo-
metrica, 29: 315±35.
Pesaran, M. Hashem and Ron Smith (1992) ``The Interaction Between Theory
and Observation in Economics,'' The Economic and Social Review, 24(1):
1±23.
Pickering, Andrew (1992) ``From Science as Knowledge to Science as Practice,''
in Andrew Pickering (ed.), Science as Practice and Culture, Chicago:
University of Chicago Press.
Pickering, Andrew and Adam Stephanides (1992) ``Constructing Quaternions:
On the Analysis of Conceptual Practice,'' in Andrew Pickering (ed.), Science
as Practice and Culture, Chicago: University of Chicago Press.
358 Methodological implications of economic ontology

Pool, Robert (1989) ``Strange Bedfellows,'' Science, 245: 700±3.

Roll, Richard (1970) The Behavior of Interest Rates, New York: Basic Books.
Sargent, Thomas J. (1968) ``Interest Rates in the Nineteen-Fifties,'' Review of
Economics and Statistics, 50(2): 164±72.
(1969) ``Commodity Price Expectations and the Interest Rate,'' Quarterly
Journal of Economics, 83(1): 127±40.
(1971) ``Expectations at the Short End of the Yield Curve: An Application of
Macaulay's Test,'' in Jack M. Guttentag (ed.), Essays on Interest Rates, vol.
2, Columbia University Press, for the NBER.
(1972) ``Rational Expectations and the Term Structure of Interest Rates,''
Journal of Money, Credit, and Banking, 4(1): 74±97.
(1976) ``The Observational Equivalence of Natural and Unnatural Rate
Theories of Macroeconomics,'' Journal of Political Economy, 86(6):
1009±44.
(1979) ``Estimating Vector Autoregressions Using Methods Not Based On
Explicit Economic Theories,'' Federal Reserve Bank of Minneapolis Quarterly
Review, 3(3): 8±15.
(1981a) ``Interpreting Economic Time Series,'' Journal of Political Economy,
89(2): 213±48.
(1981b) ``Lecture Notes on Filtering, Control, and Rational Expectations,''
unpublished manuscript, University of Minnesota, Minneapolis.
(1984) ``Autoregressions, Expectations, and Advice'' (with discussion), Amer-
ican Economic Review, 74(2): 408±21.
(1987a) Macroeconomic Theory, 2nd edition, Boston: Academic Press.
(1987b) Dynamic Macroeconomic Theory, Cambridge, MA: Harvard Univer-
sity Press.
(1987c) ``Rational Expectations,'' in John Eatwell, Murray Milgate, and Peter
Newman (eds.), The New Palgrave, London: Macmillan.
(1993) Bounded Rationality in Macroeconomics, Oxford: Oxford University
Press.
Sargent, Thomas J. and Christopher A. Sims (1977) ``Business Cycle Modeling
without Pretending to Have Too Much A Priori Theory,'' in Christopher A.
Sims (ed.), New Methods in Business Cycle Research: Proceedings From a
Conference, Federal Reserve Bank of Minneapolis.
Sent, Esther-Mirjam (1998a) ``Engineering Dynamic Economics,'' History of
Political Economy, 29, special issue: 41±62.
(1998b) The Evolving Rationality of Rational Expectations: An Assessment of
Thomas Sargent's Achievements, Cambridge: Cambridge University Press.
Sims, Christopher A. (1980) ``Macroeconomics and Reality,'' Econometrica, 62:
540±52.
Tirole, Jean (1982) ``On the Possibility of Speculation under Rational Expecta-
tions,'' Econometrica, 50(5): 1163±82.
18 Two models of idealization in
economics

ALAN NELSON1

Economics and social sciences in general can be troubling because of

their lack of utility in making accurate quantitative predictions. The
trouble cannot be simply written off to the complexity of the phenomena
to be predicted because physical phenomena are at least as complex on
most sorts of complexity metrics that one might try to imagine. Modern
physics is able to work through complexity by means of idealizations
which enable us to simplify our analyses to the point where they have
quantitative results that actually prove useful. Let us examine the
suggestion that idealization in economics is different from idealization in
physics (which is not yet to suggest that the difference implies a fault) to
see whether we might get some perspective on the quantitative perform-
ance of economics.
The nature of idealization in science is, of course, not fully understood
± not even in physics. There are, however, some aspects which have been
well understood since the time of Newton and Leibniz. Physics, for
example, postulates such ideal entities as point masses, ideal gases, and
frictionless planes. One reason that the name ideal is appropriate for
these entities is that they clearly differ from any real massive body,
sample of gas, or surface of an object. And the difference is not restricted
to the fact that point masses have less volume than actual bodies, ideal
gases less intermolecular attraction, and frictionless planes less friction
than their real counterparts. It is quite impossible that there be a massive
body with no volume, gas molecules that exert no attractive forces on
each other, and surfaces without friction.
There is an explanation of why the impossibility of these entities does

1 Talks partly based on the material in this paper were given at the Erasmus University of
Rotterdam, and at the 1996 meeting of the American Economic Association. I am grateful
to both audiences for useful comments and to Uskali MaÈki for criticism of the written
version.

359
360 Methodological implications of economic ontology

not detract from their utility. We have quantitative laws that apply in a
perfectly exact way to these ideal entities. These laws do not work so well
± sometimes hardly at all ± in the same form for many actual entities.
Why do we nevertheless regard them as true and, in some ways,
fundamental? The answer is that the behavior of real entities that are
very similar to the ideal ones in the relevant respects is very closely
approximated by the exactly calculated behavior of the ideal. What is
more, the closer the real entities are in the relevant respects to the ideal,
the more closely the ideal approximates their own behavior. And ®nally,
if we examine the relationship between, on the one hand, the degree of
similarity of the real to the ideal and, on the other hand, the accuracy
with which the behavior of the ideal approximates the behavior of the
real, we ®nd a function that is continuous and may have other nice
properties as well. Suppose, for example, that we do experiments with
surfaces having a wide range of coef®cients of friction and record the
error we get by applying the law for a frictionless surface. We will ®nd
that the error smoothly approaches zero as the coef®cient of friction
approaches unity. The great seventeenth-century philosopher Leibniz
thought that these remarkable facts could be deduced from a great
metaphysical Principle of Continuity, but for physicists, it is enough that
they are true.
The concept of idealization under discussion has it that the behavior of
real objects nicely approaches the behavior prescribed by ideal laws as
the real objects nicely approach ideal objects in the relevant respects. Let
us call this Standard Idealization. It is very tempting to suppose that
Standard Idealization operates in economics just as it does in the physical
sciences. Indeed, comparisons with frictionless planes and the like are
frequently encountered in textbooks. One reads that the perfectly max-
imizing consumer is like an ideal gas molecule, or that the assumption of
perfect information is like the assumption of a perfect vacuum, or that
economic general equilibrium is like the equilibrium of a sample of gas,
and so on. It is generally realized that people are not perfect maximizers,
that perfect information is impossible, and that economies are never in
general equilibrium (though this last one is, perhaps, somewhat contro-
versial). It also seems to be generally assumed, however, that in common
situations where people behave nearly like perfect maximizers, etc., that
the results of idealized economic reasoning will make a good quantitative
®t with reality. In other words, it seems to be generally assumed that
Standard Idealization is prevalent in the application of economic theory
or in the theory itself.
There is another concept of scienti®c idealization that is important to
consider even though it does not seem to play a role in modern physical
 Two models of idealization in economics 361

sciences.2 It should be called Cartesian Idealization because it was

described and developed by the great scientist, mathematician, and
philosopher Rene Descartes in the seventeenth century, though it has not
received much attention. I shall argue that this neglect is unwarranted
because its application to economics is illuminating.
In his work Principles of Philosophy, Descartes derives three funda-
mental laws of physics ± they bear some resemblance to Newton's more
famous (and successful) three laws. The laws are derived from metaphy-
sical principles concerning matter (it has only the geometrical properties
of extension) and God (He is immutable). In retrospect, we easily see this
commitment to substantive metaphysical principles as the root of
Descartes's eventual failure in physics. It was largely through attention
to observed phenomena that Newton found the critical concept of mass
and the computational device of the center of mass which is so important
for the quantitative success of Newton's laws.
Descartes's laws are miserably inadequate for empirical application.3
Descartes himself was well aware that his laws could not be applied to
give quantitative predictions, but this left him quite undaunted. There
were at least two reasons for his attitude. One reason is that Descartes
thought it completely impossible to have quantitative laws exactly
describing the kinematics of bodies. His universe is absolutely full of
inde®nitely small particles whose effects on the bodies we actually
observe are not at all negligible. The mechanical world is much too
complex for any actual behavior to be described in quantitative detail ±
not even by laws that are known to be true. It is ironic that Descartes's
laws could apply exactly and quantitatively to observable bodies only if
these were situated in the complete simplicity of a vacuum. Vacuums,
too, turn out to be utterly impossible in Cartesian metaphysics because
of the famous dictum that any extension whatsoever is nothing other
than body. Why do the ``laws'' have the status of laws if they are
empirically useless? This brings us to the second reason that Descartes
was unperturbed by the quantitative inapplicability of his laws. It is that
such calculations were not even among the goals of the scienti®c
enterprise. This requires some explanation.
Descartes was a mechanistic scientist. This means that all phenomena
which admit of any scienti®c explanation at all must be explicable solely
2 There is a large literature on the methodology of idealization in economics. A good place
to ®nd recent contributions is Hamminga and de Marchi (eds.) (1994).
3 They seem to predict, for example, that a speeding bullet will be repulsed by a ripe
watermelon. Descartes's problem, of course, is with his concept of matter as consisting in
extension. This means that he cannot have the concept of density (quantity of matter per
unit volume) and hence cannot have the crucial concept of mass.
362 Methodological implications of economic ontology

in terms of matter and its motion. Mechanistic science was, in part, a

reaction to scholastic philosophy which posited various forces and
in¯uences that were regarded as ``occult'' by the mechanists. One of the
principal challenges to mechanism was the existence of phenomena that
seem to defy purely mechanical explanation such as gravitation, mag-
netism, and the functioning of sensory organs. Descartes was usually
satis®ed in his scienti®c writings and correspondence if he could tell a
mechanistic story about such phenomena which showed how it was
possible that they were the manifestations of nothing more than matter in
motion. So when he proposed that magnets attract iron because they
emit invisibly small screw-shaped particles which screw into appropri-
ately shaped ori®ces in the iron, he thinks with the famous Cartesian
perfect certainty that this account had to be completely true. Since he
had given a metaphysical proof that some kind or other of mechanical
story must be true, establishing the mere possibility of a particular
account suf®ced to defeat the claim that it was impossible for mechanism
to deal with this or that phenomenon.
One result of this project of furthering the mechanistic revolution in
science is that Cartesian physics runs on two tracks. The ®rst track
consists of laws that are known to be true of the world since they are
deduced with perfect certainty from metaphysical ®rst principles. These
laws are quite literally ideal since they follow from our divinely guaran-
teed innate ideas of body and of God. They are, in a later philosophical
terminology, a priori. Even though the laws are true they are, as already
noted, completely useless for accurate predictions of the behavior of
physical objects. Since these ideal laws do not directly explain anything,
Cartesian science requires another track. This consists in qualitative and
schematic accounts of mechanisms which could underlie the phenomena
of interest, and which are not downright inconsistent with the ideal
laws.4
This can all be summarized by saying that Cartesian Idealization
makes the fundamental laws absolutely true since they are derived from
certainly true ideas. The truth of the laws does not mean, however, that
they can be used to produce accurate quantitative descriptions. Euclidean
geometry is true of three-dimensional shapes, but one does not use it to
produce the most perspicuous description of the shape of a rose. It is, in
fact, physically impossible that there be a physical object that is a perfect
Euclidean solid. Similarly, one does not use the laws of motion to
describe accurately the mechanical universe; one instead gives qualitative
descriptions of mechanisms that are compatible with the laws. It is,

4 This interpretation of Descartes is developed in Nelson (1995).

 Two models of idealization in economics 363

moreover, impossible in Cartesian physics that the fundamental laws of

physics yield accurate quantitative results.5
Now it might seem disingenuous to suggest that Cartesian Idealization
can help us as we turn to questions about how to understand economics.
Descartes's physics, despite the popularity it enjoyed well into the
eighteenth century, did not stand up to comparison with Newton's.
Today, we might even hesitate to confer upon it the honori®c term,
Science, because of its explicit reliance on metaphysical principles.
Nevertheless, the failure of Descartes's physics does not necessarily
re¯ect badly on its attendant conception of idealization. Unless we are
committed to the assumption that any science must resemble physics in
all methodological respects, we must consider independently which
concept or concepts of idealization ®t best with economics.
The project of understanding economics through Cartesian Idealiza-
tion is not as outlandish as it might seem; in fact, in some respects it is
not even entirely original. The illustrious economic methodologist Lionel
Robbins can be interpreted as advocating something quite similar.
Robbins has been described as a ``platonist'' and an ``a priorist'' (and
worse!), and we are now in a position to understand how there is some
justice in these attributions. But we are also in a position to understand
his contribution more sympathetically. Let us reexamine what his view
was. One prominent aspect of Robbins's methodological doctrine was his
insistence that economics should form a deductive system.6 For this
reason he is also sometimes considered to be allied with logical postivists.
In the second, 1935 edition of An Essay on the Nature and Signi®cance of
Economic Science, Robbins wrote:
As we have seen, [the propositions of Economics] are deductions from simple
assumptions re¯ecting very elementary facts of general experience. If the premises
relate to reality the deductions from them must have a similar point of reference.
(Robbins 1935, p. 104, emphasis added)
Deductive systems in science have virtues of elegance and clarity, but
they are, perhaps, mainly motivated by the consideration mentioned here
by Robbins. If we can be sure of the truth of the axioms, or premises,
then we can be similarly sure of what we deduce from them. In
economics, there is a special concern about whether the deductions from
5 The laws could apply exactly only if the bodies in question were perfectly isolated from all
others (i.e. in a vacuum) and perfectly elastic (i.e. absolutely hard). Both of these
conditions are physically impossible ± in fact, they involve logical contradiction (see
Nelson 1995).
6 It is unlikely that Robbins was directly in¯uenced by Descartes's presentation of physics.
Blaug describes the connection between Robbins's methodological writings and those of
J. S. Mill and, especially, J. E. Cairns (Blaug 1980, pp. 86±91).
364 Methodological implications of economic ontology

the premises relate to reality because economics performs so badly in

quantitative predictions.
This concern leads, through an interesting twist, to the methodo-
logical doctrine associated with Milton Friedman. This doctrine has it
that the job of economic theory is to ®nd premises for deductions which
in fact result in good predictions. These premises, or ``assumptions'' are
then instrumentally justi®ed whether or not they are realistic. Here,
``realistic'' means something like ``quantitatively applicable.'' The
concept of Cartesian Idealization separates quantitative applicability (or
realism) from truth. So economists in¯uenced by Friedman and em-
ploying Standard Idealization might say that good economic premises
are false. An economist impressed by Cartesian Idealization wants
premises that are true even though they might not be realistic or
quantitatively applicable.
Robbins fully realized these concerns about quantitative prediction,
but did not think they re¯ected badly on either the premises or the
deductions of economics. He wrote:
But to recognise that Economic laws are general in nature is not to deny the
reality of the necessities they describe . . . having delimited the nature and the
scope of such generalizations we may proceed with all the greater con®dence to
claim for them a complete necessity within this ®eld.
Economic laws describe inevitable implications. If the data they postulate are
given, then the consequences they predict necessarily follow . . . Granted the
correspondence of [the analytic method's] original assumptions and facts, its
conclusions are inevitable and inescapable. (Robbins 1935, pp. 121±2)

The laws of economics are, therefore, absolutely true of real

economic phenomena. Robbins did not mean, however, that they can
be quantitatively applied in any exact way. This is only something we
have been thoroughly conditioned to expect because of faulty compar-
isons between physics (and Standard Idealization) and economics (and
Cartesian Idealization). True laws that have no quantitative application
seem paradoxical if one has in mind Standard Idealization, but after
studying Descartes we know better. Descartes shows under what
conditions it is possible to have true laws that are quantitatively
almost worthless. Robbins also knew better. Concerning money, for
example, he wrote:
We are not justi®ed in asserting, however, as has been so often asserted in recent
years, that if the exchanges fall, in¯ation must necessarily follow . . . We know
that governments are often foolish and craven and that false views of the
functions of money are widely prevalent. But there is no inevitable connection
between a fall in the exchanges and a decision to set the printing presses working.
 Two models of idealization in economics 365

A new human volition interrupts the chain of ``causation.''7 (Robbins 1935,

p. 128)

This example shows that the goals of economic science do not

generally include accurate quantitative prediction. Instead, the real goals
of economics could be taken to be threefold:
(1) To provide deductions of propositions that are absolutely true of the
world.
(2) To provide qualitatively predictions that are accurate most of the
time.
And perhaps,
(3) To provide post hoc reconstructions of historical events, usually
recent events. Sometimes these reconstructions can be expected to ®t the
historical record with considerable quantitative precision.
Robbins himself (and Descartes for that matter) does not seem to have
recognized (3), but it is not inconsistent with this concept of idealization
and since it does describe a very large proportion of research in
economics, it should be included in the descriptive methodological
concept being developed. Robbins also does not have much to say about
(2), though we have seen that Descartes does. Here, however, it would
seem that Descartes's important objective of providing qualitative
mechanistic explanations has a clear analogue in much of contemporary
economics. A good deal of macroeconomics has qualitative explanation
and prediction as an explicit goal. And microeconomics is often defended
against critics by pointing out that it too can sometimes provide useful
qualitative predictions ± as useful anyway as those provided in such
sciences as meteorology and seismology.
If (1)±(3) can be accepted as the goals of economics, then envy of
(post-Cartesian) physics should not discourage economists faced with
these failures of quantitative prediction. As Robbins noted:
It may be admitted that our knowledge of the facts which are the basis of
economic deductions is different in important respects from our knowledge of the
facts which are the basis of the deductions of the natural sciences. It may be
admitted, too, that from this reason the methods of economic science ± although
not the tests of its logical consistency ± are often different from the methods of
7 The reference here to ``human volition'' strikingly parallels one of the sources of
unmanageable complexity in Cartesian physics. Descartes made thinking substances and
physical substance utterly distinct, but was notoriously very vague on how the two kinds
of substance interact. Insofar as physical substance is affected by minds, the laws of
physics will fail to be accurate in particular applications.
366 Methodological implications of economic ontology

natural sciences . . .
In Economics, as we have seen, the ultimate constituents of our fundamental
generalizations are known to us by immediate acquaintance . . . There is much
less reason to doubt the counterpart in reality of the assumption of individual
preferences than that of the assumption of the electron. (Robbins 1935,
pp. 104±5)
So economics is actually in a better situation than physics with respect to
its basic assumptions. It is not even necessary to justify them either by
direct empirical test, or by Friedman-style reference to the usefulness of
the theorems that can be proved from them.
These [the most basic postulates of economics] are not postulates the existence of
whose counterpart in reality admits of extensive dispute once their nature is fully
realized. We do not need controlled experiments to establish their validity: they
are so much the stuff of our everyday experience that they have only to be stated
to be recognized as obvious. (Robbins 1935, p. 79)
Moreover, since the structure of economics is deductive, the deductive
consequences of the premises also have superior epistemological status to
those of physics. Physics comes off better only with respect to quanti-
tative applications. Robbins repeatedly advises caution about the pro-
spects of quantitative laws in economics, writing:
Perhaps, indeed, this is another of the methodological differences between the
natural and the social sciences. In the natural sciences the transition from the
qualitative to the quantitative is easy and inevitable. In the social sciences, for
reasons which have already been set forth, it is in some connection almost
impossible, and it is always associated with peril and dif®culty. (Robbins 1935,
p. 111)
Let me summarize the positive part of this paper's argument. Consider
what economics is indisputably strong at. One would include rigorous
deductions from elegant mathematical premises, qualitative predictions
that are sometimes used to good effect by governments and ®rms, and
quantitative retrodictions. These are exactly the things one would expect
if we understand economics as making use of Cartesian Idealization.
Now consider where economics is indisputably weak, or at least not so
strong as physics etc.
Economics is weak at making accurate quantitative predictions con-
cerning phenomena of interest to economic agents. Given Standard
Idealization, this is a very serious and puzzling problem for the entire
enterprise of economic science. In Cartesian Idealization, however,
accurate quantitative prediction is not to be expected. On the contrary,
we are free to embrace the powerful arguments that have been given (by
Robbins and many others) for why we should not expect accurate
 Two models of idealization in economics 367

quantitative prediction: crucial parameters change unpredictably, exo-

genous causal factors intrude inevitably, and so on. In short, Cartesian
Idealization seems clearly superior as a descriptive tool for understanding
what actually gets done by economists. Saying it is a ``descriptive tool''
means that it makes sense of the activity of economists. It explains why
they are interested in (and good at) (a) ®nding elegant premises that
permit the deduction of interesting theorems and (b) describing rough
mechanisms that enable qualitative forecasting. Saying that Cartesian
Idealization describes economics is not to say that it ®ts with most
economists' self-image, or that most economists would naturally formu-
late it as an answer to the interviewer's question: ``What are you doing?''
Many, perhaps most economists, who consider the question explicitly
would naturally come up with the picture of Standard Idealization which
describes physics so well.
One can also imagine that an understanding of Cartesian Idealization
could have prescriptive value for economics. It could help focus the
efforts of economists in those arenas where there is progress to be made:
proving interesting theorems, and suggesting mechanisms that produce
good qualitative predictions. Such a prescription would not be taken
kindly, however, by those whose careers are invested in trying to do what
the Cartesian concept entails is not part of the goal of economics in the
®rst place ± making quantitative predictions.
It is possible to put a negative spin on the argument of this paper.
There is some tension in both Descartes's and Robbins's thought
between the idea that science proceeds by deduction from principles that
are extremely simple and evident, and the fact that the principles are not
at all trivial to discover. Descartes convinced very few, for example, that
the easily introspectible innate idea of body reveals that the essence of
body is nothing but geometrical extension. Worse, in historical retrospect
we see that his physics was superseded by another that showed the ideas
that seemed so certain and innate to Descartes himself were ¯atly wrong.
Regarding the simple fact that people have well-behaved preferences,
Robbins writes:
The business of discovery consists not merely in the elucidation of given premises
but in the perception of the facts which are the basis of the premises. The process
of discovering those elements in common experience which afford the basis of our
trains of deductive reasoning is economic discovery just as much as the shaking
out of new inferences from old premises . . . But the great discovery, the
Mengerian revolution, which initiated this period of progress, was the discovery
of the premises themselves . . . The perception and selection of the basis of
economic analysis is as much economics as the analysis itself. (Robbins 1935,
pp. 105±6)
368 Methodological implications of economic ontology

I suppose it is true that some principles which are completely obvious

once stated are extremely dif®cult to discover, but there is a competing
hypothesis. It is always possible that the obviousness of the principles
that are discovered with so much labor is more a matter of ideology and
indoctrination than a matter of absolute truth. Three hundred years
later, we can see that is how it was with Descartes, anyway. Perhaps we
should ask ourselves the same kinds of hard questions about the basic
postulates of economics that Newton asked when faced with the basic
postulates of Cartesian physics. Consider the propositions that Robbins
himself considered fundamental. Do individuals really have ``prefer-
ences'' that can be ``arranged in order'' in the senses of those terms that
are required by economics? Can we separate and count ``factors of
production'' in the same sense and then see that they must be subject to
diminishing returns? A consideration of the Cartesian, or Robbinsian
concept of idealization would ultimately refocus methodological investi-
gations on these kinds of questions.8

References
Blaug, Mark (1980) The Methodology of Economics, Cambridge: Cambridge
University Press.
Descartes, R. (1644/1983) Principles of Philosophy (trans. V. Miller and
R. Miller), Dordrecht: Reidel Publishing Company.
Hamminga, B. and N. de Marchi (eds.) (1994) Idealization in Economics, special
issue of Poznan Studies in the Philosophy of the Sciences and the Humanities,
38.
Nelson, Alan (1990) ``Are Economic Kinds Natural?,'' Minnesota Studies in the
Philosophy of Science, vol. 14: 102±35.
(1995) ``Micro-chaos and Idealization in Cartesian Physics,'' Philosophical
Studies, 77: 377±91.
Robbins, Lionel (1935) An Essay on the Nature and Signi®cance of Economic
Science (2nd edn.), London: Macmillan.

8 One approach to answering these questions is developed in Nelson (1990).

19 The way the world works (www):
towards an ontology of theory choice

È KI
USKALI MA

1 Introducing the ontology of theory choice

Economists choose theories and they choose ways of pursuing theories,

and they leave others unchosen. Why do economists choose the way they
do? How should economists choose? What are the objectives and what
are the constraints? What should they be? The questions are both
descriptive and prescriptive.
There are two broad classes of ``criteria of choice'' that have been
somewhat systematically considered in the recent literature on economic
methodology:
1. Empirical criteria. There are several possible ways of incorporating
empirical criteria in one's theory of science. The respective method-
ology of theory assessment may be static or dynamic, it may be
deductivist or inductivist, it may include various ideas of what
constitutes empirical evidence, and so on. What they all share is the
general idea that scienti®c theories are, or are to be, checked against
empirical evidence according to some rules, and that this determines
the choice of theory.
2. Social criteria. Again, there are several options. The social criteria
may be related to the social interests of scientists or larger social
collectives, they may be based on the persuasiveness and tradition-
boundedness of theories, they may involve social or moral norms,
they may be derived from various costs and bene®ts of holding a
theory in a given research community, and so on. If they involve

Earlier versions of this paper have been presented at a seminar at the High Council of
Scienti®c Research of Spain and at Erasmus Institute for Philosophy and Economics.
Acknowledgements to the two audiences for inspiring reactions and in particular to Jack
Vromen and Frank Hindriks for detailed comments.

369
370 Methodological implications of economic ontology

empirical data, it is the social aspects of the data that matter. What all
these views share is that scienti®c theories are taken to have social
attributes (functions, consequences) that play or should play a major
role in theory choice.
In the ®eld of economic methodology, most of the research and discus-
sion during the era of ``Popperian dominance'' (see MaÈki 1990a) was
devoted to examining the empirical criteria and to assessing theories and
larger theoretical constellations in terms of such criteria (e.g., Latsis
1976; Blaug 1980; Weintraub 1985). This research led to the discovery
that empirical criteria play only a limited role in theory development and
in discriminating between rival fundamental theories (e.g., chapters in de
Marchi 1988; de Marchi and Blaug 1991; de Marchi 1993). Theories turn
out to be severely underdetermined by empirical evidence so that the
Duhem±Quine problem is taken to be particularly dif®cult in economics.
Partly in response to the failure to establish a systematic relationship
between theory choice and empirical criteria, attempts have recently been
made to look at the social dynamics of theory choice in economics. Given
that theory choice is taken to be radically underdetermined by empirical
tests, it is suggested that various social factors adopt the role as determi-
nants, complementing or replacing or shaping the conventional empirical
criteria. Some of the key categories used for depicting the allegedly powerful
social factors in play have been ``conversation,'' ``rhetoric,'' ``social con-
struction,'' and more speci®cally, ``sunk costs,'' ``path dependence,''
``physics envy,'' ``gender bias,'' and others (e.g., Klamer 1983; McCloskey
1985; Mirowski 1989; Weintraub 1991; MaÈki 1992d, 1993a, 1999; Hands
1994; Ferber and Nelson 1993; Zamora 1999). Many of the possible
directions and issues within this class of factors remain to be examined.
My view is that there is an important third class of criteria which
should be invoked both in descriptive and normative considerations of
theory choice in economics:
3. Ontological criteria. These may be related to various conceptions
about the basic constituents of social reality, their causal capacities,
relations of causal and other kinds of dependence between them, and
mechanisms of change among them. More particularly, these criteria
may be based on some more or less fundamental and more or less
well-articulated visions concerning human beings in regard to notions
of rationality, and the social arrangements of their lives in regard to
individualism, collectivism, as well as conceptions of social causation
and evolution. What all such ideas share is that consistency with
conceptions of the structure and functioning of the world serves as a
criterion of, or a constraint on, theory choice.
 The way the world works 371

I believe there is no doubt that, as a matter of actual fact, ontological

criteria play an important role in constraining actual theory choices
made by economists. I also believe they should play such a role. Given
these two claims, a third one is suggested: too little systematic analysis
has been done on the ontological grounds of theorizing by economic
methodologists. Some interesting work has been done on the ontology of
economics, though. Among the examples are the argument from the folk
psychological foundations of microeconomics (Rosenberg 1992); work
on causal powers and tendencies in the economy (Cartwright 1989; MaÈki
1990b, 1992c; Lawson 1997); work on causal processes and mechanisms
(MaÈki 1992c, 1998a; Salmon 1998); study on Dennettian real patterns
(Ross 1995); and a study on universals in economics (MaÈki 1997). One
may also translate notions such as ``physics envy'' and ``gender bias'' into
ontological claims. Few of the works in the above list explicitly construe
ontological notions as criteria of theory choice or as constraints on
theorizing. This is where the focus of the present essay lies (this draws
upon earlier work with a similar focus: MaÈki 1998d; 1998e; see also MaÈki
1992c).
I will introduce a generic ontological notion, that of the way the world
works, or www for short. My descriptive claim is that many economists
invoke this idea with an intention of providing a constraint on theorizing.
I will call this ontological constraint the www constraint. Economists
invoke this constraint either implicitly or explicitly. I am also inclined to
hold the normative claim that economists are on the right track when
they, at least in some important instances, invoke such a constraint. My
arguments in this paper are for the descriptive claim.
Let us suppose that it is of the essence of science that scientists pursue
an understanding of the world. Let us also suppose that understanding
the world amounts to understanding how the world works, the way the
world functions. Let us further suppose that at least to a large extent, the
world's workings are a matter of causal processes being in place. Let us
®nally suppose that an understanding of the world is sought by means of
theories. All this means that scienti®c theories are ± and are to be ±
constructed and employed with the purpose of depicting the causal
processes that constitute the ways the world works. One of the prominent
recent philosophical statements of this time-honored general idea has
been put forth by Wesley Salmon: ``To understand the world and what
goes on in it, we must expose its inner workings. To the extent that
causal mechanisms operate, they explain how the world works'' (Salmon
1984, p. 133). Explanatory understanding is dependent on description, or
theoretical redescription: the events and their co-occurrences and se-
quences, as well as the entities involved in them, are redescribed in terms
372 Methodological implications of economic ontology

of theory as what they are believed to be, namely as manifestations of

``underlying'' entities and processes. Thus answers to why? questions are
dependent on answers to how? questions (see MaÈki 1990c, 1992c). In
Salmon's words: ``Causal processes, causal interactions, and causal laws
provide the mechanisms by which the world works; to understand why
certain things happen, we need to see how they are produced by these
mechanisms'' (Salmon 1984, p. 132). There is a sense in which adequate
explanations are adequate descriptions of goings-on in the world.
Here is how the www constraint often works: if it is the case that we
understand the world by way of theories which describe its causal
functioning, then any doubt about the capacity of a given theory to
expose the major elements of such causal functioning amounts to a doubt
about its capacity to render parts of the world understandable. Such
doubts may lead to various responses, ranging from outright rejection to
relative lack of attention. In such cases, the constraint functions nega-
tively: beliefs about its non-satisfaction serve as grounds for exclusion.
Even though this negative role of the constraint in providing grounds for
exclusion may be particularly important, this does not rule out its
positive role in providing supportive grounds for a theory. The general
point is that the www constraint serves a function in determining the
merits and demerits of a theory.
We will next look at an example of the negative function of the
constraint. The model of perfect competition will serve as our case. While
this model enjoys an established status in most textbooks of economics,
there are economists who believe the model to be fatally misguided as a
picture of the competitive market economy. We will see how skepticism
about the model of perfect competition as expressed by certain econo-
mists is based on invoking the www constraint. These economists have
doubts about the non-satisfaction of the constraint, and these doubts
serve as grounds for exclusion. Their reasoning may be put in terms of
unrealisticness: all models are unrealistic; this model is unrealistic in a
wrong way.

2 The way the world won't work: perfect competition

The purpose of the present section is not to add anything novel to some
of the customary complaints about the model of perfect competition, nor
to try to persuade the advocates that the criticisms are sound. My
purpose is rather to show that some of these criticisms amount to what I
have set out to argue in this paper: some such criticisms are a matter of
 The way the world works 373

(1) invoking an ontological www constraint on economic theorizing, and

to (2) arguing that the perfect competition model does not meet this
constraint. I cite three economists ± George Richardson, Ronald Coase,
and James Buchanan ± as evidence in support of the two claims. Notice
also that I will not have to be very precise about what the target of such
criticisms is (nor is the textual evidence I employ very precise about this);
it is enough for my case to show that the www constraint is being invoked
in an argument against something.
In each case, the reasoning is the same, exhibiting the same pattern:

1. Acknowledge the unavoidability of excluding much in one's theory:

any theory or model is necessarily relatively narrow and simple.
2. Argue that the model of perfect competition is narrow in a wrong
way, in that it excludes items that should not be excluded.
3. Suggest that the grounds for the claim of faulty narrowness, and, more
generally, for judging what to include and what to exclude in one's
model, are based on one's conception of the way the world works.

Narrowness of theory

It is a standard criticism of a model or theory to blame it for excluding

something from consideration. This suggests a common form of unrealis-
ticness: a theory is unrealistic if it is unjusti®ably ``narrow'' ± that is, if it
excludes from consideration factors that are deemed important. This
notion is entertained also by Coase and Richardson. Both think that
conventional textbook theory is unrealistic in this sense: it excludes
factors that should be included in the theory. To put it in other words,
conventional neoclassical theory isolates from factors that should be
explicitly theorized; thus, the theory should be de-isolated so as to
incorporate these factors (see MaÈki 1992a, 1994a, 1994b).
However, and this is where Coase and Richardson part company, they
do not share the details of where to start de-isolating the theory, that is,
what precisely to include that was excluded from consideration, and in
what order. Any theory excludes an enormous number of elements in
reality. From this set of excluded elements, one may then only choose for
inclusion a small subset comprising those that are found most important.
Coase and Richardson choose somewhat differently. For Coase, the
most important element to be included is transaction costs; for Richard-
son it is the process of information acquisition that will ensure the
satisfaction of what he calls the ``informational requirements.''
374 Methodological implications of economic ontology

For Coase, the incorporation of positive transaction costs required the

relaxation of the assumption of zero transaction costs, which ``is, of
course, a very unrealistic assumption'' (Coase 1960, p. 15). By relaxing
this assumption, the theory could be broadened by way of theoretical de-
isolation. The item thereby incorporated is what he calls ``the missing
element'' in economic theory (Coase 1993c, p. 62). This is not the only
missing item, however: ``No doubt other factors should also be added''
(Coase 1988b, p. 30).
Richardson stresses his main theme by saying that economists ``should
take account of informational considerations that have generally been
neglected'' and then goes on by suggesting that ``there are yet other
relevant considerations which cannot be brought into formal analysis''
(Richardson 1998, p. 21). There is a lot that has been neglected, but not
all of these factors can be incorporated into formal theory. ``Economic
theory is an indispensable instrument of analysis, but effective only when
we are aware of its limitations'' (p. 21). Yet, much that has been
excluded, can and should be included. Among them are product and
process innovations. Richardson blames standard theory for ``[c]hoosing
to set aside, or assume away, the fact that products and processes are
subject to continuous development'' (p. 3). This is not an innocent
exclusion, since ``the habit of abstracting from product development, the
pace of which has been accelerating, may have led economists to view the
real world as much less competitive than business men know it to be''
(p. 14).
Coase is explicit that some of the excluded factors are those that
characterize the internal organization of the business ®rm, thus leading
to a notion of the ®rm as a black box:1
The concentration on the determination of prices has led to a narrowing of focus
which has had as a result the neglect of other aspects of the economic system.
[. . .] What happens in between the purchase of the factors of production and the
sale of the goods that are produced by these factors is largely ignored. [. . .] The
®rm in mainstream economic theory has often been described as a ``black box.''
And so it is. This is very extraordinary given that most resources in a modern
economic system are employed within ®rms, with how these resources are used
dependent on administrative decisions and not directly on the operation of a
market. Consequently, the ef®ciency of the economic system depends to a very
considerable extent on how these organizations conduct their affairs, particularly,
of course, the modern corporation. Even more surprising, given their interest in

1 The exclusion of internal characteristics of the objects under study (such as the internal
organization of business ®rms) amounts to ``internal isolation'' in contrast to ``external
isolation'' which is a matter of excluding characteristics of the system surrounding the
object (such as other markets in partial equilibrium analysis) (see MaÈki 1992a for details).
 The way the world works 375

the pricing system, is the neglect of the market or more speci®cally the
institutional arrangements which govern the process of exchange. (Coase 1993
[1992], p. 229)
Richardson shares the belief that price theory gives an overly narrow if
not distorted picture of economic reality; he also thinks that its promi-
nence may be based on the mechanical analogy borrowed from physics.
This suggests that conventional price theory is set up to conform to a
wrong ontological constraint:
if we are led to study the informational aspects of social systems only in terms of
a rigid conceptual framework borrowed from physics, we shall certainly obtain a
distorted picture. Prices, for example, and particularly the current values of
prices, assume from this point of view an undeserved prominence, because it is in
terms of them that a quasi-physical signalling mechanism can be elaborated and
given mathematical expression. (Richardson 1960, p. 41)
Coase puts some of the blame on the conception of economics as a
theory of choice which has contributed to the exclusion of the human
and institutional ``substance'' of the economy from theoretical considera-
tion. The resulting picture is ontologically suspect:
This preoccupation of economists with the logic of choice [. . .] has nonetheless
had, in my view, serious adverse effects on economics itself. One result of this
divorce of the theory from its subject matter has been that the entities whose
decisions economists are engaged in analyzing have not been made the subject of
study and in consequence lack any substance. The consumer is not a human
being but a consistent set of preferences. The ®rm to an economist [. . .] ``is
effectively de®ned as a cost curve and a demand curve [. . .]''. Exchange takes
place without any speci®cation of its institutional setting. We have consumers
without humanity, ®rms without organization, and even exchange without
markets. (Coase 1988b, p. 3)
No doubt economists have an amazing variety of grounds for thinking of
a given theory as unrealistic, but a major one is certainly the perception
of the theory as excessively narrow or partial or isolative. As we have
seen, this is also one of Coase's and Richardson's critical perceptions of
standard neoclassical theory. They insist on the de-isolation of the
theoretical picture of the economy by incorporating neglected elements
into the theory.2 But we have not said enough; our picture of their views
on narrowness is still too narrow.

2 More precisely, we are here talking about ``horizontal de-isolation'' ± de-isolation at a

given level of abstraction or vertical isolation (MaÈki 1994a). In contrast to Coase's quest
for horizontal de-isolation, he also insists on vertical de-isolation, that is, lowering the
level of abstraction by engaging oneself in empirical case studies (MaÈki 1998b).
376 Methodological implications of economic ontology

The missing ``imperfections'': Richardson

Having listed some successes of the market system in adjusting to

changes, and before providing an exposition of aspects of general
equilibrium theory to account for these successes, Kenneth Arrow makes
this claim: ``All these phenomena show that by and large and in the long
view of history, the economic system adjusts with a considerable degree
of smoothness and indeed of rationality to changes in the fundamental
facts within which it operates'' (Arrow 1974, p. 254). Richardson agrees
that the market system has the capacity of coordinating activities and
adjusting to changes. However, referring to the theory of general
equilibrium, Richardson makes a somewhat blunt statement:
We therefore need to ®nd a better explanation of how, and in what conditions,
coordination through market transactions, which we know from experience can
work, does in fact work. Men stir up a dust . . . and then complain that they
cannot see. And it is indeed through the creation of a model with perfect
competition, perfect mobility (and, sometimes, perfect knowledge, whatever that
means) that we obscure the actual process of adjustment. (Richardson 1995,
p. 1492)
``Men stir up a dust and then complain that they cannot see.'' This
sounds like a rather harsh judgement about economists holding certain
theories. In what follows, we try to understand what Richardson means
by the expressions ``stirring up a dust'' and ``cannot see.'' This much
should be clear to begin with: one does not ``stir up a dust'' by making
just any unrealistic assumptions. The very ®rst page of Richardson's
Information and Investment hints at the possibility of harmless unrealistic-
ness in the case of the perfect competition model:
The conditions which de®ne it, as everyone knows, are rarely, if ever, character-
istic of the real world, but it can be argued that this divergence represents no
more than the normal degree of abstraction associated with general theoretical
models. (Richardson 1960, p. 1)
Coase is more explicit about what makes unrealisticness justi®ed, or, in
Richardson's words, what determines ``the normal degree of abstraction'':
. . . our assumptions should not be completely realistic. There are factors we
leave out because we do not know how to handle them. There are others we
exclude because we do not feel the bene®ts of a more complete theory would be
worth the costs involved in including them. [. . .] Again, assumptions about other
factors do not need to be realistic because they are completely irrelevant. (Coase
1988c, p. 66)
In other words, Coase is comfortable with a theory being isolative in
 The way the world works 377

general, provided there are good reasons for this such as the ones he cites
above.3 However, there are limits to unrealisticness: beyond these limits,
one would ``stir up a dust.'' There are some items that just cannot be
excluded without obscuring the view: if you exclude them, you ``cannot
see.'' This is the fundamental message endorsed by Richardson and
Coase. They also hold a shared idea of what determines these limits.
These limits, they believe, are determined by our conception of the way
the world works. They not only believe that the conventional picture is
narrow, but that it is harmfully narrow in unduly excluding important
factors. The question is to ask, what is it that makes those factors
important? Why are they claimed to be unduly excluded? Why should
precisely they, and not some other factors, be included? In order to
answer these questions something else is needed. In the case of Coase and
Richardson this something else amounts to the invocation of the
ontological constraint. Plenty of documentation can be provided in
support of this suggestion.
The subtitle of Richardson's main work, Information and Investment, is
accurate: A Study in the Working of the Competitive Economy. Indeed, it
is a recurring theme of the book that it is the task of economics to give an
account of how the economy ``works.'' Richardson's persistent critique
of the perfect competition model is telling. The second page of Infor-
mation and Investment summarizes his critique of the model precisely for
failing to account for the world's workings:
Perfect competition, I shall af®rm, represents a system in which entrepreneurs
would be unable to obtain the minimum necessary information; for this reason, it
cannot serve as a model of the working of actual competitive economies.
(Richardson 1960, p. 2; emphases added)
Here is the claim put in more speci®c terms:
I feel convinced that one of the essential elements of any adequate account of the
attainment of equilibrium has not been provided; for the most part, indeed, the
need for it has been ignored. No explanation has been given of how, in the
conditions which de®ne the perfectly competitive model, entrepreneurs could
obtain the information on which their expectations, and therefore the investment
plans required for equilibrium, would have to be based. (Richardson 1960, pp.
23±4; emphases added)
Note that Richardson says that considerations of information are
supposed to provide ``one of the essential elements'' involved in the
process of ``the attainment of equilibrium'' and that the model under

3 For arguments that show why an economist espousing realism as a theory of theories is
entitled or even required to use unrealistic assumptions, see MaÈki (1994a, 1994b).
378 Methodological implications of economic ontology

criticism fails to incorporate this essential element. Richardson puts the

idea explicitly in terms of the working of the world:
Whatever insights the theory of perfect competition may have given us, it fails to
provide . . . a convincing account of how free enterprise economies in fact work.
. . . And this failure results from its neglect of what we may call the informational
requirements of any workable system, the conditions that must be realized, that
is, in order that those within it can have suf®cient knowledge to take investment
decisions. (Richardson 1998, p. 2)
He goes on by explaining why he thinks the model of perfect competition
fails in accounting for the way the system works ± why its advocates have
``stirred up a dust'':
By abstracting from some of the circumstances of real economic life, we have,
paradoxically, made it more dif®cult, if not impossible, to explain how the system
works. I have argued elsewhere that markets generally operate not despite, but
because of, some ``imperfection'' of competition, because of the existence, that is,
of circumstances which, although excluded by de®nition from the perfect
competition model, full®l the informational requirements of the system by
endowing the business environment with a high degree of stability suf®cient to
make informed investment decisions possible. (Richardson 1998, p. 2)
The ``imperfections'' that Richardson refers to ± in his 1960 book he
called them ``restraints'' ± are institutional features of economies, in-
cluding forms of information sharing such as price agreements, vertical
integration, and signaling, as well as reputation and trust. Such ``im-
perfections'' reduce the cost of information and constitute commitments,
and thereby facilitate the coordination of competitive and complemen-
tary investments. The ``imperfections'' are among the essential elements
in the way the world works. To understand the workings of the economic
world one has to understand such institutional features. Such under-
standing is not provided by the model of perfect competition:
Coordination through market transactions is only possible . . . by virtue of the
existence, in the real world, of circumstances which set bounds to what can
happen. And these enabling circumstances, which exist naturally (such as differ-
ential capabilities) or which may be contrived through collective action (such as
market sharing arrangements), all represent deviations from the pure competition
model. (Richardson 1995, p. 1492; emphases added)4

4 ``Perfect competition earned its reputation as an ideal market structure because of the
belief that, to the extent that the conditions de®ning it were realized, resources would be
allocated so as to exhaust all pro®t opportunities. No one, in my view, ever provided a
fully satisfactory explanation of how this was to come about . . . the conditions
favourable to successful adjustment are not those laid down in the perfectly competitive
model'' (Richardson 1964, pp. 160, 161).
 The way the world works 379

Richardson argues that to understand (even the possibility of ) successful

adjustment through market transactions, we need to incorporate various
institutional features into our picture of the economy. This can be done
by relaxing assumptions such as those of perfect information, atomistic
®rms, and homogeneous goods, and of paying attention to the ways in
which the ``informational requirement'' is met. In the imperfect infor-
mation world, institutions matter. There is a parallel logical structure in
Coase's account. By relaxing the zero transaction cost assumption, that
is, by incorporating positive transaction costs into his account Coase was
not only able but also forced to incorporate institutional features that
were previously neglected from systematic analysis. In the positive
transaction cost world, institutions matter.
We cannot yet pretend to fully understand the notion of the world's
working as it appears in Richardson. What does it mean, precisely? Why
does one theory fail and why does another theory succeed as a potentially
adequate representation of the way the world works? A complete account
would be impossible to pursue here, but one observation can be provided.
An element of ontic necessity ± necessity de re rather than just de dicto ±
appears to involved. Richardson puts the idea variously, including
references to ``conceivable'' systems and to some elements being ``essen-
tial'' or ``necessary'' for the functioning of the system. Here is an example:
Irreducible uncertainty, as a factor in any conceivable economic system, owes its
existence, in part to incomplete information about preferences and production
functions. In much of economic theory, this incompleteness is ignored . . . But
where the object is to study the working of a competitive economy, the question
of the availability of information cannot thus be pushed aside. (Richardson 1960,
p. 81; emphases added)
He also says that ``some market imperfections may be essential to the
process of successful economic adjustment'' (p. 38; emphasis added) and
that ``the conditions which de®ne the system of perfect competition are
not such as would permit the economic adjustments required'' (p. 10;
emphasis added). As Brian Loasby puts it, Richardson's ``conclusion is
that in perfecting the model of perfectly competitive equilibrium, econo-
mists have re®ned away the essential mechanism'' (Loasby 1986, p. 152).
Here is yet another way of formulating the idea:
By assuming, overtly or tacitly, that [the optimum strength for these restraints] is
zero, and therefore by neglecting the whole problem of information, the perfect
competition model condemns itself not only to unrealism but to inadequacy even
as a hypothetical system. (Richardson 1960, p. 69; emphasis added)
The following may be taken as an explanation of what Richardson
means by ``inadequacy even as a hypothetical system'':
380 Methodological implications of economic ontology

It is no defence to appeal, moreover, to the analogy of mechanical statics which,

though neglecting friction, can still identify the equilibrium position of a system
of forces, for we cannot demonstrate that economic systems have such positions
of rest without reference to expectations and information which could not be
presumed to be available in the absence of restraints. (Richardson 1960, p. 69;
emphasis added)
Thus, ``inadequacy even as a hypothetical system'' can be taken to mean
something like ``ontic impossibility''. This can then be taken to imply
that there are some ontically necessary features that a real system has to
possess in order for its theoretical picture to count as adequate ``even as a
hypothetical system.'' This suggests that it is in the character of hypo-
thetical pictures that they represent systems that are at least possible ±
even though perhaps not fully actual. One further piece of evidence
appears to support this reading. Ontic necessities and impossibilities have
entailments regarding empirical actualities. The following can be read in
this light:
It is not by accident that the markets in the real [actual] world that are closest to
pure competition, namely commodity markets, are usually subject to some kind
of regulation, or to disarray or, very commonly, to both. There is absolutely no
presumption that either the presence of a large number of competitors, or the
homogeneity of the products they offer, are, as the theory of perfect competition
would lead us to suppose, conditions favourable to ef®cient allocation. (Ri-
chardson 1995, pp. 1492±3)
To sum up, there seem to be two ontically necessary connections
envisaged by Richardson. One is that information is necessary for
adjustment. The other is that ``restraints'' are necessary for information.
And since these restraints appear in the form of institutions ± ``customs,
conventions, and the laws'' (1960, p. 69) ± this implies that a theory
about the working of the competitive economy has to incorporate
institutions. This is a theoretical necessity suggested by the two ontic
necessities.

Missing category of costs, missing institutions: Coase

Consider then Coase's version of the idea of the world's workings. We

have already cited Coase's complaint about economics which theorizes
``consumers without humanity, ®rms without organization, and even
exchange without markets'' (Coase 1988b, p. 3). Let us try to see what
grounds Coase has for this complaint. Such grounds are far from
intuitively obvious given that there are highly respected scienti®c theories
 The way the world works 381

that study planets without extension, planes without friction, and

molecules without color. Why are some exclusions suspicious while some
others are not?
Coase explains that it took him a long time to realize that ``the whole
of economic theory would be transformed by incorporating transaction
costs into the analysis'' (Coase 1993c, p. 62). The somewhat revo-
lutionary tone in this judgement can only be understood as re¯ecting the
idea that transaction costs constitute a major factor in economic reality
and that its inclusion in theory has major consequences for economics.
But what does ``major'' mean here? It does not just designate the idea
that here we have another ``factor'' which has a large impact on economic
phenomena and therefore had better be included in explanations. More is
involved than just causal relevance, namely what we might call ontic
indispensability: the element to be added is not just causally in¯uential, it
is necessary for the functioning of the system. The introduction of
positive transaction costs in the theory brings with it new kinds of
entities, namely institutions, such as legal rules and contractual struc-
tures. If it is held that such institutions play both an indispensable and a
powerful role in the functioning of the economy, it would be inexcusable
to exclude them from the analysis.
This suggestion is based on a central idea that occurs frequently in
Coase's writings, namely that it is the task of economic theory to provide
``insight into how the system works'' (Coase 1988c, p. 64). He argues
that ``realism [i.e. realisticness] of assumptions is needed if our theories
are ever to help us understand why the system works in the way it does''
(p. 65). Of course, by this he must mean realisticness subject to the
ontological constraint ± and this leaves room for plenty of legitimate
unrealisticness.
The idea that there is a way in which ``the economic system works''
suggests that the complaint about a ``missing element'' in theory is in
effect a complaint about an indispensable missing link in the working of
the real system; without this link, the system would not function as it
does ± or would not function at all. Therefore, this link has to be
theorized in order to understand how the system works. Coase's view
here is akin to Richardson's, amounting to the suggestion that some sort
of necessity is involved: ``The solution was to realize that there were costs
of making transactions in a market economy and that it was necessary to
incorporate them in the analysis'' (Coase 1993a, p. 46; emphasis added).
In this somewhat essentialist spirit, he says that the point of theorizing is
``to get to the essence of what [is] going on in the economic system''
(Coase 1988c, p. 68).
382 Methodological implications of economic ontology

The missing endogenous process: Buchanan

Our ®nal witness is James Buchanan. He puts forth a familiar complaint:

The basic ¯aw in this model of perfect competition is not its lack of correspon-
dence with observed reality; no model of predictive value exhibits this. Its ¯aw
lies in its conversion of individual choice behavior from a social-institutional
context to a physical-computational one . . . but surely this is nonsensical social
science, and the institutionalist critics have been broadly on target in some of
their attacks. (Buchanan 1979, p. 29)
Buchanan then goes on to put forward his favored position; it is in terms
of endogenous process. Here is a lengthy passage:
A market is not competitive by assumption or by construction. A market
becomes competitive, and competitive rules come to be established as institutions
emerge to place limits on individual behavior patterns. It is this becoming process,
brought about by the continuous pressure of human behavior in exchange, that is
the central part of our discipline, if we have one, not the dry rot of postulated
perfection. A solution to a general-equilibrium set of equations is not predeter-
mined by exogenously determined rules. A general solution, if there is one,
emerges as a result of a whole network of evolving exchanges, bargains, trades,
side payments, agreements, contracts which, ®nally at some point, ceases to
renew itself. At each stage in this evolution toward solution there are gains to be
made, there are exchanges possible, and this being true, the direction of move-
ment is modi®ed. (Buchanan 1979, p. 29)
Buchanan concludes that it is ``for these reasons that the model of perfect
competition is of such limited explanatory value except when changes in
variables exogenous to the system are introduced. There is no place in
the structure of the model for internal change, change that is brought
about by the men who continue to be haunted by the Smithean
propensity [to truck, barter, and exchange]'' (Buchanan 1979, pp.
29±30). We may paraphrase these remarks by saying that, according to
Buchanan, the model of perfect competition fails to meet an important
www constraint, namely it fails to depict the way the competitive
economy works in terms of endogenous causal process.

4 Concluding remarks

I have used a class of criticisms against the model of perfect competition

as evidence in support of the descriptive claim that at least some
economists invoke ontological constraints in their arguments for or
 The way the world works 383

against particular theories or lines of theorizing. This limited evidence

does not yet give us much of an idea about the proportion, weight, and
likely context of ontological arguments in the argumentative practices of
economists. It remains a task for empirical investigation to determine
such matters. For example, it is obvious that the ontological arguments
in relation to the model of perfect competition far from exhaust all the
relevant and popular arguments there are about this model. Indeed,
many economists are unimpressed by such arguments and justify their
use of the model in non-ontological terms. To them, the model was never
intended as an account of the way the competitive economy works. And
they often say this explicitly so as to discourage others to assess the
model by what they regard as wrong standards. Does this mean that
these economists do not have conceptions of the www, or more precisely,
consequential conceptions of the www? Not at all, on the contrary. The
very fact that they often explicitly speak out their view that the perfect
competition model is not, and is not intended as, a description of the way
the competitive economy works implies that they do have consequential
views of the workings of the world.
All theories are bound to be unrealistic in the trivial sense of excluding
much, in being isolative or narrow. All economists think that there are
limits to narrowness. Many economists think that the appropriate
isolations have to meet an ontological constraint provided by concep-
tions of the way the economic system works. Some economists believe
that the model of perfect competition does not meet the constraint, and
therefore cannot be taken as an adequate representation of the core or
essence of the competitive economy ± not even as a hypothetical
possibility. The need for de-isolation ± that is, the need for incorporating
institutional ``imperfections'' ± is ontologically grounded. The idea is not
the simple one that these ``imperfections'' are real and causally in¯uen-
tial, and therefore should be included ± for example in order to improve
the ®t of the predictive implications of the model. The idea is rather the
stronger one that the imperfections play a necessary or essential role in
the working of the world, and that therefore they should play an
indispensable role in theory.
No complete account of the workings of the www constraint can be
attempted here. I will just list a few features that would appear to be
characteristic about it. It is the task of future research to revise and re®ne
these suggestions.

1. The www constraint is an ontological constraint on theory choice. It is

to the merit of a theory to meet the constraint, and to its demerit to fail
to meet it. Good theories are believed to depict the way the world works
384 Methodological implications of economic ontology

and thereby to make the world understandable to us. Reasons to think

that a given theory has failed to conform to the constraint are reasons to
think that the theory will not help us understand. Economists have
beliefs about the www, and these beliefs serve as, or give rise to, reasons
to think of a theory favorably or unfavorably. Not all economists invoke
such reasons all the time, but some economists invoke them some of the
time.
2. The sources of such reasons may range across a broad spectrum of
ideas, such as the contents of the theory itself, other scienti®c theories,
metaphysical theories, everyday experience, commonsense accounts, and
systematically gathered empirical evidence. The www constraint on
theories draws upon such sources which may be either explicitly acknowl-
edged or implicitly presupposed. These sources will constitute something
like a hierarchy. On the top, there may be some general metaphysical
visions such as a generic causal process theory (think of Einstein's
determinism and the locality principle); below it, there may be alternative
more speci®c metaphysical theories such as versions of the generic causal
process theory at various levels of speci®city (think of Salmon's formula-
tions of the notion of causal process close to the highest such level; and
Richardson's ideas about competition close to the lowest level). These
may set boundaries within which other sources selectively take over ±
such as other scienti®c theories, commonsense experience, previous
research, and so on. These ``lower-level'' sources may, in turn, feed back
to the higher-level sources of ontological beliefs.
3. Theorizing involves making choices, including choices between
theories. Every theory candidate includes ``unrealistic'' elements. It is
therefore never going to be a sound criticism of any given theory to claim
that the theory is unrealistic. However, supposing that all theories are
not equally bad, or equally good, it must be that some elements of
unrealisticness are all right, while some others are less so. The problem of
theory choice is also a problem of choosing between good and bad
unrealistic elements. The suggestion pursued here is that the www
constraint is there to help scientists resolve this choice problem. Putting
the thought in terms of truth, we may say the following. Falsehood is not
suf®cient for making a theory bad. Truth is not suf®cient for making a
theory good. Yet, truth and falsehood should matter. For them to
matter, we have to be able to distinguish between harmful and harmless
falsehoods on the one hand, and between signi®cant and insigni®cant
truths on the other. Good theories avoid harmful falsehoods and contain
signi®cant truths. These distinctions are based on the www constraint.
Good theories are true about the www and can survive falsehoods that
 The way the world works 385

are not big enough to stop the world (see MaÈki 1998e). In our example,
Richardson believes that the assumptions of very large numbers, ato-
mistic ®rms (price signals in the market being the only vehicle of contact
between them), and perfect information, are harmful falsehoods: under
these assumptions, the competitive economy cannot possibly work.

4. The www constraint is characteristically not able to constrain the

theory choice so that only one unique optimal theory option remains to
be ``chosen.'' Rather, the constraint typically functions as a weaker
exclusion device: it helps exclude theory candidates which depict the
world in such a way that we have reason to believe that the world does
not function that way, or, more strongly, that it cannot function that
way, or, still more strongly, that it cannot function at all, given what we
know about it. This is the thrust of the arguments against the model of
perfect competition that we have dealt with; they serve the purpose of
excluding the model but do not uniquely determine the best alternative
model. We may say that theory choices are underdetermined by the
ontological constraint just as they are underdetermined by empirical tests
and by various social criteria. We may also say that the three classes of
constraints ± empirical, social, and ontological ± may combine in various
permutations and with different weights of impact so as to reduce the
degree of overall arbitrariness in theory choice.
5. Among other things, two philosophical positions are involved in this
notion. One is ontic realism: the world and its constituent sub-systems
have a characteristic way of functioning. The solar system, the earth's
ecological system, the human organism, your computer ± they all have a
de®nite way of working. We may say that they function in the way they
do because they are what they are ± because they are made of certain
kind of stuff, because they have such and such constituents and such and
such a structure. We may also say, at least in some cases such as
Richardson's conceptions about the workings of the competitive
economy, that modal elements are involved: our conceptions of the www
suggest that there are ontic necessities and impossibilities in the world.
This view is ontic realism, since it characterizes a feature of reality. It is
ontic realism, since this feature ± the way the world works (or the way it
necessarily works or cannot possibly work) ± is supposed to be a
characteristic the world has independently of what we believe of it or
how we represent it.

6. The other philosophical position is theoretical realism: good theories

are purportedly true descriptions of the way the world works. This runs
counter to instrumentalist conceptions of theories according to which
386 Methodological implications of economic ontology

theories are not attempts to capture the inner workings of the world, but
rather ± and no more than ± instruments of organizing data or of
attaining some practical goals. The realist will allow for unrealistic
assumptions, but will hold that it is inexcusable not to be realistic about
those elements that are necessary for the functioning of a given system,
while it is permissible and advisable to leave out some others that are
not.
In his critique of Coase's methodology, Posner put forth an instrumen-
talist view, namely that ``[a] model can be a useful tool of discovery even
if it is unrealistic, just as Ptolemy's astronomical theory was a useful tool
of navigation [. . .] even though its basic premise was false'' (Posner 1993,
p. 77). Given Coase's interest in how the system under study functions, he
would not be content with Ptolemyan theory. He would prefer Coper-
nican heliocentrism to the false geocentrism, even if the Copernican
theory has many minor details wrong and may therefore fail in predic-
tions. ``Faced with a choice between a theory which predicts well but
gives us little insight into how the system works and one which gives us
this insight but predicts badly, I would choose the latter [. . .]'' (Coase
1988c, p. 64). Coase's rejection of the instrumentalist conception of
theories is explicit: ``But a theory is not like an airline or bus timetable.
We are not interested simply in the accuracy of its predictions. A theory
also serves as a base of thinking. It helps us to understand what is going
on [. . .]'' (p. 64). While bus timetables may help us predict the behavior
of buses and thus to serve as ``inference tickets,'' they fail to give us any
idea about the mechanisms and processes that keep buses running as they
do.5 As Salmon puts it, ``[a] detailed knowledge of the mechanisms may
not be required for successful prediction; it is indispensable to the
attainment of genuine scienti®c understanding'' (Salmon 1984, p. 133).

References
Arrow, Kenneth J. (1974) ``General Economic Equilibrium: Purpose, Analytic
Techniques, Collective Choice,'' American Economic Review, 64: 253±72.
Blaug, Mark (1980) The Methodology of Economics, Cambridge: Cambridge
University Press.
Buchanan, James (1979) What Should Economists Do? Indianapolis: Liberty
Press.
Cartwright, Nancy (1989) Nature's Capacities and their Measurement, Oxford:
Clarendon Press.

5 For a critique of Posner's critique of Coase, see MaÈki (1998c).

 The way the world works 387

Coase, R. H. (1937) ``The Nature of the Firm,'' Economica, 4 (n.s.): 386±405.

(1960) ```The Problem of Social Cost,'' Journal of Law and Economics, 3: 1±44.
(1988 [1946]) ``The Marginal Cost Controversy,'' Economica, 13 (n.s.). Refer-
ences are to the reprint in Coase (1988a).
(1988a) The Firm, the Market, and the Law, Chicago: University of Chicago
Press.
(1988b) ``The Firm, the Market, and the Law,'' in Coase (1988a): 1±31.
(1988c) ``How Should Economists Choose?'' in Ideas, Their Origins, and Their
Consequences, Washington DC: American Enterprise Institute: 63±79.
(1993 [1992]) ``The Institutional Structure of Production,'' American Economic
Review, 82. References are to the reprint in Williamson and Winter (1993):
227±35.
(1993a) ``The Nature of the Firm: Origin,'' in Williamson and Winter (1993):
34±47.
(1993b) ``The Nature of the Firm: Meaning,'' in Williamson and Winter (1993):
48±60.
(1993c) ``The Nature of the Firm: In¯uence,'' in Williamson and Winter (1993):
61±74.
de Marchi, Neil (ed.) (1988) The Popperian Legacy in Economics, Cambridge:
Cambridge University Press.
(1992) Post-Popperian Methodology of Economics, Boston: Kluwer.
de Marchi, Neil and Mark Blaug (eds.) (1991) Appraising Economic Theories,
Aldershot: Edward Elgar.
Ferber, Marianne and Julie Nelson (1993) Beyond Economic Man: Feminist
Theory and Economics, Chicago: University of Chicago Press.
Hands, Wade (1994) ``The Sociology of Scienti®c Knowledge,'' in New Develop-
ments in Economic Methodology, ed. Roger Backhouse, London: Routledge:
75±106.
Klamer, Arjo (1983) Conversations with Economists, Totowa: Rowman and
Allenheld.
Latsis, Spiro (ed.) (1976) Method and Appraisal in Economics, Cambridge:
Cambridge University Press.
Lawson, Tony (1997) Economics and Reality, London: Routledge.
Loasby, Brian (1986) ``Competition and Imperfect Knowledge: The Contribution
of G. B. Richardson,'' Scottish Journal of Political Economy, 33: 145±58.
MaÈki, Uskali (1989) ``On the Problem of Realism in Economics,'' Ricerche
Economiche, 43: 176±98. Reprinted in The Philosophy and Methodology of
Economics, ed. Bruce Caldwell Aldershot: Edward Elgar.
(1990a) ``Methodology of Economics: Complaints and Guidelines,'' Finnish
Economic Papers, 3: 77±84.
(1990b) ``Mengerian Economics in Realist Perspective,'' History of Political
Economy, Annual Supplement to vol. 22: 289±310.
(1990c) ``Scienti®c Realism and Austrian Explanation,'' Review of Political
Economy, 2: 310±44.
(1992a) ``On the Method of Isolation in Economics,'' Idealization IV: Intellig-
388 Methodological implications of economic ontology

ibility in Science, ed. Craig Dilworth, special issue of Poznan Studies in the
Philosophy of the Sciences and the Humanities, 26: 319±54.
(1992b) ``Friedman and Realism,'' Research in the History of Economic
Thought and Methodology, 10: 171±95.
(1992c) ``The Market as an Isolated Causal Process: A Metaphysical Ground
for Realism,'' in Austrian Economics: Tensions and New Developments, ed.
Bruce Caldwell and Stephan Boehm, Kluwer Publishers: 35±59.
(1992d) ``Social Conditioning of Economics,'' in Post-Popperian Methodology
of Economics, ed. N. de Marchi, Kluwer Publishers: 65±104.
(1993a) ``Social Theories of Science and the Fate of Institutionalism in
Economics,'' in Rationality, Institutions and Economic Methodology, ed.
U. MaÈki, B. Gustafsson, and C. Knudsen, London: Routledge: 76±109.
(1994a) ``Isolation, Idealization and Truth in Economics,'' in Idealization in
Economics, ed. Bert Hamminga and Neil de Marchi, special issue of Poznan
Studies in the Philosophy of the Sciences and the Humanities, 38: 147±68.
(1994b) ``Reorienting the Assumptions Issue,'' in New Developments in
Economic Methodology, ed. Roger Backhouse, London: Routledge: 236±56.
(1996) ``Scienti®c Realism and Some Peculiarities of Economics,'' Boston
Studies in the Philosophy of Science, 169: 425±45.
(1997) ``Universals and the Methodenstreit: A Reexamination of Carl Menger's
Conception of Economics as an Exact Science,'' Studies in History and
Philosophy of Science, 28: 475±95.
(1998a) ``Mechanisms, Models, and Free Riders,'' in Economics and Method-
ology: Crossing the Boundaries, ed. R. Backhouse, D. Hausman, U. MaÈki,
and A. Salanti, London: Macmillan: 98±112.
(1998b) ``Is Coase a Realist?'' Philosophy of the Social Sciences, 28: 5±31.
(1998c) ``Against Posner Against Coase Against Theory,'' Cambridge Journal
of Economics, 22: 587±95.
(1998d) ``On the Issue of Realism in the Economics of Institutional Organiza-
tion: Themes from Coase and Richardson,'' in Beyond Keynes, ed. S. Dow
and J. Hillard.
(1998e) ``Aspects of Realism About Economics,'' Theoria, 13: 301±13.
(1999) ``Science as a Free Market: A Re¯exivity Test in an Economics of
Economics,'' Perspectives on Science, 7: 486±509.
McCloskey, Donald (1985) The Rhetoric of Economics, Madison: University of
Wisconsin Press.
Mirowski, Philip (1989) More Heat than Light, Cambridge: Cambridge Univer-
sity Press.
Posner, Richard A. (1993) ``The New Institutional Economics Meets Law and
Economics,'' Journal of Institutional and Theoretical Economics, 149:
73±87.
Richardson, G. B. (1960) Information and Investment. A Study in the Working of
the Competitive Economy, Oxford: Oxford University Press.
(1964) Economic Theory, London: Hutchinson.
(1972) ``The Organization of Industry,'' Economic Journal, 82: 883±96.
 The way the world works 389

(1995) ``The Theory of the Market Economy,'' Revue eÂconomique, 6: 1487±96.

(1998) ``Innovation, Equilibrium and Welfare,'' in Beyond Keynes, ed. S. Dow
and J. Hillard.
Rosenberg, Alex (1992) Economics ± Mathematical Politics or a Science of
Diminishing Returns, Chicago: University of Chicago Press.
Ross, Don (1995) ``Real Patterns and the Ontological Foundations of Micro-
economics,'' Economics and Philosophy, 11: 113±36.
Salmon, Pierre (1998) ``Free Riding as Mechanism,'' in Economics and Method-
ology: Crossing Boundaries, ed. R. Backhouse, D. Hausman, U. MaÈki, and
A. Salanti, London: Macmillan: 62±87.
Salmon, Wesley (1984) Scienti®c Explanation and the Causal Structure of the
World, Princeton: Princeton University Press.
Weintraub, E. Roy (1985) General Equilibrium Analysis, Cambridge: Cambridge
University Press.
(1991) Stabilizing Dynamics. Constructing Economic Knowledge, Cambridge:
Cambridge University Press.
Williamson, Oliver E. and Sidney G. Winter (eds.) (1993) The Nature of the Firm:
Origins, Evolution, and Development, Oxford: Oxford University Press.
Zamora Bonilla, Jesus P. (1999) ``Verisimilitude and the Strategy of Economic
Theory,'' Journal of Economic Methodology, 6: 331±50.
 Name index

Aigner, D. J. 63 Cartwright, Nancy 11, 13, 127, 131,

Alchian, A. A. 13, 143, 151, 190, 194, 244, 254, 255, 271, 276, 284, 292,
196±9, 201, 204, 221 299±301, 307, 314, 327, 331, 371,
Alesena, A. 260, 271 386
Anand, P. 133, 143, 151 Casson, M. 109, 113
Aoki, M. 109, 113 Chen, Ping 189, 222
Aristotle 33±5, 41±45, 49±52, 159, Clark, Andy 198, 221
160±5 Coase, R. H. 373±7, 379, 380±1, 386,
Arrow, Kenneth 30, 59, 61, 66, 335, 387
349, 355, 376, 386 Collingwood, R. G. 5, 7, 14
Axelrod, Robert 83, 96 Cournot, A. 227, 229, 231, 238, 239,
244
Becker, G. S. 58, 59, 67, 84, 96, 120, Culler, J. 101, 113
130, 221 Curley, S. P. 133, 151
Begg, D. K. H. 339, 355 Cutler, T. 157
Bennett, Fred 8, 13 Cyert, R. M. 150, 151, 200, 201, 207,
Berkowitz, M. K. 265, 271 221
Bewley, T. 139, 149, 150, 151
Bhaskar, R. 299, 307 Darwin, Charles 175, 209
Binmore, Ken 212±15, 218, 221 Davidson, P. 139, 151
Blattberg, Robert 340, 353, 355 Davis, John 11, 13, 119, 120, 131
Blaug, Mark 227, 228, 244, 368, 370, Dawkins, Richard 212, 221
386, 387 Day, R. H. 63, 189, 222
Boland, L. 227, 244 DeCanio, S. 335, 356
Braybrooke, David 177 Deci, E. 112, 113
Brennan, H. G. 80, 81, 96 de Finnetti, B. 151
Buchanan, James 5, 6, 9, 14, 96, 307, de Marchi, N. 370, 387
373, 382, 386 Dennett, D. 251±3 passim, 271,
Burgess, D. 264, 271 371
Descartes, Rene 361±5, 367, 368
Caldwell, B. 165, 190, 221, 387, 388 Dore, R. 173
Campbell, D. T. 319, 331 Dosi, G. C. 204, 222

390
 Name index 391

Douglas, M. 108, 113 Haavelmo, T. 279

Dow, J. 148, 151 Hacking, Ian 138, 152, 231, 240, 244
Downs, Anthony 58, 59, 61, 97 Hahn, F. 231, 244
DupreÂ, John 11, 13, 247, 251, 252, 257, Halpern, P. J. 265, 271
261, 271, 331 Hammond, J. D. 197, 222
Hands, Wade 370, 387
Earman, J. 277 Hansen, Lars 343±5, 347, 353±5
Eells, Ellery 76, 97 passim, 356
Ellis, B. 231, 244 Hansen, R. G. 215, 222
Ellsberg, D. 145, 146, 148, 149, 151 Hardin, Russell 8, 13, 61, 64, 65, 69,
Elster, Jon 181 72, 74
Epstein, L. G. 133, 148, 151 Harding, S. 248, 271
Everett, M. J. 189, 222 Hargreaves Heap, Shaun 9, 13, 99,
113
Falkenburg, B. 337, 356 HarreÂ, R. 164
Ferber, M. 370, 387 Hart, Oliver 285±90
Ferejohn, John 89, 97, 198, 223 Hausman, Daniel 11, 13, 18, 127, 131,
Fine, Arthur 30 187, 298, 307
Finley, M. I. 53 Hayek, F. A. 165±8, 173, 219, 222,
Fishburn, P. C. 133, 140, 151 228, 229, 239, 244
Fodor, J. 250, 271 Hegel, G. W. F. 164, 169, 171, 256
Fogel, R. W. 93, 97 Heisenberg, W. 175
Foreman-Peck, J. 291 Hempel, C. G. 275
Foss, N. 219, 222 Hicks, J. R. 139, 152
Frank, R. E. 323, 331 Hindess, B. 157, 166±8
Frege, G. 118, 131 Hirschleifer, J. 152, 222
Frey, B. S. 112, 113 Hobbes, Thomas 42, 52, 60, 61, 63, 66,
Friedman, B. 335, 356 69, 74
Friedman, Michael 249, 271 Hodgson, G. 150, 152, 189, 222
Friedman, Milton 13, 143, 152, 186, Hogarth, R. M. 133, 152
190, 192, 194, 196±8, 199, 201, Hoover, Kevin D. 18, 30, 130, 131,
204, 205, 222, 244, 253, 271, 331, 227, 231, 238, 242, 243, 244, 259,
364, 366 271, 336, 356
Houthakker, H. S. 185
Gagnier, R. 322, 331 Hull, D. L. 206, 222
Galison, P. 315, 331 Hume, David 34, 69, 119, 209, 274,
GaÈrdenfors, P. 148, 152 284±5
Gauthier, David 77, 78, 97 Hutchinson, T. 307
Gilbert, M. 122, 131
Gilboa, I. 148, 152 Janssen, Maarten 226, 227, 245
Gilovich, T. 323, 331 Jenkins, G. 264, 265, 271
Glauber, R. R. 340, 356 Jones, L. 244, 263, 271
Granger, C. W. J. 345, 346, 356
Gray, J. 166 Kelsey, D. 133, 148, 152
Guzzardi, W. 337, 356 Keynes, John Maynard 34, 40, 41, 52,
392 Name index

Keynes, John Maynard (cont.) 229, 240, 242, 245, 270, 299, 307,
53, 133, 135±9, 146, 148±51 331, 336, 356, 370, 372, 374, 375,
passim, 152, 226, 227, 245 385, 387±8
Kincaid, Harold 11, 12, 192, 222 Mangum, S. L. 328, 331
Kim, Jaegwon 237±9, 245, 249, 251, Marcet, Albert 350, 351, 354, 355,
271, 331 357
Kim, Kyun 356 March, J. G. 150, 151, 200, 201, 207,
Kirman, A. P. 214, 221, 228, 245 221
Kirzner, Israel 4, 14 Marimon, R. 357
Kitcher, Philip 31 Marras, A. 249, 251, 272
Klamer, A. 339, 343, 346, 356, 370, Marshall, Alfred 4, 14, 37, 38, 40, 42,
387 203, 226, 245
Knight, F. H. 133±9 passim, 146, 147, Marx, Karl 34, 35, 38, 40, 41, 49, 52,
152 53, 69, 70, 71, 164, 165, 166, 173,
Koopmans, T. 204, 222 259
Kreps, D. M. 140, 143, 152 Maynard Smith, J. 210, 211, 223
Kripke, S. 118, 131, 163 Mayr, Ernst 223
Kyburg, H. E. 136, 141, 148, 152 McCloskey, D. 174, 370, 388
McGrattan, E. 352, 357
LaCasse, C. 251, 257, 269, 270, 272 Meikle, Scott 8, 12, 164
Lachmann, L. 228, 245 Mellor, H. 285
Lakatos, I. 177 Metcalfe, S. 190, 223
Langlois, R. N. 147, 152, 189, 222 Meyer, John 350, 353, 356
Latsis, S. 370, 387 Mill, J. S. 36, 42, 49, 66, 127, 279,
Lawson, T. 114, 127, 131, 132, 139, 293±7, 299, 301±6, 307
147, 152, 299, 300, 307, 308, 331, Mirowski, P. 178, 223, 331, 338, 340,
356, 371, 387 341, 357, 370, 387
LeRoy, S. F. 147, 153 Mises, Ludwig von 228, 239, 245
Lester, R. A. 195, 196, 222 Mongin, P. 126, 131
Levy, David 228, 238, 239, 245 Moore, John 285±90
Levy, I. 153 Morgan, M. S. 223, 290
Lewis, David 90, 92, 93, 97, 125, 283 Morgenstern, O. 103, 185, 192, 197,
Loasby, B. 379, 387 210
Locke, John 42, 60, 66, 160, 312 Morris, S. 146, 153
Lucas, Robert E. 139, 141, 153, 245, Muellbauer, John 30
227, 228, 245, 346, 347, 354, 355, Mukerji, S. 150, 153
356 Muth, J. F. 357
Luce, R. D. 210, 222 Myrdal, G. 247, 272
Nelson, Alan 11, 13, 14, 68, 123, 131,
Machlup, F. 195±7, 205, 217, 222, 246, 368, 370
272 Nelson, R. 131, 189, 190, 199±209
Mackie, J. L. 177 passim, 212, 217±21 passim, 222,
Madden, E. 164 223
Mailath, G. J. 212, 222 Noonan, H. 119, 131
MaÈki, Uskali 7, 10, 14, 114, 116, 127, Nowak, L. 241±3 passim, 245
131, 143, 165, 204, 219, 223, 226, Nozick, Robert 91, 97
 Name index 393

O'Neill, John 8, 13, 158, 164 Rossett, C. 272

Oppenheim, P. 249, 250, 272, 275, 309, Roth, A. 107, 113
331 Roulini, N. 259, 260, 272
Ordeshook, P. C. 58, 74 Ruben, D. H. 278
Runde, Jochen 8, 13, 114, 134, 136,
Pareto, W. 177 140, 150, 153
Par®t, Derek 117, 131 Russell, B. 34, 118, 131
Pesaran, M. H. 335, 357
Petit, P. 268, 271 Sachs, J. 259, 260, 272
Pettit, Philip 10, 11, 13, 77, 79, 80, 81, Sahlin, N. E. 148, 152
83, 93, 95, 96, 97, 112, 113 Salmon, P. 389
Pickering, A. 315, 331, 332, 357 Salmon, Wesley 371, 384, 389
Pines, D. 349, 355 Samuelson, Larry 215, 221
Plato 49, 88, 178 Samuelson, P. 184, 185, 245, 332
Platts, Mark 77, 97, 178 Samuelson, W. F. 222
Polanyi, K. 41, 164±5, 169, 205 Sandel, M. J. 72
Pool, R. 349, 358 Sargent, Thomas 335±8 passim
Popkin, S. L. 64, 74 Satz, Debra 89, 97, 198, 223
Posner, R. A. 386, 388 Savage, L. J. 103, 140, 153, 197, 205,
Putnam, H. 118, 131, 163, 249, 250, 222, 356
272, 309, 331 Savellos, E. 249, 272
Saviotti, P. P. 190, 223
Quine, W. V. O. 9, 174±6, 370 Schaffer, M. E. 215, 223
Schelling, T. 106, 107, 113, 125
Raiffa, H. 210, 222 Schick, Frederic 77, 97, 141, 153
Ramsey, F. P. 136, 140, 153 Schmeidler, D. 148, 153
Reddy, W. 329, 332 Schumpeter, J. A. 5, 14, 38
Regan, D. T. 323, 331 Scott, J. C. 62, 63, 64
Reuten, G. 307 Sen, Amartya 64, 77, 78, 97, 103, 113
Ricardo, D. 35, 38, 45, 46, 265, 266, Sent, Esther-Mirjam 7, 10, 14, 338,
267, 272, 303, 304 344, 348, 358
Richardson, G.B. 373±88 passim Shapere, Dudley 31
Riker, W. H. 58 Simon, Herbert A. 87, 90, 97, 200, 207,
Riley, J. G. 132, 146, 147, 152 209, 223
Rizvi, A. 213, 223 Sims, Christopher 343, 344, 353±5
Robbins, Lionel 4, 14, 52, 226, 245, passim, 358
272, 368 Smith, Adam 35, 37, 38, 45, 53, 209,
Roll, Richard 43, 340, 342, 353, 355, 322
358 Smith, B. 165
Rosenberg, Alexander 16±20, 175, Smith, K. R. 63
190±5 passim, 198, 199±201 Smith, Ron 335, 357
passim, 207, 209, 223, 237±40 Sober, Elliott 97, 293
passim, 245, 272, 310±14 passim, Stewart, H. 99, 113
317, 321, 332, 389 Strawson, Peter 11, 14, 33
Ross, Don 8, 13, 191, 192, 223, 247, Sugden, Robert 83, 97, 103, 113, 131,
251, 257, 258, 269, 272, 371, 389 143, 153, 214, 215, 224
394 Name index

Tabellini, G. 260, 271 Watkins, J. N. 31

Tani, P. 214, 221 Weibull, J. W. 212, 215, 224
Taylor, Charles 83, 102, 113 Weintraub, E. Roy 177, 245, 370, 389
Taylor, Michael 97 Werlang, S. R. 141, 148, 151
Taylor, P. 211, 224 Weston, S. 246±8 passim, 253, 272
TeraÈsvirta, T. 346, 356 Wiggins, D. 32, 34, 38, 163
Thompson, E. P. 65 Williams, B. 48, 49, 51, 102, 113
Tirole, J. 348, 358 Williamson, O. E. 150, 153, 389
Traweek, S. 315, 332 Winter, S. 13, 123, 131, 189, 190,
199±209 passim, 212, 217±21
Vanberg, V. J. 150, 153 passim, 224, 389
Van Damme, E. 224 Witt, U. 219, 224
van Fraassen, Bas 31, 312, 332 Wittgenstein, Ludwig 5, 6, 14, 33, 158,
Von Neumann, J. 185, 224 171±2
Vriend, N. 218, 224
Vromen, Jack 7, 13, 129, 131, 195, 224, È . 249, 272
YalcËin, U
369 Young, P. 214, 224

Walliser, B. 126, 131 Zamora Bonilla, Jesus P. 370, 389

Walras, LeÂon 178, 244 Zerbe, R. O. Jr. 224
Wang, T. 133, 148, 151 Zucker, R. 265, 272
Waring, M. 323, 332
 Subject index

abstraction 187, 202, 229, 319, 326, 225, 227, 252, 293, 312, 336, 372,
330, 345, 376 384
adaptive expectations 335, 336, 337, betting 140, 144±6, 150
339, 350, 351, 353, 354, 355 boundaries of economics and the
agency 11, 48, 103, 105, 109, 112, economic realm 4, 5, 8±9
114-31, 302
aggregates 3, 8, 13, 29, 180, 185, capacities 3, 6, 44±5, 98, 160, 167, 242,
225-46 249, 299±301, 316±19, 324, 330,
general price level and GDP as 226, 370
228, 230, 232, 236, 239 capital 27, 36, 40, 42, 43, 59, 114, 120,
natural 230 196, 260, 261, 263±5, 267, 320,
real rate of interest 230, 235, 239, 329
241 cartesian 361±8
supervenience on microeconomics catallaxy 165
226, 236 causal completeness 310, 329
synthetic 230±6 causal process 371±2, 382, 384, 386
yield curve 241 causal relevance 381, 383
aggregation 39, 71, 182, 228, 229, 235, causal theory of reference 118, 121,
310, 312, 327, 347 122, 129
appreciative theory 202±5 causality 23, 129, 130, 343 (see also
Arrow's impossibility theorem 59, 61 composition of causes, disturbing
arti®cial intelligence 351±5 causes)
associationist theory of concept causation 3, 10±11, 13, 300, 302, 319,
formation 285 321, 365, 370
ceteris paribus clauses 127, 300
bargaining 82, 83, 120 ceteris paribus laws 275±92 passim
Bayesianism 81, 140, 143, 147, 148 chance 74, 132±4, 137±9, 141, 143,
behaviorism 99, 184±6 150, 151, 238, 330
belief 5, 7±11, 13, 51, 54, 76, 77, 81, 82, choice 3, 5, 6, 8±10, 13, 14, 40, 57±68,
93, 99, 100, 108, 109, 122, 125, 71±4, 76, 78, 86, 99, 100, 102±6,
126, 132, 135, 140±51, 176, 182, 113, 114, 117, 132, 140, 143,
183, 190 ±7, 202±4, 208, 218, 219, 145±7, 150, 177, 182±7, 189, 193,

395
396 Subject index

choice (cont.) emotions of shame, guilt and

205, 210, 212, 217, 219, 228, 238, embarrassment 98, 102
252, 253, 263, 265, 270, 299, 300, empiricism 187, 312±3
325, 336, 346, 352, 355 epiphenomena 313
of theory 369±71, 375, 382±6 epiphenomenalism 311, 323
class consciousness 70 equilibrium 95, 96, 106, 107, 111, 112,
commensurability 45, 48, 51 125, 178±80, 202, 203, 210, 211,
communitarianism 68, 71 213, 231, 243, 244, 327, 336±8,
competitive markets 195, 219, 220, 327 346±54, 360, 376, 377, 380, 382
complex systems 299 essence, essential 159±61, 163, 164, 167,
composition of causes 10, 13, 302, 305, 170, 242, 377±8, 379, 381, 383
306 essentialism 158, 163±73, 381
consumptions 62±6 evidential weight 146, 297
contract 3, 69, 165, 168, 171, 201, 382 evolution 8, 13, 129, 175, 178, 179, 214,
contractarianism 68, 69, 168 326, 382
convention 88, 92, 93, 95, 123, 125, evolutionary economics 13, 189±221
126, 130, 150, 214±16, 319, 380 evolutionary game theory 209±21
counterfactuals 283±4, 298 evolutionary processes 189, 204, 208,
Cournot problem 227, 229, 231, 238, 215, 219
239 exchange 37, 38, 42±4, 51, 52, 70, 82,
84, 122, 140±2, 161±3, 165, 178,
deductive method 293±7, 299, 302±7 250, 264, 347, 348, 364, 375, 382
democratic theory 61 expected utility theory 76, 140, 146,
desire 13, 43, 71, 74, 76±81, 84, 93, 148, 149
100, 111, 121, 140, 182, 183, experimentation 303
191±4, 234, 252, 264, 265, 297, explanandum 91, 96, 192
299, 327 explanation 19, 27±9, 57±9, 62, 68, 72,
determinism 175, 309, 310, 313, 319, 73, 75, 79, 83, 90±7, 100, 105, 109,
339±42, 353, 384 111, 112, 114, 125, 146, 167±70,
and constraint 316, 329 178±82, 185, 193, 196, 214±16,
and free will 313, 318 236, 238, 240, 257, 261, 292, 294,
mereological 313±16 295, 299±301, 312, 317±20, 327,
distributed lags 338±40, 342±4, 354 328, 359, 361, 362, 365, 372, 376,
disturbing causes 296, 298, 300, 301 377, 379, 381
downward causation 319, 321 exploitation 69, 70, 74
Dutch Book 141, 142, 148
fact±value distinction 324
econometrics 255, 339, 341, 343, 344, ®rm 109, 110, 374, 375, 385
347, 353 folk and scienti®c economics 32, 54, 229
economic kinds 17, 20±1, 321 folk psychology 191, 192, 194
economic ontology 3±14 passim formal theory 202±4, 374
(see also ontology) free-riding 93
economic realm 3±14 passim, 39, 263, Fregean theory of reference 118, 119,
265, 314 121
economic world view 3, 6, 7 frequency theory 137
egocentricity 76 Freudianism 85
 Subject Index 397

general equilibrium 178, 179, 360 laws 91, 183, 184, 229, 237, 239, 255,
general equilibrium theory 178, 179, 293±307, 309, 311±16, 320, 330,
231, 243, 327, 336, 338, 346±8, 341, 351, 360±4, 366, 372, 380
350, 352, 354, 376 laws of nature 275±92 passim
learning 129, 202, 211, 212, 217, 338,
hermeneutics 10 348±55
heterogeneity 39, 40, 45, 218 liberalism 66, 168
homo economicus 10, 13, 75
hysteresis 112 machines 13, 212, 215, 306, 315±18,
326±31
idealization 14, 184, 187, 229, 242, 243, macroeconomics 255, 257, 259, 269,
252, 359±68 270, 330, 350, 355, 365
imitation 212 aggregates in 226, 230, 235, 236, 241,
incentive-compatible 80 242, 243
index numbers 232, 235 de®ned 226
individualism 13 Keynesian 226, 227, 242±3
methodological 26±30, 130, 180 182, manipulation of aggregates 241
184, 185, 227±9 measurement of aggregates in
ontological 26±7, 228, 229, 236, 370 230±2, 235
induction 91, 106, 151, 253, 293, 303 microfoundations for 29, 180, 225,
inertia 92, 93, 218, 221 226±9, 238
inexact laws 297±301 new classical 243
inference to the best explanation 17, real business-cycle model 243
19±20 reality of 225, 240±3
informational requirement 373, 377±8, marginalism controversy 195
380 market 3, 5, 6, 8, 9, 13, 32±4, 39±41,
institutions 3, 53, 61, 95, 159±64, 167, 44, 45±8, 51, 53, 63, 67, 68, 70, 75,
168, 170, 171, 181, 182, 214±16, 78, 82, 86, 90±2, 94, 96, 111, 122,
329, 378±82 157±9, 161±73, 178, 180, 186,
instrumentalism 10, 143, 144, 186, 204, 187, 195, 196, 198, 199, 201±5,
252, 253, 325, 385±6 208, 209, 219, 220, 235, 241, 264,
intentionality 182±4, 238, 252 299, 312, 322, 326±31, 375, 380,
interest groups 58, 59 382, 385
isolation and de-isolation 373±6, 383 materialism 257, 312, 313
mechanical phenomena 302, 319
Keynesian economics 133, 137, 242, mechanics, mechanism 4, 6, 8, 11, 13,
243, 336 112, 129, 186, 202, 204, 241, 258,
kinds 17, 20±1, 34, 35, 37, 42, 45±9, 268, 280, 282, 299±302, 308±15,
176, 229, 312, 316, 320, 321, 381 324±9, 362, 367, 370±2, 375, 379,
386
labor/labour 38, 45±7, 196, 265±7, metaphysics 8, 33, 34, 46, 47, 50, 51,
329 131, 149, 174±88, 191, 298, 308,
labor market 327±9 309, 312, 321, 361
labor theory of value, Marxist 69±71 method a priori 293, 295, 302
labour and technological change 268 methodological individualism see
Laspeyres index 233 individualism
398 Subject index

methodological instrumentalism 143, observational equivalence 344, 345,

144, 186, 204, 253, 325, 385±6 354
microeconomics 180±2, 184, 191, 192, occurrent properties 284
251, 257, 269, 270, 365, 371 ontological categories 7±8, 11
microeconomics, de®ned 176, 226 ontological commitment 4±5, 6±7, 9,
microeconomics, supervenience of 12, 13, 116, 121, 122, 184, 189,
macroeconomics on 236±40 190, 203, 204, 214, 217, 220
microfoundations program for ontological constraint 338, 339, 341±5,
macroeconomics 225, 226±9, 239, 347, 350±5, 371, 375, 377, 381±3,
243 385
mind, common and economic 85, 87, ontology 99, 100, 115, 116, 136, 141,
90±4 149, 236, 252, 312, 321, 325
modality 300, 385 and ``metaphysics'' 8
impossibility 135, 377, 380, 385 a priori and a posteriori 12
necessity 167, 228, 256, 310, 364, as absolute presuppositions 5, 7
379, 380, 381, 383, 385, 386 as constitutional framework 5, 6, 11
possibility 218, 378, 379, 380, 383 as constraining belief and theory
models 18, 76, 78, 103±12, 120, 133, choice 9±10, 371±2, 377 (see also
201±3, 213±16, 227±9, 231, 243, ontological constraint)
300, 301, 315±18, 325, 327, 328, as fundamental 5, 249
330, 336, 337, 339, 340, 342±54, as implicit 6±7, 11
372, 373, 376±86 as vision 4, 5
models and laws 316, 330 as windows 5, 6, 9±10, 12
money 3, 8, 9, 32, 37±44, 49±53, 58±9, demysti®cation of 6
142, 144, 162, 164, 169, 227, 230, descriptive and revisionary 11±12
231, 233, 234, 263, 305, 321, 327, local or regional or special 7, 8
364 of economics 3±14 passim, 98, 112,
money-multiplier 280±2 371
moral economy 62±5 passim of economists and of economic
multiple equilibria / equilibrium theories 6±7, 10, 136, 149,
selection 106, 107, 112, 213 189±91
of theory choice 369±86 passim
natural kinds 17, 20±1, 34, 38, 192, opportunity cost 263±7
193, 194
natural selection 196, 200, 201, 204, Paasche index 233
209±11, 310, 317 Paretian 80
natures 43, 171, 173, 275±92 passim parts and wholes 318±20
neo-Kantian 118 payments system 110
non-truism, strong and weak 77, 78 perfect competition 372±83, 385
normativity 246 imperfections 378, 383
norms 93, 99, 102, 104±7 passim, 111, personal identity 115, 117, 126±7, 130
112, 214±16, 369 perspectives (epistemological) 253±6
philosophy of psychology 184
objectivity and policy 246±8, 258±68 physics 174, 176, 178±80, 186, 240,
passim, 270 255, 257, 270, 300, 309±16 passim,
 Subject Index 399

319±21 passim, 326, 329, 349, realism 114±16, 121, 226, 229, 252,
359±68 passim, 370, 371, 375 336, 337, 353, 355, 364
pluralism 327 ontic 385
policy, ®scal investment 259±62 ontological 114, 121, 226
political economy, structuralist 60 referential 116, 118
positive economics 246, 247, 249, 253, representational 116
257, 265 scienti®c 118, 258
post-modernism 158 semantical 116
prediction 58, 72, 79, 146, 191, 196, theoretical 336, 381, 385±6
197, 214, 238, 255, 294, 296, reality, at micro- and macroeconomic
300±3, 335, 336, 339, 343, 345, levels 238, 239, 262
346, 359±62 passim, 364±7 reality, macroeconomic 238, 269
passim, 386 reality, microeconomic 236, 238
preferences 3, 9, 39, 60, 63, 64, 66, 76, reduction 4, 9, 26±30, 35, 39, 105, 158,
80±1, 98±105, 119±20, 142, 145, 225, 228, 229, 238±40 passim, 244,
177±8, 181±5, 191±4, 258, 345, 250, 251, 257, 319
366±8, 375, 379 reductionism 26±30, 170, 175, 181,
principle of indifference 136, 137, 139 239, 249, 250, 257, 309±11, 313,
probability 3, 8, 13, 64, 76, 132±51 320, 326
passim, 310, 318, 348, 351 reference 34, 92, 116±22, 129, 130, 134,
a priori 134, 135, 138 137, 138, 150, 199, 269, 363, 366
interval-valued 147±9 regularities 254, 275±92 passim, 294,
logical interpretation of 135, 136, 299, 300, 317, 343
138 representative-agent model 228, 243
product and process innovation 374 resilience 75, 91±5
projectability 254±6 revealed preference 184, 185, 192
revolution 60, 62, 67, 69, 72, 73, 344,
quarks 186, 240 355, 362, 367
risk aversion 63
randomness 339±42, 353 routines 116, 123±30, 150, 205±9, 212,
rational choice 8, 13, 57±74, 99, 100, 217, 218
103, 104, 106, 132, 212
rational expectations 193, 335±55 Santa Fe Institute 346, 349, 352, 354
rational peasant 63 scarcity 3, 63, 219, 247, 248, 263,
rationality 3, 5, 8, 9, 57, 59, 63, 67, 75, 265
76, 78, 141, 146, 194, 207, 210, selection arguments 29, 190, 194±5,
214, 216, 254, 323, 346, 350, 355, 198, 199, 201±3
376 self-worth 9, 13, 98±112
rationality, bounded 210, 216, 349, self-interest 49, 61, 62, 74, 82, 87±90,
352 92, 323, 327
rationality, expressive 13, 98±113 social facts 122, 125, 126, 128, 319
(especially 99, 103) socialism 158, 165
rationality, instrumental 9, 99, 121 socio-economic machines 275±92
real patterns 8, 13, 251, 252, 255±7, passim
262, 265, 268±70, 371 Sophists 50
400 Subject index

supervenience 4, 27±8, 236±40, 249, 234, 246, 253, 264, 268, 311, 321,
250, 251, 257, 269, 270, 313, 320 337, 359, 360
of macroeconomic aggregates on utility functions 58, 66, 76, 77, 103,
microeconomics 29, 235, 236±7 145, 198
symmetry 14, 335±55 passim
value theory 57, 59, 60±2, 67±73
technical constraint 335±55 passim passim
tendencies 9, 13, 66, 198, 199, 209, vector autoregressions 342±6, 347,
293±303 passim, 317, 371 348, 351, 354
time allocation model 120 veri®cation 295, 296
transaction costs 373, 374, 379, 381 virtual control 87±9
truth 48, 116, 118, 137, 138, 172, 174,
176, 183, 184, 186, 187, 192, 247, wants/needs 177
259, 270, 297, 298, 310, 315, 336, wealth 3, 36, 39, 40, 42±5, 51, 52, 63,
362±4 passim, 368, 385 104, 161, 162, 169, 297±303
signi®cant and insigni®cant 384 passim
welfare 8, 13, 57±60, 63, 64, 66, 67, 71,
unity of science 250, 315, 320, 324 73, 157, 265, 329
unrealisticness 372±7, 381, 383, 384, welfare economics 80, 176, 270, 321
386 work 6, 61, 62, 110, 111, 214, 265,
as narrowness 373±5, 383 321±2
use value 35 www (the way the world works)
usefulness 36±8, 197, 366 371±86 passim
utility 3, 33, 36±40, 46, 47, 51, 67, 76, www constraint 371±3, 377, 381,
91, 140, 146, 148, 149, 192, 198, 383±5