The Economics of Financial

Services in Emerging Markets

Any enquiry into the nature, performance, role, demerits, growth, efficiency
or other aspects of financial services such as banking and insurance activities,
requires rigorous estimates of their economic output, that is, the economic
contributions made by these firms, as well as by the industries as a whole.
Accordingly, this book condenses several theoretical, methodological,
empirical and philosophical issues in conceptualizing, measuring and empirically
operationalizing the economic output of the banking and insurance industries.
The analytical focus is on both Global and Emerging Markets perspectives. The
book synthesizes applied and conceptual evidence to locate the chosen theme’s
analytical patterns, consensus and disagreements. The selected subject matter
is studied within the firm-level and aggregate settings, bringing literature of
varied scopes together. Contributions from various international academics,
practitioners and policymakers further enrich the narrative.
The book concludes with data-driven case studies that analyse the extent to
which the critical performance parameters of the banking and insurance industries
in the BRIICS economies – including estimation of aggregate industry-level
partial factor productivities, total factor productivity, technical efficiency and
returns to scale – vary concerning alternate measures of their output. The present
work also provides a brief note on the inputs measurement dimension, following
which there is a discussion on the limitations, future scope and conclusions.
This work will be valuable for researchers and policymakers undertaking
performance analyses related to banking and insurance activities. It shall provide
them with the examination of a plethora of analytical options and related issues
on the theory and praxis of output measurement, all finely organized into one
single volume.

Bhagirath Prakash Baria is Assistant Professor at the Department of Banking

and Insurance, Faculty of Commerce, The Maharaja Sayajirao University of
Baroda, India. The broad areas of his interest include macroeconomics of banking,
agricultural productivity and performance analysis of banking and insurance
industries. The author strongly believes in pluralism in both teaching and research.
He has a masters’ degree in business economics with two Gold Medals, has
published in several reputed national and international journals, won a Young
Researcher Award, founded a non-profit educational initiative called The Cafe
Economics (TCE), and is pursuing a PhD in the area of International Finance.
Banking, Money and International Finance

28. The Digital Disruption of Financial Services

International Perspectives
Edited by Ewa Lechman and Adam Marszk

29. The Economics and Finance of Commodity Price Shocks

Mikidadu Mohammed

30. Artificial Intelligence and Big Data for Financial Risk

Management
Intelligent Applications
Edited by Noura Metawa, M. Kabir Hassan and Saad Metawa

31. The Role of Crises in Shaping Financial Systems

From the Global Financial Crisis to COVID-19
Małgorzata Iwanicz-Drozdowska, Elżbieta Malinowska-Misiąg, Piotr Mielus,
Paweł Smaga, and Bartosz Witkowski

32. Insurance Market Integration in the European Union

Slawomir Ireneusz Bukowski and Marzanna Barbara Lament

33. The Economics of Financial Services in Emerging Markets

Measuring the Output of the Banking and Insurance Industries
Bhagirath Prakash Baria

34. Innovative Finance for Technological Progress

Roles of Fintech, Financial Instruments, and Institutions
Edited by Farhad Taghizadeh-Hesary, Roohallah Aboojafari and Naoyuki
Yoshino

35. Understanding the Polish Capital Market

From Emerging to Developed
Edited by Marek Dietl and Dariusz Zarzecki

For more information about this series, please visit: www.routledge.com/

series/BMIF
The Economics of Financial
Services in Emerging Markets
Measuring the Output of the Banking and
Insurance Industries

Bhagirath Prakash Baria

First published 2023
by Routledge
4 Park Square, Milton Park, Abingdon, Oxon OX14 4RN
and by Routledge
605 Third Avenue, New York, NY 10158
Routledge is an imprint of the Taylor & Francis Group, an informa business
© 2023 Bhagirath Prakash Baria
The right of Bhagirath Prakash Baria to be identified as author of this work
has been asserted in accordance with sections 77 and 78 of the Copyright,
Designs and Patents Act 1988.
All rights reserved. No part of this book may be reprinted or reproduced or
utilised in any form or by any electronic, mechanical, or other means, now
known or hereafter invented, including photocopying and recording, or in
any information storage or retrieval system, without permission in writing
from the publishers.
Trademark notice: Product or corporate names may be trademarks or
registered trademarks and are used only for identification and explanation
without intent to infringe.
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library
Library of Congress Cataloging-in-Publication Data
Names: Baria, Bhagirath Prakash, author.
Title: The economics of financial services in emerging markets : measuring
the output of the banking and insurance industries / Bhagirath Prakash Baria.
Description: Abingdon, Oxon ; New York, NY : Routledge, 2023. | Series:
Banking, money and international finance | Includes bibliographical
references and index.
Identifiers: LCCN 2022027726 (print) | LCCN 2022027727 (ebook) |
ISBN 9780367712136 (hardback) | ISBN 9780367712143 (paperback) |
ISBN 9781003149828 (ebook)
Subjects: LCSH: Financial services industry—Developing countries. |
Banks and banking—Developing countries. | Insurance—Developing
countries.
Classification: LCC HG195 .B367 2023 (print) | LCC HG195 (ebook) |
DDC 332.109172/4—dc23/eng/20220616
LC record available at https://lccn.loc.gov/2022027726
LC ebook record available at https://lccn.loc.gov/2022027727
ISBN: 978-0-367-71213-6 (hbk)
ISBN: 978-0-367-71214-3 (pbk)
ISBN: 978-1-003-14982-8 (ebk)
DOI: 10.4324/9781003149828
Typeset in Bembo
by Apex CoVantage, LLC
 To,
Maa Saraswati
Prof. K. Shanmugan
Ela
Contents

Acknowledgements xiv
Contributors xvi
Foreword xvii
Preface xix

1 Setting the analytical background 1

1.1. Background and introduction 1
1.2. Theory of output measurement: features of an ideal output
measure for banking and insurance activities 3
Product-mix representativeness 4
Quality-adjusted measure 4
Robustness to regulatory interventions 4
Endogenous to the production function 5
Methodological robustness 5
Availability of multiple proxies 6
Maximisation criterion 6
Based on an output specification approach rather than ad hoc 6
Reflect the size of the firm or the industry 7
High correlation with other possible alternatives 7
Existence of an optimal corresponding price index 7
Ability to handle the multi-output nature of banking
and insurance activities 8
Stable and rigid definition 8
Fully endogenous to the firm 9
Unitary influence factor 9
1.3. Final remarks 9
Notes 10
References 11
viii Contents
PART I
Survey of evidences: theoretical and empirical issues13

2 Qualitative and quantitative aspects of the survey

of literature 15
2.1. Sample versus population dimensions of literature survey 15
2.2. Review strategy 16
2.3. Quantitative summary of the selected sample of studies 17
2.4. Problems in the purposive selection of studies 23
2.5. Background on further sections 24
Notes 25
References 26

3 Review of evidence on banking output measurement:

global perspective 27
3.1. Introduction 27
3.2. Empirical sensitivity of banking performance estimates
to alternative outputs 27
3.3. Approaches to output specification in the banking literature 30
Production approach 32
Intermediation approach 34
Asset approach 36
Value-added approach 37
User cost approach 37
Index numbers approach 38
3.4. Output measurement in the banking literature 39
Loan outputs 39
Deposit outputs 43
Asset output 48
Investment output 48
Income and off-balance-sheet output 51
FISIM output 53
National income accounts and banking output 54
3.5. Thematic issues in banking output measurement 55
Deflation of banking output 55
Stock versus flow dimensions in output measurement 58
Issues in economic aggregation 58
Some econometric dimensions in the literature 59
Monetary policy and the banking production function 60
Other issues 61
Appendix: Evolution of banking output measurement in the extant
wisdom from 1950 to 2022 63
Notes 90
References 93
 Contents ix
4 Review of evidence on insurance output
measurement: global perspective 110
4.1. Introduction 110
4.2. Empirical sensitivity of insurance performance estimates
to alternative outputs 110
4.3. Approaches to output specification in the insurance
literature 111
Production approach 111
Intermediation approach 112
Value-added approach 112
National income approach 113
4.4. Output measurement in the insurance literature 113
Life versus nonlife insurance output 113
Premium output 114
Claims and loss output 116
Investment and financial assets output 119
Reserves as output 121
Income output 121
National income output of insurance 123
4.5. Thematic issues in insurance output measurement 126
Market structure, organisational structure and performance
estimation 126
Output mix and output measures 129
Regulation and output measurement 129
Insurance output measurement in advanced versus emerging
economies 130
Deflation of nominal output in insurance 130
Taxation and insurance output measurement 131
Appendix: Evolution of insurance output measurement in the extant
wisdom from 1980 to 2022 132
Notes 149
References 151

5 Extended notes on the banking and insurance output

measurement problem 162
5.1. Summarising the literature survey 162
5.2. Extended notes 162
Synthesising the concept and measure of outputs 162
Stock versus flow outputs 163
Nominal versus real outputs 163
Single versus multiple outputs 164
National income versus economic approaches 164
Dual nature of output specification problem 165
x Contents
5.3. Background for further chapters 165
Notes 165
References 165

PART II
Expert opinions and contributions: emerging markets
perspective167

6 Method and rationale 169

6.1. Motivation and rationale 169
6.2. Sampling strategy 169
6.3. Notes for the next chapter 170
Notes 170

7 Expert perspectives on output measurement in

banking and insurance 171
Barendra Kumar Bhoi 171
Issues specific to the measurement of financial services in the
system of national accounts 2008 in emerging economies 171
Difficulties in generating current price and constant price
estimates of services output through the value-added method,
including the issue of double deflation 172
Theory and measurement of financial services output versus
other services 174
The great financial recession and key performance issues of the
banking sector of the emerging economies 175
Structural changes in the composition of banking business from
intermediation services to fund-based services and their
implications for the importance of the FISIM approach 176
Further measurement issues in the output of the
banking sector 177
Some issues in the input-output debate in banking
output measurement 178
Debashis Acharya 180
Output measurement in banking services 180
Some issues with aggregation and aggregation functions in the
current context 182
Some frontier issues in output measurement theory and practice 183
Edoardo Pizzoli 185
Satellite accounts, financial services and the use of SNA in
emerging economies 185
Output of banks and some comments on its economic
aspects 186
Notes on some allied issues 187
 Contents xi
Frauke Kreuter 188
Traditional data collection systems and their emerging
alternatives in the construction of macroeconomic
statistics 188
Adoption of emerging alternative data sources in advanced
economies and the emerging market economies 190
Some challenges faced by the emerging economies in traditional
data collection 191
Justin Paul 193
Multi-output nature of banking and insurance production 193
Banking industry in the advanced versus emerging
economies 194
Further remarks on measuring the output of the banking
industry 195
Some issues in using financial statements as the fundamental
source of output data for banks 195
N. R. Bhanumurthy 197
Notes on national income and other macroeconomic
accounts 197
Financial services versus the tangibles in the estimation
of output 198
Differences in economic structures and international
comparisons 198
Input-output measurement in financial services, including the
FISIM methodology 199
Some frontier issues in financial services output
measurement 201
Pronab Sen 202
Emerging economies and the system of national accounts 202
Under- and overestimation of financial services output under
economic shocks 203
Defining the output of services versus other industries 204
Dual intermediation roles of banks and the input-output
debate 205
Finance-growth nexus in emerging economies 206
Some notes on insurance output measurement 206
Some frontiers of banking output measurement 207
R. B. Barman 208
Theoretical and empirical issues in the system of national
accounts of emerging economies 208
Nature of financial services output 210
Comments on the FISIM approach in measuring banking
output 211
xii Contents
Possible future developments in the output measurement
of financial services 212
Ram Pratap Sinha 213
Insurance production as a multi-stage activity 213
Alternative approaches to the empirical estimation of insurance
performance 215
Some issues in insurance input measurement 215
Some issues in insurance output measurement 216
Price measurement in insurance 217
Zhu Haiju 218
Measurement of output and its rationale 218
Nature of output in the services industry 219
Some empirical aspects of output measurement
in the EMEs 219
Notes 221
References 226

PART III
Empirical case studies: a BRIICS perspective229

8 Empirical case studies for the banking industry on

implications of using alternative output definitions 231
8.1. Rationale, methodologies and data 231
8.2. Empirical sensitivity of selected partial factor productivity
estimates across the banking sector of BRIICS 235
8.3. Empirical sensitivity of total factor productivity and technical
efficiency estimates across the banking sector of BRIICS 237
8.4. Empirical sensitivity of returns to scale estimates across the
banking sector of BRIICS 239
Brazil 241
Russia 241
India 241
Indonesia 242
China 242
South Africa 242
8.5. Cross-country analysis 242
Appendix 1: Estimates of labour and capital (partial factor)
productivities for section 8.2 243
Appendix 2: Estimates of residual total factor productivity index
and technical efficiency scores for the section 8.3 248
Appendix 3: Estimated models for the returns to scale coefficients
in section 8.4 257
 Contents xiii
Appendix 4: Estimates for section 8.5 275
Appendix 5: Input variables and estimates of output elasticities
of inputs used in the construction of the total output index for
section 8.3 276
Notes 279
References 283
9 Testing the empirical sensitivity of major performance
indicators of the insurance industry under alternative
output definitions: selected exercises 285
9.1. Background 285
9.2. Partial factor productivity estimates for the insurance
industry 286
9.3. Total factor productivity and technical efficiency estimates for the
insurance industry 286
9.4. Returns to scale estimates for the insurance industry of selected
EMEs 289
Appendix 1: Estimates for Section 9.2 290
Appendix 2: Estimates for section 9.3 293
Appendix 3: Estimates for section 9.4 300
Notes 305

PART IV
Extended notes and concluding remarks307

10 Inputs measurement issues 309

10.1. Introductory note 309
10.2. Inputs measurement in the banking and insurance
literature 309
10.3. Concluding remarks 312
Notes 312
References 312

11 Limitations, future scope and concluding remarks 314

11.1. Limitations and future scope 314
11.2. Final concluding remarks 316
Notes 317
References 317

Index318
Acknowledgements

No research work can be undertaken without direct and indirect contributions

from various people. Many people have contributed to the culmination of an
idea in this work. Their help and efforts are embodied in the pages that will
follow.
I first express my sincere thanks to Dr Hersch Sahay, who allowed me to
access various data sources that I otherwise couldn’t have. He was kind enough
to let me utilize some of his resources for this work. Sincere regards are also due
to Dr Meeta Rathod for helping undertake the plagiarism check for the manu-
script. My deepest regards are due to the esteemed experts who agreed to con-
tribute their rich insights and opinions to this work. These experts from across
different nations have immensely enriched this work. Despite being extremely
preoccupied, the experts put their time and effort into creating insights and
narratives that will help readers explore the theme through a practitioner’s lens.
I submit my sincere gratitude to Dr Barendra Kumar Bhoi, Dr Edoardo Piz-
zoli, Prof. Debashis Acharya, Prof. Frauke Kreuter, Prof. Justin Paul, Prof. N.
R. Bhanumurthy, Dr Pronab Sen, Dr R. B. Barman, Prof. Ram Pratap Sinha
and Prof. Zhu Haiju. They all lent their years of hard work, experience, and
expertise to the pages to follow. It would be a serious error to miss thanking
Dr Sofia Devi for helping me connect with Dr Barendra Kumar Bhoi and
Prof. Justin Paul. She was kind enough to let me utilize her academic network.
Thankfulness is also due to Prof. K. Shanmugan, who lent his time and exper-
tise to comment on several portions of the book, particularly the empirical
parts. I want to submit my gratitude to Prof. K. V. Bhanumurthy, who provided
rich insights on the problems of adopting aggregated perspective in production
analysis. His discussions have helped me recognize several strengths and limita-
tions of the empirical work undertaken in this book. Warm regards are also due
to Prof. Madhusudan Raj, who helped in connecting with Prof. Zhu Haiju
and who also provided valuable inputs on the subject matter. At this junc-
ture, I must also thank the anonymous referees from Routledge, who provided
extremely useful comments on this work in the early stages. The comments
were very insightful and were duly incorporated into this work. It must also be
humbly put on record that my University Library – Smt. Hansa Mehta Library
 Acknowledgements xv
lent great help in providing access to premium databases through which it was
possible to build the database of studies surveyed in this work.
While these wonderful people have shaped various portions of this book,
some have played a foundational role throughout the journey. Kristina Abbotts,
Senior Editor (Economics, Economics, Finance, Business & Industry), and
Christiana Mandizha, Editorial Assistant (Economics), Routledge, are among
them. Both of them guided me throughout this journey. They were always
available for help, queries and any other issues that an author might typically
face. Not only that, but both were extremely tolerant of my continuous que-
ries with long emails and frequent delays. The positive aura that they reflected
throughout this journey has played a pivotal role in my journey on this won-
derful project. I submit my most sincere thanks to both of them. It will be
incorrect to miss thanking my dear wife, Devanshi, for constantly being there
every single day and night that I worked on this manuscript. She was always
there by my side, ranging from lending help in collecting and organizing the
data to preparing infinite rounds of coffee throughout the night. Last but not
the least, I convey my earnest gratefulness to the Routledge publishers for pro-
viding an invaluable platform for me as an author.
The responsibility for errors, if any, is solely mine.
Bhagirath
Contributors

Barendra Kumar Bhoi Ex-economic policy expert for the Central Bank of
Oman; Ex-Principal Adviser at the Reserve Bank of India; Mumbai, India
Debashis Acharya Professor in the School of Economics, University of
Hyderabad, Hyderabad, India
Edoardo Pizzoli Statistician at the Italian National Institute of Statistics;
Associate Researcher at the University of Luxembourg, Esch-sur-Alzette,
Luxembourg
Frauke Kreuter Professor at the Ludwig-Maximilians-University of Munich;
Head of the Statistical Methods Group at Institute for Employment Research
(IAB); Professor at the University of Maryland, Baltimore, USA
Justin Paul Editor-in-Chief of the International Journal of Consumer Studies;
Faculty Member at the University of Puerto Rico, San Juan, USA
N. R. Bhanumurthy Visiting Fellow, Foundation Maison des Sciences de
l’Homme, and the McGill University; Vice-Chancellor, Dr B. R. Ambedkar
School of Economics University, Bangalore, India
Pronab Sen Former Chief Statistician of India; Programme Director for the
International Growth Centre’s India Central Programme, New Delhi, India
R. B. Barman Former Executive Director at the Reserve Bank of India;
Ex-Vice Chairman of the Irving Fisher Committee on Central Bank Statis-
tics, Bank for International Settlements; Advisor to the National Payments
Corporation of India, New Delhi, India
Ram Pratap Sinha Associate Professor at the Government College of Engi-
neering and Leather Technology, Kolkata, India
Zhu Haiju Professor at the Zhejiang Gongshang University, Hangzhou, China
Foreword

Economic production presumes the existence of well-defined inputs and out-

puts and their technological linkages. These linkages culminate into a produc-
tion function. Observing the classical assumptions of the firm theory, the study
of service as a means of production requires adaptation of the assumptions that
regard tangibility. Within the financial services industry, the specification of
inputs and measurement of outputs is a demanding task. In particular, bank-
ing and insurance firms further complicate this by relying on implicit services;
these violate several postulates of a well-behaved production function and raise
questions on applying traditional production theory to contemporary practice.
Such research has significant implications for both firms and governments.
There has been significant growth in the financial sector among emerging
economies, contributing to overall economic growth. The future points to the
sector expanding its impact as it rapidly advances financialization and services
to consumers in the economy. In literature, the issues, analytical dimensions,
and empirical perspectives are spread across voluminous literature that requires
proper coherence and structure. Undoubtedly, their assimilation into a single
reliable source has definitely been missing in the extant literature. Addressing
this gap, the present book synthesizes a large body of theoretical and empirical
literature on output measurement in banking and insurance, and the scope is to
relay relevant work to this puzzle in a condensed format. The book studies out-
put in the financial sector at both the firm and aggregate levels. The underlying
motivation is to summarize the wide differences in the output measurement
schools regarding banking and insurance activities. Given the special focus on
emerging market economies, the book can be utilized by a large audience in
both the developing and advanced nations as a singular reference for studying
various aspects of the central theme of this work.
While surveys of literature exclusively focus on reviewing the extant wis-
dom, this book also moves beyond it by providing fresh perspectives on the
subject matter from the viewpoint of carefully selected experts. The contribu-
tions by experts on the subject matter enrich the analysis by including debates
on the frontier in output measurement of financial services. Finally, the book
also provides empirical case studies on the sensitivity of various performance
indicators of banking and insurance industries to alternative measures of their
xviii Foreword
output. These case studies focus on the BRIICS countries and the advance-
ment of the financial services sector therein. These empirical case studies, para-
metrically and non-parametrically, examine the sensitivity of key performance
indicators such as partial factor productivities, total factor productivity, techni-
cal efficiency, and returns to scale to different output measures. Readers should
not build bias from the case studies; rather, they should find new ideas and
topics for discussion.
M. Kabir Hassan, PhD
Graduate Coordinator for PhD Program in Financial Economics, professor
of Finance, Hibernia Professor of Economics and Finance, Bank One
Professor in Business, and Hancock Whitney Chair Professor of Economics,
Department of Economics and Finance, University of New Orleans,
New Orleans, Louisiana
Preface

Sitting in the corridors of Smt. Hansa Mehta Library, my colleague Dr Sofia

Devi and I came up with a question: where do we locate a single systematic
source for issues in output measurement for the banking and insurance indus-
tries? This concern was raised as we were working on a particular theme where
identifying the correct output measures was necessary. After some preliminary
survey of the literature, it was clear that this subject was spread across a very
voluminous literature. It was unrealistic to locate any single reference work that
could summarize it in a compact form. This was when the idea of writing this
book emerged, and the pages that follow embody this fundamental concern.
After that meeting, Dr Sofia got engaged in her research commitments, and
I somehow got fascinated by this issue. It struck me that if the large body of
evidence could be systematized into a reference volume, possibly, many analysts
would find it useful. It could also serve as an introduction to output measure-
ment issues, which was definitely missing. However, a standard review of the
literature was surely not enough. This issue deserved much more than a litera-
ture survey due to the sheer diversity of thought. Accordingly, the work on
the structure of this book began. After a long series of informal consultations
with many people working in this area, I finally realized that a ‘mixed method’
strategy was necessary to do justice to this topic.
Accordingly, the work on this manuscript began with full vigour. My earlier
notes on this matter served as the beginning point for this journey. As work
progressed, the manuscript evolved. There was simply too much knowledge
to be captured into a single book! Careful assessment of the evidence, sources,
empirical data, expert consultations, and intuition helped frame the key prem-
ises of this book. The journey of writing this book was a dramatic one with
many movie-like ups and downs. However, nature has its own ways of evolving
us into the beings we are expected to be. The work kept progressing, albeit
at a different pace at different times. And finally, the manuscript emerged out
of the long months of hard work. Passion, love and sincerity have been mixed
together to present this humble work to you, my dear readers. I hope that
this book is able to help you, in some or the other way, in your journey as a
researcher and an analyst.
Bhagirath
1 Setting the analytical
background

1.1. Background and introduction

The theory of firm attempts to explain the existence of firms from diverse
analytical angles such as the sociological, administrative, political, economic,
and others. In the present context, the economic theory of the firm and its
place in the theory of production are critical. The economic theory of the
firm1 as applied to production theory may be traced to the seminal work of
Adam Smith (Curwen, 1976), who made a distinction between the produc-
tive and unproductive sectors of the economy. Services were characterised as
non-productive sectors, with the underlying notion that they consume the
surplus value rather than contribute to it. As the marginal utility theory began
replacing the old concepts of economic value, so did the theory of firm begin
recognising services as value-creating rather than only value-consuming activi-
ties.2 Despite admitting services as value-adding activities, the discourse on
production theory has found it challenging to analyse services with the theo-
retical finesse that it can achieve in the study of tangible goods.
The fundamental problem in analysing services under the traditional theory
of production and firm3 is that the production process in services does not lend
itself easily to quantification. Intangibility is the primary cause of this. The
intangible nature of services results in the unobservability of their production
processes, and further complications are created by the overlapping and inter-
locking nature of their production functions.4 Such complexities in analysing
the production function imply difficulties in conceptualising and measuring the
output of service industries. Vuorinen (1998) summarises the output measure-
ment problems in service industries as follows. First, quality is inseparable from
quantity in services (Adams et al., 2002). Quantitative movements in outputs
and inputs embody qualitative movements too. They cannot be separated eas-
ily in case of services due to the aforementioned problems. Hence, any meas-
ure of output or input in service industries will always contain information
on the quality of the same. Increases (decreases) in inputs and output can be
purely technical or caused by qualitative improvements (deteriorations). Sec-
ond, the nature of services does not permit the storability of its output, which
causes problems in defining proper economic relations between the inputs and

DOI: 10.4324/9781003149828-1
2 Setting the analytical background
outputs (Blois, 1985). Third, quality considerations override the quantity of
services for consumers and society. Thus the nature of service markets is quite
different from the tangible goods markets, where quantity and quality both
may be traced to the changes in the physical composition of the goods, which
is mainly unobservable in the case of services. Fourth, quality has to be con-
ceptualised in the technical sense and via human and social dimensions. Such
an approach naturally complicates the measurement of quality in services and
thus further obscures the derivation of the pure quantity measure of output.5
Among the more significant service sector, the emergence of intermedia-
tion6 services has raised more challenges for the theory of output in production
economics. Whether financial or otherwise, intermediation has been recog-
nised in the economics literature as a source of private and social gains because
of its comparative advantages over the direct exchange. Reduction of transac-
tion, search, monitoring, and bargaining costs, along with a better capability
to handle problems such as adverse selection, moral hazard and credible com-
mitments in trade contracts (Spulber, 1999), have enabled intermediaries to
thrive across the breadth and depth of market economies. Economic theory
recognised the nature and importance of intermediaries in a market economy
long ago. However, with the structural shift across economies towards a higher
share of services in the aggregate output, due recognition was given to the role
of intermediation-service providers as not only arbitragers but also creators
of diverse values. Consequently, the theory of the firm has also undergone
innovations and modifications to account for various kinds of intermediaries
necessary for the efficient functioning of any market economy.
Among these intermediaries, banking and insurance firms are the primary
focus of this work.7 Banking firms intermediate via the deposit-credit-asset
channel, and insurance firms perform intermediation via the premiums-
claims-asset channel, as will be explained in Chapters 3 and 4. The nature of
production by banks and insurers is further complicated because they not only
intermediate but also transform and sell services at each stage of the interme-
diation process. Thus, for illustration, banks not only transform deposits into
credit and sell loan services to borrowers, but they also sell deposit related
services to those who lend their money to banks in the form of deposits.
Similarly, insurers not only pool the premiums and transform them into claim
payments but also sell the risk coverage service to those who pay premiums to
the firm in the first place. These concerns have led to newer debates within
the theory of the firm, whereby analysts have tried incorporating the peculiar
characteristics of these firms into their traditional analysis. Accounting for risk
and uncertainty in production; imperfections in markets (Campbell & Kracaw,
1980); unobservable and indirect services; specialised input requirements at dif-
ferent stages of production (Bylund, 2017); adverse selection and moral hazard
problems; interdependent production; implicit pricing; and non-marketable
services are some of the fundamental ways in which traditional firm theory
has been adjusted to analyse the production behaviour of banks and insurers.
The need to recognise these dimensions seems more significant, particularly
 Setting the analytical background 3
when performance analysis for these firms is undertaken. Performance analysis
consists of diverse issues such as different efficiencies, partial and total factor
productivities, financial performance, profitability analysis, competition and
market power analysis, and other related matters. There is voluminous litera-
ture on the performance analysis of banking and insurance firms. This literature
provides rich insights into how banks and insurers perform on different param-
eters across diverse spaces (cross-sectional units) and times (temporally). A per-
sistent disagreement among the analysts has been to conceptualise, specify, and
measure the output of banks and insurers at the firm and industry levels.
While disagreements have also occurred on the specification of inputs and
the choice of empirical methods, the literature on banking and insurance per-
formance analysis suggests that the extent of difference of opinion is the largest
on the issue of output measurement (Eling & Luhnen, 2010). The sheer varia-
tions in the performance estimates obtained by various studies, as analysed later,
motivate one to ask to what extent are these differences due to differences in
output measurement and to what extent are these due to other considerations?
Literature in banking and insurance has shown considerable similarity in the
inputs vector required in the performance analysis and indicates a noticeable
degree of homogeneity in methodological choices. When inputs specification
and empirical methodology are held constant, the differences in performance
estimates seem primarily caused by the variations in the output vector used
across these studies. While this matter has received attention in the literature,
a singular reference that summarises the voluminous literature in this regard
is missing. The present work adopts a three-dimensional strategy to study the
output measurement issue in banking and insurance performance analysis lit-
erature with this motivation. The first strategy is to provide a detailed survey
of the output specification and measurement approaches in the literature on
banking and insurance. This is undertaken in Part I. The second strategy is to
enrich the findings from the literature survey by providing insights and opin-
ions of practitioners in the subject matter, which is provided in Part II. Finally,
the debates on output measurement in the literature are examined through the
prism of empirical case studies. These case studies offer a fresh perspective on
how the aggregate performance parameters of banking and insurance indus-
tries vary under alternative output measures, with particular attention to the
significant emerging market economies (EMEs). Part III of the present work
provides these case studies. When put together, this three-dimensional strategy
of analysing the output measurement issue in banking and insurance literature
offers a unique perspective beyond a traditional survey of the literature.

1.2. Theory of output measurement: features of an ideal

output measure for banking and insurance activities
While the chapters ahead shall explore and examine many aspects of the output
measurement debate in the banking and insurance literature on performance
analysis, the essence of these debates can be summarised in the features that the
4 Setting the analytical background
literature considers essential for an ideal output measure in the current con-
text. These features have been extracted from the survey of literature presented
in Part I. The discussion below informs the crux of the output measurement
debates in banking and insurance literature. These characteristics represent the
broad expectations from an ideal measure of output in the extant literature on
banking and insurance performance analysis. They are not the final word on the
matter but an indication of how analysts have theorised the idea of bank and
insurance output in the literature. The banking and insurance output measures
discussed in Chapters 3 and 4 have been criticised on several postulated features
below. The task of the analyst is then to attempt to build an output measure that
fulfils as many of these requirements as possible. These ideas are elaborated on
later but presented here to provide a preliminary overview of the critical prob-
lems in measuring the output of banking and insurance firms and industries.

Product-mix representativeness
The ideal output measure of a banking or insurance firm is expected to fully
reflect the changes in the composition of the output, i.e. the product mix.
At the firm level, this implies that the emergence of new products (services)
or changes in the relative share of each service in the total output must be
reflected in the chosen output measure. At the industry (i.e. aggregate) level,
intra-industry shifts in the industry’s total production should induce subsequent
changes in the selected output measure. If the output measure is invariant to
such changes, then it may not truly represent the actual production conditions
of the firm or industry.

Quality-adjusted measure
The chosen output measure for banking or insurance production should be net
of quality effects, i.e. unaffected by the quality changes. As Hughes and Mester
(1993, p. 295) state: “ideally, the y (output) vector in the production transfor-
mation should be measured as quality-adjusted output”. The rationale for this
requirement is that the changes in output embody essential information for
measuring various performance indicators such as efficiencies, productivities,
scale and scope economies, and profitability, among others. These indicators
are estimated by utilising the information on output level and its changes. Sup-
pose the chosen output measure embodies the effects of quality changes. In that
case, the difference in output will not represent the pure change in production
quantity but will be distorted by quality improvements or deterioration. This
may lead to over- or underestimation of various performance indicators.

Robustness to regulatory interventions

Ideally, the composition and determination of the chosen measure of bank-
ing or insurance output should not be distorted by non-market forces such as
 Setting the analytical background 5
regulation. This occurs in cases such as when the so-called ‘Averch and Johnson
effect’ exists, as discussed in further chapters. This requirement is necessary
because of the heavy regulation of the banking and insurance industries in most
major advanced and emerging economies. Observed output and its movements
may be induced by changes in regulations such as monetary policy in the case
of banking and solvency and reserve requirements in case of insurance. Such
variations in output measure may be independent of the underlying produc-
tion conditions and thus may distort the inferences on efficiency, productivity
and other performance indicators. A classic illustration of this phenomenon is
the loan amount, frequently used as an output measure in banking literature.
Increases (decreases) in loans disbursed may be due to loosening (tightening)
of interest rates by the central bank. However, such changes in output do not
necessarily represent any meaningful change in the quantum of production
by the banks. Instead, such changes in output occur outside the scope of the
production function and thus cannot be meaningfully analysed in performance
analysis.

Endogenous to the production function

The technological relationship implied by a production function must hold in
both economic and empirical senses for an output measure to play a meaning-
ful role in any performance analysis that directly or indirectly estimates a pro-
duction function. When an output measure is subjected to frequent and large
exogenous shocks, its viability as an element of the production, cost, profit,
revenue or other functions is questionable. At times, such exogenous changes
in output may be mistaken for technological changes.8 This implies that the
changes in the ideal output measure should represent the underlying market
forces of demand and supply rather than exogenous or non-market forces as far
as possible. Consider an illustration from the insurance literature. At times, the
amount of reserves has been used to measure the insurance firm’s output. How-
ever, reserves may change even when the numbers of policies and policyholders
and even premiums have not changed. This may simply be due to risk aversion
behaviour on the part of the insurers in the light of unexpected adverse shocks.
The higher the degree of risk aversion of an insurer, the greater such changes
may not necessarily reflect actual output movements (Wise, 2018).

Methodological robustness
Performance estimation has generally been undertaken in the banking and
insurance literature using the production frontier approach. As discussed later,
a host of parametric and non-parametric techniques have been utilised. When
different methods within the parametric or non-parametric schools are used,
differences in the performance estimates are natural. However, a desirable char-
acteristic of the chosen output measure(s) is that these differences should be
as minor as possible or within a tolerable limit. This will ensure that different
6 Setting the analytical background
empirical methods do not change the estimated results by an unacceptable mar-
gin if the same output measure is used across all the methodological innovations.

Availability of multiple proxies

The concept of output, its measure and the actual data used to proxy it may
have considerable divergences. Empirical exercises are data sensitive, preventing
an analyst from perfectly measuring the preferred concept of output. Literature
on banking and insurance performance analysis has invariably relied on proxy-
ing the theoretical output concepts by rough proxy measures. For illustration,
the amount of technical reserves maintained by insurers has been used to rep-
resent the risk-bearing services provided by insurance firms to their consumers.
While it is the quantum of risk-bearing services in which one may be inter-
ested, it must be measured using the available data. Technical reserves have been
employed as one such measure.9 However, the amount of losses paid or claims
are also possible candidates and have been used frequently in the literature,
as shown in Chapter 4. Thus, multiple proxies can help an analyst choose an
output measure that best represents the underlying output concept in a given
context. This also implies that the empirical results on various performance
indicators should remain broadly consistent irrespective of the proxy variable
representing a given output measure.

Maximisation criterion
Production theory demands that output must be subject to conscious (micro-
economics) or unconscious (macroeconomics) maximisation. Any measure of
production for banks or insurers must be such that the economic agents should
strive to maximise it within a rational framework rather than minimise it. This
point may seem trivial, but it has critical importance for banking and insur-
ance literature. As shown in Chapters 3 and 4, researchers have used measures
such as Non-Performing Loans and the amount of loss paid as output meas-
ures for banks and insurers, respectively. These output measures cannot be eas-
ily justified because their maximisation does not seem rational economically.
Some studies have dodged this problem by theorising such output measures as
‘bad output’. Still, a production function that minimises both the inputs and
the ‘output’ cannot yield economically meaningful performance parameters.
Gaganis et al. (2013, p. 431) summarise this notion by asserting “that more
output should be preferred to less” in any rational production process. Similarly,
Brockett et al., 2005 suggest that performance estimates should improve when,
ceteris paribus, output increases or inputs decrease, or both co-occur.10

Based on an output specification approach rather than ad hoc

As elaborated on in Chapter 3, the literature in the current context has
treated the problem of output measurement as first a problem of output
 Setting the analytical background 7
specification. Well-established specification approaches exist in the litera-
ture. They include production, value-added, intermediation, asset, user cost,
national income and other approaches. Ideally, the choice of the output vari-
able must emerge from the chosen specification approach rather than being
ad hoc. The approaches mentioned earlier are the fundamental theories of
banking and insurance production. Hence, choosing output measures that
correspond to a specific theory has been deemed a better strategy in the
extant wisdom than the ad hoc or the subjective approach. These matters will
be discussed in detail later.

Reflect the size of the firm or the industry

At the firm and industry levels, the quantum of output is also utilised to meas-
ure how large or small that firm or industry is. There is a strong positive cor-
relation between the size of a firm and its output level. This has been one of the
primary motivations for choosing output measures in the extant literature on
banking and insurance performance analysis (Wise, 2018). A similar observa-
tion applies at the industry level. Literature on the present theme has found it
necessary to use such measures of output that can directly signify the economic
size of the firm or industry. Economists consider the real value of production in
the goods market as the correct representation of the level of activities under-
taken by a firm or industry.11

High correlation with other possible alternatives

As discussed previously earlier, the same output concept may be captured by
different measures in the case of the banking and insurance industries. Empiri-
cally, if the alternative proxies are strongly correlated with each other, prefer-
ably positively, then the task of choosing an output variable becomes primarily
dependent on data availability. The analyst can then select that proxy variable
with higher data availability. Both the banking and insurance literature have
implied this expectation.12 For example, in insurance output literature, the risk
coverage output can be proxied by the amount of claim or loss or the reserve.
In such cases, the a priori belief is that the chosen variable strongly correlates
with other possible proxy variables. Thus, there is not much information loss
in the performance analysis. In practice, such beliefs have been presumed rather
than proven empirically. The analyst uses her subjective understanding and
prior evidence to make such a leap of faith in choosing the correct output
measure for a given output concept.

Existence of an optimal corresponding price index

One of the thorniest issues in output measurement debates on banking and
insurance performance analysis has been the deflation of the current price
value-measures of outputs. As found in Chapters 3 and 4, no single ideal price
8 Setting the analytical background
index can be used as an optimal deflator specifically for banking and insur-
ance firms or industries. In highly disaggregated studies that focus on branches
of an individual firm or a few firms within a particular banking or insurance
industry segment, it is possible to use directly observable prices of individual
output items. It may also be feasible to construct price measures using observ-
able data. However, the performance analysts have struggled to locate such a
price index and have instead employed the aggregate CPI and GDP deflators.
Theoretically, these two deflators are far from the reality of price dynamics in
the banking and insurance industries. However, the lack of data on a suitable
price index has forced such a choice in the literature. An ideal output measure
is thus expected to have a corresponding price index which is well-defined and
rigorously measured. Hardly any measurement of banking or insurance outputs
can qualify this criterion.13

Ability to handle the multi-output nature of banking

and insurance activities
An ideal output measure is expected to account for the multi-output nature
of banks and insurers (Greenbaum, 1967; Clark, 1984). This is especially
important at the aggregate levels of analysis, where industry-level output is
estimated through aggregation across firm-level outputs. The banking and
insurance industry provides many services to its consumers. When a specific
kind of service is aggregated across firms, such as aggregating the total deposit
amount across all the banks, there still can be considerable heterogeneity in
the type of products provided under each of these services. For example,
the amount of total deposits will aggregate banks’ diverse set of facilities and
products to current, savings and term deposit customers. The service vec-
tor may be different for each of these customer segments. However, when
the total deposit amount is used as an output, the service vectors are col-
lapsed and merged through aggregation procedures. Weighting factors are
used to undertake the same. Thus, an ideal output measure should possess
such aggregation weights that the full scale of heterogeneity in banking and
insurance outputs can be captured while aggregating them into industry-level
measures.

Stable and rigid definition

The chosen output measure must exhibit a stable and consistent representation
as an output variable. The methodology chosen to specify the output variable
must not be subject to temporal or spatial instability. If that is the case, the
output measure may switch to an input and vice-versa, which will not allow
dynamic or even comparative-static performance analysis. Such a problem may
occur, for example, while employing the user cost approach to measure bank-
ing or insurance outputs (Berger & Humphrey, 1992).
 Setting the analytical background 9
Fully endogenous to the firm
The preferred output measure should be freely changeable by the banking or
insurance firm. This is necessary for performance issues such as technical, cost,
profit and other efficiencies, and scale and scope economies. The variations
in the ideal output measure should be under complete control of the bank or
the insurance firm under consideration. Subsequently, the industry should be
capable of freely changing the quantity of output at aggregate levels. Thus, the
underlying production function should allow the output measure to be scalable
as per the firm’s or the industry’s choice. When this is not the case, alternative
specifications of the production/cost/profit function are required to consider
the rigidity of output. Berger and Mester (1997) discuss such issues when esti-
mating the “alternative profit function” that allows output quantity constant
while allowing variations in output prices.

Unitary influence factor

Boer (1999) and Inklaar and Wang (2013) elaborate on the influence factor
in output measurement for banking and insurance industries. They argue that
when a particular measure is employed as an output measure, the assumption
is that the underlying service content of the output measure varies proportion-
ally to the quantum of the output measure itself. The relationship between the
observed quantitative variations in the chosen output measure and the unob-
servable changes in the underlying service content is broadly defined as the
influence factor by Boer (1999), who suggested that the preferred influence
factor should be 1. Hence, a 10 percent increase in the amount of deposits
would imply that the quantum of services provided by banks has also increased
precisely in proportion. Inklaar and Wang (2013) criticised the value-based
measures of bank output on this account. They state that the value-based meas-
ure of aggregate banking sector output “assumes that every (real) euro or dollar
corresponds to a constant flow of financial intermediation services over time”
(p. 96). Similar criticisms have been raised by Wang and Basu (2005).

1.3. Final remarks

With the background laid out and the fundamental ideas briefly elaborated,
the following chapters will develop the narrative using the three-dimensional
strategy discussed in section 1.1. This monograph has been divided into three
parts, and each part focuses on the issue of output measurement in the perfor-
mance analyses literature for banking and insurance. Part I examines this matter
from the angle of a detailed survey of literature where approximately 500 high-
quality studies are analysed and compactly summarised separately for the bank-
ing (Chapter 3) and insurance (Chapter 4) literature. The methodology for the
literature review in this work is explained in Chapter 2. After that, Chapter 5
elaborates on some extended issues about the subject that are not given much
10 Setting the analytical background
space in the core review chapters. Part II then expands the scope of this litera-
ture review by incorporating the insights and opinions of selected experts in
the form of edited contributions. Chapter 6 explains the methodology adopted
for this task. Chapter 7 provides the selected contributions representing experts
with diverse policy, academic and professional backgrounds related to output
measurement for the banking and insurance industries. Part III then examines
the issue of empirical implications when different output measures are used for
studying key performance parameters of aggregate industry-level labour and
capital productivities, total factor productivity, technical efficiency and returns
to scale. Chapter 8 undertakes the exercise for the aggregate banking industry,
while Chapter 9 complements the same on the insurance industry. Finally, Part
IV examines the inputs measurement issue in the present context (Chapter 10),
while Chapter 11 summarises and concludes this study.

Notes
1 In economic terms, a firm may be defined as a nexus of contractual relations among
individuals for mutually beneficial economic gains (Jensen & Meckling, 1976).
2 Works by Carl Menger, William Stanly Jevons and Léon Walras laid the foundations of
the newer theory of firm in production economics wherein the distinction between the
tangible and the intangible sectors could be relaxed. Thereafter, Marshall, Clark, Knight
and particularly Cournot helped finalize the profit maximization perspective of firms
producing tangibles and intangibles in competitive markets. In the 1930s, however, the
earlier works of Pierro Sraffa re-emerged in the efforts of Robinson and Chamber-
lin, who modified the strict profit maximization conception of firms and allowed for
non-competitive tendencies in the behaviour of firms. Somehow, the profit maximiza-
tion framework under competitive markets has continued its dominance as the major
approach in empirical works on firm theory. See Curwen (1976) for an elaborated
perspective on these matters.
3 Bylund (2017) suggests that the theory of firm and production are two sides of the same
coin itself. However, one may subsume the theory of firm as a special case of the theory
of production because production can occur outside the firms also.
4 Overlapping production functions imply that the clean decomposition of the produc-
tion process into different stages across the value chain is not feasible. The same vector
of input-output flow may coincide across different stages in the value chain of service
industries. Interlocked production functions imply that in many cases, the individual
production functions of various services (products) being produced by a firm are so
complexly webbed together that disaggregating the production function for total output
across its components may become unrealistic.
5 An illustration with reference to banking services may be provided here. A common
practice in the literature on banking performance is to use loan-loss provisions (non-
performing assets provisions) as an indicator of loan quality. This is netted out of the
gross loans on the books of the bank, and a quality-adjusted output measure is derived.
However, this is a crude measure of loan quality and is one-dimensional because it
assumes that higher NPA provisions imply lower loan quality. In many cases, the loan
quality may not be determined only by the loan-loss provisions; rather the loan losses
themselves may be caused by exogenous shocks which are unrelated with loan quality
per se. This measure also focuses on outcome rather than output. Banks might be highly
diligent in screening and monitoring loans but might face loan losses nonetheless. In
this case, the outcome of the screening and monitoring services is negative, though the
output might still be positive given that the loan losses recorded could have been higher
 Setting the analytical background 11
if the given level of diligence was not followed. Agu (1988) examines the problems in
output measurement in services with special reference to commercial banking firms.
6 The general concept of intermediation is captured by the so-called ‘intermediation
hypothesis’ as examined in Spulber (2009). The author explains that intermediation
occurs when consumers do not exchange directly but via firms. Such a viewpoint con-
siders firms as intermediaries by definition. However, intermediation has a more spe-
cific conception wherein not all firms are intermediaries. Under this perspective, firms
that transform the resources purchased from one set of economic agents and sell these
transformed services to another set of economic agents are regarded as intermediaries.
Banking and insurance firms qualify for this definition.
7 Banking and insurance firms are inherently different. Despite both of them being inter-
mediaries in the larger financial system, the nature of their production and services are
very different. This fact is recognized throughout this work. Thus, both the industries
are treated separately as can be ascertained from the index of contents itself.
8 Endogeneity implies that the output variable is determined within the production func-
tion and that the variations in output are not explained significantly by any factors other
than inputs and residual technology. However, this may not always be the case.
9 The input-output dilemmas in banking and insurance output measurement are elabo-
rated on in Chapters 3 and 4, respectively.
10 This feature is also required by various frontier methods, such as the isotonic assumption
in DEA which states that outputs and inputs must possess positive correlation (Nourani
et al., 2017).
11 Finance and accounting disciplines generally utilize other measures, such as market capitali-
zation, for example, as a representation of the level of activities and size of a firm/industry.
Economists, on the other hand, prefer measures that closely correspond to the real produc-
tion by the firm or the industry. This is due to the recognition by economists of the fact that
financial size and economic size may be different on account of many non-real forces (such
as market speculation, boom-and-bust cycles, and other sources of noise in the financial
system) that may cause divergence between the economic and financial scales.
12 This is also important because many researchers use a single measure of output in bank-
ing and insurance literature rather than a vector. If the chosen variable has a strong posi-
tive correlation with the other measures that have not been included, the performance
parameters could still be helpful and provide meaningful, practical insights without
much of an information loss.
13 Some researchers have stressed the need for output measures that do not contain any
price component. Such volume measures of output are generally difficult to locate in the
case of many advanced economies, let alone the emerging market economies (EMEs).
In reality, the value-based measures dominate the banking and insurance ­literature and
these are mainly composed of price and quantity elements. The removal of the effects
of the price component from a composite value measure is the problem of deflation in
banking and insurance performance literature.

References
Adams, R. M., Röller, L.-H., & Sickles, R. C. (2002). Market power in outputs and inputs:
An empirical application to banking. Wissenschaftszentrum Berlin (Discussion Paper FS IV
02–33). https://nbnresolving.org/urn:nbn:de:0168-ssoar-113399
Agu, C. C. (1988). On the performance analysis: The definition and measurement of bank
output in Nigeria. Savings and Development, 12(2), 107–126.
Berger, A. N., & Humphrey, D. B. (1992). Measurement and efficiency issues in commercial
banking. In Output measurement in the service sectors (pp. 245–300). University of Chicago
Press. www.nber.org/books/gril92-1
12 Setting the analytical background
Berger, A. N., & Mester, L. J. (1997). Inside the black box: What explains differences in the
efficiencies of financial institutions? Journal of Banking & Finance, 21, 895–947.
Blois, K. J. (1985). Productivity and effectiveness in service firms. Service Industries Journal,
4(3), 45–60.
Boer, S. D. (1999). A volume index for the output of the Dutch banking industry based on quantity
indicators. Statistics Netherland (National Accounts Occasional Paper No. NA-094).
Brockett, P. L., Cooper, W. W., & Golden, L. L. (2005). Financial intermediary versus
production approach to efficiency of marketing distribution systems and organisational
structure of insurance companies. The Journal of Risk and Insurance, 72(3), 393–412.
Bylund, P. L. (2017). The problem of production: A new theory of firm. Routledge.
Campbell, T. S., & Kracaw, W. A. (1980). Information production, market signalling, and
the theory of financial intermediation. The Journal of Finance, 35(4).
Clark, J. A. (1984). Estimation of economies of scale in banking using a general-
ized functional form. Journal of Money, Credit and Banking, 16(1), 53–68. https://doi.
org/10.2307/1992648
Curwen, P. J. (1976). The theory of the firm (Vol. 1). MacMillan. https://doi.org/10.1007/
978-1-349-15645-0
Eling, M., & Luhnen, M. (2010). Efficiency in the international insurance industry:
A cross-country comparison. Journal of Banking & Finance, 34, 1497–1509. https://doi.
org/10.1016/j.jbankfin.2009.08.026
Gaganis, C., Hasan, I., & Pasiouras, F. (2013). Efficiency and stock returns: Evidence from
the insurance industry. The Journal of Productivity Analysis, 40, 429–442. https://doi.
org/10.1007/s11123–013–0347-x
Greenbaum, S. I. (1967). Competition and efficiency in the banking system – empirical
research and its policy implications. Journal of Political Economy, 75(4), 461–479.
Hughes, J. P., & Mester, L. J. (1993). A quality and risk-adjusted cost function for banks:
Evidence on the ‘too big-to-fail’ doctrine. The Journal of Productivity Analysis, 4, 293–315.
Inklaar, R., & Wang, J. C. (2013). Real output of bank services: What counts is what
banks do, not what they own. Economica, 80(317), 96–117. https://doi.org/10.1111/j.
1468–0335.2012.00942.x
Jensen, M. C., & Meckling, W. (1976). Theory of the firm: Managerial behavior, agency
costs and ownership structure. Journal of Financial Economics, 3, 305–360.
Nourani, M., Devadason, E. S., & Chandan, V. (2017). Measuring technical efficiency of
insurance companies using dynamic network DEA: An intermediation approach. Techno-
logical and Economic Development of Economy, 24(5), 1909–1940. https://doi.org/10.3846/
20294913.2017.1303649
Spulber, D. F. (1999). Market microstructure: Intermediaries and the theory of the firm. Cambridge
University Press.
Spulber, D. F. (2009). The theory of the firm: Microeconomics with endogenous entrepreneurs, firms,
markets, and organisations (1st ed.). Cambridge University Press.
Vuorinen, I., Järvinen, R., & Lehtinen, U. (1998). Content and measurement of productiv-
ity in the service sector: A conceptual analysis with an illustrative case from the insurance
business. International Journal of Service Industry Management, 9(4), 377–396. http://dx.doi.
org/10.1108/09564239810228876
Wang, J. C., & Basu, S. (2005). Risk bearing, implicit financial services, and specialisation in the finan-
cial industry. Federal Reserve Bank of Boston Public Policy (Discussion Paper No. 06–3).
Wise, W. (2018). Life insurance company efficiency: Best method and proxies. Insurance
Markets and Companies, 9, 6–19. http://dx.doi.org/10.21511/ins.09(1).2018.02
Part I

Survey of evidences
Theoretical and empirical issues
2 Qualitative and quantitative
aspects of the survey
of literature

2.1. Sample versus population dimensions of literature survey

Theoretically, a literature review can broadly be conducted under the ad hoc
and the systematic approaches. Ad hoc methods primarily rely on the subjec-
tive understanding of the reviewer regarding the various critical elements of
a survey of the literature. These elements include the sources of the stud-
ies, choice of databases for performing search queries, keywords to be used,
inclusion-exclusion criteria to be employed, and the issues analysed under the
review exercise.1 On the other hand, systematic approaches establish a replica-
ble format for constructing the sample of studies which shall then be subjected
to rigorous theoretical and empirical analysis.2 In both approaches, a sample of
studies is used to derive insights into the significant agreements, disagreements,
theoretical issues, empirical methodologies, causal connections among varia-
bles of interest and other prevalent dimensions in the extant wisdom. However,
the reliability of the insights from both these approaches differs considerably
due to the role of subjectivity and associated biases that would inevitably be
embodied in the nature of the sample that is selected for review. While both
the approaches are subject to such problems, the systematic approaches allow
the readers to gauge the broad extent of these biases due to their replicability
and use of well-defined procedures. In the ad hoc methods, a reader cannot
locate the significant sources of biases due to the lack of any structured process
in selecting and analysing studies. Thus, systematic approaches are more reliable
when the literature survey is the primary focus. In contrast, ad hoc approaches
may be used in empirical works that wish only to provide a broad glimpse of
major works undertaken earlier. Given the scope and objectives of this work, a
broadly systematic approach to the literature survey is adopted herein. Further
discussion shall clarify the nature of the review work undertaken in this work.
Irrespective of the approach to literature review adopted, any review work
engages in constructing a set of studies that the reviewer believes truly and suf-
ficiently represents the extant wisdom on the topic of interest. Unless an issue
is highly specialised,3 it is challenging to undertake a census of the studies, and
thus, the reviewer must inevitably select a sample of studies for assessment. This
is particularly important for social sciences and economics, where elimination

DOI: 10.4324/9781003149828-3
16 Survey of evidences
by scientific experiments of older models is essentially not permissible.4 Thus,
out of the population of studies on a particular topic, a set of studies is selected
using an ad hoc or a systematic procedure. We may term that as the sample of
interest. Recognising this population-sample dimension in a review exercise
allows one to consciously construct a set of studies that minimises the biases that
a high degree of subjectivity can bring into the review.5 While such biases will
not be eliminated, their extent can be kept under tolerable limits, thus allowing
the reviewer to summarise the extant wisdom with more excellent reliability.
The underlying population of ideas that a reviewer aims to summarise can be
better captured by the selected sample if an explicitly defined procedure is fol-
lowed from the very beginning of the exercise. Moreover, such an approach
allows the reviewer and their readers to recognise the limits of the insights drawn
from the review work. Despite both the theory and technology for conducting
such reviews, minimal studies have followed such an approach to assessing the
chosen theme. One of the objectives of the present work is to fulfil this gap.

2.2. Review strategy
The importance of a scientifically undertaken review of literature cannot be
overstated. As Gough et al. (2012, p. 3) note, “Reviews can inform us about
what is known, how it is known, how this varies across studies, and thus also
what is not known from previous research”. The extent to which these expec-
tations can be fulfilled depends on how well the review exercise is framed from
the very beginning. Thus, this section clarifies the stages through which the
sample of studies was constructed.6
The entire process of building the sample was divided into the following
stages. In stage I, the keywords, databases7 and inclusion criteria were specified.
The primary concern was to collect studies focused on various performance
measures of banking and insurance activities. This included studies centred on
productivity, efficiency, profitability, competition and related aspects of banking
and insurance activities. The first set of studies was collected in stage II, total-
ling approximately 500 studies for banking and insurance-related issues com-
bined. In stage III, a rolling sample was formed by selecting studies from the
reference list of each of these 500 studies. This was done using the pre-defined
inclusion criteria. This increased the sample size to approximately 1,000 stud-
ies, covering an extensive range of banking and insurance performance analysis
issues and output measurement. The sample was cleaned for duplication, rep-
etition, and irrelevance in stage IV. This yielded 872 studies. In stage V, a fixed-
weighted quality score was developed by quantifying five main parameters so
that the scores ranged between 0 and 100 on an ordinal scale. The parameters
included the publisher (weight assigned was 0.3), database where the source
was listed (weight given was 0.2), citations for the individual literature as per
popular databases (weight assigned was 0.1) and whether the literature focused
on issues of interest (weight given was 0.1) and the depth of discussion on the
chosen output variables (weight assigned was 0.3).
 Qualitative and quantitative aspects of survey of literature 17
The 872 studies were reviewed in detail and ranked on these five param-
eters. The weighted score for each of these studies was estimated. After that,
studies were grouped into seven groups based on the score value: very high,
high, moderately high, moderate, moderately low, low, and very low.8 Stud-
ies scoring below the moderate group were eliminated from the sample. This
yielded approximately 650 studies for banking and insurance issues combined.
As the scores also varied within each group, a threshold score of 450 was
selected as a cut-off value. All studies below this score from within the moder-
ate group were further eliminated. This yielded 450 studies – 260 on banking
performance analysis and 190 on insurance performance analysis. This work
has included these studies, though all 872 pieces of evidence were reviewed.
This was done due to space restrictions.9 The 872 studies on various aspects
assumed critical for output measurement analysis were examined in detail.
The quantitative summary of the finally selected 450 studies is presented in
the next section in Tables 2.1 and 2.2. Some of the primary criteria used are
reflected therein.

2.3. Quantitative summary of the selected sample

of studies
Tables 2.1 and 2.2 summarise the nature of studies that have been selected for
analysing banking and insurance output measurement.10 The range of cover-
age in terms of sources is visible in the tables. Not only journal articles but
also PhD theses and books are covered. However, journal articles dominate
the sample. For both the samples, as shown in Tables 2.1 and 2.2, studies that
focus on industry-level analysis using firm-level data dominate the sample.
This is possibly a reflection of the extensive use of panel data in the literature.
Firms as a unit of study also correspond closely to the main empirical methods
employed in the literature. Performance analysis is generally applied to data on
individual decision-making units, and this is reflected in the analytical approach
used in the literature. Efficiencies, productivities and scale economies dominate
the sample on banking related studies. Methodologically speaking, traditional
data envelopment analysis (DEA), stochastic frontier approach (SFA), ordinary
least squares (OLS) method and ratio-based empirical approaches dominate
the selected sample, as reflected in Table 2.1. In contrast, analysis of efficien-
cies and productivity occupies the insurance sample’s largest share, as shown
in Table 2.2. Regarding empirical methodologies for insurance performance
analysis, the distribution of studies in the insurance sample is mainly similar to
that of the banking sample. Regarding the geographical distribution, advanced
and emerging economies are the main focal points for the selected studies in
both the samples, as shown in Tables 2.1 and 2.2. For the banking sample,
advanced economies have a sizeably larger share, though coverage of emerg-
ing economies is quite broad for both the samples. The geographical spread
achieved here is extensive and has not been covered in any literature review in
the present context.11
18 Survey of evidences
Table 2.1 Quantitative summary of the sample studies on banking performance

Dimensions Sub-dimension Frequency

Source Journal 229 (88%)

Working Paper 14 (5%)
Book Chapter 5 (2%)
Discussion Papers 4 (2%)
PhD thesis 3 (1%)
Others (conference papers; edited volume; 5 (2%)
occasional paper; unpublished)
Analytical Industry level 16 (6%)
approach Industry-level (using firm-level and other unit- 202 (78%)
level data)
International (using country-level aggregate data) 2 (1%)
International (using firm-level and other unit- 10 (4%)
level data)
Unit level (branch and individual accounts) 3 (1%)
Sectoral 14 (5%)
Review of evidence 7 (3%)
Others (theoretical analysis, mathematical 6 (2%)
economics, and other approaches)
Rating based on Very high 89 (34%)
ranking scale High 95 (37%)
Moderately high 60 (23%)
Moderate 16 (6%)
Decadal 1950–1959 1 (0.4%)
distribution 1960–1960 5 (1.9%)
1970–1979 14 (5.4%)
1980–1989 19 (7.3%)
1990–1999 59 (22.7%)
2000–2009 79 (30.4%)
2010–2019 61 (23.5%)
2020 and later 22 (8.5%)
Issue of analysis Technical efficiency 60
Cost efficiency 59
Economies of scale (including returns to scale) 48
Total factor productivity 41
Bank output measurement 23
Profit efficiency 17
Economies of scope 13
Performance analysis 13
Profitability 13
Allocative efficiency 11
Scale efficiency 11
Partial factor productivity 9
Others 8
X-efficiency 6
Technical change 6
Pure technical efficiency 5
Revenue efficiency 5
FISIM (financial intermediation services 3
indirectly measured)
Market power 3
 Qualitative and quantitative aspects of survey of literature 19

Dimensions Sub-dimension Frequency

General efficiency analysis 3
Cost elasticity 3
Overall technical efficiency 2
Competition 2
Technology gaps ratio 2
Cost gap ratio 1
Quality efficiency 1
Fuzzy efficiency 1
Empirical Non-parametric
framework DEA (data envelopment analysis)
Traditional DEA 71
Emerging DEA 11
DFA (distribution free approach) 8
TFA (thick frontier approach) 2
Parametric
SFA (stochastic frontier approach) 32
MLE (maximum likelihood)
One-stage MLE 21
Two-stage MLE 1
Regression
OLS (ordinary least squares) 24
SUR (seemingly unrelated regression) 22
Panel regression 13
Fuzzy regression 1
COLS (corrected OLS) 1
GMM (generalised method of moments) 2
Others
Index numbers
Traditional 3
Emerging 3
Ratios 24
Others 18
Geographical Advanced economies 148
distribution Emerging economies 96
Developing and non-emerging 6
Underdeveloped 2
Transition 5
Global 10
Stock-flow Only stock 135 (52%)
dimension of Only flow 31 (12%)
the chosen Both 76 (29%)
output Ratio of flow to stock 14 (5%)
variables Ratio of stock to stock 2 (1%)
Others 2 (1%)
Nominal-real Nominal 169
dimension of Real (constant price) 70
the chosen Real (volume) 23
output
variables

(Continued)
20 Survey of evidences
Table 2.1 (Continued)

Dimensions Sub-dimension Frequency

Nature of data Cross sectional 58
Time series 26
Panel
Unbalanced 18
Balanced 4
Unclear 119
Pooled 15
Temporal Mean years 10
dimension of Median years 8
the sample Minimum years 2
periods (No. Maximum years 37
of years)
Frequency of Annual 187
sample period Quarterly 8
Monthly 2
Daily 2
Deflators Consumer Price Index 25
employed for Gross Domestic Product deflator 28
the chosen Gross National Product deflator 4
output Wholesale Price Index 4
variables Others 4
Output Loans 196 (75%)
measurement Deposits 107 (41%)
dimensions Investments 83 (32%)
Income measures 71 (28%)
Non-traditional measure (off-balance-sheet 66 (25%)
output)
Assets 84 (32%)
Sectoral and national income measures 14 (5%)
Notes: 1. The total number of studies is 450, of which 260 are on banking performance analysis. 2. In
some cases, the percentage (%) values shall not be equal to 260 due to multiple counting of a given study
in several sub-dimensions and rounding off of percentage values.
Source: Author’s survey of the literature.

A contrasting feature between both the samples is that the banking sample
has primarily employed the stock measures of banking output while the insur-
ance sample uses flow measures. This is probably a reflection that stock output
measures are preferred in banking literature, given the historical development
of banking output measures. However, nominal output measures are the most
preferred choice in both samples. This again reflects the lack of an optimal
price index for banking and insurance services. There also rages a debate on
conceptualising the price of banking and insurance outputs and the optimal
method to measure the same using micro-level data. This is reflected in the
dominance of nominal output variables in the literature. Lastly, as summarised
in Table 2.1, the banking sample suggests that loans and deposits are the most
frequently employed output measures for banking. At the same time, premiums
 Qualitative and quantitative aspects of survey of literature 21
Table 2.2 Quantitative summary of sample studies on insurance performance

Dimensions Sub-dimension Frequency

Source Journal 158 (83%)

Working paper 7 (4%)
Book chapter 9 (5%)
Full book 5 (3%)
PhD thesis 4 (2%)
Others (conference papers; monograph; other 5 (3%)
dissertations; unpublished)
Scope Industry level using firm-level data 136 (72%)
Others 18 (10%)
Industry-level 8 (4%)
Sectoral level 8 (4%)
Industry-level using other units of analysis (policies; 6 (3%)
intermediaries; groups of firms; geographical areas)
Firm-level 6 (3%)
Review of evidence 4 (2%)
Sectoral using firm-level data 2 (1%)
Rating Very high 49 (26%)
High 45 (24%)
Moderately high 42 (22%)
Moderate 52 (28%)
Decadal Before 1989 4 (2%)
distribution 1990–1999 12 (6%)
2000–2009 37 (20%)
2010–2019 125 (67%)
2020 onwards 9 (5%)
Main issue Cost efficiency 43
Technical efficiency 55
Advertising efficiency 1
Revenue efficiency 13
Profit efficiency 6
Profitability 5
Economies of scale and scope 8
Total factor productivity 51
Partial factor productivity 13
X-efficiency 1
Performance analysis 10
Insurance output measurement 3
Underwriting cycle 7
Review of evidence 1
Allocative efficiency 9
Others 7
Empirical Non-parametric
framework DEA
Traditional 86
Bootstrapped 5
Dynamic DEA Network 2
Fuzzy analytic hierarchic process 1
TFA 1

(Continued)
22 Survey of evidences
Table 2.2 (Continued)

Dimensions Sub-dimension Frequency

Parametric
SFA 18
Bayesian SFA 1
Time series models 4
OLS 6
Other regression models 4
Others
Index-based methods (including ratio-based 25
methods)
Geographical Advanced 91
Distribution Emerging 81
Developing and non-emerging 6
Underdeveloped 3
Gulf Cooperation Council 1
Islamic 1
Global 1
Others 9
Stock-Flow Stock 61
dimension of Flow 153
the output Ratio of flow to flow 2
measure Ratio of flow to stock 14
Ratio of stock to stock 6
Nominal-real Nominal 85
dimension Real (constant price) 58
of output Real (constant currency) 1
measure Real (volume) 9
Unclear 4
NA (not applicable) 25
Nature of data Cross sectional 20
Time series 15
Panel 123
Unbalanced 9
Balanced 6
Unclear 108
Pooled 4
Deflators Aggregate CPI 51
employed GDP deflator 4
for chosen Aggregate WPI 1
output Self-constructed 1
variables Output-specific deflators 1
Unclear 6
Sectoral Focus Life insurance 102 (54%)
Nonlife insurance 58 (31%)
Specific individual industries 18 (10%)
Property-liability 30 (16%)
Aggregate 20 (11%)
Takaful 4
Financial services sector 2
 Qualitative and quantitative aspects of survey of literature 23

Dimensions Sub-dimension Frequency

Output Premiums 92 (49%)
variables Investments/financial assets 96 (51%)
Losses 41 (22%)
Claims 37 (20%)
Investment income 34 (18%)
Benefits paid 26 (14%)
Technical reserves 17 (9%)
Number of policies sold 7
Expenses 3
Sectoral and national income 3
Amount of time 2
Number of products 2
Number of accidents 1
Number of contracts 1
Number of claims 1
Notes: 1. The total number of studies is 450, of which 190 are on insurance performance analysis. 2.
In some cases, the percentage (%) values shall not be equal to 190 due to multiple counting of a given
study in several sub-dimensions and rounding off of percentage values.
Source: Author’s survey of the literature.

and investments are the most commonly used output measures for insurance
activities. The choice of these variables must largely be seen in terms of data
availability. Unsurprisingly, the sectoral and national income measures of bank-
ing and insurance outputs are not very popular. This is possibly a reflection
of the economic versus national income-based output measurement debates
ongoing since the early 1980s (Wolff, 1991).

2.4. Problems in the purposive selection of studies12

The ad hoc (or traditional) approaches to literature review are characterised by
an unstructured and largely subjective selection of studies. They also generally
lack recognition of the sample-versus-population dimension in their survey.
This lends the insights derived from such reviews subject to many biases. These
biases include, but are not limited to, the researcher bias (Machi & McEvoy,
2022), whereby the reviewer tends to select studies that conform to her sub-
jective ideas and preconceived notions; survivor bias, whereby the reviewer
tends to choose studies that have significant citations and thus end up ignoring
conflicting views that may not have garnered equal attention; and publication
bias (Andrews & Kasy, 2019) whereby statistically significant findings have a
higher chance of being included in the sample. Such biases may distort the
true underlying consensus and disagreements in the literature. They also make
the selected sample unrepresentative of the underlying population of studies
that it tries to summarise. While this may not be a problem for empirical
researchers aiming to provide a brief commentary on the significant patterns
24 Survey of evidences
in the literature, it can be a serious limitation for reviews that aim to provide
rigorous analysis of the research gaps in the extant wisdom on a topic. Specifi-
cally, in banking and insurance output measurement, very few studies exist that
undertake a full-scale survey. Furthermore, those existing reviews are largely
unclear on their literature selection and sample construction strategies. This
makes deriving any kind of generalisation from such reviews very difficult and
subject to errors. While it is impossible to eliminate such problems, they can
be minimised if the reviewer adopts a structured process and states the same
explicitly in clear and straightforward terms.
One must be careful not to equate subjectivity with bias. There are situations
when subjective selection of studies can be superior to structured selections.
Highly specialised issues are generally well-captured by the expertise of the
reviewers in that area. However, the subject matter of this work is a more gen-
eralised issue and has been discussed and debated by a large number of research-
ers from diverse disciplines and ideological schools. Thus, a structured selection
of studies with well-defined parameters can help reduce the highlighted biases
and possibly allow better generalisations for future research.

2.5. Background on further sections

The following two chapters examine the major output measurement issues in
banking and insurance activities. The focus of these chapters is to condense
the voluminous literature into broad generalisations that can provide meaning-
ful insights into the consensus and disagreements on this topic. The primary
output variables and output specification approaches are analysed. After that,
thematic discussions on various issues derived from the survey are discussed.
The idea that a survey of the literature is sufficient in itself is challenged in
Part II by incorporating contributions from experts in the areas related to the
theme of this work. These experts were carefully selected and interviewed and
then were requested to provide their opinions on the central issues in banking
and insurance output measurement. The expert contributions serve as first-
hand accounts of the beliefs of those involved in critical positions in policy-
making and academic research. The experts’ profiles are provided in Chapter 7,
and the selection and interview strategies are examined in Chapter 6. An analy-
sis of output measurement issues in banking and insurance shall be incomplete
without examining how alternative measures change the critical performance
parameters. Six major emerging market economies are selected for this pur-
pose: Brazil, Russia, India, Indonesia, China, and South Africa (the so-called
BRIICS). Aggregate banking and insurance industries are analysed instead of
firms to give a fresh perspective on this issue as the literature has not paid much
attention. The same is also motivated by data constraints.13 After that, Part IV
begins by discussing another debated issue in banking and insurance perfor-
mance literature – namely, the measurement of inputs. The same is discussed in
Chapter 10, along with some additional issues. Finally, Chapter 11 concludes
the work by recognising the limitations, suggesting the possible directions for
 Qualitative and quantitative aspects of survey of literature 25
future research and deriving tentative conclusions only to be challenged by
more able researchers in the coming time.

Notes
1 Given that an increasingly large volume of literature in social sciences is available in
digital form, the discussion here assumes that digital literature search is the leading
source for selecting studies. In some cases, mainly when the topic is highly specialized,
searching in physical databases such as libraries shall be required. The discussion here
can broadly be applied in such cases, though considerations of time and costs are not
delved deeper here.
2 See Machi and McEvoy (2022) for a compact analysis of the significant elements of a
systematic literature review. The authors divide the literature review process into six
stages covering various critical aspects of undertaking such an exercise. They stress cri-
tiquing the literature to locate research gaps. However, this aspect has not been under-
taken in this work. The primary aim of the present work is to summarize and condense
the voluminous literature into a single reference. Hence, critiquing it in detail would
have expanded the scope beyond the permissible space.
3 This implies that ad hoc approaches shall be more efficient in selecting the set of
studies to be analysed because the literature on such specialized topics is expected to
be ­distributed relatively compactly across a limited number of sources. Problems in
using such an approach emerge when the subject of interest is more generalized and
multidisciplinary.
4 Economic analysis can be approached from two perspectives – the absolutist and the
relativist (Blaug, 1990). The absolutist approach considers economic theory from a pure
empiricist angle. In this approach, older models are slowly discarded with empirical
evidence, and more refined, and scientifically sounder models remain in practice. How-
ever, the relativist approach allows for the fact that current empirical proofs may not be
sufficient to eliminate one model of economic behaviour in favour of another. Older
models may re-emerge later in the same or slightly modified forms to explain new data.
The issue of banking and insurance output measurement must be approached from a
relativistic angle rather than an absolutist one. Given the lack of universal consensus on
even the core set of services provided by banks and insurers, elimination by empiricism
is not feasible in this context. Older approaches may very well explain newer data, and
current strategies may be unsuitable for explaining present reality.
5 This also allows a reviewer to use quantitative meta-analytic methods, such as meta-
regression, to summarize the consensus and disagreement in the literature.
6 The present discussion only highlights this process in brief due to space constraints. The
detailed methodology may be obtained from the author upon request.
7 The initial databases included WorldCat, Google Scholar, Scopus, Web of Science and
JSTOR. Multiple people and institutions enabled access to these databases and other
premium sources, which has been acknowledged in the Acknowledgements section.
8 The rankings are not a final statement of the actual quality of the work. This is a sub-
jective matter. This ranking score method is used only as a tool to classify the more
relevant studies from the less relevant ones. Hence, one must interpret these scores and
categories in terms of relevance for this study rather than as a measure of their research
quality. Some degree of subjectivity is inevitably involved in such a scoring process, but
the attempt has been made to reduce this to the best possible extent.
9 The detailed scoring process and scores for all 872 studies and additional notes are hap-
pily available from the author on request.
10 The distribution of literature across sources as found here has been corroborated by
other studies. Studies of such scope are limited and include the one by Sharma et al.
(2013). They, too, find the same journals in their list of top published journals.
26 Survey of evidences
11 Review works on banking and output measurement and studies that undertake an
empirical sensitivity analysis of performance estimates to alternative output variables are
discussed in Chapters 3 and 4.
12 This section has been included based on the valuable suggestions by the anonymous
reviewer, for which the author is deeply thankful.
13 The data source called BankFocus (earlier BankScope) of Moody’s Analytics has been
increasingly used for obtaining firm-level data for banking and insurance industries
across the advanced and emerging economies. This data source is constructed from
official data and probably other sources. This database has extensive coverage. However,
there are concerns about using this database. See Bhattacharya (2003). This was one
reason for avoiding the use of this database. Moreover, performance analysis at aggregate
levels is largely missing in the literature. Hence, this was another motivation to utilize
aggregate industry-level data from official public sources. Future works by the author
shall incorporate more disaggregated databases for empirical sensitivity analysis, such as
that by BankFocus.

References
Andrews, I., & Kasy, M. (2019). Identification of and correction for publication bias. Ameri-
can Economic Review, 109(8), 2766–2794. https://doi.org/10.1257/aer.20180310
Bhattacharya, K. (2003). How good is the bankscope database? A cross-validation exercise with cor-
rection factors for market concentration measures. BIS (Working Paper No. 133).
Blaug, M. (1990). Economic theory in retrospect (4th ed.). Cambridge University Press.
Gough, D., Oliver, S., & Thomas, J. (2012). An introduction to systematic reviews (1st ed.). SAGE.
Machi, L. A., & McEvoy, B. T. (2022). The literature review: Six steps to success. Corwin, SAGE.
Sharma, D., Sharma, A., & Barua, M. K. (2013). Efficiency and productivity of banking
sector: A critical analysis of literature and design of conceptual model. Qualitative Research
in Financial Markets, 5(2), 195–224. http://dx.doi.org/10.1108/QRFM-10-2011-0025
Wolff, E. N. (1991). Productivity growth, capital intensity, and skill levels, in the U.S. Insur-
ance industry, 1948–86. The Geneva Papers on Risk and Insurance. Issues and Practice, 16(59),
173–190. www.jstor.org/stable/41952062
3 Review of evidence
on banking output
measurement
Global perspective

3.1. Introduction
Concerns with bank output measurement are not new. One can trace the dis-
cussions on this matter to as early as the late 1950s. Notable early work on this
subject matter can be found in Speagle and Kohn (1958), who summarize the
main challenges in constructing an optimal measure of bank output. The fac-
tors highlighted by them include the heterogeneity of products, change in the
composition of the output mix, product innovations and changes in the qual-
ity of the banking services. These concerns remain as profound today as they
were decades ago. These concerns emerge due to the multi-product nature of
banking firms. More concretely, such concerns are a result of the larger philo-
sophical issues in the measurement of economic phenomenon (Boumans, 2007).
The intangibility of production further complicates the matters because cleanly
separable production functions cannot be immediately located in the banking
value chain. The production process is essentially interlocked, and the underly-
ing production function is itself a product of many interrelated and overlapping
production functions spread across different stages in the value chain. Such issues
have been well-recognized in the banking literature, which has given rise to
alternative approaches to output specification. These approaches are examined
in section 3.3. Underlying these specification approaches are different measure-
ments of bank output.1 Some methods favour volume measures – those that are
in quantity terms without any explicit price element. Others profess value meas-
ures – those that are in monetary terms with both quantity and price elements.
Within both these approaches, different lines of thought have evolved. Such
aspects are reviewed and examined in section 3.4. Finally, the most pressing theo-
retical and empirical issues that have occupied the intense attention of researchers
are enlisted and discussed in section 3.5, after which an appendix highlights the
evolution of bank output measurement from 1957 to 2022, covering 260 studies.

3.2. Empirical sensitivity of banking performance

estimates to alternative outputs
The earliest literature highlighted two significant measures of bank output:
balance sheet-based and income account-based (Kalish & Gilbert, 1973;

DOI: 10.4324/9781003149828-4
28 Survey of evidences
Agu, 1988).2 Since then, balance sheet measures have dominated the literature,
corroborating the large share of stock measures in the studies sample, as shown
in Table 2.1. Despite early warnings against the sole reliance on balance sheet
items, their use has continued to be pervasive in banking literature. Speagle and
Kohn (1958) provided a way to reconcile this dichotomy of stock-flow meas-
ures by suggesting the use of index measures of bank output. While such efforts
have been used in literature, their use is sparse at best. Since then, literature on
banking performance has employed a rich set of variables that have been justi-
fied as the correct measures of the underlying output concept.
While these matters are reviewed in the following sections, one exciting
development has been the tests of empirical sensitivity of performance esti-
mates to alternative bank output measures. By the early 1980s, it was explicitly
recognized that there could not be a single measure of bank output due to
the multi-product nature of banking activities. This led to the need to test the
variations in performance estimates brought by alternative output measures
employed in the literature. Clark (1984) was one of the earliest attempts in
this direction. The author used lending output, lending plus non-lending out-
put and earning assets output to estimate the sensitivity of economies of scale
estimates to these alternative output measures. Significant differences in scale
economies were found on this account. An early systematic attempt to compare
the effects of economic versus national income measures3 on performance esti-
mates was undertaken by Subramaniam (1985).4 The author undertook a sen-
sitivity analysis of scale economies in the Australian banking industry using the
national income measure, total deposits, total assets (physical plus financial) and
total financial assets. In level form, the estimates of scale economies using these
alternative measures displayed a strong positive correlation, while in growth
rate form, the correlation reduced drastically. Since then, a sizeable set of litera-
ture has undertaken such sensitivity analysis. In the 1980s, authors were gener-
ally concerned with testing the sensitivity of scale and scope economies using
different balance sheet measures. This included notable works by Berger et al.
(1987) and Lawrence (1989), while Hunter and Timme (1986) is a noteworthy
work with regards to technical efficiency. In the 1990s, the authors’ attention
was expanded to undertake similar exercises in other advanced economies such
as Finland (Kolari & Zardkoohi, 1990) and Norway (Berg et al., 1991). The
US continued to garner the most attention in this respect, and notable works
were undertaken by Hunter and Timme (1995) and Berger and Humphrey
(1997). Berger and Humphrey (1997, p. 198) summarize the crux of the mat-
ter when they assert that “inferences regarding efficiency may be importantly
affected by how the output is measured, a result usually less dependent upon
investigator choice than the availability of data”.
The 2000s saw a continuation of this analytical preoccupation. Athanas-
sopoulos and Giokas (2000) undertook empirical sensitivity tests to estimate
branch-level intermediation efficiency for alternative volume and value meas-
ures of branch output. A notable attempt during this period was made by
Tortosa-Ausina (2002), who examined the sensitivity of non-parametric cost
 Review of evidence on banking output measurement 29
efficiency estimates under the intermediation and value-added approaches.5
The author studied the distribution of the cost efficiency estimates under dif-
ferent output specifications and found that the mean and variability of the effi-
ciency distributions differ substantially across alternative output specifications.
In terms of emerging economies, works by Kumbhakar and Sarkar (2003) and
Sathye (2003) undertake a sensitivity check for productivity and efficiency
estimates respectively of Indian commercial banks. An essential feature of the
former study is that it used the number of branches as an output variable and
traditional stock variables. However, the latter study used flow output variables
regarding interest and non-interest incomes and net loans (gross loans net of
non-performing assets). Kumbhakar and Sarkar found significant differences
in the estimates of total factor productivity (TFP) for public and private sector
banks when branches were included as output variables. Sathye (2003) moves
beyond output sensitivity and studies the differences in technical efficiency esti-
mates under alternative inputs-output combinations. Among the two models
that they utilize, which are defined under the intermediation approach, the
efficiency of the public sector is lower than that of foreign banks in one case
and vice-versa in another. A similar attempt was made by Hughes et al. (2000)
for scale economies of US commercial banks, which focused on the issue of
empirically testing the input or output status of deposits. They assert that the
choice of deposits as an input or an output “is a technological question that can
be answered by testing whether the data are consistent with the different tech-
nological roles of outputs and inputs” (p. 6). Their study concludes that depos-
its must be treated as an input. However, this finding is tentative and subject to
change when studying other periods or countries. Some studies also analysed
the differences in technical efficiency estimates using one flow and one stock
output variable. De (2004) was undertaken in this spirit. The author undertook
sensitivity tests for technical efficiency estimates using the Cobb-Douglas sto-
chastic frontier approach (SFA) with two output measures – viz. gross income
(flow) and amount of loans and investments (stock).
The phase of the 2010s saw an improved focus on output sensitivity analysis
and the implications of using alternative parametric and non-parametric meth-
ods in the production analysis of banks. Lozano-Vivas and Pasiouras (2010)
examined the effect of including non-traditional output measures (off-balance-
sheet and non-interest income) on cost and profit efficiencies. They found that
cost efficiency results improved sizably on having non-traditional outputs, but
profit efficiency results showed only a marginal improvement. Similar efforts
were undertaken by Feng and Serletis (2010), who checked for the robustness
of their efficiency and TFP estimates for US banks by additionally including
non-traditional output in their output vector. Both the studies found improve-
ments in their performance estimates, including the off-balance-sheet output.
Guarda et al. (2013) brought a fresh perspective on this matter. They used a
directional distance function to specify inputs and outputs empirically and then
examined the sensitivity of estimated distance function parameters to alterna-
tive specifications of the selected output variables. While these studies generally
30 Survey of evidences
used output variables from both asset and liability sides, Oluitan (2014) ana-
lysed the empirical sensitivity of cost efficiency scores to three different asset
side variables: the amount of loan, amount of investment, and other earning
assets non-interest income. Their findings broadly resonated with the conclu-
sions of past studies. Most of these studies combined loans and investments.
However, Mamonov and Vernikov (2017) checked for the sensitivity of cost
efficiency estimates by adding securities investments as a separate output in
their trans-log cost function. They, too, found that estimates are quite sensitive
to changes in the output vector. Some countries focused on non-advanced
economies. This included Thilakaweera (2016), who examined the various
economic efficiencies and the Malmquist TFP of Sri Lankan commercial banks
under the DEA approach. The stock-flow dimension of output measures was
explicitly analysed by adopting both the intermediation approach (stock) and
the operating profit approach (Flow). The efficiency and productivity estimates
sensitivity to these two output specification approaches are then analysed. Size-
able and statistically significant differences in efficiency and productivity esti-
mates were found between both approaches. Finally, Anwar (2019) examined
the sensitivity of parametrically estimated cost efficiency under SFA to alter-
native asset side outputs in three models. The author found that the cost effi-
ciency estimates were the highest when the standard output specification was
used – namely total loans, investments and non-interest incomes compared to
more disaggregated measures of loans, investments and non-interest incomes.
In recent times, Boda and Piklová (2021) compared the technical efficiency
scores for nine input-output specifications of the intermediation approach, nine
specifications of the production-like approaches and three network-integrated
specifications of the modelled production function. They found that “inter-
mediation input-output specifications tend to produce higher efficiency scores
than production-like specifications” (p. 1551). One can thus conclude that the
empirical sensitivity of performance estimates to different outputs in banking
has been an area of debate for a long time. This has remained unsettled, though
some consensus has begun emerging. These matters will be discussed later,
but it is interesting to note that very few studies have undertaken this exercise
using alternative outputs and alterative empirical methods. This gap is seri-
ous because output sensitivity is expected to change under different empirical
methods. This research gap is broadly addressed in Part III of this work.

3.3. Approaches to output specification in the

banking literature
This section elaborates on the different lines of thought that have developed
on the specification of the “correct” output measure in banking literature. At
the outset, one must accept that output specification is contextual and driven
by different considerations. Primary among them are the researcher’s beliefs
concerning the main functions of banks and the availability of data. While the
latter is outside of the control of an analyst, the former has induced a large
 Review of evidence on banking output measurement 31
amount of debate and disagreements on how to represent banking output cor-
rectly. Different lines of thought have different views. It is not plausible to
collapse the differences in beliefs into water-tight compartments of thoughts.
However, literature has found it necessary to classify authors’ beliefs on out-
put measurement issues into different groups. Each has a specific procedure to
ascertain which variables shall be used as output and the rationale. While many
researchers elaborated on the theoretical basis underlying these approaches,
there have been ad hoc choices on this account too. In several cases, research-
ers simply adopted a particular output specification approach without really
delving into the rationale for their choice. Complete justification is not always
provided, and general beliefs are utilised to define the outputs.6
The schools of thought on output specification in banking literature have
evolved and changed with time and experience. The banking industry has
undergone a vast evolution in its services, institutions that provide those ser-
vices, and policies that govern them. Consequently, different approaches have
emerged in the literature. Early traces of this account can be found in Gor-
man (1969), who delineated two frameworks for output specification in bank-
ing: liquidity and transaction approach. The former “measures real output as a
weighted average of deflated dollars of demand and term deposits” (Geehan &
Allen, 1978, p. 670). The annual average rate earned on bank assets minus the
average annual interest incurred on deposits is used as the weighting factor.
Alternatively, “the transactions approach measures output as a weighted aver-
age of the deflated dollar volume of transactions for the two types of deposits,
the weights being base-year interest forgone on deposits of each type” (Gee-
han & Allen, 1978, p. 671). The value-volume dichotomy is embedded among
these two approaches. In a broader sense, the stock-flow debate in bank output
measurement seems to have flown from the debates of the 1970s on liquidity
versus transaction approaches. After that, another novel7 classification of output
specification approach emerged in Clark (1984), who suggests the following
approaches: un-weighted stock outputs, “technically independent produc-
tion functions” (p. 54) and “weighted index of bank output using information
from the income statement as well as the balance sheet” (p. 54). The author
suggested these approaches to tackle the issue of non-homogeneous bank
production functions. Recognition of value measures, i.e., price and quan-
tity components, increased with time in the literature. Despite the academic
strengths of the volume measures, i.e. those having only a quantity compo-
nent, value measures of output began garnering popularity. Consequently, the
specification approaches that recommended their use became more frequent
in empirical applications. The 1980s concluded with the suggestion of using
both the physical (volume) and value measures together in an indexed measure
of bank output. Todhanakasem et al. (1986) suggested two major approaches
in this context – physical measures-based production functions or production
functions employing multiple output measures via indexing.
During the 1990s, one of the most influential works was undertaken by
Berger and Humphrey (1992). They classified output specification approaches
32 Survey of evidences
into the asset, user cost, and value-added approaches. Coupled with the produc-
tion and intermediation approaches already in use, these five approaches laid the
foundations for all future empirical works on banking performance analysis. In
the same spirit, Grigorian and Manole (2002) examined the major approaches
as suggested in Berger and Humphrey (1992). Several works emerged during
the 1990s and 2000s using these approaches in defining output vectors in bank
production analysis. However, the gap between these theoretical economic
approaches and the national income accounts remained wide. Hence, in the
2010s, Daisuke et al. (2011a, 2011b) contextualized the economic approaches
regarding the national income approach, culminating in the FISIM method
of the System of National Accounts (SNA) 2008. Another strongly emerging
trend during the 2010s was the shift from single-stage to a multi-stage theo-
rization of the banking and insurance production process. While many works
may be quoted on this account, Kouki et al. (2014) provided a fresh perspec-
tive by undertaking vertically integrated modelling of the production process
in financial services firms such as banks and insurers.8 The broad consensus in
literature post-2020 may be summarised by Shah et al. (2022, p. 8), who assert
that “Production approaches are more applicable for branch-level data, while
intermediation approaches are for bank-level data”. These approaches are sum-
marised in Figure 3.1.
As highlighted in Figure 3.1, specification approaches to bank output can be
classified into volume, value, hybrid and national income approaches. The so-
called production approach dominates the volume specification, while inter-
mediation (including the asset approach) and value-added approaches dominate
the value specifications. The combinations of these approaches yield the hybrid
approaches, as explained in the figure. Finally, the national income approach is
used for sectoral and aggregate analyses.

Production approach
Underlying any output specification is a specific kind of input-output rela-
tionship. The inputs and outputs must be defined consistent with the chosen
specification approach. In the production approach, the inputs are labour, capi-
tal, and, increasingly also, equity capital. However, labour and capital are the
most frequently used inputs. Both deposits and loans are specified as outputs
in either volume or value forms.9 Generally, volume measures of bank output
are preferred under this approach. However, the literature has been flexible in
specifying the output vector under the production function. Value measures
have been used frequently. Both flow and stock measures have been employed,
though the early literature stressed that the production approach should employ
flow measures of volume output (Colwell & Davis, 1992).
The essential idea in this approach is to theorise banks as typical produc-
tion units that use standard inputs (labour and capital) to produce traditional
outputs (deposits and loans). The current consensus suggests that the produc-
tion approach is the best choice for branch-level analysis. Intermediation, its
 Bank Output
specifications

Volume output Value output Hybrid output National Income

Volume and

Review of evidence on banking output measurement 33

Production Intermediation Production Value-added
Value measures Output Index Direct Output FISIM
approach approach approach approach
together

Loans and
Pure
Asset approach Deposits Unweighted Weighted Risk-free
Intermediation
separately

Loans and Loans and Loans and

Invesments Invesments Deposits Output price Revenue Share Risk-adjusted
separately combined combined

Employment
Others
share

Figure 3.1 Major approaches to banking output specification in the extant literature
Source: Author’s analysis based on the survey of the literature.
34 Survey of evidences
primary counterpart, is hardly a process under the control of retail branches.
The production approach specifies the most basic form of the bank produc-
tion function, which the literature has modified in different directions to yield
alternative approaches. A major criticism of this approach is the utter ignorance
of the intermediation function that banks use “the mismatch between lending
and borrowing to generate . . . profit” (Daisuke et al., 2011b, p. 7). Because
deposits are treated only as output, the liability-asset transformation undertaken
by banks is wholly ignored. Essentially, the production approach disregards
the interest costs paid on deposits (Shah et al., 2022). This approach is also
very much at odds with the core idea of banking, which implies transforming
public money into loanable funds. Despite its limitations, this approach pro-
vides computational advantages as it can be easily incorporated into a single-
stage production function. Its data requirements are also not very strict.10 This
approach has also been utilised as the first stage in a multi-stage production
function wherein deposits are produced initially using labour and capital. Sub-
sequently, loans and investments are produced using deposits and other inputs.
However, one aspect that seems to be missing in the literature is recognising the
production approach as a valuable tool for sectoral levels of analysis. Banks need
to be seen as service providers to depositors and borrowers at sectoral levels.
The production approach allows the use of deposits and loans both as outputs.
Other approaches tend to underplay the deposit services as output and prob-
ably overemphasize them as inputs.
In conclusion, the popularity of this approach can be ascribed to lenient
data requirements and possibly also to the increasing share of off-balance-sheet
output (generally defined as non-interest incomes of banks) in the revenue
portfolios of banks. Non-interest incomes include fees, commissions earned,
and other such directly charged services of banks. Suppose their share indeed
is increasing. In that case, banks could be theorised as production units that
process the basic inputs of branches, labour, and equity capital into directly
charged services without much theoretical loss.11

Intermediation approach
In general, conceptualizing banks as intermediaries accounts for most of the
services rendered by banks.12 It also implies that banks function as transformers
of funds rather than simply suppliers of goods, as Hancock (1985) suggested.
Under this approach, “banks intermediate deposited and purchased funds into
loans and other assets” Berger et al. (1987, p. 508). The purest form of this
approach treats deposits strictly as inputs. No, a priori empirical test is con-
ducted to decide whether to treat deposits as input or output. “Whenever
deposits appear on the input side, commercial banks are viewed as intermedia-
tion links between surplus-fund units (depositors) and deficit-fund units (debt-
ors)” (Boda & Piklová, 2021, p. 1553). Hence, this framework sees banks as
transformers of deposits (liabilities) into loans and advances (assets). The under-
lying belief in this approach is that banks possess a competitive advantage in
 Review of evidence on banking output measurement 35
minimizing delegation and monitoring costs compared to other lending mod-
els.13 Thus, their primary source of revenue is assumed to be intermediation
services, and this aspect must thus be reflected in the hypothesized production
function. This approach implies that banking institutions are primarily engaged
in transforming liabilities into assets at the macroeconomic level. Their crucial
role in the economy is to allocate scarce loanable funds efficiently.
At least four problems emerge in using the pure form of intermediation
approach.14 First, the services rendered by banks to depositors are ignored in
the output vector. Banks provide various fee-based and non-fee based ser-
vices to the depositors. The quantum of these services is generally assumed
to be positively correlated with the deposit amount. Thus, treating deposits as
an input contradicts this observation. Moreover, the treatment of deposits as
inputs is all the more troublesome, given that “a very substantial and growing
portion of the industry’s total revenue is received in the form of fee income”
(Radecki, 1999, p. 53). Second, at the macroeconomic level,15 to whom should
the intermediation services be ascribed? If depositors are the choice here, then
the very idea of intermediation loses context. Deposits being treated as inputs,
depositors gain the benefits of the intermediation services, which seems theo-
retically invalid. If borrowers are assigned the total value of intermediation
output, then one ignores the services supplied by banks to the depositors. As
discussed later on, these concerns are also at the heart of the national income
approach to bank output measurement. The allocation problem for financial
intermediation services indirectly measured (FISIM) output directly results
from these concerns.
Three, the intermediation process is inherently a two-stage process (Fukuy-
ama et al., 2020). In the first stage, variable and fixed inputs are used to pro-
duce deposits, while in the second stage, the deposits are used to produce
loans and investments. However, most selected literature studies have theo-
rised intermediation as a single-stage process. This is one matter that has moti-
vated the emergence of newer models such as network DEA (e.g. Yu et al.,
2019), dynamic DEA (e.g. Zha et al., 2016), stochastic DEA, two-stage DEA
(e.g. Khan et al., 2018) and simultaneous non-linear parametric models (e.g.
Kumbhakar & Sarkar, 2003). Fourth and last is the role of monetary policy
in the intermediation production function. Banks’ transformation of deposits
and other funds into assets is conditional on the monetary policy rates in an
economy. These rates can induce exogenous shifts in the production function
without changing the underlying performance parameters such as efficiency,
productivity, etc. Hartwick (1996) provides an overview of this aspect. Stud-
ies using the intermediation approach have not accounted for this dimension.
Further discussion on this aspect is provided later in this chapter. However, an
attempt was made by Triplett and Bosworth (2004a, 2004b) to incorporate
monetary reserves in the measurement of bank output. Other than this study,
the author could not locate any high-quality work.
Finally, incorporating risk into the intermediation process has been chal-
lenged. “Typically in the literature, the cost and profit functions or frontiers are
36 Survey of evidences
measured without considering the bank’s capital structure or bank’s choice of
risk” (Hughes & Mester, 2008, p. 9). A notable review on this issue was made
in this direction by Kenjegalieva et al. (2009), who proposed a risk-adjusted
intermediation approach-based output specification. The study highlighted
three competitive advantages of banks in terms of managing risks: assessment
(information advantages), diversification (pooling of deposits) and risk-taking
(deposits as a check on bank’s speculative investments). Accordingly, they
stressed incorporating three vital explanatory variables: technology, property
rights and organizational form. The previous discussion clarifies that a strict
specification of production function under the pure form of intermediation
approach will cause many theoretical and empirical problems. Researchers
have resolved this issue by evolving further extensions of the intermediation
approach. The asset approach is one of them and is explored now.

Asset approach
The limitations of the pure intermediation approach have motivated output
analysts to incorporate more dimensions of the asset-side of a bank’s balance
sheet into the output vector. The so-called asset approach is the result of these
analytical innovations. The primary difference between the pure intermedia-
tion and the asset approach is captured by Leong et al. (2003, p. 198), who
state that, “In essence, this stream of thought is a variant of the intermedia-
tion approach, but instead defines outputs as the stock of loan and investment
assets”. Inclusion of investments, specifically financial investments, improves
the coverage of the output vector under the intermediation approach and
permits the recognition of the shareholders as major consumers of the bank’s
intermediation process. Along with depositors and borrowers, an additional
consumer of the intermediation services is introduced by the asset approach –
viz., the shareholders. Banks are not only transformers of deposit funds into
loanable funds. They also undertake wealth generation and maximization for
their shareholders. This fact is recognized under the banking literature’s asset
approach to output specification. By including investments as an output in the
production function, banks are viewed as transformers of liabilities into loans
and liabilities into investments. Two measures of bank output are used under
this approach: 1. amount of loans; 2. amount of investments. These two vari-
ables are either used separately or combined in the literature. Table 2.1 shows
the frequency of use of these variables in the sample studies. Deposits, however,
continue to be treated as inputs and the usual problems of the intermedia-
tion approach continue to plague this approach. Another improvement in this
approach is that non-deposit funds (generally termed as other purchased funds
in literature) are recognised as inputs and traditional deposit funds. While the
pure intermediation approach considers only deposits as the source of loanable
funds, the asset approach can also be extended to include non-deposit sources
of funds. In practical terms, this framework generally provides a better fit for
productivity and efficiency estimations using bank-level data.
 Review of evidence on banking output measurement 37
Value-added approach
Concerns with intermediation and production approach gave rise to a new
line of thought on bank output measurement, which introduced a data-driven
approach to output-input specification of a given variable in the bank’s produc-
tion function. The production approach was improved upon, and instead of
ad hoc criteria in choosing the outputs, an empirical approach was suggested
by Berger and Humphrey (1992, p. 247), who assert that the value-added
approach considers “outputs as those activities that have substantial value-added
(i.e., large expenditures on labour and physical capital)”. Compared to the
intermediation approach, the value-added approach “considers all liability and
asset categories to have some output characteristics rather than distinguishing
inputs from outputs in a mutually exclusive way” (Berger & Humphrey, 1992,
p. 250). Output characteristics are established by operating cost allocations for
each liability- and asset-side component. Benston et al. (1982, p. 440) state,
“Output should be measured in terms of what banks do that cause operating
expenses to be incurred”. Hence, if an activity is causing the bank to incur
operating costs, it may be termed an output irrespective of whether it is a
liability or an asset.16 This approach is essentially a reframing of the production
approach but with the additional restriction of operating cost allocations to
each liability and asset item. The value-added approach has been modified to
incorporate more aspects of the output vector in banking. Literature has also
used value-added specifications, including interest costs or price of deposits,
as inputs while considering the deposit amount as an output. This modified
value-added approach has perplexed many analysts who criticize the inclusion
of deposits on both sides of the production function. Despite some of its limita-
tions, this approach has been the most favoured within the literature. However,
very few studies undertake any empirical assessment of the operating costs
function to locate those items that should be considered outputs and those that
should be defined as inputs. The value-added approach has primarily been ad
hoc in the literature. Researchers claim that the selected items (from both the
deposits and asset sides) have positive operating cost shares and thus must be
considered as outputs. However, hardly any of the reviewed studies empirically
justified such an assertion. Concurrent with this approach, another data-driven
approach has been frequently used in literature to specify bank outputs, namely
the user cost approach, as discussed in the next section.

User cost approach

The emergence of the idea of user cost can be traced back to the works of
Donovan (1977) and Barnett (1978), who attempted to construct index meas-
ures for the aggregate stock of money supply. The foundation of this approach is
well-captured by Hartwick (1996, p. 5), who states that “This approach derives
from the practice in the theory of general equilibrium where inputs have nega-
tive prices and outputs have positive prices”. It is thus clear that the user cost
38 Survey of evidences
approach utilizes the net contribution of an activity to the total revenue of a
banking firm to determine whether it is an input or an output.17 Positive user
cost would imply that the activity is output, while a negative sign would imply
that the activity is an input (Hancock, 1985). This approach requires data on
net revenue for each banking firm and its disaggregation across various banking
services. Individual income and balance sheet items can be specified as input or
output in the production analysis. Furthermore, even if such data are available,
one needs to estimate the opportunity costs,18 which will require implicit or
shadow pricing methods. The data requirements are strict in this approach, and
barring a few advanced economies, such data are generally difficult to obtain,
particularly in emerging economies.
Two problems have frequently been discussed in the literature under this
approach. First is the instability of the user cost sign (Banker et al., 2010). Sec-
ond is the problem of accounting for implicit earnings (Berger & Humphrey,
1992). Regarding the first problem, it should be noted that the user cost sign
is an empirical issue. It can differ depending on the country, banks, regions,
and other aspects. Both temporal and spatial variations in the user cost sign are
possible. This makes the output specification using this approach somewhat
tentative and unstable. In this case, the output measure may well switch over
to input and vice-versa with variations in time and space. The temporal and
spatial stability of output measures is difficult to establish in this case. Regarding
the second problem, the user cost approach can examine only the observed and
direct revenue. This is because the data for such explicit revenues are readily
available in banks’ income statements. However, there are implicit revenues too
that banks earn on intermediation, risk management, wealth maximization
and other services. The user cost approach will ignore these services unless the
total revenue is modified to include these implicitly priced services too. This
is a data-intensive job and will generally be challenging in emerging econo-
mies with weak data infrastructure.19 A criticism often directed at the user cost
approach is the lack of explicit specification of the services provided by finan-
cial services firms. The user cost sign from the underlying revenue function is
used to specify the output. However, there is no guarantee that a particular asset
or liability item will always be an input or an output. As mentioned earlier,
the user-cost sign is subject to temporal and spatial variations. This makes the
input-output specification using user cost of approach ad hoc and possibly even
biased.20 This approach estimates FISIM output under the SNA framework,
but its application at the bank and branch levels is complicated. Few studies
which have used this approach have not undertaken any empirical estimation
of the user-cost sign of the chosen output variables. Instead, they use extant
literature and their beliefs in specifying outputs under the user-cost approach.

Index numbers approach

The index numbers approach is a novel and less frequently used method for
specifying bank output in the literature. The multi-product nature of banking
 Review of evidence on banking output measurement 39
firms, which was stressed as early as 1959 by Speagle and Kohn, has motivated
researchers to use an index of outputs rather than to use single or multiple out-
puts separately. Noteworthy works using this approach to bank output speci-
fication include Benston et al. (1982), Brand and Duke (1982), Clark (1984),
Kim (1985), Colwell and Davis (1992), Athanasoglou et al. (2009), and Royster
(2012).21 All these studies undertake weighted indexing. The choice of weights
differs from study to study. The popularly used weighting factors22 in the litera-
ture include unit costs, unit prices, gross yields (in the case of assets measures of
output), revenue shares, and labour requirements per unit of output. The basic
idea behind this approach is that a single output or even multiple outputs sepa-
rately do not capture the output vector correctly. The multiple outputs should
be combined using theoretically consistent index numbers.23 The choice of
outputs can be undertaken using any previously stated approaches. However,
the chosen output variables should be combined into an index. Two popular
output index numbers in the literature are the Divisia index and the Tornquist
index. These indexes are generally considered to be possessing important sta-
tistical properties such as being superlative. This allows theoretically consistent
aggregation across the chosen outputs. The application of this approach has
been limited in the literature, though.

3.4. Output measurement in the banking literature

This section elaborates on and reviews the major output measures used in
the literature. Each output measure is presented in a flow diagram that cap-
tures the significant mechanics behind its construction and the essential issues
highlighted.24

Loan outputs
The critical dimensions of the output measures based on loans and advances in
banking performance literature are captured in Figures 3.1, 3.2, and 3.3. The
figures are prepared in a self-explanatory manner. The fundamental categories
of loan-based outputs are classified into three types: 1. stock-based measures, 2.
flow-based measures, and 3. other loan-based measures, which is a reflection of
how the literature has approached this output measure. Given the distribution
of sample studies, as shown in Table 2.1, the stock-based measures, in general,
have dominated the banking literature. Generally, the loan amount, either gross
of NPAs or net of NPAs, has been employed as a bank output measure. This
amount is derived from the balance sheet and is a year-end figure.
The amount of loan and number of loan accounts are the most frequently
used measures of bank output in terms of loans-based measures. With regards
to the amount of loan, which is by far the most frequently used measure, it has
been used in five different ways, namely, as an aggregated measure (weighted
or un-weighted), disaggregated measure, in combination with non-loan
outputs, in terms of adjustments for interbank loans or as a “bad” output (in the
 40 Survey of evidences
Amount of Loans

Weighted measures Consumption

Maturity- sector-based
Price Revenue weights based
weights
Un-
weighted Individual
Share in total lines Average ticket
measures size-based
Effective loans
interest rate

Risk-adjusted
interest rate
Frequency-based

Month-end
Year-end Inc. of
Exc. NPA
NPA

Exc. Inter-
bank loans Exc. loan
With investments Inc. loan
losses
Inc. Interbank losses
With Deposits Exc. NPA loans Inc.
reserves NPA
Both deposits
Exc. NPA and investments Amt. of
reserves Inc. NPA NPA Amt. of
Exc. Loan Exc. Loan
losses NPA Losses
Exc. Loan Exc. NPA
losses reserves
Exc. NPA Amt. of Interbank
Exc. Loan losses loans

Figure 3.2 Stock measures of loan output in banking literature

Note: “Inc.” implies inclusive of, “Exc.” implies exclusive of, and “Amt.” implies amount.
Source: Author’s analysis based on the survey of the literature.
 Flow measures of Loan
output

Value
Measures Volume Measures

Number of new
Gross interest Net interest accounts (Stock or
income income Flow)

Without netting out Combined with new

After netting out deposit accounts
interest expenses interest expenses

Weighted Un-weighted

Review of evidence on banking output measurement 41

Interest on summation
deposits summation
Interest on
deposits
Share in un-
Interest on other weighted total
borrowings

Interest on other Share in total

borrowings revenue
Combined with non-
interest income
Average interest
Interest rate rate
Income from
investments

Income from
deposit services

Both

Figure 3.3 Flow measures of loan output in banking literature

Source: Author’s analysis based on the survey of the literature.
 42 Survey of evidences
Other measures

National
Heterodox Income
measures

Amount of loan
net of Flow amount of loans
securitization Price x Quantity
measure
Adjusted for effective
interest rate
Islamic measures
Both
Index measures
Adjusted for flow of
deposits

Volume Measures
Hybrid
Value Measures
Revenue share weighted

Interest rate weighted Interest rate weighted

Revenue share
weighted

Figure 3.4 Other measures of loan output in banking literature

Source: Author’s analysis based on the survey of the literature.
 Review of evidence on banking output measurement 43
form of NPAs). As examined in the previous section, these approaches to loan
output measurement have been used under almost every specification frame-
work. Aggregate measures are at the level of the entire banking industry or the
banking firm. Weighted aggregates, in this case, have generally been built using
price (interest-rate) weights, revenue-share weights, effective (market interest
rate net of reference rate) interest rate weights, share in un-weighted total loans,
and risk-adjusted interest rate (where risk is by an interest rate or return on
assets or equity).25 These measures are aggregate in the sense that they represent
the sum of all loans with different maturities (short-term and long-term), dif-
ferent lines (real estate, personal, etc.), different ticket sizes (retail and corpo-
rate) and diverse consumers (firms, individuals, governments, etc.). Literature
has also disaggregated the aggregate amount of loans across these categories
and used them separately. This can be examined by looking at the appendix to
this chapter. Another frequently used approach is to combine loan output with
non-loan outputs. The combination of loans with investments is the most fre-
quently used measure. The aggregation of loans and investments has generally
been un-weighted in the literature, i.e. a simple sum of both the components
is undertaken to estimate the total bank output.
Further adjustments are made, as shown in Figure 3.2. Adjusting for NPAs
has been a serious concern for researchers. When this is done, one has to
deal with how to treat the NPAs, i.e. whether to remove them from total
loans or treat them separately as “bad” output (subject to minimization rather
than maximization).26 Literature has primarily used the gross loan amount (i.e.
including NPAs). This seems to be that incurring losses is a service that banks
render to generate good loans (Hartwick, 1996). Another approach has been
to combine loans and deposits27 together.28 This is a frequently used output
measure under the value-added method for output specification, though even
the national income approach29 employs this measure. There has been a lot
of debate on whether loans and deposits should be combined. The consensus
seems to favour using them separately rather than combining them.30 Lastly,
some novel issues in loan output measurement are treating interbank loans and
NPAs as “bad” output in the production function. Studies dealing with the for-
mer aspect can be found in Mertens and Urga (2001), Isik and Hassan (2002),
Pasiouras et al. (2009), Olson and Zoubi (2011), Kouki et al. (2014), Mamonov
and Vernikov (2017), Khan et al. (2020), and Bansal et al. (2022). For the lat-
ter issue, some recent illustrations include Fukuyama et al. (2020), and Zaman
et al. (2022), among others.

Deposit outputs
Deposits represent the most crucial source of loanable funds for the bank-
ing system. Increasingly, however, non-deposit funds such as equity have been
gaining increasing attention. Figures 3.5, 3.6, and 3.7 present a systematic rep-
resentation of the ways in which this output variable has been conceptualized
and empirically operationalized in the literature.
 Amount of

44 Survey of evidences
deposits

Total Deposits Total Deposit plus Core deposits Interest rate

Time Demand
deposits Investments deposits
Demand Loans
deposits Savings
Saving Gross deposits
deposits Loans
Retail time
deposits
Checking
account Net
deposits Loans
Both loans
and
Money Stock- investments
Weighted based

Other
deposits
Interest rate
Certificate of
Adjusted for deposits
Foreign
Interbank currency
Share in total deposits deposits
revenue
Insured vs
Monetary Uninsured
policy reserve
Share in direct ratio Adjusted for
fee-based turnover rate
revenue (velocity of
money)
Effective
interest rate

Figure 3.5 Amount of deposits as bank output in the literature

Source: Author’s analysis based on the survey of the literature.
 Volume measures of
Deposit output

Time Transactions Accounts

Spent on
transactions Demand deposits Time Deposit
Demand Deposit

Review of evidence on banking output measurement 45

Other deposit Core deposit Others
services services Time deposits
Combined

Interest rate
weighted
Current
Non- Interest- accounts
Interest-bearing
bearing
Interest income
Turover rate share weighted

Figure 3.6 Volume measures of deposits as bank output in literature

Source: Author’s analysis based on the survey of the literature.
46 Survey of evidences
Generally, the sum of demand and time deposits has been used to measure
bank output in the literature. However, the use of demand and time deposits
separately is also possible, given that the deposit structure may be an essential
concern in modelling the production process of banks. When used together,
demand and time deposits are aggregated using interest rate, share in total rev-
enue, or direct fee-based revenue as the weighting factor. Banks’ deposit ser-
vices can be associated with two components: those related to maintaining the
amount of deposit with the bank and those enabling the amount of customers’
transactions through and with the bank (Gorman, 1969). The value of the
deposits generally captures these services in the studies on bank performance.
In the early studies, analysts have also used other measures of deposits, such as
certificates of deposits (Mullineaux, 1978; Todhanakasem et al., 1986; Law-
rence, 1989; Noulas et al., 1990; Elyasiani & Mehdian, 1992; Sharpe, 1997;
Muldur & Sassenou, 1993; Resti, 1997), and foreign currency deposits (Kim,
1986; Kim & Weiss, 1989; Jagtiani & Khanthavit, 1996). Another novel but
recently popular deposit type being included in the output vector of banks
is insured deposits (Ohlson et al., 2021). Researchers have used the deposit
amount individually, separately with other non-deposit measures or in com-
bination. Loans and investments have been used as non-deposit outputs in the
output vector, and the amount of deposit and these three output measures are
either combined or used separately. When used together, they have generally
been used in an un-weighted form. The value-added approach to output speci-
fication generally recommends this kind of output measure. Even the produc-
tion approach has utilised this kind of bank output measure in the literature.
Finally, the most crucial debate in using deposits in the bank’s produc-
tion function is the so-called input-output dilemma (Agu, 1988). Figure 3.7
examines the significant dimensions of the input-output dilemma in the bank-
ing literature. The main issue of contention is whether to treat the amount
of deposit as an input,31 quasi-fixed input32 or an output.33 The production
approach recommends treating deposits as output, while the intermediation
approach stresses its role as an input. The value-added approach recommends
appropriately using it on both sides (e.g. interest costs as input and amount of
deposits as output). Analysts have also combined them with loans to measure
bank output under the value-added approach. User-cost34 and operating cost
approaches recommend that empirical tests be performed to decide whether
the amount of deposits is an input or an output in a given context. These two
approaches are data-driven and reject any a priori conception of deposits being
input or output.35 Finally, in the national income approach, while estimating
FISIM, deposits are treated as an output but in terms of the flow of deposit
services rather than its stock.36
In two-stage models under the intermediation approach, deposits are generally
treated as intermediate output and input in the entire production process. In other
words, there is no single ideal prescription on how to resolve this dilemma. An
alternative input-output dilemma in the literature is whether to treat the number
of branches as an input (such as in Adams et al., 2002; Sinha & Chatterjee, 2008;
 Deposits as input or output

Branches as
input or output

Single-stage Multi-stage National income

Ad-hoc Theory-driven Data-driven Intermediation Flow

Review of evidence on banking output measurement 47

Structure-based User cost User cost

Production Intermediation Value Added Operating cost

Asset

User cost

Figure 3.7 Approaches generally employed in literature to resolve the input-output dilemma
Source: Author’s analysis based on the survey of literature
48 Survey of evidences
Chen et al., 2018, among others) or as an output (such as in Focarelli & Panetta,
2003; Kumbhakar & Sarkar, 2003; Humphrey, 2009, among some others). This
dilemma also stands unresolved. Lastly, as shown in Figure 3.6, the literature
has also utilized the volume measures37 of deposit output. Three measures are
employed in this context. First is the amount of time spent on handling deposit
services. Second is the number of deposit transactions (disaggregated to individ-
ual deposit services). Third and last is the number of deposit accounts (combined
or disaggregated into demand and time deposit accounts).

Asset output
Amount of loans (as discussed earlier), amount of investments and amount
of other assets are the critical asset measures of bank output employed in the
literature. Figure 3.8 shows the key permutations and combinations made in
defining the asset output of banks in the literature. Assets are those items in
the bank balance sheet that create the possibility of future earnings for the
banks. Returns may be generated on these items, and thus they are classified as
assets. While the amount of loan is the most common asset measure, additional
asset items are included and combined with loans to measure the asset output
of banks. The amount of bank total assets has generally been defined in the
literature as the sum of loans, financial investments, physical assets and other
assets (such as loans to the financial sector, unused assets, non-earning assets
and interbank assets). The so-called asset approach to output specification, as
discussed in section 3.3, recommends using the amount of total assets as the
correct measure of bank output. Naturally, deposits are not given a space in the
output vector under this framework. However, studies under the value-added
approach have used deposits alongside the amount of total assets as bank out-
puts. More supplementary discussion on this output measure can be obtained
from the author upon request.

Investment output
The amount of investments is another frequently employed measure. Fig-
ure 3.9 exemplifies the critical aspects of this measure in the literature. Invest-
ments in this context are financial investments, i.e. investments made by banks
in financial instruments and securities. The amount of investment is used either
individually or in combination with loans in the literature. Individually, it is
either used as a total consisting of various individual investment items or in
disaggregated terms. Literature has adopted different categorizations of invest-
ments. They are enlisted in Figure 3.9. The debate seems to hover around the
inclusion of risky versus debt investments. How risky investments are included
in the investment output impacts the amount of volatility introduced into the
output vector. Generally, literature has undertaken un-weighted aggregation
of risky and debt investments. Another investment output measure used is the
value of liquid assets, i.e. investments with a high degree of liquidity. Some of
the notable recent literature using this measure includes Berger and Bouwman
 Assets

Gross Total Assets Liquid assets

Financial Investments
Loans
Interest-based Illiquid
Plus OBS assets
Others output
Aggregation

Bond
Loans to Financial Holdings Non-int.
sector Int. earning
earning
Unused assets
assets
Liquid and

Review of evidence on banking output measurement 49

Interest rate
Illiquid
Effective interest rate
Earnings-based
Operations-
Share in revenues based
Earning Non-earning
assets assets
Competitive interest NPA Non-
rate operating
Other earning assets
assets
Interbank Activity
Operating
Interbank Net of Loans assets
assets Interbank
Liabilites
Net of Physical
Assets
Value of Physical Financial Interbank
assets assets Assets

Figure 3.8 Asset measures of bank output in the literature

Source: Author’s analysis based on the survey of the literature.
 Investment
Output

50 Survey of evidences
Approved Un-approved
Amount

Combined
Total
with
Authorization
-based
Loans

Gross Loans Net Loans

Liquidity- Security- Others
based based

Assets held in
Short term Government Risky Others Country-based
trading account

Equity Other
Long term Short term Long term Entrusted Domestic
instruments enterprises

Money
Bonds Inter-bank Foreign
market

Capital
Debentures
market

Others (repo,
Others
etc.)

Figure 3.9 Investment outputs used in banking literature

Source: Author’s analysis based on the survey of the literature.
 Review of evidence on banking output measurement 51
(2007), Shih et al. (2007), Banerjee (2012), Davies and Tracey (2014), and San-
tosa (2022). However, its use can be found in literature from 2000 onwards, as
shown in the appendix to this chapter.

Income and off-balance-sheet output

As discussed earlier, a production function requires flow outputs rather than
stock measures. However, literature on bank output measurement has found it
inevitable to use stock variables. Yet, flow measures of output have evolved in
the literature. An early systematic attempt was the SNA (1953) which intro-
duced the FISIM method to estimate the banking sector’s output per national
income account. At the microeconomic level, some researchers have also
attempted to estimate flow measures of deposits and loans. Other than these
flow measures, income measures of output have been employed that account
for the explicit direct fee commissions charged by banks from depositors, bor-
rowers and other customers. Figure 3.10 explains the effective approaches in
this context.
The basic idea behind this approach is to capture the non-traditional ser-
vices and consequent revenues generated by banks. As per the intermediation
approach to output specification, the traditional function of banks is to trans-
form deposits into assets. The interest income captures this and implies that
banks generate implicit revenues, which are accounted for through indirect esti-
mation methods such as the FISIM. However, as noted in the literature and
will also explained by expert contributions38 in Chapter 7, the contribution
of fee-based output is increasing in the bank’s total revenues, and this is being
driven by income-based off-balance-sheet items such as fees, commissions, bro-
kerage charges, and other such incomes of banks which are directly recorded
in the income statement. One can find the use of this output measure as early
as Mullineaux (1978). As shown in the appendix to this chapter, several studies
have used this measure. An advantage of this measure is that it is a flow measure-
ment and thus closely corresponds to the demands of the neoclassical produc-
tion theory. They are termed ‘non-traditional’ because these services diverge
from the traditional notion that the primary function of banks is to transform
deposits into assets. The income measure includes both interest and non-interest
incomes. In contrast, the off-balance-sheet or the non-traditional measure only
includes the non-interest incomes as per the extant literature. At times, non-
interest income is adjusted for service charges on deposits and net gains (profit
or loss) on speculative investments to yield net non-interest income (such as in
Rogers, 1998). Such a measure has been scarcely used in the literature.
Other income measures include the amount of profit (recent examples
include Khan et al., 2018; Le & Ngo, 2020; Owusu & Alhassan, 2021; Bansal
et al., 2022, among others) and amount of loan adjusted for unearned income
(for example, Hughes et al., 2000). Using the off-balance-sheet measures of
bank output implies recognising significant structural changes ongoing in the
business model of modern banks. This measure is also quite suitable for output
 52 Survey of evidences
Income
measures
Other income
measures
Revenue
Unearned
income
Security and currency Interest Non-interest
trading revenue income income

Also called Off-

balance sheet
Profit output
Before Tax
After Tax
Gross Interest Non-interest
Operating income
income inome

Adjusted Fee and Adjusted From From Both

for charges for Lending Deposits combined

Interest Other Service charges Unweighted Weighted

expenses operating From on deposits aggregation aggregation
expenses Lending
Net Interest And net gains on Interest- Share in
Income From rate total assets
Deposits speculative investments

From From Market

Lending Investments interest rate
From Net Non-interest
Deposits income Effective
interest rate

Other
measures

Share in total Share in total

operating cost cost

Figure 3.10 Income and off-balance-sheet bank output in the literature

Source: Author’s analysis based on the survey of the literature.
 Review of evidence on banking output measurement 53
analysis of the non-banking financial sector as non-interest income occupies a
larger portion of their total revenue.

FISIM output39
The essence of the national income approach to bank output measurement is
the FISIM approach.40 This measure tries to capture the output of banks that is
implicit41 in their intermediation activities rather than directly measurable from
their income statements or balance sheets. This measure of banking output can
be used mainly at the sectoral level and does not directly lend itself to branch-
level, firm-level, and intra-industry-level analyses. The main reason for this
limitation is that disaggregated measures of this output are not available. How-
ever, for macroeconomic analysis of the banking industry, this output measure
is quite helpful and provides a reliable summary of the value-added by banks.
The primary debate underlying this approach is the choice of the reference
rate,42 also known as the Barnett benchmark rate and the Hancock opportunity
cost rate of money (Fixler & Zieschang, 1999). The fundamental debates on
the choice of reference rate are summarised in Figure 3.11.
Analysts have used the concept of reference rate to estimate both the FISIM
output43 and disaggregated flow measures of output. In the latter case,44 the
deposit and loan amounts are adjusted for the effective interest rate (market
interest rate minus the reference rate). The flow of deposit and loan services
is estimated. They are then combined to measure the total FISIM output.
FISIM captures the implicit portion of banking output in the national income
accounts. While its measurement and accounting procedures have a consen-
sus in the SNA framework, the allocation of the net interest margin (interest
received on loans minus interest paid on deposits) to various “business (inter-
mediate consumers) versus households, government, and the rest of the world

Reference
Rate

Risk-free Risk- adjusted

Money market
Rent- adjusted Asset- specific
interest rate

Others Single Rate Dual Rate

Figure 3.11 Debates on the references rate under the FISIM approach
Source: Author’s analysis based on the survey of the literature.
54 Survey of evidences
(final consumers)” (Fixler & Zieschang, 1999, p. 547)45 is still unsettled. An
issue that has received considerable attention within the reference rate debates
is risk treatment. Traditional SNA-based FISIM estimation relies on a risk-free
or minimum risk-based reference rate.46
However, several authors have disagreed with this notion and have attempted
to incorporate ‘risk-incorporated’ reference rates into the FISIM framework.
Their fundamental argument is that ignoring risk overestimates the FISIM out-
put due to choosing a risk-free (or minimum risk) reference rate such as the
treasury bill rate. They also assert that the measurement of opportunity cost
with a risk-less reference rate does not capture the actual behavioural dynam-
ics at play. It implicitly assumes risk-aversion on the depositors and borrowers,
which may not be applied universally. However, the SNA (1993) defends its
use in a risk-free form by stating that “the reference rate to be used represents
the pure cost of borrowing funds – that is, a rate from which the risk premium
has been eliminated to the greatest extent possible and which does not include
any intermediation services”. The SNA (2008) goes further and states that “the
reference rate should contain no service element and reflect the risk and matu-
rity structure of deposits and loans”. Thus, the essential features of a reference
rate as per the SNA approach are 1. the underlying asset whose interest rate is
used as a reference rate must be risk-free; 2. the underlying asset should not
provide any financial intermediation services. One peculiar aspect of the SNA
approach to reference rate is that it uses a single aggregate rate for deposits and
loans.47 This has been criticized in the literature, as a single rate will inevitably
fail to capture the interest rate structure on deposits and loans.
Furthermore, a single reference rate treats depositors and borrowers as agents
with similar risk perceptions, which is not true theoretically or empirically.
The very use of a concept like a reference rate has been criticized in literature
by arguing that the quantity of financial services and the financial instruments
in the market are not necessarily related to each other. The primary assertion
has been “that the implicit price of financial services bears no definitive rela-
tionship with any reference rate” (Wang & Basu, 2005, p. 1). They also argue
that “not all of a bank’s net interest income is compensation for its implicit
services” (p. 5). The FISIM method is a valuable tool for obtaining a macroeco-
nomic accounting-based nominal measure of banking output but lacks reliable
foundations for deriving real banking output (Wang & Basu, 2005).

National income accounts and banking output

Other than FISIM, national income accounts are utilised to estimate a mac-
roeconomic measure of banking output. The same has been presented in Fig-
ures 3.12 and 3.13. The primary debate in this regard is whether to use gross
output or value-added measures. The main difference between both of them is
that of intermediate consumption. Generally, the literature has favoured using
the value-added approach for sectoral-level analysis in banking. The broad con-
sensus on this choice seems to be that estimating intermediate consumption
 Review of evidence on banking output measurement 55
for the banking industry is complicated. The gross-output measure is generally
available for the financial services industry rather than the banking industry.48
Lastly, a major challenge in using sectoral measures of banking output has
been the matter of deflation of the nominal output. The main methods used
in the literature to deflate nominal banking sector output are highlighted in
Figure 3.12.

3.5. Thematic issues in banking output measurement

Given the literature survey undertaken in the above sections, several thematic
issues can be extracted, and the same is present in very brief terms below. The
readers can access detailed notes on this in the online resources.

Deflation of banking output

Nominal output has been frequently used in banking literature, as explained
in Table 2.1. However, many researchers have employed real measures of bank
output also. The real measures include volume output measures and constant-
price output measures. The volume measures are in pure quantity terms, as
discussed earlier. The main problem is converting the nominal (current-price)
output to constant-price output. There is no consensus on the ideal deflator for
bank output to date. One is presented with a menu of analytical choices, and
given the context and purpose, the most feasible deflator is chosen. A reflec-
tion of the significant deflators in banking literature has been provided in Fig-
ure 3.12. Deflation removes the effects of inflation from the first difference
form of a nominal output variable. This means that when the growth of con-
stant-price output is measured, it should represent purely the quantity move-
ments rather than quantity and price movements, as is the case with the growth
rate of nominal output. Inflation embodies itself in banking activities under
different dynamics not correctly captured by aggregate price indexes such as
the Consumer Price Index (CPI), Wholesale Price Index (WPI), GDP deflator
and GNP deflator. The frequent use of these indexes reflects the fact that opti-
mal industry-specific deflator for banking services does not yet exist. As dis-
cussed by Prof. R. B. Barman in Chapter 7 and Barman and Samanta (2007),
many efforts have been undertaken by official agencies in the EU and US to
build such a price index. However, there is still no optimal deflator, which is
the case with emerging economies. The literature has primarily used CPI and
WPI for deflation of nominal bank output at the firm and intra-industry-level
and has preferred GDP deflator at the sectoral level of analysis.
Different prices can measure the constant-price value of banking output
and the aggregation functions of disaggregate banking outputs: 1. the user-cost
price is frequently estimated within the SNA framework (Fixler & Zieschang,
1999); 2. CPI and WPI as crude proxies for banking output price; 3. observed
market interest rates (this approach treats observed interest rates as prices of the
concerned output measure; deposit interest rate poses a problem here. One can
 Sectoral
Measures

56 Survey of evidences
Deflation

Explicit Implicit

General

GDP
CPI Specialized Deflator GNP
Deflator
WPI
Banking Service
Price Index

Gross
Output
Value
Added
Financial
Implicit Flow of Services
Loans
Explicit
FISIM Banking
industry
Adjusted for
Flow of EIR
Income Deposits
Fee Disaggregation
Actual
Off-balance Adjusted for interest rate
Lending- sheet EIR Across Space
based Reference
Actual rate Geographical
interest rate Units
Deposit-
based Reference Individual
rate services

Figure 3.12 National income measures of macro-level banking output

Note: EIR is Effective Interest Rate.
Source: Author’s analysis based on the survey of the evidence.
 Review of evidence on banking output measurement 57

Financial
Services

Financial
Explicit Insurance
Intermediation

Financial
Banking
Markets

FISIM

SNA interest Bank interest

(bank to non- (bank to non-
bank) bank)

Reference rate- Market interest

based inflow of rate-based inflow
interests on loans of interests on
loans

Reference rate- Market interest

based outflow of rate-based inflow
interests on of interests on
deposits loans

Figure 3.13 Conceptualizing banking output in the SNA framework

Source: Author’s analysis based on the SNA 2008 manual.

use the difference between the opportunity cost of holding deposits and the
deposit interest rate as a proxy here); 4. operating cost plus margin-based price
measures can be used as are traditionally used in manufacturing industry output
measurement. The literature has primarily relied on established price indexes
such as the aggregate CPI and aggregate GDP deflator. Another thorny issue
is the single versus double deflation method.49 The double deflation method
uses two different price indices for output and input deflation, while the single
deflation method uses the same price index. Theoretically, double deflation
is a superior method as it accounts for more extensive information on the
price dynamics in banking – at the side of outputs and inputs. Double defla-
tion deflates outputs and inputs separately; “deflated value-added results from
deflated production value minus deflated intermediate consumption of goods
and services” (Boer, 1999, p. 5). In other words, inputs and outputs are deflated
58 Survey of evidences
differently and separately. However, data constraints are more pressing in this
method as one requires reliable disaggregated measures of input costs and mar-
ket prices.50 Literature has generally used the single deflation approach.51

Stock versus flow dimensions in output measurement

Production theory requires the use of flow measures of output. This is irre-
spective of whether the output is for tangible or intangible goods. In banking
literature, stock measures have been dominant. Their use is relatively lesser in
the case of insurance firms where flow output data are available and economi-
cally meaningful. The use of stock measures implies that the entire amounts
of loans and deposits active in a given year are the bank’s output and not only
the new loans and deposits added during the year. Clark (1984, p. 54) asserted
that “a stock concept is appropriate because the relevant stock can only be
maintained by the continuous production of intermediation services”. How-
ever, there are many warnings against using stock measures. Literature has
stressed that using stock measures may cause specification errors in the mod-
elled production function (Subramaniam, 1985; Kim & Weiss, 1989). They
stress the need for flow measures rather than stock measures, whereas literature
has broadly used stock measures found by the survey undertaken in the pre-
sent work. Hartwick (1996) provides a way out of this by proposing a method
to convert stock amounts of deposits and loans into flow measures through
adjustment by respective interest rates. This approach is quite similar to what is
done under the FISIM method. Another solution is to use both stock and flow
variables together in a production function for banks (Resti, 1997). However,
this approach does not resolve the matter but finds a middle way. The use of
stock variables to measure the output of banks assumes fixed proportionality
between the underlying flow of services and each unit of the stock variable,
or whatBoer (1999) calls the ‘influence factor’. This, too, is a severe criticism
levied against this practice. Data constraints continue to maintain the popular-
ity of stock measures of output.

Issues in economic aggregation

Other than the studies that focus on the most disaggregated units of analysis,
such as a single banking firm or an individual branch, most analysts have had
to face the problem of aggregation. Aggregation is undertaken across space and
time (such as converting higher frequency data into annual data) in the litera-
ture on banking. Space may be branches, banks, regions, products, etc. The
essence of this problem is captured in Benston et al. (1982, p. 440), who state
that “the literature on aggregation theory and index numbers concludes that
the appropriate way to aggregate quantities of bank output is to (a) estimate the
parameters of a ‘flexible’ aggregator function which captures the substitution
and complementarity relationships between all deposit and loan categories or,
equivalently, (b) use a statistical index number formula which approximates the
 Review of evidence on banking output measurement 59
results that would have been obtained from a flexible aggregator function”.
Thus, underlying the issue of aggregation of outputs in banking is the choice
of the optimal aggregation function. Generally, linear aggregation functions
have been employed in the literature with different weights. The aggregation
has been used in banking output literature for two essential purposes: first is
to estimate the total output as a composite of heterogeneous sub-components
and second for building an index of outputs. Combining outputs of different
kinds into an aggregate measure of bank output requires fulfilling various nec-
essary and sufficient conditions, as elaborated in Kim (1986). Very few studies
undertake aggregation of banking services within a rigorously defined aggre-
gation procedure while fulfilling the conditions as explained in Kim (1986) or
Fisher (1922). The most fundamental condition for meaningful aggregation of
bank outputs is separability. It “implies that the marginal rate of transforma-
tion between any two elements of the aggregate must be independent of the
levels of the variables outside the aggregate” (Kim, 1986, p. 184). Achieving
such independence for measures of aggregate output in banking is very dif-
ficult given the high degree of interdependencies of banking activities with
exogenous non-output factors and the high correlation of bank outputs with
general macroeconomic variables. Another major challenge is the problem
of aggregation bias, which has a long history in macroeconomics literature.
Crudely speaking, aggregation bias occurs when micro-level data patterns are
distorted and not represented consistently in the macro aggregate constructed
from them. In other words, incorrect specification of variables and the func-
tional relationship between inputs and outputs due to incorrect or theoreti-
cally inconsistent aggregation functions can roughly be called aggregation bias.
This problem violates the condition of consistent aggregation (Kim, 1986),
resulting in a biased measure of aggregate output, which will be incapable of
genuinely resenting the level and variations of the underlying individual output
measures. As many studies in the literature have used individual bank-level data
for analysis, this might not have been a major concern therein. However, this
issue deserves sizeable attention for sectoral and aggregate studies. The national
income measures of output undertake aggregation while dealing with these
problems. Butter (2007) summarises the main procedures used under the SNA.

Some econometric dimensions in the literature

Performance analysis of banks has been undertaken using a diverse set of
empirical methods, data types and underlying issues, as examined in Table 2.1.
Estimation of the production or cost function has been the most frequently
undertaken exercise in the banking literature. Depending on the type of data,
researchers have faced econometric challenges. The use of panel and cross-
sectional data has been widespread in the literature. For cross-sectional data, dealing
with heteroskedasticity has been a concern. Studies have generally adopted
two approaches to this problem: the first has been to normalize all the inputs
and outputs (all the variables in the cost, revenue, profit or production function)
60 Survey of evidences
by total assets (e.g. Vennet, 2002, among others), and second has been to use
weighted least squares method wherein weights are specified as the total asset
variable (e.g. Bos & Kolari, 2005). Another approach has been to use heter-
oskedasticity-adjusted standard errors in the OLS estimation. For panel data,
the main problem has been the lack of balanced panels. This is managed by
researchers using appropriate panel econometric models. Generally, production
and cost functions have been estimated using panel OLS for panel data.
Researchers have often had to use unbalanced panels due to data constraints.
Some studies have used industry-level aggregate data. Using aggregate data,
industry-level studies estimate the average production functions rather than
firm-specific ones. Generally, these studies employ an OLS or a modified ver-
sion. But “proximity to an OLS production function does not necessarily mean
productivity is maximized” (Colwell & Davis, 1992, p. 119). Yet, researchers
have had to accept this constraint due to data limitations and the use of single-
stage production functions. For studies using time series data, the usual issues
of non-stationarity, unit root and other related matters have been faced by
researchers. Given many innovations in time-series econometrics and compu-
tation technology, handling econometric challenges for time-series data has not
been a major problem for the literature.
Several researchers have used large sample sizes. Given the large volume
of data, cleaning the same has been necessary. This is particularly important
because outliers can bias the performance estimates. Weill (2004) recom-
mended a novel approach, which used the turkey-box plot method to clean
their data before empirically estimating the cost function. Lastly, simultaneity
bias has been a problem recognized by many analysts in the banking output
literature. Inputs and outputs may be determined simultaneously, and hence
their separation in the production or cost function might be incorrect. In this
case, literature has recognized at least six approaches: use of trans-log or the
more flexible Fourier production and cost functions, use of innovative meth-
ods such as the doubly indirect inference method of Kouki et al. (2014) or
Fuzzy methods or the indirect inference method of Gourieroux et al. (1993),
redefining the variables in the function in different terms (such as in ratio form
rather than level form), adopting a multi-stage estimation procedure rather
than single-stage estimation procedure (Kouki et al., 2014), avoiding the OLS
method which is unsuitable when face with this issue and finally using panel
data instead of time-series data.

Monetary policy and the banking production function

Banks are constrained in their intermediation process by monetary policy
interventions. Thus, under the intermediation framework, a production func-
tion that links the inputs and outputs without accounting for monetary policy
might result in biased estimates of performance parameters such as efficiency,
productivity, and others. Across the nations, Central Banks are actively engaged
in banking sector management, which can induce exogenous changes in the
 Review of evidence on banking output measurement 61
production function of the banks that might be mistaken for productivity and
efficiency changes rather than a result of monetary policy shocks. Many analysts
state that reserve requirements can shift the production function exogenously,
thereby producing residual productivity changes that result from monetary pol-
icy shock rather than any genuine advance in technology or efficiency (Green-
baum, 1967). Hence the inclusion of monetary policy as a controlling factor
in the production process of the banking sector seems inevitable. Researchers
have asserted that efficiency, productivity and profitability changes in the com-
mercial banking sector are mainly explained by monetary policy (Jacob, 1964).
Lastly, a very remotely highlighted issue in the literature is the impact of the
self-interest maximizing behaviour of regulators on the banking output. Boot
and Thakor (1993) analyse the theoretical implications for bank output meas-
urement of such economic behaviour on the part of the banking regulators.

Other issues
The first issue here is true versus observed interest rates in banking literature.
Interest rates are used as the price of banking output and hence employed as
weights in aggregation functions for building measures of total output. The
true measure of interest rate should be free of imperfections in the banking
market, particularly should not be affected by monetary policy interventions.
Such an interest rate will reflect the actual market price of banking output.
However, it is not observable in the real world, and hence one has to depend
on observed interest rates, which are strongly affected by non-market forces
(Schweitzer, 1972). Market imperfections generally mar observed rates, and
one must accept the same as given. These imperfections are not solely due to
monetary policy interventions. “Even if there is no central bank regulation on
the deposit or lending rates of banks . . . , these rates fail to be market-clearing,
owing to various kinds of imperfection in the credit market” (Das & Maiti,
1998, p. 3084). The second issue is the rise of the NBFC sector as a substitute
for commercial banking lending.
NBFC segment is increasingly competing with the banking industry as far
as the loan segment is concerned and the time deposits are concerned. Ignor-
ing their activities while measuring the performance of the banking segment
may not be correct and might distort the results. Similarly, the output of the
banking industry might be better measured if the effects of the output of the
NBFC industry are taken into account. Limited progress has been made in this
regard. Lastly, the banking industry’s treatment of payments service output has
received some attention in the literature, and some analysts have recommended
treating it as a separate output in itself. However, its measurement is chal-
lenging and has more extensive data requirements. In this regard, two studies
worth noting are Radecki (1999) and Athanasoglou et al. (2009). Readers can
refer to these works for more analysis on how to measure the payments service
output of commercial banks. Finally, a somewhat lesser debated issue has been
the measurement of central bank output and its inclusion in the traditional
62 Survey of evidences
banking output measures. Bhuyan (2016) summarizes the debate quite well in
the context of emerging economies. Accounting for central bank output into
the estimates of aggregate output of commercial banks introduces aggrega-
tion problems, particularly in determining the optimal weights for aggregation.
More fundamentally, both institutions produce very different utilities for the
economy and individual agents. Hence, literature has generally treated central
bank output measurement as a separate issue in itself rather than subsuming it
into the ambit of traditional banking output measurement. Thus, this aspect
has not been delved into in detail in this work.
This concludes the survey of literature on bank output measurement. The
following section provides an appendix on the evolution of bank output meas-
urement from 1957 to 2022.
Appendix
Evolution of banking output measurement in the extant wisdom from 1950
to 2022

No. Citation Issue Output Specification F or S N or R Approach

1 Farrell (1957) TE Value Measure F NOM NA

2 Speagle and LP Amount of Deposits; Number of cheques handled; Number of debit F, S R (CP); R PRA
Kohn (1958) transactions; Amount of loans and investments; Number of deposit (VL)
accounts
3 Benston (1965) CE Average number of month-ending deposit; Average number of month- S R (VL) PRA
ending loans accounts
4 Greenbaum CE Total earning assets (loans plus investments); current operating earnings; F; S R (VL), R PRA
(1967) Lending output was defined as the gross yield-weighted sum of the (CP)
diverse earning assets in each bank’s portfolio;
5 Gorman (1969) BO; LP Real Gross Domestic Product (GDP) of Commercial Banking industry F R (CP) PRA
6 Powers (1969) CE Sum of amount of lending, non-lending and interest-differential S NOM PRA
weighted time deposits; For majority firms: Correct measure of
Quantity of output = Value of output divided by price; For banking
services: Not necessarily so because physical measures will give biased
picture as the interest rates (price) component is an inherent part
of the banking output; Lending (loans) and non-lending (deposits)
outputs need to be estimated using both the price (interest rate) and
volume
7 Khusro et al. OE Amount of Deposits S NOM Reverse
(1971) INA
8 Benston (1972) EOS Highlights following output variables from literature: demand deposits, F; S NOM PRA;
time deposits, real estate loans, instalment loans, business loans and INA
securities, number of deposit and loans accounts; gross operating
income; total deposits; total assets; number of deposit and loan
accounts

(Continued)
(Continued)

64 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
9 Murphy (1972) CE number of: demand deposit accounts, time deposit accounts, instalment S R (VL) PRA
loans, business loans, real estate loans, safe deposits (banking activity);
average size of these accounts (within each activity)
10 Schweitzer EOS Financial Assets (investments in equity, government securities, interbank S NOM INA
(1972) instruments, money market instruments)
11 Coyne (1973) PRF Demand deposits; Time deposits; Total Capital accounts S NOM PRA
12 Kalish and OE Loans plus investments of banks; revenue of banks adjusted for their F; S NOM INA
Gilbert (1973) market power
13 Swamy (1973) EOS Total deposits S NOM PRA
14 Singh (1974) PRF Balance Sheet total; Deposits and Borrowings combined; Loans and S NOM PRA
Investments combined
15 Longbrake and CE Amount of total deposits S NOM PRA
Haslem (1975)
16 Sato (1976) ROM NOM Value added = Value of gross output – Cost of materials or NA NA NIA
intermediate inputs; Real Value added = NOM Value Added deflated
by a Price Index where the choice of price index is the most serious
challenge; Single deflation versus Double deflation issues arise;
Highlights the properties required in the ideal deflator of value added;
Suggests that R value added is an intermediate output and not a final
output; Favours Divisia price index for deflation rather than Laspeyer’s
price index
17 Sealey (1977) CE Amount of Loans; Amount of Securities (investments); Amount of S NOM INA
Excess Reserves
18 Geehan and ROM Index of the volume (quantity-based) of 74 non-loan activities (deposit- F; S R (VL), R PRA
Allen (1978) side); Index of the amount of five major types of loans (CP)
19 Mullineaux PRF Real Estate Loans, Consumer Instalment Loans, Commercial & S NOM INA
(1978) Agricultural Loans, Safe-Deposit Rental Fees
20 Ojo (1979) PA Performance indicators used: Amount of Deposits; Amount of Loans and F; S NOM PRA
Advances; Total Earnings; Net Profit Before Tax; Net Profit After Tax
21 Benston et al. EOS Divisia index: 1. of average number of deposit and loan accounts, 2. of S NOM PRA
(1982) average account size for deposits and loans; numbers of deposit and
loan accounts; amount of deposits and loans
22 Brand and Duke LP Output Index = Number of demand deposit transactions; number of F; S NOM PRA
(1982) transactions on stock exchanges (which imply increased usage of
banking services for payments and settlement); amount of time deposits;
amount of savings deposits; number of long-term loan contracts;
consumer credit operations; real estate loans; trust department services;
23 Angadi and LP Total working funds (credits and deposits) F; S NOM PRA
Devaraj (1983)
24 Gilbert (1983) EOS Amount of demand deposits S NOM PRA
25 Verghese (1983) PRF Gross Profit; Net Profit; Operating Margin; Gross Yield; Average S; Ratio of NOM NA
Earnings on earnings assets; Average cost of funds; Average excess F to S

Review of evidence on banking output measurement 65

CRR, SLR; Opportunity cost of CRR, SLR; Cost of Funds
26 Clark (1984) EOS Lending output: gross yield-weighted sum of earning assets; Lending S NOM INA
plus Non-lending output: lending output + the difference between
the bank’s total operating income and its operating income from
lending; Earning assets output: un-weighted sum of the amounts of
bank’s earning assets
27 Kim (1985) EOS Divisia aggregate of total Bank Output as done in Benston et al. (1982) S NOM PRA
28 Nelson (1985) EOS Amount of total deposits produced by each branch S NOM INA
29 Hunter and TC; EOS Main output variable: Sum of all loans, total investment securities, and S NOM PRA
Timme (1986) total deposits less reserves for losses on loans and securities; Alternative
measures for sensitivity analysis: (1) total loans, (2) total net loans and
deposits, and (3) total loans and securities less reserves for losses.
30 Kilbride et al. EOS Same as in Clark (1984) S NOM INA
(1986)
31 Kim (1986) OA Output is specified in terms of demand deposits, foreign currency S; S*F R (CP) PRA
deposit securities, and loans, all of which are calculated as yearly
average of monthly balances in inflation deflated terms (by applying
the CPI); Deposit related variables are estimated in terms of the
“moneyness” concept and thus is defined as the product of money
stock [stock variable] and the turnover rate [flow variable])
32 Owualah (1986) EOS Amount of total assets S NOM INA

(Continued)
(Continued)

66 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
33 Todhanakasem EOS Rate of interest on total loans S (PR) NOM INA
et al. (1986)
34 Berger et al. EOS Production approach: number of following accounts: 1) demand S R (VL); PRA;
(1987) deposits, 2) time and savings deposits, 3) Real estate loans, 4) NOM INA
commercial loans, 5) instalment loans; Intermediation approach:
Intermediation approach: amount intermediated (meaning unclear;
probably implies the amount of assets)
35 Humphrey EOS Amount of total assets S NOM INA
(1987)
36 Agu (1988) BO Amount of Deposits; Amount of earning assets; Number of cheques F; S NOM; R PRA
processed; Amount of cheques processed (CP); R
(VL)
37 Evanoff et al. EOS Amount of Total Assets; Total Loans; Total Loans and Deposits S NOM PRA
(1989) combined; Total Loans, Deposits and Investment Securities
38 Kim and Weiss TFP Output is specified in terms of demand deposits, foreign currency S; S*F R (CP) PRA
(1989) deposit, securities and loans, all of which are calculated as yearly
average of monthly balances in inflation deflated terms (by applying
the CPI); Deposit related variables are estimated in terms of the
“moneyness” concept and thus is defined as the product of money
stock [stock variable] and the turnover rate [flow variable])
39 Lawrence (1989) EOS; Amount of Deposits (includes demand deposits and time deposits, S NOM PRA;
EOSCO including certificates of deposit, $100,000 and over); Amount of loans INA
(includes real estate loans made and serviced, commercial, consumer,
construction, agricultural and other loans); Amount of investments
(includes short-term money market instruments and long-term
securities held)
40 Elyasiani and TE; TC Loans: (1) real estate loans, (2) commercial and industrial loans, and (3) S NOM INA
Mehdian other loans; Investments: includes all the securities other than those
(1990) held in the bank’s trading accounts
41 Aly et al. (1990) TE; SE; Amount of: real estate loans; commercial and industrial loans; consumer S NOM INA
AE loans; all other loans; demand deposits
42 Kolari and EOS; Amount of advances outstanding (credit outstanding); Amount of bills S NOM INA
Zardkoohi EOSCO outstanding (written orders requiring the bank to extend a specified
(1990) amount of credit to the borrower)
43 Noulas et al. RTS Loans to individuals for household, family, and other personal expenses; S NOM INA
(1990) loans secured by real estate; commercial and industrial loans; federal
funds sold, securities, total investment securities and assets held in
trading accounts
44 Oral and Yolalan TE; PRF Service efficiency outputs: amount of time spent on general service F; S R (VL); PRA
(1990) transactions (accounting, control, information, transfers, payments); NOM (Effi-
amount of time spent on credit transactions (contracts, guarantees, ciency);
credit and risk related procedures); amount of time spent on deposit INA
transactions (commercial accounts, saving accounts); amount of time (Profit-

Review of evidence on banking output measurement 67

spent on foreign exchange transactions; Profitability outputs: interests ability)
earned on loans; non-interest income
45 Berg et al. (1991) TE Amount of: demand deposit services; time deposit services; short-term S NOM PRA
loan services; long-term loan services; other services (brokerage
services, property management and the provision of safe deposit boxes)
46 Goldberg et al. EOS; Revenue from brokerage operations (commissions, margin interest F NOM PRA
(1991) EOSCO and fees from mutual funds sales); revenue from underwriting and
capital positioning operations (underwriting fees and gains on trading
and investments for firms own accounts); revenue from account
supervision (account management fees, research services, mergers and
acquisitions and other revenue)
47 Gropper (1991) EOS Amount of: investments; total loans; trust accounts S NOM INA
48 Hunter and TCH; Amount of loans (Real estate, agricultural, commercial and industrial, S R (CP) INA
Timme (1991) RSE; personal, credit card and other loan); Amount of deposits (demand
TSB deposits, interest-bearing checking accounts, savings accounts and
small-time deposits)
49 Kolari and EOS; Model 1: Loans [Real estate; Instalment; Commercial and others]; S NOM PRA
Zardkoohi EOSCO Model 2: Deposits [Demand; Time]; Model 3: Loans and Deposits
(1991) together

(Continued)
(Continued)

68 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
50 Tachibanaki EOS; Output 1: Amount of lending adjusted for interest differential between S NOM PRA
et al. (1991) EOSCO interest charged and interest paid (interest charged is measured by (though
the interest received through lending and interest paid is measured claimed
by the interest paid divided by the total debt minus some receivables, to be
and amount of lending is measured by the amount of total lending; INA
Output 2: Expected profit adjusted for output 1, total cost, discount by the
rate assumed by shareholders, share value of the bank and rate of authors)
interest on deposits
51 Berg et al. (1992) MTFP Amount of loans: short-term; long-term loans; Amount of deposits: S R (CP) PRA
produced (non-bank) deposits [i.e. excludes interbank deposits]
52 Berger and CE Amount of – demand deposits; time and savings deposits; Real estate S R (CP) VAA
Humphrey loans; commercial and industrial loans; instalment loans
(1992)
53 Colwell and BO; PFP; Value Measures (Amount of loans, deposits, investments); Volume F; S R PRA;
Davis (1992) TFP Measures (Number of deposit and loan accounts); Index based INA;
measures using appropriate weights for aggregating various NIA
heterogeneous outputs
54 Elyasiani and OE; AE; Amount of: Commercial and industrial loans; real estate loans; other S NOM INA
Mehdian OTE; loans; investment securities
(1992) PTE; SE
55 Hughes and EOS; Amount of loans: Commercial real estate loans; commercial loans (C&I S NOM INA
Mester (1993) EOSCO loans and loans for securities); consumer loans; other loans; securities,
assets in trading accounts, fed funds sold, and total investment
securities
56 Mester (1992) CE Amount of different information produced by banks (proxied by amount S NOM INA
of loans)
57 Dietsch (1993) EOS; Total Deposits; Total of loans to firms and households; Long-term S NOM INA
EOSCO securities (assets); Interbank market activity (interbank liabilities net of
interbank assets)
58 Muldur and EOS; Cobb-Douglas case: Amount of total assets (Commercial banks); S NOM INA
Sassenou EOSCO Number of accounts (Savings institutions); Trans-log case: Amount of
(1993) commercial loans, interbank loans, securities transactions (Commercial
Banks); Amount of customer deposits, commercial loans, securities
and cash management transactions (Savings institutions)
59 Shaffer (1993) CE; CMP Amount of assets (loans, investments, etc.) S R (CP) INA
60 Subrahmanyam BO Amount of deposits; Amount of loans S NA PRA
(1994)
61 Clark and EOS; Transactions deposits; Purchased Funds; Time Deposits; Investments; S NOM PRA
Speaker (1994) EOSCO Real estate and mortgage loans; Instalment loans; Credit card loans;
Commercial and other loans
62 Keshari and Paul TE Amount of total deposits + Amount of total advances S NOM PRA

Review of evidence on banking output measurement 69

(1994)
63 Sherwood (1994) BO Services produce changes and thus the unit of output may be intangible; NA NA NA
Highlights the following major difficulties: (1) enumerating the
elements of a complex bundle of services, (2) choosing among
alternative representations of an industry’s output, (3) accounting
for the consumer’s role in the generation of a service output, and
(4) specifying clearly whether what is being worked with is a value-
added model or a gross output model; Suggests that service prices are
included implicitly in many cases and thus decomposing the observed
prices into explicit and implicit components is important; Consumer
involvement is inevitable in production (quantum of services versus
quantum of outcomes); Highlights the problems in measuring service
output when product and process innovations occur
64 Evanoff and CE Amount of total loans and securities S NOM PRA
Israilevich
(1995)

(Continued)
(Continued)

70 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
65 Hunter and CE (1) wholesale loans (average money balance of all commercial and S NOM VAA
Timme (1995) industrial and security loans); (2) consumer loans (average money
balance of credit cards and other personal loans, excluding loans
secured by residential real estate); (3) Real estate and other loans
(average money balance of all loans secured by real estate, agricultural,
and other loans not included in other variables); (4) other BO; (5)
average money balance of transaction deposit accounts; (6) average
money balance of retail non-transaction accounts less than $100,000
66 Wheelock and TE Amount of total earning assets (sum of amount of total loans and bond S NOM INA
Wilson (1995) holdings); Amount of demand deposits
67 Zaim (1995) TE; AE Total balance of demand deposits; Total balance of time deposits; Total S NOM INA
balance of short-term loans; Total balance of long-term loans
68 Batra (1996) PRF Gross rates on: loans and advances; investment in govt. and other F (PR) NOM INA
approved securities; investment in other securities; bills purchased and
discounted
69 Clark (1996) CE; SE Commercial and industrial loans (commercial loans, direct leasing, S NOM VAA;
foreign loans, and loans to financial institutions); consumer and real INA
estate loans; total securities (investment securities and trading account
securities); core deposits (demand deposits, savings deposits and retail
time deposits)
70 Hartwick (1996) BO Net interest income (Interest income from loans and investments F R (CP) INA
combined and net of interest expenses) adjusted for – 1. Service
charges, 2. Loan losses (as input ex-ante, and net out of output
ex-post), 3. Capital losses (when accounting for equity)
71 Humphrey and PRF Amount of core deposits (demand deposits plus savings and small S R (CP) INA
Pulley (1997) denomination time deposit); Amount of loans (the sum of the R
values of real estate, commercial and industrial, and instalment,
including credit card, loans)
72 Jagtiani and TCH; Amount of: deposits; loans and investments; OBS products (OBS S R (CP) INA
Khanthavit RSE; guarantees, interest rate and foreign currency swaps, options, and
(1996) EPSE; futures and forward contracts)
EPSA
73 Kaushik and PRF Annual growth rate in Assets – Total and individual items in balance S NOM NA
Lopez (1996) sheet such as interest earning assets, loans and non-interest earning
assets; Annual growth rate in Liabilities – Total, Foreign, Domestic
(Demand, Other Checkable, Savings), Subordinated notes, Equity
capital, Loss provisions
74 Sharpe (1997) BOP Amount outstanding in: six-month certificate of deposits; money-market F NOM PRA
deposit accounts
75 Berger and CE Commercial loans; consumer loans; real estate loans; transactions deposits F; S NOM PRA
DeYoung (transactions deposits include demand deposits, NOW accounts,
(1997) automatic transfer service accounts, and telephone and pre-authorized
transfer accounts); fee-based income (fee-based income equals gross
non-interest income less both service charges on deposit accounts and

Review of evidence on banking output measurement 71

gains/losses from securities and foreign exchange trading)
76 Berger and EFFS Production approach (recommended for branch-level data): Number F (Ideal); S NOM PRA;
Humphrey of deposit accounts, loan accounts or deposit transactions, loan INA
(1997) transactions, produced in a given year (flow concept); Intermediation
approach (recommended for institutions-level data): Flow amount of
deposits, loans combined in a given year (flow concept); However,
data constraints force the use of stock variables instead of flow
variables; Highlights the asset, user cost, and value-added methods of
specifying output and notes the debate on treatment of deposits as an
input or an output.
77 Berger and CE; PE Amount of: consumer loans (instalment and credit card and related S R (CP) INA
Mester (1997) plans); business loans (all other loan); securities (gross total assets less
loans and physical capital, so that all financial assets are considered to
be outputs)
78 Bhattacharyya TE Amount of advances (loans); Amount of deposits; Amount of S NOM INA
et al. (1997) investments
79 Chaffai (1997) ISTE Amount of total loans S NOM INA
80 Das and Maiti PA Change (First difference) in the amount of credit flowing from the F NOM PRA
(1998) organised market, particularly from the commercial banks; Change in
the amount of deposits with organized commercial banks

(Continued)
(Continued)

72 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
81 Mester (1997) CE Amount of: Real estate loans; Commercial loans (commercial and S NOM INA
industrial loans, lease financing receivables, agricultural loans, loans to
depository institutions, acceptances of other banks, loans to foreign
governments, obligations of states and political subdivisions, and other
loans); loans to individuals
82 Resti (1997) CE Amount of loans: all forms of loans to customers except non-performing F; S R (CP) PRA
loans; Amount of deposits: checking accounts, savings accounts,
certificates of deposit with retail customers (excluding interbank
CDs); Non-interest income: Net fee-based income, net revenues from
security and currency trading
83 Anthanassopou- ME; CE Market efficiency: Liability sales; Loans & mortgages; Insurances & F; S R (VL); PRA
los (1998) securities; Cost Efficiency: Number of transactions; Liability sales; NOM
Loans and Mortgages; Insurance and securities; Number of cards
84 Ayadi et al. TE Amount of: total loans; interest income; non-interest income S; F NOM INA
(1998)
85 Berger and CE Amount of: Demand deposits; time and savings deposits; Real estate S NOM PRA
Hannan (1998) loans; commercial loans; instalment loans
86 DeYoung and PE Amount of: total loans; transactions deposits; fee-based financial services F; S NOM INA
Hasan (1998) (fee-based financial services equals non-interest income less service
charges on deposit accounts)
87 Durkin and EOS; CEL Number of loans outstanding (that is, serviced) S R (VL) PRA
Elliehausen
(1998)
88 Freixas (1998) EFFS Payment volume; Payment value; Net payments system is defined as F NA NA
interbank payments system with net clearing of funds on daily basis;
Gross payments system is the settlement of claims by banks through
intense use of liquidity; Suggest that net payments system generates a
larger risk; Net system saves liquidity (central bank money available
with the banks), while Gross system reduces the risk of contagion by
one-to-one settlement but intensely uses liquidity
89 Grifell-Tatjé MTFP Amount of: loan accounts; savings accounts; checking accounts S R (Constant) VAA
et al. (1998)
90 Hughes and CEL Real estate loans; business loans (i.e., commercial and industrial loans, S NOM INA
Mester (1998) lease financing receivables, and agricultural loans); loans to individuals;
other loans; amount of earning assets (securities, assets in trading
accounts, fed funds sold and securities purchased under agreements
to resell, and total investment securities); amount of off-balance-sheet
output [credit-risk equivalent amount of off-balance-sheet items (e.g.,
commitments, letters of credit, derivatives, etc.)]
91 Rogers (1998) CE; RE; Amount of: demand deposits; time and savings deposits; Real estate F; S NOM INA
PE loans; other loans; Net non-interest income
92 Boer (1999) FISIM Mortgages; Consumer credit; Credits granted to enterprises; New credit F R (VL); R PRA

Review of evidence on banking output measurement 73

granted to private enterprises; Total savings account business; Deposit (CP)
business; Money transfer services (excluding commercial money
transfers)
93 Camanho and TE; PTE; Number of general service transactions performed by branch staff; F; S R (VL); PRA
Dyson (1999) SE Number of transactions in external ATMs; Number of all types of NOM
accounts at the branch; Value of savings; Value of loans
94 Fixler and MTFP Asset side: Cash and Balances Due from Depository Institutions; F NOM PRA
Zieschang Amount of Loans; Amount of Securities; Amount of Federal Funds
(1999) Sold and Securities Purchased; Liability side: Amount of Deposits;
federal funds purchased and securities sold
95 Radecki (1999) BO Amount of non-interest revenue (it contains the fee-based income from F NOM PRA
payments services)
96 Srivastava (1999) EOS; CEL Amount of total deposits; amount of the sum of total advances and total S NOM VAA
investments
97 Thanassoulis TE; TFP Finnish bank: Number of transactions processed by human tellers; F R (VL) PRA
(1999) Number of cash withdrawals; Number of loans processed; Number
of transactions effected at automatic teller machines; British bank:
Number of mortgage applications; Number of insurance policies sold;
Number of new savings accounts; Number of transactions
98 Wheelock and TE; MTFP Real estate loans; commercial and industrial loans; consumer loans; all S R (CP) INA
Wilson (1999) other loans; total demand deposits

(Continued)
(Continued)

74 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
99 Altunbas et al. TE; SE; Amount of total loans; Amount of total securities; Amount of off- S R (CP) INA
(2000) XE balance-sheet output
100 Athanassopoulos PEF; INE; Production Efficiency: Number of deposit and transfer transactions; F R (VL) PRA
and Giokas XEF Number of credit transactions; Foreign-receipt transactions;
(2000) Intermediation Efficiency: savings deposits; current deposits; demand
deposits; time deposits; total loans; and non-interest income;
101 Hughes et al. EOS Amount of: liquid assets; short-term securities; long-term securities; S NOM INA
(2000) loans and leases net of unearned income; other assets
102 Tokle and Tokle CMP Amount of total deposits S NOM PRA
(2000)
103 Alam (2001) MTFP Amount of: Securities; Real estate loans; commercial and industrial S NOM INA
loans; instalment loans
104 Altunbaş et al. CE; EOS Amount of: total aggregate loans (all types of loans); total aggregate S R (CP) INA
(2001) securities (short-term investment, equity and other investments and
public sector securities)
105 Karim (2001) CE Amount of: commercial and industrial loans; other loans; time deposits; S NOM INA
demand deposits; securities and investments
106 Mertens and CE; PE; Amount of: Interbank loans; Consumer loans; Other investments S NOM INA
Urga (2001) EOS; (including government and risky securities, and investment in other
EOSCO enterprises)
107 Okuda and Saito EOS Amount of: interest income (from loans, securities and deposits); Fee- F NOM INA
(2001) based income (sum of commissions and fees)
108 Adams et al. MPR Amount of: commercial and industrial loans; instalment loans; real estate S R (CP) INA
(2002) loans
109 Chaudhuri PRF Yield from: loans and advances; Investments Ratio of F NOM INA
(2002) to S
110 Clark and Siems CE; PE; Amount of total credit equivalent amount of OBS transactions S NOM INA
(2002) XE (constructed following the guidelines set out by the Basel Committee);
aggregate measure of asset equivalence that utilizes the rate of return
balance-sheet assets to capitalize the non-interest income from OBS
activities; non-interest income; commercial loans, direct leasing,
foreign loans, and loans to financial institutions; consumer and Real
estate loans; other on-balance-sheet assets (defined as Total Assets
minus commercial loans minus consumer and real estate loans)
111 Das (2002) MTFP Amount of: Bank credit; total investments S R (CP) INA
112 Devaney and PE Amount of: Real estate loans; four types of commercial and industrial S NOM UCA
Weber (2002) loans (larger than $1million, between $250,000 and $1 million,
between $100,000 and $250,000, $100,000 or less); personal loans;
securities (investments); transaction account deposits
113 Grigorian and TE Set 1: revenues (Revenues are defined as the sum of interest and non- S NOM VAA
Manole (2002) interest income), net loans (Net loans are defined as loans net of loan
loss provisions), liquid assets (Liquid assets include cash, balances with
monetary authorities, and holdings of treasury bills); Set 2: deposits,
net loans, liquid assets
114 Isik and Hassan CE; APE Short-term loans (with less than one year of maturity); long-term loans F; S NOM INA
(2002) (loans with more than a year maturity); Risk-adjusted off-balance- (Converted

Review of evidence on banking output measurement 75

sheet items (guarantees and warranties, letters of guarantee, bank to US
acceptance, letters of credit, guaranteed pre-financing, endorsements dollars)
and other commitments, foreign exchange and interest rate derivatives
as well as other off-balance-sheet activities); Other earning assets (loans
to special sectors, interbank funds sold, and investment securities
which includes treasury bills, government bonds and other securities)
115 Kumbhakar et al. LP Amount of: loans to the public; guarantees; deposits S R (CP) PRA
(2002)
116 Lozano-Vivas TE; CE Amount of: Loans; Deposits; Other Earning assets S NOM PRA
et al. (2002)
117 Mohan (2002) PA Interest spread; Amount of net profit F NOM NA
118 Tortosa-Ausina CE; BO Intermediation: Amount of – loans (all forms of loans to all categories S R (CP) INA;
(2002) of customers); Other earning assets (securities and loans to financial VAA
institutions); Value-added: Amount of – loans (all forms of loans to all
categories of customers); Other earning assets (securities and loans to
financial institutions); Deposits (savings, time and transactions)
119 Vennet (2002) CE; PE Intermediation approach – Amount of: total loans, total securities F; S Converted to INA;
(investments); Value added – total interest income, total non-interest ECU VAA
revenues
120 Bhattacharya Bank Statistical distribution of: the log of total assets; market shares; Finds NA NA NA
(2003) Scope selectivity bias
(Continued)
(Continued)

76 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
121 Fixler et al. BO Total Imputed Output = [Implicit output (Amount of loans effective
*
F NOM VAA
(2003) interest charged + Amount of Deposits*effective interest paid, effective
being net of reference rate) + User cost of own funds (Reference
rate*Net Assets, where net assets = total assets – total liabilities)]; User
cost of own funds compensates the banks for using equity capital or
any other form of self-acquired financial capital
122 Focarelli and M&A Size is measured by the amount of total assets; Service quality measures: S NOM; R INA
Panetta (2003) number of branches per 100,000 inhabitants, average number of (VL)
employees per branch in each local market, size of the bank
123 Kumbhakar and CE; TFP Amount of: fixed deposits; savings deposits; current deposits; S R (CP); R VAA
Sarkar (2003) investments; loans and advances; Branch output – Number of rural, (VL)
urban and semi-urban, and metropolitan branches
124 Leong et al. TE Amount of: Loans, Interest income, Other income, Risk-weighted S NOM INA
(2003) financial assets (investments plus OBS output with items weighted
as per Basel convention with weights being 0, 10, 20, 50 and 100%
percent for five categories of assets)
125 Mohan (2003) PA Stock market returns of the banks Ratio of F NOM NA
to S
126 Mukherjee et al. QE; PE; Quality efficiency: Service quality output variable (constructed using F; S NOM; PRA
(2003) OE Likert scale responses of 185 customers of nationalized banks Subjective
tangibility, responsiveness, reliability, assurance and empathy); Profit Ranking
and overall efficiency: Amount of: deposits (demand, savings and term
deposits); advance (bills purchased, cash credit, overdraft, loans); non-
interest income (comprising commission, exchange, brokerage, profit
from investments, assets deployed and exchange transactions)
127 Rushdi and PRF; LP Amount of: interest income from bank loans, fee and other income; F; S NOM VAA
Tennant total R assets (Assets)
(2003)
128 Sathye (2003) TE Model 1 – Amount of: net interest income, non-interest income; Model F NOM INA;
2 – Amount of: net loans, non-interest income VAA
129 Agu (2004) PA Amount of deposits; Amount of loans; Operating asset ratio; Operating S; Ratio of NOM PRA
income ratio F to S
130 Bossone and Lee EOS Sum of amount of total loans and other earning assets (the average S R (CP) INA
(2004) amount at the end of each year)
131 Casu et al. MTFP Amount of: total loans; other earning assets; off-balance-sheet activities S; F NOM INA
(2004) at NOM value
132 Cuesta and TE Loans (to non-bank sector); Investments (Bonds, cash and other assets); S; F R (CP) INA
Zofío (2005) Off-balance-sheet output (non-interest income)
133 De (2004) TE Gross income (equivalent to interest and discount earned plus F; S R (CP) INA
commission, exchange and brokerage plus other receipts); Amount of
total earning assets (total loans and investments at the end of the year)
134 Fixler (2004) BO Deposit-side services; Borrowing-side services; Implicit + Explicit F NA VAA
outputs should be included

Review of evidence on banking output measurement 77

135 Girardone et al. EOS; XE Amount of: total loans; securities (investments) S R (CP) INA
(2004)
136 Maghyereh PE Amount of: earning assets (loans and liquid assets & investments); Off- S; F R (CP) INA
(2004) balance-sheet business; other services
137 Mohan and Ray RE Amount of: loans; investments; other incomes S R (CP) INA
(2004)
138 Triplett and BO; IO Banking: Gross Output of non-depository financial institutions, Value F NOM PRA
Bosworth Added; Insurance: Premiums minus Claims, Risk-adjusted premiums
(2004a)
139 Triplett and BO; FISIM Criticize the one-stage production functions for ignoring the financial F NA INA
Bosworth constraints; Inclusion of reserves in the production function is
(2004b) desirable; Incorporate implicit services to depositors in the overall
output measure
140 Weill (2004) TE Amount of: loans; investments S NOM INA
141 Bos and Kolari XE; EOS; Amount of loans (aggregate of commercial and industrial, real estate, S; F R (CP) INA
(2005) EOSCO consumer, agriculture, and other outstanding credit); Amount of
investments (securities, equity investments, and all other investments
reported on the balance sheet); Off-balance-sheet output (Off-
balance-sheet activities are credit items and other guarantees, loan
commitments, derivative securities, and other loan and securities
exposures not reflected on the balance sheet)

(Continued)
(Continued)

78 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
142 Camanho and CE Production approach: Total value of deposits; Total value of loans; S; F0 NOM; R PRA;
Dyson (2005) Total value of off-balance-sheet business; Number of general service (VL) VAA
transactions; Value-Added approach: Total value of deposits; Total
value of loans; Total value of off-balance-sheet business.
143 Das et al. (2005) TE; CE; Amount of: investments, performing loan assets, other non-interest fee- S; F NOM INA
RE; PE based incomes
144 Sensarma (2005) CE Amount of: fixed deposits; saving deposits; current deposits; loans; S R (CP); R PRA
investments; and Number of branches (VL)
145 Tirtiroğlu et al. TFP Sum of amount of real deposits and loan balances S R (CP) VAA
(2005)
146 Wang and Basu FISIM Stress on the need to incorporate certain minimum levels of risk into NA NA NA
(2005) the measure of reference rate rather than searching for a pure risk-free
reference rate; Reject the practice of using a single reference rate in
NIPA and recommend using multiple reference rates to reflect the
true nature of FISIM output better; Recommend use of separate
reference rates for deposit side and loans side FISIM components
147 Beccalli et al. TE Amount of: loans; earning assets S NOM INA
(2006)
148 Dick (2006) PA Performance indicators used: Loan interest rate; Deposit interest rate; F NOM NA
Service fees; Spread; Operating costs; Credit risk; Profit margin
149 Havrylchyk CE; AE; Amount of: Loans; Treasury bonds; Off-balance items S NOM INA
(2006) TE;
PTE; SE
150 Howcroft and MTFP Loan-based model: Amount of – loans, investments; Income-based S; F R (CP) INA
Ataullah model: Amount of – interest income, non-interest income
(2006)
151 Kirkwood and CE Technical Efficiency: Amount of – Interest-bearing assets, Non-interest S; F R (CP) INA
Nahm (2005) income; Profit Efficiency: Profit before tax and abnormal items
152 Lyroudi and MTFP Amount of: total deposits; total customer loans; investments (equity + S NOM VAA
Angelidis government securities held by banks)
(2006)
153 Nakane et al. MPR Amount of: time deposits; demand and savings deposits; loans S NOM PRA
(2006)
154 Reddy (2006) MTFP; Amount: liquid assets; total advances; total deposits; total income S; F NOM VAA
TE; SE (interest plus non-interest income)
155 Tadesse (2005) EOS; TC Sum of amounts of Personal loans + industrial loans + investments in S NOM INA
cash dues securities
156 Allen and Liu CE; EOS Consumer loans (amount of personal loans for non-business purposes); S R (CP) INA
(2007) Non-mortgage loans (amount of secured call and other loans to
investment loans dealers and brokers + loans to regulated financial
institutions + loans to domestic and foreign governments + lease
receivables + reverse repurchase agreements + loans to individuals and
others for business purposes); Mortgage loans (amount of residential

Review of evidence on banking output measurement 79

and non-residential mortgage loans); Investments (amount of other
financial assets on a bank’s balance sheet); OBS output (non-traditional
Asset-equivalent measure of non-interest income-generating activities)
157 Ariff and Can CE; PE Amount of: Loans (net loans and interbank lending); Investments (short, S R (CP) INA
(2008) long-term and entrusted investments)
158 Asaftei and CE Amount of: current loans; portfolio investment; fee and commission S; F R (CP); R INA
Kumbhakar income; and number of branches (VL)
(2008)
159 Berger and BO Four measures of liquidity created by banks: cat fat, mat fat, cat nonfat, S NOM PRA;
Bouwman mat nonfat VAA
(2007)
160 Casolaro and TFP Output 1: Composite of three services – i. turnover of current accounts S R (CP); R VAA
Gobbi (2007) as a proxy for payment services, ii. total amount of deposited (VL)
securities as a proxy for both safekeeping and asset management, iii.
loan guarantees as a proxy for off-balance intermediation activity,
Weighting factor for Output 1 is shares in total fees and commission
income as weights; Output 2 – outstanding amounts of short-term
loans, medium-term loans, long-term loans, deposits
161 Chen et al. (2007) CE; PE Number of investor accounts S R (VL) PRA

(Continued)
(Continued)

80 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
162 Dick (2007) PA; MPR Major explanatory variables: Amount of deposits; Amount of assets; Real S; F NOM PRA
estate loans; Commercial and industrial loans; Loans to individuals;
Leases; Non-interest income; Demand deposits; Time deposits; Rate
of individual loans; Service charges on deposits
163 Ray (2007) SZE Amount of: total credit (net of non-performing loans); Investments; S NOM INA
Other income
164 Shih et al. (2007) PA Four core measures of bank performance: overall performance, liquidity Ratio of F NOM NA
management, credit risk management, and capital profitability [Core to S
capital ratio; Capital risk ratio; Asset profitability; Doubtful loan ratio;
Short-term liquid asset-liability ratio; Risk-adjusted capital adequacy
ratio; Ratio of short to long-term debt; Lost loan ratio; Overdue loan
ratio; Capital profitability]
165 Harimaya (2008) EOS; Amount of: loans and bills discounted; investment securities and trading S R (CP) INA
EOSCO account securities; total liabilities in trust accounts
166 Pelosi (2008) MTFP Models for cost minimization function for Banks: Amount of – core S; F (non- R (CP) INA
deposit (transaction and savings deposits); loans and securities interest
(investments); Model for deposit-taking institutions: Amount of – income)
loans, non-interest income
167 Rezvanian et al. OE; AE; Amount of: Loans and Advances; securities (investments); other earning S NOM INA
(2008) OTE; assets
PTE; SE;
MTFP
168 Sinha and MTFP Amount of deposits + loans and advances S R (CP) PRA;
Chatterjee VAA
(2008)
169 Sufian (2008) TE; AE; Amount of: Total Loans; Investment and dealing securities S NOM INA
OE
170 Athanasoglou PFP; TFP Total Output Index (Tornquist): 1. Financial intermediation output – S; F R (CP); R PRA
et al. (2009) Sum of the amount of loans and deposits adjusted for their effective (VL)
interest rate (net of reference rate); 2. Payments services output:
(a) direct fees for payment services provided and (b) free payment
services provided through demand deposits, both weighted through
opportunity cost of deposits; Alternative variables used here were
number of transactions on a series of payments, including payments
through ATMs, credit and debit cards, credit transfers, direct debits
and cheques; 3. Other incomes (non-interest income – from fees and
securities); All three components are weighted by the percentage share
of each component in the value of total BO
171 Humphrey EOS Point of sale transactions (card and check); bill payments (electronic and S; F R (CP) but PRA

Review of evidence on banking output measurement 81

(2009) paper giro); ATM; Size-standardized branch office networks PPP in
current
price
172 Illueca et al. MTFP Amount of: Loans (to firms and households); Other loans (Cash and S; F NOM INA
(2009) balances with the Bank of Spain, interbank loans); Fixed-income
securities (Fixed-income securities, public debt); Other securities
(Shares and other equities such as quoted shares, unquoted shares,
other equity, mutual funds shares); Non-traditional output (non-
interest income)
173 Kao and Liu TE Amount of: demand deposits; short-term loans; long-term loans S NOM INA
(2009)
174 Kenjegalieva PA For incorporating risk in production behaviour: Amount of profit, F; Ratio of NOM INA
et al. (2009) Return on equity F to S
175 Koutsomanoli- CE; PE Amount of: loans (net of provisions); other earning assets (government S NOM INA
Filippaki and securities, bonds, equity investments, CDs, T-bills, etc.)
Mamatzakis
(2009)
176 Pasiouras et al. CE; PE Amount of: Loans; other earning assets; total deposits (customer and S R (CP) VAA
(2009) interbank)

(Continued)
(Continued)

82 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
177 Seelanatha and PA Amount of: Loans; Deposits; Investments S; Ratio of NOM NA
Wickremasin- S to S
ghe (2009)
178 Altman et al. PA Amount of loans; Returns on loans S; Ratio of NOM NA
(2010) S to S
179 Banker et al. TE Amount of: interest revenue; other operating revenue F R (CP) INA
(2010)
180 Chronopoulos TE Amount of: total customer loans; other earning assets; non-interest S; F NOM INA
et al. (2010), income
in Fiordelisi
et al. (2010)
181 Dong (2010) CE Amount of: total loans; Other earning assets (investments, interbank S; F R (CP) INA
assets); Non-interest income
182 Feng and Serletis TE; EC; Amount of: securities (all non-loan financial assets, i.e., all financial S R (CP) INA
(2010) EOS; and physical assets minus the sum of consumer loans, non-consumer
TFP loans, and equity); consumer loans; non-consumer loans (industrial,
commercial, and real estate loans)
183 Hsiao et al. OE Amount of: interest revenue (interest on loans, income from government S NOM INA
(2010) bonds and corporate bonds); non-interest revenue (services charges on
loans and transactions, income from renting and fiduciary activities,
commissions, and other operating income); total loans (short-term and
medium-term loans)
184 Lozano-Vivas CE Amount of: loans, other earnings assets, off-balance-sheet (OBS) items, S; F R (CP) INA
and Pasiouras non-interest income
(2010)
185 Margono et al. CE; EOS; Amount of: total aggregate loans; total aggregate securities S NOM INA
(2010) TFP
186 Murillo-Melchor MTFP Amount of: customer loans (all forms of loans performed by banks); S; F NOM INA;
et al. (2010) deposits (excluding interbank deposits); Securities and equity VAA
investments; other earning assets; non-interest income
187 Ray and Das CE; PE Investments (in approved and non-approved securities); Earning advances S NOM INA
(2010) (NPA adjusted bank credit); Other income (Fee income emanated
from the commission, exchange, brokerage, etc.)
188 Weyman-Jones BO Kenjegalieva et al. (2009) F; Ratio of NOM INA
et al. (2010), F to S
in Fiordelisi
et al. (2010)
189 Andries (2011) MTFP Amount of: total loans; total investments; Other income (non-interest S; F NOM INA
income)
190 Curi et al. (2013) TE Amount of: interbank loans; customer loans; securities; Off-balance- S; F R (CP) INA
sheet output (non-interest income)
191 Daisuke et al. PFP Output 1: Gross Output = Interest Receipt – Interest Payment; Output F NOM INA

Review of evidence on banking output measurement 83

(2011a) 2: B/S Adjusted Output = [Interest Receipt − Interest Payment] x
(Loan Outstanding/Deposit Outstanding); Output 3: B/S Adjusted
Output – expected average loan losses for next three years
192 Hon et al. (2011) TE; MTFP Amount of Loans and Advances combined S NOM INA
193 Huang et al. TE; AE Amount of: total loans; investments; non-interest income S; F R (CP) INA
(2011)
194 Rajan et al. TE; TFP Amount of: total loans; investments S R (CP) INA
(2011)
195 Banerjee (2012) REV Popularly used: Amount of: deposits; loans; Other earning assets; Less S; F NOM PRA;
frequently used: net loans; liquid assets INA
196 Diewert et al. BO Measures of output in terms of services rendered by banks to: 1. F NOM NA
(2011) Household sector, 2. Banking sector, 3. Non-financial sector and 4.
Owner-occupied housing sector
197 Royster (2012) BO Tornquist Index of output: Revenue-share weighted index of loan S NOM; R INA
quantity index and deposit quantity index (VL)
198 Boucinha et al. EOS; CE; Amount of: total loans (adjusted for securitization); other earning assets S NOM INA
(2012) CFP
199 Guarda et al. BO Amount of: loans; investments S R (CP) INA
(2013)
200 Huang and Fu CE; CGR; Amount of: total loans; total investments S R (CP) INA
(2013) MFCE

(Continued)
(Continued)

84 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
201 Inklaar and BO Count-based (US) versus Balance-based (EU); Quality-adjusted Counts- F NA INA
Wang (2012) based output; Reject stock measures in favour of flow measures
202 Lee and Kim MTFP Amount of: interest income; fee income F NOM INA
(2013)
203 Lee et al. (2013) CE; EOS; Amount of: total loans; investment; non-interest income S; F R (CP) INA
EOSCO
204 Sharma et al. REV Key findings: 1. Intermediation approach is the most popular approach NA NA NA
(2013) (56 out of 106 used it in their survey); 2. Production approach was
the second-highest used (21 out of 106); 3. Stock measures of output
dominated the literature reviewed by them; 4. DEA is the most
popular methodological choice (66 out of 106 studies)
205 Westhuizen MTFP Amount of: loans and investments; deposits S R (CP) PRA
(2013)
206 Davies and EOS Amount of: Liquid assets (cash and equivalents); Total investments S; F NOM INA
Tracey (2014) (includes securities, trading asset securities, and mortgage-backed
securities); Net loans (includes loans given to banks and customers,
financial leases, and total portfolio loans, and deducts allowance for
loan losses and other adjustments); Other non-loan assets (Total other
assets); Off-balance-sheet output (Non-interest income – service
activities charge of deposits)
207 Kouki et al. EOS Amount of: total loans (except the interbanks loans); Total securities; S NOM INA
(2014) Total amount of interbank loans
208 Oluitan (2014) CE Amount of: Loans; Other earning assets; Non-interest income S NOM INA
209 Tan and Floros PRF Profitability indicator: ROA (ratio of net income to total assets); ROE Ratio of F NOM NA
(2014) (ratio of net income to equity) to S
210 Tzeremes (2015) TE Amount of: Loans; Other earning assets S R (CP) INA
211 Allen et al. BO Key lending variable (dependent variable): R growth in total loans Ratio of F R (CP) INA
(2015) to S
212 Beccalli et al. EOS Amount of: Loans (Total amount of loans granted by banks, as expressed S; F R (CP) INA
(2015) in the balance sheet); Securities (Total amount of securities, as
expressed in the balance sheet); Off-balance-sheet items (comprising
managed securitized assets reported off-balance-sheet, other off-
balance-sheet exposure to securitization, guarantees, acceptance
and documentary credits, committed credit lines, other contingent
liabilities, as expressed in the balance sheet, the exact definition used
pre-and post-IFRS adoption)
213 Grover and TE Net-interest margin (i.e., spread); other income; bancassurance income F R (CP) PRA
Arora (2015) (income earned from -selling insurance policies as a corporate agent)
214 Huang et al. MTFP Amount of: total loans; investments S NOM INA
(2015)

Review of evidence on banking output measurement 85

215 Mamatzakis et al. TE; AE Amount of: Loans; Other earning assets (government securities, bonds, S NOM INA
(2015) equity investments, CDs, T-bills, etc.)
216 Mamonov and CE Amount of loans to households and nonfinancial firms; Amount of retail S; F NOM PRA
Vernikov and corporate deposits (excluding government and interbank accounts);
(2017) Amount of non-interest income (fee and commission income)
217 Mosko and CE Amount of: total loans; other earning assets S NOM INA
Bozdo (2016)
218 Phan et al. (2016) XE Amount of Loans S NOM INA
219 Salim et al. TE Desirable outputs: Amount of loans, Amount of earning assets; S NOM INA
(2017) Undesirable outputs: non-performing assets
220 Thilakaweera TE; MTFP Intermediation approach: Amount of loans and advances, Amount of S; F R (CP) INA;
(2016) investments; Operating profit approach: Amount of interest income, OPA
Amount of non-interest income
221 Zha et al. (2016) TE; CE Desirable outputs: amount of – interest income, non-interest income; F NOM INA
Undesirable output – non-performing loans
222 Abdul-Majid CE; TGR Conventional banks – Amount of loans, other earning assets; Islamic S NOM INA
et al. (2017) banks: Amount of financing, other earning assets
223 Badunenko and CE; TC Amount of: interest-bearing assets; other earning assets (investments and S; F R (CP) INA
Kumbhakar securities); fee-based income
(2017)

(Continued)
(Continued)

86 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
224 Batir et al. TE; AE; Total loans (Sum of long-term and short-term loans); Off-balance- S; F NOM INA
(2017) CE sheet items (Sum of guarantees, commitments and financial derivative
instruments)
225 Chen et al. TE Amount of: Total Assets; Fixed Assets; Gross Loans; Total Securities; S; F NOM PRA
(2018) Total Customer Deposits; Pre-tax Profit; Net Interest Income; Non-
Interest Operating Income
226 Feng et al. TE; RTS; Amount of: Consumer loans; Securities (non-loan financial assets, i.e., S; F R (CP) INA
(2017) TFP all financial assets minus the [sum of all loans, securities, and equity]);
non-consumer loans (industrial, commercial, and real estate loans);
off-balance-sheet items
227 Wanke et al. TE; FZE Ratios of equity to: assets, net loans, short-term funding, liabilities; ratio Ratio of F NOM PRA
(2018) of cost to income; ratio net loans to total assets; net interest margin; to S
return on average assets; return on average equity; recurring earning
power; interbank ratio; loans; total earning assets; total assets; deposits
and short-term funding; other incomes (non-interest bearing); net
interest revenue; other operating income; profit before tax; net income
228 Babu and Kumar TE Amount of operating profit F NOM PRA
(2018)
229 Khan et al. TE Total Loans (all the net loans and advances to the customers – net of S; F NOM INA
(2018) provision for non-performing loans); Interest/Profit (net interest
income for conventional banks and profit/markup for Islamic banks);
Investments (investments in government bonds, debentures, treasury
securities, and shares of other entities; Other Incomes (non-markup or
non-interest income)
230 Robin et al. PA Performance indicators: Net Interest Margin (Ratio of net interest income Ratio of F R (CP) INA
(2018) to total assets); ROA (Ratio of net income to total assets); ROE (Ratio to S
of net income to total equities); Yield (Yield on earning assets)
231 Wu et al. (2018) PE Amount of: Bad loans (undesirable output); Total loans (desirable output) S NOM PRA
232 Anwar (2019) CE Model 1: Total finance, Securities and investment, Other income; Model S; F NOM INA
2: small business finance, Other finance, Securities and investment,
Other income; Model 3: Other finance, Securities and investment,
Other income
233 Glass and TFP Amount of: loans; securities; net interest income S; F R (CP) INA
Kenjegalieva
(2019)
234 Dincer et al. SE Return on Equity (Net Profit (Losses)/Total Shareholders’ Equity); Ratio of F NOM INA
(2019) (Return on Assets (Net Profit or Losses/Total Assets); Asset Quality to S
(Total Loans and Receivables/Total Assets); Income Structure – 1
(Interest Income/Total Assets); Income Structure – 2 (Non-Interest
Income (Net)/Total Assets)
235 Nartey et al. MTFP Amount of: loans and advances (total loans and advances offered to S NOM INA
(2019) customers); other earning assets (securities investments)
236 Otero et al. CE Amount of: Loans; Off-balance-sheet activities S; F NOM INA
(2020)

Review of evidence on banking output measurement 87

237 Yu et al. (2019) TE Stage I: Amount of Deposits; Stage II: Loans, Investments, Non-interest S; F NOM PRA;
income, Undesired output (NPAs) INA
238 Al-Khasawneh CE; SE; Amount of – net loans (net of NPAs); other earning assets (loans to S NOM INA
et al. (2020) PTE; special sectors and interbank loans); investment securities (treasuries
AE; and other securities)
MTFP
239 Colesnic et al. TE Amount of: Earning assets; Net Loans; Impaired loans S NOM INA
(2020)
240 Fukuyama et al. TE; CE Good outputs – Amount of trading securities, Amount of net loans (net S NOM INA
(2020) of NPAs); ‘bad’ output – Amount of non-performing loans (‘bad’
output)
241 Huang et al. LMS Amount of total assets S R (CP) NA
(2020)
242 Khan et al. EFFS Amount of – Total loans (both short-term and long-term loans); Other S; F NOM INA
(2020) earning assets (loans to special sectors – directed and specialized loans,
interbank funds sold, and investment securities – treasury and other
securities); Off-balance-sheet items (Guarantees and warranties –
letters of guarantee, bank acceptance, letters of credit, guaranteed pre-
financing, endorsements and others, commitments, foreign exchange
and interest rate transactions, as well as other off-balance-sheet activities)

(Continued)
(Continued)

88 Survey of evidences
No. Citation Issue Output Specification F or S N or R Approach
243 Mansour and TE; AE; Amount of: total customer deposits; impaired loans S NOM PRA
Moussawi CE;
(2020) MTFP
244 Phan et al. PA Performance indicators: Net Interest Margin (Ratio of net interest income Ratio of F NOM INA
(2020) to total assets); ROA (Ratio of net income to total assets); ROE (Ratio to S
of net income to total equities); Yield (Yield on earning assets)
245 Boda and Piklová Output Major output variables across specifications: interest income, other S; F R (CP) PRA;
(2021) earning assets, total other operating expense, non-interest expense, INA
total loans, total deposits
246 Chen et al. TFP Amount of: investments (equity, securities, other financial assets); total S NOM INA
(2022) loans
247 Darb and MTFP Amount of: Murabaha loan, investment loan, Al-Hassan loan and S NOM INA
Mohammed Al-Musharaka loan
(2021)
248 Le and Ngo PRF Profitability measures: Net Interest Margin; ROA; ROE; Implicitly F; Ratio of NOM INA;
(2020) includes interests earned on loans plus service charges and fee from F to S PRA
depositors)
249 Ohlson et al. MDS Amount of: insured deposits; uninsured deposits S R (CP) PRA
(2021)
250 Olson and Zoubi CE; PE Amount of: net loans (net of NPA provisions); securities and other S NOM INA
(2011) earning assets
251 Owusu and PRF Profitability measures: Net Income (net operating income plus securities F NOM INA;
Alhassan gains or losses and extraordinary credits or charges less income taxes); PRA
(2021) Net Interest Income (the difference between interest income and
interest expense); Implicitly includes interests earned on loans plus
service charges and fee from depositors
252 Shair et al. (2021) TE; MTFP Amount of: loans; interest income S; F NOM INA
253 Syed (2021) NPAs Amount of net income; amount of assets; ROA S; F; Ratio NOM INA
of F to S
254 Aksoy et al. TE Amount of: Interest income (from credits, banks and securities); Non- F NOM INA;
(2022) interest income (Fees and service charges) PRA
255 Bansal et al. TFP Stage 1: Amount of deposits; Stage 2: Unused assets of the previous S; F R (CP) PRA;
(2022) period (total assets minus the sum of required reserves, fixed capital, INA
loans, and securities investments), net profits; Stage 3: Net-interest
income, Non-interest income
256 Milenković et al. TE Amount of: loans; investments Ss NOM INA
(2022)
257 Sang (2022) TE Amount of: total loans; interest income; non-interest income (including S; F NOM INA
net income from services, net income from trading activities of
securities trading, investment and net income from other activities)
258 Santosa (2022) TE Amount of: total loans; liquid assets S NOM INA
259 Zaman et al. MTFP Amount of: performing loans (gross loans – NPAs); Investments S; F R (CP) INA
(2022) (government and other approved securities both in India and abroad

Review of evidence on banking output measurement 89

by the banks); Fee-based income (income from exchange transactions,
commission, securities, and underwriting)
260 Shah et al. TE; TGR Amount of: net interest income; non-interest income F NOM INA
(2022)
Notes: 1. LP – Labour Productivity, BO – Bank Output, OA – Output Aggregation, PRF – Profitability, TC – Technical Change, TE – Technical Efficiency, PE – Produc-
tive Efficiency, Production Efficiency, CE – Cost Efficiency, AE – Allocative Efficiency, PTC – Pure Technical Change, EOS – Economies of Scale, EOSCO – Economies of
Scope, TCH – Technological Change, RSE – Ray Scale Economies, TSB – Technological Scale Bias, OE – Overall Efficiency, OTE – Overall technical Efficiency, PTE –
Pure Technical Efficiency, SE – Scale Efficiency, CMP – Competition, EPSE – Expansion Path Scale Economies, EPSA – Expansion Path Subadditivity, BOP – Bank Output
Pricing, EFFS – Efficiencies, ISTE – Input-specific Technical Efficiency, ME – Market Efficiency, CEL – Cost Elasticity, XEF – X-Efficiency, PEF – Production Efficiency,
INE – Intermediation Efficiency, MPR – Market Power, PE – Profit Efficiency, APE – Alternate Profit Efficiency, M&A – Mergers and Acquisitions, QE – Quality Effi-
ciency, IO – Insurance Output, SZE – Size Efficiency, PA – Performance Analysis, REV – Review of Evidences, CFP – Cost Function Parameters, CGR – Cost Gap Ratio,
MFCE – Metafrontier CE, TGR – Technology Gap Ratio, FZE – Fuzzy Efficiency, LMS – Loan Market Switching, TFP – Total Factor Productivity or Malmquist TFP or
Luenberger TFP or Malmquist-Luenberger TFP, MDS – Market Discipline, PRA – Production Approach, INA – Intermediation Approach, VAA – Value Added Approach,
NIA – National Income Approach, UCA – User Cost Approach, OPA – Operating Profit Approach, F – Flow variables, S – Stock variables, N – NOM, R – R, R (CP) – R
(Constant Price), R (VL) – R (Volume), S (PR) – S (Price), F (PR) – Flow (Price); 2. A further detailed review sheet with more analytical dimensions on each of the studies
above and on a larger sample of studies can be availed from the author on request.
Source: Survey of evidence by the author.
90 Survey of evidences
Notes
1 Once the appropriate approach to output specification is determined, as discussed in
section 3.3, the issue becomes a measurement problem. The analyst must choose the
best possible proxy for the underlying output concept as prescribed by the chosen speci-
fication approach. This seems to be how literature has conceptualised the bank output
measurement problem.
2 Agu (1988) reviews the significant output measures used under the balance sheet versus
income statements approaches in the literature on bank performance spanning from the
1950s to the late 1970s.
3 Bankers, economists and national income accountants tend to view the banking activi-
ties from different angles. See Stauffer (2004) for an overview of these perspectives.
4 The University of Adelaide, Australia, made this study available and was kind enough
to send the soft copy of this work on request. Also, see Tripplet and Bosworth (2004a,
2004b), for more details on this issue.
5 The author studied the distribution of the cost efficiency estimates using the Kernel
density method. The author suggested this non-parametric distribution to avoid intro-
ducing subjectivity by having to choose a specific functional form, as is the case with
parametric distributions. The author asserts that “the need to know the functional form
makes the parametric approach barely useful as a descriptive tool” (p. 206).
6 The approaches discussed here define not only the output but inputs as well. Hence, these
approaches can be better referred to as production function specification approaches in
banking rather than output specification approaches. However, as discussed in Chap-
ter 10, there is a reasonable consensus on the significant inputs employed in a banking
production function. Hence, the major differences among these approaches result from
incongruence in output specification. That is the approach followed in this work.
7 Another novel approach on this account was the one Leland and Pyle (1977) sug-
gested, which approached bank output measurement under a risk-based approach. They
incorporated dimensions such as risk management, adverse selection and information
asymmetry into their measurement of banking output. After the Great Recession, this
approach gained increased attention, particularly in advanced economies. Another nota-
ble study in more recent times on broadly similar lines is Saeed et al. (2020).
8 The discussion here is brief due to space constraints. More works on the approaches
highlighted here can be located in the appendix to this chapter and the additional mate-
rial provided in the online resources.
9 Several approaches use these same variables but are different from the production
approach. The basic idea is that both the inputs and outputs should be consistent with
what a particular approach suggests. Different approaches may use the same output vec-
tor. These dimensions will emerge clearly in further discussion.
10 This statement is true to the extent that value measures are used in the production
approach. Data availability on volume measures is challenging in both advanced and
emerging economies.
11 This statement can be confirmed only under the assumption that the intermediation
component of total bank output is falling or lower than the direct component. How-
ever, for aggregate sectoral analysis, if one wishes to focus on the performance of the
banking sector in terms of its direct output, the production approach can serve as a help-
ful tool. In Chapter 8, section 8.3, sectoral analysis of returns to scale for the banking
and insurance industries of the BRIICS countries is performed wherein both the pro-
duction and intermediation approaches are employed. The returns to scale coefficient
does not diverge much under both approaches.
12 The fundamental essence of the intermediation approach in the banking and insurance
literature is that these intermediaries have a comparative advantage in minimization of
the transaction, search and monitoring costs relative to non-banking and non-insurance
firms. However, with the increasing digitalization of both these industries, transactions
and search costs may be brought down considerably in the coming years. Does that
 Review of evidence on banking output measurement 91
mean that their intermediation function will become redundant? See Scholtens and
Wensveen (2003) for a possible answer to this concern.
13 Diamond (1984) warned at a very early stage that the emergence of non-banking finan-
cial companies (NBFC) and the increasing problem of NPAs indicate the falling compet-
itive advantages banks enjoy in resolving the issues of monitoring costs and information
asymmetries. This warning still holds, and the implications are examined in section 3.5.
14 See Radecki (1999) for a criticism of the pure intermediation approach.
15 For a compact introduction to intermediation theory, see Inklaar and Wang (2011,
2012). Interestingly, the authors suggest that the flow of funds between depositors
and borrowers via the banks should not be considered a measure of output under this
approach. Broadly, similar line of analysis is adopted by Colangelo and Inklaar (2010).
16 This also implies that the operating cost shares can be used to aggregate various outputs
into a single aggregate/composite index.
17 A compact introduction to user costs concerning banking can be found in Hancock
(1985).
18 This is required because net contribution can be estimated only after adjusting for all
economic costs, including accounting and opportunity costs.
19 Such an exercise is undertaken across many advanced and emerging economies in the
form of FISIM estimation under the SNA framework. However, FISIM estimates are
estimated and available at the sectoral level – bank-level and branch-level information.
20 See Wang and Basu (2005) for more details on this dimension. They also provide a way
out of this limitation of the user-cost approach. Their work is vital because the imputa-
tion of the implicit output of the commercial banking sector in the SNA is based mainly
on the user cost framework.
21 Further details on these studies can be found in the appendix to this chapter and the
extended review sheet, which may be requested from the author directly.
22 Weights are essential elements of an index number. In his works, Fisher established that
rather than the averaging procedure, the choice of weights and other considerations are
more serious challenges in the construction of index numbers. While reviewing the
classic work of Fisher, namely ‘the making of index numbers’, Synder (1923, p. 421)
comments that “the choice of weights, of material, of the accuracy of the data, of the
number of quotations, is of far greater importance, and it is these and not the method
of averaging which makes practically all the remarkable differences”.
23 This can be undertaken using the so-called acid tests laid down by Fisher (1922).
24 Further detailed notes on each of these output measures for both the banking and
insurance activities may be supplied in the online resources or may be availed from the
author upon request. Due to space constraints, only the most essential aspects have been
presented here.
25 At this juncture, readers might be interested in knowing the specific studies that employ
the measures highlighted here. A thematic bibliography and a list of studies for the vari-
ous output measures discussed throughout this work can be obtained from the author
upon request. This was inevitable due to space restrictions. The readers may kindly
forgive the author for this presentation arrangement.
26 See Boda and Piklová (2021) for a brief note on this matter. Also see Setiwan et al.
(2017).
27 A common criticism against using this approach to output measurement has been the
so-called divisibility constraint. “Deposits are typically divisible, liquid and riskless,
whilst loans are indivisible, illiquid and risky” (Dong, 2010, p. 131). Thus, combining
them can distort the true underlying output, and further allocation of the output to
consumption sectors, different lines of business, etc., may not be feasible.
28 Needless to say that this measure of output is inconsistent with the pure intermediation
and asset approach to output specification.
29 See Hartwick (1996) for an elaboration on this aspect.
30 This implies that deposits and loans are treated as homogenous services, which is theo-
retically incorrect. If they are to be weighted, the correct aggregation function must be
92 Survey of evidences
defined, and valid weighting factors must be ascertained. This is a debate that has not
been settled till date. The broad consensus is to treat them separately as far as possible.
31 Sealey and Lindley (1977) proposed this issue for the first time for a rigorous analysis
of deposits as inputs. Similarly, Humphrey and Pulley (1997) defend using deposits as
inputs. However, Tripplet and Bosworth (2004b) outrightly reject the notion of deposits
being an input.
32 For treatment of deposits as quasi-fixed inputs, please see Flannery (1982). “A quasi-
fixed factor of production is one whose cost to the firm in any production period is
partly fixed (sunk) and partly variable” (Flannery, 1982, p. 530). The author rejects
the notion that deposits are variable inputs and contends that they must be included
as quasi-fixed inputs to capture the true production process of banks better. This con-
ception is also based on the premise that demand contraction or external shocks do
not allow the banks to freely change the amount of deposit received, especially if the
negative demand or external shock is temporary. Banks can’t freely change the deposit
amount as per extant economic conditions, which requires their treatment as a quasi-
fixed input rather than variable input.
33 Deposits are treated as output due to various reasons. One of them is that the explicit
revenues on deposits are small, and thus there must be implicit earnings for banks which
require deposits to be treated as output (Berger & Humphrey, 1992). Hughes and Mester
(1993) have proposed a basic test to decide whether deposits are input or output in a
given context. They estimate a variable cost function with deposits as an explanatory
variable. If the estimated sign is positive, then deposits are output or else they may be
treated as an input.
34 “In the user cost framework the sign of the user cost provides the answer: a negative sign
indicates an output status and a positive sign indicates input status” (Fixler & Zieschang,
1999, p. 548). They conclude that as per the evidence, deposits must be treated as output.
35 “It is a technological question that can be answered by testing whether the data are
consistent with the different technological roles of outputs and inputs” (Hughes et al.,
2000, p. 6).
36 The stock amount of deposits is adjusted for the effective interest rate for inclusion in
the FISIM.
37 More details on this matter can be found in the online resources or can be obtained from the
author on request. Also see Rakshit and Bardhan (2022) for some dimensions on this aspect.
38 See the contribution by Prof. Barendra Bhoi in Chapter 7.
39 Detailed Notes on the estimation procedures of FISIM output and other supplementary
notes are provided in the online resources. Only the most pressing issues in the literature
are elaborated on in this section.
40 See Akritidis (2017) and Triplett and Bosworth (2004a, 2004b) for the history and evo-
lution of the FISIM framework in the national income accounting.
41 Several authors have criticized this implicit approach to measuring the implicit output.
Instead, they focus on directly estimating the real flow of services. See Wang and Basu
(2005) for an illustration of such an approach.
42 For an overview of alternative approaches to estimating the reference rate, see Davtyan
(2018).
43 In the most fundamental terms, the equation for estimating FISIM output is (RL – RR)
YL + (RR – RD)YD, where RL is the chosen rate of interest on loans, RD is the chosen
rate of interest on deposits, YL is the amount of loan, YD is the amount of deposit and
RR is the chosen reference rate.
44 See Daisuke et al. (2011a, 2011b), who construct a risk-adjusted measure of FISIM
output for individual banks and examine FISIM output productivity using data on 368
to 1251 bank-firm groups of Japan across 1976–2005.
45 The authors provide an extensive discussion on the various aspects of bank output
regarding the requirements and principles of the national income accounts, particularly
the SNA, 1993 manual released shortly before their study.
 Review of evidence on banking output measurement 93
46 The reference rate was initially conceptualised as the minimum rate of return that
accounts for a given risk exposure on an asset, but the SNA adopted a risk-free inter-
pretation (Fixler et al., 2003).
47 Disaggregated reference rates can be used to represent the benchmark rate for each bank
with each FISIM-consuming sector (e.g. reference rate between banks and the house-
hold sector, banks and manufacturing business, banks and agriculture businesses, banks
and services, banks and other banks, banks and Government, etc.). Such efforts are
possible only for advanced economies. EMEs are still a long way from undertaking such
exercises. For a brief overview on this problem in the context of emerging economies,
see Das and Jangili (2017).
48 An emerging database on sectoral and industry-level output measures is the KLEMS
project. However, it has not been developed enough to be used on a larger scale. The
author could not find any high-quality study on the present theme that employed the
KLEMS estimates of value-added and gross output. Moreover, the KLEMS database
provides estimates only for the financial services sector and not for the banking industry.
49 For a detailed discussion on double deflation methodology, especially concerning
national income accounts, see Vanoli (2005).
50 While this may be possible for labour variables, the controversy on the correct price of
capital input makes such an exercise subject to many disagreements. Regulatory data on
banking statistics in the emerging economies generally do not get deeper into capital
input pricing. Moreover, the pricing of capital input has been a complicated matter and
has been raged by many debates. The so-called “fundamental problem of accounting”
(Diewert, 2001, p. 3) persists. The analyst has to estimate the same using disparate data
sources, which again can be a possible source of measurement error.
51 A unique approach for single deflation of banking inputs and outputs has been suggested
by Diewert et al. (2012). The price index of non-monetary assets is proposed as defla-
tors for the liability components in the banking sector’s balance sheet constructed by the
authors. Even earlier attempts on broadly similar lines may be found in Diewert et al.
(2004, 2009).

References
Abdul-Majid, M., Falahaty, M., & Jusoh, M. (2017). Performance of Islamic and conven-
tional banks: A meta-frontier approach. Research in International Business and Finance, 42,
1327–1335. https://doi.org/10.1016/j.ribaf.2017.07.069
Adams, R. M., Röller, L.-H., & Sickles, R. C. (2002). Market power in outputs and inputs: An
empirical application to banking. Social Science Research Network (SSRN Scholarly Paper
No. 355100). https://doi.org/10.2139/ssrn.355100
Agu, C. C. (1988). On the performance analysis: The definition and measurement of bank
output in Nigeria. Savings and Development, 12(2), 107–126.
Agu, C. C. (2004). Efficiency of commercial banking in the Gambia. African Review of Money
Finance and Banking, 31–50.
Akritidis, L. (2017). Financial intermediation services indirectly measured (FISIM) in the U.K.
Revisited. Office for National Statistics. www.ons.gov.uk/economy/grossdomesticproduct
gdp/articles/financialintermediationservicesindirectlymeasuredfisimintheukrevisited/
2017-04-24
Aksoy, E. E. A., Dirik, C., & Göker, İ. E. K. (2022). Opening the black-box of bank effi-
ciency in Turkey with two-stage data envelopment analysis: A study on capital adequacy
ratio. Ege Academic Review, 22(1), 75–91. https://doi.org/10.21121/eab.1064816
Alam, I. M. S. (2001). A nonparametric approach for assessing productivity dynamics of
large U.S. Banks. Journal of Money, Credit and Banking, 33(1), 121–139. https://doi.
org/10.2307/2673875
94 Survey of evidences
Al-Khasawneh, J. A., Essaddam, N., & Hussain, T. (2020). Total productivity and cost effi-
ciency dynamics of U.S. Merging banks: A non-parametric bootstrapped analysis of the
fifth merger wave. The Quarterly Review of Economics and Finance, 78, 199–211. https://
doi.org/10.1016/j.qref.2020.02.002
Allen, F., Jackowicz, K., Kowalewski, O., & Kozłowski, Ł. (2015). Bank lending, crises,
and changing ownership structure in central and Eastern European countries. Journal of
Corporate Finance. https://doi.org/10.1016/j.jcorpfin.2015.05.001
Allen, J., & Liu, Y. (2007). Efficiency and economies of scale of large Canadian banks. The
Canadian Journal of Economics, 40(1), 225–244.
Altman, E. I., Gande, A., & Saunders, A. (2010). Bank debt versus bond debt: Evidence
from secondary market prices. Journal of Money, Credit and Banking, 42(4), 755–767.
Altunbaş, Y., Gardener, E. P. M., Molyneux, P., & Moore, B. (2001). Efficiency in Euro-
pean banking. European Economic Review, 45(10), 1931–1955. https://doi.org/10.1016/
S0014-2921(00)00091-X
Altunbas, Y., Liu, M.-H., Molyneux, P., & Seth, R. (2000). Efficiency and risk in Japanese
banking. Journal of Banking & Finance, 24(10), 1605–1628.
Aly, H. Y., Grabowski, R., Pasurka, C., & Rangan, N. (1990). Technical, scale, and alloca-
tive efficiencies in U.S. Banking: An empirical investigation. The Review of Economics and
Statistics, 72(2), 211–218. https://doi.org/10.2307/2109710
Andries, A. M. (2011). The determinants of bank efficiency and productivity growth in the
central and Eastern European banking systems. Eastern European Economics, 49(6), 38–59.
Angadi, V. B., & Devaraj, V. J. (1983). Productivity and profitability of banks in India. Eco-
nomic and Political Weekly, 18(48), M160–M170.
Anthanassopoulos, A. D. (1998). Nonparametric frontier models for assessing the market
and cost efficiency of large-scale bank branch networks. Journal of Money, Credit and Bank-
ing, 30(2), 172–192. https://doi.org/10.2307/2601208
Anwar, M. (2019). Cost efficiency performance of Indonesian banks over the recovery
period: A stochastic frontier analysis. The Social Science Journal, 56(3), 377–389. https://
doi.org/10.1016/j.soscij.2018.08.002
Ariff, M., & Can, L. (2008). Cost and profit efficiency of Chinese banks: A non-parametric
analysis. China Economic Review, 19, 260–273. https://doi.org/10.1016/j.chieco.2007.
04.001
Asaftei, G., & Kumbhakar, S. C. (2008). Regulation and efficiency in transition: The
case of Romanian banks. Journal of Regulatory Economics, 33(3), 253–282. https://doi.
org/10.1007/s11149-007-9041-0
Athanasoglou, P. P., Georgiou, E. A., & Staikouras, C. C. (2009). Assessing output and pro-
ductivity growth in the banking industry. The Quarterly Review of Economics and Finance,
49(4), 1317–1340. https://doi.org/10.1016/j.qref.2009.08.003
Athanassopoulos, A. D., & Giokas, D. (2000). The use of data envelopment analysis in bank-
ing institutions: Evidence from the commercial bank of Greece. Interfaces, 30(2), 81–95.
Ayadi, O. F., Adebayo, A. O., & Omolehinwa, E. (1998). Bank performance measurement in
a developing economy: An application of data envelopment analysis. Managerial Finance,
24(7), 5–16. https://doi.org/10.1108/03074359810765589
Babu, M. R., & Kumar, A. M. (2018). Evaluating the nationalization & privatization effect:
A case of Indian banking industry. Banks and Bank Systems, 13(1), 11–21. https://doi.
org/10.21511/bbs.13(1).2018.02
Badunenko, O., & Kumbhakar, S. C. (2017). Economies of scale, technical change and per-
sistent and time-varying cost efficiency in Indian banking: Do ownership, regulation and
heterogeneity matter? European Journal of Operational Research, 260(2), 789–803. https://
doi.org/10.1016/j.ejor.2017.01.025
 Review of evidence on banking output measurement 95
Banerjee, B. (2012). Banking sector efficiency in new E.U. Member states: A survey of
cross-country evidence. Eastern European Economics, 50(6), 81–115.
Banker, R. D., Chang, H., & Lee, S.-Y. (2010). Differential impact of Korean banking
system reforms on bank productivity. Journal of Banking & Finance, 34(7), 1450–1460.
https://doi.org/10.1016/j.jbankfin.2010.02.023
Bansal, P., Kumar, S., Mehra, A., & Gulati, R. (2022). Developing two dynamic Malmquist-
Luenberger productivity indices: An illustrated application for assessing productiv-
ity performance of Indian banks. Omega, 107, 102538. https://doi.org/10.1016/j.
omega.2021.102538
Barman, R. B., & Samanta, G. P. (2007). Measuring banking intermediation services: Issues
and challenges for India. Economic and Political Weekly, 42(37), 3754–3763.
Barnett, W. A. (1978). The user cost of money. Economics Letters, 1(2), 145–149. https://doi.
org/10.1016/0165-1765(78)90051-4
Batir, T. E., Volkman, D. A., & Gungor, B. (2017). Determinants of bank efficiency in Tur-
key: Participation banks versus conventional banks. Borsa Istanbul Review, 17(2), 86–96.
https://doi.org/10.1016/j.bir.2017.02.003
Batra, A. (1996). Bank profitability with a hybrid profit function – the Indian case. Indian
Economic Review, 31(2), 223–234.
Beccalli, E., Anolli, M., & Borello, G. (2015). Are European banks too big? Evidence on
economies of scale. Journal of Banking & Finance, 58, 232–246. https://doi.org/10.1016/j.
jbankfin.2015.04.014
Beccalli, E., Casu, B., & Girardone, C. (2006). Efficiency and stock performance in Euro-
pean banking. Journal of Business Finance & Accounting, 33(1–2), 245–262. https://doi.
org/10.1111/j.1468-5957.2006.01362.x
Benston, G. J. (1965). Branch banking and economies of scale. The Journal of Finance, 20(2),
312–331. https://doi.org/10.2307/2977238
Benston, G. J. (1972). Economies of scale of financial institutions. Journal of Money, Credit
and Banking, 4(2), 312–341. https://doi.org/10.2307/1991041
Benston, G. J., Hanweck, G. A., & Humphrey, D. B. (1982). Scale economies in banking:
A restructuring and reassessment. Journal of Money, Credit and Banking, 14(4), 435–456.
https://doi.org/10.2307/1991654
Berg, S. A., Førsund, F. R., & Jansen, E. S. (1991). Technical efficiency of Norwegian banks:
The non-parametric approach to efficiency measurement. Journal of Productivity Analysis,
2(2), 127–142.
Berg, S. A., Førsund, F. R., & Jansen, E. S. (1992). Malmquist indices of productivity growth
during the deregulation of Norwegian banking, 1980–89. The Scandinavian Journal of Eco-
nomics, 94, S211–S228. https://doi.org/10.2307/3440261
Berger, A. N., & Bouwman, C. H. S. (2007). Bank liquidity creation. Federal Deposit Insur-
ance Corporation, U.S. (Conference Paper).
Berger, A. N., & DeYoung, R. (1997). Problem loans and cost efficiency in commercial
banks – sciencedirect. Journal of Banking & Finance, 21(6), 849–870.
Berger, A. N., & Hannan, T. H. (1998). The efficiency cost of market power in the bank-
ing industry: A test of the ‘quiet life’ and related hypotheses. The Review of Economics and
Statistics, 80(3), 454–465.
Berger, A. N., Hanweck, G. A., & Humphrey, D. B. (1987). Competitive viability in bank-
ing: Scale, scope, and product mix economies. Journal of Monetary Economics, 20, 501–520.
Berger, A. N., & Humphrey, D. B. (1992). Measurement and efficiency issues in commercial
banking. In Output Measurement in the Service Sectors (pp. 245–300). University of Chi-
cago Press. www.nber.org/books-and-chapters/output-measurement-service-sectors/
measurement-and-efficiency-issues-commercial-banking
96 Survey of evidences
Berger, A. N., & Humphrey, D. B. (1997). Efficiency of financial institutions: International
survey and directions for future research. European Journal of Operational Research, 98(2),
175–212.
Berger, A. N., & Mester, L. J. (1997). Inside the black box: What explains differences in the
efficiencies of financial institutions? Journal of Banking & Finance, 21(7), 895–947. https://
doi.org/10.1016/S0378-4266(97)00010-1
Bhattacharya, K. (2003). How good is the bankscope database? A cross-validation exercise with cor-
rection factors for market concentration measures. Bank for International Settlements (Working
Paper No. 133). https://doi.org/10.2139/ssrn.901147
Bhattacharyya, A., Lovell, C., & Sahay, P. (1997). The impact of liberalization on the pro-
ductive efficiency of Indian commercial banks. European Journal of Operational Research,
98(2), 332–345.
Bhuyan, P. (2016). Measurement of central bank output – methodological issues for India. Reserve
Bank of India (Working Paper No. 06/2016).
Boda, M., & Piklová, Z. (2021). Impact of an input-output specification on efficiency scores
in data envelopment analysis: A banking case study. RAIRO – Operations Research, 55,
S1551–S1583. https://doi.org/10.1051/ro/2020040
Boer, S. D. (1999). A volume index for the output of the Dutch banking industry based on quantity
indicators. Statistics Netherland.
Boot, A. W. A., & Thakor, A. V. (1993). Self-interested bank regulation. The American Eco-
nomic Review, 83(2), 206–212.
Bos, J. W. B., & Kolari, J. W. (2005). Large bank efficiency in Europe and the United States:
Are there economic motivations for geographic expansion in financial services? The Jour-
nal of Business, 78(4), 1555–1592. https://doi.org/10.1086/430869
Bossone, B., & Lee, J.-K. (2004). In finance, size matters: The ‘systemic scale economies’
hypothesis. IMF Staff Papers, 51(1), 19–46.
Boucinha, M., Ribeiro, N., & Weyman-Jones, T. (2012). An assessment of Portuguese
banks’ efficiency and productivity towards Euro area participation. Journal of Productivity
Analysis, 39(2), 177–190. https://doi.org/10.1007/s11123-012-0310-2
Boumans, M. (2007). Measurement in economics: A handbook. Emerald Group Publishing
Limited.
Brand, H., & Duke, J. (1982). Productivity in commercial banking: Computers spur the
advance. Monthly Labor Review, 105(12), 19–27.
Butter, F. A. G. den. (2007). National accounts and indicators, Chapter 8. In M. Boumans
(Ed.), Measurement in economics: A handbook. Elsevier.
Camanho, A. S., & Dyson, R. G. (1999). Efficiency, size, benchmarks and targets for bank
branches: An application of data envelopment analysis. Journal of the Operational Research
Society, 50, 903–915. https://doi.org/10.1057/palgrave.jors.2600792
Camanho, A. S., & Dyson, R. G. (2005). Cost efficiency, production and value-added
models in the analysis of bank branch performance. The Journal of the Operational Research
Society, 56(5), 483–494.
Casolaro, L., & Gobbi, G. (2007). information technology and productivity changes in the
banking industry. Economic Notes, 36(1), 43–76.
Casu, B., Girardone, C., & Molyneux, P. (2004). Productivity change in European banking:
A comparison of parametric and non-parametric approaches. Journal of Banking & Finance,
28(10), 2521–2540. https://doi.org/10.1016/j.jbankfin.2003.10.014
Chaffai, M. E. (1997). Estimating input-specific technical inefficiency: The case of the Tuni-
sian banking industry. European Journal of Operational Research, 98(2), 314–331. https://
doi.org/10.1016/S0377-2217(96)00350-5
 Review of evidence on banking output measurement 97
Chaudhuri, S. (2002). Some issues of growth and profitability in Indian public sector banks.
Economic and Political Weekly, 37(22), 2155–2162.
Chen, G., Kim, K. A., Nofsinger, J. R., & Rui, O. M. (2007). Trading performance, dispo-
sition effect, overconfidence, representativeness bias, and experience of emerging market
investors. Journal of Behavioral Decision Making, 20(4), 425–451. https://doi.org/10.1002/
bdm.561
Chen, X., Zhang, X., Wang, Y., & Jia, H. (2022). The risk-adjusted profit productivity
change in Chinese banks: A comparative analysis of the different type banks. Managerial
and Decision Economics, 43(2), 523–532. https://doi.org/10.1002/mde.3399
Chen, Z., Matousek, R., & Wanke, P. (2018). Chinese bank efficiency during the global
financial crisis: A combined approach using satisficing DEA and support vector machines.
The North American Journal of Economics and Finance, 43, 71–86. https://doi.org/10.1016/j.
najef.2017.10.003
Chronopoulos, D., Girardone, C., & Nankervis, J. (2010). Post-merger bank effi-
ciency and stock market reaction: The case of the U.S. versus Europe. In New issues
in financial institutions management (pp. 122–135). Palgrave Macmillan. https://doi.
org/10.1057/9780230299153_8
Clark, J. A. (1984). Estimation of economies of scale in banking using a general-
ized functional form. Journal of Money, Credit and Banking, 16(1), 53–68. https://doi.
org/10.2307/1992648
Clark, J. A. (1996). Economic cost, scale efficiency, and competitive viability in banking.
Journal of Money, Credit and Banking, 28(3), 342–364. https://doi.org/10.2307/2077979
Clark, J. A., & Siems, T. F. (2002). X-efficiency in banking: Looking beyond the balance
sheet. Journal of Money, Credit and Banking, 34(4), 987–1013.
Clark, J. A., & Speaker, P. J. (1994). Economies of scale and scope in banking: Evidence
from a generalized Translog cost function. Quarterly Journal of Business and Economics,
33(2), 3–25.
Colangelo, A., & Inklaar, R. (2010). Banking sector output measurement in the Euro
area: A modified approach. SSRN Electronic Journal, 1–51. https://doi.org/10.2139/
ssrn.1610200
Colesnic, O., Kounetas, K., & Michael, P. (2020). Estimating risk efficiency in middle east
banks before and after the crisis: A metafrontier framework. Global Finance Journal, 46,
100484. https://doi.org/10.1016/j.gfj.2019.100484
Colwell, R. J., & Davis, E. P. (1992). Output and productivity in banking. The Scandinavian
Journal of Economics, 94, S111–S129. https://doi.org/10.2307/3440252
Coyne, T. J. (1973). Commercial bank profitability by function. Financial Management, 2(1),
64. https://doi.org/10.2307/3665102
Cuesta, R. A., & Zofío, J. L. (2005). Hyperbolic efficiency and parametric distance functions:
With application to Spanish savings banks. Journal of Productivity Analysis, 24(1), 31–48.
Curi, C., Guarda, P., Lozano-Vivas, A., & Zelenyuk, V. (2013). Is foreign-bank efficiency in
financial centers driven by home-country characteristics? Journal of Productivity Analysis,
40, 367–385. https://doi.org/10.2139/ssrn.1833327
Daisuke, M., Tomohiko, I., & Keishi, S. (2011a). Bank efficiency and client firms’ productivity.
Economic and Social Research Institute (Discussion Paper No. 273).
Daisuke, M., Tomohiko, I., & Shoji, K. (2011b). Productivity of banks and its impact on the
capital investments of client firms. RIETI (Discussion Paper Series 11-E-016).
Darb, J. A.-K., & Mohammed, A. A. (2021). Measuring and analyzing the productivity of
Islamic banks in Iraq according to the Malmquist productivity index. Review of Interna-
tional Geographic Education, 11(7), 1646–1658.
98 Survey of evidences
Das, A. (2002). Risk and productivity change of public sector banks. Economic and Political
Weekly, 37(5), 437–440 + 443–445 + 447–448.
Das, A., & Jangili, R. (2017). Financial intermediation services indirectly measured (FISIM):
The role of reference rate. Statistical Journal of the IAOS, 33, 515–524. https://doi.
org/10.3233/SJI-160280
Das, A., Nag, A., & Ray, S. (2005). Liberalisation, ownership and efficiency in Indian bank-
ing: A nonparametric analysis. Economic and Political Weekly, 40, 1190–1197. https://doi.
org/10.2307/4416367
Das, P. K., & Maiti, P. (1998). Bank credit, output and bank deposits in West Bengal and
selected states. Economic and Political Weekly, 33(47/48), 3081–3088.
Davies, R., & Tracey, B. (2014). Too big to be efficient? The impact of implicit subsidies on
estimates of scale economies for banks. Journal of Money, Credit and Banking, 46(1), 219–253.
Davtyan, A. (2018). Estimation of financial intermediation services indirectly measured in Armenia’s
external accounts. Irving Fisher Committee – Central Bank of Armenia Workshop on
‘External Sector Statistics’.
De, P. K. (2004). Technical efficiency, ownership, and reforms: An econometric study of
Indian banking industry. Indian Economic Review, 39(1), 261–294.
Devaney, M., & Weber, W. (2002). Small-business lending and profit efficiency in com-
mercial banking. Journal of Financial Services Research, 22(3), 225–246. https://doi.
org/10.1023/A:1019733226515
DeYoung, R., & Hasan, I. (1998). The performance of de novo commercial banks:
A profit efficiency approach. Journal of Banking & Finance, 22(5), 565–587. https://doi.
org/10.1016/S0378-4266(98)00025-9
Diamond, D. W. (1984). Financial intermediation and delegated monitoring. The Review of
Economic Studies, 51(3), 393–414. https://doi.org/10.2307/2297430
Dick, A. A. (2006). Nationwide branching and its impact on market structure, quality, and bank
performance. The Journal of Business, 79(2), 567–592. https://doi.org/10.1086/499131
Dick, A. A. (2007). Market size, service quality, and competition in banking. Journal of
Money, Credit and Banking, 39(1), 49–81.
Dietsch, M. (1993). Economies of scale and scope in French commercial banking industry.
Journal of Productivity Analysis, 4(1/2), 35–50.
Diewert, E., Fixler, D., & Zieschang, K. (2011). The measurement of banking services in the
system of national accounts. Vancouver School of Economics (Working Paper No. 2011–23).
Diewert, E., Fixler, D., & Zieschang, K. (2012). Problems with the measurement of banking
services in a national accounting framework. In Discussion Papers (No. 2012-25; Discus-
sion Papers). School of Economics, The University of New South Wales. https://ideas.
repec.org/p/swe/wpaper/2012-25.html
Diewert, W. E. (2001). Measuring the price and quantity as capital services under alternative
assumptions. Department of Economics, The University of British Columbia (Working
Paper No. 01–24).
Diewert, W. E., Greenlees, J. S., & Hulten, C. R. (2004). Incorporating financial services in
a consumer price index. NBER Conference on Price Index Concepts and Measurement,
Vancouver, Canada. https://doi.org/10.7208/chicago/9780226148571.001.0001
Diewert, W. E., Greenlees, J. S., & Hulten, C. R. (2009). Price index concepts and meas-
urement. University of Chicago Press. www.nber.org/books-and-chapters/price-index-
concepts-and-measurement
Dincer, H., Hacioglu, U., Tatoglu, E., & Delen, D. (2019). Developing a hybrid analytics
approach to measure the efficiency of deposit banks. Journal of Business Research, 104,
131–145. https://doi.org/10.1016/j.jbusres.2019.06.035
 Review of evidence on banking output measurement 99
Dong, Y. (2010). Cost efficiency in the Chinese banking sector: A comparison of parametric and non-
parametric methodologies [Ph.D. Thesis]. Loughborough University. https://repository.lboro.
ac.uk/articles/thesis/Cost_efficiency_in_the_Chinese_banking_sector_a_comparison_
of_parametric_and_non-parametric_methodologies/9496763/1
Donovan, D. J. (1977). Consumption, leisure, and the demand for money and money substitutes
[Ph.D. Thesis]. University of British Columbia, Department of Economics.
Durkin, T. A., & Elliehausen, G. E. (1998). The cost structure of the consumer finance
industry. Journal of Financial Services Research, 13(1), 71–86. https://doi.org/10.1023/
A:1007906527814
Elyasiani, E., & Mehdian, S. M. (1992). Productive efficiency performance of minority
and nonminority-owned banks: A nonparametric approach. Journal of Banking & Finance,
16(5), 933–948. https://doi.org/10.1016/0378-4266(92)90033-V
Elyasiani, E., & Mehdian, S. M. (1990). A nonparametric approach to measurement of
efficiency and technological change: The case of large U.S. Commercial banks. Journal of
Financial Services Research, 4, 157–168.
Evanoff, D. D., & Israilevich, P. R. (1995). Scale elasticity versus scale efficiency in banking.
Southern Economic Journal, 61(4), 1036–1046. https://doi.org/10.2307/1060739
Evanoff, D. D., Israilevich, P. R., & Merris, R. C. (1989). Technical change, regulation, and
economies of scale for large commercial banks: An application of a modified version of Shephard’s
Lemma. Federal Reserve Bank of Chicago (Working Paper No. 89–11). https://ideas.
repec.org/p/fip/fedhfi/89-11.html
Farrell, M. J. (1957). The measurement of productive efficiency. Journal of the Royal Statistical
Society, 120(3), 253–290. https://doi.org/10.2307/2343100
Feng, G., Peng, B., & Zhang, X. (2017). Productivity and efficiency at bank holding com-
panies in the U.S.: A time-varying heterogeneity approach. Journal of Productivity Analysis,
48(2), 179–192.
Feng, G., & Serletis, A. (2010). Efficiency, technical change, and returns to scale in large
U.S. Banks: Panel data evidence from an output distance function satisfying theoreti-
cal regularity. Journal of Banking & Finance, 34(1), 127–138. https://doi.org/10.1016/j.
jbankfin.2009.07.009
Fiordelisi, F., Molyneux, P., & Previati, D. (2010). New issues in financial institutions man-
agement (1st ed.). Palgrave Macmillan. https://link.springer.com/book/10.1057/
9780230299153
Fisher, I. (1922). The making of index numbers. Houghton Mifflin Company.
Fixler, D. (2004). Incorporating financial services in a consumer price index. Bureau of Economic
Analysis (Working Paper).
Fixler, D., Reinsdorf, M., & Smith, G. (2003). Measuring the services of commercial banks in the
NIPAs changes in concepts and methods. OECD (OECD National Accounts Experts Meet-
ing STD/NAES(2003)20).
Fixler, D., & Zieschang, K. (1999). The productivity of the banking sector: Integrating
financial and production approaches to measuring financial service output. Canadian Jour-
nal of Economics, 32(2), 547–569.
Flannery, M. J. (1982). Retail bank deposits as quasi-fixed factors of production. The American
Economic Review, 72(3), 527–536.
Focarelli, D., & Panetta, F. (2003). Are mergers beneficial to consumers? Evidence from
the market for bank deposits. American Economic Review, 93(4), 1152–1172. https://doi.
org/10.1257/000282803769206241
Freixas, X., & Parigi, B. (1998). Contagion and efficiency in gross and net interbank payment
systems. Journal of Financial Intermediation, 7, 3–31.
100 Survey of evidences
Fukuyama, H., Matousek, R., & Tzeremes, N. G. (2020). A Nerlovian cost inefficiency
two-stage DEA model for modeling banks’ production process: Evidence from the Turk-
ish banking system. Omega, 95, 102198. https://doi.org/10.1016/j.omega.2020.102198
Geehan, R., & Allen, L. (1978). Measuring the real output and productivity of savings
and credit institutions. The Canadian Journal of Economics/Revue Canadienne d’Economique,
11(4), 669–679. https://doi.org/10.2307/134372
Gilbert, R. A. (1983). Economies of scale in correspondent banking: Note. Journal of Money,
Credit and Banking, 15(4), 483–488. https://doi.org/10.2307/1992163
Girardone, C., Molyneux, P., & Gardener, E. P. M. (2004). Analysing the determinants of
bank efficiency: The case of Italian banks. Applied Economics, 36(3), 215–227. https://doi.
org/10.1080/0003684042000175334
Glass, A. J., & Kenjegalieva, K. (2019). A spatial productivity index in the presence of effi-
ciency spillovers: Evidence for U.S. Banks, 1992–2015. European Journal of Operational
Research, 273(3), 1165–1179. https://doi.org/10.1016/j.ejor.2018.09.011
Goldberg, L. G., Hanweck, G. A., Keenan, M., & Young, A. (1991). Economies of scale and
scope in the securities industry. Journal of Banking & Finance, 15(1), 91–107.
Gorman, J. A. (1969). Alternative measures of the real output and productivity of com-
mercial banks. In Production and productivity in the service industries (pp. 155–199). NBER.
www.nber.org/books-and-chapters/production-and-productivity-service-industries/
alternative-measures-real-output-and-productivity-commercial-banks
Gourieroux, C., Monfort, A., & Renault, E. (1993). Indirect inference. Journal of Applied
Econometrics, 8, S85–S118.
Greenbaum, S. I. (1967). Competition and efficiency in the banking system – empirical
research and its policy implications. Journal of Political Economy, 75(4), 461–479.
Grifell-Tatjé, E., Lovell, C. A. K., & Pastor, J. T. (1998). A quasi-Malmquist productivity
index. Journal of Productivity Analysis, 10(1), 7–20.
Grigorian, D. A., & Manole, V. (2002). Determinants of commercial bank performance in tran-
sition: An application of data envelopment analysis. International Monetary Fund (Work-
ing Paper No. W.P./02/146). www.imf.org/en/Publications/WP/Issues/2016/12/30/
Determinants-of-Commercial-Bank-Performance-in-Transition-An-Application-
of-Data-Envelopment-15965
Gropper, D. M. (1991). An empirical investigation of changes in scale economies for the com-
mercial banking firm, 1979–1986. Journal of Money, Credit and Banking, 23(4), 718–727.
https://doi.org/10.2307/1992706
Grover, N., & Arora, N. (2015). Revisiting the technical efficiency gains of Indian commer-
cial banks from bancassurance. Review of Finance and Banking, 7(1), 19–29.
Guarda, P., Rouabah, A., & Vardanyan, M. (2013). Identifying bank outputs and inputs with
a directional technology distance function. Journal of Productivity Analysis, 40(2), 185–195.
Hancock, D. (1985). The financial firm: Production with monetary and nonmonetary
goods. Journal of Political Economy, 93(5), 859–880.
Harimaya, K. (2008). Impact of nontraditional activities on scale and scope economies:
A case study of Japanese regional banks. Japan and the World Economy, 20(2), 175–193.
https://doi.org/10.1016/j.japwor.2006.11.003
Hartwick, J. M. (1996). The financial capital constraint and the valuation of commercial bank-
ing activity. Queen’s University – Department of Economics (Working Paper No. 942).
https://ideas.repec.org/p/ags/queddp/273374.html
Havrylchyk, O. (2006). Efficiency of the Polish banking industry: Foreign versus domes-
tic banks. Journal of Banking & Finance, 30(7), 1975–1996. https://doi.org/10.2139/
ssrn.499922
 Review of evidence on banking output measurement 101
Hon, L. Y., Tuck, C. E., & Yu, K. L. (2011). Efficiency in the Malaysian banking industry.
ASEAN Economic Bulletin, 28(1), 16–44.
Howcroft, B., & Ataullah, A. (2006). Total factor productivity change: An examination
of the commercial banking industry in India and Pakistan. The Service Industries Journal,
26(2), 189–202. https://doi.org/10.1080/02642060500369305
Hsiao, H.-C., Chang, H., Cianci, A. M., & Huang, L.-H. (2010). First financial restruc-
turing and operating efficiency: Evidence from Taiwanese commercial banks. Journal of
Banking and Finance, 7(34), 1461–1471. https://doi.org/10.1016/j.jbankfin.2010.01.013
Huang, J., Matthews, K., & Zhou, P. (2020). What causes Chinese listed firms to switch
bank loan provider? Evidence from a survival analysis. Emerging Markets Review, 43,
100678. https://doi.org/10.1016/j.ememar.2020.100678
Huang, M.-Y., & Fu, T.-T. (2013). An examination of the cost efficiency of banks in Tai-
wan and china using the metafrontier cost function. Journal of Productivity Analysis, 40(3),
387–406.
Huang, M.-Y., Juo, J.-C., & Fu, T. (2015). Metafrontier cost Malmquist productivity index:
An application to Taiwanese and Chinese Commercial banks. Journal of Productivity Analy-
sis, 44(3), 321–335.
Huang, T.-H., Shen, C.-H., Chen, K.-C., & Tseng, S.-J. (2011). Measuring technical and
allocative efficiencies for banks in the transition countries using the fourier flexible cost
function. Journal of Productivity Analysis, 35(2), 143–157.
Hughes, J. P., & Mester, L. J. (1993). A quality and risk-adjusted cost function for banks:
Evidence on the ‘too-big-to-fail’ doctrine. Journal of Productivity Analysis, 4(3), 293–315.
Hughes, J. P., & Mester, L. J. (1998). Bank capitalization and cost: Evidence of scale econ-
omies in risk management and signaling. The Review of Economics and Statistics, 80(2),
314–325.
Hughes, J. P., & Mester, L. J. (2008). Efficiency in banking: Theory, practice, and evidence.
Draft Manuscript prepared for the Oxford Handbook of Banking.
Hughes, J. P., Mester, L. J., & Moon, C.-G. (2000). Are scale economies in banking elusive or
illusive? Evidence obtained by incorporating capital structure and risk-taking into models of bank
production. Federal Reserve Bank of Philadelphia (Working Paper No. 00–4). https://
ideas.repec.org/p/fip/fedpwp/00-4.html
Humphrey, D. B. (1987). Cost dispersion and the measurement of economies in banking.
FRB Richmond Economic Review, 73(3), 24–38.
Humphrey, D. B. (2009). Payment scale economies, competition, and pricing. European Central
Banking (Working Paper No. 1136). https://doi.org/10.2139/ssrn.1522022
Humphrey, D. B., & Pulley, L. B. (1997). Banks’ responses to deregulation: Profits, tech-
nology, and efficiency. Journal of Money, Credit and Banking, 29(1), 73–93. https://doi.
org/10.2307/2953687
Hunter, W. C., & Timme, S. G. (1986). Technical change, organizational form, and the
structure of bank production. Journal of Money, Credit and Banking, 18(2), 152–166.
https://doi.org/10.2307/1992199
Hunter, W. C., & Timme, S. G. (1991). Technological change in large U.S. Commercial
banks. The Journal of Business, 64(3), 339–362.
Hunter, W. C., & Timme, S. G. (1995). Core Deposits and Physical Capital: A Reexamina-
tion of Bank Scale Economies and Efficiency with Quasi-Fixed Inputs. Journal of Money,
Credit and Banking, 27(1), 165–185. https://doi.org/10.2307/2077857
Illueca, M., Pastor, J. M., & Tortosa-Ausina, E. (2009). The effects of geographic expan-
sion on the productivity of Spanish savings banks. Journal of Productivity Analysis, 32(2),
119–143.
102 Survey of evidences
Inklaar, R., & Wang, J. C. (2011). Real output of bank services: What counts is what banks do, not
what they own. Federal Reserve Bank of Boston (Working Paper No. 11–1).
Inklaar, R., & Wang, J. C. (2012). Measuring real bank output: Considerations and com-
parisons. Monthly Labor Review, 18–27.
Isik, I., & Hassan, M. K. (2002). Technical, scale and allocative efficiencies of Turkish bank-
ing industry. Journal of Banking & Finance, 26(4), 719–766. https://doi.org/10.1016/
S0378-4266(01)00167-4
Jacobs, D. (1964). The impact of examination practices upon commercial bank lending poli-
cies. In Committee on banking and currency. US Government Printing Office.
Jagtiani, J., & Khanthavit, A. (1996). Scale and scope economies at large banks: Includ-
ing off-balance sheet products and regulatory effects (1984–1991). Journal of Banking &
Finance, 20(7), 1271–1287.
Kalish, L., & Gilbert, R. A. (1973). An analysis of efficiency of scale and organizational form
in commercial banking. The Journal of Industrial Economics, 21(3), 293–307. https://doi.
org/10.2307/2098151
Kao, C., & Liu, S.-T. (2009). Stochastic data envelopment analysis in measuring the efficiency
of Taiwan commercial banks. European Journal of Operational Research, 196(1), 312–322.
https://doi.org/10.1016/j.ejor.2008.02.023
Karim, M. Z. A. (2001). Comparative bank efficiency across select ASEAN countries.
ASEAN Economic Bulletin, 18(3), 289–304.
Kaushik, S. K., & Lopez, R. H. (1996). Profitability of credit unions, commercial banks and
savings banks: A comparative analysis. The American Economist, 40(1), 66–78.
Kenjegalieva, K., Ravishjankar, G., Shen, Z., & Weyman-Jones, T. (2009). Modelling risk in
efficiency and productivity analysis of banking systems. Undefined, 1–36. www.semantics
cholar.org/paper/Modelling-risk-in-efficiency-and-productivity-of-Kenjegalieva-Ravish
jankar/075104d8e6f89e2133afda7c243b32fed8ccd6e8
Keshari, P. K., & Paul, M. T. (1994). Relative efficiency of foreign and domestic banks.
Economic and Political Weekly, 29(9), M31–M36.
Khan, A., Hassan, M. K., Maroney, N., Boujlil, R., & Ozkan, B. (2020). Efficiency, diver-
sification, and performance of U.S. Banks. International Review of Economics & Finance, 67,
101–117. https://doi.org/10.1016/j.iref.2019.12.010
Khan, I. U., Ali, S., & Khan, H. N. (2018). Market concentration, risk-taking, and
efficiency of commercial banks in Pakistan: An application of the two-stage double
bootstrap DEA. Business & Economic Review, 10(2), 65–96. https://doi.org/10.22547/
BER/10.2.4
Khusro, A. M., Raghavan, S. N., Ram, K., & Siddharthan, N. S. (1971). Banking efficiency
and banking growth. Economic and Political Weekly, 6(23), 1150–1152.
Kilbride, B. J., McDonald, B., & Miller, R. E. (1986). A reexamination of economies of
scale in banking using a generalized functional form: Note. Journal of Money, Credit and
Banking, 18(4), 519–526. https://doi.org/10.2307/1992471
Kim, M. (1985). Scale economies in banking: A methodological note. Journal of Money,
Credit and Banking, 17(1), 96–102. https://doi.org/10.2307/1992509
Kim, M. (1986). Banking technology and the existence of a consistent output aggregate.
Journal of Monetary Economics, 18(2), 181–195.
Kim, M., & Weiss, J. (1989). Total factor productivity growth in banking: The Israeli bank-
ing sector 1979–1982. Journal of Productivity Analysis, 1(2), 139–153.
Kirkwood, J., & Nahm, D. (2005). Australian banking efficiency and its relation to stock returns.
Macquarie University, Department of Economics (Research Papers No. 0508). https://
ideas.repec.org/p/mac/wpaper/0508.html
 Review of evidence on banking output measurement 103
Kolari, J., & Zardkoohi, A. (1990). Economies of scale and scope in thrift institutions: The
case of finnish cooperative and savings banks. The Scandinavian Journal of Economics, 92(3),
437–451. https://doi.org/10.2307/3440010
Kolari, J., & Zardkoohi, A. (1991). Further evidence on economies of scale and scope in
commercial banking. Quarterly Journal of Business and Economics, 30(4), 82–107.
Kouki, M., Park, S. S., & Renault, E. (2014). Estimating scale economies in financial inter-
mediation: A doubly indirect inference. Journal of Productivity Analysis, 41(3), 351–365.
Koutsomanoli-Filippaki, A., & Mamatzakis, E. (2009). Performance and Merton-type
default risk of listed banks in the E.U.: A panel VAR approach. Journal of Banking and
Finance, 11(33), 2050–2061. https://doi.org/10.1016/j.jbankfin.2009.05.009
Kumbhakar, S. C., Heshmati, A., & Hjalmarsson, L. (2002). How fast do banks adjust?
A dynamic model of labor-use with an application to Swedish banks. Journal of Productivity
Analysis, 18(1), 79–102.
Kumbhakar, S. C., & Sarkar, S. (2003). Deregulation, ownership, and productivity growth
in the banking industry: Evidence from India. Journal of Money, Credit and Banking, 35(3),
403–424.
Lawrence, C. (1989). Banking costs, generalized functional forms, and estimation of econo-
mies of scale and scope. Journal of Money, Credit and Banking, 21(3), 368–379. https://doi.
org/10.2307/1992419
Le, T. D. Q., & Ngo, T. (2020). The determinants of bank profitability: A cross-country
analysis. Central Bank Review, 20(2), 65–73. https://doi.org/10.1016/j.cbrev.2020.04.001
Lee, J. Y., & Kim, D. (2013). Bank performance and its determinants in Korea. Japan and the
World Economy, 27, 83–94. https://doi.org/10.1016/j.japwor.2013.05.001
Lee, T.-H., Liang, L.-W., & Huang, B.-Y. (2013). Do mergers improve the efficiency of
banks in Taiwan? Evidence from stochastic frontier approach. The Journal of Developing
Areas, 47(1), 395–416.
Leland, H. E., & Pyle, D. H. (1977). Informational asymmetries, financial structure,
and financial intermediation. The Journal of Finance, 32(2), 371–387. https://doi.
org/10.2307/2326770
Leong, W. H., Dollery, B., & Coelli, T. (2003). Measuring the technical efficiency of banks
in Singapore for the period 1993–99: An application and extension of the Bauer et al.
(1997) technique. ASEAN Economic Bulletin, 20(3), 195–210.
Longbrake, W. A., & Haslem, J. A. (1975). Productive efficiency in commercial banking: The
effects of size and legal form of organization on the cost of producing demand deposit ser-
vices. Journal of Money, Credit and Banking, 7(3), 317. https://doi.org/10.2307/1991625
Lozano-Vivas, A., & Pasiouras, F. (2010). The impact of non-traditional activities on the
estimation of bank efficiency: International evidence. Journal of Banking & Finance, 34(7),
1436–1449. https://doi.org/10.1016/j.jbankfin.2010.01.006
Lozano-Vivas, A., Pastor, J. T., & Pastor, J. M. (2002). An efficiency comparison of european
banking systems operating under different environmental conditions. Journal of Productivity
Analysis, 18(1), 59–77.
Lyroudi, K., & Angelidis, D. (2006). Measuring banking productivity of the most recent
European union member countries: A non-parametric approach. Journal of Economics and
Business, IX(1), 37–57.
Maghyereh, A. (2004). The effect of financial liberalization on the efficiency of financial
institutions: The case of Jordanian commercial banks. Savings and Development, 28(3),
235–270.
Mamatzakis, E., Tsionas, M. G., Kumbhakar, S. C., & Koutsomanoli-Filippaki, A. (2015).
Does labour regulation affect technical and allocative efficiency? Evidence from the
104 Survey of evidences
banking industry. Journal of Banking & Finance, 61, S84–S98. https://doi.org/10.1016/j.
jbankfin.2015.06.012
Mamonov, M., & Vernikov, A. (2017). Bank ownership and cost efficiency: New empiri-
cal evidence from Russia. Economic Systems, 41(2), 305–319. https://doi.org/10.1016/j.
ecosys.2016.08.001
Mansour, R., & Moussawi, C. E. (2020). Efficiency, technical progress and productivity of
Arab banks: A non-parametric approach. The Quarterly Review of Economics and Finance,
75, 191–208. https://doi.org/10.1016/j.qref.2019.02.002
Margono, H., Sharma, S. C., & Melvin, P. D. (2010). Cost efficiency, economies of scale,
technological progress and productivity in Indonesian banks. Journal of Asian Economics,
21(1), 53–65. https://doi.org/10.1016/j.asieco.2009.06.001
Mertens, A., & Urga, G. (2001). Efficiency, scale and scope economies in the Ukrainian
banking sector in 1998. Emerging Markets Review, 2(3), 292–308.
Mester, L. J. (1992). Traditional and nontraditional banking: An information-
theoretic approach. Journal of Banking & Finance, 16(3), 545–566. https://doi.
org/10.1016/0378-4266(92)90044-Z
Mester, L. J. (1997). Measuring efficiency at U.S. Banks: Accounting for heterogeneity is
important. European Journal of Operational Research, 98(2), 230–242.
Milenković, N., Radovanov, B., Kalaš, B., & Horvat, A. M. (2022). External two stage DEA
analysis of bank efficiency in West Balkan countries. Sustainability, 14(2), 978. https://doi.
org/10.3390/su14020978
Mohan, T. T. R. (2002). Deregulation and performance of public sector banks. Economic and
Political Weekly, 37(5), 393–397.
Mohan, T. T. R. (2003). Long-run performance of public and private sector bank stocks.
Economic and Political Weekly, 38(8), 785–788.
Mohan, T. T. R., & Ray, S. C. (2004). Comparing performance of public and private sector
banks: A revenue maximisation efficiency approach. Economic and Political Weekly, 39(12),
1271–1276.
Mosko, A., & Bozdo, A. (2016). Modeling the relationship between bank efficiency, capital
and risk in Albanian banking system. Procedia Economics and Finance, 319–327.
Mukherjee, A., Nath, P., & Pal, M. (2003). Resource, service quality and performance triad:
A framework for measuring efficiency of banking services. The Journal of the Operational
Research Society, 54(7), 723–735.
Muldur, U., & Sassenou, M. (1993). Economies of scale and scope in French banking and
savings institutions. Journal of Productivity Analysis, 4(1/2), 51–72.
Mullineaux, D. J. (1978). Economies of scale and organizational efficiency in banking:
A profit-function approach. The Journal of Finance, 33(1), 259–280. https://doi.org/
10.2307/2326364
Murillo-Melchor, C., Pastor, J., & Tortosa-Ausina, E. (2010). A bootstrap approach to ana-
lyse productivity growth in European banking. The Journal of the Operational Research
Society, 61(12), 1729–1745.
Murphy, N. B. (1972). A reestimation of the Benston-bell-murphy cost functions for a
larger sample with greater size and geographic dispersion. The Journal of Financial and
Quantitative Analysis, 7(5), 2097–2105. https://doi.org/10.2307/2329957
Nakane, M. I., Alencar, L. S., & Kanczuk, F. (2006). Demand for bank services and market power
in Brazilian banking. Central Bank of Brazil (Working Paper No. 107). https://ideas.repec.
org/p/bcb/wpaper/107.html
Nartey, S. B., Osei, K. A., & Sarpong-Kumankoma, E. (2019). Bank productivity in Africa.
International Journal of Productivity and Performance Management, 69(9), 1973–1997. https://
doi.org/10.1108/IJPPM-09-2018-0328
 Review of evidence on banking output measurement 105
Nelson, R. W. (1985). Branching, scale economies, and banking costs. Journal of Banking &
Finance, 9(2), 177–191.
Noulas, A. G., Ray, S. C., & Miller, S. M. (1990). Returns to scale and input substitution
for large U. S. Banks. Journal of Money, Credit and Banking, 22(1), 94–108. https://doi.
org/10.2307/1992130
Ohlson, M. P., Lima, G. A. S. F. de, & Takeda, T. (2021). Deposit insurance and brokerage
firms: Impacts on the market discipline of the Brazilian banking industry. Central Bank of Brazil
(Working Paper No. 542).
Ojo, A. T. (1979). Economies of scale in banking: The Nigerian case. Savings and Develop-
ment, 3(4), 277–295.
Okuda, H., & Saito, J. (2001). The production technology of philippine domestic banks in
the pre-Asian crisis period: Estimation of cost function in the 1990–96 period. Hitotsub-
ashi Journal of Economics, 42(2), 81–101.
Olson, D., & Zoubi, T. A. (2011). Efficiency and bank profitability in MENA countries.
Emerging Markets Review, 12(2), 94–110. https://doi.org/10.1016/j.ememar.2011.02.003
Oluitan, R. (2014). Bank efficiency in Africa. The Journal of Developing Areas, 48(2),
205–221.
Oral, M., & Yolalan, R. (1990). An empirical study on measuring operating efficiency and
profitability of bank branches. European Journal of Operational Research, 46(3), 282–294.
https://doi.org/10.1016/0377-2217(90)90002-S
Otero, L., Razia, A., Cunill, O. M., & Mulet-Forteza, C. (2020). What determines efficiency
in MENA banks? Journal of Business Research, 112, 331–341. https://doi.org/10.1016/j.
jbusres.2019.11.002
Owualah, S. I. (1986). Scale economies and efficiency of bank operations: Further evidence
from Nigeria. Savings and Development, 10(4), 419–431.
Owusu, F. B., & Alhassan, A. L. (2021). Asset-liability management and bank profitabil-
ity: Statistical cost accounting analysis from an emerging market. International Journal of
Finance & Economics, 26(1), 1488–1502. https://doi.org/10.1002/ijfe.1860
Pasiouras, F., Tanna, S., & Zopounidis, C. (2009). The impact of banking regulations on
banks’ cost and profit efficiency: Cross-country evidence. International Review of Financial
Analysis, 18(5), 294–302. https://doi.org/10.1016/j.irfa.2009.07.003
Pelosi, T. (2008). Risk, efficiency and industry dynamics in the Australian banking sector[Thesis].
UNSW. https://doi.org/10.26190/unsworks/17934
Phan, D. H. B., Narayan, P. K., Rahman, R. E., & Hutabarat, A. R. (2020). Do financial
technology firms influence bank performance? Pacific-Basin Finance Journal, 62, 101210.
https://doi.org/10.1016/j.pacfin.2019.101210
Phan, H. T. M., Daly, K., & Akhter, S. (2016). Bank efficiency in emerging Asian countries.
Research in International Business and Finance, 38, 517–530. https://doi.org/10.1016/j.
ribaf.2016.07.012
Powers, J. A. (1969). Branch versus unit banking: Bank output and cost economies. Southern
Economic Journal, 36(2), 153–164. https://doi.org/10.2307/1056432
Radecki, L. J. (1999). Banks’ payments-driven revenues, staff reports 62. Federal Reserve Bank
of New York.
Rajan, S. S., Reddy, K. L. N., & Pandit, V. N. (2011). Efficiency and productivity growth in
indian banking. Centre for Development Studies (Working Paper No. 199).
Rakshit, B., & Bardhan, S. (2022). Does bank efficiency enhance bank performance?
Empirical evidence from Indian banking. Buletin Ekonomi Moneter Dan Perbankan, 25,
103–124. https://doi.org/10.21098/bemp.v25i0.1844
Ray, S. C. (2007). Are some Indian banks too large? An examination of size efficiency in
Indian banking. Journal of Productivity Analysis, 27(1), 41–56.
106 Survey of evidences
Ray, S. C., & Das, A. (2010). Distribution of cost and profit efficiency: Evidence from
Indian banking. European Journal of Operational Research, 201(1), 297–307. https://doi.
org/10.1016/j.ejor.2009.02.030
Reddy, A. A. (2006). Productivity growth of regional rural banks. Economic and Political
Weekly, 41, 1079–1086. https://doi.org/10.2139/ssrn.1537538
Resti, A. (1997). Evaluating the cost-efficiency of the Italian banking system: What can be
learned from the joint application of parametric and non-parametric techniques. Journal
of Banking & Finance, 21(2), 221–250.
Rezvanian, R., Rao, N., & Mehdian, S. M. (2008). Efficiency change, technological pro-
gress and productivity growth of private, public and foreign banks in India: Evidence
from the post-liberalization era. Applied Financial Economics, 18(9), 701–713. https://doi.
org/10.1080/09603100701222317
Robin, I., Salim, R., & Bloch, H. (2018). Financial performance of commercial banks in
the post-reform era: Further evidence from Bangladesh. Economic Analysis and Policy, 58,
43–54. https://doi.org/10.1016/j.eap.2018.01.001
Rogers, K. E. (1998). Nontraditional activities and the efficiency of U.S. Commercial banks.
Journal of Banking & Finance, 22(4), 467–482.
Royster, S. E. (2012). Improved measures of commercial banking output and productivity.
Monthly Labor Review, 3–17.
Rushdi, M., & Tennant, J. (2003). Profitability of Australian banks: 1985–2001. Agenda:
A Journal of Policy Analysis and Reform, 10(3), 229–243.
Saeed, M., Izzeldin, M., Hassan, K., & Pappas, V. (2020). The inter-temporal relationship
between risk, capital and efficiency: The case of Islamic and conventional banks. Pacific-
Basin Finance Journal, 62. https://doi.org/10.1016/j.pacfin.2020.101328
Salim, R., Arjomandi, A., & Dakpo, K. H. (2017). Banks’ efficiency and credit risk analy-
sis using by-production approach: The case of Iranian banks. Applied Economics, 49(30),
2974–2988. https://doi.org/10.1080/00036846.2016.1251567
Sang, N. M. (2022). Impact of the COVID-19 Pandemic on bank efficiency in Vietnam.
Banks and Bank Systems, 17(1), 13–23. https://doi.org/10.21511/bbs.17(1).2022.02
Santosa, B. (2022). Indonesian banking efficiency in the post-Asian monetary crisis. Proceedings of
the First Lekantara Annual Conference on Public Administration, Literature, Social Sci-
ences, Humanities, and Education (LePALISSHE 2021). https://eudl.eu/doi/10.4108/
eai.3-8-2021.2315127
Sathye, M. (2003). Efficiency of banks in a developing economy: The case of India.
European Journal of Operational Research, 148(3), 662–671. https://doi.org/10.1016/
S0377-2217(02)00471-X
Sato, K. (1976). The meaning and measurement of the real value added index. The Review
of Economics and Statistics, 58(4), 434. https://doi.org/10.2307/1935875
Scholtens, B., & Wensveen, D. van. (2003). The theory of financial intermediation: An essay
on what it does (not) explain. In SUERF studies. SUERF – The European Money and
Finance Forum. https://ideas.repec.org/b/erf/erfstu/23.html
Schweitzer, S. A. (1972). Economies of scale and holding company affiliation in banking.
Southern Economic Journal, 39(2), 258–266. https://doi.org/10.2307/1056596
Sealey, C. W. (1977). A further reconsideration of optimal reserve management at com-
mercial banks. Southern Economic Journal, 44(1), 117. https://doi.org/10.2307/1057305
Sealey, C. W., & Lindley, J. T. (1977). Inputs, outputs, and a theory of production and cost
at depository financial institutions. The Journal of Finance, 32(4), 1251–1266. https://doi.
org/10.2307/2326527
Seelanatha, L., & Wickremasinghe, G. (2009). Financial reforms in Sri Lanka and their
influence on the banking industry. Bank and Bank Systems, 4(4), 23–35.
 Review of evidence on banking output measurement 107
Sensarma, R. (2005). Cost and profit efficiency of Indian banks during 1986–2003: A sto-
chastic frontier analysis. Economic and Political Weekly, 40(12), 1198–1209.
Setiwan, C., Hassan, T., Hassan, M. K., & Mohamed, S. (2017). Non-performing loans and
bank efficiency of conventional and Islamic banks in the organization of Islamic coopera-
tion (OIC) countries. Journal of Islamic Economics, Banking and Finance, 13(4). https://doi.
org/10.12816/0051000
Shaffer, S. (1993). A test of competition in Canadian banking. Journal of Money, Credit and
Banking, 25(1), 49–61. https://doi.org/10.2307/2077819
Shah, W. U. H., Hao, G., Zhu, N., Yasmeen, R., Padda, I. U. H., & Kamal, M. A. (2022).
A cross-country efficiency and productivity evaluation of commercial banks in South
Asia: A meta-frontier and Malmquist productivity index approach. PLoS One, 17(4),
1–17. https://doi.org/10.1371/journal.pone.0265349
Shair, F., Shaorong, S., Kamran, H. W., Hussain, M. S., Nawaz, M. A., & Nguyen, V. C.
(2021). Assessing the efficiency and total factor productivity growth of the banking
industry: Do environmental concerns matters? Environmental Science and Pollution Research,
28(16), 20822–20838. https://doi.org/10.1007/s11356-020-11938-y
Sharma, D., Sharma, A. K., & Barua, M. K. (2013). Efficiency and productivity of banking
sector: A critical analysis of literature and design of conceptual model. Qualitative Research
in Financial Markets, 5(2), 195–224. https://doi.org/10.1108/QRFM-10-2011-0025
Sharpe, S. A. (1997). The effect of consumer switching costs on prices: A theory and its
application to the bank deposit market. Review of Industrial Organization, 12(1), 79–94.
Sherwood, M. K. (1994). Difficulties in the measurement of service outputs. Monthly Labor
Review, 117(3), 11–19.
Shih, V., Zhang, Q., & Liu, M. (2007). Comparing the performance of Chinese banks:
A principal component approach. China Economic Review, 18(1), 15–34. https://doi.
org/10.1016/j.chieco.2006.11.001
Singh, S. P. (1974). Profitability of nationalised banks. Economic and Political Review, 9(35),
M67+M69–M72.
Sinha, R. P., & Chatterjee, B. (2008). Fund-based activity of Indian commercial banks:
A Malmquist approach. Indian Economic Review, 43(1), 83–102.
SNA. (1953). System of national accounts. United Nations Statistics Division.
SNA. (1993). System of national accounts. United Nations Statistics Division.
SNA. (2008). System of national accounts. United Nations Statistics Division. https://unstats.
un.org/unsd/nationalaccount/docs/sna2008.pdf
Speagle, R. E., & Kohn, E. (1958). Employment and output in banking, 1919–1955. The
Review of Economics and Statistics, 40(1), 22–35. https://doi.org/10.2307/1926477
Srivastava, P. (1999). Size, efficiency and financial reforms in Indian banking. Indian Council for
Research on International Economic Relations (Working Paper No. 49).
Stauffer, P. (2004). A tale of two worlds: How bankers and national accountants view banking.
International Association for Research in Income and Wealth.
Subrahmanyam, G. (1994). Measurement of output, scale economies and productivity in
banks: A methodological note. Indian Economic Review, 29(2), 223–227.
Subramaniam, I. (1985). A study of the measurement of bank output[Bachelor Degree Disserta-
tion]. The University of Adelaide.
Sufian, F. (2008). Risks and efficiency in Malaysian banking. Savings and Development, 32(1),
7–29.
Swamy, D. S. (1973). Economies of scale in banking: A critique. Economic and Political Weekly,
8(8), M13+M15–M16.
Syed, A. (2021). Assessing the impact of banking efficiency, operations, and regulation on
banking performance: Fresh insight using dynamic correlated framework on the data set
108 Survey of evidences
of Russia and Ukraine. Journal of Eastern European and Central Asian Research (JEECAR),
8(1), 89–99. https://doi.org/10.15549/jeecar.v8i1.514
Synder, C. (1923). Fisher’s ‘the making of index numbers’. The American Economic Review,
13(3), 416–421.
Tachibanaki, T., Mitsui, K., & Kitagawa, H. (1991). Economies of scope and shareholding of
banks in Japan. Journal of the Japanese and International Economies, 5(3), 261–281.
Tadesse, S. (2005). Consolidation, scale economies and technological change in Japanese
banking. Journal of International Financial Markets, Institutions & Money, 16, 425–445.
https://doi.org/10.2139/ssrn.681534
Tan, Y., & Floros, C. (2014). Risk, profitability, and competition: Evidence from the Chi-
nese banking industry. The Journal of Developing Areas, 48(3), 303–319.
Thanassoulis, E. (1999). Data envelopment analysis and its use in banking. Interfaces, 29(3),
1–13.
Thilakaweera, B. (2016). Efficiency and productivity in Sri Lanka’s banking sector: Evidence from
the post-conflict era [Ph.D. Thesis]. University of Wollongong. https://ro.uow.edu.au/
theses/4938
Tirtiroğlu, D., Daniels, K. N., & Tirtiroğlu, E. (2005). Deregulation, intensity of competi-
tion, industry evolution, and the productivity growth of U.S. Commercial banks. Journal
of Money, Credit and Banking, 37(2), 339–360.
Todhanakasem, W., Lynge, M. J., Primeaux, W. J., & Newbold, P. (1986). Economies of scale
and organization efficiency in banking. Managerial and Decision Economics, 7(4), 255–261.
Tokle, R. J., & Tokle, J. G. (2000). The influence of credit union and savings and loan com-
petition on bank deposit rates in Idaho and Montana. Review of Industrial Organization,
17(4), 427–439.
Tortosa-Ausina, E. (2002). Bank cost efficiency and output specification. Journal of Productiv-
ity Analysis, 18(3), 199–222.
Triplett, J. E., & Bosworth, B. P. (2004a). Measuring banking and finance: Conceptual
issues. In Productivity in the U.S. Services sector: New sources of economic growth (pp. 177–210).
Brookings Institution Press. www.jstor.org/stable/10.7864/j.ctt12879w8.13
Triplett, J. E., & Bosworth, B. P. (2004b). Productivity in the US services sector: New sources of
economic growth. Bibliovault OAI Repository, The University of Chicago Press.
Tzeremes, N. G. (2015). Efficiency dynamics in Indian banking: A conditional directional
distance approach. European Journal of Operational Research, 240(3), 807–818. https://doi.
org/10.1016/j.ejor.2014.07.029
Vanoli, A. (2005). A history of national accounting. IOS Press.
Vennet, R. V. (2002). Cost and profit efficiency of financial conglomerates and universal
banks in Europe. Journal of Money, Credit and Banking, 34(1), 254–282.
Verghese, S. K. (1983). Profits and profitability of Indian commercial banks in seventies.
Economic and Political Weekly, 18(48), M145–M157.
Wang, J. C., & Basu, S. (2005). Risk bearing, implicit financial services, and specialization in the
financial industry. Social Science Research Network (SSRN Scholarly Paper No. 917427).
https://doi.org/10.2139/ssrn.917427
Wanke, P., Azad, A. K., & Emrouznejad, A. (2018). Efficiency in BRICS banking under
data vagueness: A two-stage fuzzy approach. Global Finance Journal, 35, 58–71. https://
doi.org/10.1016/j.gfj.2017.05.001
Weill, L. (2004). Measuring cost efficiency in European banking: A comparison of frontier
techniques. Journal of Productivity Analysis, 21(2), 133–152.
Westhuizen, G. (2013). Bank productivity and sources of efficiency change: A case of the
four largest banks in South Africa. International Business & Economics Research Journal,
12(2), 127. https://doi.org/10.19030/iber.v12i2.7625
 Review of evidence on banking output measurement 109
Weyman-Jones, T., Boucinha, M., Kenjegalieva, K., Ravishankar, G., Ribeiro, N., & Shen, Z.
(2010). Incorporating risk in the efficiency and productivity analysis of banking systems. In
New issues in financial institutions management (pp. 91–105). Palgrave Macmillan. https://ideas.
repec.org/h/pal/pmschp/978-0-230-29915-3_6.html
Wheelock, D. C., & Wilson, P. W. (1995). Explaining bank failures: Deposit insurance,
regulation, and efficiency. The Review of Economics and Statistics, 77(4), 689–700. https://
doi.org/10.2307/2109816
Wheelock, D. C., & Wilson, P. W. (1999). Technical progress, inefficiency, and productivity
change in U.S. Banking, 1984–1993. Journal of Money, Credit and Banking, 31(2), 212–234.
https://doi.org/10.2307/2601230
Wu, M., Li, C., Fan, J., Wang, X., & Wu, Z. (2018). Assessing the global productive effi-
ciency of Chinese banks using the cross-efficiency interval and VIKOR. Emerging Markets
Review, 34, 77–86. https://doi.org/10.1016/j.ememar.2017.10.005
Yu, M.-M., Lin, C.-I., Chen, K.-C., & Chen, L.-H. (2019). Measuring Taiwanese bank
performance: A two-system dynamic network data envelopment analysis approach.
Omega. https://doi.org/10.1016/j.omega.2019.102145
Zaim, O. (1995). The effect of financial liberalization on the efficiency of Turkish commercial
banks. Applied Financial Economics, 5(4), 257–264. https://doi.org/10.1080/758536876
Zaman, M. S., Valiyattoor, V., & Bhandari, A. K. (2022). Dynamics of total factor productivity
growth: An empirical analysis of Indian commercial banks. Social Science Research Network
(SSRN Scholarly Paper No. 4040619). https://doi.org/10.2139/ssrn.4040619
Zha, Y., Liang, N., Wu, M., & Bian, Y. (2016). Efficiency evaluation of banks in China:
A dynamic two-stage slacks-based measure approach. Omega, 60, 60–72. https://doi.
org/10.1016/j.omega.2014.12.008
4 Review of evidence
on insurance output
measurement
Global perspective

4.1. Introduction
Similar to Chapter 3, this chapter surveys and reviews the literature on insur-
ance output measurement by focusing on the selected studies on insurance
performance analysis. Insurance is primarily a tool for protection against for-
tuitous risks whose outcomes are dichotomous with either a non-profit-no-
loss outcome or a loss after an event. Thus, in crude terms, “the purpose of
insurance is to protect risk-averse individuals from suffering the full conse-
quences of those actions on the part of nature which affect them unfavourably”
(Spence & Zeckhauser, 1971, p. 380). Analysis of various performance param-
eters of insurance firms and industry, such as their efficiencies, productivities,
financial performance, profitability and others, has a long history. A crucial
analytical ingredient in the evolution of thought on insurance performance
measurement has been the insurance output measurement.1 There is a wide
variety of thought and approaches on this matter.2 The same are presented in
the pages to follow.

4.2. Empirical sensitivity of insurance performance

estimates to alternative outputs
Unlike the literature on banking, studies on insurance performance analysis
have not delved deeper into the sensitivity of performance estimates to alterna-
tive output measures. Even though a few studies were found to have analysed
this dimension, it is scarce compared to its banking counterpart, as examined
in section 3.2. While the insurance literature has used multiple outputs in their
analysis, examination of the sensitivity of those estimates to different output
measures does not seem to have been a major concern. This is surprising given
that the debates on output measurement are intense in insurance and banking
performance analysis. Two works could be found in the literature that assessed
this dimension. The first is the work of Leverty and Grace (2010), who exam-
ined the empirical sensitivity of properly-liability insures’ efficiency estimates
to alternative output measures using DEA and range adjusted measure (EAM)
of efficiency. They specified outputs in terms of the value-added approach

DOI: 10.4324/9781003149828-5
 Review of evidence on insurance output measurement 111
and the flow approach. They measured insurance output regarding the present
value of real losses incurred and average real invested assets (for the value-added
approach) and rate of return on investments, amount of liquid asset and sol-
vency score (for the flow approach). They find that “firms identified as highly
efficient by the value-added approach are less likely to fail, while firms with
high flow efficiency are more likely to fail” (p. 1552). The second work is by
Lee (2013), who studies the interrelationship between the insurance market’s
development and the real aggregate output while controlling banking activi-
ties. The author uses alternative measures of real insurance premiums as proxies
of insurance activity measured by life, nonlife and total insurance premiums.
The author finds that insurance output as measured by life and nonlife outputs
separately provides better results than combining them.

4.3. Approaches to output specification in the

insurance literature
The qualifications and concerns raised in section 3.3 are as valid here as they
are in the case of bank output measurement. The approaches examined below
have been summarised based on the selected literature on insurance output
measurement, as represented in Table 2.2.

Production approach
As discussed in section 3.3, the production approach theorises a firm as a
transformer of primary inputs of labour, capital and funds into outputs. This
approach specifies inputs like labour, physical capital and possibly equity in
insurance literature.3 In contrast, outputs are specified in value-form in terms
of premiums received and claims or benefits paid. The discussions about the
fundamental properties of the production approach in section 3.3 are equally
applicable here. The inputs are generally defined in terms of labour, physical
capital and financial capital.4 The outputs for life insurers are mostly defined
as the amount of premium and benefit paid. Those for nonlife insurers are
defined as the amount of premium and loss/claim. Insurers are seen purely in
their risk transfer and risk coverage functions. There is no recognition of the
financial and risk intermediations undertaken by the insurance industry under
this approach. Under the production approach, the service vector of insurers is
sufficiently captured by Carrington et al. (2011). They highlight risk pooling
and bearing, direct financial services and indirect financial services as the pri-
mary functions of insurers. Literature suggests that such a conceptualisation of
insurance production function is ideally suitable for branch-level analysis rather
than industry-level analysis. “The production approach is preferable for evalu-
ating the efficiency of single branches, while the intermediation approach may
be more appropriate for evaluating entire financial institutions” (Fiordelisi &
Ricci, 2012, p. 184). Even though the authors suggested this about the banking
industry, these words seem equally representative of the consensus in insurance
112 Survey of evidences
literature. This approach to insurance output specification looks at insurers as
typical production units whose primary function is to transform resources into
consumption while producing directly visible and measurable output. Insurers
are not seen as intermediaries of any sort under this framework. A large portion
of the sample studies has employed this approach (84 out of 190), though the
most popular approach is an extension of this approach called the value-added
framework. It is discussed later in this chapter.

Intermediation approach
The single-most raised criticism against the production approach in insurance
output specification has been its utter lack of recognition of the intermediation
functions5 performed by insurers. On the one hand, insurers collect premiums
and under risk pooling to generate and fulfil claim liabilities. On the other
hand, insurers use premiums to pool the funds and generate wealth and capital
appreciation for policyholders and shareholders. These functions are essentially
intermediation functions (Cummins et al., 2009). The former may be called
the risk intermediation function, wherein insurers transform funds (premiums)
into payment for liabilities (claims/losses). The latter may be called ‘financial
intermediation’6 wherein insurers utilise the premiums to generate a surplus,
which is then invested into various assets to maximise shareholders’ wealth and
improve the investment returns for policyholders in the case of life insurance.
These essential functions of insurers are ignored in the production approach.
However, the intermediation approach allows the recognition of this fact and
thus specifies outputs and inputs in a suitable manner. As noted by Eling and
Luhnen (2010b, p. 229): “intermediation approach views the insurance com-
pany as a financial intermediary that manages a reservoir of assets, borrowing
funds from policyholders, investing them on capital markets, and paying out
claims, taxes, and costs”. Thus, the main difference between the production
and intermediation approaches in insurance literature emanates from the dif-
ference in the treatment of premiums as an input rather than an output (Barros
et al., 2005). Criticisms of premiums as output have been widely recognised
in the literature.7 The intermediation approach poses amount of premium as
an input while treating claims as output for capturing risk intermediation and
financial assets as output for capturing financial intermediation. Some studies
have also recognised equity capital as an essential input for undertaking risk
intermediation. This is done mainly by recognising the fact that equity capital8
supplements premiums as a source for meeting future claim liabilities. In sum-
mary, the intermediation approach in insurance output specification recognises
the amount of claim and financial assets as the correct output measures.9

Value-added approach
One of the dissatisfactions with the production approach was its ad hoc speci-
fication of output measures. This was addressed via the so-called value-added
 Review of evidence on insurance output measurement 113
approach in insurance as laid down by Cummins and Weiss (2000).10 In basic
terms, “the value-added approach is derived from the micro-economic theory
of the firm and is based on the theoretical premise that firms maximise profits
by jointly minimising costs and maximising revenues” (Leverty & Grace, 2010).
As explained in the case of bank output specification in section 3.3 earlier, the
value-added approach considers all those items as output that produce revenues
for the firm.11 Literature has argued that financial intermediation is a tiny pro-
portion of the total services provided by insurers. Thus the production/value-
added approach is the correct representation of the theoretical production
function of the insurance firms (Cummins & Weiss, 2000; Leverty & Grace,
2010; Carrington et al., 2011, among others).12 In the value-added approach,13
insurers’ activities are recognised as value-producing for the final consumers,
i.e. the policyholders: Risk-pooling and risk-bearing,14 real financial services
relating to insured losses and financial intermediation15 (Mahlberg & Url, 2010;
Cummins & Xie, 2008). Hence, the value-added approach treats the amount
of losses/claims, assets and premiums as the correct output measures and is an
extension of the traditional production approach.

National income approach

The SNA manual provides the extant wisdom on measuring life and nonlife
insurance outputs under the national income accounting framework. Unlike
its banking counterpart, insurance output measurement at the aggregate level
is not complicated by the FISIM method. Most of the services of insurers are
generally reflected in their balance sheets mainly because of the nature of pric-
ing in insurance. Insurers, unlike bankers, do not need to charge implicitly and
instead charge for all their services in the premium amount itself. This allows
macro accountants to capture the majority of insurance services from the bal-
ance sheet and income statements of insurers which is not feasible in the case
of banking firms. The output measures under this approach are examined later
in this chapter.

4.4. Output measurement in the insurance literature

Life versus nonlife insurance output

At the outset, it is necessary to recognise that insurance output measurement is
inherently different for the life and the nonlife segments. Both these segments
have a very different set of services for their consumers. Both the segments are
also governed by different laws and are affected by factors that may be mutually
exclusive. The nature of risks covered in these segments is different, and thus
their outputs must be analysed separately. However, these segments have been
combined to measure total insurance output in the literature, though such stud-
ies are relatively lesser. The differences are not very large either. Some measures
are used differently for these segments, mainly due to different accounting
114 Survey of evidences
conventions and balance sheet forms across both these segments. The discus-
sions below examine the output measures and approach considering both the
segments separately but do not necessarily specify it explicitly. It is assumed that
the readers will interpret the different sections while considering this.

Premium output
Premiums represent the most critical source of funds for insurers. By pooling
premiums, insurers can promise claims and fulfil their liabilities. As suggested
in the literature, the amount of premium collected by insurance companies can
possibly represent their technical activity. The ability of insurers to fulfil their
liabilities is directly conditional upon the amount of premium that they collect.
Premiums have been theorised as a proxy for risk transfer and management
services rendered by insurers to their consumers. It is used as a measure for
output because it is assumed to be strongly and positively correlated with most
services provided by insurers. This is a natural result that premiums signify the
chief source of funds out of which insurers can undertake all their economic
activities.
In the literature, premiums have been used in three different forms as meas-
ures of insurance output: first in the form of the amount of premium, second
as a ratio measure and third in other forms, as explained in Figure 4.1. Use of
premiums in the form of the amount in the literature has been either in an
aggregate form for a firm or the entire industry or in disaggregated form across
product groups, coverage groups or segments. The primary debate in premium
output measurement has been the treatment of reinsurance claims received by
the insurance firms. The gross measure of premium amount includes reinsur-
ance benefits received by the insurance firm and removing its impact from
the gross premium amount yields net premiums. Netting has also been done
using other methods such as subtracting the amount of claims paid from the
premiums received (life insurance) and subtracting losses incurred from pre-
miums received (nonlife insurance), and adding financial returns to premiums
to measure the total available funds with the insurers. Premiums in themselves
have been defined differently in the literature. Studies have conceptualised
the amount of premiums as earned premiums (Weiss, 1991; Brockett et al.,
2005; Rao, 2016, among others), unearned premiums (Fenn et al., 2008; Lev-
erty & Grace, 2010; Chen et al., 2011; Zhong & Sun, 2011; Leverty & Grace,
2012); returned premiums (Mahlberg & Url, 2003), lapsed premiums16 and
retained premiums (Ai et al., 2015). Literature has also used the premium and
the amount of investments as a single output measure. This has generally been
practiced to capture the total funds available with insurers to undertake inter-
mediation activities. Some researchers have recommended adding equity capi-
tal to this measure output, but it has probably not been practised.
Another line of approach to premium output measurement is ratio meas-
ures. Two measures have been used popularly in the extant literature: the first
is the ratio of gross or net premiums to losses or claims (such as in Trufin
 Premium

Amount
Ratio
Aggregate measures
Others
Premiums Premiums
Gross to losses to assets
(Direct)
Other PCA-based
Plus Reinsurance additions components
assumed Dsiaggregated
Investment

Review of evidence on insurance output measurement 115

income
Less reinsurance Bancassurance
Based on ceded premium
realization Segment-
Net wise
Premium Products-wise
Earned Reinsurance
Other premium premiums
Netting Coverage
approaches Unearned scope wise Non-life
premium Unearned
premiums Individual Group
Returned reserves policies policies Premiums per
premium policyholders
Financial
returns
Lapsed
premium
Claims Life First-year
premium
Retained
Losses premium
incurred
Year-end
premium
(Stock)

Figure 4.1 Premium output measures in insurance literature

Source: Author’s analysis based on the literature survey.
116 Survey of evidences
et al., 2009), and the second is the ratio of premiums to total assets or only the
financial assets. These two measures of insurance output have been used mainly
in analysing underwriting cycles in insurance or for financial and operational
performance analyses. Lastly, some novel approaches are categorized as ‘others’
in Figure 4.1. Among these, the Principal Components Analysis (PCA)–based
measures of premium amount have been employed by Jeng et al. (2007). Some
researchers have focused on the reinsurers rather than direct insurers as well
on the reinsurance activities of nonlife insurers (such as Kao & Hwang, 2008;
Wu & Zeng, 2011; Anandarao et al., 2018; Almulhim, 2019). Some other
measures too have been used and the same may be located in the appendix to
this chapter.

Claims and loss output

Conceptually, claims are the losses that insurance firms have to bear due to the
liabilities arising out of insurance contracts.17 Literature has used different con-
cepts in its place, such as ‘losses’ and ‘benefits paid’ (in life insurance). There is a
belief in the literature that there is a strong positive correlation between claims
paid and insurers’ losses (Wise, 2018). The frequency and severity of losses dif-
fer between life and nonlife segments and among different lines of products
within the segments. This results in differences in the amount of claims or the
number of claims paid by insurance firms. The term ‘claim’ has been used in a
general sense in the literature.18 It can be defined as claims incurred (Fenn et al.,
2008) and claims paid. Incurred claims are composed of the claims paid and the
additions to technical reserves.19 Both represent claim liabilities of an insurer
but in different ways. The former signifies the amount of claims due for pay-
ment by the firms (including its expectations during the current year), while
the latter shows the amount paid during settlement in a given year. Similar to
the premiums measure, there is a gross form and net form of the amount of
claims. The gross form is unadjusted for the reinsurance20 claims received by
the firms. The net for then is the adjusted measure. There are other adjustment
approaches in the literature also, as explained in Figure 4.2. These include the
changes in loss reserves, loss adjustment expenses and bonuses and rebates paid
by the firms.
The difference between claims and losses arises when one approaches insur-
ers’ liabilities in an ex-ante framework. Insurers expect some amount of losses to
occur, whether they are life or nonlife claims inevitably. Literature has termed
it expected loss (Weiss, 1990; Cummins & Xie, 2008). Studies have used two
measures of expected losses21: first is the implicit measure which is estimated as
the product of the amount of gross or net premiums and the average loss ratio
(such as in Leverty & Grace, 2010); second is the present value of losses (Cum-
mins & Nini, 2002; Choi & Weiss, 2005 across a given period. Another con-
cept used has been ‘technical losses’ (Ferro & León, 2017). These concepts are
different ways of approaching the financial liabilities of insurers. The choice on
this account largely depends on the segment of focus and the data availability.
 Claims and
Losses

Financial losses Total losses

Technical
Claims Losses

Net (Gross Paid

Gross
adjusted for) Incurred

Review of evidence on insurance output measurement 117

Un-adjusted Reinsurance Current
for received Loss adjustment period
reinsurance expenses
received
Changes in
loss reserves Cumulative
Bonus and
rebates
Loss
Reserves First-report

Expected Underwriting and

Incurred administrative
(Due) expenses

Present value of
Implicit expected loss

Premiums Average
earned Loss Ratio
Expected loss
over a duration Benefits
incurred
(Life)

Figure 4.2 Claims and losses as insurance output in the literature

Source: Author’s analysis based on the review of the literature.
118 Survey of evidences
In terms of popularity, the so-called real present value of losses incurred22
and paid is the most preferred choice for capturing the financial liabilities
of insurers and measuring their output. As noted by (Cummins et al., 2010,
p. 1529): “Proxying output by losses incurred is appropriate because the objec-
tive of risk-pooling is to collect funds from all policyholders and redistribute
them to those who incur losses”. This measure is used as it is or discounted to
reflect its present value. This may be necessary because one of the components
of expected losses is the estimated value of liabilities for the future. This future
value component requires discounting to represent its current value correctly.
The choice of the discount rate and method can change the quantum of out-
put and thus change performance estimates. Discounting is necessary23 “for the
revenue and output elements of the insurance price measure to consistently
reflect present value concepts” (Cummins & Xie, 2008, p. 15). Losses have
been criticised as measures of insurance output on various grounds (Cum-
mins & Weiss, 2000; Carrington et al., 2011). The first criticism is that, like
claims, this represents a bad output rather than a good output. The second
issue is that it is a random variable and may have statistical noise24 (Cummins &
Weiss, 2000). The third problem with this output measure is that it can be
induced by internal inefficiencies rather than external problems. As noted by
(Carrington et al., 2011, p. 127): Variations in incurred losses can also be due to
“lax company risk mitigation procedures or enforcement of policy conditions
or both could contribute to incurred losses”. The fourth criticism is that losses
represent past obligations rather than current output (Carrington et al., 2011).
Finally, literature has argued that losses ignore the quality of loss control and
risk management. Yaisawarng et al. (2012) noted that “an insurer with excel-
lent loss-prevention practices would have a lower amount of losses incurred and
therefore produce lower output” while, in fact, it has improved.
Lastly, a scarcely used measure of loss output in the literature is the inverse
loss ratio.

The inverse loss ratio (premiums divided by losses) shows the price that an
insurance firm gets for each unit of losses paid and is taken as an approxi-
mation of the unit price. It gives an indication of what customers get back
from insurers for the premiums they have paid and may be regarded as a
proxy for the output of the insurance industry.
Trufin et al., 2009, p. 387

Shifting from value to volume measures, they include the number of claims
received (e.g. Segovia-Gonzalez et al., 2009) and the number of claims paid/
settled (e.g. Manikowski & Weiss, 2013). These have been utilised sparsely in
the literature, though. The use of claims as an insurance output measure is gen-
erally justified because it reflects the risk pooling services undertaken by insur-
ers. Both premiums and claims are used as proxies for measuring the output of
this service in the literature. This measure of insurance output is not without its
own set of criticisms. The first criticism is that claims cannot be treated as goods
 Review of evidence on insurance output measurement 119
but must be treated as ‘bads’. This is because, in terms of economic theory, an
output variable must be subject to maximisation, and maximising claims seems
irrational from a rational firm’s perspective. Thus, this output measure does not
satisfy the critical characteristic of an ideal output measure as identified in the
literature, i.e. more output must be preferred over less (Gaganis et al., 2013).
The second criticism has been that “claims do not represent current expenses;
they measure past activity” (Wise, 2018, p. 13). The author also states that “as
claims represent the end of the policy, the outcome is a loss of future profits
and probably negative profits now” (p. 13). However, as stated earlier, the belief
in a strong positive association between claims and losses seems to be the chief
motivation for using this output measure in the literature.
The amount of losses as a measure of insurance output has a long his-
tory in the insurance literature. Notable works in this regard include Weiss
(1991), Cummins and Weiss (2000), Cummins et al. (2004), Cummins and Xie
(2008, 2013), Yaisawarng et al. (2012), Schlesinger (2000), Carrington et al.
(2011), and Wise (2017, 2018). Three measures of prevalent in the literature.
Actual losses are the losses paid by the insurers in a given period; incurred
losses, broadly defined as the losses paid plus changes in technical reserves and
expected losses, generally defined as the amount of liabilities to be paid ex-ante
using statistical forecasting models. A few studies have also used loss adjustment
expenses to measure insurance output. Pal et al. (2017) is an illustration of the
same. Another novel measure in the literature is the amount of incurred losses –
either on a first-reported or developed basis.25 A vital feature of loss measures
of insurance output is that these are generally flow measures.

Investment and financial assets output

The intermediation and value-added approach have popularly utilised the
amount of financial assets to measure insurance output. The essence of this
measure is that it proxies the wealth maximisation efforts undertaken by insur-
ers for its owners and policyholders and represents the additional funds available
with the insurers (in terms of returns generated) to handle future liabilities.26 It
is a balance sheet item and thus is a stock measure of insurance output. Stud-
ies have used flow and stock measures together in many cases. A debate in the
literature has been whether to proxy the intermediation services of insurers
by the amount of financial assets or to use returns on investments. The extent
of financial intermediation services provided by insurers must account for the
value of financial investments and the amount of funds used by the insurers
to generate these investments. In that light, return on investments seems to
be a relatively superior measure. Assets under management/Invested assets/
Financial assets are the different terms this output measure is characterised
in the literature. The essence of this output variable is that the surplus from
the policyholder’s money is invested in financial markets to earn a return on
the idle money until claims are paid. These returns are then passed on to the
owners and policyholders – whether in life or nonlife segments.27 Figure 4.3
 Assets

120 Survey of evidences

Policyholder
surplus

Change in surplus Assets under

(flow) Management

Level of surplus (stock)

Amount of Investment Yield on

Others
investments income investments

In Financial Physical On physical Return on

Lending
markets assets assets investments

On financial
Traditional Others Real estate Bullion Others assets Loans to
policyholders
Capital Money Foreign Rent income Capital gains
market market exchange
market To others
Interest
Capital gains
income
Derivatives
market Returns on Liquidity-
investments based
Dividends
Bank
deposits Yield on
Liquid assets
investments

Others Returns on Illiquid

investments assets

Figure 4.3 Investments and assets as insurance output in the literature

Source: Author’s analysis based on the review of the literature.
 Review of evidence on insurance output measurement 121
summarises the primary ways in which literature has employed this variable to
measure insurance output.

Reserves as output
The consensus in the literature on the use of reserves as an insurance output
is presented in Figure 4.4. Another prevalent input-output dilemma in the
literature has been whether to treat technical reserves as an input or an output.
Essentially, the consensus has been to treat it as an input into the insurance
production function. Some studies have, however, used it as an output. Notable
works include Mahlberg and Url (2003) and Gaganis et al. (2013). Some nota-
ble recent studies treating it as input include Parida and Acharya (2017); Ksenija
(2017); Alshammari et al. (2019); Krupa et al. (2019); Yakob et al. (2014); and
Sun (2020), among many others before them. The amount of reserves is a
stock variable. This is one way in which the literature has employed it. Another
way literature has used this concept is through ‘additions to reserves’, a flow
variable. The term ‘reserve’ can include concepts other than technical reserves,
such as Incurred But Not Reported (IBNR) reserves, reinsurance reserves and
reserves for primary insurance contracts. Chiefly, however, this term is used for
representing technical reserves.
Technical reserves are generally statutory (and may also include voluntary)
provisions made by insurers to ensure that future expected losses are fulfilled.
This measure has been conceptualised as representing the risk coverage services
produced by insurers. The level and movements of this measure are assumed to
be strongly correlated with the amount of risk coverage that the insurance firm
can provide. It may be considered a measure of insurers’ productive capacity
to fulfil their future liabilities. Its use as a measure of input, particularly debt
capital, has been the most prevalent practice. This is because it is criticised as
a measure of output on several pressing grounds. The first has been that its
estimation may differ across insurers, and thus use of the same in performance
analysis can generate biased and unreliable results. Another criticism has been
that this variable may change due to exogenous shocks even when the amount
of premiums and the number of policies issued have not changed. This can
happen due to changes in risk assessment by the underwriters and risk manag-
ers in the firm. Studies have also avoided using technical reserves as an output
measure due to high cross-sectional heterogeneity and differences in regulatory
norms across regions.28 Lastly, analysts have argued that technical reserves rep-
resent the future, and thus its use as a measure of current output is inconsistent
with its fundamental nature (Wise, 2018).

Income output
Income output is conceptualised as operating income in the insurance litera-
ture. It has two major components: premium income (as discussed earlier) and
investment income. Investment income is the flow of returns generated by the
 Reserves

122 Survey of evidences

Claims
Premiums

Past
claims Past
Other premiums
Liabilities Expected
premiums
Current Current
claims premiums
Unplanned
Expected expenses
claims
Present
Unexpected Value
claims
Present
Value Other Assets
Discounting
Forecasting Method factor
Investment elements
Discounting income
factor
Unplanned Actuarial
incomes
Pure
mathematical

Others Statistical

Reinsurance
reserves
IBNR reserves Type of Tail
For primary
insurance contracts

Figure 4.4 Reserves as a measure of insurance output

Source: Author’s analysis based on the extant literature.
 Review of evidence on insurance output measurement 123
financial investments undertaken by insurers. Given that surplus will gener-
ally be available, it is invested in various financial instruments subject to the
risk-preference of insurers constrained under extant regulatory norms. Invest-
ments made in financial assets will generate a flow of earnings, broadly called
investment income. Given its flow nature, its use is preferable against stock
variables. Investment income is generated in interest income and volatile spec-
ulative returns. Insurers tend to invest in risky (e.g. equity) and interest-bearing
(debentures, bonds, etc.) assets. The different ways in which these investments
can be classified are explained in Figure 4.5.
Literature has recognised the role of investment income in substituting pre-
mium income. With larger investment incomes, insurers can suppress premium
rates below the competitive levels to achieve higher market share (Mahlberg &
Url, 2010), mainly if the price elasticity of insurance demand is high. The
portfolio of invested assets and their composition also plays a vital role in deter-
mining whether this measure can be used empirically for output analysis of
insurance. There can be excessive volatility and instability if an insurer’s expo-
sure to equity and other risky assets is high. Exogenous factors might induce
changes in this measure, producing biased estimates of performance parameters
such as efficiency, productivity and profitability. Literature has also recognised
the income from physical assets, but this dimension has not been employed
frequently. Data availability and problems in computing the price and returns
on physical capital might be the main reasons.

National income output of insurance

The SNA manual provides helpful notes on how national income accountants
approach the issue of insurance output measurement. The broad framework is
presented in Figure 4.6.
In national income accounting, the insurance sector is looked at from at least
four perspectives: direct insurance, reinsurance, retrocession (insurance assumed
by reinsurers) and social insurance. The largest share is occupied by direct insur-
ers and reinsurers in the sector output of the insurance industry in any economy.
The output estimation procedures for reinsurers are precisely the same for direct
nonlife insurers irrespective of “whether it is life or nonlife policies that are
being reinsured” (SNA, 2008, p. 119). SNA (2008) defines the output of direct
insurers for the life segment as Premiums earned + premium supplements –
benefits due – increases (+ decreases) in life insurance technical reserves.
Four components are there in this measurement for the life insurance indus-
try. First is the amount of premiums earned. Second is the total premium
supplement, which is additional income earned by life insurers in returns on
investments, fees and other sources. The third component is the benefits (life
insurance claims) due in a given year. This consists of both the death and matu-
rity claims. Finally, the last component is the change in technical reserves. Data
on all these components are estimated using accounting data reported by insur-
ers to their regulators. The final result is a flow measure of insurance output.
 124 Survey of evidences
Income

Operating
income

Premium Investment
income income

Liquidity-
Risk-based
Direct based

High risk
Gross Liquid assets
assets

Net Low risk assets Illiquid assets

Tangibility-
Maturity-based
based
Reinsurance

Short-term
Physical assets
Ratios assets

Long-term Financial
assets assets

Figure 4.5 Conceptualising income output in the insurance literature

Source: Author’s analysis based on the extant literature.
 SNA measures
of output

Direct insurance Reinsurance Retrocession Others

Non-Life Life Social insurance

Premiums Premiums Premiums Premums

earned supplements earned supplements

Review of evidence on insurance output measurement 125

Paid during given Changes in
accounting period Benefits due Technical
reserves

Adjusted Claims
incurred

Technical
Expectations Accounting Reserves
method method

Adjusted
claims

Equalization
provisions

Cost method

Figure 4.6 Conceptualising insurance output within the SNA framework

Source: Author’s analysis based on the SNA, 2008 Manual.
126 Survey of evidences
As one may ascertain, the SNA uses a net premium measure of insurance out-
put wherein gross premiums are netted out for claims and reserves. This meas-
ure is used for sectoral analysis and provides the recommended method for
national economies. The same subtleties of using this measurement will differ
depending on the economy. Advanced economies such as the US have very
well-developed granular data on the financial service industry.29
Similarly, SNA recommends output measurement for the nonlife insurance
industry as Total premiums earned + premium supplements – less adjusted
claims incurred. This measure is also based on the framework of the net pre-
mium, where the gross premium amount is adjusted for supplementary income
and incurred claims. It may also be noted that the SNA recommends using
incurred claims for the nonlife segment instead of only the claims paid. As
Figure 4.6 shows, the incurred claims may be estimated using the expectations
method, accounting method or the cost method. The debate has been ongoing
on the correct method for estimating incurred claims. Incurred claims consist
of paid plus changes in reserves. The reserve component requires forecast-
ing, and thus the methods mentioned earlier are needed. Advanced economies
can utilise the accounting method relatively efficiently due to well-established
data collection conventions. Emerging economies generally have to choose
between the cost and expectations methods. The latter is an ad hoc method
that extrapolates past patterns into the future. Given the advances in forecast-
ing models, the expectations method might provide valuable estimates of the
reserves and expected claims.

4.5. Thematic issues in insurance output measurement

Given the literature review as elaborated above, some important theoretical and
empirical themes arise that may be presented as follows.

Market structure, organisational structure and performance estimation

It is well-recognised in neoclassical production theory that the market structure
in which output is traded has direct implications for the quantity of that output
and the method to measure it. Each insurance industry has a broadly specific
market structure and showcases different degrees of integration of the firms and
different intensities of competition among the firms. The way the industrial
and firm-level production functions are conceptualised depends fundamentally
on the industry’s market structure. Aspects of a firm-level production func-
tion include substitutability among the inputs, price elasticity of demand for
output, marginal rates of transformation and substitutions, exogenous factors
that may affect the production function and others profoundly shaped by the
market structure industry. This aspect has received limited empirical attention
in the insurance literature. Firm- and industry-level production functions are
generally estimated by assuming a specific market structure a priori. While this
 Review of evidence on insurance output measurement 127
may provide empirical ease, it may be costing theoretical inconsistencies. The
assumption of price-taking behaviour in inputs and even outputs markets, for
example, has been common in insurance performance studies using trans-log
cost and revenue functions.30 Such an assumption is generally imposed on the
context via stylised facts about the industry or the analyst’s subjective under-
standing. It may so occur that if such assumptions were to be relaxed, the esti-
mates could improve. This is an empirical question and requires more effort in
the coming periods. Recognising market structure implies taking cognisance
of the industry’s structure and allowing that structure to shape the measure of
output used in the analysis.
The implications of competition for output measurement have received
some attention in the vast literature. Insurers have been found engaging in both
price and non-price competition. Price competition will imply that output
measures with high price elasticity of demand should be used after controlling
price movements. Otherwise, excessive variability might enter into the data
that could wrongly be associated with non-price factors, and the estimates of
performance indicators too might be biased. As an illustration, under a con-
tinuous increase in insurance premium (the price of insurance products), the
amount of total premium collected by insurers might show a large fall. Suppose
these price movements were not accounted for in the production function. In
that case, underestimating productivity is dangerous because the output vari-
able (amount of total premium) is not adjusted for price movements. In case
the price elasticity of demand for the chosen output variable is low, i.e., it is
highly inelastic, then one would be in a better position to analyse such output
variables within a production function approach, as price considerations will
not be a constraint. However, demand’s degree of price elasticity is an empiri-
cal question and requires estimation rather than any a priori assumptions. Such
evidence is limited in the context of both advanced and emerging economies.
Regulators are keenly observing the market behaviour of insurers to prevent
collusive behaviour. However, when collusive behaviour does occur, output
determination has to account for it. The output measurement in insurance is
undertaken mainly by defining each firm (in the firm- and industry-level stud-
ies) as independent output producers. Independence is conceptualised as a lack
of strong interdependency between the firms’ production functions. Hence, all
the firms are assumed to be functioning under uncertain environments, and
all the firms are assumed to be behaving as per the conditions prevailing in the
market. However, at times, this may not be true. An example of this is Turkey,
“where in 2000 the Antitrust Authority imposed a conspicuous fine on 39 com-
panies for their anti-competitive behaviour deriving from ad-hoc information
exchange from 1994 onwards” (Coccorese, 2010, p. 55). Kühn and Vives (1995)
argue that “it is hard to find plausible business reasons, other than collusion, that
might justify the exchange of individual price and output data” (p. 115). The
theorisation of production functions of these colluding firms as independent
would have yielded incorrect results due to simultaneity bias in determining
128 Survey of evidences
their performance with separate and independent production functions. Under
such situations, the output of the independent-appearing firms is actually col-
lusive, and conceptual problems would arise, as examined by Coccorese (2010).
First, at the industry level, the behaviour of colluding firms will not be inde-
pendent of each other and will have a high degree of correlation. Comparisons
among these firms might produce estimates with solid relationships, distorting
the findings on competition and efficiency and productivity comparisons. Sec-
ond, at the aggregate level, aggregation of the firm-level outputs into a single
measure of industry-level output requires a well-defined aggregation function
with theoretically correct aggregation weights. In literature, these weights are
generally the price of each output. Given that price decisions might be collusive,
they are expected to be highly correlated. Under such a scenario, these observed
market prices or revenue shares will not provide much meaningful information
on the unique weight that should be assigned to each output in the aggregate
output variable. Using such weighting factors that are highly correlated and
not independently determined will produce biased results. The aggregation will
become all the more challenging in this case. Third, when collusion exists, price
signals become inefficient and may not reflect the actual market conditions, at
least for the agents other than the colluding firms. Their use as aggregation fac-
tors would probably be rendered meaningless. The fourth and last issue is that
aggregation weights such as revenue shares might also reflect cost shares due to
interdependent production across the colluding firms when collusion occurs.
While revenue shares are suitable for output aggregation, cost shares are used
for inputs aggregation. If both were highly correlated (positively or negatively),
the aggregated outputs and inputs would not represent any unique information.
The estimated total input and total output variables shall reflect similar kinds of
information rather than information specific to the product and factor markets.
An advantage of this is that either could be used for both outputs and inputs
aggregation. However, such an approach in all likelihoods, will be deemed theo-
retically incorrect.
Another dimension in this context is the role of organisational structure in
insurance performance analysis. One school of thought disregards the impor-
tance of organisational structure in determining the performance of insurers.
The best consensus of this school of thought may be located in the so-called
‘Demsetz and Lehn hypothesis’, which “states that organisational form is not
important for determination of technical efficiency, in view of the fact that
regulators act as monitors” (Fukuyama, 1997, p. 490). Evidence does not show
consensus on this, and this seems to hold in some advanced economies, but
the findings for EMEs are weak on this account. This may become an agenda
for future research. Cummins et al. (2004) have elaborated on four different
hypotheses on the importance of organisational structure for the performance
of insurance firms. These are the efficient structure hypothesis, expense prefer-
ence hypothesis, managerial discretion hypothesis and the maturity hypothesis.
Each of them is empirically tested by them in their 2004 study.
 Review of evidence on insurance output measurement 129
Output mix and output measures
Firm size can be a factor that affects the output mix and thus can affect the
choice of output in insurance. Smaller firms might be more prone to entering
into more contracts and undertaking a relatively larger amount of reinsurance
to manage risks. In this case, measures such as the number of policies (life or
nonlife or combined) may be useful proxies for output that can help capture
the extent to which the small insurance firms can manage and allocate risks
efficiently (via diversification and reinsurance). However, the same may not be
accurate for larger firms, and thus the output mix can be very different in their
case. For example, larger insurance firms may achieve a lower price elasticity of
demand and thus may be able to charge higher prices than the smaller firms.
In this case, dependency on diversification of contracts and reinsurance might
not be the sole ways to manage risks efficiently. In such a case, one would find
value measures of insurance output as better than volume measures. Moreover,
a common criticism of volume measures in insurance is that “the number of
policies does not capture meaningful differences in insurers’ services” (Weiss,
1990, p. 24). However, this is true at the level of firms and groups of firms. At
an industrial level (aggregate), such differences can be suppressed and averaged
out to obtain a larger macroeconomic perspective.

Regulation and output measurement

Does regulation affect insurance output? Are there any implications of regula-
tory norms and interventions for measuring insurance output? Such concerns
have been raised in the literature. An early attempt in this line of thought was
in terms of the so-called ‘Averch-Johnson hypothesis’. The “Averch-Johnson
hypothesis” suggests links between the degree of regulation of firms and their
economic behaviour. It was initially propounded by Averch and Johnson (1962)
to analyse the inefficient capital accumulation behaviour of electricity utility
firms due to the regulated rate of return (as against market-oriented rate of
return on invested capital), which the Government guarantees. However, this
hypothesis assumes single output firms and thus cannot be directly applied to
banking and insurance firms which are inherently multi-output. Multi-output
firms may face regulations on some outputs or different regulations on different
outputs. This can make it difficult to disentangle the distortions created by such
differential regulations and hence makes it difficult to estimate a pure market-
oriented output measure.
Furthermore, “regulation can affect firms’ incentives and costs and thereby
influence productivity growth” (Weiss, 1990, p. 15). Similarly, it is possible that
the regulated rate of return (e.g. premium rate charged by insurers as regulated
by the regulator) may not be greater than the price of capital invested in the
business. The Averch-Johnson effect may be in the form of under accumula-
tion/utilisation of resources. Many insurance regulatory bodies stress the “fair”
130 Survey of evidences
rate of premium, which might result in equalisation of the rate of return on
capital and the cost of capital.
The nature of regulation and its underlying motivation can affect the output
of banking and insurance firms and hence can make it difficult to estimate the
market-oriented output. If public interests drive the regulations, the market
orientation of the output variables will be weak and exogenous (not caused by
market forces). However, if private interests dominate, then the market orien-
tation of the output measures is higher even when the interests of the industry
giants drive regulations. One can then use various models of market structures
such as oligopoly and monopoly to measure and analyse the output behaviour
of such firms. As an illustration, consider the standard solvency regulations:
capital adequacy for banks and reserve determination norms for insurers. Both
are generally assumed to be motivated by public interests in the EMEs. How-
ever, the efficacy of this belief is questionable and hardly any empirical pieces
of evidence exist that can corroborate such an a priori assumption.

Insurance output measurement in advanced versus emerging economies

Compared to the emerging economies, advanced economies have several data
related advantages. One of them is the availability of detailed disaggregated
accounting data at the firm level for the insurance industry. As Cummins and
Xie (2009, p. 135) note: “Having detailed operating data enables us to provide
more accurate estimates of inputs and outputs and to account more fully for
heterogeneity among the firms in the industry”. This is a significant constraint
for analysts focusing on emerging economies. Another advantage in analysing
advanced economies is the availability of price indexes for the insurance indus-
try and different lines of insurance. This, too, is missing in the case of EMEs.

Deflation of nominal output in insurance

The most frequently used price indexes for deflation of insurance firms and
industry nominal output are CPI, WPI, insurance-specific price indexes (such
as in Carrington et al., 2011) and the GDP deflator. Deflation is a necessary
procedure for lending data comparable across time. This is necessary as far as
time series and panel data are concerned. For cross-section data, one may use
nominal variables. As expected, aggregate price indexes have been employed
due to a lack of reliable and rigorous insurance-specific price indexes. To this
extent, CPI has been the most frequent choice in insurance literature, followed
by the GDP deflator. However, such aggregate deflation methods are not jus-
tifiable if one is interested in deriving the quantity changes of the concerned
variable. They might help establish the comparability of data from different
regions or countries. The use of such measures implicitly assumes a strong
correlation between the aggregate price level and the price levels in the insur-
ance industry. This generally does not hold true. Deflation of input data is at
times done using input-specific price indexes. These are mainly seen in the
 Review of evidence on insurance output measurement 131
case of the US, where firm-level disaggregated data are reliably available. In
their study, Xie et al. (2011) apply such an approach to the deflation of inputs.
A similar approach is also possible in the case of outputs, but hardly any study
has adopted this framework. Either the CPI or the aggregate GDP deflator has
been used.

Taxation and insurance output measurement

Most countries levy taxes on premium amounts. These should ideally be
accounted for in the measurement of output for the industry as a whole and
individual firms. Taxes are levied in indirect taxes as a proportion of the under-
writing premium amount quoted by the insurer. Suppose financial statement
data are used to determine the total premium amount. In that case, the value
of the tax amount could be calculated, and the same be adjusted for (removed
from) the observed data if already included. This is because the amount of tax
on the premium is not an insurer’s income, nor does it represent any value-
addition by the insurer.
The following section presents the appendix on the evolution of output
measurement in insurance literature from 1980 to 2022.
Appendix
Evolution of insurance output measurement in the extant wisdom from
1980 to 2022

No. Citation Issue Output Specification Flow or Nominal or Real Underlying

Stock Approach

1 Daly and Rao Total Factor Productivity, Amount of Premium Flow, Real Production
(1985) Economies of Scale and Stock approach
Technical Progress
2 Weiss (1990) Total Factor Productivity Real expected losses proxied by relevant Flow Real (Constant Intermediation
Growth expenses Price) approach
3 Diacon (1990) Insurance Output Ideal output measure is number of Flow Real (Volume) Production
Measurement policies sold or a similar volume
measure
4 Hornstein and Insurance Output Traditional measure: Net premiums Flow Real (Constant National income
Prescott (1991) Measurement (Gross premiums net of claims paid); Price)
Suggestion: product of price and
quantity of insurance contracts using
hedonic pricing model
5 Weiss (1991) Output, Input and TFP Incurred losses and loss reserves Flow Real Intermediation
estimation approach
6 Wolff (1991) Labour Productivity Gross National Product of insurance Flow Real Production
growth and TFP growth industry approach
7 Cummins and Cost Efficiency Direct Premiums; Net Premiums; Losses Flow Nominal Production
VanDerhei incurred
(1992)
8 Fecher et al. Technical efficiency Gross Premium; Financial Returns Stock Nominal Production
(1993) for life and nonlife approach
insurance industries
9 Kumbhakar and Labour-use efficiency Individual group of services provided by Flow Real (Volume) Production
Hjalmarsson (Labour Productivity social insurance offices approach
(1995) or Technical Efficiency
in use of labour input)
of Swedish local social
insurance offices
10 Fukuyama (1997) Productive Effiiency and Reserves; Loans Stock Nominal Intermediation
TFP growth approach
11 Toivanen (1997) Cost Efficiency Working time; Number of accident; Flow Real (Volume); Production
Premium Real (Constant approach
Price)
Oulton (1998) Labour productivity of Aggregate Value-added Flow Nominal Production

Review of evidence on insurance output measurement 133

12
aggregate economy and approach
individual industries
including insurance
industry
13 Vuorinen et al. Partial and Total Factor Number of Applications; Turnaround Flow Real (Volume) Production
(1998) Productivity time Approach
14 Cummins and Zi Technical Efficiency Incurred Benefits; Additions to Reserves Flow Real (Constant Production
(1998) Price)
15 Sherwood (1999) Insurance output Difference between amount of Flow Nominal, Real National income
measurement premiums (adjusted for rate of return (Constant Price)
on investments) and total claims paid
16 Acemoglu and Effects of change in Amount of claims paid Flow Nominal Insurance
Shimer (2000) Unemployment as welfare
Insurance level on enhancing
aggregate output level service
17 Fukuyama and Technical efficiency Reserves; Loans; Investments Flow, Real (Constant Production
Weber (2001) of nonlife insurance Stock Price) Approach
industries with different
efficiency measures

(Continued)
(Continued)

134 Survey of evidences

No. Citation Issue Output Specification Flow or Nominal or Real Underlying
Stock Approach
18 Cummins and Capital Utilisation in Present value of losses incurred; Amount Stock Real (Constant Production
Nini (2002) property liability of invested assets Price) approach
insurance
19 Mahlberg and Technical Efficiency and Sum of expenses on claims, change in Flow Nominal Production
Url (2003) Malmquist Total Factor reserves and returned premiums
Productivity
20 Ennsfellner et al. Technical Efficiency Incurred benefits net of reinsurance; Both Nominal Intermediation
(2004) Changes in reserves net of
reinsurance; Total invested assets;
Claims incurred net of reinsurance
21 Cummins et al. Technical, cost, and Assets; Premium; Income (Total, Life, Both Real (Constant Value Added
(2004) revenue frontiers Nonlife separately) Price)
22 Arvanitis (2005) Firm productivity Amount of Sales Flow Nominal Production
Approach
23 Barros et al. Technical Efficiency and Claims and Profits Stock Real (Constant Intermediation
(2005) Malmquist Total Factor Price)
Productivity
24 Brockett et al. Technical Efficiency Solvency; Liquid Assets/Liabilities; ROA Ratio of Nominal Intermediation
(2005) Flow to approach
Stock;
Ratio of
Stock to
Stock
25 Choi and Weiss Cost and Revenue Losses; Assets Flow, Real (Constant Value Added
(2005) Efficiencies Stock Price)
26 Jeng and Lai Technical and Cost Policies; Assets Stock Real (Constant Value Added
(2005) Efficiencies Price)
27 Tone and Sahoo Cost Efficiency and Claims Settled, Ratio of Liquid Assets to Flow, NA NA
(2005) Returns to Scale Liabilities Ratio of
Stock to
Stock
28 Bertschek et al. Labour Productivity Sum insured Stock Nominal Production
(2006) of Service industries approach
including insurance
29 Cummins and Cost, technical, and Losses; Reserves; Assets Flow and Real (Constant Value Added
Rubio-Misas allocative efficiency, and Stock Price)
(2006) Malmquist total factor
productivity (TFP)
change
30 Jeng et al. (2007) Technical, Allocative and Benefits paid; IRIS indicators, Principal Flow Real (Constant Value Added and
Cost Efficiencies Components Analysis-based Price) Intermediation
components
Sinha (2007a) Technical Efficiency Benefits paid, Operating Income; Net Flow NA NA

Review of evidence on insurance output measurement 135

31
Premium
32 Sinha (2007b) Technical Efficiency and Net Premium; Gross Income Flow NA NA
Malmquist Total Factor
Productivity
33 Barros et al. Total Factor Productivity Profit and loss account; net premiums; Flow Real Production
(2008) settled claims; outstanding claims; approach
investment income
34 Cummins and Total Factor Productivity, Expected Losses; Assets Flow Real (Constant Production
Xie (2008) Technical Efficiency and Price) approach
Scale Efficiency
35 Maré (2008) Labour Productivity Value added Flow Nominal Production
approach
36 Chen et al. Economies of scale and Amount of Premium Flow Real (Constant Intermediation
(2008) scope Price) approach
37 ECB (2008) Underwriting cycle in EU Amount of Premium Flow NA NA
38 Bikker and Cost X-inefficiency and Premiums; Policies; Insured Capital; Flow, Nominal Production
Leuvensteijn Economies of Scale Insured Annuities Stock
(2008)
39 Fenn et al. (2008) Cost Efficiency Claims Flow Real (Constant Production
Price)

(Continued)
(Continued)

No. Citation Issue Output Specification Flow or Nominal or Real Underlying

136 Survey of evidences

Stock Approach
40 Kao and Hwang Technical Efficiency Profits (Underwriting and Investment); Flow Nominal Production
(2008) Premiums (Direct and Reinsurance)
41 Weiss and Choi Cost and Revenue Losses; Assets Flow, Nominal Production
(2008) Efficiencies Stock
42 Cummins et al. Cost Efficiency Amount of assets; Amount of Asset Both Real (Constant Both
(2009) Turnover; Return on Investment (all Price)
three for Financial Intermediation);
Amount of Surplus (for Risk
Management function); Amount of
incurred losses (for Risk Management
function)
43 Cummins and Cost and Revenue Incurred Losses and Invested Assets Stock Real (Constant Production
Xie (2009) Efficiencies Price) approach
44 Eling and Luhnen Underwriting cycle Amount of Premium; Amount of Losses Flow Nominal NA
(2009) incurred; Amount of developed
incurred losses
45 Jawadi et al. Underwriting Cycle Direct Premiums written Flow Nominal NA
(2009)
46 Luhnen (2009) Technical Efficiency and Amount of Claims; Amount of Invested Flow, Real (Constant Production
Malmquist Total Factor Assets Stock Price)
Productivity
47 Garg and Deepti Total Factor Productivity Premium; Operating income Flow Nominal Production
(2009)
48 Segovia-Gonzalez Efficiency Gross and Net Premiums Flow Nominal Production
et al. (2009)
49 Kasman and Total Factor Productivity Losses paid; Benefits paid Flow Nominal Production
Turgutlu
(2009b)
50 Trufin et al. Underwriting Cycle Inverse loss ratio NA NA NA
(2009)
51 Cerchiara and NA NA NA NA NA
Lamantia
(2009)
52 Kasman and Cost Efficiency and Losses; Benefits; Assets Flow, Real (Constant Value Added
Turgutlu Economies of Scale Stock Price)
(2009a)
53 Mahlberg and Technical Efficiency and Amount of Claims (Losses) incurred; Flow Real (Constant Intermediation
Url (2010) Malmquist Total Factor Amount of Technical (Actuarial) Price)
Productivity Reserves; Net Return on Financial
Assets
54 Trigo-Gamarra Cost and Profit Incurred Benefits/Claims paid; Flow Real (Constant Production
and Growitsch Efficiencies Premiums Price)
(2010)

Review of evidence on insurance output measurement 137

55 Eling and Luhnen Cost and Technical Claims [Nonlife claims + additions to Flow Real (Constant Production
(2010a) Efficiency reserves]; Benefits paid [Life benefits Price)
+ additions to reserves]; Investments
56 Lin et al. (2011) Technical Efficiency and Premiums and Financial Income Flow Nominal Production
Malmquist Total Factor
Productivity
57 Eslake (2010) Labour Productivity Gross Value Added as per SNA Flow Nominal Production
58 Bates et al. (2010) Technical Efficiency Percentage of people reporting being Stock Real (Volume) NA
healthier
59 Kader et al. Cost Efficiency Premiums Flow Nominal Production
(2010)
60 Cummins et al. Economies of scope Incurred Benefits (claims paid); Losses; Flow, Real (Constant Production
(2010) Assets Stock Price)
61 Nektarios and Malmquist Total Factor Invested assets, incurred losses, Flow, Real (Constant Production
Barros (2010) Productivity reinsurance reserves, own reserves Stock Price)
62 Lei and Schmit Financial Performance Risk-adjusted ROA Ratio of Nominal NA
(2010) Flow to
Stock
63 Erhemjamts and Technical and cost Incurred benefits Flow Real (Constant Value Added
Leverty (2010) efficiency Price)

(Continued)
(Continued)

138 Survey of evidences

No. Citation Issue Output Specification Flow or Nominal or Real Underlying
Stock Approach
64 Leverty and Pure Technical, Allocative, Value-added approach: Present Value of Flow, Real (Constant Value Added and
Grace (2010) Cost, Revenue and Real Losses Incurred; Average Real Stock Price) Flow
Scale Efficiencies Invested Assets; Flow Approach: Rate and
of return on investments, Liquid Ratio of
assets, Solvency score Flow to
Stock
65 Xie (2010) Technical, Allocative, Losses; Assets Flow and Real (Constant Value Added
Cost, Revenue and Stock Price)
Scale Efficiencies
66 Chakraborty and Total Factor Productivity Net Premium, Number of Products Flow N NA
Sinha (2010) Launched
67 Owusu-Ansah Technical Net Premium, Claims Settled, Income Flow NA NA
et al. (2010) on Investment
68 Carrington et al. Partial and Total Factor Amount of Premium Flow Real (Constant Production
(2011) Productivity Price)
69 He et al. (2011) Cost and Revenue Incurred Losses; Invested Assets Flow, Real (Constant Production
Efficiencies Stock Price)
70 Pottier (2011) Cost, Revenue and Profit Claims Incurred; Invested Assets Flow, Nominal Production
Efficiencies Stock
71 Berry-Stölzle Cost and Revenue Incurred Losses; Invested Assets Flow, Real Production
et al. (2011) Efficiencies Stock
72 Xie et al. (2011) Cost and Technical Incurred Benefits; Invested Assets Flow, Real (Constant Production
Efficiency Stock Price)
73 Eckles and Pottier Cost Efficiency Incurred Losses; Invested Assets Flow, Not Clear Production
(2011) Stock
74 Biener and Eling Technical, Allocative and Incurred Benefits; Total Investments; Flow; Real (Constant Value Added
(2011) Scale Efficiencies Insured Population proportion Stock Price)
75 Bikker and Cost Efficiency Incurred Losses; Premiums; Investments Flow, Real (Constant Production
Gorter (2011) Stock Price)
76 Chen et al. Technical, Allocative and Losses; Assets; Surplus; Capitalisation; Flow, Real (Constant Value Added;
(2011) Cost Efficiencies Yield; Liquid Assets Stock Price) Intermediation
and
Ratio of
Flow to
Stock
77 Choi and Cost and Cost Scale Losses; Assets Flow, Real (Constant Value Added
Elyasiani (2011) Efficiencies Stock Price)
78 Huang et al. Technical and Cost Losses; Assets Flow, Real (Constant Value Added
(2011) Efficiencies Stock Price)
79 Zhong and Sun Cost Efficiency Assets; Claims; Reserves Flow, Real (Constant Production
(2011) Stock Price)

Review of evidence on insurance output measurement 139

80 Wu and Zeng Technical and Scale Premium; Profits; Reserves; Investment Flow Nominal Value Added
(2011) Efficiencies Income
81 Bawa and Technical and Scale Net Premiums Flow Nominal Production
Ruchita (2011) Efficiencies
82 Saad and Idris Malmquist Total Factor Net Premium, Net Income on Flow Nominal Production
(2011) Productivity Investment
83 Singh and Rajpal Technical Efficiency Net Premium, Income on Investment, Flow Nominal Production
(2011) Claims
84 Yaisawarng et al. Technical Efficiency, Number of policies issued; Total Flow, Real (Constant Hybrid of
(2012) Economies of Scale invested assets Stock Price) Production and
and Total Factor Intermediary
Productivity approaches
85 Chakraborty Malmquist Total Factor Premium; Number of Products Flow Real (Constant Production
and Sengupta Productivty Price)
(2016)
86 Jackson (2012) Malmquist Total Factor Claims Incurred; Invested Assets Flow, Real (Constant Value Added
Productivty Stock Price)
87 Leverty and Cost, revenue, scale, Losses; Assets Flow, Real (Constant Value Added
Grace (2012) technical, and allocative Stock Price)
efficiencies

(Continued)
(Continued)

No. Citation Issue Output Specification Flow or Nominal or Real Underlying

140 Survey of evidences

Stock Approach
88 Al-Amri et al. Technical Efficiency and Premiums; Investment Income Flow Nominal Value Added
(2012) Malmquist Total Factor
Productivity
89 Chaiyawat (2014) Technical Efficiency Losses; Assets Flow, Nominal Value Added
Stock
90 Cummins and Technical Efficiency, Present value of losses incurred; Amount Stock Real (Constant Production
Xie (2013) Economies of Scale of invested assets Price) approach
and Total Factor
Productivity
91 Chakraborty Technical Efficiency and Benefits paid; Income from Investment Flow Real (Constant Value Added
et al. (2013) Malmquist Total Factor Price)
Productivity
92 Sinha and Khan Malmquist Total Factor Gross Investment income and Net Flow Nominal Production
(2013) Productivty Premium
93 Sinha (2013) Cost Efficiency Sum Assured; Incremental Assets Under Flow Nominal Production
Management
94 Gaganis et al. Cost and Profit Premiums; Invested Assets Flow, Nominal Production
(2013) Efficiencies Stock
95 Cummins and Review of Evidences Life insurance: premiums; sum of NA NA NA
Weiss (2013) incurred benefits and additions to
reserves; average invested assets;
Nonlife: present value of real losses
incurred;
96 Oulton and Labour Productivity Gross Domestic Product for financial Flow Not Clear Macro
Sebastiá-Barriel services
(2013)
97 Sauian et al. Labour Productivity Gross Domestic Product for financial Flow Not Clear Macro
(2013) services
98 Akotey et al. Profitability Investment income; underwriting profit Flow Nominal NA
(2013)
99 Olaosebikan Profitability Net Profit Flow Nominal NA
(2013)
100 Vencappa et al. Total Factor Productivity Premiums; Claims Flow Real (Constant Production
(2013) Growth Price)
101 Manikowski and Underwriting Cycle Claims, Premiums, Loss Ratios, Rates- Flow Nominal NA
Weiss (2013) on-Line, Capacity
102 Lee (2013) Insurance Market Premiums NA NA NA
Development
103 Dutta (2013) Technical Efficiency Premiums Earned; Income from Flow NA Reverse
Investments Intermediation
104 Bawa and Profitability Return on Assets Ratio of Nominal NA
Chattha (2013) Flow to
Stock

Review of evidence on insurance output measurement 141

105 Chen and Technical and Cost Benefits paid; Invested Assets Flow, Real (Constant Value Added
McNamara Efficiencies Stock Price)
(2014)
106 Rahmani et al. Financial Performance Earnings per share; Current ratio; Total Flow, and Nominal Production
(2014) (Efficiency) income; Operating income Ratio of
Stock to
Stock
107 Porrini (2014) Development of Insurance Premiums, Claims Flow Nominal NA
Market
108 Jarraya and Bouri Profit Efficiency Claims; Investments Flow Nominal Value Added
(2014)
109 Lorson and Sales Success Premiums Flow Nominal NA
Wagner (2014)
110 Grmanová (2014) Technical Efficiency Premiums Flow Nominal NA
111 Ma et al. (2014) Cost, Revenue and Losses; Assets Flow, Unclear as not Value added
Malmquist Total Factor Stock mentioned
Productivity clearly
112 Kulvinskienė and Labour Productivity Number of meetings, Number of Flow Real (Volume) Production
Apčelauskaitė concluded contracts
(2014)

(Continued)
(Continued)

No. Citation Issue Output Specification Flow or Nominal or Real Underlying

142 Survey of evidences

Stock Approach
113 Javaheri (2014) Malmquist Total Factor Profits; Claims Flow Nominal Production
Productivity
114 Mandal and Technical Efficiency Net Premium, Claims Processed Flow NA NA
Dastidar (2014)
115 Nandi (2014) NA Net Premium, Net Benefits paid Flow NA NA
116 Grmanová (2014) Technical Efficiency Premiums Flow Nominal Production
117 Rao and Technical Efficiency and Net premium; Investment income Flow Real (Constant Production
Venkateswarlu Malmquist Total Factor Price)
(2015) Productivity
118 Ai et al. (2015) Technical Efficiency Premiums; Policies Flow Real (Constant Production
Rate based)
119 Wasseja and Technical Efficiency and Incurred benefits; Assets Flow, Nominal Value Added
Mwenda Malmquist Total Factor Stock
(2015) Productivity
120 Biener et al. Cost, Technical, Losses; Benefits; Assets Flow, Real (Constant Value Added
(2015) Allocative, Scale, Stock Price)
Revenue and
Malmquist Total Factor
Productivity
121 Alhassan and Technical Efficiency and Premiums; Claims; Investment income Flow Real (Constant Value Added
Biekpe (2015) Malmquist Total Factor Price)
Productivity
122 Grmanová (2015) Technical Efficiency Premiums; Investment Income Flow Nominal Intermediation
123 Cole et al. (2015) Profitability Premiums; Losses incurred Flow Nominal NA
124 Bruneau and Underwriting Cycle Premiums written; Losses incurred Flow Nominal NA
Sghaier (2015)
125 Bawa and Bhagat Technical Efficiency Net Premiums, Number of Policies Sold Flow Nominal Production
(2015)
126 Janjua and Akmal Technical Efficiency and Net Premium, Investment Income, Net Flow Nominal Production
(2015) Profitability Claims
127 Sinha (2015) Technical Efficiency Premiums Collected, Sum Assured Flow NA NA
128 Noreen and Technical, Allocative, Net premium; Invested Assets Flow, Nominal Production
Ahmad (2016) Cost Efficiency, and Stock
Malmquist Total Factor
Productivity
129 Fier and Pooser Advertising Efficiency Premiums Flow Nominal NA
(2016)
130 Viverita et al. Cost Efficiency and Claims; Investments Flow, Real (Constant Value Added
(2016) Malmquist Total Factor Stock Price)
Productivity
131 Grmanová and Technical Efficiency Premiums; Investment Income Flow Nominal Production
Hošták (2016)

Review of evidence on insurance output measurement 143

132 Grmanová (2016) Technical Efficiency Premiums; Investment Income Flow Nominal Production
133 Suárez-Álvarez Technical Efficiency Premiums; Reserves; Assets Flow, Nominal Value Added
et al. (2016) Stock
134 Chakraborty Technical and Scale Net Premiums; Investment Income Flow Real (Constant Production
(2016a) Efficiencies and Price)
Malmquist Total Factor
Productivity
135 Cummins and Technical, Scale, Cost, Losses; Invested Assets Flow. Real (Constant Value Added
Xie (2016) Revenue and Profit Stock Price)
efficiencies; and
Malmquist Total Factor
Productivity
136 Chakraborty Financial Performance Financial Soundness Indicators Ratio of Nominal NA
(2016b) Flow to
Stock,
Stock to
Stock
137 Sinha (2016a) Technical Efficiency Benefits paid; Asset Under Management Flow, Real (Constant Value Added
Stock Price)
(AUM)

(Continued)
(Continued)

144 Survey of evidences

No. Citation Issue Output Specification Flow or Nominal or Real Underlying
Stock Approach
138 Fernandes (2016) Technical and Scale Premiums; Benefits paid Flow Nominal Value Added
Efficiencies
139 Rao (2016) Financial Performance Gross Premium; Earned Premium; Flow Nominal NA
Profit
140 Zhang (2016) Technical Efficiency Net benefits plus additions to reserves; Flow, Real (Constant Value Added
average invested assets Stock Price)
141 Sinha (2016b) Malmquist Total Factor NA NA NA NA
Productivity
142 Parida and Technical Efficiency and Benefits; Liquidity ratio Flow, Real (Constant Value Added
Acharya (2017) Malmquist Total Factor Ratio of Price)
Productivity Stock to
Stock
143 Ferro and León Technical Efficiency Losses Flow Real (Constant Production
(2017) currency)
144 Nourani et al. Technical Efficiency Premiums; Claims; Investment income; Flow Nominal Intermediation
(2018) Net profit approach
145 Ksenija (2017) Financial Performance Equity and reserves; Business assets; Flow Nominal NA
Provision and liabilities; Financial
incomes; Cost of insurance
146 Sinha (2017) Technical Efficiency Operating income; Investments, Benefits Flow, Nominal Production
paid Stock
147 Grmanová and Technical Efficiency and Premiums earned; Investment income Flow Nominal Production
Strunz (2017) Profitability
148 Pal et al. (2017) Underwriting Cycle Losses paid; Loss Adjustment expenses; Flow Nominal NA
Premium income
149 Chakraborty and Malmquist Total Factor Net premium written; Income from Flow Nominal Production
Harper (2017) Productivity investment
150 Wise (2017) Technical Efficiency Technical reserves and claims; Premiums Flow NA Production, Value
and investment income Added
151 Muthulakshmi Financial Performance Claims; Investment income; Net Ratio of Nominal NA
(2018) Premiums; Invested Assets flow to
flow,
stock to
stock
and flow
to stock
152 Uddin et al. Degree of Competition Gross premium; total assets; gross claims Flow, Nominal NA
(2018) Stock
153 Chakraborty Technical and Scale Net Premiums Earned; Income from Flow Real (Constant Production
(2018) Efficiency, and Investments Price)
Malmquist Total Factor

Review of evidence on insurance output measurement 145

Productivity
154 Reyna and Cost Efficiency Number of Policies; Number of Claims Flow Real (Volume) Production
Fuentes (2018)
155 Grmanová and Technical Efficiency Premiums earned; income from financial Flow Nominal Production
Pukała (2018) investments
156 Anandarao et al. Technical Efficiency Total Profits (excluding investment Flow Nominal Intermediation
(2018) profits); Investment Profits
157 Wise (2018) Technical Efficiency Reserves; Claims; Premiums; Investment Flow NA NA
Income; Premiums are the best
measure of insurance output as per the
authors
158 Adachi (2018a) Cost Efficiency Policyholders Flow, Real (Volume) Production
Stock
159 Adachi (2018b) Economies of Scale Number of Primary insured persons Flow, Real (Volume) Production
Stock
160 Pawirosumarto Labour Productivity NA NA Real (Volume) Production
and Iriani
(2018)
161 Born and Sirmans Underwriting Premiums; Claims Flow Nominal Production
(2018) performance
(Continued)
(Continued)

146 Survey of evidences

No. Citation Issue Output Specification Flow or Nominal or Real Underlying
Stock Approach
162 Alshammari et al. Cost Efficiency Claims; Assets Flow, Nominal Production
(2019) Stock
163 Taib et al. (2018) Technical Efficiency and Amount of premium income; net Flow Nominal Production
Malmquist Total Factor investment income
Productivity
164 Krupa et al. Investment Management Return on investments Ratio of Nominal Intermediation
(2019) Efficiency Flow to
Stock
165 Yakob et al. Risk Management and Benefits paid; Investment Returns Flow, Nominal Intermediation
(2014) Investment Management Ratio of
Efficiencies Flow to
Stock
166 Zhu (2019) Technical Efficiency Premiums Flow Nominal Production
167 Yang and Li Technical Efficiency and Profitability; Operating Ability; Flow, Nominal Production
(2019) Malmquist Total Factor Innovation Ability; ROE Ratio of
Productivity Flow to
Stock
168 Oke and Abere Operational Efficiency Operating efficiency Flow Nominal Production
(2019)
169 Kozak (2019) Cost Efficiency Premiums; Assets Flow, Nominal Production
Stock
170 Sinha (2019a) Technical Efficiency, Benefits paid; Asset Under Management Flow, Nominal Flow and value
Malmquist Total Factor Stock added
Productivity and
Returns to Scale
171 Ilyas and Technical Efficiency, Net Claims; Invested Assets Flow, Real (Constant Value Added
Rajasekaran Malmquist Total Factor Stock Price)
(2019a) Productivity and
Returns to Scale
172 Ilyas and Total Factor Productivity Net Claims; Invested Assets Flow, Real (Constant Value Added
Rajasekaran Stock Price)
(2019b)
173 Hasan et al. Financial Performance Return on Asset (ROA); Return on Ratio of Nominal NA
(2019) Equity (ROE) Flow to
Stock
174 Lei (2019) Financial Performance Return on Equity Ratio of Nominal NA
Flow to
Stock
175 Sinha (2019b) Malmquist Total Factor Benefits paid; Asset under management Flow, Real (Constant Intermediation
Productivity Stock Price)
Almulhim (2019) Technical Efficiency Premiums earned; investment and Flow Nominal Production and

Review of evidence on insurance output measurement 147

176
management fee income Intermediation
177 Masud et al. Malmquist Total Factor Premium; Investment Income Flow Nominal Production
(2019) Productivity
178 Fusco and Porrini Performance Premiums Flow Nominal NA
(2020)
179 Marjanovic and Profitability Return on Asset Ratio of Nominal NA
Popovic (2020) Flow to
Stock
180 Naushad et al. Technical Efficiency Premium; Investment income Flow Nominal Production
(2020)
181 Zhengyin et al. Malmquist Total Factor Premium income; compensation Flow, Nominal Production
(2020) Productivity expenses; return on investment Ratio of
Flow to
Stock
182 Lei and Zhanxin Technical Efficiency and Premium income; investment income Flow Nominal Production
(2020) Malmquist Total Factor
Productivity
183 Sun (2020) Malmquist Total Factor Investment income Flow Nominal Production
Productivity

(Continued)
(Continued)

148 Survey of evidences

No. Citation Issue Output Specification Flow or Nominal or Real Underlying
Stock Approach
184 Eling et al. (2020) Malmquist Total Factor Losses; Invested Assets Flow, Real (Constant Value Added
Productivity Stock Price)
185 Majid et al. Malmquist Total Factor Premium; Net investment income Flow Nominal Production
(2017) Productivity
186 Klumpes (2022) Technical Efficiency Losses; Claims; Assets Flow Real (Constant Production
(Losses Price)
and
Claims
and
Stock
Note: Detailed review sheet with various dimensions of output and inputs measurement used in the literature above can be availed from the author on request.
Source: Author’s analysis of the literature.
 Review of evidence on insurance output measurement 149
Notes
1 Outreville (1998) elaborates on the history of the economic theory of insurance output.
2 Perhaps, the current work is the first of its kind to survey, summarize and organize a
large volume of literature on this issue into a single work and similar analysis of bank
output measurement.
3 As suggested by the traditional neoclassical production theory, only the essential labour
and physical capital inputs are included in the performance analysis that employs the
production approach (Berger & Humphrey, 1997). However, several papers have
expanded the traditional input set in banking and insurance literature to include meas-
ures of financial capital like debt and equity (Carrington et al., 2011).
4 The financial capital of insurers consists of two components – equity capital and debt
capital. Most of the studies either focus on equity capital per se or combine them into
single financial capital input (Yaisawarng et al., 2012). Combining them implies aggre-
gation, and the usual aggregation problems emerge.
5 “Insurers are financial intermediaries who borrow money from their policyholders in
premium payments and invest the funds raised in financial assets” (Cummins et al., 2009,
p. 145).
6 “The period between the date of the claim occurrence and the date of the claim pay-
ment depends on the type of insurance policy. Financial intermediation activities invest the
amount of premiums received until the claim payment date” (Cummins et al., 2009, p. 149).
7 The same is discussed in the next section.
8 Equity capital is a crucial variable in the production process of insurance firms. It reflects
the insolvency risk of the insurance firm and thus its ability to meet its ‘promises’.
Maintaining an optimal amount of equity capital has both benefits and problems. See
Cummins and Nini (2002) for more details. The authors suggest that insurers tend to
hold a safe or adequate amount of equity rather than an economically optimal level due
to the costs of holding equity capital. This can be a source of economic shocks if claims
are substantially higher than the expected level.
9 See Brockett et al. (2005) for a detailed analysis of issues involved in output choice under
the financial intermediation approach and also how that approach is superior to the
traditional production approach.
10 Also, see Cummins and Weiss (2013) for a review of the developments in inputs and
outputs specifications under the value-added school.
11 Ideally, this should be examined empirically before permitting any item into the output
vector of the insurance firm. However, many studies have used the value-added without
any solid empirical justification.
12 “However, proponents of the intermediary approach . . . suggest that the objective
of profit maximization associated with the value-added approach is inadequate” (Car-
rington et al., 2011).
13 See Cummins and Xie (2008, 2009, 2013, 2016) for empirical applications of the value-
added approach and various theoretical issues in insurance output measurement under
this framework.
14 For example: “Life insurers collect premiums and annuity considerations from custom-
ers and redistribute most of the funds to those policyholders who sustain losses (the risk
bearing/risk pooling service)” (Cummins & Zi, 1998, p. 135).
15 For example: “Funds are collected in advance of paying benefits and invested in assets
until claims are paid” (Cummins & Zi, 1998, p. 135).
16 Any of the studies have not directly used the concept of lapsed premium, but it has been
proposed here as one possible measure of a ‘bad output’. Policies that lapse generally
result in the insurance firms’ loss of future earnings. Hence, minimizing it may be a
rational objective of an insurer, particularly in the case of life insurance.
17 Claims or losses have generally been treated as an output in insurance literature. But
there have been arguments on treating them as an input. One of the earliest attempts in
150 Survey of evidences
this regard was by Müller (1981). Since then, several studies have theorised claims as an
input, though the literature has primarily treated it as an output.
18 “[O]ut of the 80 studies that use the value-added approach, 46 specify output as claims
(property liabilities)/benefits (life), whereas another 32 specify output as premiums/
sums insured; however, there is no clear trend over time as to whether either of the two
main proxies is preferred in the literature” (Gaganis et al., 2013, p. 431).
19 “The inclusion of the net change in reserves guarantees that the expected future loss
from underwriting risk is appropriately accounted for as output” (Mahlberg & Url,
2003, p. 824).
20 Value measures can be adjusted for reinsurance claims received by the insurance com-
panies to reflect the actual amount of direct losses borne by the insurers. However, the
ideal adjustment should be: [Claims received by the insurer + Reinsurance premiums
paid by the insurer – Reinsurance claims]. This is because if the reinsurance claims are
removed from the claims received, then the reinsurance premium paid by the insurer
must also be included so that the “net reinsurance gains” are accounted for. No studies
surveyed have done this kind of adjustment to the claims variable.
21 Measurement of Expected losses requires knowledge about the risk preferences of the insured,
while actual losses can be known directly from the financial statements. This is one reason
why the actual loss measure has been more frequently used than its expected counterpart.
22 Losses incurred as a measure of output reflect “the total amount of losses expected to
be distributed by an insurer as a result of their providing insurance coverage for a given
year” (Cummins & Xie, 2008, p. 14).
23 “Discounting incurred losses allows for a more meaningful measure of profitability than
reported (undiscounted) losses incurred since premiums reflect the expected present
value of losses” (Choi & Weiss, 2005, p. 644).
24 Output smoothing procedures can be used to handle this problem. See Cummins and
Xie (2008) for its empirical application. Studies that do so have generally found that
smoothing of output variable (when measured by losses) does not change the final
results on efficiency, productivity etc. much.
25 Developed incurred losses are defined as “losses incurred for a given year as reported
several years after the close of the calendar year” (Cummins & Xie, 2009, p. 141). First
incurred losses are defined as the amount of losses as reported on the first day of the
year immediately succeeding the year under consideration. Differences in both these
incurred loss measures represent errors in estimation of technical errors by the under-
writers of the company as explained in Cummins and Xie (2009).
26 Total invested (financial) assets of insurance firms are used to measure their financial
intermediation function. See Yaisawarng (2012) and Cummins and Xie (2013) for an
extended perspective on this matter.
27 One aspect of this variable is that the composition of this variable can be different for life
insurers and nonlife insurers. In advanced economies like the USA, nonlife insurers are
more bent toward riskier instruments like stocks. In contrast, life insurers tend to invest
more in relatively less risky instruments like bonds. This may or may not be accurate for
emerging economies and can differ from economy to economy, region to region within
a country and firm to firm. It is an empirical question and requires empirical assessment.
28 “Using reserves as an output proxy additionally leads to the significant difficulty that
valuation assumptions and methods vary greatly between insurers” (Wise, 2018, p. 12).
29 Please refer to the expert contribution by Prof. Frauke Kreuter in Chapter 7 for further
elaboration on this and its related matters.
30 The assumption of profit maximization has mainly been adopted without actually jus-
tifying its validity except for a cursory defence. Estimation of profit, cost and revenue
efficiencies in insurance literature are among such areas. While it is necessary to assume
profit maximization to make meaningful use of techniques such as DEA in emerging
economies, the welfare-maximization objective may overshadow the profit maximiza-
tion objective due to active regulatory control of the rate-making process. In that case,
 Review of evidence on insurance output measurement 151
alternative constraints on the objective function may be required to account for non-
profit objectives.

References
Acemoglu, D., & Shimer, R. (2000). Productivity gains from unemployment insurance.
European Economic Review, 44, 1195–1224.
Adachi, Y. (2018a). Expanding the scope and improving the efficiency of long-term care
insurance operations. In The economics of tax and social security in Japan (pp. 219–237).
Springer Nature Singapore Pvt. Ltd. https://doi.org/10.1007/978-981-10-7176-8
Adachi, Y. (2018b). Expanding the scope and improving the efficiency of national health insur-
ance administrative work. In The economics of tax and social security in Japan (pp. 205–217).
Springer Nature Singapore Pvt. Ltd. https://doi.org/10.1007/978-981-10-7176-8
Ai, J., Brockett, P. L., Golden, L. L., & Pitaktong, U. (2015). How to set rates if you must:
An efficiency-based methodology for setting promulgated insurance rates with an applica-
tion to title insurance. Journal of Insurance Regulation, 34(7), 1–39.
Akotey, J. O., Sackey, F. G., Amoah, L., & Manso, R. F. (2013). The financial perfor-
mance of life insurance companies in Ghana. The Journal of Risk Finance, 14(03), 286–302.
http://dx.doi.org/10.1108/JRF-11-2012-0081
Al-Amri, K., Gattoufi, S., & Al-Muharrami, S. (2012). Analyzing the technical efficiency
of insurance companies in GCC. The Journal of Risk Finance, 13(4), 362–380. https://doi.
org/10.1108/15265941211254471
Alhassan, A. L., & Biekpe, N. (2015). Efficiency, productivity and returns to scale economies
in the non-life insurance market in South Africa. The Geneva Papers on Risk and Insurance-
Issues and Practice, 40(3), 493–515. https://doi.org/10. l057/gpp.2014.37
Almulhim, T. (2019). Analysis of takaful vs. Conventional insurance firms’ efficiency: Two-
stage DEA of Saudi Arabia’s insurance market. Cogent Business & Management, 6(1), 1–18.
https://doi.org/10.1080/23311975.2019.1633807
Alshammari, A. A., Syed Jaafar Alhabshi, S. M. bin, & Saiti, B. (2019). The impact of com-
petition on cost efficiency of insurance and takaful sectors: Evidence from GCC markets
based on the Stochastic Frontier Analysis. Research in International Business and Finance, 47,
410–427. https://doi.org/10.1016/j.ribaf.2018.09.003
Anandarao, S., Durai, S. R. S., & Goyari, P. (2018). Efficiency decomposition in two-stage
data envelopment analysis: An application to life insurance companies in India. Journal of
Quantitative Economics. https://doi.org/10.1007/s40953-018-0148-1
Arvanitis, S. (2005). Computerization, workplace organization, skilled labour and firm pro-
ductivity: Evidence for the Swiss business sector. Economics of Innovation and New Technol-
ogy, 14(4), 225–249. https://doi.org/10.1080/1043859042000226257
Averch, H., & Johnson, L. L. (1962). Behavior of the firm under regulatory constraint. The
American Economic Review, 52(5), 1052–1069.
Barros, C. P., Barroso, N., & Borges, M. R. (2005). Evaluating the efficiency and productiv-
ity of insurance companies with a Malmquist index: A case study for Portugal. The Geneva
Papers on Risk and Insurance – Issues and Practice, 30, 244–267. https://doi.org/10.1057/
palgrave.gpp.2510029
Barros, C. P., Ibiwoye, A., & Managi, S. (2008). Productivity change of Nigerian insurance
companies: 1994–2005. African Development Review, 20(3), 505–528.
Bates, L. J., Mukherjee, K., & Santerre, R. E. (2010). Medical insurance coverage and
health production efficiency. The Journal of Risk and Insurance, 77(1), 211–229. https://
doi.org/10.1111/j.1539-6975.2009.01336.x
152 Survey of evidences
Bawa, S. K., & Bhagat, N. (2015). Efficiency of life insurance companies operating in Pun-
jab. Pacific Business Review International, 7(9), 76–85. www.pbr.co.in/2015/2015_month/
March/10.pdf
Bawa, S. K., & Chattha, S. (2013). Financial performance of life insurers in Indian insurance
industry. Pacific Business Review International, 6(5), 44–52.
Bawa, S. K., & Ruchita. (2011). Efficiencies of health insurance business in India: An appli-
cation of DEA. American Journal of Social and Management Sciences, 2(2), 237–247.
Berger, A. N., & Humphrey, D. B. (1997). Efficiency of financial institutions: International
survey and directions for future research. European Journal of Operational Research, 98(2),
175–212.
Berry-Stölzle, T. R., Weiss, M. A., & Wende, S. (2011). Market structure, efficiency, and perfor-
mance in the European property-liability insurance industry (Working Paper).
Bertschek, I., Fryges, H., & Kaiser, U. (2006). B2B or not to be: Does B2B E-commerce
increase labour productivity? International Journal of the Economics of Business, 13(3), 387–405.
https://doi.org/10.1080/13571510600961395
Biener, C., & Eling, M. (2011). The performance of microinsurance programs: A data
envelopment analysis. The Journal of Risk and Insurance, 78(1), 83–115. https://doi.
org/10.1111/j.1539-6975.2010.01404.x
Biener, C., Eling, M., & Wirfs, J. H. (2015). The determinants of efficiency and productiv-
ity in the Swiss insurance industry. In Working Papers on Finance (Working Paper No. 1502;
Working Papers on Finance). University of St. Gallen, School of Finance. https://ideas.
repec.org/p/usg/sfwpfi/201502.html
Bikker, J. A., & Gorter, J. (2011). Restructuring of the Dutch nonlife insurance industry:
Consolidation, organizational form, and focus. The Journal of Risk and Insurance, 78(1),
163–184. https://doi.org/10.1111/j.1539-6975.2010.01369.x
Bikker, J. A., & Leuvensteijn, M. V. (2008). Competition and efficiency in the Dutch
life insurance industry. Applied Economics, 40, 2063–2084. https://doi.org/10.1080/
00036840600949298
Born, P. H., & Sirmans, E. T. (2018). Individual market underwriting profitability in health
insurance. Journal of Insurance Regulation, 37(7), 1–13.
Brockett, P. L., Cooper, W. W., Golden, L. L., Rousseau, J. J., & Wang, Y. (2005). Financial
intermediary versus production approach to efficiency of marketing distribution systems
and organizational structure of insurance companies. The Journal of Risk and Insurance,
72(3), 393–412. https://doi.org/10.1111/j.1539-6975.2005.00130.x
Bruneau, C., & Sghaier, N. (2015). Cyclicity in the French property – liability insurance
industry: New findings over the recent period. The Journal of Risk and Insurance, 82(2),
433–462. https://doi.org/10.1111/j,1539–6975.2013.12027.x
Carrington, R., Coelli, T., & Rao, D. S. P. (2011). Australian private health insurance pro-
ductivity growth: Is there scope to limit premium increases? The Economic Record, 87(276),
125–139. https://doi.org/10.1111/j.1475-4932.2010.00684.x
Cerchiara, R. R., & Lamantia, F. (2009). A dynamic analysis of the underwriting cycle in non-life
insurance. The Actuarial Society of Finland, Helsinki, Finland (Colloquium).
Chaiyawat, T. (2014). Corporate governance, efficiency, and firm performance in Thai
property-casualty insurers. The Journal of Risk Management and Insurance, 18(1), 70–85.
Chakraborty, J. (2016a). Efficiency and productivity analyses of public sector general insur-
ance firms in India. The IUP Journal of Applied Economics, 15(4), 74–92.
Chakraborty, J. (2016b). Financial efficiency of the public sector general insurance firms in
India. Pacific Business Review International, 1(1), 177–188.
Chakraborty, J. (2018). Efficiency and productivity analyses of general insurance firms in
India: A Malmquist approach. The IUP Journal of Applied Finance, 24(2), 45–56.
 Review of evidence on insurance output measurement 153
Chakraborty, J., & Sengupta, Dr., P. P. (2016). Measuring performance and efficiency growth
of the selected Indian life insurance companies: A total factor productivity approach. Arth
Prabandh: A Journal of Economics and Management, 1(6), 1–20.
Chakraborty, J., & Sinha, R. P. (2010). Measuring productivity and efficiency of the Indian
life insurance industry: A total factor productivity approach. SURVEY (A Quarterly Man-
agement Research Journal), 50(3–4), 67–76.
Chakraborty, K., Dutta, A., & Sengupta, P. P. (2013). Efficiency and productivity of Indian
life insurance industry. Asia-Pacific Journal of Risk and Insurance, 7(1), 1–26. https://doi.
org/10.1515/2153-3792.1164
Chakraborty, K., & Harper, R. K. (2017). Dynamic productivity analysis of insurance firms –
the effect of firm-specific and macroeconomic characteristics. International Journal of
Applied Economics, 14(2), 37–55.
Chen, L.-R., Lai, G. C., & Wang, J. L. (2011). Conversion and efficiency performance
changes: Evidence from the U.S. Property-liability insurance industry. The Geneva Risk
and Insurance Review, 36, 1–35. https://doi.org/10.1057/grir.2010.3
Chen, L. R., & McNamara, M. J. (2014). An examination of the relative efficiency of fra-
ternal insurers. Journal of Insurance Issues, 37(1), 1–31.
Chen, L. R., Peng, J. L., & Wang, J. L. (2008). Financial integration and economies of
scope: A cross-industry analysis of bancassurance in Taiwan. Journal of Insurance Issues,
31(2), 120–151.
Choi, B. P., & Elyasiani, E. (2011). Foreign-owned insurer performance in the US
property-liability markets. Applied Economics, 43, 291–306. https://doi.org/10.1080/
00036840802552353
Choi, B. P., & Weiss, M. A. (2005). An empirical investigation of market structure, effi-
ciency, and performance in property-liability insurance. The Journal of Risk and Insurance,
72(4), 635–673. https://doi.org/10.1111/j.1539-6975.2005.00142.x
Coccorese, P. (2010). Information exchange as a means of collusion: The case of the Ital-
ian car insurance market. Journal of Industry, Trade and Competition. Springer, 10, 55–70.
https://doi.org/10.1007/s10842-008-0047-9
Cole, C. R., He, E., & Karl, J. B. (2015). Market structure and the profitability of the U.S. Health
insurance marketplace: A state-level analysis. Journal of Insurance Regulation, 34(4), 1–30.
Cummins, J. D., Dionne, G., Gagne, R., & Nouira, A. H. (2009). Efficiency of insurance
firms with endogenous risk management and financial intermediation activities. Journal of
Productivity Analysis, 32, 145–159. https://doi.org/10.1007/s11123-009-0131-0
Cummins, J. D., Misas, M.-R., & Zi, H. (2004). The effect of organizational structure on
efficiency: Evidence from the Spanish industry. Journal of Banking and Finance, 28(12),
3113–3150.
Cummins, J. D., & Nini, G. P. (2002). Optimal capital utilization by financial firms: Evi-
dence from property-liability insurance industry. Journal of Financial Services Research, 21,
15–53. https://doi.org/10.1023/A:1014369617192
Cummins, J. D., & Rubio-Misas, M. (2006). Deregulation, consolidation, and efficiency:
Evidence from the Spanish insurance industry. Journal of Money, Credit and Banking, 38(2),
323–355. https://doi.org/10.1353/mcb.2006.0029
Cummins, J. D., & VanDerhei, J. (1992). A note on the relative efficiency of property-
liability insurance distribution systems. In Foundations of Insurance Economics: Readings
in Economics and Finance (Soft Copy of 1st ed., pp. 1–724). Springer Science+Business
Media, LLC.
Cummins, J., & Weiss, M. (2013). Analyzing firm performance in the insurance industry
using frontier efficiency and productivity methods. In G. Dionne (Ed.), Handbook of insur-
ance (2nd ed., pp. 795–861). Springer. https://doi.org/10.1007/978-1-4614-0155-1_28
154 Survey of evidences
Cummins, J. D., & Weiss, M. A. (2000). Analyzing firm performance in the insurance indus-
try using frontier efficiency and productivity methods. In G. Dionne (Ed.), Handbook of
insurance (pp. 767–829). Springer. https://doi.org/10.1007/978-94-010-0642-2_24
Cummins, J. D., Weiss, M. A., Xie, X., & Zi, H. (2010). Economies of scope in financial
services: A DEA efficiency analysis of the US insurance industry. Journal of Banking and
Finance, 34(7), 1525–1539. https://doi.org/10.1016/j.jbankfin.2010.02.025
Cummins, J. D., & Xie, X. (2008). Mergers and acquisitions in the US property-liability
insurance industry: Productivity and efficiency effects. Journal of Banking & Finance, 32(1),
30–55. https://doi.org/10.1016/j.jbankfin.2007.09.003
Cummins, J. D., & Xie, X. (2009). Market values and efficiency in US insurer acquisitions
and divestitures. Managerial Finance, Emerald Group Publishing Limited, 35(2), 128–155.
https://doi.org/10.1108/03074350910923482
Cummins, J. D., & Xie, X. (2013). Efficiency, productivity, and scale economies in the
U.S. Property-liability insurance industry. Journal of Productivity Analysis, 39(2), 141–164.
https://doi.org/10.1007/s11123-012-0302-2
Cummins, J. D., & Xie, X. (2016). Efficiency and productivity in the US property-liability
insurance industry: Ownership structure, product and distribution strategies. In J. Zhu
(Ed.), Data envelopment analysis: A handbook of empirical studies and applications (Vol. 238).
Springer International Publishing AG.
Cummins, J. D., & Zi, H. (1998). Comparison of frontier efficiency methods: An applica-
tion to the U.S. Life insurance industry. Journal of Productivity Analysis, 10, 131–152.
Daly, M. J., & Rao, P. S. (1985). Productivity, scale economies and technical progress in the
Canadian life insurance industry. International Journal of Industrial Organization, 3, 345–361.
Diacon, S. (1990). Insurance company output and costs. In S. Diacon (Ed.), A guide
to insurance management (pp. 158–173). London: Palgrave Macmillan. https://doi.
org/10.1007/978-1-349-07495-2_10
Dutta, A. (2013). Efficiency and productivity analysis of Indian insurers in post reform era [Ph.D
Thesis]. National Institute of Technology.
ECB. (2008). The Euro area financial system. Financial Stability Review, 113–115.
Eckles, D. L., & Pottier, S. W. (2011). Is efficiency an important determinant of A.M. Best
property-liability insurer financial strength ratings? Journal of Insurance Issues, 34(1), 18–33.
Eling, M., Hoyt, R. E., & Schaper, P. (2020). The impact of capacity on price and produc-
tivity change in insurance markets: New firm-level evidence. Journal of Insurance Issues,
43(1), 22–58.
Eling, M., & Luhnen, M. (2009). Underwriting cycles in German property-liability insurance.
UNIVERSITAT ULM.
Eling, M., & Luhnen, M. (2010a). Efficiency in the international insurance industry:
A cross-country comparison. Journal of Banking and Finance, 34(7), 1497–1509.
Eling, M., & Luhnen, M. (2010b). Frontier efficiency methodologies to measure perfor-
mance in the insurance industry: Overview, systematization, and recent developments.
The Geneva Papers on Risk and Insurance – Issues and Practice, 35(2), 217–265. https://doi.
org/10.1057/gpp.2010.1
Ennsfellner, K. C., Lewis, D., & Anderson, R. I. (2004). Production efficiency in Aus-
trian insurance industry: A Bayesian examination. The Journal of Risk and Insurance, 71(1),
135–159. https://doi.org/10.1111/j.0022-4367.2004.00082.x
Erhemjamts, O., & Leverty, J. T. (2010). The demise of the mutual organizational form: An
investigation of the life insurance industry. Journal of Money, Credit and Banking, 42(6),
1011–1036. https://doi.org/10.2139/ssrn.1558536
Eslake, S. (2010). An analysis of Victoria’s labour productivity performance. Presentation to a forum
hosted by Victorian Department of Innovation, Industry and Regional Development.
 Review of evidence on insurance output measurement 155
Fecher, F., Kessler, D., Perlman, S., & Pestieau, P. (1993). Productive performance of the
French insurance indus. The Journal of Productivity Analysis, 4, 77–93.
Fenn, P., Vencappa, D., Diacon, S., & Klumpes, P. J. M. (2008). Market structure and the
efficiency of European insurance companies: A stochastic frontier analysis. Journal of Bank-
ing & Finance, 32, 86–100. https://doi.org/10.1016/j.jbankfin.2007.09.005
Fernandes, P. R. (2016). A comparative study of performance evaluation of selected private life
insurance companies and their role in economic development of India (2001–10) [Ph.D. Thesis].
Savitribai Phule Pune University.
Ferro, G., & León, S. (2017). A stochastic frontier analysis of efficiency in Argentina’s non-
life insurance market. The Geneva Papers on Risk and Insurance – Issues and Practice, 43,
158–174. https://doi.org/10.1057/s41288-017-0058-z
Fier, S. G., & Pooser, D. M. (2016). Advertising effectiveness for financial services firms:
Evidence from the life insurance industry. Journal of Insurance Issues, 39(2), 137–168.
Fiordelisi, F., & Ricci, O. (2012). Bancassurance in Europe: Past, present and future. Palgrave
Macmillan.
Fukuyama, H. (1997). Investigating productive efficiency and productivity changes of Japa-
nese life insurance companies. Pacific-Basin Finance Journal, 5, 481–509.
Fukuyama, H., & Weber, W. L. (2001). Efficiency and productivity change of non-life
insurance companies in Japan. Pacific Economic Review, 6(1), 129–146.
Fusco, G., & Porrini, D. (2020). Less discrimination, more gender inequality: The case of the
italian motor-vehicle insurance. The Journal of Risk Management and Insurance, 24(1), 1–11.
Gaganis, C., Hasan, I., & Pasiouras, F. (2013). Efficiency and stock returns: Evidence
from the insurance industry. Journal of Productivity Analysis, 40(3), 429–442. https://doi.
org/10.1007/s11123-013-0347-x
Garg, M. C., & Deepti, G. (2009). Malmquist indices of productivity change in Indian gen-
eral insurance sector. The IUP Journal of Management Research, 8(3), 21–36.
Grmanová, E. (2014). Efficiency of insurance lines in the Slovak Republic. In V. Kajurová,
O. Deev, & J. Krajíček (Eds.), European financial systems (pp. 208–214). Masarykova uni-
verzita. www.ceeol.com/search/chapter-detail?id=842987
Grmanová, E. (2015). Efficiency of national life insurance markets in Europe. In European
financial systems 2015: Proceedings of the 12th international scientific conference (pp. 151–157).
Masarykova univerzita. www.ceeol.com/search/chapter-detail?id=840261
Grmanová, E. (2016). Influence of selected factors on the efficiency of insurance companies.
VADYBA – Journal of Management, 29(2), 71–76.
Grmanová, E., & Hošták, P. (2016). Influence of selected environmental factors on the
efficiency of commercial insurers. In European financial systems 2016: Proceedings of the 13th
international scientific conference (pp. 199–206). Masarykova univerzita. www.ceeol.com/
search/chapter-detail?id=839625
Grmanová, E., & Pukała, R. (2018). Efficiency of insurance companies in the Czech
Republic and Poland. Oeconomia Copernicana, 9(1), 71–85. https://doi.org/10.24136/
oc.2018.004
Grmanová, E., & Strunz, H. (2017). Efficiency of insurance companies: Application of
DEA and Tobit analyses. Journal of International Studies, 10(3), 250–263. https://doi.
org/10.14254/2071–8330.2017/10–3/18
Hasan, M. B., Islam, S. M. N., & Wahid, A. N. M. (2019). The effect of macroeconomic
variables on the performance of non‑life insurance companies in Bangladesh. Indian Eco-
nomic Review. https://doi.org/10.1007/s41775-019-00037-6
He, E., Sommer, D. W., & Xie, X. (2011). The impact of CEO turnover on property –
liability insurer performance. The Journal of Risk and Insurance, 78(3), 583–608. https://
doi.org/10.1111/1539-6975.2011.01411.x
156 Survey of evidences
Hornstein, A., & Prescott, E. C. (1991). Measures of the insurance sector output. The
Geneva Papers on Risk and Insurance. Issues and Practice, 16(59), 191–206.
Huang, L.-Y., Lai, G. C., McNamara, M., & Wang, J. (2011). Corporate governance and
efficiency: Evidence from U.S. Property – liability insurance industry. The Journal of Risk
and Insurance, 78(3), 519–550. https://doi.org/10.1111/j.1539-6975.2011.01410.x
Ilyas, A. M., & Rajasekaran, S. (2019a). An empirical investigation of efficiency and pro-
ductivity in the Indian non-life insurance market. Benchmarking: An International Journal,
26(7), 2343–2371. https://doi.org/10.1108/BIJ-01-2019-0039
Ilyas, A. M., & Rajasekaran, S. (2019b). Productivity of the Indian non-life insurance sec-
tor: New evidence employing Färe – Primont index approach. International Journal of
Productivity and Performance Management, 69(4), 633–650. https://doi.org/10.1108/
IJPPM-04-2019-0147
Jackson, M. I. (2012). Total factor productivity change in the non-life insurance sector, Kenya: 2005–
2009 (A Post-Graduation Research Project). Kenyatta University.
Janjua, P. Z., & Akmal, M. (2015). A comparative analysis of economic efficiency of conven-
tional and Islamic insurance industry in Pakistan. Pakistan Business Review, 21–44.
Jarraya, B., & Bouri, A. (2014). Optimal production plan and profit efficiency in Euro-
pean non-life insurance companies. Procedia Economics and Finance, 13, 69–81. https://doi.
org/10.1016/S2212-5671(14)00431-6
Javaheri, A. S. (2014). Productivity evaluation of Iranian insurance industry: A non-
parametric Malmquist approach. Iranian Journal of Economic Research, 18(57), 85–95.
Jawadi, F., Bruneau, C., & Sghaier, N. (2009). Nonlinear cointegration relationships
between non-life insurance premiums and financial markets. The Journal of Risk and Insur-
ance, 76(3), 753–783.
Jeng, V., & Lai, G. C. (2005). Ownership structure, agency costs, specialization and effi-
ciency: Analysis of keiretsu and independent insurers in the Japanese non-life insur-
ance industry. Journal of Risk and Insurance, 72(1), 105–158. https://doi.org/10.1111/
j.0022-4367.2005.00118.x
Jeng, V., Lai, G. C., & McNamara, M. J. (2007). Efficiency and demutualization: Evidence
from the U.S. Life insurance industry in the 1980s and 1990s. The Journal of Risk and Insur-
ance, 74(3), 683–711. https://doi.org/10.1111/j.1539-6975.2007.00230.x
Kader, H. A., Adams, M., & Hardwick, P. (2010). The cost efficiency of takaful insurance
companies. The Geneva Papers, 35, 161–181. https://doi.org/10.1057/gpp.2009.33
Kao, C., & Hwang, S. N. (2008). Efficiency decomposition in two-stage data envelopment
analysis: An application to non-life insurance companies in Taiwan. European Journal of
Operational Research, 185(1), 418–429. https://doi.org/10.1016/j.ejor.2006.11.041
Kasman, A., & Turgutlu, E. (2009a). Cost efficiency and scale economies in the Turkish
insurance industry. Applied Economics, 41(24), 3151–3159. http://dx.doi.org/10.1080/
00036840701367663
Kasman, A., & Turgutlu, E. (2009b). Total factor productivity in the Turkish insurance
industry. International Journal of the Economics of Business, 16(2), 239–247. http://dx.doi.
org/10.1080/13571510902917558
Klumpes, P. J. M. (2022). Consolidation and efficiency in the major European insurance markets.
Aalborg University.
Kozak, S. (2019). Efficiency of non-life insurance companies and its determinants. Acta
Scientiarum Polonorum. Oeconomia, 17(4), 87–95.
Krupa, T., Farbarzevics, K., & Salame, B. (2019). Efficiency of German life insurance com-
panies: Application of a two-stage SBM approach. Współczesna Gospodarka-Contemporary
Economy Review, 10(1), 79–91. https://doi.org/10.26881/wg.2019.1.08
 Review of evidence on insurance output measurement 157
Ksenija, M., Boris, D., Snežana, K., & Sladjana, B. (2017). Analysis of the efficiency of insur-
ance companies in Serbia using the fuzzy AHP and TOPSIS methods. Economic Research-
Ekonomska Istraživanja, 30(1). http://dx.doi.org/10.1080/1331677X.2017.1305786
Kühn, K.-U., & Vives, X. (1995). Information exchanges among firms and their impact on
competition—Publications Office of the EU. European Commission. https://op.europa.eu/en/
publication-detail/-/publication/90fa6d91-de13-4c87-be75-ccb175427a61/language-en
Kulvinskienė, V. R., & Apčelauskaitė, G. (2014). Insurance intermediaries’ work-related
stress: Connections with labour productivity. Ekonomika, 93(1), 146–159.
Kumbhakar, S. C., & Hjalmarsson, L. (1995). Labour-use efficiency in Swedish social insur-
ance offices. Journal of Applied Econometrics, 10, 33–47.
Lee, C.-C. (2013). Insurance and real output: The key role of banking activities. Macroeco-
nomic Dynamics, 17, 235–260. https://doi.org/10.1017/S1365100511000101
Lei, Y. (2019). Reinsurance and insurers’ risk-return profile. Journal of Insurance Issues, 42(1),
37–65. www.jstor.org/stable/10.2307/26608367
Lei, Y., & Schmit, J. T. (2010). Influences of organizational structure and diversification on
medical malpractice insurer performance. Journal of Insurance Issues, 33(2), 152–177.
Lei, Z., & Zhanxin, M. A. (2020). Analysis on the total factor productivity of Chinese life
insurance enterprises and its influence factors: Based on DEA-malmquist index and its
decomposition. SCIREA Journal of Management, 4(2), 16–27.
Leverty, J. T., & Grace, M. F. (2010). The robustness of output measures in property-liability
insurance efficiency studies. Journal of Banking and Finance, 34(7), 1510–1524. https://doi.
org/10.1016/j.jbankfin.2009.08.015
Leverty, J. T., & Grace, M. F. (2012). Dupes or incompetents? An examination of manage-
ment’s impact on firm distress. The Journal of Risk and Insurance, 79(3), 751–783. https://
doi.org/10.1111/j.1539-6975.2011.01443.x
Lin, H. D., Lee, Y., & Shih, M. L. (2011). A study on technical efficiency and productivity
changes of Taiwan’s life insurance industry. Journal of Global Business Management, 7(2), 1–8.
Lorson, J., & Wagner, J. (2014). Sales efficiency in life insurance: The drivers for growth
in the German market. The Geneva Papers, 39, 493–524. https://doi.org/10.1057/
gpp.2013.35
Luhnen, M. (2009). Determinants of efficiency and productivity in German property-
liability insurance: Evidence for 1995–2006. The Geneva Papers on Risk and Insurance – Issues
and Practice, 34, 483–505. https://doi.org/10.1057/gpp.2009.10
Ma, Y. L., Pope, N., & Xie, X. (2014). Insurer performance and intermediary remunera-
tion: The impact of abandonment of contingent commissions. The Geneva Papers on Risk
and Insurance – Issues and Practice, 39(2), 373–388. https://doi.org/10.1057/gpp.2013.12
Mahlberg, B., & Url, T. (2003). Effects of the single market on the Austrian insurance
industry. Empirical Economics, 28, 813–838. https://doi.org/10.1007/s00181-003-0164-6
Mahlberg, B., & Url, T. (2010). Single market effects on productivity in the German insur-
ance industry. Journal of Banking & Finance, 34, 1540–1548. https://doi.org/10.1016/j.
jbankfin.2009.09.005
Majid, A. M. S., Hamid, A., & Faradilla. (2017). Assessing the productivity of insurance
companies in Indonesia: A non-parametric approach. Journal of Applied Economic Sciences,
12(6), 1593–1605.
Mandal, S., & Dastidar, S. G. (2014). A DEA-investigation of efficiency of the Indian general
insurance during recession. Journal of Advances in Management Research, 11(1), 115–136.
https://doi.org/10.1108/JAMR-07-2012-0030
Manikowski, P., & Weiss, M. A. (2013). The satellite insurance market and underwriting cycles.
The Geneva Risk and Insurance Review, 38(2), 148–182. https://doi.org/10.1057/grir.2013.2
158 Survey of evidences
Maré, D. C. (2008). Labour productivity in Auckland firms. Motu Economic and Public Policy
Research, 1–55.
Marjanovic, I., & Popovic, Z. (2020). Profitability determinants of insurance companies in the
republic of Serbia. In Economic and financial challenges for Balkan and Eastern European countries
(pp. 133–158). Springer. https://link.springer.com/book/10.1007/978-3-030-39927-6
Masud, M. M., Rana, M. S., Mia, M. A., & Saifullah, M. K. (2019). How productive are
life insurance institutions in Malaysia? A Malmquist approach. Journal of Asian Finance,
Economics and Business, 6, 241–248. https://doi.org/10.13106/jafeb.2019.vol6.no1.241
Müller, W. (1981). Theoretical concepts of insurance production. The Geneva Papers on Risk
and Insurance, 6(21), 63–83.
Muthulakshmi, P. (2018). Are general insurance companies efficient in their perfor-
mance? – A comparison. Ushus-Journal of Business Management, 17(2), 1–7. https://doi.
org/10.12725/ujbm.44.1
Nandi, J. K. (2014). Relative efficiency analysis of selected life insurers in India using data
envelopment analysis. Pacific Business Review International, 6(8), 69–76.
Naushad, M., Faridi, M. R., & Faisal, S. (2020). Measuring the managerial efficiency of
insurance companies in Saudi Arabia: A data envelopment analysis approach. The Jour-
nal of Asian Finance, Economics and Business, 7(6), 297–304. https://doi.org/10.13106/
jafeb.2020.vol7.no6.297
Nektarios, M., & Barros, C. P. (2010). A Malmquist index for the Greek insurance industry.
The Geneva Papers on Risk and Insurance. Issues and Practice, 35(2), 309–324. https://doi.
org/10.1057/gpp.2010.2
Noreen, U., & Ahmad, S. (2016). Cost efficiency and total factor productivity: An empiri-
cal analysis of Pakistan’s insurance sector. The Lahore Journal of Economics, 21(1), 123–150.
Nourani, M., Devadason, E. S., & Chandran, V. (2018). Measuring technical efficiency of
insurance companies using dynamic network DEA: An intermediation approach. Techno-
logical and Economic Development of Economy, 24(5), 1909–1904. https://doi.org/Techno-
logical and Economic Development of Economy
Oke, M. O., & Abere, M. (2019). Does consolidation have any effecton the operational effi-
ciency in Nigerian insurance firms? Acta Universitatis Danubius. Economica, 15(5), 54–72.
www.ceeol.com/search/article-detail?id=855779
Olaosebikan, O. (2013). The determinants of the profitability of micro-life insurers in Nige-
ria. The Geneva Papers on Risk and Insurance – Issues and Practice, 38(1), 140–159. https://
doi.org/10.1057/gpp.2012.18
Oulton, N. (1998). Labour productivity and foreign ownership in the UK. NIESR (Discussion
Paper No. 143, 1–22). https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.156.
7349&rep=rep1&type=pdf
Oulton, N., & Sebastiá-Barriel, M. (2013). Long and short-term effects of the financial crisis on
labour productivity, capital and output. Bank of England.
Outreville, J. F. (1998). Theory and practice of insurance. Springer. https://doi.
org/10.1007/978-1-4615-6187-3
Owusu-Ansah, E., Dontwi, I., Seidu, B., Abudulai, G., & Sebil, C. (2010). Technical effi-
ciencies of Ghanaian general insurers. American Journal of Social and Management Sciences,
1(1), 75–87. https://doi.org/10.5251/ajsms.2010.1.1.75.87
Pal, S., Bhattacharjee, K., & Pal, S. (2017). Forecasting performance of Indian non-life insur-
ance industry – an analysis of underwriting experience of public and private insurance
companies. Asian Journal of Empirical Research, 7(1), 10–18. https://doi.org/10.18488/
journal.1007/2017.7.1/1007.1.10.18
Parida, T. K., & Acharya, D. (2017). Analysing the technical efficiency and productivity
change of life insurance companies in India. In The life insurance industry in India: Current
 Review of evidence on insurance output measurement 159
state and efficiency (pp. 33–72). Palgrave Macmillan imprint published by Springer Nature.
https://doi.org/10.1007/978-981-10-2233-3_2
Pawirosumarto, S., & Iriani, D. (2018). The influence of work stress, working cost, compen-
sation and work discipline on employee’ productivity. International Journal of Economics and
Business Administration, 6(4), 62–75.
Porrini, D. (2014). The Italian motor insurance market: Will the recent interventions solve
the old efficiency problems? The Geneva Papers, 39, 545–568. https://doi.org/10.1057/
gpp.2014.15
Pottier. (2011). Life insurer efficiency and state regulation: Evidence of optimal firm
behavior. Journal of Regulatory Economics, 39, 169–193. https://doi.org/10.1007/
s11149-010-9139-7
Rahmani, I., Barati, B., Dalfard, V. M., & Hatami-Shirkouhi, L. (2014). Nonparametric fron-
tier analysis models for efficiency evaluation in insurance industry: A case study of Iranian
insurance market. Neural Computing and Applications, 24, 1153–1161. https://doi.org/English
Rao, G. S. S. B., & Venkateswarlu, R. (2015). A study on efficiency and productivity of
Indian non-life insurers using data envelopment analysis. International Journal of Applied
Operational Research, 5(3), 1–15.
Rao, P. H. (2016). Performance analysis of public sector general insurance companies Oper-
ating in India. Pacific Business Review International, 9(5), 73–82.
Reyna, A. M., & Fuentes, H. J. (2018). A cost efficiency analysis of the insurance industry in Mex-
ico. Journal of Productivity Analysis, 49, 49–64. https://doi.org/10.1007/s11123-017-0521-7
Saad, N. M., & Idris, N. E. H. (2011). Efficiency of life insurance companies in Malaysia and Bru-
nei: A comparative analysis. International Journal of Humanities and Social Science, 1(3). 111–122.
Sauian, M. S., Kamarudin, N., & Rani, R. M. (2013). Labor productivity of services sec-
tor in Malaysia: Analysis using input-output approach. Procedia Economics and Finance, 7,
35–41. https://doi.org/10.1016/S2212-5671(13)00215-3
Schlesinger, H. (2000). The theory of insurance demand. Handbook of Insurance, 22, 131–151.
https://doi.org/10.1007/978-94-010-0642-2_5
Segovia-Gonzalez, M. M., Contreras, I., & Mar-Molinero, C. (2009). A DEA analysis of risk,
cost, and revenues in insurance. Journal of the Operational Research Society, 60(11), 1483–1494.
Sherwood, M. K. (1999). Output of the property and casualty insurance industry. The
Canadian Journal of Economics/Revue Canadienne d’Economique, 32(2), 518–546. https://
doi.org/10.2307/136434
Singh, M., & Rajpal, R. K. (2011). Efficiency analysis of the public sector general insur-
ance companies: A comparative study of pre- and post-reform period. The IUP Journal of
Risk & Insurance, 8(4), 28–46.
Sinha, A. (2016b). Productivity growth assessment of public and private sector general insur-
ance companies in India: A Malmquist index approach. International Journal of Applied Man-
agement Sciences and Engineering, 3(2), 50–61. https://doi.org/10.4018/IJAMSE.2016070103
Sinha, A., & Khan, T. L. (2013). The roadmap of productivity and productivity changes
of Indian non-life insurance companies: A study using bootstrapped malmquist model.
Prestige International Journal of Management & IT – Sanchayan, 3(2), 28–41.
Sinha, R. P. (2007a). Operating efficiency of life insurance companies: An assurance region
model. Artha Vijana, XLIX(3–4), 305–320.
Sinha, R. P. (2007b). Premium income of Indian life insurance industry: A total factor pro-
ductivity approach. The IUP Journal of Financial Economics, V(1), 61–69.
Sinha, R. P. (2013). Are Indian life insurance companies cost efficient? Some recent empiri-
cal evidence. Prajnan, 41(3), 181–201.
Sinha, R. P. (2015). A dynamic DEA model for Indian life insurance companies. Global
Business Review, 16(2), 258–269. https://doi.org/10.1177/0972150914564418
160 Survey of evidences
Sinha, R. P. (2016a). Dynamic performance benchmarking of Indian general insurers.
Journal of Quantitative Economics, Springer, 15(3), 543–564. https://doi.org/10.1007/
s40953-016-0061-4
Sinha, R. P. (2017). Efficiency-solvency linkage of Indian general insurance companies:
A robust non-parametric approach. Eurasian Economic Review, 7, 353–370. https://doi.
org/10.1007/s40822-017-0080-2
Sinha, R. P. (2019a). Efficiency, productivity and returns to scale of Indian general insurance
industry: Evidence based on panel data. The Journal of Institute of Public Enterprise, 42(2), 1–25.
Sinha, R. P. (2019b). Total factor productivity growth of Indian general insurance compa-
nies in the recent period: A bootstrapped approach. Journal of Infrastructure Development,
11(1–2), 59–80. https://doi.org/10.1177/0974930619872103
SNA. (2008). System of national accounts. https://unstats.un.org/unsd/nationalaccount/docs/
sna2008.pdf
Spence, M., & Zeckhauser, R. (1971). Insurance, information, and individual action. The
American Economic Review, 61(2), 380–387. www.jstor.org/stable/1817017
Suárez-Álvarez, E., Quiroga-García, R., Manzano-Pérez, I., & López-Mielgo, N. (2016).
Efficiency and financial health in the European insurance market. Journal of International
Business and Economics, 4(2), 29–40. https://doi.org/10.15640/jibe.v4n2a4
Sun, Z. (2020). Research on the factors affecting the total factor productivity of Chinese life
insurance companies. Archives of Business Research, 8(1), 38–50. https://doi.org/10.14738/
abr.81.7426
Taib, C. A., Ashraf, M. S., & Razimi, M. S. (2018). Productivity change in the efficiency of
the insurance and takaful industry of Pakistan. Academy of Accounting and Financial Studies
Journal, 22(Special Issue).
Toivanen, O. (1997). Economies of scale and scope in the Finnish non-life insurance indus-
try. Journal of Banking and Finance, 21, 759–799.
Tone, K., & Sahoo, B. K. (2005). Evaluating cost efficiency and returns to scale in the Life
Insurance Corporation of India using data envelopment analysis. Socio-Economic Planning
Sciences, 39(4), 261–285. https://doi.org/10.1016/j.seps.2004.06.001
Trigo-Gamarra, L., & Growitsch, C. (2010). Comparing single- and multichannel distribu-
tion strategies in the German life insurance market: An analysis of cost and profit effi-
ciency. Schmalenbach Business Review, 62, 401–417. https://doi.org/10.1007/BF03396813
Trufin, J., Albrecher, H., & Denuit, M. (2009). Impact of underwriting cycles on the sol-
vency of an insurance company. North American Actuarial Journal, 13(3), 385–403. https://
doi.org/10.1080/10920277.2009.10597564
Uddin, G. E., Oserei, K. M., Oladipo, O. E., & Ajayi, D. (2018). Industry competitive-
ness using firm-level data: A case study of the Nigerian insurance sector. In Advances in
Panel Data Analysis in Applied Economic Research (pp. 473–492). Springer. https://doi.
org/10.1007/978-3-319-70055-7_37
Vencappa, D., Fenn, P., & Diacon, S. (2013). Productivity growth in the European insurance
industry: Evidence from life and non-life companies. International Journal of the Economics of
Business, 20(2), 281–305. http://dx.doi.org/10.1080/13571516.2013.782979
Viverita, W. S., & Cabanda, E. (2016). Determinants of cost efficiency and productiv-
ity growth of the Indonesian insurance industry. International Journal of Knowledge-Based
Organizations, 6(2), 70–82. https://doi.org/10.4018/IJKBO.2016040105
Vuorinen, I., Järvinen, R., & Lehtinen, U. (1998). Content and measurement of productiv-
ity in the service sector: A conceptual analysis with an illustrative case from the insurance
business. International Journal of Service Industry Management, 9(4), 377–396. http://dx.doi.
org/10.1108/09564239810228876
 Review of evidence on insurance output measurement 161
Wasseja, M. M., & Mwenda, S. N. (2015). Analysis of the efficiency of life assurance com-
panies in Kenya using the DEA-model. American Journal of Mathematics and Statistics, 5(2),
60–71. https://doi.org/10.5923/j.ajms.20150502.03
Weiss, M. A. (1990). Productivity growth and regulation of P/L insurance: 1980–1984. The
Journal of Productivity Analysis, 2, 15–38.
Weiss, M. A. (1991). International P/L insurance output, input, and productivity compari-
sons. The Geneva Papers on Risk and Insurance Theory, 16(2), 179–200.
Weiss, M. A., & Choi, B. P. (2008). State regulation and the structure, conduct, efficiency
and performance of US auto insurers. Journal of Banking & Finance, 32(1), 134–156.
https://doi.org/10.1016/j.jbankfin.2007.09.010
Wise, W. A. (2017). A survey of life insurance efficiency papers: Methods, pros & cons,
trends. Accounting, 3, 137–170. https://doi.org/10.5267/j.ac.2016.11.002
Wise, W. A. (2018). Life insurance company efficiency: Best method and proxies. Insurance
Markets and Companies, 9(1), 6–19. http://dx.doi.org/10.21511/ins.09(1).2018.02
Wolff, E. N. (1991). Productivity growth, capital intensity, and skill levels, in the U. S. Insur-
ance industry, 1948–86. The Geneva Papers on Risk and Insurance Theory, 16(59), 173–190.
Wu, H., & Zeng, X. (2011). Application of computer technology in efficiency analysis
of China life insurance company. Journal of Computers, 6(9), 1832–1841. https://doi.
org/10.4304/jcp.6.9.1832-1841
Xie, X. (2010). Are publicly held firms less efficient? Evidence from the property liabil-
ity insurance industry. Journal of Banking and Finance, 34, 1549–1563. https://doi.
org/10.1016/j.jbankfin.2010.01.007
Xie, X., Lu, W., Reising, J., & Stohs, M. H. (2011). Demutualisation, control and effi-
ciency in the U.S. Life insurance industry. The Geneva Papers, 36, 197–225. https://doi.
org/10.1057/gpp.2011.6
Yaisawarng, S., Asavadachanukorn, P., & Yaisawarng, S. (2012). Efficiency and productivity
in the Thai non-life insurance industry. Journal of Productivity Analysis, 41(2), 291–306.
https://doi.org/10.1007/s11123-012-0317-8
Yakob, R., Yusop, Z., Radam, A., & Ismail, N. (2014). Two-stage DEA method in identify-
ing the Exogenous factors of insurers’ risk and investment management efficiency. Sains
Malaysiana, 43(9), 1439–1450.
Yang, X., & Li, X. (2019). Evaluation research on optimization efficiency of human resource
allocation in Marine insurance industry based on Malmquist index model. Journal of
Coastal Research, Special Issue, Selected Topics in Coastal Research: Engineering, Industry, Econ-
omy, and Sustainable Development (SUMMER 2019),94(Special Issue), 677–681.
Zhang, Y. (2016). Reinsurance counterparty analysis in life insurance industry: The impact on firm
performance/mergers and acquisitions in global insurance industry [Ph.D. Thesis]. The Temple
University Graduate Board.
Zhengyin, W., Yan, W., & Yuna, M. (2020). Research on total factor productivity of Chi-
nese and foreign life insurance industry analysis based on DEA and Malmquist model.
IETI Transactions on Social Sciences and Humanities, 8, 19–23. https://doi.org/10.6896/
IETITSSH.202006_8.0003
Zhong, C., & Sun, W. J. (2011). Insurance intermediary and cost X-efficiency of China’s
insurance firms. International Conference on Management and Service Science, 1–4. https://doi.
org/10.1109/ICMSS.2011.5998449
Zhu, J. (2019). China’s insurance companies’ efficiency: An empirical study [Ph.D. Thesis]. ISCTE
University Institute of Lisbon.
5 Extended notes on the
banking and insurance
output measurement
problem

5.1. Summarising the literature survey

The previous chapters have presented the key output specification approaches
and output measures that have been employed in the banking and insurance
literature. The fundamental aim of those chapters was to summarise and con-
dense the voluminous works in a compact form and allow the readers access to
major issues, strategies and challenges faced by output analysts in the present
context. The literature survey method adopted in this work broadly corre-
sponds to a systematic review. As highlighted in the literature, the postulates
of an ideal output were presented in section 1.6 earlier. While the theory
demands such rigorousness from the chosen output measures, data and empiri-
cal considerations result in divergence from such requirements. Literature on
both banking and insurance has given rise to a rich set of measurements that
reflect the innovation and labour work put in by analysts in bridging the gap
between the theory of output measurement and its practice. The diversity of
opinions and issues as presented earlier reflect the depth and breadth of the
efforts made by economists and national income accountants in capturing the
true economic contributions of the banking and insurance industries. Some
key aspects require further commentary to summarise the seemingly most
essential debates in output measurement literature on banking and insurance.
The same are presented below.

5.2. Extended notes

Synthesising the concept and measure of outputs

The theory of output measurement1 postulates certain prerequisites, as high-
lighted in section 1.6. However, the data availability and practical considerations
result in modifications to the concept of output as postulated by the underlying
theory. Yet the concept of output is critical and requires clarification before any
empirical measurement is undertaken. Indeed “without a conceptual measure
it is not clear what data should be collected and how they should be used to
compute an output measure” (Hornstein & Prescott, 1991, p. 197). Despite

DOI: 10.4324/9781003149828-6
 Extended notes on banking and insurance output measurement 163
there being clear trends in the choice of output variables, one must be careful
not to confuse popularity with validity (Eling & Luhnen, 2010). The concerns
of Clark (1984, p. 54) that “no general consensus seems to have arisen regard-
ing the appropriate definition” seems as accurate today as it was then. At the
same time, one must not fall prey to the error of believing that the existence
of a diversity of opinions is a reflection of a lack of agreement. A fine balance
between these extreme views is necessary to make sense of the patterns and
trends in the literature. Given the complexity of banking and insurance ser-
vices, it is unrealistic to demand a single ideal concept and thus a single ideal
measure of their outputs. Context, beliefs and data are required to choose the
practically most feasible output measure in any performance analysis of the
banking and insurance industries. Measurement is challenging in banking and
insurance, particularly because observational data are used whose information
content has to be extracted using theory rather than experimentation. There is
no a priori control of the analyst on the behaviour of firms or industries under
scrutiny. Quantitative movements in a potential output measure may emanate
from multiple sources. Thus, deciding the efficacy of a particular measure for
a given context has to be determined using the knowledge about the industry
and the ability to make sense of the available data. These demands widen as the
analysis becomes more aggregative.

Stock versus flow outputs

As examined earlier and corroborated by Tables 2.1 and 2.2, banking litera-
ture seems to have preferred the stock measures much more than the flow and
ratio measures. In contrast, insurance literature has generally preferred the flow
measures of output. The stock-flow dichotomy is a common theme in the
banking and insurance literature. As discussed earlier, proponents of both
the approaches exist, and the choice on this matter is primarily governed by the
output specification approach chosen and the data availability. Another debate
has been about using value or volume measures within the stock measures.
Furthermore, within the value framework, there is a debate on whether to use
constant or current price measures. Reconciliation of these debates suggests
that, ideally, the literature would prefer flow measures for both banking and
insurance analysis. Still, difficulties in locating optimal flow measures and the
need to derive them from their stock counterparts have motivated the frequent
use of stock measures.

Nominal versus real outputs

Do banks and insurers produce in real terms or nominal terms? This question
has not been resolved yet. Factoring in that both nominal and real (volume and
constant price) output measures have been frequently used, there seems to be
no clear consensus on this. Some of the experts surveyed as a part of Chapter 7
held a strong view that deflation of banking and insurance output is against the
164 Survey of evidences
nature of production in these industries. They argued that banks and insurers
do not produce in real but nominal terms, and nominal considerations drive
their decisions. Moreover, they argued that if deflation is undertaken for other
empirical reasons, industry-specific deflators or output-specific price indexes
should be employed.
Given that these are missing in the literature, they argue that leaving the data
in nominal terms is the most plausible strategy.2 Those in favour of deflation
have argued that a generalised deflator is better than no deflation at all. They
believe that economic theory demands the output be measured in real terms
(ideally in volume terms). As quantity measures are not readily available in
banking and insurance analysis, deflation with the most relevant price index
is the best strategy. Among these proponents, a large majority argue for single
deflation, while some have argued for double deflation. The trend in the litera-
ture surveyed in this work is evident. Single deflation is preferred over double
deflation, though few studies have adopted the latter approach. Data availability
in terms of the lack of industry-specific price indexes seems to be the primary
determinant of the choices by analysts on this matter.

Single versus multiple outputs

As argued several times earlier in this study, banks and insurers are inherently
multi-output firms. Literature has duly recognised this fact and has employed
multiple measures in performance analysis – whether non-parametric (e.g.
DEA-based production functions) or parametric (e.g. Stochastic Cost Fron-
tiers). Using multiple outputs is the best strategy when analysing performance
parameters such as efficiencies, productivity and others for banking and insur-
ance firms. Modern statistical and econometric methods allow multiple out-
puts and inputs in production analysis. A sparsely used approach has been the
index numbers measure of outputs. This may be understood in the context of
the complexities involved in building output indexes for banking and insurance
outputs and the data availability.

National income versus economic approaches

Economic and national income approaches to output measurement in banking
and insurance have always been at odds in terms of their conceptual framework,
methodology, data requirements and applicability in empirical analysis. Given
that the national income approach is meant for systematising the aggregate pic-
ture of banking and insurance industries, there is considerable rigidity in this
measurement approach. This is required to establish international comparabil-
ity. Economic approaches, on the other hand, have been diverse and innovative.
As comparability is not a concern, the economic approach to output specifica-
tion has accommodated an extensive menu of choices. These alternatives have
been developed to tackle different contexts and different performance issues.
However, this diversity itself has led to a lack of any clear consensus and thus,
 Extended notes on banking and insurance output measurement 165
synthesising the findings remains a difficult challenge. The present work has
attempted to push forward the progress on this account by a small step.

Dual nature of output specification problem

An exciting aspect of the output specification problem and consequently the
output measurement problem is that inputs and outputs must be defined inter-
dependently. Banking and insurance outputs may be conceptualised as inde-
pendent of inputs in purely theoretical terms. But their practical use requires
establishing congruence between the chosen outputs and inputs. The out-
put specification approaches discussed in Chapters 3 and 4 make this point
abundantly clear. Literature has used these output specification approaches to
choose the relevant output and related input measures. This also implies that
the measurement of inputs needs to be theoretically consistent with the chosen
specification approach. Otherwise, the estimating production, cost, revenue,
profit or other functions will produce spurious and unreliable estimates. Input
measurement issues are briefly highlighted in Chapter 10.

5.3. Background for further chapters

Given the narrative so far, it is now time to expand the scope of this work by
providing opinions of selected experts on the central theme of this work. While
a literature survey is necessary to summarise the extant wisdom, it is not a suf-
ficient condition. Traditional literature reviews restrict themselves to analysing
what has been done in the sample studies. Unlike the traditional approach, this
work goes one step further to allow the readers to gain insights from the prac-
titioner’s point of view. Expert opinions can supplement and improve upon the
issues analysed through the literature survey. It is instrumental in highlighting
the debates and discussions at the frontier of a theme. With this motivation, the
next part presents the contributions from carefully selected experts who were
first interviewed through a semi-structured interview and were then requested
to contribute their insights on the theme.

Notes
1 The theory of output measurement in economics is subsumed under the ambit of metrol-
ogy, i.e. the science of measurement. See Link (2021) for a synthesis of metrology and
economic measurement theory.
2 Prof. K. V. Bhanumurthy in particular held a strong view on similar lines. His contri-
bution can be obtained from the author on request or may be included in the online
resources.

References
Clark, J. A. (1984). Estimation of economies of scale in banking using a generalized functional
form. Journal of Money, Credit and Banking, 16(1), 53–68. https://doi.org/10.2307/1992648
166 Survey of evidences
Eling, M., & Luhnen, M. (2010). Efficiency in the international insurance industry:
A cross-country comparison. Journal of Banking & Finance, 34, 1497–1509. https://doi.
org/10.1016/j.jbankfin.2009.08.026
Hornstein, A., & Prescott, E. C. (1991). Measures of the insurance sector output. The
Geneva Papers on Risk and Insurance, 16(59). www.jstor.com/stable/41952063
Link, A. N. (2021). The economics and science of measurement: A study of metrology (1st ed.).
Routledge. https://doi.org/10.4324/9781003186953
Part II

Expert opinions and

contributions
Emerging markets perspective
6 Method and rationale

6.1. Motivation and rationale

The fundamental objective of Chapters 6 and 7 is to supplement and elaborate
on the issue of output measurement in banking and insurance from the prac-
titioner’s perspective. The valuable contributions made by the experts located
and specifically interviewed for this study are contained in the next chapter.
Instead of directly asking for contributions, as would be the case in a typical
edited volume, the selected experts were first interviewed through a semi-
structured questionnaire. The interviews were recorded, and the experts were
then requested to submit their responses to the questionnaire. After that, each
expert was requested to contribute a write-up on their insights and opinions
on the theme under consideration. This draft contribution was then edited
using the pre-recorded interview and response sheet. The final draft was then
sent to the respective experts for their approval. The contributions received
from the experts after their final approval have been presented below.1 The
chief motivation to undertake this exercise is to move beyond a review of his-
torical evidence and present the state-of-the-art ideas and analysis directly from
those engaged in these issues at both policy and academic levels.

6.2. Sampling strategy
The review of evidence undertaken in the previous chapters provided a set of
analysts whose works have been regularly cited. It also allowed the author of
this study to locate researchers who have been regularly contributing scholarly
works on output measurement in banking and insurance performance analysis.
While this method allowed locating possible contributors from academics, the
policy angle also needed to be accounted for. Accordingly, using judgemental
sampling, the author prepared a list of possible experts involved with bank-
ing and insurance-related regulators. This included professionals working in
national income accounting, central banks across major countries, and other
institutions. Both the advanced and emerging economies were targeted. The
final set of potential contributors was then prepared, and each of them was
contacted for this exercise. The response rate was not very high but, at the

DOI: 10.4324/9781003149828-8
170 Expert opinions and contributions
same time, allowed bringing experts from diverse backgrounds on board. Once
they agreed, semi-structured interviews were conducted online or offline, and
the same was recorded. In the second stage, the same experts were requested
to provide their responses to a detailed questionnaire which covered critical
dimensions and issues on bank and insurance output measurement.2 In the
third stage, the experts were requested to contribute a draft write-up on the
theme, and then edited using the pre-recorded interview and response sheet.
The experts were requested to provide final comments and approval on the
edited contributions in the last stage. While several potential experts were
approached, the response rate was limited to around 50%.3

6.3. Notes for the next chapter

The next chapter presents the contributions made by each of the experts in
the final sample. Each contribution is unique and covers different dimensions
of the theme under consideration. There are insights on some frontier issues
that can have important implications for the output measurement problem in
banking and insurance literature. The narratives provided in the next chapter
are a set of rich analyses by those in academics and policymaking. Each of
these results from the selected expert’s vast experience, subjective beliefs and
wisdom. Each contribution is original and has not been published elsewhere
in any form.4

Notes
1 It may again kindly be noted that the views expressed here are strictly personal and
belong only to the expert under consideration. They are not representative of their pro-
fessional opinion or the stance/opinion/beliefs of their current or any of the past institu-
tions or employers. There is no conflict of interest of any manner.
2 The detailed methodological notes including the questionnaire used for this chapter may
be made available in the online resources or can happily be availed directly from the
author.
3 This was understandable given the nature of this exercise. The list of potential experts
who were approached by the author and their response may be availed directly from the
author on request.
4 The contact details of the experts have not been provided here. The same may be availed
from the author of this book directly.
7 Expert perspectives on output
measurement in banking
and insurance

Barendra Kumar Bhoi

Brief Profile: Dr Barendra Kumar Bhoi is a career central banker and an
economist. He served the Reserve Bank of India for more than three decades
since 1984 and held several responsible positions, including Principal Adviser
and Head of the Monetary Policy Department at RBI from 2014 to 2016. He
also served as Visiting Fellow at IGIDR (2018–2020), Chief General Manager/
Head of the Research Department at SEBI (2005–2007) and an Economic
Policy Expert at the Central Bank of Oman (2009–2012). Currently, he holds
the position of RBI Chair Professor at Utkal University, Bhubaneswar, India.

Issues specific to the measurement of financial services in the system

of national accounts 2008 in emerging economies
The SNA, 2008 is a foundational document for laying down the data com-
pilation and aggregation procedures for macroeconomic accounts, and these
are expected to be followed by the member countries of the United Nations
(UN). This vital document recommends the principles that need to be fol-
lowed for achieving a high degree of comparability of national income accounts
data. Official national income accounts provide the most important estimates
of the level and growth of economic activities and are critical for international
macroeconomic policies and coordination. While the member countries of
the UN put in a lot of effort to harmonize their statistical systems to suit the
recommendations of the SNA, there is substantial inequality in the quality of
data collection systems between the emerging and the developed economies.
The reasons for this are well-debated in the literature. However, an essential
dimension in this regard is the localized nature of data compilation by various
countries. While advanced economies are in a position to implement the SNA
procedures efficiently, the local differences across EMEs are far more expansive
and quite heterogeneous. With tight fiscal space and relatively higher priority
for welfare programmes, the financial resources for statistical system improve-
ments take a secondary role in emerging economies. However, this also results
in divergences in the institutional arrangements adopted by various EMEs for

DOI: 10.4324/9781003149828-9
172 Expert opinions and contributions
data collection systems depending on their domestic budgets. Institutional
arrangements for official statistical systems also differ due to considerable dif-
ferences in the structures of these economies. This is all the more true for the
financial services industry across the EMEs as the nature of financial interme-
diation1 differs widely across these countries, unlike the advanced economies,
which have experienced a more sizeable extent of financial globalization and
harmonization of their data collection practices. Despite these facts, the SNA
does allow for at least partial comparability for macroeconomic accounts of the
EMEs. However, this is achieved only partially, and the above-raised factors
remain areas requiring further improvements.

Difficulties in generating current price and constant price estimates

of services output through the value-added method, including the issue
of double deflation
It is well-recognized that there are three major methods for estimating GDP.
They are the income method, the expenditure method and the value-added
method. Theoretically sound estimates of industry-level output require reli-
ance on the value-added method, which demands suitable data collection sys-
tems. Financial markets are more complicated than the real sector due to the
intangibility of what they produce and the existence of indirect intermediation
services. These services are not directly observable and are implicit in the activ-
ities undertaken by financial intermediaries, particularly in the case of banks
and insurance firms. While the banking intermediation services are broadly
proxied by the FISIM methodology, other intermediaries such as insurance
firms remain only partially measured. Along with indirect services implicit in
their operations, such services are non-marketable and require non-traditional
pricing procedures. Traditionally, the observed market prices are the preferred
measures for estimating current price values of output, aggregation across
product segments within an industry and deflating the current price values.
However, generally, banking and insurance prices cannot be ascertained simply
from their observed prices in the financial markets. This is because observed
prices represent the directly observable services provided by the intermediaries,
while implicit services are also not accounted for by these observed prices. The
FISIM approach is the best procedure that we have today, and yet it too fulfils
its aims only in a limited case of the EMEs Brazil, India, Russia, Saudi Arabia,
Turkey, Argentina, etc. These economies do not possess the depth and maturity
of financial markets observed in advanced economies.
Consequently, deriving the implicit prices through explicitly observed data
becomes a challenge for policymakers. FISIM also relies heavily on the qual-
ity of financial markets data, and that also remains a grey area for the largest
of the emerging economies, particularly for the so-called BRIICS countries.
Thus, the sum of value-added across all banking and insurance industries does
not represent its total output under the value-added approach, as the indirect
component of their total output remains unaccounted for.
 Expert perspective on output measurement in banking, insurance 173
Along with these issues, another critical dimension is deriving constant price
estimates of banking and insurance outputs in emerging countries. Advanced
economies utilize double deflators, while the developing economies still rely
on single deflators. Microeconomic theory suggests that the demand and sup-
ply functions of the factor markets and product markets emerge in different
ways, and hence the price behaviour of these two markets is subject to dif-
ferent factors. This is particularly important in the case of the determinants
of inflation in factor markets versus the same in the product markets. Prices
of one sector are not representative of another sector in this case due to dif-
ferent economic structures of both markets, different levels of development
of these markets and differences in the institutional arrangements that govern
the behaviour of these two markets, among others. Moreover, developing
economies rely considerably on filling the gaps and insufficiencies in their
official administrative data through surveys. This approach is only helpful if
the surveys are undertaken in a statistically ideal sense, which is difficult to
follow practically on the ground when data collectors interact with the sam-
ple respondents. Surveys are only a way of making up for the insufficiency
of administrative data. They cannot substitute a matured administrative data
collection system.
With a lack of sufficient statistical infrastructure to take care of the data
needs for adopting the double deflation method of financial services inputs
and outputs data, which is what is recommended by the SNA, 2008, emerg-
ing economies rely on the single deflation method and employ similar price
indexes for both inputs and outputs data. This distorts the representativeness
of the prices for estimating current price data and makes the constant price
outputs and inputs data of financial services challenging to interpret meaning-
fully. Both the current price and constant price data are affected by using a
single deflation method. This is because the countries that do not have mutu-
ally independent price indexes for inputs and outputs generally rely on the
single deflation procedure. Consequently, both the current price and constant
price data depend on market price indexes, which are the same for inputs
and outputs for economies that rely on single deflation procedures. Estimates
of financial services data, such as the amount of bank credit disbursed by the
banking system, amount of premium underwritten by insurers, the value of
assets owned by the banking and insurance firms and others, are critical in
the estimation of the performance of the financial services. These data require
deflation to be economically meaningful. The direction and quantitative extent
of deflation effects will linearly affect both inputs and outputs with a single
deflation approach. In reality, however, this effect should be different for inputs
and outputs of the financial services industry. This is problematic because the
changes in constant price estimates will be distorted across time and space to
the extent that the inflation rates of inputs and outputs of financial services dif-
fer. Over time, this will particularly get blown up, resulting in incorrect policy
inferences from performance indicators of financial services that use both input
and output variables.
174 Expert opinions and contributions
Theory and measurement of financial services output versus other services
Financial services, particularly the banking and insurance industries, present
problems of output measurement that are unique to them compared to other
services. There are both advantages and limitations in the output measurement
of these industries. Due to their complexity on account of direct and indi-
rect intermediation services, considerable methodological developments have
occurred on their measurement issues. A well-defined FISIM approach exists
for estimating the intermediation component of banking services. A similar
procedure does not quite exist for other financial services, including insur-
ance. In the case of non-banking financial services, imputations are used where
needed in order to capture their non-marketable output, which is not as rig-
orous as that of the FISIM approach. The SNA, 2008 recognizes the indirect
component of banking services and recommends the FISIM approach. How-
ever, sample estimates are blown up for other financial services to account for
their non-marketable outputs. This approach is the only route for emerging
countries that do not have a mature enough statistical system to handle the
financial system’s data complexities. For other countries such as China and
Saudi Arabia, where the Government has a tight hold on the official statistics,
undertaking a census may be feasible, but their data collection practices are
generally not transparent to the international community.
Banks produce both marketable and non-marketable outputs. The market-
able portion of their total output is well-captured by their accounting data
and observed prices, such as their fees for various services and interest rates
on loans and advances. However, the non-marketable portion of the output
cannot be ascertained from their accounting data and thus requires external
sources of information for computation through indirect methods. FISIM
remains the gold standard for the measurement of this indirect component.
In reality, though, the countries differ vastly in their abilities to execute the
official data collection principles required for FISIM estimation. This results
in a large degree of divergence in the underlying methods and data used to
estimate the FISIM components across various emerging economies, making
sound international comparisons difficult. Even if an international compari-
son is overlooked, the inability of a country to capture the true output of its
banking sector can distort its understanding of the finance-growth nexus and
can probably result in incorrect policy inferences so far as the growth-improving
capacity of the banking sector is concerned. Despite these aspects, banks’
accounting and financial data to their respective regulators in emerging nations
is much more reliable than the data reported by other entities. This is mainly
because of strict regulation of the banking sector in these economies. Almost
every country in the category of EMEs has a central bank, and collecting and
reporting data from the banking system is an essential function. The real issue
arises when the estimation of the indirect services of banks is concerned, and it
is here that the depth and maturity of the official statistical systems play a deter-
mining role. National income accountants handle such indirect estimations
 Expert perspective on output measurement in banking, insurance 175
and thus largely fall outside the purview of the central banks. This may explain
emerging economies having a mature central banking system but a less devel-
oped national income accounting system.
Insurance output measurement also poses unique problems for national
income accountants in emerging nations. An essential element of insurance is
the spreading of risk through the pooling of the same. Life insurance firms also
provide investment management services, which are well-recognized by the
SNA, 2008. However, countries collect data based on their local realities and
practical problems, which often do not quite correspond to the actual require-
ments of the SNA manual. One of these aspects is the issue of treating returns
versus claims as to the actual output of the insurance industry. While the SNA
manual recommends measuring output through data on premiums collocated
by insurance firms while adjusting the same for their technical reserves and
investment services, an alternative approach can be treating the claims settled
by the insurers as their true output. Both the premiums collected and the
claims settled may also be employed as measures of output, but such approaches
require additional official data, and this can be a problem area for emerging
countries with constrained fiscal space. Compared to banking, insurers do not
undertake any indirect intermediation service per se. However, they are inter-
mediate as risk aggregators and thus help achieve a better distribution of risk
that otherwise would not have been possible if these entities were absent in the
economy. Such aspects are well-recognized by the SNA manual, but again, the
maturity of a nation’s official statistical system is the determining factor.
An essential feature of the insurance markets in emerging economies is their
low penetration and density relative to banking services. This gap in the spread
of insurance and banking services is a constraint on achieving the aim of finan-
cial inclusion, which is a prerequisite to ending the problem of poverty in
emerging and developing nations. Both banking and insurance interact with
each other and across other financial services to create an economic environ-
ment conducive to eliminating poverty, which is the very first Sustainable
Development Goal (SDG) among the seventeen goals currently laid down by
the United Nations. Inclusive growth is the need of the hour for many emerg-
ing economies marred by the severe rural-urban divide. Moreover, emerging
markets display downwards rigidity of insurance premiums, which explains
the observed low levels of its penetration. Nonlife insurance is particularly at
a disadvantage because there is no investment component in these insurance
products. The price elasticity of demand is exceptionally high for nonlife insur-
ance products due to the lack of an investment component over and above risk
coverage offered by most life insurance products.

The great financial recession and key performance issues of the banking
sector of the emerging economies
The financial recession of 2007–08 disrupted advanced economies so that its
aftereffects continue to ripple even today. The emerging economies displayed
176 Expert opinions and contributions
considerable immunity to the same, though they had faced a series of recessions
and recessionary tendencies due to their business cycles and the world eco-
nomic slowdown that began much before the COVID-19 pandemic emerged.
The financial recession increased risk-aversion among banks across major
developed and emerging economies. The regulatory requirements of capital
adequacy, liquidity, and others have become tighter in China, India, Russia,
and Brazil. Banks in the EMEs are further burdened with large NPAs that drain
the productive financial resources of these firms, thereby constraining their
growth and, via the finance-growth nexus, also slowing down the aggregate
economic growth rate and coupled with the world economy and trade slow-
down, the banking industry in the economies mentioned earlier. Hence, three
chief sources of weakening performance of the banking sector may be pointed
out: first, the regulatory stringency post-crisis on banking entities across their
assets and liabilities management; second, the slowdown in the world economy
roughly occurring in the mid-2010s; and third, the increasing burden of NPAs
on their balance sheets. These issues are all the more worrying when one
observes the Non-Performing Loans Ratio (NPLR) of some major emerging
economies such as Russia and India. Furthermore, having a lower NPLR does
not in itself imply that the banking sector is safe and healthy. Large write-offs
by the banks through restructuring by the Governments are also a matter of
concern in countries like China and India, which are heavily dependent on the
public sector banks.

Structural changes in the composition of banking business from

intermediation services to fund-based services and their implications for
the importance of the FISIM approach
Traditionally, banking has been associated with financial intermediation. The
FISIM approach attempts to translate the intermediation services within the
national income accounting framework. However, modern banking is increas-
ingly shifting its income base from indirect to fee-based services. Many of these
services are explicitly charged by banks. Such a trend may result from the net
interest margin (NIM) tendency to be on the lower end for many banks in
EMEs. Low NIM is a significant constraint for any bank as it squeezes its will-
ingness to transform deposits into credit and may result in lower profitability
and possibly even weaker monitoring and selection norms for loan applications.
Such factors are generally the ones that sow the seeds of the financial crisis.
Banks’ services across both emerging and advanced economies have diversi-
fied considerably in the last decade. Financial innovations have been a vital
source of this development. The diversity of fee-based services has expanded
enormously, and so has its volume in the past decade. This begs the question
of whether there has been a change in the composition of the output of com-
mercial banks in terms of the shares of fee-based and intermediation services.
General macroeconomic observations do suggest that this has been the case.
Fund-based services have seen a large increase with the increasing spread of
 Expert perspective on output measurement in banking, insurance 177
universal banking in emerging economies that are generally more dependent
on bank-based financial system designs than capital market-dependent designs.
With such developments, output measurement that emerges from fee-based
services also needs an overhaul to account for the larger share of these services
in the income sources of banks. Traditional intermediation services continue to
be the most crucial output generated by the banking system, but the existing
national income accounting conventions in emerging nations need to adopt
methodological changes to suit the proper measurement of fee-based services.
A lower real rate of NIM could be considered a factor that may explain the shift
in the financial dependency of banks from intermediation to fee-based services.
Stringency in the regulation of the credit creation process is also a possible fac-
tor in this regard. Further research is required to understand how the banking
systems in emerging economies adopt such changes and their implications for
the conventional SNA, 2008-based recommendations.

Further measurement issues in the output of the banking sector

The capital formation process in emerging economies is primarily driven by the
funds supplied by the banking system. Capital markets are not matured or deep
enough to meet the financial needs of the entire economic system. Bank-based
financial systems are thus much more sensitive to the problems of output meas-
urement. One can theorize three different kinds of outputs produced by a bank:
While the above discussion points out a few aspects of the measurement of
non-fee and fee-based output of banks, the non-fee and non-intermediary ser-
vices also need to be accounted for in the macroeconomic estimates of bank-
ing output. Consider, for example, the yield curve produced or discovered by
the banking system. It plays a pivotal role in monetary policy formulation and
efficiently enables capital markets to allocate financial resources. Such services
are like positive externalities and thus result in underestimation of the true
output produced by the banking system. While well-developed pricing pro-
cedures impute such output for public goods, even banking services need to

Output composition of
commercial banks

Non-fee output Fee-based output Non-marketable output

[Measured by interest rate [Measured by revenue [Requires imputation as in
differences] collection] the case of public goods]

Figure 7.1 Composition of commercial bank output

Source: Analysis of the contributing expert.
178 Expert opinions and contributions
be recognized as suppliers of positive externalities. Such inclusions will most
probably increase the value of nominal output produced by the banks.
Similarly, the non-marketable output contributed by the central banks in
terms of the regulatory and supervisory services rendered for the financial
system needs more rigorous accounting procedures. While the central banks
are treated separately from commercial banks, the overall output of the bank-
ing sector must include that of the central bank, as it is a critical component
of the banking industry. To undertake this inclusion, possibly the user-cost
approach is the best. However, emerging and developing nations are marred
by constraints on the statistical system. Such computation methods demand
highly disaggregated data that is observed rather than imputed. Financial sec-
tor data in developing nations, even in large and fast-growing countries such
as China, India, Brazil, etc., are not collected at the degree and depth of
disaggregation required for such computations. Very rigorous input-output
tables at very high degrees of industrial disaggregation are required to ascertain
various externalities associated with the banking sector. Input-output tables
are generally more biased towards industries producing tangible outputs, espe-
cially in developing economies. Such issues need to be addressed by the poli-
cymakers on a war footing.

Some issues in the input-output debate in banking output measurement

The physical distinction between inputs and outputs is critical for empirical
work on production analysis, including estimating total factor productivity.
Unless this is undertaken, empirically estimating a production function for any
economic output will be difficult.
In banking services, this distinction is largely blurred due to multi-stage pro-
duction undertaken by banks wherein deposits, for example, maybe considered
as output in one case and inputs in another. Because banks rely on financial
inputs much more than physical inputs, it is tricky to measure inputs. The
lack of optimal banking price indices further accentuates the problem of input
measurement, as deflation of financial inputs is necessary before estimating the

Alternative functions
of a banking firm

Risk Management
Maturity transfer Payments facilitation Intermediation by
through risk
through asset through banking transforming deposits
reallocation and
reallocation infrastructure into credit
distribution

Figure 7.2 Key functions of commercial banking firms

Source: Analysis of the contributing expert.
 Expert perspective on output measurement in banking, insurance 179
production functions of banking output. However, this problem is dodged by
the developing economies through the use of proxies, which are generally inef-
ficient in representing the actual inflation rate of banking services. Most of
such proxy deflators are price indices for other service industries or general
wholesale or consumer price indices. The construction of an optimal and effi-
cient banking price index is a critical challenge for emerging economies, and
the performance of even the larger ones among them has been unsatisfactory.
Even an aggregate financial services deflator could do more justice than proxy
measures from non-financial services if not a banking price index.
Despite such issues, Total Factor Productivity (TFP) is estimated for the
banking industries of emerging economies. Such studies generally define
inputs from the deposit side and output from the credit side. However, it is
only an operational approach, and like any other empirical adjustment, it is
highly imperfect. One reason that may be immediately cited for this is that
input requirements in a physical sense are always lesser for financial services due
to their higher dependency on financial inputs in digitalization. Banking firms
can serve a larger market with lower fixed costs than manufacturing and agri-
cultural firms. This implies that traditional neoclassical production functions
with labour and capital as inputs might overestimate the TFP. Hence, deflation
of nominal financial inputs before their use in a production function of bank-
ing output is critical. If this deflation procedure is incorrect, one can end up
with biased estimates of TFP even after accounting for financial inputs because
the deflator is not optimal. This same dimension of optimal deflation is equally
applicable to the measurement of the real output of the banking industry.
180 Expert opinions and contributions
Debashis Acharya
Brief Profile: Debashis Acharya is currently a Professor in the School of Eco-
nomics, University of Hyderabad, India. Prior to joining UoH, he taught at
IIT Madras, IIT Hyderabad, mentored by IIT Madras and Panjab University,
Chandigarh. He was also a UGC Visiting fellow at the Gokhale Institute of
Politics and economics. His research interests are macro-monetary econom-
ics, financial economics, and inclusive finance. Some of his recent publica-
tions include Financial inclusion across major Indian states with spatial panel
econometric evidence, the role of inflation targeting in the context of mon-
etary policy and financial stability, the credibility of inflation targeting in Asian
countries, near-stock properties of Bitcoin, and the future of crypto in India
in International Journal of Social Economics, Australian Economic Review, Economic
Change and Restructuring, International Journal of Managerial Finance and EMER-
iCs, KIEP. He has published about 50 papers in refereed national and inter-
national journals, including European Journal of Operational Research, Journal of
Economic Studies, Economic Change and Restructuring, Journal of Financial Economic
and Policy, Indian Economic Journal, Economic and Political Weekly, Journal of Quan-
titative Economics to name a few. Debashis has completed the macro research
award project on financial inclusion funded by the Indian Institute of Banking
and Finance. A few other sponsored research projects completed by Debashis
are funded by ICSSR, UGC, DFID, UK, SICI, Calgary, and IMTFI, Univer-
sity of California, Irvine.

Output measurement in banking services

Banks produce a complex bundle of services that cannot be neatly divided into
inputs and outputs. Depending on the perspective from which one approaches
a particular service, one may consider it an input or an output. An input-output
distinction is dynamic in banking, and this is both a strength and a challenge for
analysts interested in measuring the output of banking services. Distinguish-
ing inputs from outputs is necessary to establish the theoretical foundations
for production analysis. Given that neoclassical theorizing essentially informs
production analysis in current literature, the production frontier approach plays
a dominant role in the performance2 analysis of the banking industry. Studies
focusing on technical efficiency, TFP, and other aspects of banking perfor-
mance require theoretically sound measures of inputs and outputs. Irrespective
of whether the theorized production frontier is parametric or non-parametric,
separating observed banking services into inputs and outputs is necessary. Neo-
classical production theory seems to display a broad consensus on the inputs
that must be included in the production analysis. Labour and Capital are the
two significant inputs on this account. As far as banking firms and industry are
concerned, this should generally not be a problem for a researcher.
However, the major challenge is identifying the correct measure of out-
put, and the mainstream pure theory of production does not directly indicate
 Expert perspective on output measurement in banking, insurance 181
the banking firms specifically. Experience, literature, wisdom, and theoretical
debates need to be carefully synthesized to identify an observed variable as an
output in the case of banking services. While the pro-intermediation analysts
suggest that deposits must be considered an input and credit as an output,
the production approach analysts would prefer to identify both deposits and
credit as outputs at different stages in the banking supply chain. Performance
estimates can differ radically when output is measured from these alternative
angles. Establishing the robustness of performance estimates to alternative out-
put measures is necessary to obtain economically valid inferences about how
the banks and insurance firms have performed and what policy interventions
might be necessary to improve the same.
A concept that plays a fundamental role in macroeconomic research on
banking and insurance performance analysis is that of value-added. This meas-
ure is derived from firm-level data and the various estimates arrived at by the
official data generating agencies in a country. Generally, value-added is used for
industry- and aggregate-level analysis rather than firm-level analysis. Locating
the value-added for manufacturing and even agriculture is a well-researched
issue, and there are many schools of thought and perspectives on this account.
However, value-added accounting is quite challenging for the financial service
industry, particularly for banking. The foremost reason might be that the sup-
ply chain in banking is interlaced across different verticals and horizontals and
thus does not allow a clean division of the production process into various
stages. This also implies that the separability conditions required for justifying
aggregate output measures in an inherently multi-output industry are difficult
to establish in the case of banking services.3 Locating value-added and thus
measuring output at the industry level is an identification problem. This prob-
lem has been there since a long time, and disagreement remains even to date on
the optimal method to undertake it with regards to the financial intermediation
services, such as banking and insurance.
The EMEs under consideration in this book have all had a broadly similar
evolutionary path as far as their banking and insurance industries are concerned.
These economies have been transforming from a predominantly public sector-
dominated industries to increasingly market-oriented private entities-driven
industries. This may be considered a rough stylized fact about these econo-
mies. To this day, the public sector dominates the scene; the determination of
output is straightforward and could be explained by government policies and
the behaviour of the central banks. From a structural perspective, the param-
eters required to estimate an output function could broadly be specified by
studying the policy movements. As these economies began shifting to market-
oriented competition and private ownership of banking and insurance firms,
the number of parameters required to estimate the output of these industries
has increased sizeably. One can no more regard the output of banks and insurers
as majorly driven by the government and Central Bank’s will. An extensive set
of factors affect the demand and supply sides of the industrial output function,
and accounting for them has become a serious challenge more than any time
182 Expert opinions and contributions
in the past. The output level and volatility of banking and insurance industries
have increased due to the move toward privatization and market competition.
The number of parameters required to specify a realistic output determination
function is more prominent than ever before due to multiple determinants of
both the level and volatility of the outputs in these industries.
Lastly, one must bear in cognizance that at higher levels of aggregation, the
choices available for measuring outputs in the banking and insurance industries
reduce to a manageable level. However, many candidates at the microeconomic
level are considered outputs and even inputs. This is due to the sheer complex-
ity and richness of activities performed by banks and insurers. For example,
each fee-based service offered by a bank may be analysed as an output that the
bank produces using a set of inputs. In this manner, one may locate many out-
put variables. However, their relative importance in the aggregate production
function reduces. One is left with some special measures such as the amount
of deposits, amount of loans and advances and the total net worth of the firms.
These observations can be corroborated by the diversity of output and even
input variables employed for firm-level studies as against for industry-level
studies in the present context. Corresponding to these issues is the construction
of an optimal price deflator for banking and insurance outputs. Deflation con-
tinues to be a serious limitation of both disaggregate and particularly aggregate
level studies. One may refer to Barman and Samanta (2004) for a brief crux of
this matter in the banking context.

Some issues with aggregation and aggregation functions in the

current context
Aggregating individual quantities, amounts, and entities is necessary for micro-
economic and macroeconomic analyses. In the study by Acharya and Kamaiah
(2001), I reflect on this point in detail. From the standpoint of microeconomic
theory, demand functions faced by individual banks may need aggregation to
derive a market demand function. From the macroeconomic point of view,
the amount of loans disbursed across all the commercial banking firms may
be aggregated to study the impact of monetary policy rates on credit flow. In
either case, aggregation across individual units is required, and hence the nature
of the aggregation function employed for this task must be clearly specified.
Another critical dimension is the so-called hermeneutic aggregation problem
(Schlicht, 1985).
Simply put, an output variable that, for example, represents the total amount
of deposits across all the banking firms in the entire economy must necessarily
have a microeconomic foundation. Such a variable must have been derived from
the underlying deposit-collecting behaviour of individual banks. If that is not
the case, then possibly such macroeconomic measures might be mere figments
of our imagination rather than sound economic measurements (Menger, 1985).
Three issues deserve special attention in the current narrative. First, eco-
nomic aggregation theory has to be utilized to decide the correct method for
 Expert perspective on output measurement in banking, insurance 183
aggregation, i.e., the correct form of the aggregation function. Aggregation
functions in output measurement for banking and insurance firms inevitably
take the form of multiple functions that need to be themselves simultaneously
aggregated.4 Second, the method of estimating the chosen aggregation func-
tion must be carefully decided. There are generally two methods used in the
literature on the present theme – parametric and non-parametric. Paramet-
ric methods require pre-defined aggregation factors to undertake aggregation
across the inputs and outputs. An advantage of this method is its ability to
measure aggregate input and output for performance analysis, such as required
in residual Total Factor Productivity analysis. Non-parametric methods (such as
Data Envelopment Analysis) can handle multiple inputs and outputs to estimate
performance measures (such as technical efficiency) but cannot yield a direct
measure of aggregate input and aggregate output, as is the case in the paramet-
ric case. However, one can derive non-parametric aggregation factors by using
additional information related to the inputs and outputs of banks and insur-
ance firms. For example, the ratio of individual input cost to total cost may be
used to aggregate the individual inputs into aggregate input. The aggregation
factor, in this case, is the ratio of the cost of individual input to the total cost
and fulfils Menger’s qualification above. Third, there will always be multiple
choices in choosing the optimal aggregation factor – i.e., the weightage factor
shall allow summation or multiplication of the individual outputs into a single
output. There is no consensus on how to decide the optimal factor for aggre-
gation in the case of output measurement in the banking and insurance indus-
tries. Using different methodologies, one may use non-parametric measures or
derive many parametric aggregation factors. Output elasticities of inputs are
one possible alternative in this regard, and they can be arrived at using deter-
ministic or stochastic frontier techniques. However, in the case of parametric
methods, the analyst has some amount of control over the aggregation factors.
This may be considered an advantage or a possible source of bias, depending
on one’s perspective.

Some frontier issues in output measurement theory and practice

Here I would like to put forward some aspects pertaining to the debate at the
frontier of output measurement theory and practice in a general context, rather
than focusing specifically on the EMEs.
Imputed quantities and amounts are necessary for making the macroeco-
nomic accounting measures of output theoretically consistent. Sectoral esti-
mates of Gross Value Added (GVA) and Gross Domestic Product (GDP) for
the financial services industry are frequently used by macro analysts interested
in performance analysis at the industry or sectoral levels. Regarding the bank-
ing industry, there are services whose value cannot be explicitly determined
by the financial statement data of banks. The intermediation services5 that the
banking system undertakes occupy a sizeable portion of the value added by the
banks. However, this is not an explicit output within the production function
184 Expert opinions and contributions
of a typical bank and instead emerges from the inter-linkages and cross-inter-
actions among all the banks. It is inherently an aggregate measure and thus
cannot be explicitly explained via microeconomic production functions and
frontiers. This issue is handled with the help of the so-called FISIM approach
(Financial Intermediation Services Indirectly Measured), whose derivation is
well-narrated in the literature. The basic idea underlying FISIM is that when
the behaviour of individual banking firms is aggregated, their interrelationships
produce additional value-added, which cannot be ascertained simply through
an aggregation of their financial statements data. As noted above, the inter-
mediation6 between lenders and savers through mutually beneficial contracts
is facilitated by the banking7 firms. This requires measurement at the industry
level by accounting for the interactions among the banks and looking at the
banks as a system of firms working in cooperation to channelize scarce loan-
able funds rather than looking at them as individual profit-maximizing agents.
Debates have primarily been undertaken on the choice of the reference rate
in the estimation of the FISIM. Traditionally, a pure interest rate serves as the
reference rate on this account. Locating such an interest rate is a challenge, and
the most frequently employed measure is a typical risk-free interest rate such as
the yield on Government bonds. However an alternative is to utilize the user
cost approach, but the data requirements are even stricter in this case. Given
these aspects, the user-cost approach8 seems unsuitable for the EMEs, and the
FISIM method, no matter its limitations, seems to be the best available bet in
capturing the intermediation services of banks.
The emergence of Big Data is a well-known phenomenon, and the same is
being slowly utilized for performance analysis in banks at the firm level. Large
multi-national corporate banks are the ones who have initiated the inclusion of
Big Data analytics in business decision making.9 Another kind of data analytics
that is emerging is that of thick data, “where the emphasis is not in the quantity
of information but on depth and richness of insights” (Strömmer, 2019). Such
emerging technologies that can be utilized for consumer behaviour analysis
are banking and insurance. Notably, estimating demand functions for different
outputs might become a little easier with more extensive (Big Data) and deeper
(thick data) information on people’s banking behaviour. Whether big or thick
data, these are information in the end, and economic theory will be required to
make sense of them. However, the existing data environment has considerable
scope to improve and include information from such emerging technologies.10
This might help further bridge the gap between what information the theory
of output measurement demands and the knowledge that the existing data
ecosystem can provide.
 Expert perspective on output measurement in banking, insurance 185
Edoardo Pizzoli
Brief Profile: Dr Edoardo Pizzoli is a research fellow at the University of
Luxembourg, Germany, and is also working as a professional statistician at the
Italian National Institute of Statistics, Italy. He has a long hands-on experience
with the national income accounting process, particularly with structural sta-
tistics and has been dealing with disaggregated data across different industries
in Italy.
This work on output measurement of banking and insurance activities is
a welcome contribution to the existing discourse. I appreciate this opportu-
nity to provide some notes on this theme. The discussion below is based on
my experience as both a professional statistician dealing with national income
accounts in Italy and my academic exposure to the University of Luxembourg
and other institutions that I had the opportunity to work with. The views
expressed are strictly mine and are based on my interview and discussions con-
ducted by the author. This contribution has no association with the views of
any of my past or current institutions and is purely personal.

Satellite accounts, financial services and the use of SNA

in emerging economies
In terms of the National Income and Product Accounts (NIPA), emerging
economies are converging to the standards of data assimilation as laid down by
the SNA, 2008. Ample scope for improvement exists, but the upgrades required
differ across different industries. Specific sectors such as the manufacturing sec-
tor are reasonably well-approximated by the existing data collection systems in
nations such as Russia, Brazil, and others. However, significant problems arise
in front of the service industry, specifically financial intermediation services.
Unlike traditional services, intermediation services are not only intangible but
are also difficult to track across space and time. This is mainly due to the indi-
rect services rendered by these firms, especially the banking firms. Such services
require imputation, which is one of the challenges for emerging economies.
Imputation occurs primarily through existing data sources. Available informa-
tion is used to produce estimates of what cannot be clearly tracked and directly
observed in the economy. However, the SNA is too aggregated. While they
account for the indirect intermediation services through the FISIM approach,
additional data sets can help produce more disaggregated measures of the indi-
rect intermediation services of the banking industry. Satellite accounts are a
valuable means to overcome some of the constraints indicated above.11 Achiev-
ing a healthy amount of disaggregation in banking and insurance industries
will probably allow researchers to interpolate sub-industry and even firm-level
FISIM estimates. While the FISIM determination is essentially an aggregate
procedure, its disaggregated measures can add more dimensions to the micro-
level economic analysis of banking output measurement.
186 Expert opinions and contributions
Output of banks and some comments on its economic aspects
Financial services are functional to the very process of production in capital-
istic organizations, i.e., organizations that work under decentralized private
ownership of resources. Compared to the service industries such as education
or tourism, financial services produce direct observable gains for services and
non-service industries. In terms of economic theory, one may always argue that
all economic activities produce externalities in some way or another; here, the
direct observable effects are being focused upon. Banking services, for exam-
ple, are part and parcel of all the industries that require funds. Thus, the value-
addition12 of the banking industry is not limited to itself or some pre-selected
set of industries. Typical input-output analysis is also not helpful mainly because
such disaggregated input-output tables are not available13 as the value addition
of banks may be inherent to the output of other industries and thus may not be
directly observable. The FISIM approach roughly captures such an unobserv-
able portion of their value-addition. However, measuring the directly observ-
able value-addition by the banking and insurance firms is also a challenge due
to the immense diversity in the use of these services. Banking, in particular,
presents a stricter challenge due to the higher penetration of banking services
in the EMEs than insurance penetration. This implies higher interconnections
of the banking industry with other industries and thus a more complex web
of value-chains that need to be delineated for estimating the final value-added
by the banking industry. The use of financial statements of the banking firms
is the most frequently used approach in determining their output. However, as
suggested above, there are many intricate interconnections of banking activi-
ties with other industries, making it challenging to trace all the value-additions
produced for different sectors. The output of non-banking industries might be
overstated due to the value-added by the banking services in those industries.
Such measurement issues require imputation, and even the FISIM estimation
will be insufficient as it focuses only on the indirect intermediation services
rather than the direct services embodied in the outputs of other industries that
directly utilize banking services.
Keeping aside the empirical issues in measuring banking output, it is nec-
essary to conceptualize what banks supply in a typical emerging economy.
Smooth circulation of value14 is a prerequisite for a healthy real sector of an
economy. The networks of banks, their technological sophistication, the spread
of core banking services, the depth of the financial markets and many such
dimensions will determine how well the goods and services that represent the
value produced in the economy circulate throughout the economy. Banks
play a pivotal role in enabling this value circulation and thus allow the real
sector can establish fast, efficient and stable interconnections that otherwise
would be difficult to achieve. Similarly, banks play a critical role in physical
and financial capital accumulation. Emerging economies generally display a
higher debt-to-equity ratio, particularly for the firms in the unorganized sector
of the economy. With more extensive dependence on banks for funds rather
 Expert perspective on output measurement in banking, insurance 187
than on capital markets, the nature of the credit distribution by banks across
the different industries will have a long-term impact on the structure of the
economy and its growth path. Banks allow funds-scare firms to borrow money
and enable them to expand their current output and invest in improving physi-
cal capital to increase future production. On similar lines, banks also provide
investment opportunities to savers, thus enabling them to shift their future
consumption to higher levels through financial accumulation.15 The banking
industry is an essential element in the growth story of the emerging economies
that are slowly catching up with the advanced economies. They are also a solid
policy tool to improve the growth process in these economies.

Notes on some allied issues

I would now like to point out some caveats and concerns for output measure-
ment analysts and the larger audience. These are based on my experience as a
professional statistician working on national income accounting as well as an
academic.
One aspect that must be appreciated is that the banking industry is under-
going a tremendous transformation in terms of technological sophistication.
It may not be incorrect to suggest that the banking industry will be primar-
ily digitalised in the next decade, with physical infrastructure playing only a
secondary role. In the economic sense, Output must somehow capture the
value generated by the production process for its consumers. While consumer
valuation is subjective, as established by the marginalist school in econom-
ics, performance analysts and national income accountants cannot fulfil their
work under such a belief. An estimate of what has been produced is needed to
inform the policy interventions designed to promote growth and development
in fast-growing economies such as the EMEs. This task will become more
complicated as the value chain shifts towards digital channels. Digitalization is
not a problem for output measurement analysts at the firm and industry levels.
The changes that this process brings in the value production by banks require
more research and much more debate in the academic literature. Will it be
feasible to specify the production function of the banking industry with physi-
cal capital as a determinant at all? What kind of technology will represent this
sector’s residual total factor productivity in the coming decade? Will labour,
as traditionally defined, play a sizeable role as an input in the production pro-
cess of this industry? How will the output be defined if banks enter an era of
super-specialization due to ever-expanding market segmentation? Several such
questions arise, and the output analyst will need to embrace these dimensions
of banking output while expanding the scope of what is considered inputs and
outputs in the banking industry of the next decade.
188 Expert opinions and contributions
Frauke Kreuter
Brief Profile: Professor Frauke Kreuter is Co-director of the Social Data Sci-
ence Center (SoDa) and faculty member in the Joint Program in Survey Meth-
odology (JPSM) at the University of Maryland, USA; Professor of Statistics and
Data Science at the Ludwig-Maximilians-University of Munich, Germany, and
head of the statistical methods group at the Institute for Employment Research
(IAB) in Nuremberg, Germany. She is an elected fellow of the American Sta-
tistical Association and the 2020 recipient of the Warren Mitofsky Innovators
Award of the American Association for Public Opinion Research. In addition
to her academic work, Dr Kreuter is the Founder of the International Program
for Survey and Data Science, developed in response to the increasing demand
from researchers and practitioners for the appropriate methods and right tools
to face a changing data environment; Co-Founder of the Coleridge Initia-
tive, whose goal is to accelerate data-driven research and policy around human
beings and their interactions for program management, policy development,
and scholarly purposes by enabling efficient, effective, and secure access to
sensitive data about society and the economy; and Co-Founder of the German
language podcast Dig Deep.

Traditional data collection systems and their emerging alternatives in the

construction of macroeconomic statistics
Timely and error-free macroeconomic data are the most critical needs for mac-
roeconomic policy formulation in both advanced and emerging economies.
In reality, however, the data collected by official agencies are subject to errors –
both sampling and non-sampling, as well as being constrained by a plethora
of issues such as the state of the data collection infrastructure, the political will
to invest in improving the same, the degree of granularity that the agencies
can achieve in collecting data, among others. Despite such constraints, official
macroeconomic data are the best available sources for researchers, policymak-
ers, and other economic agents. Traditionally, physical survey-based data have
focused on data collection agencies in both developed and emerging nations.
Though practical considerations have made sample-based surveys the preferred
approach, such data can be collected through census or sample surveys. Sample
surveys are utilised from data on key aggregate economic indicators such as
Gross Domestic Product (GDP), employment and unemployment, and infla-
tion to specialized data sets on district-level development indicators. These
surveys typically consist of pre-defined questionnaires through which a ran-
domly selected16 sample is approached physically. Their responses are collected,
recorded and appropriately organised for further aggregations. In the case of
industries such as banking and insurance, a census might be a feasible option
as these industries are typically well-regulated and are required to report their
economic activities to the designated official authorities regularly.17 However,
a large portion of economic information is obtained from survey data. Even in
 Expert perspective on output measurement in banking, insurance 189
cases where access to population data is possible, the entire population might
be too heterogeneous to be covered in a single census. Survey-based informa-
tion thus remains necessary, though not a sufficient, source of information for
policymaking.
While sample-based information shall continue to be the chief source of
knowledge about the economic issues for macroeconomic policy, alternative
sources of data are emerging in developed economies, which may help either
substitute or complement the survey-based data sources. The traditional data
and the data collection methodologies are now being fundamentally trans-
formed due to the emergence of various innovative economic and financial
information sources. These sources range from satellite imagery to cell phone
data and have a high degree of granularity. The volume and frequency of such
data are substantial compared to the traditional survey-based data. Such data are
generally observational, as against experimental,18 and require different statisti-
cal philosophies for optimal decision-making. A notable feature of these data
sets is that while many statistically significant correlations are observable, cau-
sality is much more challenging to locate due to the highly complex nature of
these data and the underlying data generating processes (Shriffin, 2016). A few
years ago, this was true even for traditional survey data. The nature of problems
associated with survey data took several decades to be understood. While some
grey areas still need further research, there is also a sizeable consensus on what
such data are, how their underlying data generating process is to be modelled
and what various sampling and non-sampling errors may be encountered while
dealing with them in official macroeconomic statistics. Today, we are facing
the same situation with reference to the emerging data types and sources as
highlighted above.
Merging multiple data types and data sources into macroeconomic policy-
making requires an overhaul of the data collection systems and the data col-
lection agencies that have been traditionally employed in official information
assimilation, especially across emerging economies due to various well-known
limitations in their information technology (IT) infrastructure and data process-
ing capabilities.19 It will not be a hyperbole to assert that the future of informed
policymaking in various areas of macroeconomic issues shall be driven by the
use of multiple data sources rather than relying primarily on survey data. Such
a development has already been initiated in several corners of the world, par-
ticularly in the advanced economies. In the U.S., for example, I was a part of
the US National Academy of Sciences Panel on “Improving Federal Statistics
for Policy and Social Science Research Using Multiple Data Sources and State-
of-the-Art Estimation Methods”, out of which two reports emerged – namely
Groves and Harris-Kojetin (2017a, 2017b). These two reports provide a valua-
ble introduction to the issues that official agencies need to be prepared to handle
if they wish to diversify the portfolio of information used for decision-making.
One may also extrapolate these reports’ ideas to the financial services indus-
tries, where observational data are the most practical way to assess economic
activities. Increasingly, financial markets are moving toward digital platforms,
190 Expert opinions and contributions
even in EMEs. With increasing financial integration and the ongoing shift
to digital delivery platforms, financial services can now be observed through
multiple angles, rather than only relying upon survey methods. Data obtained
through digital recording of banking services through various platforms such
as mobile banking, internet banking and others are valuable information for
monetary and financial authorities. Deeper patterns can be unearthed using
data sources that would never be feasible in survey-based approaches. Achiev-
ing a high degree of granularity is only possible when highly disaggregated data
can be reliably recorded, which is possible only through the various alternative
sources as touched upon here. Survey data shall be highly costly if one attempts
to achieve a high level of disaggregation. Another issue with survey data is the
secular decline in the response rates (Groves & Harris-Kojetin, 2017b). The
reliability of survey-based information where response rates are gradually but
consistently falling is open to heavy criticism. Hence, combining alternative
data sources to analyse the same economic issue is urgently needed, particularly
for emerging economies with enormous scope for further economic growth
and development.
The challenge of using multiple data sources is further intensified by the
complexities involved in combining such sources into a single analytical frame-
work. For illustration, consider the issue of monetary policy formulation. The
central bank will typically use a macroeconomic model with well-defined
structural parameters. These parameters are generally estimated using data col-
lected using survey methods by the concerned agencies. Using multiple data
sources will pose a challenge because the specified theoretical model assumes
comparable and consistent data. Standardizing the same shall be needed with
different data sources, which requires innovative statistical approaches beyond
the traditional standardization procedures employed in official discourses on
this matter.20 Such issues present fascinating scope for further research. The
US National Academy of Sciences has been working on these matters through
various colloquiums, panel reports and other means. Many such organizations
and agencies are working on such issues with particular reference to official
macroeconomic accounts of countries.

Adoption of emerging alternative data sources in advanced economies

and the emerging market economies
Another project that I have been working on is the BERD@NFID platform –
namely the Business, Economic and Related Data project at the National
Research Data Infrastructure, Germany. Large unstructured data sets from
varied businesses and industries are becoming common today. It would be
unrealistic for national income accountants, industrial economists and financial
services professionals not to utilize the same. Consumer behaviour data from
digital marketplaces to social media analytics – such data are not necessarily
numerical. Such data break the traditional belief that a datum is a numerical
piece of information. However, that is not the case anymore. Videos, images,
 Expert perspective on output measurement in banking, insurance 191
clicks by people on online platforms, likes and dislikes on social media plat-
forms, biometric information collected by official agencies and a host of such
inherently non-numerical data sources have emerged. The BERD@NIFD
platform is an ambitious attempt to systematize such information into easily
usable data sets for businesses, governments and others.
Let us contextualize these ideas regarding industrial output estimation at
the macroeconomic level. Theoretically, the measurement of production in
any industry requires the knowledge of the cost, revenue, demand and supply
functions. Without the information on these, empirical estimation of output,
productivity, efficiency and other parameters is not feasible. This problem is
exemplified in the case of financial services, where physically tracing these
functions is not possible. The value-addition process in industries such as bank-
ing and insurance is interspersed across the supply chain. Hence, delineating
the inputs and outputs at various stages of production is difficult. However,
such constraints may be partially overcome if survey-based official data sources
and alternative data sets such as digital purchase behaviour and social media
analytics are used. The data for estimating the demand function, for example,
can very well be sourced from unstructured data sets that online marketplaces
such as Amazon and Walmart usually collect. While the data are large, they
can be structured using AI and ML algorithms. Such data can be merged with
traditional data sources. This has both economic and statistical advantages.21
Such ideas seem fascinating, but the data infrastructure of EMEs needs an
overhaul to account for such emerging information sources. With massive
development agendas on hand, the governments in emerging markets such
as China, Russia, Indonesia, India and others face high costs when they may
divert scarce funds from politically popular developmental projects to improve
the statistical infrastructure. Political will plays a crucial role here, and it will
take several years more for developing and emerging countries to garner the
popular support to spend financial resources on projects that do not seem criti-
cal today in the eyes of the public. Financial resources are required because the
data collection agencies need to be trained to handle novel types of informa-
tion. The analytical capabilities need to be improved by investing in Informa-
tion Technology.22

Some challenges faced by the emerging economies in traditional

data collection
The opportunity cost of not protecting sensitive economic and financial data
increases with the increase in the volume and flow of data. Hence, when
developing, nations begin merging multiple data sources to construct mac-
roeconomic statistics, the need to protect the microdata will increase. Such
disaggregated data can also include sensitive information. Thus, safeguarding
them from misuse while also allowing access to researchers, policymakers and
other stakeholders is a genuine concern that the advanced economies have
already been dealing with and the emerging economies are expected to deal
192 Expert opinions and contributions
with sooner or later. In the U.S., I have been a part of the Coleridge Initia-
tive. This project aims to bridge the gap between how and what kind of data
informs public policy decisions and how much the public knows about these
policy decisions. Informing the public about the data-dependency of policy
decisions is vital to flourishing a truly democratic political system. Most emerg-
ing economies, except a few, practice one or the other form of democracy.
Hence, public policy measures are well-informed is a matter of interest even in
developing nations.
However, the dynamics of official policy formulation are very different in
the developing economies compared to their developed counterparts. Trans-
parency in policymaking and the extent to which rigorous data is utilized to
make decisions are only scarcely visible in emerging nations. This issue is pro-
nounced when bureaucracies manage official data collection systems. Political
interference and even administrative interference and errors can restrict the
protection of sensitive data collected through novel sources, as highlighted ear-
lier. Such matters are real and need a debate across the board in EMEs so that
the fast-emerging data sources can be smoothly merged into their traditional
data systems.
 Expert perspective on output measurement in banking, insurance 193
Justin Paul
Brief Profile: Prof. Justin Paul currently serves as Editor-in-Chief of the
International Journal of Consumer Studies (IJCS), a 46-year-old global aca-
demic journal ranked an “A” grade by the Australian Business Deans Council
(ABDC). A former faculty member with the University of Washington, he
is a full professor of PhD and MBA programs, at the University of Puerto
Rico, USA, with a 3-Year Visiting Professor appointment at the University
of Reading Henley Business school, England. He holds two honorary titles as
“Distinguished Scholar/Distinguished Professor” with the Indian Institute of
Management (IIM-K) and SIBM, the premier business school in South Asia.
I welcome this opportunity to examine some aspects of the theme of this
book. The views expressed below are based on my experience in academics
and corporate offices of the banking firms in the US, India, and other places.

Multi-output nature of banking and insurance production

Measurement of what is produced by an individual firm is an essential prereq-
uisite for strategizing its scarce resources in a competitive environment, assess-
ing its performance, changing the scale of operations, and many other areas of
business decision-making. In a highly competitive market, the firms face con-
tinuous pressures to innovate and provide newer products and services. Bank-
ing and insurance firms are no exception on this account. Banks and insurers
invest many resources in technological upgrades on retail and corporate deliv-
ery models. This is a fact that is true for both advanced and emerging econo-
mies. It is vital for the retail segments in EMEs due to their peculiar nature.
On the one hand, these economies, such as Brazil, Russia and India, are
bank-dependent on the supply of funds. Capital markets are still relatively
underdeveloped, and banks play the dominant role in funding new businesses
and older and established businesses. On the other hand, the retail segment is
considerably more prominent in these countries than its advanced counterparts
because of larger populations and market sizes. The services produced by bank-
ing firms continuously undergo fundamental changes at the firm level. Market
segmentation is a fact of life in the banking industries of emerging economies.
These observations imply an ever-evolving nature of the output produced by
banks and even insurers. Looking from the banking angle, the value proposi-
tions embedded in the services of banks are not stable anymore, and static
definitions of bank output with a univariate focus are not good enough to lend
justice to the production dynamics of this industry.
Let me examine these observations further. Treating the amount of loans
disbursed by a bank as its output may be correct in terms of accounting and
financial senses but might not be valid from economic and marketing23 senses.
First, because each banking firm is an inherently multi-output producing
entity at the microeconomic level, single-output variables will produce biased24
estimates of performance parameters. A multi-output framework is the most
194 Expert opinions and contributions
realistic approach to this.25 Second, because the value generated by these output
variables are interdependent and in a single-output context, the actual value
provided to consumers cannot be ascertained without accounting for the inter-
dependence with other output variables in value formation for consumers.
Thus, a multi-output approach is the most reliable approach to performance
analysis of banks and insurers. Accordingly, output measurement for banking
and insurance industries must be concentrated on combining the information
from different output candidates into a single measure, or one should explicitly
use multiple outputs in the empirical exercises on performance analysis.

Banking industry in the advanced versus emerging economies

Two factors are generally believed to differentiate emerging economies’ bank-
ing industries from advanced ones. One is the pace of technological sophistica-
tion and its adoption. Second is the depth and quality of the prudential norms
in the banking industry. Compared to the insurance industry, the regulatory
restrictions are more prominent in the banking industry in emerging econo-
mies. The depth of these regulations can be roughly understood by the extent
of the banking industry they cover,26 and the quality of these regulations can be
broadly understood in terms of the outcomes they generate over time. Emerg-
ing economies are converging technologically with advanced economies. The
operational technologies, which may be considered the set of technologies that
help increase the productivity of back-office operations in banks, in the EMEs
are rapidly catching up with the extant practices in the more matured banking
industries. Regulators in countries such as Brazil, India, Russia and even China
have increasingly become aggressive in promoting state-of-the-art practices for
improving operational efficiencies of their banking industry. Banks, too, have
become more aware of the competitive advantages of such technologies due to
competitive pressures in both the national and international markets. The use
of Artificial Intelligence in risk analytics by banks and insurers is one example
of operational and technological advances. Similarly, the data collection prac-
tices in the EMEs, for example, are driven mainly by digitalization, though
the depth and width of the firm-level data are not as considerable as in the
advanced economies.
As far as the depth and quality of prudential norms are concerned, EMEs
have several challenges to be resolved. One is the issue of adverse selection
which seems to be stubbornly persistent even when more efficient AI and
ML algorithms are used in selecting, monitoring and pricing the products for
both banks and insurers. Specifically, three areas require urgent policy over-
haul in emerging nations – credit norms, retail loan assessment and corporate
loan screening. The corporate segment especially brings a large share of non-
performing assets (NPAs) for the banks in EMEs. NPAs directly reduce the
economic value generated by various outputs of banking firms for their con-
sumers and other stakeholders. If the observed amount of loans has increased
in a particular country, but the same has been accompanied by an even larger
 Expert perspective on output measurement in banking, insurance 195
percentage increase in NPAs, then simply inferring that the output (amount
of loans in this case) has increased will be misleading. The so-called economic
‘bads’ must also be measured and accounted for when the economic ‘goods’ of
the banking industry are measured.

Further remarks on measuring the output of the banking industry

The correct way to theorize about bank output, in my opinion, is to consider
deposits as inputs and the loans disbursed by banks as the output. The standard
inputs of labour and fixed capital shall be added to the production function
along with deposits, and the amount of credit disbursed should represent the
output. This is the so-called intermediation approach, which seems valid from
an industrial perspective. However, at the firm level, one must be wary of the
intermediation approach and instead adopt a more flexible approach depending
on which aspect of banking activities is being analysed. From a branch-level
perspective, the output can be measured from multiple perspectives, such as
the amount of loans and deposits, the number of depositors and borrowers and
various other measures. From the standpoint of the shareholders, the return on
equity generated by a bank would serve as the correct measure of its output
because that is the variable in whose maximization the investors would be
interested. Thus, at the firm level, the diversity of output variables is large, and
the banks are not limited to simply intermediating but also provide value-added
services that occupy a sizeable portion of their total costs.27

Some issues in using financial statements as the fundamental source

of output data for banks
Central banks across major economies – whether emerging or advanced, assim-
ilate data on the banking activities from the financial statements reported by the
banking companies to the authorities. The firm-level data pass through a series
of aggregations until they reach the public domain through the official agen-
cies. Central banks are generally entrusted with the responsibility of putting the
best possible “estimates” in the public domain. Hence, there is strict monitor-
ing of the quality of these data by the Central banks as well as by the internal
auditors of the banking firms. Thus, the incentives to manipulate or unjustifi-
ably transform financial data are significantly less at the firm level, mainly due
to strict internal and regulator monitoring.
Given that large publicly listed commercial banks characterize EMEs, public
pressure also ensures that data are reliably reported due to fears of penalties, stock
market crashes and other issues. In my humble opinion, it may be correct to
believe that commercial banks do not manipulate data due to both strict inter-
nal and regulatory monitoring. This lends the financial statements of banks as a
valuable source of information on the production dynamics in the industry. This
is one reason why premium data sources provide firm-level data on the bank-
ing industry exist, for which researchers and institutions are prepared to spend
196 Expert opinions and contributions
a hefty fee for access. Moreover, such databases allow researchers to build their
output measures depending on the scope of their cross-section samples of banks.
Lastly, it should be pointed out that once the financial data leave the cor-
ridors of banks’ corporate offices, one can never be truly sure about the quality
of data. Simply looking at the extent of statistical noise in data and inferring
data quality based on such a criterion will be misleading. One way to avoid
such a problem is to use firm-level data to estimate aggregative measures such
as industry- or sub-industry-level output measures. Literature has also used
national income data for industrial performance analysis of financial services.
The Gross Value Added (GVA) data is the most frequently used measure of
output in this context. This measure of output is highly aggregative and is gen-
erally available only for the financial services industry.28 Instead of using such
pre-aggregated data, one can focus on firm-level information, even if limited
in nature and use it for building aggregative measures for themselves. However,
at the banking industry level as a whole, one is then faced with the problem of
estimating the indirect output produced by banks, which is generally captured
by the so-called FISIM approach in the national income accounts. Firm-level
FISIM estimates are not available and difficult to derive from the sectoral GVA
series and its components. Firm-level data in banking will generally fail to cap-
ture this component, and it is here that probably the superiority of output data
from macroeconomic accounts is undoubtedly established.
 Expert perspective on output measurement in banking, insurance 197
N. R. Bhanumurthy
Brief Profile: Professor N R Bhanumurthy is currently the Vice-Chancellor,
Dr B. R. Ambedkar School of Economics University (BASE University), India.
He is also a professor at the National Institute of Public Finance and Policy
(NIPFP), India. Prior to joining NIPFP, he worked as an Assistant and Associate
Professor at the Institute of Economic Growth (IEG), India. He has worked as a
macroeconomist at UNESCAP, Bangkok, and the UNDP Regional Centre for
the Asia-Pacific region, Colombo. He has been a Visiting Fellow at the Founda-
tion Maison des Sciences de l’Homme, Paris, France, and the McGill University,
Montreal, Canada. He has also been a consultant to the International Labour
Organization, the World Bank, and the Asian Development Bank.

Notes on national income and other macroeconomic accounts

With the increasing complexity of modern economies and the increased need
for welfare interventions to induce growth and development, collecting and
analysing large volumes of data are critical for sound policy formulation. Among
all the different types of data collected by official authorities, the national income
accounts possibly play the most important role in connecting the ground-level
economic reality with those at the helm of the affairs. Different countries have
different data collection capabilities, so a common standard of official data col-
lection and reporting system is required for international comparisons. Probably,
such a common international standard is also needed for supplying different
countries with a statistically sound set of criteria that they should aim to follow if
they wish to improve their official statistical systems. These functions are fulfilled
by the System of National Accounts of the UNSC.
A concern that commonly worries the macroeconomic accountants globally,
particularly in developing and emerging economies, is the measurement errors
that inevitably occur in such large-scale data assimilation efforts. The more
advanced economies can minimize these errors, though they cannot avoid them
altogether. The developing and emerging economies are instead in a transitory
stage in the evolution of their official statistical systems. Measurement errors,
however, remain a matter of deep policy deliberation among macro accountants
to date. One of the kinds of errors generally encountered by official statistical
systems is subjective errors. Errors in recording data, using data from dissimi-
lar sources and incorrect aggregation of unit-level information are examples of
objective errors that occur mainly due to factors outside the control of the data
collectors. However, a considerable amount of imputation, interpolation and
extrapolation is used to adjust and organize the national income accounts before
they are presented. These accounts are very keenly observed both economically
and politically, and hence, the possibility of subjective errors is problematic.
These errors cannot be entirely eliminated in EMEs until their economic struc-
ture shifts completely towards the organized sector.29
198 Expert opinions and contributions
Financial services versus the tangibles in the estimation of output
The identification and quantification of unit-level information, i.e. the infor-
mation on the so-called elementary units, is relatively more accessible in the
case of goods with well-defined value chains and reliably differentiated pro-
duction stages. Due to a clearly observable production process, determining
the inputs and output at different stages in the production chain is possible in
the case of Goods. This is a prerequisite to deriving estimates of value-addition
at elementary levels. These estimates are then further aggregated to derive
larger estimates at the industry, sector and national levels. Such a possibility is
constrained in the case of services because identifying the value chain in the
production of services is complex and even indeterminate in some cases. The
financial services industry is perhaps the best example of this aspect. Universal
banking has been an essential feature of market-oriented EMEs such as Brazil,
Russia, India and others since the advent of financial globalization that roughly
gained momentum in the mid-2000s. Multiple services are being provided
through the same banking entity, and these services also cut across many other
non-banking services. This makes it challenging to locate a single output of the
banking firm; rather, its output is a combination of various activities that it per-
forms simultaneously. This lends considerable dynamism to the very concept
of banking, insurance and other financial services’ outputs. A static data collec-
tion methodology that stresses identifying single output rather than consider-
ing a plethora of simultaneous activities as a measure of output will be unable
to derive reliable estimates as far as financial services are concerned. Financial
services need to be treated with as much aggregation as plausible given the
context. Deriving highly disaggregated estimates, for example, the output from
the merchant banking services of banks, will not do justice to this inherent
dynamism like financial services. The interlocked nature of the production of
financial services in particular and the service industries in general, does not
lend itself quite well to the typical neoclassical production empirics. Thus the
traditional measures of the output of financial services must be looked at simul-
taneously as inputs and outputs depending on the context of output analysis.
The fee generated by commercial banks through merchant banking services is a
measure of output so far as the total services rendered by a bank are concerned.
However, they might also be theorized as inputs if one is interested in look-
ing at the total factor productivity of a particular bank’s retail banking segment
because banks may channel the surplus funds from their corporate segment
into their retail segment to generate loans. This is probably one reason why the
literature on the measurement of banking and insurance industry output has
been much more diverse than the estimation of individual firm-level output.30

Differences in economic structures and international comparisons

Consider India and China. Historically, both these countries have had differ-
ent occupational structures of their aggregate output. The Indian economy has
generally seen a larger share of services in its GDP, while China has grown on
 Expert perspective on output measurement in banking, insurance 199
a more extensive manufacturing base. Furthermore, both these economies cur-
rently have quite the opposite policy priorities. While the Indian policymakers
are incentivizing an increase in the share of manufacturing to GDP, China is
pushing its service sector to contribute an increasing share to the same. Since
1980, there has been a clear difference in the service sector’s contribution to
their respective GDPs. India has experienced a slowdown in the share of services
to GDP, while China has experienced the opposite. Both economies are well-
regulated, and this difference in the way their service sectors have changed since the
1980s partly reflects the differences in their development policies. Thus, are these
economies comparable even after the customary adjustments in conversion to
common currency and deflation? With very different structures, macroeconomic
data require more adjustments for standardization. As discussed previously, the
more such adjustments are made, the higher the probability of subjective errors.
This is only one example of the ample number of differences in the eco-
nomic structures across emerging economies. The structural differences in the
economies are more prominent in the case of emerging countries mainly due
to large differences in the level of economic development and the kind of
resource endowments with which these countries began their development
journeys. These differences are multiplied in the case of their financial services
industry, where the EMEs have huge differences in the depth, development and
penetration of this industry. In contrast to this, the advanced economies present
a more coherent structure with broadly similar levels of financial development,
and hence their data lend themselves to more robust international comparisons.
These observations also highlight another essential feature of the EMEs: their
higher dependency on the services industry for propelling economic growth
compared to the commodity sector.

Input-output measurement in financial services, including the

FISIM methodology
Measurement of the inputs and output of any industry are necessary elements
in estimating their economic performance parameters such as their productiv-
ity, efficiency, profitability and others. In the case of services, particularly in
financial services, the distinction between the two is quite subtle and generally
a challenge for any economic analyst. A common approach to circumvent this
problem is to estimate the output produced by the financial service firms, such
as banking and insurance companies, using the data on salary and wages paid to
the employees of these firms as an estimate of the output or rather the value-
added of the firm. This dependency on factor market data to estimate a prod-
uct market variable is probably one of the most frequently employed strategies
by national income accountants across the developing economies. However,
a sizeable shadow financial services segment can make such an approach to
output measurement pretty unreliable. However, the EMEs are experiencing a
continuous increase in the share of the organized segment to their value-added,
especially for financial services.31
200 Expert opinions and contributions
Moreover, financial services, particularly banking, do not produce only one
kind of output but rather produce many different services that interchange
their roles as inputs and outputs depending on the value chain stage one is
looking at in this industry. For the same bank, its deposits are a product it must
sell and promote with various services and features to its customers. At the
same time, it must also use those deposits to produce loans and advances that
its customers will utilize. From a marketing perspective, deposits are as much
of a product to be sold as are various credit products. Hence, it is impossible to
provide an ideal definition of input and output for a banking firm. Instead, one
should adopt a flexible approach and measure inputs and output depending on
the value chain stage that one is analysing.
Another actively debated matter is the preference given to the FISIM meth-
odology in imputing the intermediation component of the value-added by the
banking sector. While the FISIM was prepared to be used as a last resort, the
lack of theoretical clarity on the input-output distinction has somehow made
FISIM the first preference among national income accountants. Given the cur-
rent state of the official statistical systems of major emerging economies, it does
not seem that the FISIM will lose its dominance anytime soon. Theoretical
developments on the nature of production in financial services are required
before traditional or innovative approaches are employed to estimate the inter-
mediation services supplied by the banking sector.
Despite these constraints, there are some alternatives to choose from as far
as the measurement of banking output is concerned. Profit is one of the vari-
ables used to measure banks’ production. Again, this approach has its limita-
tions, mainly because of the dominance of public sector banks in major EMEs.
Public sector banks are inherently driven by welfare motives rather than profit
maximization motives, which is the primary aim of private sector banks. In
this case, the changes in profit levels will not reflect the true extent of value-
addition by the banking sector because a reduction in profit of public sector
banks does not imply a reduction in the value-added by them. On the contrary,
in economies with a bank-driven financial system design, a decline in profits
may signal these banks’ developmental role at the cost of their profit. Another
approach is the use of wages paid to the employees of the banks as a measure of
bank output. This approach is used chiefly by economies with a large organ-
ized labour market for financial services, making it easier to observe the market
wages to estimate output.
Similarly, some alternative measures of insurance output are also used actively
in academic and policy analysis. Insurance payouts during claim settlement are
one possible candidate in this regard. Suppose the underlying theoretical belief
is that risk coverage is the main output of insurance firms. In that case, the
claim settlement amounts paid by them are an appropriate measure of their
output. However, there is an ex-ante and an ex-post dimension to insurance
output. In an ex-post sense, claim settlement payouts are a valid measure of
insurance output, but in an ex-ante sense, they are not. In this case, the under-
lying theory would be that what insurance firms promise to cover is the output
 Expert perspective on output measurement in banking, insurance 201
produced by them. Then the use of the total sum insured or assured would be
a more relevant candidate. Both these approaches are challenging to apply in
real time. Hence, the SNA recommends the amount of premium collected as
a measure of insurance output and appropriate adjustments for the other non-
insurance functions performed by the insurance firms. Currently, the premium
approach is recommended by the SNA, but the forthcoming revisions of the
SNA might allow recommended procedures for other approaches, too, I hope.

Some frontier issues in financial services output measurement

Emerging economies are characterized by widespread public-sector involve-
ment in the financial services industry. The private and public firms have very
different economic objectives. The very nature of their production differs con-
siderably. The output of privately owned financial services firms can be exhaus-
tively estimated using their observed market behaviour. However, this may not
be the case with the public sector firms which the government majorly owns.
The output measurement strategy for both sectors should be different even
when they belong to the same industry. Thus, separate output accounting pro-
cedures for public and private financial services firms are a need of the hour.
Observed market data is generally not representative of the true value added
by the public firms. Public sector firms also produce non-marketable output
that cannot be measured by simply using the financial market data. The FISIM
approach too relies on observed market data to estimate the reference rate,
which is the heart and soul of this methodology. However, it does not account
for the non-marketable portion of the output produced by the public sector
banks, for example. Estimating output using the neoclassical production theory
relies heavily on observed market data such as the reference rate in FISIM,
wages paid to employees by banks and insurance firms, and the profit earned
by these firms as per their financial statements, etc. The output of public sector
firms can only partially be estimated using the typical neoclassical production
framework. More profound deliberations on these dimensions are needed.
202 Expert opinions and contributions
Pronab Sen
Brief Profile: Dr Pronab Sen is the Programme Director for the Interna-
tional Growth Centre’s (IGC) India Central Programme. He received his PhD
in Economics from Johns Hopkins University, specializing in open-economy
macroeconomic systems, international economics and public finance. He has
served as the first Chief Statistician of India, acting as the functional and tech-
nical Head of the national statistical system in India and Secretary, Ministry of
Statistics & Programme Implementation, Government of India (2007–2010).
He recently served as the Principal Adviser, Power and Energy, at the Gov-
ernment of India’s erstwhile Planning Commission. As a representative of the
Planning Commission, he was the principal author and coordinator of the
Mid-term Appraisal of the Eighth Five Year Plan, the Ninth Five Year Plan,
the Mid-term Appraisal of the Ninth Five Year Plan, the Tenth Five Year Plan
and the Mid-term Appraisal of the Tenth Five Year Plan.

Emerging economies and the system of national accounts

The SNA is a set of accounting principles and recommendations laid down by
the United Nations Statistics Division (UNSD). The divergence between the
expectations of the UNSD from the official statistical systems of the member
nations and the actual data assimilation systems currently in place in major
EMEs is well known both in the international policy circles and in the general
academic discourse. The critical issue here is that the ability of member coun-
tries to measure their national income optimally depends not on the quality of
these accounting principles but the availability of unit-level economic data.32
An essential advantage of the SNA manual is its flexibility for member coun-
tries to adopt computation methodologies suitable to their data availability.
However, despite this flexibility in methodological approaches, the differences
in the availability of high-quality unit-level data are vast both among develop-
ing and developed nations on the other and among the developing nations
themselves. Thus, the pertinent question is not whether the SNA is sufficient
to capture the intricacies of output measurement but whether emerging coun-
tries have the infrastructure to collect the required elementary data in the first
place. Methodological innovations can always be undertaken if sufficient data
inputs are available. EMEs currently face the challenge of generating statistically
“good” unit-level data rather than their ability or inability to adopt the SNA
in the whole spirit.
The advanced economies have an inherent advantage when it comes to col-
lecting unit-level data on the economic activities of various industries, par-
ticularly in the case of financial services. The corporate sector accounts for
the bulk of its economic activities. This makes their balance sheet data quite
representative of the actual economic value addition at the ground level. In
the case of market-oriented EMEs such as Brazil, India, Indonesia and others,
the non-corporate sector plays a sizeable role. Its contribution to the aggregate
 Expert perspective on output measurement in banking, insurance 203
value added differs across EMEs, but it dominates in several large emerging
economies. This lends their corporate sector data unrepresentative of the actual
value addition undertaken by the industries, including the financial services.
Another issue that arises in developing economies is that of the inter-sectoral
differences in data collection systems. Agriculture data in India, Brazil, and
possibly China are collected considerably longer than in other industries, given
the historical dominance of agriculture until structural changes took place in
these economies. Hence, the quality of non-agricultural output data in econo-
mies like India is a significant constraint.33 The UNSC can only recommend
the best practices. Its execution is local and determined by domestic factors
outside its control. However, it may be noted that as far as financial services
are concerned, the large bulk of aggregate value-addition comes from the cor-
porate firms and thus, the issue of accounting for non-corporate sector output
is limited only in the cases where the unorganized sector has a sizeable domi-
nance in the provision of financial service.

Under- and overestimation of financial services output

under economic shocks
In regular times, the method used to generate estimates of the non-corporate
or the unorganized or the informal sector of an economy work well. This
is generally in the form of extrapolation of the estimates from the corporate
sector to the non-corporate sector. Hence, without any external shocks, the
output measurement of the unincorporated sector is broadly representative of
the true picture. This representativeness breaks down when any kind of shock is
propagated through the economy. An economic shock such as the COVID-19
pandemic that has adversely affected most developing economies can impact
the corporate and non-corporate sectors. Both the general observations and the
economics literature suggest that this is the case. In a macroeconomic model
where international economic shocks propagate freely throughout the nooks
and corners of the economic structure, output measurement methodologies
that proxy the value addition in the non-corporate sector break down. With
differential impacts of such economic shocks and the lack of precise knowledge
about the essential econometric parameters of the effects of such shocks on
the non-corporate sector, extrapolation and interpolation methods will fail to
capture the true size and changes in the output of the non-corporate sector.
One can readily apply this logic to the financial services industries also. These
problems are further intensified for financial services that provide intermedia-
tion services. Here, not only the estimates of the direct services of the non-
corporate sector are distorted due to shocks, but also the estimates of their
indirect intermediation services as generally measured by the FISIM approach.
Such shocks result in underestimating and overestimating the output of
industries that comprise a large share of the non-corporate sector. For exam-
ple, in the case of the COVID-19 shock, developing countries probably over-
estimate the value added by financial services as the true adverse impact on
204 Expert opinions and contributions
the non-corporate financial services sector is largely missing from the policy
arsenal. Similarly, in the case of the 2007–08 financial recession, possibly the
countries were underestimating their GDP mainly because the total impact of
the same on the aggregate economy was overstated by the lack of knowledge
about the structural parameters of the unorganized sector.

Defining the output of services versus other industries

Output measurement in the services sector has another challenge to overcome.
While the definition of a product is quite tightly defined in the case of non-
services industries, that is not the case for the services sector. What do banks
actually produce? Do they produce deposit accounts held by lending customers
who consume the deposit account services over many years? Do banks produce
loans and advances that the borrowing customers consume? Do banks produce the
transformation of deposits into credit? Similarly, do insurance firms produce
coverage of risk that provides possible protection to the policyholders? Do they
rather produce claim settlement services when policyholders actually face a loss?
Do banks and insurance firms produce risk transformation services? Or do they
rather produce efficient financial asset allocation through the financial markets?
Such questions complicate the nature of the output of financial services.
With a lack of clear-cut tangible definition of what they produce, researchers
and analysts have a lot of freedom to tweak the specification of the production
function as per their subjective understanding of what banks and insurers pro-
duce. This makes estimates of total factor productivity, efficiency, profitability
and other performance parameters difficult to synthesize for informing policy
formulation. The specification of the output of financial services is further
complicated by the indirect intermediation services that they supply. Not only
the intermediation services but many of the positive and negative externali-
ties of their existence are generally not well-captured by both the emerging
and the advanced economies alike. Such estimation requires knowledge of the
parameters of a structural macroeconomic model that can accurately specify the
coefficients of the relationship between the aggregate economy and the indi-
rect, intermediary and non-marketable output of the financial services sector.
Such an expectation is far from true today. Even the best macroeconometric
models used today by the OECD, EU or even the US do not account for such
issues. In large part, this may be because their corporate sector is the largest
contributor to the aggregate output. Thus, resolving these issues is on emerg-
ing economies and their blocs such as the BRIIC, ASEAN, etc. With proper
support from the UNSC, these blocs can work deeply on these matters and
possibly develop more innovative methods to account for the non-corporate
sector in their national income accounts.
These problems also make it challenging to generate a well-defined price
index for financial services. Index measurement is undertaken to account for
the movements in a variable over time. With the lack of any commonly agree-
able definition of financial service output, measurement of changes in the price
 Expert perspective on output measurement in banking, insurance 205
of such services over time will be biased and untenable from an economic
point of view. Technological shocks are perhaps the strongest and quickest in
the financial services industry. For example, the same service, the home loans
disbursed by banks, does not remain the same over even short time intervals.
In this case, essentially, a sectoral measure that can be defined so that its value
accounts for such dynamic changes in the composition of the output can do
justice. Problems in the price measurement of financial services are as compli-
cated as those on the side of their output measurement.
Another dimension in this context is the implicit pricing of various
non-intermediation services provided by financial service suppliers, par-
ticularly banks. Banks offer a whole set of direct and non-fee-based and non-
intermediation services. Structurally, the share of such services is different for
different emerging economies. However, most EMEs have a banking sector
that thrives on such services due to high competition. With the exception
of perhaps China, most EMEs have both public and private banking firms in
continuous competition for market share. Such services are an essential tool for
the marketing strategists of banks to attract and retain customers. These services
are not free of charge but are rather priced implicitly. Such implicit pricing
makes it very difficult to measure an optimal Fisherian34-style price index. In
this case, with reference to output measurement, the FISIM shall account for
the implicitly priced intermediation services, but it does not account for such
services that are charged implicitly but are not intermediary in nature.

Dual intermediation roles of banks and the input-output debate

As noted earlier, there are many unresolved questions about what banks pro-
duce. However, it is agreed that banks are intermediaries. They transform
deposits into credit. However, while doing so, they also transform risk expo-
sures and redistribute risks much more efficiently compared to the situation
that would have prevailed in their absence. Depositors are assured of a return
irrespective of how their funds perform in the loans market. Indeed, this is
not always the case, as continuously loss-making banks with low net inter-
est margins can be squeezed to the extent of insolvency. However, given that
most EMEs have a well-established banking sector regulator,35 the occurrence
of such an unfortunate event at a sectoral level is complicated. This dual inter-
mediation role of banking firms, namely the intermediation as transformers
of deposits into credit and transformers of individual risk exposures into more
efficiently redistributed risk exposures across depositors, requires more innova-
tive approaches than the FISIM approach as in place right now.
Yet another debate has raged on the output measurement of banking ser-
vices. It is the so-called input-output dilemma in the banking output. Are
deposits inputs or output? This is a particularly persistent matter of debate. This
dilemma is not a dilemma at all, as the balance sheet of a bank very well reveals
the truth. Deposits are liabilities, and loans and advances are assets. Banks are
essentially liability-asset transformers. Their core source of income is the extent
206 Expert opinions and contributions
to which they can extract net interest margins while undertaking this transfor-
mation role.
Increasingly, countries are pushing forward the digitalization of their bank-
ing and insurance industries. The COVID-19 pandemic has further given an
unanticipated push to this agenda. If such a trend continues, the issue of
accounting for non-corporate output and the under- or overestimation of
financial services output might very well be overcome to a large extent.

Finance-growth nexus in emerging economies

Does finance cause economic growth, or is it the other way round? Or is it
two-way? This is a classic matter of debate in development economics. Econo-
metric evidence points toward a two-way causal relationship. Theory suggests
that finance must react to the needs of the real sector. In terms of my experi-
ence, these questions are futile. The very nature of an economy is circular. It is
further complicated by highly complex webs of inter-sectoral and intra-sectoral
feedback mechanisms that the economists do not fully understand yet. With
growth rate maximization being imperative for policymakers in developing
nations, it is clear that whichever channels can be used to increase the growth
rate must be employed.
A well-developed financial system is a necessary but not sufficient condi-
tion for economic growth. A nationwide banking system, for example, can
shift funds back and forth throughout the regions of the country depending
on where the funds are being generated and where they are being demanded.
This can lubricate the functioning of the real sector and allow increases in the
levels of “real” production and productivity. However, financial efficiency in
itself does not create the factors required for economic growth, which gener-
ally emerge from the supply side. While the financial system is a vital sup-
plier of financial capital, the ultimate supplier of human and physical capital
in terms of other financial resources is the real sector. With high levels and
quality of human and physical capital, finance may induce growth, but without
them, finance and growth may not display any meaningful causal relationship
whatsoever.36

Some notes on insurance output measurement

Insurance firms produce risk coverage. The amount of premium underwrit-
ten by them is generally a reliable measure of the amount of risk they cover.
The non-corporate sector does not play much of a role here when it comes
to insurance. Historically, insurance has been supplied by the organized sector.
Thus the balance sheet data of these firms is well-representative of what they
produce, unlike in the case of banking. Moreover, insurance companies may
also provide risk coverage to banking firms, and in a sense, insurance firms
share the risks of banks. Insurance firms also perform intermediation func-
tions by creating a pool of funds from the insured and investing it in financial
 Expert perspective on output measurement in banking, insurance 207
markets. However, it does not call for a FISIM-style measurement problem
because the SNA, 2008 already accounts for such dimensions.

Some frontiers of banking output measurement

A final issue to be highlighted is that the nature of the output of banks when
they lend for working capital versus when they lend for fixed capital is very
different. The underlying production functions themselves may be considered
non-aggregative. Simply adding the output data from their working capital
lending with long-term lending may actually be economically incorrect. Even
if such is the need, appropriate aggregation factors need to be derived that can
give some meaning to such an aggregation. Short-term lending by banks is
less of a concern as far as their intermediation role is concerned. Banks do not
generally lend short-term working capital loans to maximize their net interest
margins. These are lent more so based on the direct interest they charge and
the various fee-based services they lend. Such loans are not going to be a part
of the long-term assets of a bank, and hence their longer horizon decisions will
be more focused on the long-term loan portfolio.
Similarly, not all deposits received by banks are long-term. Current account
advances and even Savings Bank and Term deposits are not guaranteed to be sta-
ble for the banks. Variations in the withdrawal-deposit behaviour of customers
are inevitable with changing economic conditions. Banks thus require a reliable
long-term source of funds to underwrite long-term loan contracts such as that
home loans in the retail segment and possibly project financing in the corporate
segment. The extent to which short-term versus long-term funds contributes
to the banks’ intermediation services is an important area of further research.
208 Expert opinions and contributions
R. B. Barman
Brief Profile: Dr R. B. Barman, former Executive Director, Reserve Bank
of India, holds a Master’s degree in Statistics and a PhD in Industrial Eco-
nomics. The main areas of his professional interest include Business Informa-
tion Technology and Payment Systems. Dr Barman was Vice Chairman of
Irving Fisher Committee on Central Bank Statistics, Bank for International
Settlements (BIS) and a Member of the International Data Forum. He has also
served as an Advisor to the National Payments Corporation of India, which is
entrusted with the responsibility of setting up a state-of-the-art retail payment
system for India. He spearheaded the Business Intelligence initiative of the
Reserve Bank of India, which in 2001 was one of the earliest such initiatives.
The subject orientation and the data model thereon was a novelty, which many
central banks have followed thereafter. Dr Barman has contributed richly in
the area of Official Statistics – particularly on price and output statistics. His
contribution in this area spans several articles. Post-retirement, Dr Barman was
a Visitor at Indira Gandhi Institute of Development Research and Adviser
(Research), Indian Institute of Capital Market. He is also the Adjunct Faculty
of the C. R. Rao Advanced Institute for Mathematics, Statistics and Computer
Applications, Hyderabad. He was the former Chairman of the National Sta-
tistical Commission and former President of The Indian Econometric Society
(TIES). Dr Barman received the prestigious National Award in Official Statis-
tics – 2021, Government of India.

Theoretical and empirical issues in the system of national accounts of

emerging economies
The foundation of the System of National Accounts was laid down by the 1947
Report of the sub-committee on National Income Statistics of the League of
Nations Committee of Statistical Experts entitled “Measurement of National
Income and the Construction of Social Accounts”. Especially important were
the recommendations made by Richard Stone in the memorandum to the
same report. This document provided the basic building blocks for estimation
of national income accounts that most the nations today are widely undertaken.
A noteworthy objective of constructing these accounts was to use these data
to understand the welfare implications of development policy interventions.
In the post–World War II era, many nations were burdened by the aftermath
of the war. Massive development works were required to re-build these coun-
tries.37 Accordingly, relevant data were required to understand various policy
interventions’ expected and actual effects. The SNA provided the basis for a
systematic collection and aggregation of data to obtain vital macroeconomic
estimates essential for development policies and promoting a higher economic
growth rate. However, despite the focus of this report on collecting information
on elementary38 data, aggregation remained the ultimate purpose of its recom-
mendations. Aggregation at the scales at which national income accountants
 Expert perspective on output measurement in banking, insurance 209
undertake is extensive and presupposes many assumptions, only after which
the aggregates can make any economic sense. One of these assumptions is that
of symmetrical economic distributions,39 which is assumed in the aggregation
procedures. For example, actual economic distributions are generally skewed,
particularly in emerging and developing economies. Adjustments in the ele-
mentary data for this skewness during aggregation into higher levels of groups
such as industrial-, sectoral- and national-level estimates are not undertaken
in national income accounting procedures laid down by the SNA. This was a
problem with the 1949 manual, which was the first one, and it continues to
be a limitation of this manual to date. Aggregation without consideration of
micro-level economic distributions ends up suppressing the non-normal skew-
ness and kurtosis in the underlying economic data.
The second aspect of the SNA procedures is the economic reductionism
it implies in the macro accounting identities that it pre-supposes for national
income estimation. Accounting identities are meant to resolve accounting
problems and enable official agencies to construct statistically coherent esti-
mates. However, the use of these accounting identities as representations of
actual economic processes distorts the true underlying structure of the econ-
omy and possibly creates policy biases that can prove to be very costly for the
welfare economic terms. While national income accountants do not advocate
the accounting identities as an accurate representation of the actual economic
processes, the popular political and economic discourses use such accounting
ideas to interpret the success or failures of economic policies.
The third crucial theoretical issue is the blow lent to macroeconomic aggre-
gation by the celebrated Lucas Critique, as Lucas (1976) explained. Lucas notes
that “the unquestioned success of the forecasters should not be construed as
evidence for the soundness or reliability of the structure proposed in that the-
ory” (p. 23). Rightly so, this criticism led to the emergence of the Dynamic
Stochastic General Equilibrium (DSGE) models for macroeconomic policy
formulation. Many nations are currently building large-scale macroeconomic
models for their economies to undertake policy simulations and derive the
possible impact of proposed interventions on the economy and its compo-
nents. Fully understanding the microeconomic foundations of aggregate data
employed for macro policymaking is critically needed, especially with nations
pushing for higher economic growth rates. As Lucas pointed out in his sem-
inal work, without understanding how individual economic agents interact
and their implications for macro policy, a well-specified and well-estimated
macroeconomic model will not possess much economic meaning whatsoever.
Not only the microeconomic foundations but inherently stochastic model-
ling of economic processes is required for effective macroeconomic policies.
The Walrasian General Equilibrium model provides a solid theoretical tool
for analysing actual economic interactions. However, economic behaviour is
always undertaken under conditions of uncertainty, and this requires account-
ing for various probabilities associated with the modelled economic behav-
iour. Empirical work requires such an approach because the actual economy
210 Expert opinions and contributions
functions on stochastic footings rather than deterministic behaviour, which
may be easily forecasted using an econometric model. The fourth issue in the
present context is the emergence of Big Data, Cloud Computing and Artificial
Intelligence–based data collection and analysis technologies. With large-scale
disaggregated and unit-level data now available in an increasing number of
emerging economies, the need for cleanly defining the micro-foundations of
macro models is all the more urgent. The need for a stochastic-kind of Walrasian
general equilibrium models to explain macroeconomic phenomena such as
capital accumulation, aggregate GDP growth, etc., is the most pressing today
than any other policy challenge.
The SNA, 2008 is now under intense pressure for revision, especially after
the resolutions passed in the last few years by the United Nations Statistical
Council (UNSC), particularly in 2014. It is expected that the future version of
the SNA shall incorporate many of these issues. In particular, for sound meas-
urement of output and income, three issues need to be urgently addressed by
the UNSC for the revised SNA: first is the issue of economic heterogeneity in
various emerging economies, particularly economies that have high occupa-
tional diversity; second is that of non-linearity in economic relationships that is
observed in actual data but is generally avoided in econometric modelling; and,
third is the issue of path dependency which is quite well-recognized in mac-
roeconometric modelling nowadays. These characteristics of actual economic
data demand appropriate adjustments in aggregation procedures of the SNA,
and hopefully, they will be fulfilled in future revisions.

Nature of financial services output

The fundamental purpose of macroeconomic accounting is the optimal valua-
tion of economic activities undertaken by various firms, industries, sectors and
the aggregate economy. The unique contribution of each economic agent and
factor in the production process is the fundamental basis of macroeconomic
output measurement. In these terms, the nature of output is similar for goods
and non-intermediation services. Both use explicitly defined inputs to produce
explicitly observable output. In the case of intermediation services such as
banking and insurance, especially for banking, the output cannot be estimated
simply using the observable data.
Take the FISIM approach that relies on interest rate spreads between lending
and deposit rates to proxy the amount of intermediation output generated by
the banking sector. First, it is a sectoral or rather an industrial-level estimate.
It is tough to distribute FISIM across each constituent bank. Second, larger
spreads should imply larger value-additions. Still, if that is the case, then micro-
finance institutions in countries such as Bangladesh, Brazil, India and oth-
ers must be contributing the highest value-addition and hence should receive
large budgetary allocations in development plans. However, that is not the
case because the amount of this spread is low, but the financial risks and costs
involved in the microfinance industry are pretty large. What this implies is that
 Expert perspective on output measurement in banking, insurance 211
measurement of intermediation services through an interest-margin approach
(which inevitably relies on a reference-rate methodology) can capture the true
quantum of intermediation but only limitedly. Third, the dependency between
interest rate spreads and the risk profile of assets required to achieve those
spreads demands special attention in the output measurement for intermedia-
tion services rendered by banks. For example, larger spreads are achievable only
when banks presume higher risks while lending loans and advances. But such
an approach can boomerang the industry and perhaps the economy itself. See
for illustration the sub-prime crisis that emerged from the high-risk lending by
banking institutions in the US to maximize the amount of their spreads.
The short note above highlights that statistical measurement of output under
the macroeconomic accounting framework does not explain by itself the extent
of value added by a firm or an industry in the economy. The implications of
what is produced by these entities are equally important. In the case of interme-
diaries like banks, this aspect is further blurred due to indirect services rendered
by them requiring indirect estimation but using directly observable data. The
use of observable data for an inherently implicit activity needs a more rigorous
approach than the reference rate version of FISIM, which is currently used.
Lastly, a widespread debate in the discourse on banking output measure-
ment is identifying the inputs and outputs of a banking production function.
From the perspective of direct services rendered by banks, this issue does not
pose much of a problem for their output measurement. The observed data on
the revenue generated by the banks for these services is a reliable indicator of
value-added by the bank through these direct fee-based services. However, the
production function becomes quite complicated when it comes to intermedia-
tion, as the observable bank-level accounting data is insufficient to capture the
full extent of the same. In this case, a reference rate-based FISIM approach is
used to measure the difference in the deposit and lending rates of the banks
compared to a risk-free reference rate. In this framework, deposits act as inputs
and credit lent by the banks as output. But the FISIM suggests that deposits
in themselves generate returns for the depositors and hence can be considered
output in that sense. The distinction between input and output is more flexible
in banks’ intermediation services than in direct fee-based services.

Comments on the FISIM approach in measuring banking output

FISIM is a logically coherent framework. It is internally consistent also. How-
ever, as noted above, the equalization of value-addition with interest-rate
spreads is not empirically sound. As a statistician working on policymaking
issues for several years, I firmly believe that the current FISIM approach needs
revision to better measure the intermediation services provided over and above
those proxied by interest-rate differentials. One approach to account for this
could be to design macroeconometric models that can analyse the impact of
the movements in interest rate spreads on economic activities and thereby
adjust the estimate of FISIM for such estimated parameters. In the era of Big
212 Expert opinions and contributions
Data and AI-driven software capabilities, it is not unreal to expect such models
to emerge in the coming years.

Possible future developments in the output measurement

of financial services
Stochastic dynamic models will and must increasingly occupy the traditional
national income estimation procedures. FISIM estimation should be made
more rigorous by adopting DSGE models that can dynamically adjust the tra-
ditional estimates for their complex feedback effects on other sectors. Such an
approach also demands rigorous input-output tables for emerging and even
advanced economies that can explain the interdependencies at higher levels of
disaggregation. With large-sized unit-level banking data and the emergence of
Big Data analytics in banking and finance, such an expectation is not very far-
fetched from reality. Small sample constraints, for example, are not much of an
issue now for advanced economies due to the large volume of unit-level data
being generated for the financial services and other industries of these econo-
mies. However, this is still a constraint for emerging economies, such as China,
that either do not reveal unit-level data or do not collect it altogether due
to weak statistical systems. The SNA, 2008 puts a lot of expectations on the
official statistical systems, and this is a significant problem for developing and
emerging economies where data collection methods are not refined enough to
suit the needs of such improvements. Brazil, India, Russia, Indonesia, Thai-
land, and other top emerging economies need to drastically overhaul their data
collection systems. In particular, these nations need to produce high-frequency
unit-level data for all major industries if the above improvements are adopted
in the coming years. Hopefully, the SNA-2025 is underway, and such issues are
expected to be included in the next revision of the SNA.
 Expert perspective on output measurement in banking, insurance 213
Ram Pratap Sinha
Brief Profile: Dr Ram Pratap Sinha is an Associate Professor of Economics
with more than 32 years of (undergraduate and post-graduate) teaching expe-
rience. He is presently posted at the Government College of Engineering and
Leather Technology. He has several journal publications to his credit and has
contributed to 19 edited volumes, including on various issues in the perfor-
mance analysis of insurance and banking industries. Besides, he is the author of
two books. He is a reviewer of many reputed international and national jour-
nals in the area of Economics and Finance. He has taught in numerous Faculty
Development Programmes and workshops in Data Envelopment Analysis and
Econometrics, focusing on various performance measurements of the financial
services industry of major emerging economies.

Insurance production as a multi-stage activity

What do insurance firms produce? This has been a question of policy and aca-
demic debates since long. As the insurance industry made its way into the Sys-
tem of National Accounts (SNA) as an explicitly-recognized industry with its
own unique input-output relationships, it became a well-recognized fact that
the theoretical notions adopted from the neoclassical production theory needed
an overhaul, and the estimation of insurance output required newer approaches.
Empirically speaking, estimation of insurance production function is more popu-
larly undertaken using single-stage models, which posit a linear one-way input-
output relationship. In such empirical models, production is theorized with a
clean distinction between inputs and outputs. This representation may not always
be correct and can have severe consequences if the estimates of insurance per-
formance parameters40 are used in firm-level or macroeconomic policymaking.
The theoretically more robust approach is to consider insurance as a two-stage
activity in which an intermediate output divides the total production function
into two stages – the first wherein inputs are used to produce an intermediate
output and the second stage wherein the intermediate output is used as an input
to generate the final output. In empirical terms, the first stage is where tradi-
tional labour and capital inputs are utilized to produce an intermediate output –
namely, the insurance premium collected by the firm. After that, in the second
stage, the insurance premium collected is utilized for risk pooling and financing
claims, thereby providing risk coverage, which is the final output. As elucidated
in Figure 7.3, Stage 1 is a process which can be typically accounted for using a
traditional production function approach. Inputs are processed and transformed
into an output, i.e. the amount of premium collected by the firm. However, in
the case of the insurance firm, value generation does not cease here. In Stage 2,
the firm further transforms the premium collected into the final output, which
various stakeholders consume in different forms. The policyholders consume
the risk coverage services provided by the firm through the pooling of risks and
losses; the shareholders consume the changes in their financial wealth depending
214 Expert opinions and contributions

Initial Inputs Stage 1 Stage 2

•Labour Intermediate Output Final Output
•Capital •Premium collected by •Risk Coverage
•Technology, and the Insurance firm (policyholders)
others •Financial returns
(shareholders)
•Risk allocation
(Macroeconomy)

Figure 7.3 Theoretical production function of an insurance firm

Source: Analysis of the contributing expert.

on how the insurance firm is performing in and outside the stock markets and
the aggregate economy obtains an efficient risk allocation.41
Input measurement is the first matter of concern in this context. Account-
ing for the inputs in official data is difficult, particularly in developing econo-
mies. The data infrastructure is often weak in their case, and the measurement
practices are often inconsistent with the SNA requirements. There are many
statistical practices too that render the data on inputs difficult to use for empiri-
cal studies. Furthermore, insurance is sold through many channels. Thus, many
variables across each major input need to be accounted for in the production-
based performance studies. Insurance branches, its employees, online plat-
forms, third party agents and the Bancassurance model are the major channels
through which insurance products are sold in most emerging economies today.
Accounting for these channels in terms of official data on the financial services
industry is a challenge for many countries. This is mainly due to the lim-
ited maturity of the data collection systems and a sizeable unorganized sector
that cannot be easily incorporated into the official statistical design. The best
approach is to use proxy measures for insurance inputs, but that may not reflect
the underlying variables robustly. Capital, for illustration, can be measured as
the number of branches of the insurance firm, but this does not account for
other non-branch channels through which products are sold and which are
essential capital inputs for the firm in providing its services.42
Output measurement is another and more serious challenge in the insurance
industry. There are multiple perspectives through which the output of an insur-
ance firm and the output of the entire insurance industry can be viewed. From
microeconomic and macroeconomic points of view, it is possible to consider
the output of insurance from the perspective of the policyholders, insurance
firms, intermediate agencies and policymakers. Locating a standard measure of
output that can reflect all these perspectives is difficult, primarily since insur-
ance provides different utilities from different perspectives and measuring these
using a single variable will not reflect all these different value-additions pro-
vided by insurance firms. Consider the case of insurance firm efficiency. Effi-
ciency models can be estimated from the perspectives of different stakeholders
 Expert perspective on output measurement in banking, insurance 215
such as the policyholders, shareholders, the employees and a host of others.
The measures of inputs and outputs shall be different when looking from these
different perspectives. For policymakers, efficiency measurement needs to be
undertaken within a macroeconomic perspective so as to see the larger picture
and the various interconnections among stakeholders. In their case, merging all
these perspectives is necessary, thus demanding large-scale data with an enor-
mous scope than the data used for microeconomic and intra-industry studies.
The SNA provides one possible framework on this account. However, it relies
on single variables for measuring each of the inputs and output. Such a univari-
ate approach is computation-friendly and statistically easier to analyse but will
provide a limited view of what and how insurance firms produce their outputs.
It will not be incorrect to presume the existence of multiple outputs in a typical
insurance firm’s production function.43

Alternative approaches to the empirical estimation

of insurance performance
Any empirical work on performance44 measurement of a firm, industry or sec-
tor requires reliable estimates of both inputs and outputs. Models that presume
single-output and single-input are generally not feasible in a multi–inputs-
outputs industry such as insurance. Insurance firms can produce multiple outputs
by using the same set of inputs. This complicates determining the input-output
relationships while measuring insurance performance parameters. Empirical
models based on data envelopment analysis (DEA), dynamic network models
and two-stage econometric models infuse dynamism in the estimation process
and allow modelling of multi-input–multi-output production processes to a
large extent. However, the data requirements for such econometric methods
are considerably strict, and EMEs, even the top-performing ones, cannot satis-
factorily fulfil these requirements. A good amount of statistical adjustments to
existing data are required before being used for such econometric modelling.
Important information may be lost in this process. But the trade-off between
information loss and the model’s realism is inevitable that analysts have to face.
The choice on this front largely determines the quality of the final estimates of
insurance output.

Some issues in insurance input measurement

Traditionally employed measures of inputs such as labour and capital are not
sufficient for accounting for the entire input vector used by insurance firms.
Insurance firms transform risks of various kinds into insurable “commodities”
and thus rely on various resources that might not be fully represented by the
variables commonly used for insurance inputs. In EMEs such as India, Brazil
and others, the net worth of an insurance firm plays a significant role as an
input in the production process. The behaviour of the net worth of a firm
represents the nature of Asset-Liability Management (ALM). As suggested
216 Expert opinions and contributions
by the Statistical Cost Accounting (SCA) model, the quality of ALM and
a firm’s financial performance are strongly positively correlated.45 Another
helpful variable in the context of current discussion is the total reserves of an
insurance firm,46 which can be and has been used as a proxy for debt capital
by various analysts. Capital as input is embodied differently within the insur-
ance production function, depending on whether it is debt or equity capital.
Insurance firms in emerging economies rely, to a large extent, on the bank-
ing system to finance their activities. With relatively underdeveloped capital
markets, banks are the chief source of finance for these firms. In such a case,
the role of debt and equity capitals will differ. Both will provide funds but
under different cost-benefit conditions. Aggregating both the measures into
a single measure of “capital” may not be a plausible idea in such cases.47 Thus,
accounting for them separately can help in estimating the insurance output
more robustly.

Some issues in insurance output measurement

As stated earlier in the discussion, differentiating the intermediate stage and
final stage outputs in an insurance firm’s production process is critical to
achieving a realistic empirical representation of the actual reality. A two-stage
model is generally the better empirical strategy in the case of insurance output
measurement. One has to be clear about which stage of insurance production
function they are working upon. Given that insurance firms transform risks
into insurable commodities, this transformation does not occur as a one-stage
event. For the studies focused on the first stage, the net premium collected by
the firm is the most plausible measure. The risk pooling function of an insur-
ance firm is undertaken, among others, using the funds collected from the
policyholders. Traditional inputs of labour and capital are used to “produce”
the premium income. In the second stage, the same premium income, along
with other sources of funds such as debt and equity capital, and labour, are
utilized to produce the final output, which is the risk coverage48 given to the
policyholder and possibly also the value generated for the shareholders,49 to
the extent of equity finance received by the firm. Insurance firms also perform
the risk management function in society. This is an integral part of its total
output. A helpful variable in this context is the incremental50 Assets under
management (AM) of the insurance firms. Many insurance firms in emerging
and advanced economies are actively engaged in managing the funds of poli-
cyholders both directly (through investment-based life insurance policies) and
indirectly (through the funds collected for general insurance policies). These
firms undertake risk transformation and management functions. They pool the
funds and diversify them, thereby reducing the overall variability in returns and
thus the risk associated with the investment. Incremental AM is a plausible can-
didate in this regard, not only because it is theoretically better, but also because
data are easily available in most major EMEs. AM is a critical policy lever for
insurance regulators to keep an eye on the risk-taking behaviour of insurance
 Expert perspective on output measurement in banking, insurance 217
firms. Thus data on this variable are mostly available in emerging and advanced
economies.

Price measurement in insurance

Prices are critical ingredients in constructing reliable measures of the volume
and value of the inputs and outputs used in the insurance industry. Deflation
procedures are susceptible to the measure of price employed by the analyst. Esti-
mates of constant-price and volume-based insurance output cannot be derived
unless industry-specific standards of input and output prices exist. Across the
board in EMEs, this matter is a significant constraint. Data on insurance-specific
prices are generally not directly available. Analysts undertake imputations to fill
this gap, but these imputed measures are largely imperfect, though they serve to
some extent. For example, consider the measurement of total factor productiv-
ity in insurance. Such an exercise will require deflated measures of inputs and
outputs, which in turn shall require data on input and output prices. National
income data and industry-level deflators can be used for those working on sec-
toral data within a macroeconomic accounting framework. However, the GDP
deflator is not a measure of actual observed prices but a crude aggregate proxy
for a sectoral price index. Microeconomic studies require disaggregated data
on prices across space and time. Such data are largely absent in the context of
EMEs. This is an area that needs the urgent attention of the policymakers.51
218 Expert opinions and contributions
Zhu Haiju
Brief Profile: Professor Zhu Haiju is currently working at the Zhejiang Gong-
shang University, Hangzhou, China. He specializes in monetary economics
and has been working on various issues in China’s financial services industry.
He is also an alumnus of the Ludwig von Mises Institute, Alabama, US.

Measurement of output and its rationale

Given the initial endowments and the economic problem of scarcity, the
efficiency with which a country utilizes its available resources determines its
development’s temporal and spatial paths. Divergence or convergence of devel-
opment paths among the nations is deeply affected by the differences in the
economic efficiency52 achieved by the countries. Assessment of economic effi-
ciency and performance is critical information for stakeholders involved in
the development process. The quantities of economic outputs and the inputs
are required to derive various financial performance measures. This is true in
firms, industries and the aggregate economy. At the firm level, performance
analysis is needed to channel the scarce resources efficiently due to competitive
pressures. This is required to enable economic coordination and possibly opti-
mal industrial policies at the industry level. At the aggregate level, performance
analysis is needed to achieve the traditional macroeconomic goals of growth,
low inflation and overall progress as per Keynesian policy frameworks adopted
by many emerging market economies (EMEs).
Given these necessities, firms’ performance analysis, industrial groupings, and
aggregate economy are important considerations for any emerging economy.
With the large-scale potential to increase the efficiency in resource utilization,
EMEs are characterized by sub-optimal use of resources mainly due to distortions
in the market process53 created by government interventions and various internal
and external shocks.54 The effects of such interventions and shocks may have a
long-lasting impact on the markets’ economic coordination. These effects depend
on the changes in the economic efficiency of such shocks. Studying the nature
of economic efficiency is thus an inevitable necessity for any economy, particu-
larly those on the path of economic progress but are yet to reach their potential
outputs. The measurement of the output variables determines efficiency analysis
in turn. At the same time, inputs measurement is also a matter of debate, the
theoretical consensus in quite clear on this account. In the mainstream neoclassi-
cal models, labour and capital are considered the most fundamental inputs in any
production process.55 However, the debates on the efficiency analysis of banking
and insurance industries are more tilted towards measuring outputs.
This is particularly true because the nature of the production function
underlying the observed output is sensitive to the institutional arrangements
under which a firm or industry functions. Wide differences on this account
characterize emerging economies. While interventions in the financial services
industry are shared among these economies, the degree, type and quality of
policy interventions vary sizeably. The institutional environment, in terms of
 Expert perspective on output measurement in banking, insurance 219
the participants involved, the economic freedom they enjoy in pricing their
services, the degree of competition and the intensity of interventions, varies by
a large margin across the EMEs. Such differences inevitably affect the quantity
and composition of the output produced by the financial services industry and
thus results in differences in the estimates of efficiency across these economies.
Not only inter-country but the differences in the output generate even intra-
industry and intra-firm differences in efficiency estimates.

Nature of output in the services industry

A common practice among mainstream economists is to treat the output of
different industries in the service sector as conceptually different. This induces
the need to search for the appropriate aggregation functions when analysis has
to be undertaken at the industry level. While such an approach may be justified
as an outcome of practical observations, the truth seems that all the outputs
within the financial services industry are meant to fulfil consumer demands.
Given their consumption orientation, the most practical way of aggregating
them is through the weighting factors derived from the consumption side, i.e.
the demand side rather than the supply side. This might better reflect the rela-
tive importance of each firm’s output in the aggregate industry-level produc-
tion function. However, this is easier said than done.
A reliable amount of data on various inputs and outputs is available regard-
ing the banking and insurance industries in the significant EMEs. In China,
for example, the banking and insurance industries are occupied by state-owned
companies whose daily business is consolidated by the official agency – China
Banking and Insurance Regulatory Commission (CBIRC). The official agency
generally collects data on the assets, deposits, loans, regional distribution of the
same, interest rates and other variables. Such is the case with most major EMEs
such as Brazil, India and others. The issue that permeates most of these econo-
mies is that the data on possible weighting factors for aggregation are missing.
On the demand side, the output price should best reflect each output’s relative
importance in the industry-level aggregation function. However, reliable price
indices are generally missing, and the available data requires a lot of imputations
to become empirically meaningful. Such concerns result in dependence on
proxy measures for aggregation factors that seem to diverge from the demand-
side realities rather than converging with them.

Some empirical aspects of output measurement in the EMEs

The banking industry is highly regulated. And so is the insurance industry. In
China, the banking firms are essentially state-owned, and the role of the gov-
ernment is extremely large in channelizing the loanable funds within the econ-
omy. This is done mainly by relying on broader economic planning rather than
purely through economic coordination via the market prices.56 Price discovery
is considerably distorted by state planning, though the estimates of prices for
financial services can be arrived at using various proxy-pricing methods.57
220 Expert opinions and contributions
Similarly, the insurance industry is almost monopolized by several large state-
owned insurance companies, and it is difficult for private individuals to obtain
licenses. This has resulted in a disproportionately higher share of public sector
output in China’s total financial services output. Daily banking activities such
as withdrawals from private accounts are subject to strict constraints, such as
limits on the amount of cash to be withdrawn per day. Such restrictions are not
specific to China but are generally observed in other EMEs as well. However,
the intensity of regulation differs among these economies, and this creates the
need for adjustment of observed output for the impact of inflation, subsidies
and other interventions. Moreover, government regulations, especially inflation
and policy-induced cyclical fluctuations, make it difficult to measure the true
output quantum. Such adjustments are a matter of further research, particularly
for deriving the estimates of public sector output in financial services. This is
necessary because, ideally, the economic theory requires the purest form of eco-
nomic output, undisturbed by the non-market forces. In reality, however, the
observed financial statements of banks include the impact of such interventions.
Thus it might not be theoretically sound to simply use the official data on assets,
deposits, premiums collected, claims paid and others without proper treatment
for non-market distortions. A question immediately arises: How can this be
undertaken when such impacts are embodied in the observed data themselves,
and the relevant adjustment factors are unavailable? This matter requires a more
exhaustive analysis, and it seems that the consensus in the mainstream is still
weak on this account. One point to note here is that the structure of the finan-
cial services industry matters when judging the market-representativeness of
observed financial statements. Suppose the financial system is dominated by pri-
vate players with relatively minimal policy restrictions. In that case, the financial
accounts of the banks and insurance firms should provide a largely reliable idea
about the theoretically true level of output. The observed data, in this case, shall
be least distorted by policy interventions and hence should provide a fair idea
of the market forces at play. Such estimates of output should also be consistent
with the neoclassical production theory. India and Brazil seem closer to this styl-
ized approximation, while China is the diametrical opposite of these economies.
Estimates of financial services output by official agencies in most economies are
prepared using the firm-level financial accounts. These aspects should be borne
in mind while using official industry-level data.
At this juncture, one must, however, bear in cognizance the fact that even
though the official data may not be completely reliable, as incentives to ‘adjust’
the data by government officials or executives of state-owned enterprises will
always exist, on the whole they do reflect the underlying economic perfor-
mance of a region or industry. The data and experience are generally consistent
on this account – for example, the economy of Northeast China is relatively
backward, and this is clearly reflected in the available official economic data.
However, this representativeness becomes weaker as we move towards higher
levels of disaggregation, particularly in the case of output measurement of
financial services. This is contributed, in part, by the sheer diversity of activities
 Expert perspective on output measurement in banking, insurance 221
that are to be accounted for when firm-level behaviour is analysed and due to
a large shadow sector that is not directly captured in the financial statements
reported by the banks and insurers.

Notes
1 While theoretically, financial intermediation should work exactly the same across every
economy, domestic economic conditions, the degree of financial depth, the extent of
financial development, and a host of related factors cause differences in how effectively
and efficiently financial intermediaries function. Such differences can be wide if one com-
pares economies with different political regimes. India, for example, has a relatively more
prominent presence of private financial intermediation firms than China. While China
also follows the SNA just as India does, the dominance of public intermediaries lends a
very different character to the financial services industry of China. The data on the finan-
cial services industry from both these countries thus need to be further adjusted to make
them comparable in an international context. Such differences make it difficult to rely on
SNA procedures and principles solely. More statistical and econometric adjustments are
required to make data comparable for international macroeconomic coordination.
2 In the present discussion, the term ‘performance’ includes a class of measures such as
technical efficiency, allocative efficiency, total factor productivity and profitability.
3 Chambers and Färe (1993) state that “the existence of an aggregate output for a multi-
output technology is justified by the presumption that the underlying production tech-
nology is separable in outputs”. This presumption requires rigorous justification before
individual outputs can be aggregated to undertake industry-level performance analysis.
An alternative in this regard is non-parametric methods such as Data Envelopment
Analysis, though there are problems with it too.
4 Green (2016) provides a detailed discussion on the procedures required for consistent
aggregation across different outputs and inputs to derive aggregate input and output
measures for production analysis.
5 Linking the borrowers and the savers under economically efficient terms of exchange is
possible due to the existence of banks, particularly in EMEs whose financial system is
essentially bank-dependent rather than being more tilted towards the capital markets as
is the case with the advanced economies.
6 The banking industry globally seems to be shifting towards fee-based services in modern
times. It is too early to suggest that non-intermediary services have overtaken the inter-
mediation services in the business portfolio of the banks. Still, there is a trend towards
‘disintermediation’ in the banking industry – i.e., the increasing share of fee-based non-
intermediary services in the overall business portfolio of banks.
7 Given the focus on the banking industry, the non-banking financial intermediaries are
kept outside the scope of the current narration.
8 See Supaarmorakul (2008) and Schreyer and Stauffer (2002) for more details on refer-
ence rate and user cost approaches to FISIM estimation.
9 The primary channel through which Big Data is being employed in the banking industry is
via the adoption of blockchain technology in providing banking services. However, there
is a serious lack of discussion on the challenges and ways of overcoming them, particularly
in emerging economies (Hassani et al., 2018). Also, see Srivastava and Gopalkrishnan
(2018) for significant areas where banks are harnessing the power of Big Data.
10 Technology here implies either the software or the hardware side innovations. Big and
thick data may roughly be treated as software-side innovations.
11 Traditionally, satellite accounts have been used for covering non-traded inputs and out-
puts. However, their use in emerging economies has expanded considerably due to
the limitations inherent in the developing data collection systems in such economies.
Essentially, satellite accounts “provide a framework for examining difficult-to-measure
222 Expert opinions and contributions
activities excluded, or inadequately treated, in the NIPA” (National Research Council,
2005). Output measurement of financial services, mainly the banking industry, is one
such area.
12 Output measurement can be undertaken by the income, value-added or expenditure
methods. Theoretically, all these three approaches measure the same economic phe-
nomenon – i.e. what is produced. The value-added approach is adopted for this discus-
sion. Adopting other approaches would not change the essence of the narration.
13 Available input-output tables for major emerging economies such as India, Brazil and
Russia do not provide estimates of input-output connexions at the banking industry
level but offer the same for financial services as defined in the ISIC Revision 3 or 4.
14 ‘Value’ in this context is the use-values of commodities produced by the market econ-
omy. While the measurement of value is subjective, the amount of value produced in
an economy can be approximated from the perspective of the producer. The amount of
output produced by an individual firm is the practical measure of its value in the present
context.
15 Banks have both direct and indirect roles in financial accumulation in an economy. They
provide avenues for parking the savings in interest-bearing assets such as fixed deposits
and even access to financial markets through investment account services. Banks are
also intermediate as crucial players in both the money and capital markets; in countries
such as India and China, they largely drive the movements in these markets. This allows
financial accumulation to occur through investments in financial markets, which in
themselves are enabled by the banking industry to a large extent.
16 The degree of randomization achieved in practice is a matter of debate. Even when the
best efforts are invested in selecting a random sample, “randomization biases” (Banerjee
et al., 2017) tend to distort the population-representativeness of the sample. As discussed
later, new forms of data and data collection systems might very well be able to overcome
some of these limitations.
17 One must be warned that this belief is valid only for the organized financial services
sector. In emerging economies such as Brazil, China, India and others, a large portion of
their financial services activities emerge in the unorganized sector. If the data from the
organized sector is the only available source, then such data do not reflect the underlying
population, which consists of both the organized and unorganized sectors.
18 This is not a generalized statement, and even the so-called “Big Data” sets might be gen-
erated under well-controlled experiments. However, with reference to financial services,
observational data are the primary source as there is limited control that monetary author-
ity can wield over the financial services industries. There are many aspects of the behav-
iour of banking and insurance firms that are not directly controllable by the monetary
control, even in the most advanced economies with a high degree of financial depth.
19 This does not suggest that all the Emerging Economies are subject to the same con-
straints in their data collection systems. EMEs are diverse, and there are a lot of intricate
differences in the problems faced by official agencies in these countries. For example,
while Brazil and India have relatively transparent data assimilation and reporting systems,
China, on the other hand, has a high degree of government regulation on the data that
is made accessible to the public. Even though such differences exist, the overall state of
development of the official data assimilation systems shows a broad macro similarity in
terms of their current state of evolution across the EMEs.
20 Traditional standardization procedures involve a set of statistical treatments such as inter-
polation, extrapolation, imputation using proxy values, scaling of data in similar units,
using ratios instead of level-form data, using first differences and growth rates instead
of level-form data and others. Such procedures shall be required even when alternative
data sources are employed to analyse the same variable of interest. However, when data
sources are highly dissimilar, such procedures are not sufficient, though they may be
necessary. For example, satellite imagery data, as a source of economic production in
regions of a country, cannot be merged with the existing national accounts data simply
by following the procedures mentioned earlier. One has to harmonize the definitions,
 Expert perspective on output measurement in banking, insurance 223
measurement units, and aggregation weights and translate the novel sources of informa-
tion into a single variable which can be used as part of a typical large-scale macroeco-
nomic model.
21 Economic advantage is on account of more disaggregated data availability with an enor-
mous scope than what the typical consumer survey data would provide. The statistical
advantage is on account of the more significant degrees of freedom and the ability to
utilize the asymptomatic properties of the estimators.
22 The hardware and software components of the data collection agencies in emerging
countries need to be improved for handling multiple data sources. Big Data is character-
ized not only by its large volume but also by its rapid frequency and flow. Data storage
capacity, security, recording, analysis, and other aspects require an overhaul, which has
sizeable financial implications. Emerging economies need to optimize their budgetary
constraints to handle these outlays and treat them as investments rather than expenses.
Such efforts will not only enhance the decision-making capabilities of policymakers at
all levels of the hierarchy, but it will also create human capital in the process, as many
people will get trained in using such data.
23 Given my experience as a professional in the marketing functions of several corporate
banks, I must invoke the marketing perspective and try to blend it with the economic view-
point. It is hoped that such an approach is looked at from a multidisciplinary perspective.
24 Bias in this context is the inability of the observed performance as measured by single
output, to represent the true performance, which should be measured in a multi-output
framework.
25 This is one of the reasons why many studies use non-parametric techniques such as Data
Envelopment Analysis in the performance analysis of banking firms.
26 With the large presence of unorganized and shadow banking segments, this point implies
that if regulatory actions produce desirable changes in the non-organized components,
too, then such regulations may be hypothesized as deep in terms of their impact on the
overall banking industry.
27 One must bear in mind that the fee-based services, although explicitly charged and
taxed by authorities as products of banks, are provided as simply complementary goods
and are thus not suitable for treatment as final outputs in themselves.
28 The financial services industry has a far broader scope than the banking industry. It
includes insurance services, non-banking financial services, and other financial services
other than traditional commercial banking.
29 ILO (2018) states that more than 60% of the employed workforce in the EMEs are engaged
in the informal economy. With such a significant dependency on the informal sector, the
scope for imputations and other adjustment techniques is prominent in these nations.
30 Industry-level estimates for financial services output can be interpreted more robustly
than the individual firm-level data because any given financial services industry is simul-
taneously a supplier of inputs to other financial services and a producer of output.
Whether the services rendered by a single financial service firm are input or output
is thus debatable. However, it is theoretically much more convenient to consider the
financial services industry as a producer of output as a whole and possibly avoid the
complex interdependencies of these services at the level of firms.
31 The Global Financial Development Database of the World Bank provides information on
three variables of interest in this context. First, the financial system deposits as a proportion
of the GDP; second, the proportion of total firms using banks to finance their long-term
investments; and third, the proportion of total firms using banks to finance their working
capital requirements. On all these three accounts, major EMEs such as China, Russia,
Brazil, India and Indonesia have shown a gradual and consistent increase in the last decade.
This may be considered a sign of the increasing space that the organized financial sector
occupies in these economies as against the so-called shadow financial sector.
32 Unit-level data are the data that correspond to individual, institutional units as defined
in the SNA, 2008 manual. The quality of these data is the most crucial determinant of
the ability of a nation to construct high-quality national income accounts and, through
224 Expert opinions and contributions
them, derive high-quality estimates of the key macroeconomic variable. The inequal-
ity among emerging and developing countries on this account is significant. This has
resulted in the call for more statistical adjustments at the end of the data compilers to
make them internationally comparable.
33 With the share of non-agricultural output secularly rising against the share of agriculture
in aggregate GDP in most top emerging economies, this problem becomes all the more
pronounced and worrisome.
34 An economically optimal price index must fulfil critical properties which were very
finely laid down by Fisher (1921) and Fisher (1922).
35 This is also one of the arguments favouring more robust financial sector regulation,
which is generally disliked by the pro-market economists who would like to push
forward financial deregulation instead. The financial market structure in EMEs is not
strong enough in the current period to allow for further deregulation. Look at the issue
of non-performing assets (NPA) in Russia, India and even Brazil. If unchecked, persis-
tence in NPAs can very well crumble the existence of large-scale public sector banks
that are generally more prone to this financial challenge.
36 These observations imply that the econometric evidence on bi-causality between
finance and growth and the causal role of finance on growth should be interpreted
about whether these analysts adjusted their estimated coefficients for the availability
and quality of human and physical capital. Perhaps removing the intermediating role of
human and physical capital in the finance-growth nexus might very well result in no
causal relationship between the both whatsoever.
37 The aftermath of the war was not the only factor motivating the standardization of
national income accounts through the SNA. Richard Stone put it succinctly when he
noted that “The problems of war finance, full employment and the marketing of the
produce of all forms of economic activity . . . have come to be discussed in terms of the
concepts of national income studies” (United Nations, 1947).
38 The 1947 report stressed collecting data from the so-called elementary transactions that
occur between individual accounting entities and building up the national aggregate
from such building blocks. The report reads, “Experience shows that these ideas can be
expounded and presented more lucidly if the elementary transactions of an economic
system rather than the final aggregates of transactions, such as the national income,
are made the starting-point of the enquiry”. Richard Stone and his team particularly
stressed this approach in the memorandum submitted as a part of this report, which is
still a seminal work in national income accounting practice.
39 Economic distribution here refers to economic inequality, as Sen (1973) suggested, who
defines economic inequality as a set of income, expenditure, and wealth inequalities.
Hence, economic distribution is much larger than that of the popularly used notion of
income inequality.
40 Insurance performance consists of efficiency, profitability, productivity, economies of
scale and others.
41 Risk allocation needs to be Pareto-optimal for insurance to have a rational economic
justification. There can be situations wherein a non-competitive market structure might
prevent the insurance firms from allocating risks efficiently. Emerging economies can
be prone to this issue, mainly where government-induced monopolies exist. Many
EMEs have been opening their doors to domestic and foreign private capital. However,
changes in market structure take time to occur and possibly in the coming decade, the
economic efficiency of the insurance industry in EMEs will improve further.
42 Capital is used in the insurance industry in terms of physical capital and as human and
financial capital. Accounting for these two kinds of capital in the insurance production
function is quite challenging due to stringent data requirements that the EMEs may not
be able to fulfil. For illustration, thoroughly accounting for human capital in the insur-
ance industry requires a very well-structured, organized labour market. EMEs such as
 Expert perspective on output measurement in banking, insurance 225
China, where most of the labour market activities are included within the domain of the
official statistical system, can probably account for this variable. However, for economies
such as Brazil, Russia and India, the problem of unorganized labour markets can con-
strain the success of this parameter. Unorganized labour market activities are not easily
accountable in terms of the official data collection, and thus the agencies need to rely
on imputations such as shadow pricing, which inevitably are prone to errors.
43 Output estimation becomes more challenging when looking at investment-oriented life
insurance products that produce risk coverage for their policyholders and investment
returns. Here, investment risks are borne by the policyholder, but the insurance firm
provides a pooling of fund services that help reduce the investor’s risk exposure.
44 In the present context, ‘performance’ includes measuring efficiency, total and partial
factor productivities, and economies of scale and scope in the current context.
45 In this context, Owusu and Alhassan (2021) is an interesting application of this model
in the context of the financial services industry. Even though their paper focuses on the
banking sector, their broad findings and methodological approach can also be extrapo-
lated to the insurance industry.
46 Total reserves of an insurance firm (whether life or nonlife) is generally defined as the
sum of unearned premiums and outstanding claims as underwritten by the company).
47 Such a practice can cause serious specification errors and distort the empirical founda-
tions of the work.
48 Risk coverage provided to policyholders can be measured using the company’s amount
of claims actually paid. Such data are generally readily available in the major emerging
economies through their official agencies or the firms’ financial statements.
49 One may consider using the changes in the firm’s market capitalisation in the stock
exchange as a possible measure of value generated for shareholders. Alternatively, divi-
dends and capital gains can be used during a specified period (usually a year).
50 Incremental because the Asset under Management of the insurance firm is a stock vari-
able, while its incremental value is a flow variable. Theoretically, the output is strictly a
flow concept.
51 One must bear in mind that deflation is not always required. In the case of purely cross-
sectional studies, one may proceed with current price data. However, for inter-temporal
models, deflation is a necessity. Inter-temporal studies occupy a large part of the extant
literature on insurance performance.
52 Economic efficiency refers to the sum total of technical and allocative efficiencies in the
present context. See Kalirajan and Shand (1999) for a detailed discussion on the issues
in efficiency measurement.
53 Policy interventions cause distortions in the form of constraining the speed with which
the markets can match the buyer’s and seller’s decisions, thus hampering the process of
“entrepreneurial discovery” (Kirzner, 1997).
54 Shocks in this context imply exogenous shocks. It is necessary to locate the exogenous
shocks as “only responses to an exogenous variable can measure the effects of policy-
induced changes in that variable” (Cochrane, 1994). Exogenous shocks in the present
context might include unexpected technological innovations and unforeseen changes
in international trade. However, as stressed in Cochrane (1994), the specification of an
unexpected event as an exogenous shock depends on how much the concerned eco-
nomic agent was aware of the same and had accounted for the same in their decision-
making. This is challenging for any researcher, and the slightest error in specifying the
exogenous shocks will dramatically change the final analysis.
55 This is not to suggest that input measurement is a trivial task. The choice of the vari-
ables to measure the inputs, their data availability and the quality of available data pose
a tough challenge for analysts in obtaining theoretically consistent measures of inputs.
Mainstream neoclassical theory, for example, expects volume-based metrics of labour
and capital. In reality, however, the constant-price measures are employed and the very
226 Expert opinions and contributions
derivation of constant-price variables is a complicated task. However, the definition of
inputs and their measurement also have consensus in the empirical literature.
56 Market prices are increasingly becoming essential components in the economic plan-
ning process in China. However, there are substantial lags in obtaining the data on
prices, which creates inefficiencies in economic coordination. In the financial services
industry of China, the large extent of the NPA problem and the public sector bias in the
allocation of loanable funds is one example among many.
57 Such methods are used at the firm level in banking services also, and their use is not
limited to industry-wide analysis only. An interesting application of the same can be
found in Fragnière et al. (2011).

References
Acharya, D., & Kamaiah, B. (2001). Simple sum versus Divisia monetary aggregates: An
empirical evaluation. Economic and Political Weekly, 36(4), 317–326.
Banerjee, A., Banerji, R., Berry, J., Duflo, E., Kannan, H., Mukerji, S., Shotland, M., &
Walton, M. (2017). From proof of concept to scalable policies: Challenges and solutions,
with an application. Journal of Economic Perspectives, 31(4), 73–102.
Barman, R. B., & Samanta, G. P. (2004). Banking services price index: An exploratory
analysis for India. IFC Bulletin (4), 110–124.
Chambers, R., & Färe, R. (1993). Input-output separability in production models and its
structural consequences. Journal of Economics, 57(2), 197–202.
Cochrane, J. H. (1994). Shocks. Carnegie-Rochester Conference Series on Public Policy, 41, 295–364.
Fisher, I. (1921). The best form of index number. Quarterly Publication of the American Statisti-
cal Association, 17, 533–537.
Fisher, I. (1922). The making of index numbers. Houghton-Mifflin.
Fragnière, E., Moresino, F., & Zheng, N. (2011). Development of pricing models for intan-
gible services based on conjoint analysis and mathematical programming: A case study
about the design of credit card services in China. In 2011 annual SRII global conference.
Retrieved January 6, 2022, from https://doi.org/10.1109/srii.2011.71
Green, H. A. J. (2016). Aggregation in economic analysis. Princeton Legacy Library.
Groves, R. M., & Harris-Kojetin, B. A. (2017a). Federal statistics, multiple data sources, and
privacy protection: Next steps, a consensus study report of the panel on improving federal statistics for
policy and social science research using multiple data sources and state-of-the-art estimation methods.
US National Academy of Sciences, National Academies Press.
Groves, R. M., & Harris-Kojetin, B. A. (2017b). Innovations in federal statistics combining data
sources while protecting privacy, report of the panel on improving federal statistics for policy and social
science research using multiple data sources and state-of-the-art estimation methods. US National
Academy of Sciences, National Academies Press.
Hassani, H., Huang, X., & Silva, E. (2018). Banking with Blockchain-ed big data. Journal of
Management Analytics, 5(4), 256–275. https://doi.org/10.1080/23270012.2018.1528900
ILO. (2018). World employment social outlook: Trends 2018. Report of the International
Labour Organization.
Kalirajan, K. P., & Shand, R. T. (1999). Frontier production functions and technical effi-
ciency measures. Journal of Economic Surveys, 13, 149–172. https://doi.org/10.1111/
1467-6419.00080
Kirzner, I. M. (1997). Entrepreneurial discovery and the competitive market process: An
Austrian approach. Journal of Economic Literature, 35(1), 60–85.
 Expert perspective on output measurement in banking, insurance 227
Lucas, R. E. (1976). Econometric policy evaluation: A critique, carnegie-rochester conference series on
public policy (Vol. 1, pp. 19–46). Elsevier.
Menger, C. (1985). Investigations into the method of the social sciences with special reference to eco-
nomics. New York University Press.
National Research Council. (2005). Beyond the market: Designing nonmarket accounts for the
United States. The National Academies Press. https://doi.org/10.17226/11181.
Owusu, B. F., &Alhassan, A. L. (2021). Asset-liability management and bank profitability:
Statistical cost accounting analysis from an emerging market. International Journal of Finance
and Economics, 26(1), 1488–1502.
Schlicht, E. (1985). Isolation and aggregation in economics. Springer-Verlag.
Schreyer, P., & Stauffer, P. (2002). Measuring the production of financial corporations. Progress
Report by the OECD Task Force on Financial Services (Banking Services) in National
Accounts.
Sen, A. (1973). On economic inequality. Oxford University Press.
Shriffin, R. M. (2016). Drawing causal inference from big data. Proceedings of the National
Academy of Sciences of the United States of America, 113(27), 7308–7309. www.pnas.org/cgi/
doi/10.1073/pnas.1608845113
SNA. (2008). System of national accounts. United Nations Statistics Division.
Srivastava, U., & Gopalkrishnan, S. (2018). Impact of big data analytics on banking sector:
Learning for Indian banks. Procedia Computer Science, 50(2015), 643–652.
Strömmer, K. (2019). Big or thick data? Collecting data in meaningful ways. https://wpmu.mah.se/
nmict192group6/2019/10/11/big-or-thick-data-collecting-data-in-meaningful-ways/
Supaarmorakul, P. (2008). Estimation of financial intermediation services indirectly measured
(FISIM): Thailand’s case. IFC Bulletin No. 28, 83–100.
United Nations. (1947). Measurement of national income and the construction of social accounts.
Report of the Sub-Committee on National Income Statistics of the League of Nations
Committee of Statistical Experts, United Nations, Geneva.
Part III

Empirical case studies

A BRIICS perspective
8 Empirical case studies
for the banking industry
on implications of using
alternative output definitions

8.1. Rationale, methodologies and data

Parts I and II have provided perspectives on banking and insurance output
measurement from the literature survey and insights from practitioners, respec-
tively. Part III now provides empirical exercises on the implications of alterna-
tive output measures on various performance parameters. In the present study,
performance is conceptualized in terms of efficiencies, productivities and
related issues whose estimates can throw some light on how well the firms or
the industry utilises their resources and progresses economically. This compre-
hensive conceptualization of the term ‘performance’ suits the present context
well. Literature on banking and insurance, as summarized in Tables 2.1 and
2.2, shows the wide variety of topics that have been investigated using different
output measures. Covering all of them is not feasible due to the limited space
and scope of this book. Thus, three matters are looked into in this part. First is
the implications of alternative output measures for separately aggregate partial
factor productivities of banking and insurance industries.
Furthermore, life and nonlife insurance industries are analysed separately
in the next chapter. The second is a similar exercise concerning the index-
based measure of aggregate total factor productivity and parametric and non-
parametric measures of aggregate technical efficiency.1 The third is an analysis
of the variations, if any, that are brought about by alternative output measures
in aggregate industry-level estimates of returns to scale measured via parametric
production functions. These three matters are investigated separately for the
aggregate banking and insurance industries. Data used are sectoral, i.e. aggre-
gated over the entire industry rather than firm-level.2
The choice of the three issues is primarily motivated by their popularity
in the literature. This part of the book is presented as a case study on the
empirical implications of performance estimates to alternate output measures
for the banking and insurance industries. As noted by Boda and Piklova (2021,
p. 1554)3: “the effect of adopting an approach or choosing an input-output set
within an approach upon efficiency scores and results has not been consistently
explored”. Hence, choosing the issues that have garnered sufficient research
attention seemed to be a plausible approach in this regard. The other motivat-
ing factor is the availability of data.

DOI: 10.4324/9781003149828-11
232 Empirical case studies
More elaboration is required on the choice of focusing upon aggregate
industry-level data rather than firm-level data, which has traditionally been
done in the literature. Performance estimation in banking and insurance has
largely been undertaken on firm-level data. This is quite understandable
because assessing technical efficiency, productivity and returns to scale requires
analysing how decision-making units (DMUs) act within a production func-
tion framework. Not only theory but also the empirical methods such as DEA
require data which represent individual DMUs. Firms tend to be the best rep-
resentatives of the DMUs. This is the most preferred approach in the literature.
However, the aggregated effects of the diverse decisions by individual firms are
also matters that deserve attention. Economic theory saw a lot of analytical pro-
gress when the subject incorporated pure macroeconomic theory and released
itself from its dependency on pure microeconomic logic. Such a fundamental
change in perspective is necessary for the evolutionary success of a discipline.
As Colander (1993, p. 447) notes: “the same reality can look fundamentally dif-
ferent depending on one’s perspective and that revolutions in a discipline occur
through these changes in perspectives”. The Keynesian and the New Classical
revolutions are possibly the best illustrations of such a fundamental change in
economics (Colander, 1993). The current consensus is that micro-foundations
are a necessary condition to contextualize macroeconomic aggregates, and in
the case of performance analysis of the banking and insurance industries, one
can approach industry-level aggregate as emanating from individual optimiza-
tion decisions. Such aggregation is also necessary when long-term economic
issues need to be assessed from a policy angle.4 In such scenarios, policymak-
ers would be interested in firm-level heterogeneities in the performance of
banks and insurers and their sectoral performance. Production behaviour of
individual firms is the level at which conscious decision-making is undertaken.
However, the economic environment faced by the individual banking and
insurance firms is a result of their accumulated, aggregated and interdepend-
ent actions. One way to approach these actions is to utilize a general equilib-
rium approach using firm-level data. Such an approach requires data which
are very difficult to locate in the case of the economies under consideration
here (the BRIICS countries). Another approach is to analyse how the industry
as a whole is behaving instead of focusing on the individual firms. The issue
under consideration here is the empirical sensitivity of performance estimates
to alternate output measures for the banking and insurance industries. Sup-
pose the industry-level estimates show considerable empirical sensitivity on this
account. In that case, it may serve as a plausible justification for the existence
of a similar pattern at more disaggregated levels. Hence, aggregate banking
and insurance industries are analysed separately in this part of the book. The
industry as a whole is treated as an aggregate DMU whose behaviour is a result
of the decisions taken by individual firms composing it. The aggregate industry
across each year of the time series data for the period 2000 to 2021 is treated
as separate DMUs. In other words, in this study, the aggregate industry-level
production function is estimated rather than the industrial production function
 Banking: implications of using alternative output definitions 233
using individual firm-level data on inputs and outputs. Aggregation is under-
taken over the chosen inputs and outputs variables.
The use of the aggregate production function gives rise to the so-called
“aggregation problem” (Felipe & McCombie, 2005, p. 467). It refers to the
problems in building an aggregate production function from individual micro-
level production functions. Traditionally, this problem has been referred to in
the context of economy-wide production functions that combine inputs and
outputs from heterogeneous industries. However, in the present case, one may
partially dodge this problem by arguing that inputs and outputs from the same
industry have been aggregated. Moreover, all the data employed in Chapters 8
and 9 have been obtained from international and national official agencies and
thus represent the best possible information available.5 Furthermore, firm-
level data were not available for all the economies under consideration. Finally,
efficiency and productivity analysis at the industrial level is largely missing in
the literature. All these factors necessitated and motivated the adoption of an
aggregative approach to performance analysis using an industry-level aggregate
production function.
The present and the next chapter focus on six economies: Brazil, Russia,
India, Indonesia, China and Russia (BRIICS). As shown in Tables 2.1 and 2.2,
the amount of evidence on emerging economies is considerably lesser than
those on advanced economies. Data constraints are the fundamental reasons
for this. However, data limitations should not prevent the efforts to analyse
emerging economies as the selected economies occupy, among all the major
EMEs, a large majority of the share in the World real GDP.6 Furthermore, they
represent a vast untapped potential market, as discussed in Chapter 1. Lastly, the
evidence on performance analysis for most BRIICS countries is quite limited.
Thus, there is a considerable gap in the literature that can be fulfilled by the
attempt made here in Part III.
Given the primary aim of this part is to provide a set of case studies on the
empirical implications of alternating output measures in the aggregate pro-
duction for banking, life and nonlife insurance industries, isolating the pure
effects of variations in output definitions from other possible sources is neces-
sary. Performance estimates such as industrial technical efficiency or total factor
productivity can vary due to multiple sources, as explained in Figure 8.1.
Given these possible sources that could explain the variations in performance
estimates under different output specifications, controlling for as many non-
output sources as possible is necessary to isolate the pure empirical effects of
variations in output measures. Hence, the most basic and common input vector
of labour and physical capital is used as far as possible, particularly in analysing
technical efficiency and returns to scale. The theoretical approach of specify-
ing production function is allowed to vary to find the implications of different
output measures under alternative functional forms. Both the Cobb-Douglas7
and the Trans-logarithmic8 production functions are employed in sections 8.4
and 9.4, while only the Cobb-Douglas9 model is used for sections 8.3 and 9.3.10
In terms of empirical methods, a Solow-style residual total factor productivity
234 Empirical case studies
index11 is built in section 8.3 for the banking industry and similarly in 9.3 for life
and nonlife insurance industries. DEA,12 Ordinary Least Squares (OLS)13 and
the Stochastic Frontier Approach (SFA)14 methods are employed in sections 8.3
and 9.3, while OLS, Ridge Regression and SFA15 are used in sections 8.4 and
9.4 with both the Cobb-Douglas and trans-log production functions. Ridge
regression is employed to deal with the well-known multicollinearity problem
in estimating trans-log production functions.16 Finally, the data sources, vari-
ables, and definitions are presented in the appendix to this chapter.
However, all the data used in this and the next chapter have been collected
from various official banking and insurance regulatory bodies of the chosen
economies and diverse international sources such as the IMF, World Bank, Bank
for International Settlements (BIS) and the International Labour Organization.
All the inputs and outputs data used in this and the next chapter are specific
to the banking and insurance industries respective. They are in aggregate form
across all firms in the industry. All the nominal variables are deflated using the
country-specific Consumer Price Index (CPI) available from the IMF. All the
monetary variables were converted into the same measurement unit in order to
facilitate intra-economy comparisons of the estimates. It may also be noted that
all the subsequent empirical analyses have been undertaken in a multi-input
and single output framework. This has been done to assess how each of the
chosen output measures induces variations in the efficiency, productivity, and
scale estimates. This applies to both – Chapter 8 and Chapter 9. Given the lack
of space, only the most essential methodological details have been presented
here. The rest can be availed from the author upon request.17
Lastly, a brief note on the output variables chosen in this part of the book
is called for. The following output measures have been employed in Part III of
this work.18 The first measure is deposits and is proxied by the amount of total
outstanding deposits measured in 2010 prices-based local currency units. This is
a stock variable. It reflects the amount of services rendered by the banking sector
to its depositors. The second measure is operating income and is proxied by the
amount of net operating income after extraordinary items and taxes measured in
2010 prices-based local currency units. This is a flow variable. It represents both
the interest and non-interest incomes net of interest, non-interest and other
expenses, including taxes. It may also be construed as a rough measure of the
amount of profits earned by banks. This variable represents the flow of services
generated by banks during a given year. The third measure is the amount of
loans and is proxied by the amount of total outstanding loans outstanding meas-
ured in 2010 prices-based local currency units. This measure represents the
quantum of services banks provide in terms of their traditional intermediation
role in the economy. It is a stock variable. The fourth measure is the financial
assets of banks and is proxied by the amount of total financial assets of the bank-
ing industry measured in 2010 prices-based local currency unit. It is a stock
variable. It represents both the traditional and non-traditional intermediation
roles of banks in the economy. It consists of both the loans and investments.19
The fifth measure is a volume indicator of deposits and is proxied by the number
 Banking: implications of using alternative output definitions 235

Input-output specification Theoretical approaches

Sources of variations
in performance
estimates

Empirical approaches Other sources

Figure 8.1 Sources of variations in performance estimates

Note: The ‘other sources’ represent issues such as specification errors, human error in estimation, meas-
urement errors, etc.
Source: Author’s analysis of relevant literature.

of total depositors at the end of the given year in the aggregate banking industry.
This is a stock variable and is a volume measure. The sixth measure is a volume
indicator of loans and is proxied by the number of borrowers.
The seventh measure is another volume indicator of deposits and is captured
by the number of total deposit accounts. It is a stock variable. Finally, the last
measure is another volume indicator of loans and is represented by the number
of loan accounts. It is a stock variable. As one can observe, the stock variables
dominate in this study and reflect the literature trend, as shown in Table 2.1.
Both value and volume measures are utilized. This is an improvement over the
prevalent reliance on value measures. Data on volume measures at the aggregate
level are more readily available than at the firm level. These data were obtained
from the databases of the World Bank, International Monetary Fund (IMF),
Bank for International Settlements (BIS), and where required, from the national
banking regulators of each of the chosen countries. Appropriate adjustments
were made to interpolate some of the data that were found missing.20

8.2. Empirical sensitivity of selected partial factor

productivity estimates across the banking sector
of BRIICS
Partial Factor Productivities (PFPs) provide a comprehensive overview of the
productivity movements in an industry. Given that these measures ascribe entire
output to a single input, which violates the multi-input nature of the banking
industry, their use is limited to providing preliminary findings and descriptive
analysis. With the advent of multi-factor productivity measurement, the need
for PFPs has reduced drastically. However, for the aggregate banking indus-
try, labour and capital productivity can provide some evidence of the broad
236 Empirical case studies
productivity movements in the long run. These secular movements can provide
a valuable indication of the banking industry’s ability to contribute to aggre-
gate economic growth. This section thus comments on how the estimates of
labour productivity (LP) and capital productivity (CP) of the aggregate bank-
ing industry vary as alternative output measures are employed within the exact
definition of LP and CP. LP is defined as the ratio of the chosen output variable
for the given economy to the number of employees in the banking industry for
a given economy.

LPij = (Outputij/Labourij)(1)
CPij = (Outputij/Physical Capitalij)(2)

Where i = the given country and j = chosen output measure, Labour = number
of employees in the banking industry of the given economy, Capital = Num-
ber of non-branch retail agent outlets or the number of commercial bank
branches or the number of ATM machines.21
Tables 8.1 to 8.11 in Appendix 1 provide the estimates on variations in the
distribution22 of LP and CP induced by alternative output definitions for each
of the BRIICS countries. Analysis of each of the Tables 8.1 to 8.11 reveals
some basic patterns. First, there are substantial differences in the mean levels
of labour and capital productivities across all economies. Differences in output
measures can induce large differences in the level of labour and capital produc-
tivity. Such a gap in productivity levels can cause substantial differences in how
the progress of the banking industry is interpreted and the kind of policy inter-
ventions designed to improve the same. These differences can also cause wide
variations in how the industry’s market structure is interpreted. Low produc-
tivity levels may be interpreted as signs of a high degree of imperfections and
economic barriers that could be thwarting the full realization of the industry’s
productive capacity. With inconsistency in productivity estimates across output
measures, choosing a representative output or output vector is a critical deci-
sion in any performance analysis of the banking industry. It also does not seem
that convergence could occur in these measures if sufficiently long periods
were analysed. This is primarily because the trend itself is very different across
these estimated series. The second key finding is that variability, as measured
by the CV (coefficient of variation), in estimates of LP and CP is also disparate
across the output measures but, in general, is not as heterogeneous as in the case
of mean-level estimates. Third, the economic dynamics at play in productiv-
ity behaviour are different for value versus the volume output measures. Value
measures show larger variability than the volume measures despite both having
stock dimensions. This is particularly true when compared to their third and
fourth moments of the distribution.23
In summary, it can be broadly concluded that differences in output measures
can cause wide differences in the distribution of labour and capital produc-
tivities in the banking industry. This occurs predominantly at the first, third
and fourth moments of their distribution,24 while there is a good amount of
 Banking: implications of using alternative output definitions 237
consistency in terms of the variability measured by CV. These findings are
valid across the selected emerging economies. One of the implications of these
findings is that locating the critical determinants of banking productivity and
inducing improvements in the same would be very difficult unless the distribu-
tional trade-offs in choosing alternative output measures are recognized.
The following section moves beyond partial productivity and incorporates
total factor productivity and technical efficiency into the current narrative.

8.3. Empirical sensitivity of total factor productivity

and technical efficiency estimates across the banking
sector of BRIICS
Tables 8.12 to 8.28 in Appendix 2 present the estimates of the growth rate
in the Solow-style residual TFP index,25 aggregate technical efficiency scores
using DEA26 and the technical efficiency scores using a Cobb-Douglas sto-
chastic production frontier function. Due to space constraints, only the most
relevant findings and comments are mentioned here.27
The first issue under consideration is the growth rate in the TFP index using
different output measures, as shown in Tables 8.12, 8.15, 8.18, 8.21, 8.23 and
8.26 in Appendix 2. The initial observation is that the growth rate in the TFP
index shows broadly the same trend across all output measures. In very few
cases, one observes a shift in the trend itself, such as in Tables 8.12 (TFPG3),
8.21 (TFPG1) and 8.26 (TFPG1). Second, financial assets, number of total
deposit accounts and number of total loan accounts have caused much more
variability in the distribution of estimated growth in the TFP-index compared
to other output measures. Understandably, financial assets as an output variable
could embody large fluctuations due to risky investments by inherently volatile
banks. The finding for the number of deposit and loan accounts seems some-
what surprising. It may be on account of some structural shift in these variables.
This dimension requires further research in the future. Third, the nature of
skewness is very different depending on the output measure employed. Russia,
for example, displays a highly skewed distribution of the growth in the TFP
index across all output measures, as shown in Table 8.15. Except for it, all
other banking industries show multiple types of skewness in the data ranging
from highly skewed to fairly symmetrical. As shown in Table 8.12, for Brazil,
the distribution of the volume measures-based growth in the TFP index is
highly skewed, while the value measures-based index shows moderate to low
skewness. For Russia, while all the output measures show a high degree of
skewness, the volume measure shows positive while the value measures show
negative skewness. In the case of India, as shown in Table 8.18, skewness is very
high across most of the output measures, but it is fundamentally different for
value and volume measures. As discussed earlier in the previous chapters, the
value-volume difference in output measurement has many economic implica-
tions for any kind of performance analysis in banking. For China, as shown in
Table 8.23, except for the financial assets output measure (TFPG3), the rest of
238 Empirical case studies
the TFP growth estimates seem fairly symmetrical across the sample period.
The nature of skewness is the opposite in the case of TFPG1 (amount of depos-
its) and TFPG2 (amount of loans) for South Africa, as shown in Table 8.26.
Fourth, it is also necessary to verify if the information content of individual
estimated TFP growth series is homogenous or unique, and one crude way of
doing so is the degree of linear correlation among them. Estimates for Brazil
(Table 8.12) show that the TFP growth rates under alternative output meas-
ures are not firmly correlated, and some of the series even have an almost zero
correlation.
In the case of Russia (Table 8.15), however, the TFPG series with alterna-
tive output measures have very high and statistically significant correlations.
The Indian banking industry (Table 8.18) also shows a similar pattern. The
TFP series using the number of total loan accounts as output shows a moder-
ately high to high correlation with all other series. For the Indonesian bank-
ing industry (Table 8.21), the number of deposit accounts and the amount
of deposits based TFPG series is positively and significantly correlated. The
financial assets (TFPG3) and amount of loans (TFPG2) series are also highly
positively and significantly correlated. This is true across all the economies
except China. It is expected because the financial assets include both loans and
investments, which may be purely due to definitional similarities. In the case
of China (Table 8.23), in only a few cases is, a high and significant correlation
observed. Interestingly, the loan amount (TFPG2) and the number of deposit
accounts (TFPG4) based series show a high and significant correlation. This
has generally not been observed in other economies. Finally, in the case of
South Africa (Table 8.26), all the available series show a high and significant
correlation. Alternative output measures do not seem to be adding any new
information to the productivity performance in this case.
The second matter estimates industry-level average technical efficiency
scores across the sample period using DEA as contained in Tables 8.13, 8.16,
8.19, 8.22, 8.24 and 8.27, and with SFA as contained in Tables 8.14, 8.17,
8.20, 8.22, 8.25 and 8.28. The estimates in these tables reflect the average
efficiency scores across the sample period for the aggregate banking industry.
In the case of Brazil (Table 8.13), the mean level of efficiency across the sam-
ple period shows a reasonably good performance across all output measures,
whether value or volume. Variability, however, is widely different. It seems that
assets and the number of loan accounts-based TE scores are the most volatile
(TE3 and TE5). The correlation coefficients are primarily insignificant and
small. In the case of amount of deposits (TE2), efficiency scores positively
correlate with volume measures-based TE scores, i.e. with TE4 and TE5.28
Volume measures have a high and significant correlation, i.e. between TE4 and
TE5 in the Brazilian banking industry. Regarding the SFA-based measures,
the industry has, on average, performed well and seems to have been function-
ing near the average frontier. Variability is relatively low for all the reported
measures. Correlation is low in all the cases, possibly implying independent
information content across output measures. For the Russian banking industry,
 Banking: implications of using alternative output definitions 239
DEA-based efficiency scores do not seem to vary much across output measures
in terms of the mean levels and the variability as captured by CV. However, the
correlations are negligible, even if significant, possibly indicating differences
in the underlying trend in the efficiencies. Similar conclusions seem evident
for SFA-based measures in Table 8.17. In the context of India, as contained
in Table 8.19, performance indicators are compactly spread and do not seem
to diverge much. Only a few cases show a high and significant correlation.
Again, the amount of deposits is highly correlated with volume measures, and
volume measures show high interrelation among themselves (TE4 and TE5).
Similar results are seen for SFA-based measures, as shown in Table 8.20. Cor-
relations are very high and significant, and it does not appear that alternating
the output measures impacts the broad conclusions. Estimates for the Indone-
sian mainstream (excluding Islamic banking output) commercial banking sec-
tor reflect that efficiency scores using the amount of deposits and number of
deposit accounts show a smaller level of TE as compared to loans and assets
measures, while TE scores based on the same output measures (TE1 and TE4)
show much higher variability than the loans and asset-based scores.
Similarly, in the case of China (in Tables 8.24 and 8.25), the differences in
output measures do not seem to be inducing much variation in the findings.
CV estimates show differences, but they are not wide enough to warrant any
meaningful proof of output sensitivity. Lastly, in the South African banking
industry, estimates across different output measures show quite similar behav-
iour, though there is some evidence of unique information contents in terms of
correlations. Thus, the empirical sensitivity of efficiency estimates differs across
economies but broadly suggests similarity in the final results.

8.4. Empirical sensitivity of returns to scale estimates

across the banking sector of BRIICS
Industrial progress is an outcome of a large number of factors. One of them
is the nature of input-output relations as summarized by Economies of Scale
(EOS) for the industry as a whole. EOS can be approached from either the cost
or the production function route. This is generally done under the assump-
tion of duality. Hence, this study approaches the EOS issue via the production
function route using the concept of Returns to Scale (RTS). RTS does not
necessarily imply EOS. However, given the scope of this section, attention is
restricted to estimating RTS instead without further linking it to EOS.
Output expansion is generally a conscious and active objective of the mac-
roeconomic policy in emerging economies. Given the potentially large unu-
tilized capacities across many industries, expanding inputs and improving
technology is common in these economies. The banking industry also shows
similar characteristics to the EMEs, and thus increasing inputs to expand out-
put is a preferred strategy for improving the industry’s performance. Moreover,
given a sizeable stock of unutilized resources, understanding how the output
may change due to the increase in the use of these resources is necessary for
240 Empirical case studies
designing optimal growth and development policies. RTS provides an essen-
tial piece of information on this account. At both the microeconomic and
macroeconomic levels, the nature of RTS can have profound implications for
how an industry evolves and what kind of opportunities and challenges will it
create for its participants. The pace of growth is also, to a large extent deter-
mined by the nature of RTS. Decreasing RTS, for example, implies that larger
resource expenditures have to be incurred to achieve a given output level.
This may result in a lower growth rate in output despite large resource expen-
ditures. Achieving a particular industrial growth rate will mean expending
more resources than otherwise would be necessary under constant or increas-
ing RTS, thereby probably sacrificing efficiency. However, how RTS behaves
depends largely on how the output is defined for the banking industry. Given
the debates on output measurement and alternative viewpoints in the literature,
this section attempts to provide a brief and elementary comment on the rela-
tion between output definition and the consequent RTS under a production
function approach. RTS can be approached from two angles. First is the cost
function method, wherein the total cost function would be analysed with ref-
erence to output measures to locate the nature of scale expansion. The second
is through the production function, where the combined partial impacts of
each input on the total output show the kind of RTS that exists. The second
route has been chosen for this section. As discussed earlier and explained in the
underlying estimates to the results presented here, alternative output measures
have been employed. A two-inputs and one-output production function, spec-
ified in Cobb-Douglas and trans-logarithmic29 forms, using traditional OLS,
ridge regression and SFA are estimated. Appropriate methods to estimate the
RTS coefficients are employed, and the results have been reported in the tables
that follow. The underlying estimates are presented in Appendix 3. The inputs
are restricted to labour (number of banking employees)30 and physical capital
(number of branches).31 This has been done to present the most basic produc-
tion model for this empirical case study. By keeping the framework essential in
terms of variable complexity, the focus has been diverted to the issue of how
different output measures cause changes in the RTS estimates if any.
The use of both Cobb-Douglas and trans-log production function specifica-
tions is in line with the literature. While the trans-log function has primarily
been favoured in literature, given the aggregated nature of data, Cobb-Douglas
specification seems to be a reasonably representative theorization of the under-
lying true production function. As is well known in the literature, the trans-log
production function is characterized by very high multicollinearity. This can
cause many well-recognized econometric problems in the quality of the esti-
mated coefficients. Hence, the ridge regression method was also employed to
deal with this issue. This method allows a conscious trade-off between bias and
efficiency of the OLS estimators in order to reduce multicollinearity among
the variables by penalizing the OLS estimators. Critiques have argued that the
lasso regression is a better approach due to the non-orthogonality of the ridge
estimators. However, lasso method-based estimates were also examined, and
 Banking: implications of using alternative output definitions 241
the results were broadly the same.32 Given the reasonably simple interpreta-
tion of ridge regression estimators, the same has been used in this section and
section 9.4. Time series data have been used from 2000 to 2020. Lastly, data
constraints have greatly affected the choice of input and output variables.
The specifications33 of the production functions used in this section are as
below:

In(Yi) = In(A) + b1 In(Li) + b2 In(Ki) + ui(3)

1 1
In ( Yi ) = b0 + bL In ( L ) + bK In ( K ) + bLL ( InL ) + bKK ( InK ) + bLK ( InL ) * ( InK ) ui(4)
2 2

2 2
Where i = the banking industry in a given year, A and b0 are the constant
parameters, ui is the error term, L = number of employees in the aggregate
banking industry, and K = number of branches in the banking industry.34

Brazil
Table 8.29 in Appendix 3 summarizes the findings on the variations in RTS
estimates using different measures of banking industry outputs. The underlying
estimations are also provided in Appendix 3 in Tables 8.35 to 8.39. The broad
conclusion is that the banking industry in Brazil has recorded increasing returns
to scale at the aggregate level during the sample period of 2000 to 2020. Two
perspectives can be suggested here. In terms of each output variable, the RTS
coefficient’s robustness across different models is the robustness. In terms of
each model, though, some variations in the RTS value and direction are vis-
ible. Using the financial assets, in particular, yields divergent estimates. In terms
of the quantitative size of RTS, variations are clearly evident when alternative
output measures are employed. The loan accounts measure is remarkably diver-
gent from the average behaviour across other output measures.

Russia
For the Russian banking industry as a whole, Table 8.30 shows that there
is no variation in the direction of RTS. Still, the quantitative size of the
RTS coefficient varies considerably across models and output measures even
while ignoring Model 1, which is conflated with severely high multicol-
linearity.35 Intra-model variations in the value of RTS across output meas-
ures are much wider than intra-output variations across alternative models.
Estimations underlying Table 8.30 are provided across Tables 8.40 to 8.43
in Appendix 3.

India
Table 8.31 presents the Indian commercial banking industry results with the
related empirical estimations contained in Tables 8.44 to 8.48. Compared to
242 Empirical case studies
Brazil and Russia, the results here are much more similar in terms of RTS’s
direction and its coefficient value. Using different output measures does not
seem to be causing much of a variation in RTS estimates in the case of India.

Indonesia
In the case of the aggregate Indonesian banking industry, as shown in Table 8.32
and complemented by related results in Tables 8.49 to 8.53, considerable vari-
ations in RTS estimates can be located. Whether across each model or out-
put variable, the RTS coefficient varies in terms of the direction (increasing,
decreasing or constant) and size. Compared to the economies covered previ-
ously, the Indonesian commercial banking sector shows considerable output
sensitivity in estimates of RTS.

China
The Chinese commercial banking industry also shows wide variations com-
pared to Brazil, Russia and India. Table 8.33 shows the main results, while
Tables 8.54 to 8.58 provide the supplementary information. Model 1 is pro-
ducing inconsistent results across all countries due to the issue of high multicol-
linearity. There is a good amount of variation in the RTS coefficient when the
amount of deposits and loans are used versus when assets and volume measures
are used.36

South Africa
As shown in Table 8.34, the South African banking industry does not show
much variation in the RTS estimate across different models and output meas-
ures. Its supporting estimations are contained in Tables 8.59 to 8.62.

8.5. Cross-country analysis
Lastly, the six economies were utilized as a panel37 for analysing the average
growth rate of the Malmquist TFP index DEA across the sample period. The
results are provided in Table 8.63 in Appendix 4. It seems that changing the
output measures has a considerable effect on the nature of RTS and its size.
China and India, in particular, show higher differences in RTS estimates com-
pared to the aggregate banking industry of other economies.
The empirical case studies on the difference in performance estimates caused
by alternating output measures for the aggregate banking industries across each
of the BRIICS countries are thus concluded.
Appendix 1
Estimates of labour and capital (partial
factor) productivities for section 8.2

Table 8.1 Estimated measures of Labour Productivity of the banking industry in Brazil dur-
ing the sample period

Value Measures

Statistic LP1 LP2 LP3 LP4

Mean 2.87 0.43 2.44 7.29

SD 0.55 0.07 0.85 1.64
CV 19.26 16.27 34.87 22.46

Volume measures

Statistic LP5 LP6 LP7 LP8

Mean 161.69 297.74 96.07 487.07
SD 37.75 82.28 67.28 351.13
CV 23.35 27.63 70.03 72.09
Notes: LP1 is millions of outstanding of 2010 Brazilian reals of outstanding deposits per bank employee;
LP2 is millions of 2010 Brazilian reals of net income of commercial banks per bank employee; LP3 is
millions of 2010 Brazilian reals outstanding loans from commercial banks per bank employee; LP4 is mil-
lions of 2010 Brazilian reals of financial assets of commercial banks per bank employee; LP5 is Number
of Depositors per Bank Employee; LP6 is Number of Deposit Accounts per Bank Employee; LP7 is
Number of Borrowers from Commercial Banks per Bank Employee; LP8 is Number of Loan Accounts
per Bank Employee; SD is standard deviation and CV is coefficient of variation measured in percentages.
Source: Author’s estimation.

Table 8.2 Estimated measures of Physical Capital Productivity of the banking industry in
Brazil during the sample period

Value Measures

Statistic CPR1 CPR2 CPB1 CPB2

Mean 9.76 1.56 49.49 7.49

SD 2.06 0.66 8.98 1.34
CV 21.12 42.09 18.14 17.93
(Continued)
244 Empirical case studies
Table 8.2 (Continued)

Volume Measures

Statistic CPR4 CPR5 CPB4 CPB5

Mean 540.12 988.23 4382.52 13246.01

SD 78.3 165.77 1575.59 6769.78
CV 14.5 16.77 35.95 51.11
Notes: CPR1 is millions of 2010 Brazilian reals of outstanding deposits of commercial banks per Retail
Agent Outlet; CPR2 is millions of 2010 Brazilian reals of net income per retail agent outlet; CPR3 is
millions of 2010 Brazilian reals of total assets per retail agent outlet; CPR4 is the number of depositors
per retail agent outlet; CPR5 is the number of deposit accounts per retail agent outlet; CPB1 is millions
of 2010 Brazilian reals of outstanding deposits per commercial bank branch; CPB2 is millions of 2010
Brazilian reals of net income per commercial bank branch; CPB3 is millions of 2010 Brazilian reals of
outstanding loans per commercial bank branch; CPB4 is the number of total account holder (deposit
and loans) with commercial banks per commercial bank branch; CPB5 is the number of depositors
with commercial banks per commercial bank branch; SD is standard deviation, and CV is coefficient of
variation which is measured in percentages.
Source: Author’s estimation.

Table 8.3 Estimated measures of labour productivity of the Russian banking industry during
the sample period

Statistic Value Measure Volume Measure

LP1 LP2 LP3 LP4

Mean 35728.35 39279.85 20.67 384860.07

SD 26755.01 23528.19 13.35 63723.63
CV 74.88 59.90 64.61   16.56
Notes: LP1 is million rubles of outstanding deposits per employee in financial services; LP2 is million
rubles of outstanding loans per employee in financial services; LP3 is million rubles of total bank assets
per employee in financial services; LP4 is the number of bank total deposit bank accounts per employee
in financial series; SD is standard deviation, and CV is coefficient of variation which is measured in
percentages.
Source: Author’s estimation.

Table 8.4 Estimated measures of Physical Capital Productivity of the Russian banking indus-
try during the sample period

Measure Value

Statistic CPB1 CPB2 CPB3 CPR1 CPA1

Mean 2032.26 2170.44 1.15 42607.75 523.95

SD 1891.73 1698.27 0.93 39624.09 161.12
CV 93.08 78.25 81.13 93 30.75
 Banking: implications of using alternative output definitions 245

Measure Volume

Statistic CPB4 CPR2 CPA2

Mean 19843.12 441931.72 8567.06

SD 7103.11 208026.85 4989.4
CV 35.8 47.07 58.24
Notes: CPB1 is millions of 2010 rubles of outstanding deposits per commercial bank branch; CPB2
is millions of 2010 rubles of outstanding loans per commercial bank branch; CPB3 is millions of 2010
rubles of total bank assets per commercial bank branch; CPB4 is the number of deposit bank accounts per
commercial bank branch; CPR1 is millions of 2010 rubles of outstanding deposits per non-branch retail
agent outlet; CPR2 is the number of deposit bank accounts per employee in financial services; CPA1
is millions of Rubles of outstanding deposits per ATM; CPA2 is the number of deposit bank accounts
per ATM; SD is standard deviation, and CV is coefficient of variation which is measured in percentages.
Source: Author’s estimation.

Table 8.5 Estimated measures of Labour Productivity of the Indian banking industry during
the sample period

Statistic Value Volume

LP1 LP2 LP3 LP4 LP5 LP6

Mean 43.23 31.94 55.38 869.41 115.99 985.40

SD 6.85 6.44 8.61 334.13 25.40 355.46
CV 15.85 20.15 15.54 38.43 21.90 36.07
Notes: LP1 is millions of 2010 INR of outstanding deposits per banking employee; LP2 is millions of
2010 INR of outstanding loans per banking employee; LP3 is millions of 2010 INR of total assets per
banking employee; LP4 is the number of deposit accounts per banking employee; LP5 is the number of
loan accounts per banking employee; LP6 is the number of total accounts (deposit and loans) per banking
employee; SD is standard deviation, and CV is coefficient of variation which is measured in percentages.
Source: Author’s estimation.

Table 8.6 Estimated measures of Physical Capital Productivity of the Indian banking indus-
try during the sample period

Statistic Output Perspective

Value Volume

CPB1 CPB2 CPA1 CPB2 CPA2

Mean 1659.26 1229.93 810.11 38989.5 13822.89

SD 411.25 353.41 479.79 19243.37 5776.84
CV 24.79 28.73 59.22 49.36 41.79
Notes: CPB1 is millions of 2010 INR of outstanding deposits per branch of commercial banks; CPB2
is millions of 2010 INR of outstanding loans per branch of commercial banks; CPB3 is the number of
total banking accounts (deposit and loans) per branch of commercial banks; CPA1 is millions of 2010
rupees of outstanding deposits per ATM; CPA2 is the number of deposit accounts per ATM; SD is
standard deviation, and CV is coefficient of variation which is measured in percentages.
Source: Author’s estimation.
246 Empirical case studies
Table 8.7 Estimated measures of Physical Capital Productivity of the Indonesian banking
industry during the sample period

Statistic Value Volume

CPR1 CPB1 CPB2 CPR2 CPB3

Mean 37251.1 167563.63 115600.37 2125.66 10576.15

SD 54468.94 85518.09 33424.15 2601.57 3725.2
CV 146.22 51.04 28.91 122.39 35.22
Notes: CPR1 is millions of 2010 rupiah of outstanding deposits per non-branch retail agent outlet;
CPR2 is the number of deposit accounts per non-branch retail agent outlet; CPB1 is millions of 2010
rupiah of outstanding deposits per branch of commercial banks; CPB2 is millions of 2010 rupiah of
outstanding loans per branch of commercial banks; CPB3 is the number of total accounts (deposit and
loans) per branch of commercial banks; SD is standard deviation, and CV is coefficient of variation
which is measured in percentages.
Source: Author’s estimation.

Table 8.8 Estimated measures of Labour Productivity of the Chinese banking industry dur-
ing the sample period

Statistic Value Measures Volume Measures

LP1 LP2 LP3 LP4 LP5 LP6

Mean 17.52 11.82 0.64 4.63 136.03 140.66

SD 6.89 5.08 0.19 2.26 35.90 34.99
CV 39.33 43.01 30.05 48.76 26.40 24.88
Notes: LP1 is millions of 2010 yuan of outstanding deposits per bank employee; LP2 is millions of 2010
yuan of outstanding loans per bank employee; LP3 is millions of 2010 yuan of total assets per bank
employee; LP4 is the number of depositors per bank employee; LP5 is the number of borrowers per
bank employee; LP6 is the number of customers (sum of depositors and creditors) per bank employee;
SD is standard deviation, and CV is coefficient of variation which is measured in percentages.
Source: Author’s estimation.

Table 8.9 
Estimated measures of Physical Capital Productivity of the Chinese banking
industry during the sample period

Measure Value Volume

Statistic CPB1 CPB2 CPA1 CPB3 CPA2

Mean 676.31 458.71 170.98 5027.67 43.48

SD 347.26 255.6 59.24 815.58 12.39
CV 51.35 55.72 34.65 16.22 28.5
Notes: CPB1 is millions of 2010 yuan of outstanding deposits per commercial bank branch; CPB2 is
millions of 2010 yuan of outstanding loans per commercial bank branch; CPB3 is the number of cus-
tomers per commercial bank branch; CPA1 is millions of 2010 yuan of outstanding deposits per ATM;
CPA2 is the number of depositors per commercial bank branch; SD is standard deviation, and CV is
coefficient of variation which is measured in percentages.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 247
Table 8.10 Estimated measures of Labour Productivity of the South African banking indus-
try during the sample period

Measure Value Volume

Statistic LP1 LP2 LP3 LP4 LP5

Mean 0.71 1.09 0.01 21.47 10.67

SD 0.20 0.35 0.00 5.75 4.00
CV 28.06 32.60 31.68 26.79 37.52
Notes: LP1 is millions of 2010 rands of outstanding deposits of commercial banks per bank employee;
LP2 is millions of 2010 rands of outstanding loans per bank employee; LP3 is millions of 2010 rands of
total assets per bank employee; LP4 is the number of deposit accounts per bank employee; LP5 is the
number of loan accounts per bank employee; SD is standard deviation, and CV is coefficient of variation
which is measured in percentages.
Source: Author’s estimation.

Table 8.11 Estimated measures of Physical Capital Productivity of the South African bank-
ing industry during the sample period

Measure Value Volume

Statistic CPB1 CPB2 CPA1 CPB3 CPB4 CPB5 CPA2

Mean 409.83 624.18 75.38 12412.65 6146.89 18559.53 2162.37

SD 97.94 177.87 27.30 3458.18 2156.70 2374.53 391.14
CV 23.90 28.50 36.22 27.86 35.09 12.79 18.09
Notes: CPB1 is millions of 2010 rands of outstanding deposits per commercial bank branch; CPB2
is millions of 2010 rands of outstanding loans per commercial bank branch; CPB3 is the number of
deposit accounts per commercial bank branch; CPB4 is the number of loan accounts per commercial
bank branch; CPB5 is the number of total accounts per commercial bank branch; CPA1 is millions of
2010 rands of outstanding deposits per ATM; CPA2 is the number of deposit accounts per ATM; SD is
standard deviation, and CV is coefficient of variation which is measured in percentages
Source: Author’s estimation.
Appendix 2
Estimates of residual total factor
productivity index and technical
efficiency scores for the section 8.3

Table 8.12 Residual TFP estimates under the Solow framework for the Brazilian banking
industry

Descriptive Statistics

Variable TFPG1 TFPG2 TFPG3 TFPG4 TFPG5

Mean 8.72 2.79 −1.83 1.64 8.97

SD 25.49 9.94 10.20 3.35 38.63
CV 292.21 356.55 555.92 204.05 430.55
Skewness 0.86 −0.56 0.00 −1.32 1.80

Cross-Correlations

Variable TFPG1 TFPG2 TFPG3 TFPG4

TFPG2 0.59*** - - -
TFPG3 0.34 0.46** - -
TFPG4 0.01 0.55 −0.02 -
TFPG5 0.07 0.40* 0.29 0.14
Notes: 1. All the variables are measured as yearly growth rates and are expressed in percentages; 2.
TFPG1 is the TFP growth index with the outstanding deposits of commercial banks as an output,
TFPG2 is the TFP growth index with the outstanding loans from commercial banks as an output,
TFPG3 is the TFP growth index with the financial assets of commercial banks as an output, TFPG4
is the TFP growth index with the number of deposit accounts with commercial banks as an output,
TFPG5 is the TFP growth index with the number of loan accounts with commercial banks as an out-
put; 3. SD stands for standard deviation, and CV stands for coefficient of variation; 4. Wherever CV was
negative, because of negative mean values, the absolute value of the same has been used; 5. *, **, and ***
indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 249
Table 8.13 Technical efficiency estimates using multi-stage DEA for the banking industry
in Brazil

Statistic Descriptive Statistic

TE1 TE2 TE3 TE4 TE5

Mean 0.88 0.94 0.85 0.98 0.66

SD 0.10 0.06 0.14 0.03 0.33
CV 11.79 5.95 16.99 2.78 50.09
Skewness −0.20 −1.14 −0.61 −1.58 −0.54

Variables Cross-correlations

TE1 TE2 TE3 TE4

TE2 0.07 - - -
TE3 0.49* 0.08 - -
TE4 −0.26 0.67*** −0.25 -
TE5 −0.12 0.71*** −0.29 0.52**
Notes: 1. These estimates have been worked out using multi-stage output-oriented DEA under VRS
assumption; 2. TE1 is TE index with outstanding deposits as an output variable, TE2 is TE index with
outstanding loans as an output variable, TE3 is TE index with financial or total assets of banks, TE4 is
TE index with deposit accounts with commercial banks as an output variable, TE5 is TE index with
loan accounts with commercial banks as an output variable; 3. SD is standard deviation, and CV is
coefficient of variation; 4. All the data are in index form, and hence unit-less, 5. *, **, and *** indicate
significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.

Table 8.14 Technical efficiency estimates using the Stochastic Frontier model for the bank-
ing industry in Brazil

Statistic TE1 TE2 TE3

Mean 0.96 0.95 0.89

SD 0.02 0.04 0.10
CV 2.57 4.46 10.74
Skewness −1.05 −0.85 −1.12

Variables TE1, TE2 TE2, TE3 TE3, TE1

Correlation Coefficient 0.06 −0.40* −0.39*
Notes: 1. All the statistics are in index form and hence are unitless; 2. TE1 is TE index using SFA with
deposit accounts as an output variable, TE2 is TE index using SFA with loan accounts as an output
variable, TE3 is TE index using SFA with financial assets of commercial banks as an output variable
wherein the amount of deposits is used as one of the input variables; 3. SD is standard deviation, and
CV is coefficient of variation; 4. *, ** and *** indicate significance at 10%, 5% and 1% significance levels,
respectively.
Source: Author’s estimation.
250 Empirical case studies
Table 8.15 Residual TFP estimates under the Solow framework for the Russian banking
industry

Statistic Descriptive Statistics

Mean 11.03 9.80 6.46 1.61

SD 18.63 19.72 14.35 4.79
CV 168.90 201.19 222.23 297.76
Skewness −1.52 −1.34 −1.14 1.66

Cross-correlations

Variable TFPG1 TFPG2 TFPG3

TFPG2 0.91 ***
- -
TFPG3 0.72*** 0.81*** -
TFPG4 0.54** 0.25 0.25
Notes: 1. All the variables are measured as yearly growth rates and are expressed in percentages; 2.
TFPG1 is the TFP growth index with the outstanding deposits of commercial banks as an output,
TFPG2 is the TFP growth index with the outstanding loans from commercial banks as an output,
TFPG3 is the TFP growth index with the financial assets of commercial banks as an output, TFPG4
is the TFP growth index with the number of deposit accounts with commercial banks as an output
variable; 3. SD stands for standard deviation, and CV stands for coefficient of variation; 4. CV might
be negative as the data are in growth-rate form; 5. *, ** and *** indicate significant at 10%, 5%, and 1%
significance levels, respectively.
Source: Author’s estimation.

Table 8.16 Technical efficiency estimates using DEA for the banking industry in Russia

Statistic Descriptive Statistics

TE1 TE2 TE3 TE4

Mean 0.94 0.95 0.94 0.997

SD 0.10 0.04 0.05 0.01
CV 10.26 4.07 5.80 0.69
Skewness −1.56 −0.60 −0.18 −2.36

Variable Cross-correlations

TE1 TE2 TE3

TE2 −0.16 - -
TE3 0.15 0.51** -
TE4 0.37* −0.25 −0.21
Notes: 1. These estimates have been worked out using multi-stage output-oriented DEA under VRS
assumption; 2. TE1 is TE index with outstanding deposits as an output variable, TE2 is TE index with
outstanding loans as an output variable, TE3 is TE index with financial or total assets of banks, TE4 is
TE index with deposit accounts with commercial banks as an output variable; 3. SD is standard devia-
tion, and CV is Coefficient of Variation; 4. All the data are in index form, and hence unit-less, 5. *, **,
and *** indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 251
Table 8.17 Technical efficiency estimates using the stochastic Frontier model for the banking
industry in Russia

Statistic Descriptive Statistics

TE1 TE2 TE3 TE4

Mean 0.87 0.94 0.93 0.94

SD 0.09 0.04 0.05 0.04
CV 10.32 4.47 5.54 4.05
Skewness −1.32 −0.81 −0.71 −1.55

Variable Cross-correlations

TE1 TE2 TE3

TE2 0.03 - -
TE3 −0.10 −0.06 -
TE4 0.90 −0.32 0.10
Notes: 1. All the statistics are in index form and hence are unitless; 2. TE1 is TE index using SFA with
the amount of outstanding deposits as an output variable, TE2 is TE index using SFA with the amount
of outstanding loans as an output variable, TE3 is TE index using SFA with financial assets of com-
mercial banks as an output variable and TE4 is TE index using SFA with the total number of deposit
accounts as an output variable; 3. SD is standard deviation, and CV is coefficient of variation; 4. *, **,
and *** indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.

Table 8.18 Residual TFP estimates under the Solow framework for the Indian banking
industry

Statistic Descriptive Statistics

TFPG1 TFPG2 TFPG3 TFPG4 TFPG5

Mean 1.88 1.41 0.73 5.04 1.07

SD 10.87 7.47 7.10 6.55 21.66
CV 578.75 531.64 973.43 129.93 2031.42
Skewness −1.99 −1.58 −1.67 0.11 3.28

Variable Cross-correlations

TFPG1 TFPG2 TFPG3 TFPG4

TFPG2 0.73 ***
- - -
TFPG3 0.33 0.59*** - -
TFPG4 0.60*** 0.17 0.11 -
TFPG5 −0.79*** −0.81*** −0.60*** −0.44**
Notes: 1. All the variables are measured as yearly growth rates and are expressed in percentages; 2.
TFPG1 is the TFP growth index with the outstanding deposits of commercial banks as an output,
TFPG2 is the TFP growth index with the outstanding loans from commercial banks as an output,
TFPG3 is the TFP growth index with the financial assets of commercial banks as an output, TFPG4
is the TFP growth index with the number of deposit accounts with commercial banks as an output
variable, TFPG5 is the yearly growth rate of TFP index with the total number of loan accounts with
commercial banks as an output variable; 3. SD stands for standard deviation, and CV stands for coef-
ficient of variation; 4. CV might be negative as the data are in growth-rate form; 5. *, ** and *** indicate
significant at 10%, 5%, and 1% significance levels, respectively.
Source: Author’s estimation.
252 Empirical case studies
Table 8.19 Technical efficiency estimates using multi-stage DEA for the banking industry
in India

Statistic Descriptive Statistics

TE1 TE2 TE3 TE4 TE5

Mean 0.97 1.00 0.99 0.95 0.90

SD 0.04 0.01 0.03 0.07 0.12
CV 4.50 1.13 3.18 7.31 12.83
Skewness −1.34 −2.91 −3.17 −1.47 −0.68

Variable Cross-correlations

TE1 TE2 TE3 TE4

TE2 0.36 - - -
TE3 0.24 0.25 - -
TE4 0.55 −0.01 −0.09 -
TE5 0.50 0.37 −0.08 0.78
Notes: 1. These estimates have been worked out using multi-stage output-oriented DEA under VRS
assumption; 2. TE1 is TE index with outstanding deposits as an output variable, TE2 is TE index with
outstanding loans as an output variable, TE3 is TE index with financial or total assets of banks, TE4
is TE index with deposit accounts with commercial banks as an output variable, TE5 is the TE index
with loan accounts with commercial banks as an output variable; 3. SD is standard deviation, and CV
is coefficient of variation; 4. All the data are in index form, and hence unit-less, 5. *, **, and *** indicate
significant at 10%, 5% and 1% levels of significance respectively.
Source: Author’s estimation.

Table 8.20 Technical efficiency estimates using the Stochastic Frontier model for the bank-
ing industry in India

Statistic Descriptive Statistics

TE1 TE2 TE3

Mean 0.84 0.84 0.86

SD 0.10 0.12 0.10
CV 12.23 14.25 11.94
Skewness −0.34 −0.73 −0.48

Variable Cross-correlations

TE1 TE2
TE2 0.97 ***
-
TE3 0.91*** 0.90***
Notes: 1. All the statistics are in index form and hence are unitless; 2. TE1 is TE index using SFA with
the amount of outstanding deposits as an output variable, TE2 is TE index using SFA with the amount
of outstanding loans as an output variable, TE3 is TE index using SFA with financial assets of commer-
cial banks as an output variable; 3. SD is standard deviation, and CV is coefficient of variation; 4. *, **
and *** indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 253
Table 8.21 Residual TFP estimates under the Solow framework for the Indonesian banking
industry

Statistic Descriptive Statistics

TFPG1 TFPG2 TFPG3 TFPG4 TFPG5

Mean −2.74 1.25 1.18 1.62 8.11

SD 13.52 10.15 15.96 18.33 87.50
CV −493.25 814.39 1351.58 1129.54 1079.33
Skewness −1.23 1.31 1.65 0.56 3.59

Variable Cross-correlations

TFPG1 TFPG2 TFPG3 TFPG4

TFPG2 −0.10 - - -
TFPG3 −0.36 0.83*** - -
TFPG4 0.78*** −0.13 −0.29 -
TFPG5 0.12 −0.21 −0.17 −0.04
Notes: 1. All the variables are measured as yearly growth rates and are expressed in percentages; 2. TFPG1 is
the TFP growth index with the outstanding deposits of commercial banks as an output, TFPG2 is the TFP
growth index with the outstanding loans from commercial banks as an output, TFPG3 is the TFP growth
index with the financial assets of commercial banks as an output, TFPG4 is the TFP growth index with
the number of deposit accounts with commercial banks as an output variable, TFPG5 is the yearly growth
rate of TFP index with the total number of loan accounts with commercial banks as an output variable;
3. SD stands for standard deviation, and CV stands for coefficient of variation; 4. CV might be negative as
the data are in growth-rate form; 5. *, ** and *** indicate significant at 10%, 5%, and 1% levels, respectively.
Source: Author’s estimation.

Table 8.22 Technical efficiency estimates using multi-stage DEA and SFA for the banking
industry in Indonesia

Statistic Descriptive Statistics TE using SFA#

TE1 TE2 TE3 TE4 TE5 TE2

Mean 0.74 0.98 0.97 0.70 0.97 0.97

SD 0.19 0.03 0.05 0.23 0.04 0.02
CV 26.26 2.87 4.76 32.72 3.93 1.95
Skewness −0.29 −1.79 −2.70 0.12 −1.66 −1.03

Variable Cross-correlations

TE1 TE2 TE3 TE4

TE2 0.40*** - - -
TE3 0.19 0.82*** - -
TE4 0.87*** 0.21 0.03 -
TE5 0.07 0.52** 0.71*** −0.04
Notes: 1. These estimates have been worked out using multi-stage output-oriented DEA under VRS
assumption; 2. TE1 is TE index with outstanding deposits as an output variable, TE2 is TE index with
outstanding loans as an output variable, TE3 is TE index with financial or total assets of banks, TE4
is TE index with deposit accounts with commercial banks as an output variable, TE5 is the TE index
with loan accounts with commercial banks as an output variable; 3. SD is standard deviation, and CV
is coefficient of variation; 4. All the data are in index form, and hence unitless, 5. *, **, and *** indicate
significant at 10%, 5% and 1% levels of significance, respectively; 6. # – SFA estimates were found sta-
tistically permissible only for the loan amount variable. The gamma statistic of the rest of the estimated
frontier equations was very small and statistically highly insignificant.
Source: Author’s estimation.
254 Empirical case studies
Table 8.23 Residual TFP estimates under the Solow framework for the banking industry
of China

Statistic Descriptive Statistics

TFPG1 TFPG2 TFPG3 TFPG4 TFPG5

Mean 11.74 16.80 4.66 13.14 1.24

SD 19.55 25.83 16.72 28.70 6.57
CV 166.54 153.76 358.77 218.51 529.25
Skewness −0.49 −0.41 −2.41 −0.38 −0.18

Variable Cross-correlations

TFPG1 TFPG2 TFPG3 TFPG4

TFPG2 0.93*** - - -
TFPG3 −0.32 −0.28 - -
TFPG4 0.76*** 0.87*** −0.31 -
TFPG5 −0.39* −0.18 −0.01 −0.29
Notes: 1. All the variables are measured as yearly growth rates and are expressed in percentages; 2.
TFPG1 is the TFP growth index with the outstanding deposits of commercial banks as an output,
TFPG2 is the TFP growth index with the outstanding loans from commercial banks as an output,
TFPG3 is the TFP growth index with the financial assets of commercial banks as an output, TFPG4
is the TFP growth index with the number of deposit accounts with commercial banks as an output
variable, TFPG5 is the yearly growth rate of TFP index with the total number of borrowers from com-
mercial banks as an output variable; 3. SD stands for standard deviation, and CV stands for coefficient of
variation; 4. CV might be negative as the data are in growth-rate form; 5. *, ** and *** indicate significant
at 10%, 5%, and 1% significance levels, respectively.
Source: Author’s estimation.

Table 8.24 Technical efficiency estimates using multi-stage DEA banking industry in China

Statistic Descriptive Statistics

TE1 TE2 TE3 TE4 TE5

Mean 0.92 0.96 0.97 1.00 1.00

SD 0.12 0.05 0.06 0.00 0.00
CV 12.97 4.86 6.55 0.12 0.10
Skewness −1.76 −1.11 −1.79 −1.20 −1.46

Variable Cross-correlations

TE1 TE2 TE3 TE4

TE2 −0.01 - - -
TE3 0.28 0.38* - -
TE4 0.74*** 0.32 0.71*** -
TE5 0.34 0.49** 0.82*** 0.86***
Notes: 1. These estimates have been worked out using multi-stage output-oriented DEA under VRS
assumption; 2. TE1 is TE index with outstanding deposits as an output variable, TE2 is TE index with
outstanding loans as an output variable, TE3 is TE index with financial or total assets of banks, TE4 is
TE index with deposit accounts with commercial banks as an output variable, TE5 is the TE index with
the total number of borrowers with commercial banks as an output variable; 3. SD is standard deviation,
and CV is coefficient of variation; 4. All the data are in index form, and hence unit-less, 5. *, **, and ***
indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 255
Table 8.25 Technical efficiency estimates using the Stochastic Frontier model for the bank-
ing industry in China

Statistic Descriptive Statistics

TE1 TE2 TE3

Mean 0.73 0.74 0.82

SD 0.19 0.19 0.12
CV 25.64 25.95 14.40
Skewness −0.47 −0.63 −1.42

Variable Cross-correlations

TE1 TE2

TE2 0.98 -
TE3 0.48 0.47
Notes: 1. All the statistics are in index form and hence are unitless; 2. TE1 is TE index using SFA with
the amount of outstanding deposits as an output variable, TE2 is TE index using SFA with the amount
of outstanding loans as an output variable, TE3 is TE index using SFA with financial assets of commer-
cial banks as an output variable; 3. SD is standard deviation, and CV is coefficient of variation; 4. *, **
and *** indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.

Table 8.26 Residual TFP estimates under the Solow framework for the banking industry
of South Africa

Statistic Descriptive Statistics

TFPG1 TFPG2 TFPG3

Mean −1.03 0.79 5.43

SD 10.36 17.00 15.58
CV −1008.50 2154.75 287.00
Skewness −1.24 2.42 −0.50

Variable Cross-correlations

TFPG1 TFPG2

TFPG2 −0.83*** -
TFPG3 0.68*** −0.66***
Notes: 1. All the variables are measured as yearly growth rates and are expressed in percentages; 2.
TFPG1 is the TFP growth index with the outstanding deposits of commercial banks as an output,
TFPG2 is the TFP growth index with the outstanding loans from commercial banks as an output,
TFPG3 is the TFP growth index with the number of deposit accounts with commercial banks as an
output variable; 3. SD stands for standard deviation, and CV stands for coefficient of variation; 4. CV
might be negative as the data are in growth-rate form; 5. *, ** and *** indicate significant at 10%, 5% and
1% significance levels, respectively.
Source: Author’s estimation.
256 Empirical case studies
Table 8.27 Technical efficiency estimates using multi-stage DEA for the banking industry
in South Africa

Statistic Descriptive Statistics

TE1 TE2 TE3 TE4 TE5

Mean 0.95 0.98 0.96 0.92 0.91

SD 0.05 0.03 0.04 0.12 0.08
CV 5.22 3.02 4.44 12.96 8.28
Skewness −0.71 −2.74 −0.96 −1.38 −0.46

Variable Cross-correlations

TE1 TE2 TE3 TE4

TE2 0.39* - - -
TE3 0.16 0.82*** - -
TE4 0.60*** 0.07 −0.07 -
TE5 0.07 −0.14 0.01 −0.32
Notes: 1. These estimates have been worked out using multi-stage output-oriented DEA under VRS
assumption; 2. TE1 is TE index with outstanding deposits as an output variable, TE2 is TE index with
outstanding loans as an output variable, TE3 is TE index with financial or total assets of banks, TE4
is TE index with deposit accounts with commercial banks as an output variable, TE5 is the TE index
with the total number of loan accounts with commercial banks as an output variable; 3. SD is standard
deviation, and CV is coefficient of variation; 4. All the data are in index form, and hence unit-less, 5. *, **,
and *** indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.

Table 8.28 Technical efficiency estimates using the Stochastic Frontier model for the bank-
ing industry in South Africa

Statistic Descriptive Statistics

TE1 TE2 TE3 TE4 TE5

Mean 0.89 0.92 0.94 0.83 0.91

SD 0.06 0.06 0.04 0.12 0.07
CV 7.18 6.71 4.04 14.98 7.90
Skewness −1.10 −2.18 −1.88 −0.36 −0.41

Variable Cross-correlations

TE1 TE2 TE3 TE4

TE2 0.88*** - - -
TE3 0.09 0.39* - -
TE4 0.66*** 0.56*** −0.02 -
TE5 0.09 0.01 −0.07 −0.21
Notes: 1. All the statistics are in index form and hence are unitless; 2. TE1 is TE index using SFA with
the amount of outstanding deposits as an output variable, TE2 is TE index using SFA with the amount
of outstanding loans as an output variable, TE3 is TE index using SFA with financial assets of com-
mercial banks as an output variable, TE4 is TE index using SFA with deposit accounts as the output
variable, TE5 is the TE index using SFA with loan accounts as the output variable; 3. SD is standard
deviation, and CV is coefficient of variation; 4. *, ** and *** indicate significant at 10%, 5% and 1% levels
of significance, respectively.
Source: Author’s estimation.
Appendix 3
Estimated models for the returns to
scale coefficients in section 8.4

Table 8.29 Returns to scale for the banking industry of Brazil under alternative output
measures across alternative econometric models

Output Model 1 Model 2 Model 3 Model 4

Variable
Trans-log model Trans-log model Cobb-Douglas Stochastic Frontier
under standard under Ridge model under Cobb-Douglas
OLS Regression standard OLS model

Deposits −0.92 2.52 9.67 1.45

(DEP) Negative Increasing Increasing Increasing
Loans 2.30 4.37 4.42 4.30
(LOAN) Increasing Increasing Increasing Increasing
Financial −6.62 −2.81 0.67 NA
Assets Negative Negative Decreasing
(ASSET)
Deposit 5.58 5.78 5.72 4.42
Accounts Increasing Increasing Increasing Increasing
(DACT)
Loan 6.95 15.79 20.14 NA
Accounts Increasing Increasing Increasing
(LACT)
Source: Based on the author’s estimation.

Table 8.30 Returns to scale for the banking industry of Russia under alternative output
measures across alternative econometric models

Output Variable Model 1 Model 2 Model 3 Model 4

Deposits 120.46 9.40 5.64 5.64

Increasing Increasing Increasing Increasing
Loans 83.21 7.86 5.06 5.06
Increasing Increasing Increasing Increasing
Financial Assets 122.49 8.95 5.34 5.40
Increasing Increasing Increasing Increasing
Deposit 49.83 3.20 1.79 1.79
Accounts Increasing Increasing Increasing Increasing
Source: Based on the author’s estimation.
258 Empirical case studies
Table 8.31 Returns to scale for the banking industry of India under alternative output meas-
ures across alternative econometric models

Output Variable Model 1 Model 2 Model 3 Model 4

Deposits 2.70 2.30 2.23 2.30

Increasing Increasing Increasing Increasing
Loans 3.37 2.70 2.95 2.96
Increasing Increasing Increasing Increasing
Financial 2.88 2.43 2.62 2.40
Assets Increasing Increasing Increasing Increasing
Deposit 4.15 3.44 1.27 NA
Accounts Increasing Increasing Increasing
Loan Accounts 2.34 1.95 1.52 NA
Increasing Increasing Increasing
Source: Based on the author’s estimation.

Table 8.32 Returns to scale for the banking industry of Indonesia under alternative output
measures across alternative econometric models

Output Variable Model 1 Model 2 Model 3 Model 4

Deposits 19.37 1.02 0.97 NA

Increasing Constant Decreasing
Loans 5.26 0.29 1.36 1.56
Increasing Decreasing Increasing Increasing
Financial 16.33 0.39 0.88 NA
Assets Increasing Decreasing Decreasing
Deposit −6.78 0.68 0.62 NA
Accounts Negative Decreasing Decreasing
Loan Accounts 20.07 0.76 1.07 NA
Increasing Decreasing Constant
Source: Based on the author’s estimation.

Table 8.33 Returns to scale for the banking industry of China under alternative output
measures across alternative econometric models

Output Variable Model 1 Model 2 Model 3 Model 4

Deposits −11.26 0.12 0.63 −0.29

Negative Decreasing Decreasing Negative
Loans −172.52 0.45 1.17 0.11
Negative Decreasing Increasing Decreasing
Financial −285.65 1.32 1.4 1.92
Assets Negative Increasing Increasing Increasing
Deposit −413.14 2.3 4.55 NA
Accounts Negative Increasing Increasing
Borrowers −123.75 1.7 3.27 3.00
Negative Increasing Increasing Increasing
Source: Based on the author’s estimation.
 Banking: implications of using alternative output definitions 259
Table 8.34 Returns to scale for the banking industry of South Africa under alternative out-
put measures across alternative econometric models

Output Variable Model 1 Model 2 Model 3 Model 4

Deposits 0.15 0.52 0.76 NA

(DEP) Decreasing Decreasing Decreasing
Loans 0.73 0.84 0.8 0.68
(LOAN) Decreasing Decreasing Decreasing Decreasing
Financial 0.56 0.58 0.67 0.25
Assets Decreasing Decreasing Decreasing Decreasing
(ASSET)
Deposit 1.52 0.94 1.87 1.66
Accounts Increasing Decreasing Increasing Increasing
(DACT)
Source: Based on the author’s estimation.

Table 8.35 Returns to scale with the amount of outstanding deposits as the output variable
for the Brazilian banking industry

Output Variable = In(DEP)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −3029.97*** 0.85* 7.92*** 0.26

In(CAP) 1985.97*** 3.49*** 1.75** 1.19
In(LAB)2 176.96*** 0.02
In(CAP)2 3.75 −0.16**
In(LAB)*In(CAP) −157.192*** 0.04 Statistic
RTS −0.92 2.52 9.67 1.45
R-squared 0.95 0.55 0.60 Gamma 0.66
F-statistic 58.46*** 4.01* 12.85*** Log-likelihood ratio 11.77
Notes: 1. Model 1 is the Trans-log production function using standard OLS, Model 2 is the Trans-log
Production Function by Ridge Regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the MLE estimator of the Cobb-Douglas Stochastic Frontier Model; 2. The
Cobb-Douglas functions include the first lag of labour variable and the second lag of the capital variable;
3. For Model 2, the final chosen k-value is 0.0005; 4. Trans-log returns to scale coefficient is estimated
at mean levels of inputs; 5. Values in the bracket are standard errors; 6. *, ** and *** indicate significant at
10%, 5%, and 1% significance levels, respectively
Source: Author’s estimation.

Table 8.36 Returns to scale with the amount of outstanding loans as the output variable for
the Brazilian banking industry

Output Variable = In(LOAN)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −2919.76*** −1.02 1.01 0.79

In(CAP) 1590.76*** 3.33*** 3.41*** 3.51***
In(LAB)2 162.74*** 0.04
In(CAP)2 6.69 -0.06

(Continued)
260 Empirical case studies
Table 8.36 (Continued)

Output Variable = In(LOAN)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB)*In(CAP) −131.48*** 0.10*** Statistic -
RTS 2.30 4.37 4.42 4.3
R-squared 0.98 0.90 0.90 Gamma 0.51
F-statistic 187.06*** 27.32*** 81.27*** Log-likelihood ratio 13.04
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the MLE estimator of the Cobb-Douglas Stochastic Frontier Model; 2. When
the Model 3 is estimated under the intermediation approach by including the amount of outstanding
deposits as a regressor, the R2 turns out to be 0.99. The coefficient of amount of deposits is worked out
to be 0.75 and is found to be significant at 1% level. The F-statistic is 589.60 and is also significant at 1%
level. The size of the labour coefficient is -0.39 which is insignificant even at 10% level. The estimated
coefficient of capital is 2.89 which is significant at 1% level. However, this specification has been excluded
throughout the book to maintain consistent presentation; 3. Similarly, Model 4 was re-estimated under
the intermediation approach, and the MLE coefficient of deposit variables was found to be 0.82 and is
significant at 1% level. Labour coefficient is worked out to be -0.42, and that of Capital is 2.81, and both
are significant at 1 % level. The log-likelihood ratio is 40.16 while the LR test statistic is 3.47, which is
significant at 5% significance level as per critical values given in Kodde and Palm (1986); 4. For Model 2,
the final chosen k-value is 0.0005; 5. Trans-log returns to scale coefficient is estimated at mean levels of
inputs; 6. *, ** and *** indicate significant at 10%, 5% and 1% levels of significance, respectively
Source: Author’s estimation.

Table 8.37 Returns to scale with the amount of financial assets of banks as the output vari-
able for Brazilian banking industry

Output Variable = In(ASSET)

Input Variable Model 1 Model 2 Model 3

In(LAB) −1384.10* −0.11 −2.08***

In(CAP) 1585.13*** 4.72*** 2.75***
In(LAB)2 99.68** −0.21***
In(CAP)2 0.20 −0.09#
In(LAB)*In(CAP) −120.91** −0.003
Estimated Returns to Scale −6.62 −2.81 0.67
R-squared 0.88 0.34 0.48
F-statistic 22.95*** 1.68 8.89***
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log pro-
duction function with ridge regression, Model 3 is the Cobb-Douglas production function with standard
OLS; 2. When the Model 3 is estimated under the intermediation approach by including the Amount of
outstanding deposits as a regressor, the R2 turns out to be 0.77. The coefficient of Amount of Deposits is
worked out to be 0.83 and is found to be significant at 1% level. The F-statistic is 20.201 and is also signifi-
cant at 1% level. The sign of the other coefficients remains the same though. The size of the labour coef-
ficient is -7.14 and that of capital is 2.17. Both are significant at 1% level. However, this specification has
been excluded throughout the book so as to maintain consistence of presentation; 3. Model 4 produced
a very low and an insignificant gamma statistic, and hence have been excluded. It was estimated also re-
estimated under the intermediation approach and the MLE coefficient of deposit variables was found to be
0.93 and is significant at 1% level. Labour coefficient is worked out to be -6.69 and that of Capital is 1.52
and both are significant at 1 % level. The log-likelihood ratio is 23.93 while the LR test statistic is 11.32
which is significant at 1% level of significance as per critical values given in Kodde and Palm (1986). Thus
RTS worked out to be -4.24, thus implying DRS; 4. For Model 2, the final chosen k-value is 0.0005; 5.
Trans-log returns to scale coefficient is estimated at mean levels of inputs; 6. *, ** and *** indicate significant
at 10%, 5% and 1% levels of significance respectively, and # indicates significant at 15% level.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 261
Table 8.38 Returns to scale with the number of deposit accounts as the output variable for
Brazilian banking industry

Output Variable = In(DACT)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −807.46*** 0.44 4.67*** 2.70***

In(CAP) 174.40* 0.59 1.05*** 1.72***
In(LAB)2 30.81** 0.15***
In(CAP)2 −8.69*** −0.04*
In(LAB)*In(CAP) 0.40 0.07*** Statistic -
RTS 5.58 5.78 5.72 4.42
R-squared 0.99 0.95 0.96 Gamma 0.99***
F-statistic 1039.16*** 64.20*** 213.55*** Log-likelihood ratio 37.87
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model. 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs. 3. In Model 2, the k-value is 0.0005; 4. *, ** and ***
indicate significant at 10%, 5% and 1% levels of significance respectively
Source: Author’s estimation.

Table 8.39 Returns to scale with the number of loan accounts as the output variable for
Brazilian banking industry

Output Variable = In(LACT)

Input Variable Model 1 Model 2 Model 3

In(LAB) −6141.69*** −11.85*** 14.79**

In(CAP) −85.89 1.30* 5.35**
In(LAB)2 207.08* 0.46**
In(CAP)2 −39.52### 0.07
In(LAB)*In(CAP) 69.25 0.55***
Estimated Returns to Scale 6.95 15.79 20.14
R-squared 0.97 0.66 0.74
F-statistic 92.32*** 6.27*** 25.18***
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS; 2. Model 3 has been estimated with Cochrane-Orcutt adjustment for serial correlation.
When estimated under the intermediation approach by including the Amount of outstanding deposits
as a regressor, the R2 turns out to be 0.67. The coefficient of Amount of Deposits is worked out to be
0.56 and is found to be not significant even at 10% level. The F-statistic is 12.22 and is also significant
at 1% level. The labour coefficient is -6.92 and that of capital is 9.57. Only the capital coefficient is sig-
nificant at 1% level. Other two coefficients are insignificant event at 10% level. However, this specifica-
tion has been excluded throughout the book so as to maintain consistence of presentation. 3. Model 4,
under the production approach produces a very low and highly insignificant gamma statistic. However,
when it is re-estimated using amount of outstanding deposits as an input variable, its MLE coefficient is
worked out to be 0.47 which is not significant even at 10% level. Labour coefficient is found to be -6.56
and that of Capital is 9.41, and both are significant at 1% level. The log-likelihood ratio is 18.41. This
implies RTS of 3.32 which signifies IRS; 4. Trans-log returns to scale coefficient is estimated at mean
levels of inputs; 5. *, ** and *** indicate significant at 10%, 5% and 1% levels of significance respectively.
6. For Model 2, the chosen k-value is 0.0005.
Source: Author’s estimation.
Table 8.40 Returns to scale with amount of outstanding deposits as the output variable for
the Russian banking industry

Output Variable = In(DEP)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −78.33 2.81*** 7.15*** 6.99***

In(CAP) 97.11 −1.19*** −1.51*** −1.35***
In(LAB)2 14.38 0.21***
In(CAP)2 −0.19 −0.06***
In(LAB)*In(CAP) −12.39 0.12*** Statistic -
RTS 120.46 9.40 5.64 5.64
R-squared 0.99 0.98 0.98 Gamma 0.93***
F-statistic 243.64*** 204.86*** 642.27*** Log-likelihood 15.73
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 3. All the coefficients were found to be statistically insig-
nificant in Model (1) after which It was re-estimated using only the interaction term and excluding the
polynomial terms. Another variant of the same model with cubic terms was tested for. Still, the results
did not improve. Hence, it was excluded; 4. For Model 2, the k-value is 0.002; 5. *, ** and *** indicate
significant at 10%, 5% and 1% levels of significance respectively.
Source: Author’s estimation.

Table 8.41 Returns to scale with amount of outstanding loans as the output variable for the
Russian banking industry

Output Variable = In(LOAN)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −39.16 2.66 ***

5.92 5.89***
***

In(CAP) 82.99### −0.77*** −0.86***

−0.83***
In(LAB)2 11.98* 0.16*** -
In(CAP)2 0.67 −0.03***
In(LAB)*In(CAP) −12.84** 0.08*** Statistic
RTS 83.21 7.86 5.06 5.06
R-squared 0.99 0.99 0.99 Gamma statistic 0.94***
F-statistic 488.61*** 328.41*** 1089.53*** Log-likelihood ratio 24.19
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log pro-
duction function with ridge regression, Model 3 is the Cobb-Douglas production function with standard
OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. When the Model 3 is estimated under
the intermediation approach by including the Amount of outstanding deposits as a regressor, the R2
turns out to be 0.99. The coefficient of Amount of Deposits is worked out to be 0.48 and is found to be
significant at 1% level. The F-statistic is 1633.95 and is also significant at 1% level. The size of the labour
coefficient is 2.47 which is significant at 1% level. The estimated coefficient of capital is -0.13 which is
insignificant even at 10% level. This model suffers from a high degree of multicollinearity as indicated by
the Variance Inflation Factors. Furthermore, the same model is estimated using Ridge Regression with
the k-value of 0.02. The R2 is 0.98 and the sign of the coefficients remains the same. The size of the
coefficient is reduced sizeably, however. These specifications have been excluded here so as to maintain
consistency of presentation; 3. Similarly, Model 4 was re-estimated under the intermediation approach
and the MLE coefficient of deposit variables was found to be 0.50 and is significant at 1% level. Labour
coefficient is worked out to be 2.34 which is significant at 1% level while that of Capital is -0.15 which
is highly insignificant. The log-likelihood ratio is 35.84 while the LR test statistic is 2.52 which is sig-
nificant at 10% level of significance as per critical values given in Kodde and Palm (1986); 4. For Model
2, the final chosen k-value is 0.002; 5. Trans-log returns to scale coefficient is estimated at mean levels of
inputs; 6. *, ** and *** indicate significant at 10%, 5% and 1% levels of significance respectively.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 263
Table 8.42 Returns to scale with amount of financial assets of banks as the output variable
for the Russian banking industry

Output Variable = In(ASSET)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −91.74* 2.23*** 6.35*** 6.34***

In(CAP) 117.07* −0.73** −1.01*** −0.94***
In(LAB)2 12.84### 0.19***
In(CAP)2 −2.36### −0.06***
In(LAB)*In(CAP) −8.96 0.13*** Statistic -
RTS 122.49 8.95 5.34 5.40
R-squared 0.99 0.99 0.99 Gamma 0.83***
F-statistic 383.64*** 247.80*** 771.34*** Log-likelihood 19.41
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. When the Model 3 is esti-
mated under the intermediation approach by including the Amount of outstanding deposits as an input
variable, the R2 turns out to be 0.99. The coefficient of Amount of Deposits is worked out to be 0.68
and is found to be significant at 1% level. The F-statistic is 1619.48 and is also significant at 1% level.
Labour coefficient is 1.52 and that of capital is 0.009. Only the coefficient of Labour is significant at 1%
level in this case. The coefficient of Capital is highly insignificant. However, this specification has been
excluded so as to maintain consistency of presentation; 3. Similarly, Model 4 was re-estimated under
the intermediation approach and the MLE coefficient of deposit variables was found to be 0.70 and
is significant at 1% level. Labour coefficient is worked out to be 1.41 which is significant at 5% level
while that of Capital is 0.04 which is insignificant. The log-likelihood ratio is 36.38 while the LR test
statistic is 6.65 which is significant at 1% level of significance as per critical values given in Kodde and
Palm (1986); 4. For Model 2, the final chosen k-value is 0.002; 5. Trans-log returns to scale coefficient
is estimated at mean levels of inputs. 6. *, ** and *** indicate significant at 10%, 5% and 1% levels of
significance respectively.
Source: Author’s estimation.

Table 8.43 Returns to scale with the number of deposit accounts as the output variable for
the Russian banking industry

Output Variable = In(DACT)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −39.60# 0.66*** 2.13*** 2.08***

In(CAP) 36.74 −0.30** −0.34*** −0.29***
In(LAB)2 5.07 0.07*** - - -
In(CAP)2 −0.58 −0.02***
In(LAB)*In(CAP) −3.26 0.05*** Statistic
RTS 49.83 3.20 1.79 1.79
R-squared 0.98 0.97 0.97 Gamma 0.89***
F-statistic 132.12*** 105.60*** 318.16*** Log-likelihood 34.65
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 3. In Model 2, the k-value is 0.002; 4. ** and *** indicate
significant at 10%, 5% and 1% levels of significance respectively.
Source: Author’s estimation.
264 Empirical case studies
Table 8.44 Returns to scale with amount of outstanding deposits as the output variable for
the banking industry of India

Output Variable = In(DEP)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −95.64# 0.30** 0.80** 0.81***

In(CAP) 320.98*** 0.62** 1.43* 1.49***
In(LAB)2 0.87 0.01** -
In(CAP)2 −20.27** 0.03
In(LAB)*In(CAP) 7.00 0.02*** Statistic -
RTS 2.70 2.30 2.23 2.30
R-squared 0.95 0.84 0.63 Gamma 0.99**
F-statistic 59.41*** 17.04*** 15.43*** Log-likelihood 17.58
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-
log production function with ridge regression, Model 3 is the Cobb-Douglas production function
with standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to
scale coefficient is estimated at mean levels of inputs; 3. The trans-log function was re-estimated using
only the interaction term and excluding the polynomial terms. Another variant of the trans-log func-
tion with cubic terms was tested for. Still, the results did not improve. 4. For Model 2, the k-value is
k = 0.095818.5. Model 3 has been estimated with Cochrane – Orcutt adjustment for auto correlation;
6. *, ** and *** indicate significant at 10%, 5% and 1% levels of significance respectively, and # indicates
significant at 15% level.
Source: Author’s estimation.

Table 8.45 Returns to scale with amount of outstanding loans as the output variable for the
banking industry of India

Output Variable = In(LOAN)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −87.88 0.28** 0.69* 0.66

In(CAP) 398.77*** 0.84*** 2.26*** 2.30**
In(LAB)2 0.58 0.01**
In(CAP)2 −24.05* 0.04***
In(LAB)*In(CAP) 7.03 0.02*** Statistic -
RTS 3.37 2.70 2.95 2.96
R-squared 0.94 0.85 0.85 Gamma 0.99***
F-statistic 53.06*** 18.29*** 54.17*** Log-likelihood 15.41
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. When the Model 3 is esti-
mated under the intermediation approach by including the Amount of outstanding deposits as a regres-
sor, the R2 turns out to be 0.99. The coefficient of Amount of Deposits is worked out to be 1.18 and is
found to be significant at 1% level. The F-statistic is 684.10 and is also significant at 1% level. The size
of the labour coefficient is -0.26 which is significant at 5% level. The estimated coefficient of capital is
0.51 which is significant at 5% level. It been excluded here so as to maintain consistency of presentation;
3. Similarly, Model 4 was re-estimated under the intermediation approach and the MLE coefficient of
deposit variables was found to be 1.14 and is significant at 1% level. Labour coefficient is worked out
to be -0.23 which is significant at 5% level while that of Capital is 0.51 which is significant at 1% level.
The log-likelihood ratio is 41.83 while the LR test statistic is 0.18 which is insignificant as per critical
values given in Kodde and Palm (1986); 4. For Model 2, the chosen k-value is 0.195328; 5. Trans-log
returns to scale coefficient is estimated at mean levels of inputs; 6. *, ** and *** indicate significant at 10%,
5% and 1% levels of significance respectively, and # indicates significant at 15% level.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 265
Table 8.46 Returns to scale with financial assets of commercial banks as the output variable
for the banking industry of India

Output Variable = In(ASSET)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) −148.34** 0.15 0.19 0.53***

In(CAP) 400.58*** 0.81*** 2.43*** 1.87***
In(LAB)2 −2.04 0.006# -
In(CAP)2 −32.91*** 0.04***
In(LAB)*In(CAP) 19.98** 0.02***
RTS 2.88 2.43 2.62 Statistic 2.40
R-squared 0.93 0.81 0.64 Gamma 0.99***
F-statistic 44.78*** 14.09*** 16.23*** Log-likelihood 19.35
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log pro-
duction function with ridge regression, Model 3 is the Cobb-Douglas production function with standard
OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. When the Model 3 is estimated under
the intermediation approach by including the Amount of outstanding deposits as an input variable, the
R2 turns out to be 0.98. The coefficient of Amount of Deposits is worked out to be 0.91 and is found to
be significant at 1% level. The F-statistic is 306.45 and is also significant at 1% level. Labour coefficient is
-0.39 and that of capital is 0.78. Both are significant at 1% level. This specification has been excluded so
as to maintain consistency of presentation. Researchers can access the actual data from the accompanied
e-resources and estimate further extensions and variations of the empirical case studies presented herein;
3. Similarly, Model 4 was re-estimated under the intermediation approach and the MLE coefficient of
deposit variables was found to be 0.90 and is significant at 1% level. Labour coefficient is worked out to
be -0.11 which is insignificant. Output elasticity coefficient of Capital is 0.42 which is significant at 5%
level. The log-likelihood ratio is 42.24 while the LR test statistic is 9.54, which is significant at 1% level of
significance as per critical values given in Kodde and Palm (1986); 4. For Model 2, the final chosen k-value
is 0.131441; 5. Trans-log returns to scale coefficient is estimated at mean levels of inputs; 6. Model 3 has
been estimated with Cochrane – Orcutt adjustment for auto correlation; 7. *, ** and *** indicate significant
at 10%, 5% and 1% levels of significance respectively, and # indicates significant at 15% level.
Source: Author’s estimation.

Table 8.47 Returns to scale with total loan accounts with commercial banks as the output
variable for the banking industry of India

Output Variable = In(LACT)

Input Variable Model 1 Model 2 Model 3

In(LAB) -134.04*** 0.36*** 1.38***

In(CAP) 364.31*** 0.42** 0.14
In(LAB)2 2.94** 0.01***
In(CAP)2 -21.20*** 0.02***
In(LAB)*In(CAP) 5.19 0.02***
RTS 2.34 1.95 1.52

(Continued)
266 Empirical case studies
Table 8.47 (Continued)

Output Variable = In(LACT)

Input Variable Model 1 Model 2 Model 3

R-squared 0.98 0.69 0.48
F-statistic 177.50*** 7.15** 8.36***
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. When the Model 3 is esti-
mated under the intermediation approach by including the Amount of outstanding deposits as an input
variable, the R2 turns out to be 0.83. The coefficient of Amount of Deposits is worked out to be 0.81
and is found to be significant at 1% level. The F-statistic is 27.66 and is also significant at 1% level.
Labour coefficient is 0.50 and that of capital is -1.07. Both are significant at 5% level. This implies RTS
of 0.25, i.e. DRS. This specification has been excluded so as to maintain consistency of presentation; 3.
Model 4 produced very poor results and the stochastic frontier model did not seem to be suitable. It was
also re-estimated under the intermediation approach and the MLE coefficient of deposit variables was
found to be 0.98 which is insignificant even at 10% level. Labour coefficient is worked out to be 0.85
while that of Capital is -1.27, both of which are insignificant at 10% level. The log-likelihood ratio was
found to be 23.45; 4. For Model 2, the final chosen k-value is 0.916500; 5. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 6. *, ** and *** indicate significant at 10%, 5% and 1%
levels of significance respectively, and # indicates significant at 15% level.
Source: Author’s estimation.

Table 8.48 Returns to scale with total deposit accounts with commercial banks as the output
variable for the banking industry of India

Output Variable = In(DACT)

Input Variable Model 1 Model 2 Model 3

In(LAB) -245.16*** 0.58*** 0.70**

In(CAP) 316.77*** 0.76*** 0.57
In(LAB)2 0.33 0.02***
In(CAP)2 -30.96*** 0.04***
In(LAB)*In(CAP) 23.13*** 0.03***
RTS 4.15 3.44 1.27
R-squared 0.98 0.83 0.27
F-statistic 213.48*** 16.16*** 3.38*
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 3. In Model 2, the k-value is 0.606159; 4. Model 3 has
been estimated using Cochrane-Orcutt adjustments for autocorrelation. An ARMA (1, 1) model was
also estimated for the same model and in both the cases, the results do not improve; 5. Model 4 yielded
did not produce good results. The gamma statistic was very low and insignificant and the resultant MLE
coefficients were very small and insignificant even at 10% level. Hence, it has not been presented; 6. *,
**
and *** indicate significant at 10%, 5% and 1% levels of significance respectively.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 267
Table 8.49 Returns to scale with amount of outstanding deposits commercial banks as the
output variable for the banking industry of Indonesia

Output Variable = In(DEP)

Input Variable Model 1 Model 2 Model 3

In(WAGE) -206.46*** -0.25## 0.67***

In(CAP) 67.90** 0.22*** 0.30***
In(WAGE)2 9.17*** -0.007
In(CAP)2 1.21* 0.11**
In(WAGE)*In(CAP) -6.25** 0.008***
RTS 19.37 1.02 0.97
R-squared 0.93 0.57 0.97
F-statistic 44.02*** 4.21** 232.88***
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 3. For Model 2, the k-value is 0.118012; 4. In Model
3, the labour variable has been taken at its second lag; 5. *, ** and *** indicate significant at 10%, 5% and
1% levels of significance, respectively, and ## implies significant at 13% level; 6. Values in the bracket
are standard errors. 7. Model 4 produced poor estimates. The gamma statistic was very small and highly
insignificant, while the coefficients were statistically insignificant even at the 10% level.
Source: Author’s estimation.

Table 8.50 Returns to scale with the amount of outstanding loans from commercial banks
as the output variable for the banking industry of Indonesia

Output Variable = In(LOAN)

Input Variable Model 1 Model 2 Model 3 Model 4

In(WAGE) 7.78** -1.16** 0.86*** 0.49***

In(CAP) 0.24* 0.17 0.50*** 0.15***
In(WAGE)2 -0.69*** 0.02* †
0.92***
In(CAP)2 -1.26*** 0.01 -
In(WAGE)*In(CAP) 1.53*** 0.03*** Statistic -
RTS 5.26 0.29 1.36 1.56
R-squared 0.99 0.85 0.98 Gamma 0.91***
F-statistic 250.89*** 17.71*** 333.51*** Log-likelihood 51.69
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 2. In(LAB) is with second lag in Model 3. When the
same model is estimated under the intermediation approach by including the Amount of outstanding
deposits as a regressor, autocorrelation appears to be a serious problem. Accordingly, the Cochrane-
Orcutt Serial Correlation adjustment procedure is used to tackle this problem. The resulting model
provides an R2 of 0.98. The coefficient of Amount of Deposits is worked out to be 0.80 and is found
to be significant at 1% level. The F-statistic is 235.51 and is also significant at 1% level. The size of
the labour coefficient is 0.73, which is significant at 1% level. The estimated coefficient of capital is
0.14 which is significant at 15% level. It been excluded here to maintain consistency of presentation.
Researchers can access the data from the accompanied e-resources and estimate more extensions and
variations of the empirical case studies presented herein; 3. Model 1 has been estimated with Cochrane-
Orcutt adjustment for autocorrelation, and In(CAP) is with first lag; 4. For Model 2, the final chosen
k-value is 0.002; 5. † – Model 4 is reported with intermediation approach results, and the third variable
is In(DEP); 6. Trans-log returns to scale coefficient is estimated at mean levels of inputs; 7. *, ** and ***
indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.
268 Empirical case studies
Table 8.51 Returns to scale with amount of financial assets of commercial banks as the out-
put variable for the banking industry of Indonesia

Output Variable = In(ASSET)

Input Variable Model 1 Model 2 Model 3

In(WAGE) -167.06*** -0.15 0.46**

In(CAP) 58.47** 0.18** 0.42***
In(WAGE)2 7.72** 0.002**
In(CAP)2 1.43* 0.01
In(WAGE)*In(CAP) -5.88** 0.01***
RTS 16.33 0.39 0.88
R-squared 0.93 0.83 0.93
F-statistic 45.10*** 15.52*** 46.98***
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 2. When the Model 3 is estimated under the interme-
diation approach by including the Amount of outstanding deposits as an input variable, the R2 turns
out to be 0.98. The coefficient of the amount of deposits is worked out to be 0.70 and is found to be
significant at 1% level. The F-statistic is 245.71 and is also significant at 1% level. Labour coefficient is
0.53 and that of capital is 0.05. Labour coefficient is significant at 1% level. However, the output elastic-
ity coefficient of capital is highly insignificant. However, when the first lag of the capital variable is used,
its coefficient turns out to be 0.13 and becomes significant at 5% level. The signs of other two coef-
ficients do not change. This specification has been excluded to maintain consistency of presentation; 3.
For Model 2, the final chosen k-value is 0.013861; 4. Trans-log returns to scale coefficient is estimated
at mean levels of inputs; 5. *, ** and *** indicate significant at 10%, 5% and 1% levels of significance,
respectively; 6. In(WAGE) in Model 3 is in first lag.
Source: Author’s estimation.

Table 8.52 Returns to scale with the total number of loan accounts with commercial banks
as the output variable for the banking industry of Indonesia

Output Variable = In(LACT)

Input Variable Model 1 Model 2 Model 3

In(WAGE) -99.01*** 0.27*** 0.90***

In(CAP) 55.20*** 0.09*** 0.17**
In(WAGE)2 5.09*** 0.009*** -
In(CAP)2 0.89** 0.004*** -
In(WAGE)*In(CAP) -4.79*** 0.005*** -
RTS 20.07 0.76 1.07
R-squared 0.98 0.83 0.95
F-statistic 202.62*** 15.46*** 197.41***
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log pro-
duction function with ridge regression, Model 3 is the Cobb-Douglas production function with standard
OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale coefficient is
estimated at mean levels of inputs; 2. When the Model 3 is estimated under the intermediation approach
by including the Amount of outstanding deposits as an input variable, the coefficient of Amount of Depos-
its is worked out to be 0.11 and is found to be insignificant even at 10% level. The labour coefficient is
1.02 in this case and is significant at 1% level. However, this model suffers from autocorrelation problem
and even after undertaking the Cochrane-Orcutt adjustments, the coefficients behaved similarly. 2. For
Model 2, the final chosen k-value is 0.651838; 3. Trans-log returns to scale coefficient is estimated at mean
levels of inputs; 4. *, ** and *** indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 269
Table 8.53 Returns to scale with the total number of deposit accounts with commercial
banks as the output variable for the banking industry of Indonesia

Output Variable = In(DACT)

Input Variable Model 1 Model 2 Model 3

In(WAGE) -71.59 0.12 -0.12

In(CAP) -34.16 0.17*** 0.74**
In(WAGE)2 1.29 0.005
In(CAP)2 -0.90 0.009***
In(WAGE)*In(CAP) 3.57 0.008***
RTS -6.78 0.68 0.62
R-squared 0.97 0.51 0.58
F-statistic 93.64*** 3.37** 12.91***
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 3. In Model 2, the k-value is 0.423110; 4. Model 1
has also been estimated without the intercept term. With the intercept term, all the coefficients are
statistically insignificant as shown in the table at even 10% level. By excluding the constant term, the
coefficients become significant, but the signs of these coefficients change. Hence, it was thus decided to
exclude that specification, because it deviates the sign and even the size of the coefficients considerably;
5. *, ** and *** indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.

Table 8.54 Returns to scale with the amount of outstanding deposits with commercial banks
as the output variable for the banking industry of China

Output Variable = In(DEP)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) -119.62** 1.17*** 2.78*** 2.97***

In(CAP) -129.22 -1.98*** -2.15** -3.26***

In2(LAB) 7.66** 0.04***
In2 (CAP) 7.03 -0.09***
In(LAB)*In(CAP) -2.40 0.04*** - Statistic -
RTS -11.26 0.12 0.63 -0.29
R-squared 0.87 0.75 0.82 Gamma 0.99***
F-statistic 20.99*** 9.72*** 41.42*** Log-likelihood 1.33
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 3. For Model 2, the k-value is 0.049336; 4. Models
1 and 3 have been estimated with Cochrane-Orcutt adjustments for autocorrelation problem. The
unadjusted standard OLS model showed strong evidence of autocorrelation, and hence the same was
appropriately treated for. The signs of the coefficients are the same in both the cases for both models.
Hence the Model 2 has also been estimated; 5. *, ** and *** indicate significant at 10%, 5% and 1% levels
of significance, respectively.
Source: Author’s estimation.
270 Empirical case studies
Table 8.55 Returns to scale with the amount of outstanding loans with commercial banks as
the output variable for the banking industry of China

Output Variable = In(LOAN)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) -223.23*** 0.99*** 2.90*** 3.04***

In(CAP) -266.85 -1.93*** -1.73** -2.93***
In(LAB)2 9.12*** 0.05***
In(CAP)2 14.23 -0.09***
In(LAB)*In(CAP) -4.21 0.05*** - Statistic -
RTS -172.52 0.45 1.17 0.11
R-squared 0.89 0.78 0.86 Gamma 0.99***
F-statistic 26.79*** 11.43*** 55.42*** Log-likelihood 1.85
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 2. When the Model 3 is estimated under the interme-
diation approach by including the Amount of outstanding deposits as a regressor, the resulting model
provides an R2 of 0.99. The coefficient of Amount of Deposits is worked out to be 1.02 and is found
to be significant at 1% level. The F-statistic is 1500.36 and is also significant at 1% level. The size of the
labour coefficient is 0.12 which is insignificant even at 10% level. The estimated coefficient of capital
is 0.33 which is significant at 10% level. It been excluded here to maintain consistency of presentation.
Researchers can access the actual data from the accompanied e-resources and estimate more exten-
sions and variations of the empirical case studies presented herein. However, when the fourth model is
estimated in the same context, the results are not good and the gamma statistic is highly insignificant
along with insignificant coefficients; 3. For Model (2), the final chosen k-value is 0.007755; 4. Trans-
log returns to scale coefficient is estimated at mean levels of inputs; 5. Model 3 has been estimated with
Cochrane-Orcutt adjustments for serial correlation. In either case, the signs of the coefficients do not
change and sizes are roughly similar; 6. *, ** and *** indicate significant at 10%, 5% and 1% levels of
significance, respectively.
Source: Author’s estimation.

Table 8.56 Returns to scale with the amount of financial assets with commercial banks as
the output variable for the banking industry of China

Output Variable = In(ASSET)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) -17.42 1.07*** 2.65*** 2.46***

In(CAP) -523.99 -1.11** -1.25* -0.54
In(LAB)2 0.03 0.03***
In(CAP)2 21.73 -0.04**
In(LAB)*In(CAP) 1.67 0.05*** - Statistic -
RTS -285.65 1.32 1.4 1.92
 Banking: implications of using alternative output definitions 271

Output Variable = In(ASSET)

Input Variable Model 1 Model 2 Model 3 Model 4

R-squared 0.86 0.84 0.84 Gamma 0.90***
F-statistic 19.33 16.63*** 50.32*** Log-likelihood 5.95
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 3. When the Model 3 is estimated under the interme-
diation approach by including the Amount of outstanding deposits as an input variable, the R2 turns
out to be 0.87. The coefficient of the amount of deposits is worked out to be 0.33 and is found to be
significant at 5% level. The F-statistic is 42.28 and is significant at 1% level. Labour coefficient is 1.63
and that of capital is -0.09. Labour coefficient is significant at 1% level while the coefficient of capital
is highly insignificant. If the first lag of the capital variable is used, the results deteriorate considerably.
This specification has been excluded to maintain the consistency of the presentation. Similar exercise for
Model 4 results into poor estimates. The gamma statistic is very small and highly insignificant. Labour
and Deposit coefficients are found to be positive and significant while capital coefficient is negative and
insignificant; 4. For Model 2, the final chosen k-value is 0.006076; 5. The significance level of LR-
test statistic has been undertaken using mixed-chi square distribution values given in Kodde and Palm
(1986); 6. Trans-log returns to scale coefficient is estimated at mean levels of inputs; 7. Models 1 and 2
have been estimated with the first lag of the labour variable; 8. *, ** and *** indicate significant at 10%,
5% and 1% levels of significance, respectively, and ## indicates significant at 12% level.
Source: Author’s estimation.

Table 8.57 Returns to scale with the total number of deposit accounts with commercial
banks as the output variable for the banking industry in China

Output Variable = In(DACT)

Input Variable Model 1 Model 2 Model 3

In(LAB) -287.31*** 1.19*** 3.28***

In(CAP) -672.62* -0.82 1.27*
In(LAB)2 14.05*** 0.04***
In(CAP)2 36.81** -0.03
In(LAB)*In(CAP) -11.54# 0.06*** -
RTS -413.14 2.30 4.55
R-squared 0.94 0.69 0.91
F-statistic 55.62*** 7.16*** 78.99***
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. When the Model 3 is esti-
mated under the intermediation approach by including the Amount of outstanding deposits as an input
variable, the coefficient of the amount of deposits is worked out to be 0.11 and is found to be insignifi-
cant even at 10% level. The labour coefficient is 1.02 in this case and is significant at 1% level. However,
this model suffers from the autocorrelation problem, and even after undertaking the Cochrane-Orcutt
adjustments, the coefficients behaved similarly; 3. For Model 2, the final chosen k-value is 0.249917; 4.
Trans-log returns to scale coefficient is estimated at mean levels of inputs; 5. Model 3 has been estimated
using the Cochrane-Orcutt adjustments for dealing with the problem of autocorrelation in the non-
adjusted model. The capital variable is in its second lag; 6. *, ** and *** indicate significant at 10%, 5%
and 1% levels of significance, respectively, and # indicates significant at 15% level.
Source: Author’s estimation.
272 Empirical case studies
Table 8.58 Returns to scale with the total number of borrowers from commercial banks as
the output variable for the banking industry in China

Output Variable = In(BOR)

Input Variable Model 1 Model 2 Model 3 Model 4

In(LAB) -85.25*** -0.07 0.40** 0.002

In(CAP) -210.29** 1.16*** 2.87*** 2.99***
In(LAB)2 3.21*** -0.002
In(CAP)2 9.98*** 0.05***
In(LAB)*In(CAP) -1.01 0.008** - Statistic -
RTS -123.75 1.70 3.27 3.00
R-squared 0.98 0.67 0.84 Gamma 0.99***
F-statistic 133.32*** 6.41*** 42.36*** Log-likelihood 19.11
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale
coefficient is estimated at mean levels of inputs; 3. In Model 2, the k-value is 0.603455; 4. Model 1
has also been estimated without the intercept term. With the intercept term, all the coefficients are
statistically insignificant, as shown in the table at even 10% level. By excluding the constant term, the
coefficients become significant, but the signs of these coefficients change. Hence, it was decided to
exclude the specification because it deviated from the theoretical expectations on the expected sign and
the size. Similarly, when the model fourth is estimated, the results are insignificant from an economic
point of view, though the gamma statistic is high and significant; 5. Model 3 has been estimated with
the second lag of capital variable. The model has been estimated with Cochrane-Orcutt serial correla-
tion adjustment; 6. *, ** and *** indicate significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.

Table 8.59 Returns to scale with the amount of outstanding deposits of commercial banks
as the output variable for the banking industry of South Africa

Output Variable = In(DEP)

Input Variable Model 1 Model 2 Model 3

In(WAGE) 13.61 -0.04 0.43***

In(CAP) 12.98* 0.13*** 0.33**
In(WAGE)2 -0.24 -0.002
In(CAP)2 -0.13 0.008***
In(WAGE)*In(CAP) -1.17** 0.02** -
RTS 0.15 0.52 0.76
R-squared 0.83 0.60 0.85
F-statistic 15.50*** 4.80*** 45.53***
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log pro-
duction function with ridge regression, Model 3 is the Cobb-Douglas production function with standard
OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. Trans-log returns to scale coefficient is
estimated at mean levels of inputs; 3. For Model 2, the k-value is 0.233219; 4. Models 3 has been estimated
with Cochrane-Orcutt adjustments for the autocorrelation problem. The unadjusted standard OLS model
showed strong evidence of autocorrelation, and hence the same was appropriately treated for. The sign of
the coefficients is the same in both the cases; 5. For model 3rd, the labour variable has been specified at
its 3rd lag. Even with its current value and for lags up to 3rd, the sign remains the same. The size of the
labour coefficient increases slightly, at the cost of the size of the capital coefficient; 6. *, ** and *** indicate
significant at 10%, 5% and 1% levels of significance, respectively, and # indicates significant at 15% level.
Source: Author’s estimation.
 Banking: implications of using alternative output definitions 273
Table 8.60 Returns to scale with the amount of outstanding loans from commercial banks
as the output variable for the banking industry of South Africa

Output Variable = In(LOAN)

Input Variable Model 1 Model 2 Model 3 Model 4

In(WAGE) 21.25* 0.04† 0.31*** 0.29*

In(CAP) 9.99## 0.12*** 0.49*** 0.39***
In(WAGE)2 -0.67## 0.003†
In(CAP)2 0.03 0.007***
In(WAGE)*In(CAP) -1.10** 0.03*** - Statistic -
RTS 0.73 0.84 0.80 0.68
R-squared 0.77 0.66 0.69 Gamma 0.86***
F-statistic 10.52*** 6.26*** 21.03*** Log-likelihood 22.40
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. When the Model 3 is esti-
mated under the intermediation approach by including the amount of outstanding deposits as a regres-
sor, its coefficient is worked out to be 0.62 and is found to be significant at 1% level. The rest of the
estimated coefficients are positive and highly significant. Similarly, when Model 4 is estimated similarly,
the coefficients are insignificant, and the overall model seems poor due to a very low and insignificant
gamma statistic. The individual coefficients are, however significant at 1% level; 3. For Model 2, the final
chosen k-value is 0.059198; 4. Trans-log returns to scale coefficient is estimated at mean levels of input;
5. Model 3 has been estimated with Cochrane-Orcutt adjustments for serial correlation. In either case,
the signs of the coefficients do not change, and sizes are roughly similar; 6. Models 1 and 2 have been
estimated with the first lag of the labour variable. In the case of Model 1, this allowed the suppression
of the degree of multicollinearity to a great extent, though it remained high; 7. *, **, and *** indicate
significant at 10%, 5% and 1% levels of significance, respectively, and †indicates significant at 16% level.
Source: Author’s estimation.

Table 8.61 Returns to scale with the amount of financial assets of commercial banks as the
output variable for the banking industry of South Africa

Output Variable = In(ASSET)

Input Variable Model 1 Model 2 Model 3 Model 4

In(WAGE) 21.36†† 0.03 0.22** 0.02

In(CAP) 7.20 0.15*** 0.45*** 0.23*
In(WAGE)2 -0.66 0.004
In(CAP)2 0.24 0.01***
In(WAGE)*In(CAP) -1.16* 0.01 - Statistic -
RTS 0.56 0.58 0.67 0.25
R-squared 0.72 0.61 0.62 Gamma 0.99***
F-statistic 8.23*** 4.93*** 15.37*** Log-likelihood 22.72
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log pro-
duction function with ridge regression, Model 3 is the Cobb-Douglas production function with standard
OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. When the Model 3 is estimated under the
intermediation approach by including the Amount of outstanding deposits as an input variable, the R2 turns
out to be 0.86. The coefficient of Amount of Deposits is worked out to be 0.69 and is found to be significant
at 1% level. The F-statistic is 38.57 and is significant at 1% level. Labour coefficient is 0.16, and that of capital
is 0.12. Labour coefficient is significant at 5% level while the coefficient of capital is significant at 15% level.
Similarly, the fourth model is also estimated in the same context, and the model seems to be alright. The
gamma statistic turns out to be 0.83 and is significant at 1% level. For Model 2, the final chosen k-value is
0.025251; 3. Trans-log returns to scale coefficient is estimated at mean levels of inputs; 4. *, ** and *** indicate
significant at 10%, 5% and 1% levels of significance, respectively, and †† indicates significant at 13% level.
Source: Author’s estimation.
274 Empirical case studies
Table 8.62 Returns to scale with the number of deposit accounts of commercial banks as the
output variable for the banking industry of South Africa

Output Variable = In(DACT)

Model 3 Model 1 Model 2 Model 3 Model 4

In(WAGE) 13.05 -0.30*** 0.39 0.71***

In(CAP) 1.28 0.44*** 1.48*** 0.95***
In(WAGE)2 -0.16 -0.02***
In(CAP)2 0.76 0.03***
In(WAGE)*In(CAP) -1.29 0.04** - Statistic -
RTS 1.52 0.94 1.87 1.66
R-squared 0.95 0.82 0.91 Gamma 0.99***
F-statistic 59.33 14.20*** 98.25*** Log-likelihood 14.27
Notes: 1. Model 1 is the Trans-log production function with standard OLS, Model 2 is the Trans-log
production function with ridge regression, Model 3 is the Cobb-Douglas production function with
standard OLS, Model 4 is the stochastic frontier Cobb-Douglas model; 2. When the Model 3 is esti-
mated under the intermediation approach by including the Amount of outstanding deposits as an input
variable, the coefficient of amount of deposits is worked out to be 0.11 and is found to be insignificant
even at 10% level. The labour coefficient is 1.02 in this case and is significant at 1% level. However,
this model suffers from the autocorrelation problem, and even after undertaking the Cochrane-Orcutt
adjustments, the coefficients behaved similarly; 3. For Model 2, the final chosen k-value is 0.387843;
4. Trans-log returns to scale coefficient is estimated at mean levels of inputs; 5. Model 3 has been esti-
mated using the Cochrane-Orcutt adjustments for dealing with the problem of autocorrelation in the
estimated residuals of the base model. The capital variable is in its second lag; 6. *, ** and *** indicate
significant at 10%, 5% and 1% levels of significance, respectively.
Source: Author’s estimation.
Appendix 4
Estimates for section 8.5

Table 8.63 Average change in the Malmquist Index of total factor productivity for the panel
of selected economies across the sample period

Country DMTFP LMTFP DAMTFP LAMTFP AMTFP

Brazil 0.3 2.4 2.6   3 2.6

Russia 12.3 9.4 3.7 NA 4.9
India 0.6 1.0 5.1 2.8 3.0
Indonesia -6.7 -4.8 -3.1 -1.8 -2.5
China 6.0 6.6 -0.2 -0.2 6.1
South Africa -3.1 -3.8    3 -3.3 -1.5
Notes: 1. All the variables represent the Malmquist Total Factor Productivity Index as estimated using
the DEA approach; 2. In terms of the underlying output variables, DMTFP represents the amount of
outstanding deposits with the commercial banks, LMTFP represents the amount of outstanding loans
with the commercial banks, DAMTFP represents the number of total deposit accounts with the com-
mercial banks, LAMTFP represents the number of total loan accounts with the commercial banks and
AMTFP represents the amount of financial assets owned by the commercial banks.
Source: Author’s estimation.
Appendix 538
Input variables and estimates of
output elasticities of inputs used in the
construction of the total output index
for section 8.3

Table 8.64 Cobb-Douglas production function estimates used in the estimation of weights

for total inputs index for residual TFP index for Brazil and Russia

Brazil Russia

Output Elasticities for Inputs Aggregation Output Elasticities for Inputs Aggregation

All variables are in millions of their respective All variables are in millions of their respective
Units Units
Inputs Output 1: Log Inputs Output 1: Log
of outstanding of outstanding
deposits deposits
Real Mean Wage 0.22 Bank employees 6.21
Rate
Bank Branches -1.53 Commercial Bank -0.97
in Millions Branches
Total Equity in 0.81 Number of total 0.15
Millions electronic terminals
(self-service, POS,
etc.)
Inputs Output 2: Log of Inputs Output 2: Log
Deposit Accounts of Outstanding
with Commercial Loans (in
Banks millions)
Real Mean Wage 0.08 Bank employees 6.02
Rate
Bank Branches 1.20 Commercial Bank -0.80
in Millions Branches
Total Equity in 0.43 Inputs Output 3: Total
Millions Bank Assets
Inputs Output 3: Log of Bank employees 0.97
Outstanding loans
with Commercial
Banks
Real Mean Wage 0.16 Commercial Bank 0.20
Rate Branches
Bank Branches 2.67 Outstanding Deposits 0.75
in Millions
 Banking: implications of using alternative output definitions 277

Brazil Russia
Outstanding 0.73 Inputs Output 4:
deposits Total Bank
Accounts
Inputs Output 4: Log of Bank employees 2.13
Loan Accounts
(in millions)
Real Mean Wage 2.45 Commercial Bank -0.34
Rate Branches
Bank Branches 3.69
in Millions
Number of 0.99
depositors
Inputs Output 5: Log of
Total Assets (in
millions of LCU)
Real Mean Wage -0.33
Rate
Bank Branches 4.36
in Millions
Total Equity in -0.58
Millions
Amount of 1.30
Deposits
per deposit
account

Table 8.65 Cobb-Douglas production function estimates used in the estimation of weights

for total inputs index for residual TFP index for India and Indonesia

India Indonesia

Output Elasticities for Inputs Aggregation Output Elasticities for Inputs Aggregation

All variables are in millions of their respective All variables are in millions of their respective
Units Units
Inputs Output 1: Log Inputs Output 1: Log
of outstanding of outstanding
deposits deposits
Bank employees 0.80 Mean Real Wages 0.59
Commercial 1.49 Branches of 0.51
Bank Branches commercial
banks
Inputs Output 2: Log Inputs Output 2: Log
of Outstanding of Outstanding
Loans Loans
Bank employees -0.26 Mean Real Wages -0.31
Commercial 0.51 Branches of 0.38
Bank Branches commercial
banks

(Continued)
278 Empirical case studies
Table 8.65 (Continued)

India Indonesia
Amount of 1.18 Amount of 0.84
outstanding outstanding
deposits deposits
Inputs Output 3: Log of Inputs Output 3: Total
Total Bank Assets Bank assets
Bank employees -0.39 Mean Real Wages -0.28
Commercial 0.78 Branches of 0.27
Bank Branches commercial banks
Amount of 0.91 Amount of 0.61
outstanding outstanding
deposits deposits
Inputs Output 4: Log of Inputs Output 4: Log
Deposit Accounts of Deposit
accounts
Bank employees 2.15 Mean Real Wages 1.16
Commercial 1.19 Branches of 0.85
Bank Branches commercial banks
Inputs Output 5: Log of
Loan accounts
Mean Real Wages -0.43
Branches of 0.44
commercial
banks
Deposits per 0.18
borrower

Table 8.66 Cobb-Douglas production function estimates used in estimation of weights for

total inputs index for residual TFP index for China and South Africa

China South Africa

Output Elasticities for Inputs Output Elasticities for Inputs

Aggregation Aggregation
All variables are in millions of their All variables are in millions of their respective
respective Units Units
Inputs Output 1: Log Inputs Output 1: Log
of outstanding of outstanding
deposits deposits
Bank employees 3.55 Employees in Finance 0.55
and Insurance
Commercial -5.30 Branches of 0.33
Bank Commercial Banks
Branches (2nd lag)
Inputs Output 2: Log Inputs Output 2: Log
of Outstanding of Outstanding
Loans Loans
Bank employees 3.77 Employees in Finance -0.37
and Insurance
 Banking: implications of using alternative output definitions 279

China South Africa

Commercial -5.16 Branches of 0.07
Bank Commercial Banks
Branches
Inputs Output 3: Log of Amount of deposits 0.73
Bank Assets
Bank employees 2.65 Inputs Output 3: Log
of Deposit
accounts
Commercial -1.26 Employees in Finance 1.37
Bank and Insurance
Branches
Inputs Output 4: Log Branches of 0.90
of Deposit Commercial Banks
Accounts
Bank employees 2.66
Commercial -3.53
Bank Branches
(2nd lag)
Inputs Output 5: Log of
Borrowers from
commercial
banks
Bank employees 0.94
Commercial 2.80
Bank Branches
(2nd lag)
Amount of -0.21
deposits
Inputs Output 6: Log of
Depositors with
banks
Bank employees 3.29
Commercial -1.78
Bank
Branches

Notes
1 For an overview and application of the aggregate industry-wide approach to technical
efficiency adopted here, please see Méon and Weill (2010).
2 It is duly recognized at this stage that aggregation generates some insurmountable prob-
lems. First, it gives rise to the aggregation problem as initiated by the Cambridge Capital
controversy. Second, there are very strict conditions to be fulfilled to prove that the
estimated coefficients represent technological relationships rather than mere accounting
identities. Both these matters are examined in detail in Felipe and McCombie (2014).
However, these criticisms are generally levied against the use of aggregate production
functions for deriving microeconomic inferences. No such attempt is made in the Part
III of this study. First, data constraints have forced the use of this approach. The ideal
approach is always to prefer firm-level information rather than industry-level data. If
high-quality, consistent and reliable firm-level data are unavailable, one then needs to
adopt a more accommodative approach. Second, the primary aim of Part III is to show
how empirical estimates of performance parameters change when alternative output
280 Empirical case studies
measures are used. Suppose the evidence on aggregated data shows meaningful impact
of changing output measures on selected parameters. In that case, it does provide a pre-
liminary case for such an impact at a more disaggregated level of data. Macroeconomic
literature has evolved considerably since the early critiques on aggregation. If aggrega-
tion were an economically non-useful approach, then macroeconomic theory’s entire
edifice, including national income accounting, could crumble down. As argued in Col-
ander (1993), economists have been equally vociferous on the existence of a separate
and independent macroeconomic reality, which is an outcome of interactions among
individual actions but may evolve independently of those individual actors. The present
work has adopted this rather optimistic view toward aggregation. Due to data con-
straints and providing a fresher perspective, the study has adopted the use of aggregate
industry-level data. A final reason for this choice was that the author wanted to use data
from the same source as far as possible so that comparability of results across economies
could be maintained. Thus, given the above qualifications, the analysis in Chapters 8
and 9 must be interpreted as brief case studies rather than rigorous economic analysis of
the chosen performance issues, which is far beyond the scope of this work.
3 Boda and Piklova (2021) find higher variability in efficiency scores based on the produc-
tion approach mainly caused when the number of labourer variables is used along with
the total asset variable. In the present study, similar findings are obtained concerning the
variation in output measures only rather than both outputs and inputs. This has been
done by keeping the input vector constant across all specifications.
4 Lawrence Klein’s works are generally referred as the seminal works on this issue in
applied macroeconometrics. Since being in the Cowles commission, he aimed to
link the macroeconomic issues with the microeconomic, optimizing logic of eco-
nomic agents into “properly measured” macro aggregates (Visco, 2014). To date,
the debate on this issue has continued. One school of thought asserts that macro
aggregates are meaningless without microeconomic foundations. The other states
that macroeconomic reality can be theorized independently of pure microeconomic
behaviour within minimalistic micro-foundations. “The new macro perspective
allowed one to talk about interrelationships among aggregates without specifying the
underlying individual choice theoretic framework” (Colander, 1993, p. 447). See the
author’s work for more detail on these two schools of thought in macroeconomic
theorizing.
5 A frequently used data source in the literature is BankFocus (earlier BankScope) by
Moody’s Analytics. While this is a useful source for firm-level information in banking
and insurance, criticisms have been levied on its unqualified use. The same has been
highlighted earlier in Chapter 2. Furthermore, the author could not fully access the
source to verify the complete availability of firm-level data and its quality. Hence, it
seemed best to use data from national and international official agencies.
6 There are multiple definitions of EMEs in the literature. The exact definition used here
is based on the multi-dimensional procedure and accounts for both the economic and
financial characteristics of the economies. The online resources may provide the details
on the procedure used to define and then select the emerging economies. The same can
also be obtained from the author upon request.
7 The Cobb-Douglas functional form possesses several strengths. The “Cobb-Douglas
production function is robust to duality” (Kouki et al., 2014, p. 357). Furthermore, it
has been suggested in works on macroeconometrics by Klein and others that Cobb-
Douglas is ideal for aggregate economic analysis.
8 This functional form is theorized as a second-order Taylor series approximation to an
unknown underlying true production function. For more details on its properties and
issues in constructing and using it in production analysis, see Boisvert (1982).
9 The Cobb-Douglas production function has a rich history. It has transcended many bar-
riers and has passed through many stages of conceptual metamorphosis. The evolution
of this function can be located in Felipe and McCombie (2014) and, more recently, in
Biddle (2021).
 Banking: implications of using alternative output definitions 281
10 Cobb-Douglas production function was the primary choice. The trans-log function was
adopted to provide additional insights.
11 See the seminal work of Solow (1957) for more details.
12 In sections 8.3 and 9.3, the chosen DEA model is output-oriented under Variable
Returns to Scale (VRS) on the lines of Fare et al. (1994, Chapters 4 and 6). The DEAP
Version 2.1 software was used.
13 All econometric estimations reported in this and the next chapter were tested for econo-
metric diagnostics, and due care was taken to ensure that the key econometric issues
of normality of estimated residuals, autocorrelation, multicollinearity and specification
error tests were handled appropriately. The estimated results are robust to alternative
specifications in terms of the variables. Sensitivity tests to alternative specifications were
undertaken to ensure that the estimated coefficients and results are reliable on the lines
suggested by Magnus and Vasnev (2015). Due to space restrictions, the same has not
been reported but can be happily obtained from the author on request.
14 There are some works, such as Fecher (1993), that have used a Cobb-Douglas based SFA
for estimating industry-level production functions. Similarly, Kumbhakar and Hjalmars-
son (1995) use a trans-log SFA with only interactive terms.
15 There are two key choices to be made when using the SFA: first is the functional
form, and the second is the distributional assumption for the inefficiency term (Eling &
Luhnen, 2010). In sections 8.3, 8.4, 9.3, and 9.4, the Battese and Coelli (1992) error
components model under SFA is employed using the FRONTIER 4.1 software. More
details may be availed from the author on request.
16 The detailed definitions and explanations of each of the variables used, the procedures
employed for dealing with missing data, the data source for each variable and other
important methodological issues can be obtained from the author on request. This
choice was made due to space restrictions.
17 The detailed definitions and explanations on each of the variables used, the procedures
employed for dealing with missing data, data source for each variable and other impor-
tant methodological issues can be obtained from the author on request. This choice was
made due to space restrictions.
18 The choice of these variables is largely based on the production approach; though, in
section 8.4, the intermediation approach is also tested. The production approach has
been used as it lends itself to the objective of these empirical case studies. Each of the
chosen output measures requires being included in the production function alterna-
tively to study the impact that this can have on performance estimates. The production
approach can allow such an exercise more conveniently than other output specification
approaches. Furthermore, given that the nature of the analysis is aggregative rather than
firm-level, the production approach seems to rationalize the behaviour of the aggregate
banking industry quite well. Data constraints have played a pivotal role in choosing the
final set of output variables in Part III.
19 The data on this variable were obtained from the Global Monitoring Report on Non-
Bank Financial Intermediation dataset of the Financial Stability Board. This dataset
provides rich insights into the financial structure of various intermediaries, including
commercial banks.
20 Details on the variable-wise issues, adjustments and other dimensions may be available
in the online resources. Alternatively, they can be obtained directly from the author
upon request.
21 The exact output variable and the definition of each labour and capital productivity
measure used have been clearly stated in notes to each of the supporting tables in the
appendix at the end of this and the next chapter.
22 Only the first and second moments are analysed here. The primary reason for this is that
the data are industry-level aggregates rather than firm-level estimates. Going beyond
the second moment did not seem to make much sense as the disaggregated information
is unavailable. Reading too much in too little data may not be a plausible approach, so
further moments are not presented.
282 Empirical case studies
23 The use of traditional descriptive statistics and the associated distribution theory for
social data (information derived from human behaviour) has been criticised because data
in social sciences do not easily lend themselves to the statistical assumptions demanded
by traditional distributions theory. Various conceptual and philosophical issues are raised
when mainstream statistical theory is applied to economic and social data without rec-
ognising the pitfalls. Winkler (2009) analyses these issues in detail and provides a hetero-
dox view on analysing social science data. The author argues for an alternative narrative
in analysing economic and social data that does not demand the strict assumptions
required by the traditional statistical theory for data in pure sciences.
24 Estimates of skewness and kurtosis have not been reported here but may be provided in
the online resources or can directly be obtained from the author on request.
25 The residual TFP index was built as a ratio of the total output index (wherein only one
output measure has been used at a time) and a total inputs index (with multiple inputs
that include labour, physical capital and financial capital). The main debate is on the
choice of the weights for aggregating the inputs into the index of the total input. This
study uses the output elasticities of each of the chosen inputs estimated using a Cobb-
Douglas production function under Ordinary Least Squares. Some mathematical and
economic conditions need to be fulfilled while using this approach. The first is that the
sign of the output elasticity coefficient for each chosen output should be positive. The
second and weaker condition is that failing the first requirement, at least the sum of
output elasticities of inputs should be positive. The second condition was consistently
fulfilled across most of the cases. In cases where it couldn’t be fulfilled, lags of the input
variables were tried to overcome the problem. The estimated weights used for aggregat-
ing inputs into a total inputs index are presented in Appendix 5.
26 A rigorous survey of the evolution of the DEA method and its empirical applications
can be found in Emrouznejad and Yang (2017).
27 Detailed analysis and extended comments may be obtained from the online resources
or can be availed from the author upon request. The purpose of Chapters 8 and 9 is to
indicate the broad analytical dimensions of the issue under investigation. Higher levels
of disaggregated data are required to infer any further insights than presented here.
28 Such observations are purely statistical. The economic content of the same needs further
investigation. As the primary aim of this part of the book is to provide case studies on
the empirical implications of alternative output measures, the economic issues of such
findings are much beyond the scope. Further research can address these concerns.
29 Another approach could have been to use the Fourier flexible form employed alterna-
tively in the literature. However, “a disadvantage (of Fourier flexible form of production
function) is that the number of Fourier terms can become very large, causing degrees of
freedom problems. Consequently, the method cannot be used for small datasets” (Cum-
mins & Weiss, 2013, p. 810).
30 In the case of some economies, data on the number of employees in the banking indus-
try were not available. Only in such cases have the data on the number of employees
in the financial services industry. Data on the number of branches were not available in
all the cases. Due to data constraints, the number of banking institutions was used as a
rough proxy for the size of physical capital employed by the banking industry.
31 Literature on banking has treated the number of branches as both input and output.
Use of the same as output is defended on the grounds that the number of branches is
representative of convenience (Fixler & Zieschang, 1999). However, this study uses the
number of branches to measure quasi-fixed capital input, which is presumed to vary
broadly (but slowly) in direct correlation to the demand for outputs (whether an asset,
liability or others). Technological advances and digitalisation of banking processes might
cast disagreement on such a belief. However, for the EMEs under consideration, physi-
cal infrastructure can be assumed to have a large scope for expansion due to under reach
of core banking services. Thus it might not be too unrealistic to assume that branch
 Banking: implications of using alternative output definitions 283
expansion is a key input change for inducing larger outputs of the banking industry
as a whole. For a defence of this variable as an output measure, see Srivastava (1999),
Kumbhakar and Sarkar (2003), and Asaftei and Kumbhakar (2008), among others. For
those who favour its use as an input, see the discussion and references in Chapter 10.
32 These estimates can be obtained from the author on request.
33 Please see Barua et al. (2015) for the form of the trans-log function used here and the
methodology to estimate returns to scale from the same.
34 In a few cases where consistent and continuous data on the number of branches were
not available, the data on the number of banking firms were used.
35 Multicollinearity is measured in terms of the value of Variance Inflation Factors.
36 The analysis in this section and in 9.4 has been kept basic and limited to statistical notes
rather than economic inferences. Such an exercise will require much more space and
data, which is outside the scope of this book.
37 Building international or global production frontiers in an industry such as banking
is difficult to justify theoretically. Banking activities are generally localised in terms of
the policies, institutions, markets and price determination. External shocks may impact
these elements, but in broad terms, the industry is specific to local conditions, and con-
ceptualising an international or global frontier seems unreliable. “One has to account
for country-specific differences in the environments in which firms operate in order to
make a global frontier meaningful” (Gaganis et al., 2013, p. 430). Very few studies have
adopted such an approach in the banking performance literature, maybe due to such
concerns.
38 Detailed estimated models can be obtained from the author on request.

References
Asaftei, G., & Kumbhakar, S. C. (2008). Regulation and efficiency in transition: The
case of Romanian banks. Journal of Regulatory Economics, 33(3), 253–282. https://doi.
org/10.1007/s11149-007-9041-0
Barua, A., Goldar, B., & Sharma, H. (2015). The role of technological conditions of production in
explaining India’s manufacturing growth, 1998–99 to 2007–08: Some policy perspectives. Centre
for International Trade and Development (Discussion Paper No. 15–04).
Battese, G. E., & Coelli, T. J. (1992). Frontier production functions, technical efficiency and panel
data: With application to paddy farmers in India. The Journal of Productivity Analysis, 3, 153–169.
Biddle, J. E. (2021). Progress through regression the life story of the empirical Cobb-Douglas production
function (1st ed.). Cambridge University Press. https://doi.org/10.1017/9781108679312
Boda, M., & Piklova, Z. (2021). Impact of an input-output specification on efficiency
scores in data envelopment analysis: A banking case study. RAIRO Operations Research,
55, S1551–S1583. https://doi.org/10.1051/ro/2020040
Boisvert, R. N. (1982). The Translog production function: Its properties, its several interpreta-
tions and estimation problems. New York State College of Agriculture and Life Sciences
(Working Paper No. A.E. Res. 82–28). https://ecommons.cornell.edu/bitstream/han-
dle/1813/65165/Cornell_Dyson_rb8228.pdf?sequence=1&isAllowed=y
Colander, D. C. (1993). The macrofoundations of micro. Eastern Economic Journal, 19(4),
447–457. www.jstor.org/stable/40325524
Cummins, J. D., & Weiss, M. A. (2013). Analyzing firm performance in the insurance
industry using frontier efficiency and productivity methods. In Handbook of insurance (2nd
ed., pp. 795–858). Springer.
Eling, M., & Luhnen, M. (2010). Efficiency in the international insurance industry: A cross-
country comparison. Journal of Banking and Finance, 34(7), 1497–1509.
284 Empirical case studies
Emrouznejad, A., & Yang, G. (2017). A survey and analysis of the first 40 years of scholarly
literature in DEA: 1978–2016. Socio-Economic Planning Sciences. https://doi.org/10.1016/j.
seps.2017.01.008
Fare, R., Grosskopf, S., & Lovell, C. A. K. (1994). Production frontiers (1st ed.). Cambridge
University Press.
Fecher, F., Kessler, D., Perlman, S., & Pestieau, P. (1993). Productive performance of the
French insurance indus. The Journal of Productivity Analysis, 4, 77–93.
Felipe, J., & McCombie, J. S. L. (2005). How sound are the foundations of the aggregate
production function? Eastern Economic Journal, 31(3), 467–488.
Felipe, J., & McCombie, J. S. L. (2014). The aggregate production function: ‘Not even wrong’.
Review of Political Economy, 26(1), 60–84. https://doi.org/10.1080/09538259.2013.874192
Fixler, D., & Zieschang, K. (1999). The productivity of the banking sector: Integrating
financial and production approaches to measuring financial service output. The Canadian
Journal of Economics, 32(2), 547–569.
Gaganis, C., Hasan, I., & Pasiouras, F. (2013). Efficiency and stock returns: Evidence from
the insurance industry. The Journal of Productivity Analysis, 40, 429–442. https://doi.
org/10.1007/s11123–013–0347-x
Kodde, D. A., & Palm, F. C. (1986). Wald criteria for jointly testing equality and inequality
restrictions. Econometrica, 54(5), 1243–1248.
Kouki, M., Park, S. S., & Renault, E. (2014). Estimating scale economies in financial inter-
mediation: A doubly indirect inference. The Journal of Productivity Analysis, 41, 351–365.
https://doi.org/10.1007/sl 1 123–013–0345-z
Kumbhakar, S. C., & Hjalmarsson, L. (1995). Labour-use efficiency in Swedish social insur-
ance offices. Journal of Applied Econometrics, 10, 33–47.
Kumbhakar, S. C., & Sarkar, S. (2003). Deregulation, ownership, and productivity growth
in the banking industry: Evidence from India. Journal of Money, Credit and Banking, 35(3),
403–424.
Magnus, J. R., & Vasnev, A. L. (2015). Interpretation and use of sensitivity in econometrics,
illustrated with forecast combinations. International Journal of Forecasting, 31, 769–781.
http://dx.doi.org/10.1016/j.ijforecast.2013.08.001
Méon, P.-G., & Weill, L. (2010). Does financial intermediation matter for macroeco-
nomic performance? Economic Modelling, 27, 296–303. https://doi.org/10.1016/j.
econmod.2009.09.009
Solow, R. (1957). Technical change and the aggregate production function. The Review of
Economics and Statistics, 39(3), 312–320.
Srivastava, P. (1999). Size, efficiency and financial reforms in Indian banking. Indian Council for
Research on International Economic Relations (Working Paper No. 49).
Visco, I. (2014). Lawrence R. Klein: Macroeconomics, econometrics and economic policy.
Journal of Policy Modeling, 36, 605–628. http://dx.doi.org/10.1016/j.jpolmod.2014.03.003
Winkler, O. W. (2009). Interpreting economic and social data a foundation of descriptive statistics (1st
ed.). Springer. https://doi.org/10.1007/978-3-540-68721-4
9 Testing the empirical
sensitivity of major
performance indicators
of the insurance industry under
alternative output definitions
Selected exercises

9.1. Background
Continuing the analysis undertaken in Chapter 8, a similar framework is applied
to the insurance industry in this chapter. The fundamental aims are twofold:
first is to find whether performance measures such as efficiency, productivity
and scale vary on account of different output measures of the aggregate insur-
ance industry, and the second is to know the extent of such variations if any.
This chapter investigates the above two objectives by adopting the similar meth-
odological, econometric and data-oriented approach as Chapter 8. In terms of
partial factor productivities, labour, physical capital, financial capital and opera-
tional productivities are defined using the ratios as shown in equations 1 and 2
in Chapter 8, but with output and inputs pertaining to the insurance industry as
discussed later. TFP is defined as explained in Chapter 8. Using only the Cobb-
Douglas production function, technical efficiency is measured under both the
standard OLS and SFA approaches. Unlike the banking production analysis in
the previous chapter, this chapter does not consider trans-logarithmic function.
There are two major reasons for this: first is that the true underlying production
function seemed fairly well-represented by the Cobb-Douglas specification in
both economic and econometric terms; second is that the econometric results
using the trans-log function yielded inconsistent and highly problematic esti-
mates. Using the principle of Occam’s razor, it was decided to adopt the Cobb-
Douglas specification, given its ability to explain a large portion of the aggregate
output variations in the insurance industry in the present study.1
Estimates in this chapter are reported separately for the total insurance indus-
try, life insurance industry and nonlife insurance segment. The insurance out-
put measures employed in this chapter for the total insurance industry are the
amount of total gross premiums, total net premiums, total claims, total net
claims (net of reinsurance received), total financial assets and the number of
total policies; for the life insurance industry they are amount of premiums, net
premium (net of commercial expenses), claims, financial assets/assets under
management (AUM) and the number of policies underwritten; for the nonlife
insurance industry they are amount of premiums, net premiums, claims, finan-
cial assets/assets under management and the number of policies underwritten.

DOI: 10.4324/9781003149828-12
286 Empirical case studies
The output set in this chapter largely consists of value measures. As noted in
the previous chapter, all the monetary variables are deflated in 2010 prices
using country-specific CPI and are expressed in local currencies. The data
sources are as per the discussion in the previous chapter.

9.2. Partial factor productivity estimates for the

insurance industry
Other than labour and physical capital productivities, operational productivity
(using the amount of real commercial expenses as an input) and financial capital
productivity (using the amount of real equity capital) is also estimated in this sec-
tion as and where data were available. Labour is defined as the number of employ-
ees in the concerned insurance industry (total, life or nonlife). In cases where data
were unavailable, the number of employees in the financial services sector or real
wage rate of the financial services sector was employed. Physical capital is defined
as the number of branches of the insurance industry or the number of total retail
agent outlets or insurance corporations (where data on branches were either
unavailable or not consistently available for the sample period). Tables 9.1 to 9.4
in appendix 1 present data on various partial factor productivities at the aggregate
industry level for Brazil, Russia, India and China. Due to data issues for Indonesia
and South Africa, their results have not been reported.
Table 9.1 (Brazil) presents the estimates of labour and operational produc-
tivities for total, life and nonlife segments separately. In terms of the mean
productivity levels, estimates within each industry segment and within each
productivity measure show wide differences on account of differences in output
measures.2 Financial assets have larger labour productivity than others, which
probably reflects the impact that stock versus flow output measures can have on
performance estimates. In terms of variability as shown by CV, except LPLC
and OPNLP, all other productivity estimates show fairly similar behaviour. In
the case of the Russian insurance industry, as shown in Table 9.2, when labour
productivity estimates are looked at in terms of their levels, financial assets
stand out (LPTFA). In terms of variability across all reported estimates, finan-
cial assets-based and total claims-based estimates show the largest CV (LPTFA
and LPTC). Table 9.3 (India) shows that the financial assets-based estimate has
a larger size than premiums. Using stock versus flow measures of output can
make a large difference in the final results. Surprisingly, volume-based output
measures generally show higher variability than the value measures. In the case
of China (Table 9.4), both the mean and variability of financial assets is the
highest. It seems that this output measure can induce large variations in pro-
ductivity estimates if not adjusted appropriately.3

9.3. Total factor productivity and technical efficiency

estimates for the insurance industry
Residual TFP index,4 DEA-based technical efficiency and SFA-based effi-
ciency results with regard to the insurance industry are presented in this section.
 Testing empirical sensitivity of major performance indicators 287
As explained in section 8.3 in Chapter 8, the methodological framework is
extended here to the insurance industry. As clarified earlier in this chapter, only
the Cobb-Douglas production specification is used mainly due to the aggregate
nature of the data and lack of theoretically acceptable results in using the trans-
log and even the Constant Elasticity of Substitution (CES) functions.
Table 9.5 estimates the average annual growth rate of the residual TFP index
for Brazil using alternative insurance output measures. Compared to the life
and nonlife segments individually, the total insurance segment shows a much
larger variation in the mean growth levels of the TFP index (TFPG1, TFPG2
and TFPG3). In this case (TFPG3), financial asset shows lesser variability than
premiums and claims while recording the highest average growth rate during
the sample period among the three estimates. In the case of the life insurance
segment, the mean growth rates are fairly similar, but in terms of variability,
the net life premiums-based measure (TFPG6) displays a very high variability.
Estimates for the nonlife segment (TFPG7 and TFPG8) seem fairly robust to
output specifications. In terms of the skewness of the distribution of estimated
TFPG series using different output measures, while all the series show moder-
ate to high skewness, the asset output for the total insurance segment and all the
estimates for the life segment show high skewness. Depending on the output
measure used and the insurance segment focused upon (total, life and nonlife),
skewness changes drastically from being negative to positive. One can only
appreciate the policy and analytical implications of such dramatic shifts in TFP
growths under different output measures.
Regarding the uniqueness of information content, as measured by the
correlation coefficient, it appears, at least for within-segment estimates, that
changing the output definitions adds new information due to the low and
insignificant correlation coefficient. Regarding the estimates of technical effi-
ciencies, in Table 9.6, within each segment, the first and third moments show
a fair amount of similarity, though there are some differences among them.
These differences may possess sizeable implications given that these data rep-
resent the entire segments in aggregated form. For the life segment, however,
the third moment of the distribution of the estimated TE index shows an
appreciable difference (TED3 and TED4). Using premiums versus claims can
induce significant differences in how the efficiency is concentrated around its
mean. The correlation coefficients are generally positive, high and significant,
and thus it seems that irrespective of the chosen output measure, efficiency shall
behave quite similarly.
In the case of Russia, as shown in Table 9.7, the sensitivity of mean levels of
growth in the TFP index is quite high wherein, at times, the average growth
is positive, and for some outputs, it is negative. This difference is sizeably wide
too. The differences in variability are also quite wide when different output
measures are employed. The premium output for total and nonlife segments
seems to be introducing considerable change in the distribution, as shown by
their positive and large skewness value. In terms of their cross-correlations, only
two pairs show a high correlation, namely TFPG1 (total premium) and TFPG5
(nonlife premium), and TFPG2 (gross claims) and TFPG3 (net claims). Thus,
288 Empirical case studies
the TFP growth has different paths on average when different output meas-
ures are employed. Concerning the efficiency estimates as shown in Table 9.8,
for the total insurance segment (life plus nonlife), premiums and assets show
much higher variability than claims output measures. However, claims-based
efficiency shows a much higher mean level of achievement on this account.
The correlations are largely low and signify substantial differences, and prob-
ably independent, in the temporal movements of aggregate technical efficiency
across the same period. In the Indian insurance industry context, as contained
in Table 9.9, the mean, variability and skewness in the residual TFP growth are
all considerably sensitive to the choice of output measure. Correlations being
quite weak, different output measures seem to be expressing different informa-
tion content and thus cannot serve meaningfully as simple alternatives to each
other in productivity analysis of the insurance industry. Choice of output will
deeply affect the behaviour of the productivity growth. From the perspective
of efficiency analysis, as shown in Table 9.10, variability is largely stable across
the estimated efficiencies using different output measures except in the case of
premium output measures (TED2, TED3 and TES3). Mean levels of efficien-
cies are largely high in all the cases. Correlations are also largely weak, though
it is moderately high in some cases, such as between claims (TED4) and assets
under management (TED5). Some degree of independence in the movements
of these efficiencies seems evident. For the Indonesian mainstream insurance
industry (excluding Takaful insurance), as shown in Table 9.11, average mean
and variability are widely fluctuating across output measures, and some evi-
dence of independence in the distribution of these estimates is visible. From the
angle of efficiency estimates in Table 9.12, the nature of skewness, variations in
mean levels and the variability of estimated efficiency series seem to be imply-
ing quite a similar behaviour.
In the case of China, differences in output measures seem to be inducing
considerable heterogeneity in the estimated average growth in the TFP index,
as shown in Table 9.13. Mean levels show diametrically opposite performance.
In terms of the analysis segment, the individual segments show larger variabil-
ity than when the composite of life and nonlife segments is analysed (TFPG3
and TFPG4 versus TFPG1 and TFP2). Skewness also shows some evidence of
sensitivity to the choice of output measure. On the empirical sensitivity of effi-
ciency estimates to alternative output measures as per Table 9.14, the first and
second moments show some sign of sensitivity, while at the higher moments,
this difference seems to vanish. This is reflected in the high and significant cor-
relation coefficients. There does not seem to be much impact of using different
output measures in this case. Evidence in Table 9.15 for the South African
insurance industry shows a higher impact of changes in output definitions on
mean, variability and skewness. However, efficiency estimates in Table 9.16
suggest otherwise.
 Testing empirical sensitivity of major performance indicators 289
9.4. Returns to scale estimates for the insurance industry
of selected EMEs
Extending the discussion in section 8.4 of Chapter 8, this section presents
the estimates on the returns to scale for the total insurance, life insurance and
nonlife insurance industries in Table 9.17 in Appendix 3. The underlying esti-
mations are reported in Tables 9.18 to 9.23 in appendix 3. The very nature of
RTS, i.e. whether increasing, decreasing or constant, changes for Brazil’s total
insurance industry when one switches from premiums to claims and assets. In
the case of life and nonlife segments separately, however, RTS remain increas-
ing while the size changes considerably. Premium output appears to be induc-
ing the largest RTS for Brazil. Except for the claims measure, all other outputs
yield a decreasing RTS for Russia. Claims-based RTS has the largest size. In
the case of India, RTS are negative and thus decreasing. Premiums-based esti-
mates are such across the segments. The magnitudes of the RTS coefficients
are also wide even when most of them are showing decreasing returns. Though
the RTS is largely increasing, the magnitudes vary considerably in Indonesia,
China and South Africa.
Appendix 1
Estimates for Section 9.2

Table 9.1 Partial factor productivity estimates for the insurance industry of Brazil

Statistic Aggregate insurance industry

LPTP LPTNP LPTC LPTFA

Mean 83330.59 69209.96 31229.31 147082.85

SD 34642.95 27028.52 15979.39 76972.73
CV 41.57 39.05 51.17 52.33

Statistic Life Insurance industry

LPLP LPNLP LPLC OPLP OPLC OPNLP

Mean 850.73 518.70 338.42 3.47 1.37 2.44
SD 442.80 268.30 249.54 0.81 0.49 1.86
CV 52.05 51.72 73.74 23.49 35.99 75.96

Statistic Nonlife industry

LPNP LPNNP LPNC OPNP OPNC OPNNP

Mean 82479.85 68691.26 30890.89 7.41 2.76 6.41
SD 34227.32 26783.65 15779.13 2.46 1.15 2.46
CV 41.50 38.99 51.08 33.16 41.45 38.34
Notes: 1. All the variables are measured in terms of 2010 constant Brazilian reals per unit of input,
wherein LP stands for Labour Productivity, and the concerned measures are in terms of per employee
in the insurance industry, while OP stands for Operational Productivity and the concerned measures are
in terms of per real of commercial expenses in life and nonlife segments; 2. Each variable represents a
measure of partial factor productivity with different output variables; 3. The various measures represent
the following output variables: LPTP – total premium per insurance industry employee, LPTNP – total
net premium per insurance industry employee, LPTC – total claims per insurance industry employee,
LPTFA – total financial assets of aggregate insurance companies per insurance industry employee, LPLP –
life insurance premium collected per insurance industry employee, LPNP – nonlife premium col-
lected per insurance industry employee, LPLC – life insurance claims per insurance industry employee,
LPNC – nonlife insurance claim per insurance industry employee, LPNLP – net life premium collected
per insurance industry employee, LPNC – nonlife claims per insurance industry employee; 4. OPLP –
life premium collected per Brazilian real of commercial expenses of life insurers, OPLC – life insur-
ance claims per Brazilian real of commercial expenses of life insurers, OPNLP – net life premium per
Brazilian real of commercial expenses of life insurers, OPNP – nonlife premium per Brazilian real of
commercial expenses of life insurers, OPNC – nonlife insurance claims per Brazilian real of commercial
expenses of life insurers, OPNNP – net nonlife premium per Brazilian real of commercial expenses of
life insurers; 5. All the variables are measured in 2010 Brazilian reals of the concerned output variable
per employee in the insurance industry, including health and reinsurance; 6. SD is standard deviation;
CV is the coefficient of variation measured in percentages.
Source: Author’s estimation.
 Testing empirical sensitivity of major performance indicators 291
Table 9.2 Partial factor productivity estimates for the insurance industry of Russia

Statistic LPTP LPTFA LPTC LPPOL IATP IAPOL

Mean 10.01 24.60 10.13 2137.80 3.61 697.56

SD 5.52 14.92 8.33 1251.43 1.45 164.77
CV 55.17 60.65 82.20 58.54 40.14 23.62
Notes: 1. LPTP is millions of constant 2010 Brazilian reals of total premium per employee in insurance
organisations, LPTFA is millions of constant 2010 Brazilian reals of financial assets of insurance compa-
nies per employee in insurance organisations, LPTC is millions of constant 2010 Brazilian reals of total
claims paid by insurers per employee in insurance organisations, LPPOL is the number of total policies
per employee in insurance organisations, IATP is millions of 2010 Brazilian reals of total premium per
insurance agent in the country and IAPOL is the number of total policies per insurance agent in the
country. 3. SD is standard deviation; CV is the coefficient of variation measured in percentages.
Source: Author’s estimation.

Table 9.3 Partial factor productivity estimates for the insurance industry in India

Statistic Total Insurance Life Insurance

Labour Force Labour Force

TLP TLA TLPO LLP LLPO

Mean 8.21 67.46 2364.58 8.13 2811.89

SD 3.81 30.18 422.32 3.62 414.91
CV 46.35 44.74 17.86 44.45 14.76

Statistic Total Insurance Physical Total Insurance Financial Life Insurance

Capital Capital Financial Capital

TBP TBPO FEPP FEPA LEPP

Mean 84.75 28050.75 7.05 59.89 6.67
SD 22.36 10284.70 1.77 18.60 2.26
CV 26.38 36.66 25.06 31.06 33.92

Statistic Life Insurance Physical Nonlife Insurance

Capital Labour Force

LBLP LBLPO LNP LNPO LNAUM

Mean 116.16 56862.59 8.37 1440.86 15.83
SD 30.55 45019.13 4.80 934.84 8.87
CV 26.30 79.17 57.29 64.88 56.04

Statistic Nonlife Insurance

Financial Capital

NEP NEAUM

Mean 8.03 15.37

SD 1.20 1.88
CV 14.96 12.23 -

(Continued)
292 Empirical case studies
Table 9.3 (Continued)

Statistic Nonlife Insurance

Physical Capital

NBP NBPO NBAUM

Mean 62.59 10542.10 118.88

SD 20.23 3819.15 35.71
CV 32.32 36.23 30.04
Notes: 1. TLP is millions of 2010 constant INR of total premium per insurance industry employee,
TLA is millions of 2010 constant INR of financial assets of insurance industry per insurance industry
employee, TLPO is the number of total policies per insurance industry employee, LLP is millions of
2010 constant INR of total premium per life insurance industry employee, LLPO is the number of life
insurance policies per life insurance industry employee, TBP is millions of 2010 constant INR of total
premium per branch in insurance industry, TBPO is the number of total policies per branch of the
insurance companies, FEPP is millions of 2010 constant INR of total premium per INR of equity capi-
tal of all insurers, FEPA is millions of 2010 constant INR of financial assets per INR of equity capital of
all insurers, LEPP is millions of 2010 constant INR of life insurance premium per INR of equity capital
of life insurers, LBLP is millions of 2010 constant INR of life insurance premium collected per branch
of life insurers, LBLPO is number of life insurance policies underwritten per branch of life insurers,
LNP is millions of 2010 constant INR of nonlife insurance premium per employee in the nonlife insur-
ance industry, LNPO is the number of nonlife policies underwritten per employee in nonlife insurance
companies, LNAUM is millions of 2010 constant INR of total assets under management of nonlife
insurance companies per employee in the nonlife insurance industry, NEP is millions of 2010 constant
INR of nonlife premium underwritten per INR of equity capital of nonlife insurers, NEAUM is mil-
lions of 2010 constant INR of total assets under management of nonlife insurance companies per INR
of equity capital of nonlife insurers, NBP is millions of 2010 constant INR of total premium per branch
of nonlife insurers, NBPO is number of nonlife policies underwritten per branch of nonlife insurance
companies, and NBAUM is millions of 2010 constant INR of total assets under management of nonlife
insurance companies per branch of nonlife insurers; 2. All the variables are expressed in terms of the
underlying output variable per unit of the concerned input variable. 3. SD is standard deviation, and
CV is the coefficient of variation expressed in percentages.
Source: Author’s estimation.

Table 9.4 Partial factor productivity estimates for the insurance industry of China

Statistic LTP LLP LNP LTA

Mean 2.45 1.88 1.24 7.32

SD 0.62 0.70 0.41 5.09
CV 25.18 37.06 33.45 69.55
Notes: 1. LPT is millions of 2010 constant yuans of the total premium underwritten in the insurance
industry per employee in the insurance industry, LLP is millions of 2010 constant yuans of life insurance
premiums underwritten per employee in the insurance industry, LNP is millions of 2010 constant yuans
of nonlife insurance premiums underwritten per employee in the insurance industry, LTA is millions of
2010 constant yuans of total financial assets per employee in the insurance industry; 2. All the variables
are expressed in terms of the underlying output variable per unit of the concerned input variable. 3. SD
is standard deviation, and CV is the coefficient of variation expressed in percentages.
Source: Author’s estimation.
Appendix 2
Estimates for section 9.3

Table 9.5 Residual TFP estimates for the insurance industry of Brazil for the sample period
under alternative measures of output

Descriptive Statistics Cross-correlations

Statistic TFPG1 TFPG2 TFPG3 Variables TFPIG1 TFPG2 TFPG3

Mean 3.54 6.74 20.63 TFPG2 -0.13

SD 14.36 21.62 59.02 TFPG3 0.05 0.01
CV 405.99 320.94 286.11 TFPG4 0.14 -0.49** 0.19
Skewness -0.83 0.96 1.81 TFPG5 0.16 0.29 0.19
Statistic TFPG4 TFPG5 TFPG6 TFPG6 0.68*** -0.04 0.21
Mean 14.93 16.21 17.7 TFPG7 0.99*** -0.13 0.05
SD 29.25 25.89 56.38 TFPG8 0.99*** -0.16 0.04
CV 195.87 159.7 318.57 Variables TFPG4 TFPG5 TFPG6
Skewness 1.63 1.78 1.78 TFPG5 0.25
Statistic TFPG7 TFPG8 TFPG6 0.39* 0.26
Mean 3.50 3.21 TFPG7 0.13 0.15 0.68***
SD 14.40 16.30 TFPG8 0.16 0.18 0.69***
CV 412.01 508.35 Variables TFPG7
Skewness -0.85 -0.82 TFPG8 0.99***
Notes: 1. TFPG is the yearly growth rate in the Total Factor Productivity Index; 2. In terms of the
underlying output variable, TFPG1 represents premium collected by life and nonlife insurers combined,
TFPG2 represents claims paid by all insurers, TFPG3 represents financial assets of all insurers, TFPG4
represents premium collected by life insurers, TFPG5 represents the claims paid by life insurers, TFPG6
represents net premium collected by life insurers, TFPG7 represents premium collected by nonlife
insurers, and TFPG8 represents net premium collected by nonlife insurers; 3. SD is standard deviation,
and CV is the coefficient of variation expressed in percentages; 4. ***, ** and * imply significance at 1%,
5% and 10% levels, respectively; 5. All the variables are in percentage terms.
Source: Author’s estimation.
294 Empirical case studies
Table 9.6 Technical efficiency estimates for the insurance industry of Brazil using multi-stage
DEA and Cobb-Douglas SFA under alternative measures of output

Statistic Descriptive Statistics

TED1 TED2 TED3 TED4 TED5 TED6 TES1 TES2

Mean 0.87 0.77 0.81 0.65 0.87 0.81 0.60 0.59

SD 0.15 0.17 0.21 0.25 0.15 0.17 0.23 0.24
CV 16.89 21.91 26 38.97 16.9 20.8 38.10 41.52
Skewness -0.73 -0.15 -1.12 0.56 -0.73 -0.26 0.20 0.07

Variables Cross-correlations

TED1 TED2 TED3 TED4 TED5

TED2 0.64***
TED3 0.84*** 0.48**
TED4 0.47* 0.75*** 0.21
TED5 0.99*** 0.64*** 0.84*** 0.47**
TED6 0.3 0.78*** 0.19 0.71*** 0.3
Notes: 1. TED is the Technical Efficiency Index derived using the DEA framework, and TES is the
Technical Efficiency Index derived using the SFA framework; 2. In terms of the underlying output vari-
ables, TED1 represents the total premium collected by all insurers, TED2 represents total claims paid
by all insurers, TED3 represents the amount of premium collected by life insurers, TED4 represents
the amount of claims paid by life insurers, TED5 represents the amount of premium paid by nonlife
insurers, TED6 represents the amount of claims paid by nonlife insurers, TES1 represents the amount
of premium paid by life insurers, and TES2 represents the amount of net premium paid by life insurers;
2. SD is standard deviation, and CV is the coefficient of variation expressed in percentages; 3. ***, ** and
*
imply significance at 1%, 5% and 10% levels, respectively; 4. All the variables are in percentage terms.
Source: Author’s estimation.

Table 9.7 Residual TFP estimates for the insurance industry of Russia for the sample period
under alternative measures of output

Statistic Descriptive Statistics Variables Cross-correlations

TFPG1 TFPG2 TFPG3 TFPG1 TFPG2 TFPG3

Mean 16.31 -4.81 -4.70 TFPG2 0.59***

SD 74.25 7.55 9.79 TFPG3 0.50** 0.92***
CV 455.35 -156.91 -208.19 TFPG4 -0.17 0.04 0.12
Skewness 2.84 0.04 0.15 TFPG5 0.93*** 0.43** 0.35
Statistic TFPG4 TFPG5 TFPG6 TFPG6 -0.17 -0.07 -0.10
Mean -1.10 71.36 -1.94 Variables TFPG4 TFPG5
SD 37.65 310.79 12.78 TFPG5 -0.23
CV -3421.58 435.55 -658.51 TFPG6 0.30 -0.09
Skewness 0.09 4.20 0.72
Notes: 1. TFPG is the yearly growth rate in the Total Factor Productivity Index; 2. In terms of the under-
lying output variable, TFPG1 represents the amount of premium collected by insurers, TFPG2 represents
the amount of total claims paid by insurers, TFPG3 represents the amount of net claims paid by insurers,
TFPG4 represents the amount of premium collected by life insurers, TFPG5 represents the amount of
premium collected by nonlife insurers, and TFPG6 represents the amount of financial assets of owned by
insurers; 3. SD is standard deviation, and CV is the coefficient of variation expressed in percentages; 4. ***, **
and * imply significance at 1%, 5% and 10% levels, respectively; 5. All the variables are in percentage terms.
Source: Author’s estimation.
 Testing empirical sensitivity of major performance indicators 295
Table 9.8 Technical efficiency estimates for the insurance industry of Russia using multi-
stage DEA and Cobb-Douglas production function with SFA under alternative
measures of output

Statistics Descriptive Statistics

TED1 TED2 TED3 TED4 TED5 TES1 TES2

Mean 0.81 0.98 0.98 0.72 0.6 0.86 0.85

SD 0.18 0.05 0.05 0.23 0.37 0.08 0.1
CV 22.49 4.9 4.9 32.6 62.31 9.08 12.19
Skewness -0.12 -2.45 -2.42 0.17 -0.06 -1.07 -0.6

Variables Cross-correlations

TED1 TED2 TED3 TED4

TED2 0.28
TED3 0.31 0.97***
TED4 -0.44** -0.01 -0.02
TED5 0.22 0.34 0.35 0.24
Notes: 1. TED is the Technical Efficiency Index estimated using DEA, while TES is the Technical
Efficiency Index estimated under the SFA framework; 2. In terms of the underlying output variable,
TED1 represents the amount of premiums collected by all the insurers, TED2 represents the amount of
claims paid by all the insurers, TED3 represents the amount of net claims paid by all the insurers, TED4
represents the amount of financial assets owned by all the insurers, TED5 represents the amount of
premium collected by life insurers, TES1 represents the amount of claims paid by all the insurers while
TES2 represents the amount of net claims paid by all the insurers; 3. SD is standard deviation, and CV
is the coefficient of variation expressed in percentages; 4. ***, ** and * imply significance at 1%, 5% and
10% levels respectively; 5. All the variables are in percentage terms.
Source: Author’s estimation.

Table 9.9 Residual TFP estimates for the insurance industry of India for the sample period
under alternative measures of output

Statistic Descriptive Statistics Variable Cross-correlations

TFPG1 TFPG2 TFPG3 TFPG1 TFPG2 TFPG3

Mean -0.96 14.98 7.36 TFPG2 0.52**

SD 20.93 30.83 18.03 TFPG3 -0.35 0.05
CV -2176.24 205.75 245.12 TFPG4 -0.05 -0.02 0.16
Skewness 1.00 1.61 0.28 TFPG5 -0.41* -0.30 0.28
Statistic TFPG4 TFPG5 Variable TFPG4
Mean 12.08 1.02 TFPG5 -0.04
SD 6.49 5.60
CV 53.72 550.01
Skewness 0.67 1.32
Notes: 1. TFPG is the yearly growth rate in the Total Factor Productivity Index; 2. In terms of the
underlying output variable, TFPG1 represents the amount of financial assets owned by all the insurers,
TFPG2 represents the amount of premium collected by life insurers, TFPG3 represents the amount of
claims paid by life insurers, TFPG4 represents the amount of premium collected by nonlife insurers,
TFPG5 represents the assets under management of nonlife insurance companies; 3. SD is standard devi-
ation, and CV is the coefficient of variation expressed in percentages; 4. ***, ** and * imply significance
at 1%, 5% and 10% levels, respectively; 5. All the variables are in percentage terms.
Source: Author’s estimation.
296 Empirical case studies
Table 9.10 Technical efficiency estimates using multi-stage DEA and Cobb-Douglas pro-
duction function using SFA for the insurance industry of India for the sample
period under alternative measures of output

Statistics Descriptive Statistics

TED1 TED2 TED3 TED4

Mean 0.98 0.85 0.84 0.96

SD 0.05 0.16 0.24 0.08
CV 4.83 19.26 28.09 8.63
Skewness -2.75 -0.63 -1.45 -3.05

Statistics TED5 TES1 TES2 TES3

Mean 0.99 0.91 0.90 0.81
SD 0.02 0.09 0.07 0.12
CV 2.13 9.76 7.42 15.45
Skewness -1.68 -1.06 -2.38 -0.45

Variable Cross-correlations

TED1 TED2 TED3 TED4

TED2 -0.16
TED3 -0.30 -0.46
TED4 0.18 0.12 -0.24
TED5 0.07 0.45 -0.36 0.66

Variable TES1 TES2

TES2 0.07
TES3 -0.38 -0.24
Notes: 1. TED is the Technical Efficiency Index estimated using DEA, while TES is the Technical
Efficiency Index estimated under the SFA framework; 2. In terms of the underlying output variable,
TED1 represents the amount of financial assets owned by all the insurers, TED2 represents the amount
of premium collected by life insurers, TED3 represents the amount of premium collected by nonlife
insurers, TED4 represents the amount of claims paid by life insurers, TED5 represents the amount of
assets under management of nonlife insurance companies, TES1 represents the amount of financial
assets owned by all the insurers, TES2 represents the amount of benefits paid by life insurers, and TES3
represents the amount of premium collected by nonlife insurers; 3. SD is standard deviation, and CV
is the coefficient of variation expressed in percentages; 4. ***, ** and * imply significance at 1%, 5% and
10% levels, respectively; 5. All the variables are in percentage terms.
Source: Author’s estimation.

Table 9.11 Residual TFP estimates for the insurance industr y of Indonesia for the sample
period under alternative measures of output

Statistics Descriptive Statistics

TFPG1 TFPG2 TFPG3

Mean 0.18 4.87 0.91

SD 12.22 36.02 16.13
CV 6874.77 739.98 1781.09
Skewness -0.57 1.81 1.03
 Testing empirical sensitivity of major performance indicators 297

Variable Cross-correlations

TFPG1 TFPG2
TFPG2 -0.17
TFPG3 0.14 -0.06
Notes: 1. TFPG is the yearly growth rate in the Total Factor Productivity Index; 2. In terms of the
underlying output variable, TFPG1 represents the amount of premium collected by nonlife insurers,
TFPG2 represents the amount of financial assets owned by all the insurers, and TFPG3 represents the
total number of policyholders of all insurers; 3. SD is standard deviation, and CV is the coefficient of
variation expressed in percentages; 4. ***, ** and * imply significance at 1%, 5% and 10% levels, respec-
tively; 5. All the variables are in percentage terms.
Source: Author’s estimation.

Table 9.12 
Technical efficiency estimates for the insurance industry of Indonesia using
multi-stage DEA and Cobb-Douglas production function based SFA for the
sample period under alternative measures of output

Statistic Descriptive Statistics

TED1 TED2 TED3 TES1 TES2 TES3

Mean 0.96 0.92 0.83 0.87 0.90 0.74

SD 0.10 0.09 0.19 0.10 0.07 0.19
CV 10.15 10.23 23.12 11.06 7.60 25.51
Skewness -3.22 -0.67 -1.22 -1.60 -1.49 -0.65
Variable Cross-correlations Variables Cross-correlations
TED1 TED2 TES1 TES2

TED2 -0.28 TES2 -0.12

TED3 0.16 0.32 TES3 -0.02 0.30
Notes: 1. TED is the Technical Efficiency Index estimated using DEA, while TES is the Technical
Efficiency Index estimated under the SFA framework; 2. In terms of the underlying output variable,
TED1 represents the amount of financial assets owned by all the insurers, TED2 represents the total
number of policyholders of all insurers, TED3 represents the amount of premium collected by nonlife
insurers, TES1 represents the amount of financial assets owned by all the insurers, TES2 represents the
total number of policyholders of all insurers, and TES3 represents the amount of premium collected by
nonlife insurers; 3. SD is standard deviation, and CV is the coefficient of variation expressed in percent-
ages; 4. ***, ** and * imply significance at 1%, 5% and 10% levels, respectively; 5. All the variables are in
percentage terms.
Source: Author’s estimation.
Table 9.13 Residual TFP estimates for the insurance industry of China for the sample period
under alternative measures of output

Statistic Descriptive Statistics

TFPG1 TFPG2 TFPG3 TFPG4

Mean -1.29 7.67 0.42 -0.29

SD 10.70 16.68 13.43 10.99
CV -832.38 217.43 3182.85 -3787.80
Skewness 0.47 -2.39 0.06 0.67

Variables Cross-correlations

TFPG1 TFPG2 TFPG3

TFPG2 0.27
TFPG3 0.33 -0.12
TFPG4 0.57*** -0.01 0.75***
Notes: 1. TFPG is the yearly growth rate in the Total Factor Productivity Index; 2. In terms of the
underlying output variable, TFPG1 represents the amount of premium collected by all insurers, TFPG2
represents the amount of financial assets owned by all insurers, TFPG3 represents the amount of premium
collected by life insurers, and TFPG4 represents the amount of premium collected by nonlife insurers;
3. SD is standard deviation, and CV is the coefficient of variation expressed in percentages; 4. ***, ** and
*
imply significance at 1%, 5% and 10% levels, respectively; 5. All the variables are in percentage terms.
Source: Author’s estimation.

Table 9.14 Technical efficiency estimates for the insurance industry of China using multi-
stage DEA and Cobb-Douglas production function based SFA for the sample
period under alternative measures of output

Statistic Descriptive Statistics

TED1 TED2 TED3

Mean 0.86 0.76 0.73

SD 0.16 0.35 0.27
CV 19.17 45.93 36.90
Skewness -0.50 -1.02 -0.06

Statistic TED4 TES1 TES2

Mean 0.77 0.84 0.87
SD 0.23 0.09 0.10
CV 29.27 10.78 11.01
Skewness -0.15 -1.06 -0.74
Correlation Coefficient (TES1, TES2) -0.10

Variables Cross-correlations

TED1 TED2 TED3

TED2 -0.56 ***

TED3 0.92*** -0.70***

TED4 0.95*** -0.67*** 0.98***
Notes: 1. TED is the Technical Efficiency Index estimated using DEA, while TES is the Technical Effi-
ciency Index estimated under the SFA framework; 2. In terms of the underlying output variable, TED1
represents the amount of premium collected by all insurers, TED2 represents the amount of financial
assets owned by all insurers, TED3 represents the amount of premium collected by life insurers, and
TED4 represents the amount of premium collected by nonlife insurers; 3. SD is standard deviation, and
CV is the coefficient of variation expressed in percentages; 4. ***, ** and * imply significance at 1%, 5%
and 10% levels, respectively; 5. All the variables are in percentage terms.
Source: Author’s estimation.
 Testing empirical sensitivity of major performance indicators 299

Table 9.15 Residual TFP estimates for the insurance industry of South Africa for the sample
period under alternative measures of output

Statistics Descriptive Statistics

TFPG1 TFPG2 TFPG3

Mean -4.30 1.08 -1.42

SD 11.51 12.22 11.25
CV -267.86 1133.98 -793.31
Skewness 0.79 1.90 -2.07

Variables Cross-correlations

TFPG1 TFPG2

TFPG2 0.47*
TFPG3 0.07 0.20
Notes: 1. TFPG is the yearly growth rate in the Total Factor Productivity Index; 2. In terms of the
underlying output variable, TFPG1 represents the amount of premium collected by life insurers, TFPG2
represents the amount of premium collected by nonlife insurers, and TFPG3 represents the amount of
financial assets owned by all the insurers; 3. SD is standard deviation, and CV is the coefficient of vari-
ation expressed in percentages; 4. ***, ** and * imply significance at 1%, 5% and 10% levels, respectively;
5. All the variables are in percentage terms.
Source: Author’s estimation.

Table 9.16 Technical efficiency estimates for the insurance industry of South Africa using
multi-stage DEA and Cobb-Douglas production function based SFA for the
sample period under alternative measures of output

Statistics Descriptive Statistics Variables Cross-correlations

TED1 TED2 TED3 TES1 TED1 TED2

Mean 0.88 0.92 0.91 0.91 TED2 0.27

SD 0.11 0.07 0.10 0.07 TED3 0.24 0.75***
CV 12.73 7.57 11.16 7.45
Skewness -0.85 -0.86 -1.00 -0.67
Notes: 1. TED is the Technical Efficiency Index estimated using DEA, while TES is the Technical
Efficiency Index estimated under the SFA framework; 2. In terms of the underlying output variable,
TED1 represents the amount of financial assets owned by all the insurers, TED2 represents the amount
of premium collected by life insurers, TED3 represents the amount of premium collected by nonlife
insurers, TES1 represents the amount of premium collected by life insurers; 3. SD is standard deviation,
and CV is the coefficient of variation expressed in percentages; 4. ***, ** and * imply significance at 1%,
5% and 10% levels, respectively; 5. All the variables are in percentage terms.
Source: Author’s estimation.
Appendix 3
Estimates for section 9.4

Table 9.17 Estimated returns to scale coefficients across alternative measures of output for
the insurance industry of the selected emerging economies

Perspective Output Variable Country

Brazil Russia India Indonesia China South

Africa

Total Insurance Premium 3.45 0.27 -3.39 NA 1.9 NA

Industry (TPREM)
Claims -0.63 1.94 NA
(TCLAIM)
Net Claims NA 2.01 NA
(TNCLAIM)
Financial Assets 0.12 0.96 -0.97 1.19 3.9 1.04
(TFAST)
Life Insurance Premium 3.58 -0.75 -7.49 NA 1.22 0.42
Industry (LPREM)
Net Premium 5.84 NA
(LNPREM)
Policyholders NA 1.53 NA
(LPOL)
Claims 2.32 NA 3.98 NA NA
(LCLAIM)
Nonlife Premium 3.48 -0.56 -1.42 4.17 1.86 1.35
Insurance (NPREM)
Industry Net Premium 4.02 NA
(NNPREM)
Claims -0.41 NA
(NCLAIM)
Assets under NA 0.25 NA
management
(NAUM)
Table 9.18 Estimated returns to scale for the Brazilian insurance industry under alternative output measures

Input Variable Output Variables

Total Insurance Life Insurance

In(TPREM) In(TCLAIM) In(TFAST) In(LPREM) In(NLPREM) In(LCLAIM)

In(LAB) 0.71*** -0.41*** -0.49*** 0.52*** 0.85*** 1.19***

1st lag 1st lag
In(PCAP) 2.74# - 3.06** 4.99*** -
1st lag 1st lag
In(TREStotal) - -0.22** -

Testing empirical sensitivity of major performance indicators 301

(0.087)
In(TRESlife) - 0.47***
In(PREMtotal) - 0.61
***
-
In(PREMlife) - 0.66***
1st lag
Estimated RTS 3.45 -0.63 0.12 3.58 5.84 2.32
F-statistic 6.38*** 10.25*** 33.36*** 11.16*** 19.60*** 105.50***
R2 0.87 0.81 0.79 0.98 0.97 0.85

Input Variable Output Variables

Nonlife Insurance

In(NPREM) In(NNPREM) In(NCLAIM)

In(LAB) 0.71 ***
0.86 ***
-0.47***
In(PCAP) 2.77# 3.16# -
In(TRESnonlife) -0.13**
In(PREMnonlife) - 0.19**
1st lag
Estimated RTS 3.48 4.02 -0.41
F-statistic 6.39*** 7.33*** 39.70***
R2 0.87 0.84 0.92
Notes: 1. In(LAB) is the natural log of the total number of labourers in life, nonlife, health and reinsurance industries; 2. ***, ** and * imply significance at 1%, 5% and 10%
levels, respectively.
Source: Author’s estimation.
 302 Empirical case studies
Table 9.19 Estimated returns to scale for the Russian insurance industry under alternative output measures

Input Variables Output Variable

Total Insurance Life Insurance Nonlife Insurance

In(TPREM) In(TCLAIM) In(TNCLAIM) In(TFAST) In(LPREM) In(NPREM)

In(LAB) 0.44*** 0.13* 0.17** -0.99*** 0.40* -0.96***

Real Wage Real Wage Real Wage
1st lag
In(AGENT) 0.13 -
2nd lag
In(PCAP) -0.30*** 0.88*** 0.89*** - -1.15** 0.40***
In(TREStotal) 0.93*** 0.95*** -
In(PREMtotal) - 0.58*** -
In(TPOL) - 1.37*** -
Estimated 0.27 1.94 2.01 0.96 -0.75 -0.56
RTS
F-statistic 140.01*** 98.99*** 76.23*** 43.64*** 4.04** 107.10***
R2 0.96 0.94 0.93 0.88 0.90 0.92
Notes: 1. In(LAB) is the natural log of the average number of employees in insurance organisations across the country; 2. ***, ** and * imply significance at 1%, 5% and 10%
levels, respectively.
Source: Author’s estimation.
Table 9.20 Estimated returns to scale for the Indian insurance industry under alternative output measures

Input Variables Output Variable

Total Insurance Life Insurance Nonlife Insurance

In(TPREM)††† In(TFAST)†† In(LPREMnew) In(LBEN) In(NPREM) In(NAUM)†

In(LABtotal) -4.17*** -1.74 - -0.72

1st lag

Testing empirical sensitivity of major performance indicators 303

In(LABlife) - -8.34*** 2.79### -
In(LABnonlife) - -1.92* -
In(PCAPlife) 0.55*** - 0.85*** -
In(PCAPnonlife) (life + nonlife branches) - 0.50# -
In(AGENTlife) - 0.24* -
In(TREStotal) - 0.48 -
In(TRESlife) - 0.95*** -
In(TRESnonlife) 0.18
In(FCAPtotal) 0.23*** 0.14 -
In(FCAPnonlife) - 0.30
In(PREMtotal) - 0.15 -
1st lag
In(PREMnonlife) - 0.49
Estimated RTS -3.39 -0.97 -7.49 3.98 -1.42 0.25
F-statistic 21.96*** 21.85*** 88.68*** 194.16*** 356.25*** 99.94***
R2 0.79 0.84 0.90 0.97 0.97 0.96
Notes: 1. ***, ** and * imply significance at 1%, 5% and 10% levels, respectively; 2. † – Ridge regression model due to high collinearity among regressors under the standard
OLS specification and the k-value is 0.069 which was found to be the optimum value; 3. †† – Ridge regression model due to high collinearity among regressors under
the standard OLS specification and the k-value is 0.272 which was found to be the optimum value; 4. ††† – Shows the estimates of the ridge regression model due to high
collinearity among regressors under the standard OLS specification. Chosen k-value is 0.285, which was found to be the optimum value.
Source: Author’s estimation.
304 Empirical case studies
Table 9.21 Estimated returns to scale for the Indonesian insurance industry under alternative
output measures

Input Variable Output Variables

Total Insurance Life Insurance Nonlife Insurance

In(TFAST) In(LPOLH) In(NPREM)

In(LAB) 0.97***
0.16 ***
0.38***
In(TRESnonlife) -0.25** -
In(TRESlife) 0.47** -
In(PCAPtotal) - 1.37** 3.79**
Estimated 1.19 1.53 4.17
RTS
F-statistic 123.76*** 4.98** 6.26***
R2 0.95 0.34 0.40
Notes: 1. In(LAB) is the natural log of the total number of employees in the country’s Financial Inter-
mediation sector, which includes insurance, banking, and other financial services. This variable has been
used as a proxy for employment in the insurance industry due to a lack of data; 2. ***, ** and * imply
significance at 1%, 5% and 10% levels, respectively.
Source: Author’s estimation.

Table 9.22 Estimated returns to scale for the insurance industry of China under alternative
output measures

Input Variable Output Variable

Total Insurance Life Insurance Nonlife Insurance

In(TPREM) In(TFAST) In(LPREM) In(NPREM)

In(LAB) 0.63* -0.83 2.14*** 2.33***

2nd lag 1st lag
In(PCAPtotal) 1.27*** 6.22*** -0.92*** -0.47**
In(TPREM) - -1.49*** -
Estimated 1.9 3.9 1.22 1.86
RTS
F-statistic 21.25*** 210.02*** 124.70*** 113.78***
R2 0.97 0.97 0.94 0.93
Notes: 1. In(LAB) is the natural log of the total number of employees in insurance companies; 2. ***, **
and * imply significance at 1%, 5% and 10% levels, respectively.
Source: Author’s estimation.
 Testing empirical sensitivity of major performance indicators 305
Table 9.23 Estimated returns to scale for the insurance industry of South Africa under alter-
native output measures

Input Variable Output Variable

Total Insurance Life Insurance Nonlife Insurance

In(TFAST) In(LPREM) In(NPREM)

In(LAB) 0.45*
0.86 ***
2.07***
2nd lag
In(TPREM) 0.59*** -
In(PCAPtotal) - -0.44## -0.72
Estimated 1.04 0.42 1.35
RTS
F-statistic 37.92*** 35.19*** 15.86***
R2 0.82 0.79 0.76
Notes: 1. In(LAB) is the natural log of the number of employees in the country’s service sector is avail-
able from the International Labour Organization database; 2. ***, ** and * imply significance at 1%, 5%
and 10% levels, respectively.
Source: Author’s estimation.

Notes
1 The detailed empirical results on the trans-log specification can be obtained from the
author on request.
2 This point is made here and in all further sections while keeping in mind that value and
volume measures of output cannot be compared in terms of mean levels.
3 One of the ways to deal with this can be to index both the output and input and then
calculate their ratio. In this case, one can estimate the growth in labour productivity and
make these diverse output measures possibly more comparable in terms of mean levels.
However, productivity per se is a level form-based economic concept, and its growth rate
throws very different economic dimensions than its absolute level. Hence, absolute level-
form partial factor productivity estimates are reported instead of their growth rates. The
estimates of growth in these partial factor productivity indexes can be obtained from the
author on request. The divergence in these estimates widens much more when estimat-
ing growth rates in productivities.
4 Due to space constraints, the estimated output elasticities of inputs used as weighting
factors in constructing the total inputs index are not reported here. The same may be
made available in the online resources and can happily be obtained from the author upon
request.
Part IV

Extended notes and

concluding remarks
10 Inputs measurement issues

10.1. Introductory note
The previous chapters have analysed the literature on output measurement
in banking and insurance performance analyses. Contributions from selected
experts were also presented, after which case studies were elaborated on the
empirical implications of using different output measurements for performance
estimation. Another closely related element in the larger debates on produc-
tion analysis of banking and insurance industries is the measurement of inputs.
Given that the production process cannot be conceptualised without sound
measures of both outputs and inputs, it is natural for performance analysts to
be concerned with both these dimensions. Contrasted with the diversity of
opinions on the outputs of banks and insurers and how to measure them, there
clearly appears to be a consensus on how to measure the inputs (Eling & Luh-
nen, 2010). This chapter briefly summarises the various dimensions of inputs
measurement1 in banking and insurance literature on performance analysis.

10.2. Inputs measurement in the banking

and insurance literature
The most fundamental inputs vector consists of volume measures of labour
and capital. Given the paucity of data, researchers have used value measures
instead. Labour input, in volume terms, is generally conceptualised in terms
of the employees’ working hours. The problem with this measure, other than
data availability, is that banks and insurers do not rely only on the labour of
employees who work from the physical premises of the firms but also on those
who are not employees but provide labour services from outside the prem-
ises. Insurance agents, brokers, banking correspondents, retail agent outlets,
and various other intermediaries provide valuable labour services by dedicating
their time to realising the business objectives of banking and insurance firms.
Reliable measurement of the time input of such labour services is difficult. This
has motivated analysts to conceptualise labour input in terms of the number of
labourers. Unlike the time-based measures, the headcount measure of labour-
ers in banks and insurance firms is relatively easier to locate. Reliable data are

DOI: 10.4324/9781003149828-14
310 Extended notes and concluding remarks
generally available. Labourers, in this context, include not just the employees
working physically on the premises of the firm but also the intermediaries,
as stated above. In many countries, official agencies require banks and insur-
ers to report such data. Thus, data constraints are somewhat lesser than the
time-based labour inputs. When information on the number of labourers is
also unavailable, another alternative in the literature in such cases has been
the monetary value of expenditure on labour services. This measure is a value
measure expressed in monetary terms rather than quantity terms. The use of
this approach to measure labour input reflects the amount of funds spent by
a bank or an insurance firm on compensating their employees for their time
and services. Concerning the intermediary labour services, data on commis-
sions are generally available. Thus, labour or personnel expenses have been a
frequently employed labour input measure.
The problem with this approach is that it is a nominal value and requires
appropriate deflation. Literature on firm-level studies has found it useful to use
specific input prices to deflate expenses’ nominal value. These prices are used
as deflators for each specific input variable. For more aggregative studies, the
traditional deflators of CPI and GDP deflator have been among the popular
choices. Ideally, inputs specific deflation must be adopted, but except for a
few advanced economies, hardly any economy can boast of such disaggregated
input price data for banking and insurance firms. In summary, labour has been
conceptualised in three major ways in the literature on banking and insurance
performance analysis: first is in terms of the time spent on performing labour
services; second is in terms of the number of labourers (both direct and inter-
mediary); and, third is in the form of nominal expenses on personnel deflated
by an appropriate deflator. Each has its own advantages and constraints, and the
choice on this account is largely driven by data availability.
As per the extant literature, the second element in the traditional input vec-
tor of banks and insurers is a volume measure of physical capital. The existing
literature has conceptualised physical capital mostly as fixed assets in this sense.
Generally, the expenses on or value of these assets are used as input. However,
this begs the question: what is capital for banks and insurers? It is well-
recognised in the literature discussed in previous chapters that the measurement
of capital input is quite complicated when it comes to banks. For example,
as captured by the number of branches, its physical premise was a favoured
measure when digitalisation was not a large-scale phenomenon. However, the
role of information and communications technology (ICT) as a capital input
is increasingly being recognised in the literature. ICT does not lend itself eas-
ily to quantification and further complicates the problems associated with the
measurement of traditional capital inputs. Among these problems, the issue
of incorporating quality changes in the quantity estimates is more serious for
ICT, where quality improvements are frequent and rapid. To dodge these con-
straints, researchers have largely avoided including ICT in the banking produc-
tion function and have allowed its impact on the residual technology parameter
instead.
 Inputs measurement issues 311
On the other hand, several studies have included the monetary value of
ICT assets such as computers and various payment terminals in the fixed assets
variable. The notable studies that have stressed the importance of including
ICT-related measures in the capital input variable of bank production function
include Brand and Duke (1982), Oral and Yolalan (1990), Anthanassopoulos
(1998), Thanassoulis (1999), Athanassopoulos and Giokas (2000), Allen and
Liu (2007) and Aksoy et al. (2022). In the case of insurance literature on similar
lines, some notable works on this account include Arvanitis (2005), Bertschek
et al. (2006) and Noreen and Ahmad (2016).2
The traditional input vector exhausts at this point, and new inputs have
been proposed in the literature that captures the specificities of the production
process in the banking and insurance industries. Capital has been expanded to
include financial capital3 also. Financial equity capital is considered an impor-
tant input in modern financial theory (Hughes & Mester, 1998) and was sub-
sequently incorporated into banking and insurance production analyses. This
variable has been measured by equity capital as per the balance sheet of the
individual banks and insurers.4 This is a stock variable specified in nominal
terms in most sample studies, as examined in Chapters 3 and 4. The impor-
tance of this variable in a bank’s or the banking industry’s production function
has been well argued by Hartwick (1996). Equity capital is also hypothesised
as a measure of the risk management function of banks, while it is theorised
as a representation of solvency for insurers. As discussed in the previous chap-
ters, equity capital has been accepted as a critical element of the inputs vector
of banks and insurers, which cannot be ignored. It was noted in Chapters 3
and 4 that the so-called output specification approaches also define the rel-
evant inputs. Hence, outputs and inputs are specified simultaneously. The set
of inputs discussed till now is accepted universally by all the output specifica-
tion approaches. This set includes labour, physical capital and equity capital.
Beyond these inputs, disagreements on what inputs must be included in the
production function of banks and insurers arise. The intermediation approach
in banking and insurance suggests that deposits and premiums must be treated
as inputs for banks and insurers. The same variables are treated as output in the
production and value-added approaches in banking and insurance literature.
Lastly, an input variable that has been less frequently employed in literature is
the value of materials. This variable consists of intermediate goods and services
used by insurers, such as lawyer fees, telephone use, stationary expenses, etc. It
is frequently used in the case of studies on advanced economies. However, data
on this input are not reliably available for economies like Brazil, China, Russia
and other EMEs. In insurance performance studies, there is another debate on
treating technical reserves as an input or an output.5 When treated as an input,
it is proxied as a measure of debt capital.
Another approach to input specification has been less frequently used explic-
itly in the literature. It is the so-called operating expenses approach, and it
suggests that all the inputs that occupy a sizeable share in the total operating
expenses of the insurance firm must be included because “It is misleading and
312 Extended notes and concluding remarks
incorrect to omit an input that accounts for a significant proportion of operat-
ing expenses” (Cummins & Weiss, 2013, p. 821). However, this is an empiricist
approach and requires information on each item’s total operating expense func-
tion and share. Such an approach for specifying inputs is plausible for firm-level
studies but may not suit industry or sectoral studies in banking and insurance
performance analysis. Finally, the national income approach has been used to
specify outputs rather than outputs and inputs. No studies were found that
claimed to have used the national income approach to specify inputs in their
study – either at the firm level or sectoral analysis.

10.3. Concluding remarks
The scope of discussion on inputs measurement in banking and insurance
literature is large. There are many theoretical and empirical challenges that
researchers have to deal with when specifying the inputs to be included in the
production function. Depending on the performance parameter being studied,
the economy being analysed and the data availability, different permutations
and combinations are possible that will lead to alternative input vectors. How-
ever, there is considerable agreement in the literature on the major inputs. This
agreement has been presented in this chapter. Given the primary focus of this
work on output measurement issues, the input measurement dimension has
only been briefly reviewed in this chapter.

Notes
1 See Maroto-Sánchez (2011) for a discussion on various issues in inputs measurement
when estimating performance of services industries in general.
2 Please refer to the appendices to Chapters 3 and 4 and the reference list given in those
chapters for more details on these studies.
3 Financial capital of banks and insurers consists of two major components – equity capital
and debt capital. Most of the studies either focus on equity capital per se, or combine
both of them into single financial capital input. For an illustration in insurance literature,
see Yaisawarng et al. (2012).
4 The use of equity capital in banking and insurance production processes as an input has
been briefly discussed in Chapters 3, 4, 8 and 9, and thus has not been elaborated on
here.
5 See Berry-Stölzle et al. (2011) for use of technical reserves as a measure of debt capital.
See Mahlberg and Url (2003) for use of technical reserves as a measure of output. Please
also see Alshammari et al. (2018).

References
Aksoy, E. E. A., Dirik, C., & Göker, I. E. K. (2022). Opening the black-box of bank effi-
ciency in Turkey with two-stage data envelopment analysis: A study on capital adequacy
ratio. EGE Academic Review, 22(1), 75–91. https://doi.org/10.21121/eab.1064816
Allen, J., & Liu, Y. (2007). Efficiency and economies of scale of large Canadian banks. The
Canadian Journal of Economics, 40(1), 225–244.
 Inputs measurement issues 313
Alshammari, A. A., bin Syed Jaafar Alhabshi, S. M., & Saiti, B. (2018). The impact of com-
petition on cost efficiency of insurance and takaful sectors: Evidence from GCC markets
based on the stochastic frontier analysis. Research in International Business and Finance, 1–29.
https://doi.org/10.1016/j.ribaf.2018.09.003
Anthanassopoulos, A. D. (1998). Nonparametric frontier models for assessing the market
and cost efficiency of LargeScale Bank branch networks. Journal of Money, Credit and Bank-
ing, 30(2), 172–192.
Arvanitis, S. (2005). Computerisation, workplace organisation, skilled labour and firm pro-
ductivity: Evidence for the Swiss business sector. Economics of Innovation and New Technol-
ogy, 14(4), 225–249. https://doi.org/10.1080/1043859042000226257
Athanassopoulos, A. D., & Giokas, D. (2000). The use of data envelopment analysis in bank-
ing institutions: Evidence from the commercial bank of Greece. Interfaces, 30(2), 81–95.
Berry-Stölzle, T. R., Weiss, M. A., & Wende, S. (2011). Market structure, efficiency, and perfor-
mance in the European property-liability insurance industry (Working Paper).
Bertschek, I., Fryges, H., & Kaiser, U. (2006). B2B or not to be: Does B2B E-commerce
increase labour productivity? International Journal of the Economics of Business, 13(3), 387–
405. https://doi.org/10.1080/13571510600961395
Brand, H., & Duke, J. (1982). Productivity in commercial banking: Computers spur the
advance. Monthly Labor Review, 105(12), 19–27.
Cummins, J. D., & Weiss, M. A. (2013). Analysing firm performance in the insurance indus-
try using frontier efficiency and productivity methods. In Handbook of insurance (2nd ed.,
pp. 795–858). Springer.
Eling, M., & Luhnen, M. (2010). Efficiency in the international insurance industry: A cross-
country comparison. Journal of Banking and Finance, 34(7), 1497–1509.
Hartwick, J. M. (1996). The financial capital constraint and the valuation of commercial banking
activity. Queen’s Economics Department (Working Paper No. 942).
Hughes, J. P., & Mester, L. J. (1998). Bank capitalization and cost: Evidence of scale econ-
omies in risk management and signaling. The Review of Economics and Statistics, 80(2),
314–325.
Mahlberg, B., & Url, T. (2003). Effects of the single market on the Austrian insurance
industry. Empirical Economics, 28, 813–838. https://doi.org/10.1007/s00181-003-0164-6
Maroto-Sánchez, A. (2011). Productivity in the services sector: Conventional and current
explanations. The Service Industries Journal, 32(5), 719–746. https://doi.org/10.1080/026
42069.2010.531266
Noreen, U., & Ahmad, S. (2016). Cost efficiency and total factor productivity: An empiri-
cal analysis of Pakistan’s insurance sector. The Lahore Journal of Economics, 21(1), 123–150.
Oral, M., & Yolalan, R. (1990). An empirical study on measuring operating efficiency and
profitability of bank branches. European Journal of Operational Research, 46, 282–294.
Thanassoulis, E. (1999). Data envelopment analysis and its use in banking. Interfaces, 29(3),
1–13.
Yaisawarng, S., Asavadachanukorn, P., & Yaisawarng, S. (2012). Efficiency and productiv-
ity in the Thai non-life insurance industry. The Journal of Productivity Analysis, 41(2),
291–306. https://doi.org/10.1007/s11123-012-0317-8
11 Limitations, future scope
and concluding remarks

11.1. Limitations and future scope

The analysis presented in this work is subject to constraints that must be brought
to the notice of the readers at this juncture. Every research attempt is subject
to factors that may be out of the researcher’s control or maybe controllable but
still must be sacrificed due to time and space restrictions. The insights provided
in this work are also subject to such constraints, and a short note on them may
help clarify some of the concerns in advance. Lack of conscious recognition of
limitations leads to various barriers and biases (Singh, 2015).
The first point in this regard is that this work has not attempted to build a
consensus on the issue of output measurement in banking and insurance per-
formance1 analysis. Rather, the aim has been to summarise the ideas, debates,
agreements and disagreements on the theme. While broad patterns in the lit-
erature have been highlighted and explained, locating a consensus on a topic
of such rich analytical diversity will require a work of a much larger scope.
The author of this work found that there did not exist any reference on output
measurement issues in banking and insurance performance analysis that could
be utilized not only for systematization of the thoughts on this subject but also
to obtain a summary account of the voluminous literature in this area. Such
concerns have motivated this work. The second point is that literature reviews
are also expected to critique the extant wisdom, which is an essential step in
any systematic review work (Machi & McEvoy, 2022). While providing criti-
cal comments and notes on various issues across the chapters, this work is not
a formal critique of what has been done on output measurement in banking
and insurance literature. A formal critique is beyond the scope of this book.
Furthermore, without having a compact summary of what has been done so
far, critiquing the extant literature would be unrealistic. It is hoped that this
work will provide some background and foundations for more rigorous ana-
lytical attempts.
The third matter is the sample size of studies included in this book. While the
author is motivated by the fact that the combined sample size (for banking and
insurance literature) of approximately 500 high-quality scholarly works is possibly
the first attempt of its kind, particularly when coupled with contributions from

DOI: 10.4324/9781003149828-15
 Limitations, future scope and concluding remarks 315
practitioners and policymakers and also accompanied by empirical case stud-
ies, it is also duly recognized that a larger sample could have been analysed to
incorporate more diversity in the analysis. On this account, it may kindly be
pointed out again that the number of studies reviewed was 872, but they had to
be downsized due to space restrictions.2 The fourth point is that the present work
does not attempt to delve into meta-analysis of the empirical evidence, which
is also a possible agenda for future research. Meta-analytic reviews are extremely
demanding in terms of time and prone to various errors if the samples of a large
number of diverse studies are not synthesized correctly. Such an attempt would
fall far outside the scope of this work, and the objectives of this study seem better
suited to a systematic review rather than a meta-analytic approach.
Further research can be undertaken on quantitatively synthesizing the empir-
ical sensitivity of various performance parameters to alternative output meas-
ures by using the empirical findings of the extant literature itself. Such research
will provide fascinating insights on this subject. It may also reveal the biases
that differences in output measures can bring into the estimates of performance
parameters such as efficiencies, productivity, scale and scope economies, com-
petitiveness, profitability and others. However, such a review work shall have
to build a very large sample to undertake a meaningful exercise of such scale.
The fifth observation is that the Part III of this work analysed banking and
insurance industries using aggregate industry-level data rather than firm-level
data.3 Generally, aggregation tends to smoothen out the heterogeneity in
firm-level data. This may be an information loss.4 Aggregated data also imply
that the quantitative production relations as estimated in Part III, are aver-
age relationships and thus represent the aggregate average production function.
However, it also provides a fresh vantage point to approach the banking and
insurance production process. The industry level point of view has largely been
ignored in the literature. While performance estimates from firm-level data are
extremely valuable, the insights from an industry-level aggregate point of view
are equally useful. As argued earlier in Chapter 8, industry-level performance
movements can provide worthwhile information for macroeconometric works
that incorporate banking and insurance industries in analysing the growth and
development process of the aggregate economy. Aggregate industry-wide find-
ings are also important for firm-level decision making as they largely shape the
external environment that banking and insurance firms shall face. Hence, given
the lack of attention on aggregate analysis on the subject matter in particular
and its usefulness for macroeconomic analysis in general, the present study
made the current choice.
Further research can investigate the empirical sensitivity of performance
estimates to different output measures using a large panel of firms for various
economies. Data constraints, however, shall be a major concern for such an
attempt. Sixth, this study has not delved deeper into two issues close to output
measurement: inputs measurement (which has been briefly reviewed in Chap-
ter 10) and output price measurement. Further research must also study how
variations in inputs and output prices affect empirical estimates of efficiencies,
316 Extended notes and concluding remarks
productivities and other performance parameters. Seventh and last is that the
empirical case studies in this book are confined to a multi-input–single-output
framework. Future research can also examine the issues covered in this work
within a multi-input–multi-output approach.
Given these constraints and the possibilities for future efforts, the next sec-
tion concludes this study with final remarks.

11.2. Final concluding remarks

The present work has provided a summarization of a voluminous literature
on banking and insurance output measurement. Various approaches to the
specification of output were reviewed in Part II, along with alternative output
measures by invoking the practices in the extant literature. Conceptual schemas
were constructed and elaborated on that capture the consensus and debates on
output measurement in banking and insurance performance analysis. After that,
various thematic issues were also analysed to bring attention to some of the
most pressing matters. While reviews of literature generally restrict themselves
to what has been presented in Part I, this book also attempted to enrich the
narrative by inviting and providing contributions of various academics, policy-
makers and practitioners in this area from various countries and backgrounds.
Part II of this work thus enlarged the scope of review by allowing selected
professionals to provide their insights, experience and wisdom on this matter.
Part III supplied selected empirical case studies on how the empirical estimates
of partial factor productivities, total factor productivity, technical efficiency
and returns to scale may change when the underlying production function is
modelled using alternative output measures and alternative empirical methods
methodologies. The three parts combined summarize a substantial volume of
literature by bringing together a traditional review of literature, contributions
by selected professionals and empirical case studies on the BRIICS countries
together into a single work.
Several tentative findings may be summarized as follows. The first is that
systematic approaches largely drive the output vector for banks and insurers to
output specifications rather than ad hoc choices. Second, while output specifi-
cation approaches are used, their economic rationale is generally not rigorously
defended in a sizeable portion of the sample studies. Third, the stock-flow
dimension of output measurement requires much more attention and can-
not be ignored. There are large variations in performance estimates between
stock and flow measures of banking and insurance outputs. Similarly, the value-
volume dimension and nominal versus real output measures are found to be
inducing wide dispersion in the performance estimates. Fourth, clear patterns
in output choices are evident, as highlighted in Chapters 3 and 4. In other
words, output choices are not random in the literature and have meaningful
patterns over space and time. Fifth, a clear consensus on this subject matter
is yet to be established, but given the evolution of the practices in literature,
agreements in output measurement are overcoming the disagreements, though
 Limitations, future scope and concluding remarks 317
gradually. Sixth, well-defined output measures exist in the literature, and each
measure has been innovated to account for more complexities of banking and
insurance production. These innovations have been summarized in various fig-
ures and narrations in Chapters 3 and 4. The seventh key finding is that on
changing the output measure, the empirical estimates of performance analysis
change by wide margins in the case of the BRIICS economies. Particularly, the
distributional characteristics of estimated technical efficiency and productivi-
ties change noticeably when alternative output measures are employed in the
underlying production function. The differences seem to be more pronounced
at the first and third moments of estimated productivities and efficiency dis-
tribution. Analysis of RTS shows mixed results, and there is some degree of
consistency in terms of the nature of RTS (increasing, decreasing or constant),
while the size of the scale coefficient varies widely for most of the selected
emerging economies.
The attempts made in this work are hoped to extend further the research on
the underlying theme in newer and fascinating directions. Until now, what has
been done has probably been compactly summarized in this work. What needs
to be done has been suggested across this work. What will be done shall now
be determined by the directions that future research assumes on this matter.

Notes
1 Again, kindly note that performance parameters have been defined in this study to include
efficiencies, productivities, profitability, competitiveness, scale and scope economies, and
other related matters. Thus this term is not limited to only financial performance as is
common practice in the literature.
2 The author’s review sheet with detailed comments on all 872 studies is available on
request.
3 The rationale for adopting such an approach and its advantages and problems have been
provided in Chapter 8.
4 This also raises the question of the degree of “descriptive realism” (Broer, 1987) that an
aggregate empirical analysis can achieve. The answer to this question is open-ended and
must be analysed by keeping the context of the empirical analysis in mind. Regarding
the present study, no explicit attempt is made to achieve a high degree of realism in the
economic description laid down in Part III of the book. Rather, a broad set of case stud-
ies are provided to elaborate on the issues examined in the previous sections. The issues
such as efficiency, productivity and returns to scale are very complex and will require a
more rigorous empirical strategy than the one adopted here if a full-fledged analysis is to
be undertaken.

References
Broer, D. P. (1987). Neoclassical theory and empirical models of aggregate firm behaviour. Kluwer
Academic Publisher. doi:10.1007/978-94-009-447-7
Machi, L. A., & McEvoy, B. T. (2022). The literature review: Six steps to success. SAGE.
Singh, S. (2015). Hello, limitations! The paradoxical power of limits in scientific writing.
Indian Journal of Dermatology, Venereology and Leprology, 81, 4–6. https://doi.org/10.4103/
0378-6323.148555
Index

Note: Page numbers in italics indicates figures and page numbers in bold indicates tables on
the corresponding page.

aggregation 8, 49, 58 – 59, 62, 128, econometric 59 – 60, 210 – 211, 215,
182 – 183, 207 – 210, 221n4, 279n2, 315 257 – 259
aggregation problem 233, 279n2 empirical sensitivity 27 – 29, 110 – 111
alternative output 27 – 29, 110, 231, endogeneity 11n8
279n2, 285 evolution: of banking output 63; of
amount of deposits 44, 46, 238 – 239 insurance output 132
amount of loans 40 exogenous shocks 5
amount of premium 111, 114, 123, 201, externalities 177 – 178
206, 285
asset approach: banking literature 33, 36 FISIM 53–54, 56–57, 93n47, 172, 174, 176,
asset output 48 – 49 184, 185–186, 196, 199–201, 210–212
Averch and Johnson hypothesis 5, 129
granular data see granularity
bank output: balance sheet measures granularity 126, 188 – 190
27 – 28, 38, 48
BankFocus 26n13, 280n5 income output 121, 124
bias: publication 23; researcher 23; index numbers approach 38 – 39
survivor 23 India 241 – 242, 251 – 252, 258, 264 – 266,
Big Data 184, 211, 221n9, 222n18, 223n22 275, 277 – 278, 291, 295 – 296, 300
Brazil 241, 243, 248, 257, 275, 276, Indonesia 242, 253, 258, 267 – 269, 275,
286 – 287, 289, 290, 293, 300 – 301 277 – 278, 296 – 297, 300, 304
BRIICS 229 influence factor 9
input-output dilemma 46 – 47, 121, 205
China 242, 254 – 255, 258, 269 – 272, 275, inputs aggregation 276 – 278
278 – 279, 288, 292, 298, 300, 304 intangible 10n2
Cobb-Douglas 233 – 234, 240, 257 interest rate 43, 54 – 55; actual 56; average
constant price see deflation 41; competitive 49; deposit 57; effective
current price see deflation 40, 42; market 52 – 53; reference 53 – 54,
92n43, 93n47, 184; risk-adjusted 40;
DEA 35, 232, 234, 237 – 239, 249 – 250, true vs. observed 61
252 – 254, 256, 294 – 299 intermediation approach: banking literature
deflation 7, 11n13, 55, 130, 310; double 33 – 36, 51, 90n2; insurance literature
57, 164, 172 – 173; single 57 – 58, 112, 90n2; pure 36
93n51, 173 intermediation hypothesis 11n6
deposit: output 43 – 48; volume measures intermediation services 2
of 45 investment output 48, 50
 Index 319
life and nonlife 116, 233 – 234, 286 – 289 publication bias 23
literature review: ad-hoc approach 15 – 16; purposive selection 23
sample vs. population 15; systematic
approach 15 – 16 quality-adjusted measure see quality
loan: flow measures of 41; other measures adjusted output
of 42; output 39 – 43 quality-adjusted output 4
quantitative summary 17 – 23
macroeconometric model 204, 211
macroeconometrics 280n4, 280n7 regulatory interventions 4 – 5
maximisation criterion 6 researcher bias 23
measures of output: flow 51, 53, 58, 163; returns to scale 231 – 233, 239, 241, 257
stock 58, 119, 163 ridge regression 240 – 241
methodological robustness 5 – 6 Russia 241, 250 – 251, 257, 275 – 277, 291,
moments of distribution 236, 281n22, 294 – 295, 300
287 – 288
monetary policy 35, 60 – 61, 190 satellite account 185
multi-output nature 8 services: intangibility of 1, 27; quality of
multiple proxies 6 1; quantity of 1
size of the firm 7
national income approach 32, 35, 46, SNA 51, 53 – 54, 57, 113, 123, 125,
53 – 54, 113, 164 – 165, 312 171 – 175, 185, 201 – 202, 207 – 210,
nominal vs. real outputs 183 – 184 212 – 215, 221n1
NPA 10n5, 39 – 40, 43, 176, 194 – 195 South Africa 242, 255 – 256, 259, 273,
299 – 300, 305
off-balance-sheet see off-balance-sheet stochastic frontier 183, 234, 249
output survivor bias 23
off-balance-sheet output 29, 34, 51 – 53, 56
output mix 129 technical efficiency 29, 231 – 233,
output specification: approaches 30, 111; 237 – 238, 248
bank 33 technical reserves 121, 123, 125
output specification problem 165 theory: of firm 1, 10n2, 180; of output
measurement 3 – 4, 162, 165n1; of
performance analysis 3, 16 – 17, 128, 218 production 1, 10n3
premium output 114 – 116, 287 – 289 trans-log 60, 234, 240, 266, 285
price index 7 – 8
production approach: banking literature unweighted aggregation 52
32 – 34, 46; insurance literature 111 – 112 user cost approach 37 – 38, 46, 92n34,
production function: endogenous 5; 178, 184
interlocking 1, 10n4; overlapping 1,
10n4, 27 value-added approach: banking literature
productivity 17, 30, 35, 61; partial factor 33, 37, 46; inputs measurement 309;
235 – 237, 243; total factor 187, 217, insurance literature 111 – 113
237, 248
product-mix representativeness 4 weighted aggregation 52