Applied Economics: A Critical Realist Approach (Routledge Frontiers of Political Economy)

Applied Economics and the Critical Realist Critique

Critical realism has been at the centre of much discussion within the social sciences. The approach is often critical of mainstream economics for its unrealistic assumptions and slavish dedication to mindless instrumentalism. This volume provides a real attempt to bridge the gap between mainstream applied economics and its critics. With refreshing open-mindedness, the contributors do not simply limit themselves to theoretical discussions, but in fact examine issues such as: ● ● ●

employment and unemployment critical realism and econometrics the political economy of monetary union.

With an impressive array of contributors including Philip Arestis, Sheila Dow and Malcolm Sawyer, Applied Economics and the Critical Realist Critique will greatly interest students and academics involved in economic methodology, philosophy and indeed mainstream economics. Paul Downward is Professor of Economics at Staffordshire University, UK. He is also author of The Economics of Professional Team Sports also published by Routledge.

Routledge INEM advances in economic methodology Series Edited by D.Wade Hands Professor of Economics, University of Puget Sound,Tacoma, USA

The field of economic methodology has expanded rapidly during the last few decades.This expansion has occurred in part because of changes within the discipline of economics, in part because of changes in the prevailing philosophical conception of scientific knowledge, and also because of various transformations within the wider society. Research in economic methodology now reflects not only developments in contemporary economic theory, the history of economic thought, and the philosophy of science; but it also reflects developments in science studies, historical epistemology, and social theorizing more generally. The field of economic methodology still includes the search for rules for the proper conduct of economic science, but it also covers a vast array of other subjects and accommodates a variety of different approaches to those subjects. The objective of this series is to provide a forum for the publication of significant works in the growing field economic methodology. Since the series defines methodology quite broadly, it will publish books on a wide range of different methodological subjects. The series is also open to a variety of different types of works: original research monographs, edited collections, as well as republication of significant earlier contributions to the methodological literature.The International Network for Economic Methodology (INEM) is proud to sponsor this important series of contributions to the methodological literature. 1. Foundations of Economic Method, 2nd Edition A Popperian Perspective Lawrence A. Boland 2. Applied Economics and the Critical Realist Critique Edited by Paul Downward

Applied Economics and the Critical Realist Critique

Edited by Paul Downward

First published 2003 by Routledge 11 New Fetter Lane, London EC4P 4EE Simultaneously published in the USA and Canada by Routledge 29 West 35th Street, New York, NY 10001 Routledge is an imprint of the Taylor & Francis Group This edition published in the Taylor & Francis e-Library, 2005. “To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.” © 2003 Selection and editorial matter, Paul Downward; individual chapters, their authors 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. 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 A catalog record for this book has been requested ISBN 0-203-98787-X Master e-book ISBN

ISBN 0–415–26785–4 (Print Edition)

Contents

List of illustrations List of contributors Acknowledgements

viii ix xi

PART I

Overview 1

Introduction

1 3

PAU L D OW N WA R D

2

Critical realism and economics

12

S H E I L A C . D OW

3

Conceptualizing unemployment in a period of atypical employment: a critical realist approach

27

S T E V E F L E E T WO O D

4

Critical realism, methodology and applied economics

51

B E R N A R D WA LT E R S A N D DAV I D YO U N G

PART II

Critical realism, formalism and empirical work: some conceptual issues

67

PART IIA

5

On closure

69

Critical realism and formal modelling: incompatible bedfellows?

71

MARK SETTERFIELD

vi

Contents

6 Seeking a role for empirical analysis in critical realist explanation

89

PAU L D OW N WA R D, J O H N H . F I N C H A N D J O H N R A M S AY

PART IIB

On econometric methods 7 Critical realism and econometrics: interaction between philosophy and Post Keynesian practice

109

111

PAU L D OW N WA R D A N D A N D R E W M E A R M A N

8 A pragmatic alliance between critical realism and simple non-parametric statistical techniques

129

J O H N H . F I N C H A N D RO B E RT M C M A S T E R

PART IIC

Objects, data and theory 9 Triangulation, time and the social objects of econometrics

151

153

W E N DY O L S E N

10 Theory foundation and the methodological foundations of Post Keynesian economics

170

F R E D E R I C S. L E E

PART III

Critical realism in empirical analysis

195

11 Questionnaires in realist research: a case study of rural small-scale industry in Ghana

197

W E N DY O L S E N , T H A N D I E M . H A R A A N D SAMPSON EDUSAH

12 Critical realism and applied work in economic history: some methodological implications BRIAN PINKSTONE

220

Contents vii 13

Critical realism and the political economy of the Euro

233

P H I L I P A R E S T I S , A N D R E W B ROW N A N D M A L C O L M S AW Y E R

14

Presenting demi-regularities: the case of Post Keynesian pricing

247

PAU L D OW N WA R D A N D A N D R E W M E A R M A N

15

From predictive to indicative statistics: explaining corporate borrowing

266

M U R R AY G L I C K M A N

16

Transition in Eastern Europe: critical realism and qualitative insights

279

KARL PETRICK

17

Conclusion

293

PAU L D OW N WA R D

Index

302

Illustrations

Figures 9.1 10.1 11.1 13.1 15.1 17.1

Illustration of three ontic levels in wage determination research Schema of the grounded theory method Patterns of women’s financial expenditure Daily exchange rates: US dollars per Euro Corporate borrowing and GDP in the UK Applied critical realism: a schema

154 175 206 239 270 298

Diagrams 11.1 11.2

Factors associated with employment in women’s RSSIs Factors associated with employment in men’s RSSIs

209 210

Tables 3.1 Empirical realist or ‘flat’ ontology 11.1 Variables and their means 11.2 Potential contribution of additional techniques to RSSI study 11.3 Areas for improvement over single-technique questionnaire method 14.1 A selective summary of econometric studies of UK pricing 14.2 Firms’ pricing responses 14.3 Explaining firms’ pricing responses

30 212 214 215 255 259 260

Boxes 9.1 9.2 9.3

Three routes toward a realist practice in economics Distancing in social science report-writing Realist empirical assumptions

156 158 162

Contributors

Philip Arestis is Professor of Economics at the Jerome Levy Institute, USA. Andrew Brown is Lecturer in Economics at the University of Leeds, UK. Sheila C. Dow is Professor of Economics in the Stirling University, UK. Paul Downward is Professor of Economics in the Staffordshire University, UK. Sampson Edusah is PhD Candidate at the University of Bradford, UK. John H. Finch is Senior Lecturer in Economics at the University of Aberdeen, UK. Steve Fleetwood is Senior Lecturer in Behaviour in Organisations at the University of Lancaster, UK. Murray Glickman is Principal Lecturer in Economics at the University of East London, UK. Thandie M. Hara is PhD Candidate at the University of Bradford, UK. Frederic S. Lee is Professor of Economics at the University of Missouri at Kansas City, USA. Robert McMaster is Senior Lecturer in Economics at the University of Aberdeen, UK. Andrew Mearman is Assistant Professor in Economics in the Wagner College, USA. Wendy Olsen is Lecturer in Socio-Economic Research in the Cathie Marsh Centre for Census and Survey Research at the University of Manchester, UK. Karl Petrick is Senior Lecturer in Economics in the Leeds Metropolitan University, UK. Brian Pinkstone is Associate Professor in Economics and Finance at the University of Western Sydney, Australia.

x

Contributors

John Ramsay is Reader in the Staffordshire University Business School, UK. Malcolm Sawyer is Professor of Economics at the University of Leeds, UK. Mark Setterfield is Associate Professor in Economics in Trinity College, USA. Bernard Walters is Senior Lecturer in Economics at the University of Manchester, UK. David Young is Senior Lecturer in Economics at the University of Manchester, UK.

Acknowledgements

While having the responsibility of putting together this book, it goes without saying that its existence is entirely due to the support of all of the contributors whom I would like to thank sincerely for devoting the time and effort involved. In a world of increasing pressure on non-mainstream economics, and a huge institutional research drive towards research papers rather than books there are few incentives to be involved in such projects other than the genuinely academic aspirations of critical discussion and intellectual endeavour. This said, I would particularly like to thank the following without whose help this book would not have surfaced. First,Tony Lawson for his enthusiasm and encouragement for this project and his commitment to open debate. Second, Rob Langham and his team at Routledge for their support and patience. I am sure that getting academics to meet deadlines must be like trying to herd cats. Frustrating in the extreme! Finally, I am very grateful for long-standing discussions and collaborations with Fred Lee, Andrew Mearman and Wendy Olsen. With one exception, Chapter 3, each chapter was written specifically for this book. In this case thanks go to Routledge for permission to essentially reprint Chapter 3. Elsewhere, where appropriate, as each author reflects and draws upon their other published work, care has been taken to ensure that full references are given to allow the reader to follow-up the development of thought distilled in this book. In particular, however, thanks are due to the editors and publishers of the following papers for permission to use revised and extended elements of previously published work: Downward, P.M. and Mearman, A. (2002) Critical realism and econometrics: constructive dialogue with Post Keynesian economics, Metroeconomica, 53(4): 391–415. Glickman, M. (2000–01) Seeking explanations not preserving theorems: corporate finance, confidence, asymmetric information and the UK macroeconomy, Journal of Post Keynesian Economics, 23(2): 201–234.

xii Acknowledgements Downward, P.M., Finch, J. and Ramsay, J. (2002) Critical realism, empirical methods and inference: a critical discussion, Cambridge Journal of Economics, 26(4): 481–500. Walters, B. and Young, D. (2001) Critical realism as a basis for economic methodology: a critique, Review of Political Economy, 13(4): 483–501.

Part I

Overview

1

Introduction Paul Downward

Introduction Critical realism has broad origins and has been at the centre of much discussion within the social sciences (Harre and Madden, 1975; Keat and Urry, 1975; Bhaskar, 1978, 1989; Outhwaite, 1987).This discussion also applies to economics that has, particularly through the efforts of Tony Lawson, developed into a coherent and sustained research programme. The Cambridge Workshop on Realism and Economics organised by Tony Lawson and Colleagues at the University of Cambridge, and running since October 1990, has played a central role in this regard.The success of the intiative is also manifest in Tony Lawson’s Economics as Social Theory series published by Routledge.Within this series Tony Lawson’s volumes Economics and Reality (1997) and Reorienting Economics (2003), can be seen as a synthesis of many of the ideas discussed at the Workshop and, indeed, developed through Tony’s seminal constributions. Another key publication in the Economics as Social Theory series by Routledge would include Critical Realism in Economics edited by Steve Fleetwood and published in 1998. Of course, there are many other contributions to the critical realist literature, however, their scope is much broader than economics. Extending debate in the context of economics is the main motivation for this book. Its particular theme is how to engage in empirical work in economics while embracing critical realism. A brief outline of critical realism follows before a more detailed discussion of the scope of the book.

Critical realism In his book, Economics and Reality, Lawson offers a critique of mainstream economics drawing upon and extending Bhaskar’s transcendental realism and, by implication, making a strong plea for the explicit discussion of ontology in economics. Economics and Reality takes as its starting point a set of inconsistencies with mainstream economic analysis.These reside at the levels of method, social theory and methodology. In the former case Lawson notes that gaps exist between the theory and practice of economics. While both economic theory formulation and the logic of inference draw upon the precision of mathematics,

4

Paul Downward

practical testing of propositions with econometrics appears to be proceeded in an ad hoc manner. In the case of social theory, the power of mathematics to elucidate economic choices is championed and, yet, paradoxically, the determinism implied belies choice. Finally, economists are inconsistent in essentially ignoring the need to formally engage in methodological discussion as part of research projects or training yet noting the inability of econometrics to produce robust theory choices. These are, of course, not inconsistencies uniquely identified by Lawson. However, what Lawson does propose is that they can be both understood and resolved by adopting a realist orientation. Part and parcel of this argument is, thus, to redefine and rework these problems in realist terms. In brief the problems with mainstream economics are identified as following from the central role played by a deductivist mode of explanation. This embraces ‘covering’ laws wherein constant conjunctions of events are postulated such that ‘whenever event x then event y’ (Lawson, 1997, p. 17). Prediction and explanation are logically equivalent according to this approach. These strict constant conjunctions are central to the axioms, assumptions and, ultimately, predictions of mainstream economic analysis. Further this mode of explanation embraces a closed-system ontology. Closure implies that causes produce the same effects and effects can always be understood, uniquely, in terms of the same causes. The empirical relevance of this approach is characterised by Lawson as empirical realism because the approach entails a determinate account of phenomena developed by deductive methods that can be fully described empirically at the level of events. In particular, mathematical theory is typically employed to generate predictions, which are then tested by econometric methods. In this way a form of positivism, that Lawson links to Hume, becomes enmeshed in orthodox economic analysis wherein sense experience and knowledge become conflated. Critical realists generally accept, therefore, that this approach is underpinned by an epistemic fallacy …by the supposition that statements about being can always be rephrased as statements about knowledge (of being), that ontology can be reduced to epistemology. With ontology unavoidable, this error functions only to cover the generation of an implicit ontology, in which the real is collapsed onto the actual which is then … identified with … human experience, measurement or some other human attribute. (Lawson, 1997, p. 62, parentheses in original) In contrast, critical realism embraces an open-system view that human agency is embedded in an organic social context. The real causes of phenomena will thus lie in complex codetermination both ontologically distinct from events and our empirical experience of them. Access to real causes must proceed through a different logic of inference than either the deductive or inductive emphases implied above. Rather, than deducing specific events from general postulates, or generalising events from particular examples, critical realism

Introduction 5 embraces ‘retroduction’. This implies moving, ontologically, from the level of events to the real causes of them. The upshot is that the empirical analysis of causes is problematic and, that in general, mainstream inferences will be misleading. Particular concern has been expressed about econometric inferences. Nonetheless, much of the critical realist literature is primarily concerned with philosophical discussion and its relevance to broad approaches to economics.What is less well explored is the implications of critical realism for more detailed epistemological issues and, in particular, the operation of retroduction and relatedly the role of different methods of empirical work. In this respect ‘econometrics’ has largely been defined in terms of, or conflated with, the use of particular statistical methods by mainstream economists. In contrast, Political Economy broadly defined is concerned with adequately capturing real elements of economic processes, and working towards concrete analysis and policy prescription. Consequently, this book not only aims to explore and articulate both the problems associated with this endeavour, but also provide some basis upon which such applied work can proceed by drawing upon contributors working within strands of Political Economy.

A cautionary note At this juncture two important caveats are worth noting before discussing the outline and structure of this book. The first is that the title may give the impression that the book carries a strong prescriptive intent and, or the impression that it presents a simple set of guidelines upon how to ‘apply critical realism’.This is simply not the case. In contrast this book explores and articulates how various scholars have attempted to both articulate critical realist empirical inferences, or specifically deliberated upon the role of empirical work in critical realist epistemology.The emphasis is very much upon a constructive dialogue between philosophy and practice.This dialogue is ongoing and evolving.This book captures a snapshot of research and ideas to the current date. Second, the dialogue essentially draws upon researchers working within Political Economy through the vehicles of the Workshop on Realism and Economics,The Post Keynesian Economics Study Group and the Association of Heterodox Economics. In part, this is because of the interests of the editor but, more importantly, it reflects the fact that the latter organisations have welcomed discussion on critical realism, generally, and critical realism and empirical methods, specifically.1 This said, it is sincerely hoped that this volume will be of interest to all of those interested in exploring critical realism. For established academics and postgraduates alike the intention is to provoke thought and further discussion upon how economic research could proceed, through reference to how others are addressing this issue. For postgraduates in particular, working to tight time constraints and yet interested in the philosophical importance of critical realism, this latter aspiration might be of particular value.

6

Paul Downward

Outline of chapters The discussion that follows is organised into three main parts. Part I provides an overview of both the origin and nature of critical realism as well as the challenges that critical realism faces in providing an effective epistemology, and the constraints that it poses for effective empirical enquiry. In Chapter 2, Sheila Dow explores the contribution which critical realism makes to our understanding of economics, as well as the reality of the economy. Beginning with a review of the debate over the realism of assumptions and the subsequent treatment of the issue by the ascendant New Classical Economics in the 1980s and the constructivist approach (found in the postmodern, rhetoric and sociology of scientific knowledge approaches), critical realism is given context.The critical realist approach is then discussed in terms of methodology: its use of the concepts of open and closed systems, fallibilism, and the retroductive method.The chapter concludes with a discussion of the implications of critical realism for economics. Chapters 3 and 4 then set out the challenges facing critical realist empirical researchers from the perspectives of both within and outside critical realism. In the former case, in Chapter 3, Steve Fleetwood uses discussion of the relationship between unemployment and employment to show that designing theoretical concepts that adequately express reality, rest (minimally) upon the adoption of appropriate ontological, philosophical and methodological foundations. Fleetwood uses critical realism to argue that this is not the case in mainstream economics. The deductive method of mainstream economics means that theoretical concepts have to be constructed in such a way as to reduce the multi-dimensional, qualitative reality of employment and unemployment to the quantitative, single dimension of variables, whereupon they cease to be adequate expressions of the reality they are designed to investigate. A radical rethink of the interpretation of contemporary unemployment figures is thus suggested, as well as, of course, mainstream empirical methods. In contrast, and perhaps more fundamentally, Bernard Walters and David Young attack what they perceive to be major problems with critical realism in terms of its ontological critique of mainstream economics as well as its undeveloped epistemology. They begin by arguing that critical realism works with a tendentious definition of positivism that belies the fact that it simply commits the researcher to concern with establishing truth from evidence and not necessarily covering laws. Walters and Young then argue that mainstream economics is not positivist per se but embraces a diversity of positions ranging from the ‘cartesian’ approaches that treat theory as inviolate and calibrate empirical findings to fit to theory, to instrumentalists not concerned with truth to realists.To an extent echoing Dow, they argue that Hume is dismissed too readily by critical realists. Their argument is that Hume was sceptical about establishing causes from constant conjunctions. Finally, they argue that critical realism lacks robust rules for uncovering

Introduction 7 causes and deep structures and postulate that some form of empiricism is needed. However, The problem for critical realists is that …[critical realism]… accepts at least some form of empiricism without addressing the issue of how to use empirical evidence. (pp. 56–57 below, parentheses added) The upshot is that critical realism needs to be more circumspect in its ontological boldness as much as its epistemological caution, at least in attacking mainstream economics. It is properly conceived as a critique of a covering law approach to explanation, yet lacks a clear epistemology. The points made by Walters and Young are serious and require serious deliberation. Indeed it is to debating the role of empirical work that the remainder of the book turns in Parts II and III. Part II addresses some conceptual issues associated with formalism and empirical work in critical realism. Chapters 5 and 6 begin with a discussion of closure. In Chapter 5, Mark Setterfield investigates the extent to which formal modelling is, or can be made, consistent with the ontology and methodology of critical realism. He argues that formal modelling can be separated out from the practices of mainstream economists. This makes it possible to develop a non-mainstream approach to formal modelling that yields tractable formal models of economic processes rendering formal modelling broadly congruent with critical realism under the guise of an open systems-ceteris paribus approach to formal model building that makes use of a concept of conditional closure. Chapter 6, in many respects, presents a similar argument yet starts from an empirical rather than theoretical perspective. Paul Downward, John Finch and John Ramsay attempt to set out some guidelines for making empirical inferences, starting from the critical realist ontological understanding that economic phenomena are characterised by open and complex social systems.They argue that crucial to their investigation, and the legitimacy of using empirical methods in critical realism, is to accept the presence of quasi-closure in social systems. This exists and is created in specific contexts by agents who require a degree of stability within their decision-making. Drawing upon Grounded theory, Keynes as well as critical realism they discuss how inference in critical realism is likely to involve social discourse in establishing acceptable interpretations as well as the simple triangulation of qualitative and quantitative insights. The next two chapters extend the conceptual discussion of empirical methods and critical realism with discussions of econometrics. In Chapter 7, Paul Downward and Andrew Mearman argue that econometric practice can play a role in analysis framed within a critical realist perspective. They argue that a degree of tension exists between the philosophical pronouncements of critical realism and the consequences of referring to empirical concepts in the practice of inference, which is something that critical realists wish to do. Thus, while critical realism has epistemological implications for the practice of

8

Paul Downward

economics, there is an iterative relationship between philosophy and practice. The chapter carefully reviews econometric methods to juxtapose against Lawson’s (1997) redefinition and critique of econometrics. Downward and Mearman then show that Lawson’s redefinition suffers from the same problems, thus they offers some proposals for empirical inference that shares the broad concerns of critical realism, yet embraces econometric work. In Chapter 8, John Finch and Robert McMaster adopt a more cautious approach and instead of emphasising econometrics as regression they argue that researchers in economics who wish to develop and assess the reliability and adequacy of knowledge claims may draw upon simple non-parametric statistical analyses such that theoretical development may be informed by empirical analysis and vice versa. They thus propose a pragmatic alliance between nonparametric techniques and, essentially, an open-system ontology, in interpreting and analysing stylised facts or demi-regularities across classes or categories of events. Simple non-parametric techniques offer the prospect of a loose, incremental and iterative coupling in undertaking theoretical development and also assessing the practical adequacy and reliability of knowledge claims drawn from such development. Chapters 9 and 10 conclude Part II with further discussion of the nature of inference within critical realism. In Chapter 9, Wendy Olsen argues that Econometrics tends to assume that the objects it describes exist. Over time this assumes, ontologically, an existence that is unchanging in the relevant characteristics. Wendy, explores the problems associated with assuming that things exist over time by exploring five main topics associated with time-series data. These topics include how time-series data should be augmented as part of a triangulated economic research strategy. In Chapter 10, Fred Lee explores the role of grounded theory in more detail. By drawing upon the Post Keynesian literature, Fred argues that the method of grounded theory provides the best set of guidelines for theory creation in critical realism because it allows for an elaboration of causal mechanisms.Yet, Fred argues that the historical economic theory that results from integrating realism, critical realism, and the method of grounded theory will result in critical realism and grounded theory being interactively modified. Realism requires real entities and grounded theory operates on the principle of empirically grounding real entities or categories; hence the structural and causal mechanisms of critical realism in the context of Post Keynesian economics are observable and real as opposed to being unobservable and imaginary. Reciprocally, critical realism establishes the objectives of the Post Keynesians using grounded theory, that is the grounding of structures and causal mechanisms. Moreover, since Post Keynesians see human agency in all causal processes and historical narratives require human agency, the causal mechanisms embedded in their theories must also contain human agency. Thus, the historical economic theories of Post Keynesian economics are a result of a particular concatenated integration of common sense propositions, realism and critical realism, and grounded theory.

Introduction 9 Both Wendy Olsen’s and Fred Lee’s chapters, by referring to their actual research provide a useful link with the remaining chapters in Part III. In this part the chapters shift in emphasis to providing examples of critical realism in empirical economic analysis. In Chapter 11, Wendy Olsen, Thandie M. Hara and Sampson Edusah provide an example of empirical work based on fieldwork in Ghana.The chapter reviews realist approaches to the use of the survey method and discusses regression analysis conducted upon the cross-sectional data. In Chapter 12, Brian Pinkstone examines the role of econometric methods in economic history. He argues that the central role played by broad socio-economic contextual factors for the explanation of historical events and the complimentary belief that good research required a sound acquaintance with the widest possible range of primary sources has been replaced. Instead the use of statistical methods and hard quantitative data combined with economic and/or econometric models to provide precise quantitative apportionment of causal significance to independent variables is common. Drawing upon the critique of econometrics in the critical realist literature, Brian uses examples from his own research in the field of economic history to illustrate an alternative approach. In Chapter 13, Philip Arestis, Andrew Brown and Malcolm Sawyer discuss some of the main problems that critical realism raises for empirical work and inferences with reference to a research project on the analysis of the single European currency. The authors argue that critical realism provides a method that can help understanding of concrete levels of analysis – illustrated by the example of the attempt to explain the falling value of the Euro. However, they argue that this analysis is partial and inappropriate to the most abstract and fundamental levels of theory. These ideas are illustrated with reference to the fundamental propositions of Keynes, Kalecki and Marx. In Chapter 14, Paul Downward and Andrew Mearman illustrate their arguments of Chapter 7, by drawing upon applied work in Post Keynesian pricing. A historical case is offered that is suggestive of the merits of econometrics and an exemplar of the triangulation strategy that they advocate, referred to above. More contemporary material is then referred to by drawing upon both a broader econometric literature and from uniting different methods of empirical analysis in a specific context. Retaining the Post Keynesian theme, in Chapter 15, Murray Glickman presents an example of Post Keynesian empirical work based on a broadly critical realist understanding of the nature of economics. Murray, rejects the stochastic paradigm of econometrics entirely in an empirical piece of work supporting his earlier, published, theoretical critique of the Modigliani–Miller (MM) propositions on capital structure. Murray’s work is based on a ‘facts first’, explanation-oriented, approach in contrast to the ‘facts last’, deductive/predictive approach that dominates mainstream economic analysis. In Chapter 16, Karl Petrick finishes the part by providing an illustration of how qualitative research methods, embedded in a critical realist perspective, can yield insights into the process of transition. An example from the Czech Republic is presented. The chapter critically reviews the central elements of

10

Paul Downward

the critical realist perspective, contrasting these with mainstream economic theory which is then contrasted with empirical economic research. It is argued that while mainstream theory may, in some sense, never be compatible with empirical investigation, critical realism offers a constructive logic within which to bridge the gap between empirical investigation and theory, generally. Karl argues that one of the criticisms of the critical realist literature is that it offers very little in terms of advice for how to conduct research. One way to solve this problem is to examine research methods texts and evaluate their advice regarding methodology using the tenets of critical realism as criteria.The chapter suggests that this will not lead to a single critical realist research method, but it is possible to identify what critical realist research should entail, while leaving the exact research methods chosen to be determined by this question – what methods are appropriate for the research question posed? Karl’s research project into banking in Eastern European transition is then provided as an example of how this interpretation of critical realist empirical research can be put into practice showing how knowledge of causal mechanisms emerges at the same time as empirical observation.With hindsight Karl identifies that the approach adopted was consistent with a grounded theory approach to social science research.

Some final comments To conclude it is worth reiterating that this book was deliberately conceived and constructed as an attempt to draw together different scholars’ perspectives on critical realism and empirical methods. There was no strong steer by the editor to force contributors into a template that circumscribed their perspective. Each chapter, noted above, will thus contain some statements about individual author’s perspectives on the issues raised by critical realism for either empirical work generally or their specific research projects. In what is hoped reflects an open-system perspective, the aim was very much to encourage a project within which creative tension could be generated. As well as an overview of the chapters, the final chapter of the book contains a personal reflection from the editor on points of commonality contained within the volume and which generate issues for further deliberation. It is very much hoped that the book does stimulate this discussion. Moreover, it is hoped that the book provides constructive material for the deliberations of economists and other social researchers who are grappling with the need for sensible empirical analysis to underpin theoretical development and policy discussion and yet feel frustrated by the stranglehold of mainstream (economic) epistemology.

Note 1 It should be noted that these are broad churches of Political Economic thought and welcome dialogue between economists working within all Schools of Thought. For example, there has been much discussion of the links between critical realism and Post Keynesian economics (see the Symposium Journal of Post Keynesian Economics,

Introduction 11 22(1), 1999).The annual conference of the AHE, for example, is organised thematically to promote dialogue between a plurality of schools and has explicitly debated critical realism and empirical methods. Moreover, the AHE has currently run two Advanced Post Graduate Research Methods workshops sponsored by the ESRC, within which critical realism has played a prominent role. Indeed the editor’s motivation for the volume began with participation in a session on empirical methods at a Post Keynesian conference organised by Sheila Dow and John Hillard. This presentation was of ideas that were employed in the editor’s PhD applied work which was later refined and published as Downward, P.M. (1999) Pricing Theory in Post Keynesian Economics:A Realist Approach, Aldershot: Edward Elgar. It was here that the need for a clear and methodologically consistent approach to empirical inferences first became apparent to the editor and which has been of interest and exploration since then.This volume thus provides an opportunity to draw together the ideas and thoughts of other researchers who have symbiotially developed similar interests.

References Bhaskar, R. (1978) A Realist Theory of Science (Brighton, Harvester Press). —— (1989) Reclaiming Reality (London,Verso Press). Harré, R. and Madden, E.H. (1975) Causal Powers: A Theory of Natural Necessity (Oxford, Blackwell). Keat, R. and Urry, J. (1975) Social Theory as Science (London: Routledge and Paul). Outhwaite,W. (1987) New Philosophies of Social Science: Realism, Hermeneutics and Critical Theory (London, Macmillan).

2

Critical realism and economics Sheila C. Dow

Introduction: economics and reality To many it might seem obvious that economics is concerned with reality. Its Greek roots are found in the management of the household.The modern day subject matter is more aggregated, but still refers to real actions: buying, selling, employing, working, using up and so on. If it is not concerned with reality, what can it be concerned with? Most economists would, if questioned,1 agree that our subject matter is real in some sense, although we shall see that there is some doubt as to how we pin down reality. Indeed, once we start studying any real subject matter, collecting evidence and constructing theories, we are putting a distance between ourselves as economists and reality. This is inevitable. It is the purpose of gathering evidence and of constructing theory to identify patterns in reality which allow us to make sense of it and, possibly, change it. In that sense all study of real subject matter itself is ‘unrealistic’; it quite deliberately differs from an exact representation of reality. But then what do we make of the relationship between theory and reality? Particularly if the ultimate purpose is to propose policies to change reality, and particularly if the change involves transformation, we need to understand that reality and how change might impact upon it. But we also need to be sure we understand the relationship between economic theory and economic reality. The critical realist critique of mainstream economics has brought these important issues to the surface of debate. The purpose of this chapter is to explore the contribution which critical realism makes to our understanding of economics, as well as our understanding of the economy. We start by placing critical realism in context.The issue of the relationship between economics and reality is a long-standing one. In the next section we sketch a brief account of how the issue has been addressed in the past.We focus on the most recent prior debate in the 1960s and 1970s over the realism of assumptions and the subsequent treatment of the issue by the ascendant New Classical Economics in the 1980s. We devote a separate section to the emergence of a very different approach in the 1980s and 1990s which raised much more fundamental questions about our access to reality and about the accounts we construct

Critical realism and economics 13 about it: the constructivist approach (found in the postmodern, rhetoric and sociology of scientific knowledge approaches). Against this background, we can then set out the critical realist approach. Critical realist methodology is derived from a philosophical approach called transcendental realism. We discuss this, and other possible philosophical foundations in the following section. We then explore in some detail what the critical realist approach consists of in terms of methodology: its use of the concepts of open and closed systems, fallibilism and the abductive method. We conclude with a discussion of the implications of critical realism for economics.

The realism of assumptions debate The issue of realism was brought back to the surface of debate in economics, fifty years ago, when Friedman (1953) challenged the idea that there needs to be a direct correspondence between theory and reality. He argued that the connection with reality should only occur at the point at which predictions from a theory are compared with actual outcomes.The assumptions of the theory itself need not correspond to reality, indeed he argued that more unrealistic assumptions are likely to yield better predictions. Theories were simply instruments for predicting the values of variables. Friedman himself did not practice what he preached – he proceeded to justify his theories in terms of their structure and assumptions. But he had brought to the surface an issue which had slid into the background, namely the realism of economic theory.2 This had been an important issue over the history of economics, but one which had subsided with the burgeoning empirical activity which followed the emergence of modern macroeconomics and the construction, for example, by the NBER in the US, of large datasets. The debate sparked off by Friedman had several outcomes which are relevant to the subsequent development of critical realism. One outcome was a closer specification of the different senses in which assumptions might be unrealistic (see Nagel, 1963), which suggested that some types of unrealism were more problematic than others. Nagel argued that assumptions could be unrealistic in three senses: falsehood, fiction and simplification.The issue then is whether any or all of these senses of unrealism lead, as Friedman argued, to better prediction. Critical realism takes a particular stand against fictional assumptions in that they preclude any hope of identifying causal mechanisms. Another outcome was a reflection on the principles which should govern theory construction, if they go beyond predictive success. Samuelson (1963) challenged Friedman’s advocacy of unrealistic assumptions, arguing that the goal of theory, rather than predictive success, should be descriptive content, or empirical validity. The difficulty, given that theories inevitably involve some unrealism, lay in establishing operational criteria for empirical validity. Samuelson advocated what is known as ‘conventionalism’ (see Boland, 1982,

14 Sheila C. Dow chapter 9), namely recourse to what are conventionally accepted as criteria for choosing theories. The result was a rather unsatisfactory falling-back on convention as a guide to theory construction. But, particularly given the popularity with economists of Popper and Lakatos’s methodology, the ultimate criterion remained empirical.3 The way to decide whether a particular theoretical development was a good one or not was by empirical testing. The third outcome was a more general questioning of the purpose of theory, whether it is addressed more to prediction or to explanation. There was an understanding that there might be a trade-off between the two. Some theories are more suited to one than the other; for example, the theory of gravity is better for prediction than explanation, while the reverse is true of the theory of evolution. The trade-off is not complete, however – if there was a correspondence between the structure of a theory and reality, then there was some chance of working out why a theory did not predict successfully. Further, it was possible to adapt theory to a change in economic structure (such as a change in exchange rate regime) if there was some correspondence with reality. If, however, theory were treated as a black box, there was no mechanism for taking account of structural change. Understanding may therefore provide a better basis for prediction. New Classical Economics attempted to cut through the problems raised by Friedman’s instrumentalism by conflating theory with reality.4 The criterion for theory choice was still predictive success, not only for economists, but also for economic agents who were depicted as employing exactly the same models. Agents were not understood in reality as knowing econometric models, but behaving as if they knew them.This ‘as if ’ argument had been a common one to justify unrealistic assumptions (like entrepreneurs equating Marginal Cost (MC) and Marginal Revenue (MR)) – an example of a fictional assumption. But the difference with New Classical Economics was that, since rational behaviour (defined in the specific way of the axioms of orthodox microeconomics) meant that no opportunity was ever left unexploited except by choice, the economy and all its components were always in equilibrium.This argument meant a much tighter correspondence between theory and reality than in neo-classical theory, where it arises only at the end of a disequilibrium process.The problem with the latter view was knowing whether observations referred to equilibrium or disequilibrium (for which there was little theoretical equipment). Whether the choice was Friedman’s instrumentalism, New Classical instrumentalism or Samuelson’s conventionalism, the tenor of the debate was that there was one best set of procedures for economists; it was just a matter of settling the argument as to which set. All agreed that the ultimate criterion for choosing a theory was empirical success (understood in econometric terms). This provided the ultimate touchstone of reality. But there was a general change in intellectual and social climate in the 1980s, which challenged, not only the idea that there was one best procedure for science, but also that there was one reality by which to assess it. This encouraged what is known as the constructivist approach to economics.

Critical realism and economics 15 Constructivism Constructivism grew out of the philosophy of science of Kuhn (1962). Like Popper, he focused on the significance of scientific community. He offered a descriptive account of science which was ‘realistic’ in setting out what scientists actually do, rather than prescribing what they ought to do as Popper had done. He addressed the observed fact that scientists do not in fact abide by the empirical criterion, but rather continue according to criteria that is internal to that particular scientific community. Reality breaks through when a disparity between theory and reality can no longer be sustained, and when there is an alternative approach (what he called a paradigm) which appears to fit better with reality. A classic example is the Great Depression in the 1930s which could no longer be ignored in spite of the inability of economic theory at the time to account for it. The conventionalism discussed above fits well into this framework. But Kuhn demonstrates that it is possible to have a range of paradigms with their own sets of conventions, and there is no ultimate set of criteria by which to decide which is best. The empirical criterion is no use in any absolute sense. What constitutes a good empirical test may differ from one paradigm to another. But more fundamentally, the view of what constitutes reality may differ. Constructivism questions the whole notion of ‘facts’. Rather, all we have are subjective perceptions which differ from person to person, and over time. Indeed Kuhn’s starting point was the different ways in which language is used, effectively making communication – about ‘facts’ or anything else – difficult between paradigms. What then transpired was a dualistic alternation between the view that there is (in principle at least) one best way of going about economics, and an ultimate arbiter in the facts, on the one hand, and the view that there was a multitude of ways of going about economics, and no objective means of choosing between them, on the other hand. In practice economists continued to develop and test theories. Oddly enough, constructivism was liberating in that orthodox conventions apparently no longer required justification. But the result was an absence of methodological awareness, and an unwillingness to discuss the principles by which economics was developing (see Lawson, 1994a). At the same time, the particular conventions for orthodox theorising were being opened increasingly to question. The constructivist approaches of postmodernism and rhetoric played an important part in this questioning, but seemed unable to offer any alternative guidance; indeed to have done so would have gone against the conclusion that no guidance could be justified. It was in this context that critical realist ideas emerged.

Critical realism: philosophical foundations Critical realism has struck a chord with many economists who are dissatisfied with mainstream economic methodology in practice. While much of the

16

Sheila C. Dow

support for critical realism therefore comes from those who approach it from the direction of practice, it is set out very explicitly as an approach which has been developed from the direction of philosophy. We consider its philosophical foundations in this section. Critical realism is founded on transcendental realism, as expounded by Bhaskar (1975). It is a philosophical argument as to what the real world must be like, given the nature of scientific activity. The crucial distinction is drawn between ontology, which is concerned with the nature of reality, and epistemology which is concerned with the nature of knowledge.The object of realist science is to build knowledge about the real world. But what can we say about the real world other than in terms of our knowledge of it? Bhaskar warns of confusing the two, that is, of committing the epistemic fallacy. If the real world is something other than our knowledge of it, then it exists at different levels, not all of which are directly accessible.Transcendental realism suggests that there are three levels: the actual, which we experience directly, the empirical, which is an attempt to measure the actual, and the real, to which we do not have direct access. It is at the level of the real that causal mechanisms operate; it is the purpose of realist science to uncover them. But we can only access the real at the empirical and actual levels; science focuses particularly on the empirical as a means of systematising knowledge of the actual. What we can tell about the real from the empirical level depends on the nature of the causal mechanisms at the level of the real.The crucial question is whether these causal mechanisms operate within a closed system or an open system. A closed system is one in which there is both extrinsic closure – no unaccounted for external forces – and intrinsic closure – no unaccounted for interrelationships between the parts of the system. Closed systems allow for law-like causal mechanisms, which in turn generate empirical regularities by which the laws may be identified. An open system, by contrast, is subject to outside influences which cannot be anticipated (even as being random, which requires some prior knowledge of their nature) and evolution and interaction within the system which also cannot be anticipated. Further, causal mechanisms take the form of powers, or tendencies, which may or may not be active at any one time, and which when active may operate simultaneously and in ways which may counteract each other. The key argument within transcendental realism is the observed fact that physical science proceeds by means of experimentation. An experiment consists of isolating variables from outside forces, that is, of fabricating external closure. Drawing conclusions from repeated experiments in turn presumes that internal closure is satisfied – that the causal mechanism identified in one experiment can be assumed to be repeated in all others. If the real world itself constituted a closed system, there would be no need to close it artificially in an experiment.The regularities would be present in observed values of variables. Therefore the real physical world must be open. Since it is widely acknowledged to be extremely difficult to construct experiments at all in the social world, then that justifies even more strongly the

Critical realism and economics 17 conclusion that the social world is an open system. And indeed there are powerful secondary arguments which support this philosophical conclusion: arguments which refer to the creativity of human behaviour, the exercise of human agency, the evolution of social institutions, which bear out an argument that the real social world must be open. In summary, then, transcendental realism first puts the spotlight on ontology as being prior to, rather than subsumed in, epistemology – the nature of reality is important for how we construct knowledge about it, and exists independent of the knowledge we construct about it. Second, the observed identification of science with abstraction and experimentation shows that there are forces from which science must abstract. Were the real world a closed system, this would not be necessary; it would be feasible for science to identify all causal mechanisms. While transcendental realism is the only philosophical route to critical realism identified in the critical realist literature (as exemplified by Lawson, 1997), there are other philosophical routes. In particular, the philosophy of the Scottish Enlightenment offers an alternative route (see Dow, 2002a). Hume had concluded that existence could not be demonstrated by reason alone; ontology could not be accessed purely by epistemology, understood as reason applied to observation. But, just as Bhaskar observes that science proceeds by experiment, Hume observed that science (and conduct more generally) proceeds on the basis of common sense belief in existence.6 The need for such a basis arises from the inaccessibility of the underlying causal mechanisms which generate what we observe and experience. In the terms we have been developing, it is because the world is an open system that we cannot hope to identify the underlying causal mechanisms and prove by reason and observation their existence. The way in which we then proceed to build up knowledge follows from the inaccessibility of the real. As far as critical realism is concerned, Hume’s philosophy takes us to the same starting-point as transcendental realism. No doubt there are other possible philosophical routes. But, as far as the implications for economic methodology are concerned, the important issues are the same: how to build knowledge about a real world which constitutes an open system.We turn in the next section to consider what critical realism implies for economic methodology.

Critical realist methodology Critical realists present their arguments as being in support of an approach to economics rather than a specific methodology, suggesting that the approach can support a range of methodologies (see, e.g. Lawson, 1994b).The approach is one which puts the focus on ontology first and epistemology second, with ontology determining epistemology. How we construct knowledge depends on the nature of the subject matter. The critical realist approach helps us to understand existing bodies of work in economics by analysing it in terms of ontology and epistemology, for

18

Sheila C. Dow

example, the work of Menger (Lawson, 1999). But the most powerful use of the approach is to offer a critique of mainstream economics (Lawson, 1997). The predominant approach to economics, as exemplified by the earlier debate over realism of assumptions, is to construct theories based on fictional assumptions about human choice, and set them in a context which implies a given (universal) institutional framework, in order to yield general propositions. In practice these propositions may not be tested empirically; rather their truth value is assessed in terms of the internal consistency of the theory. Realism then rests on the slender thread of the realism of the rationality axioms of orthodox microeconomic theory, which we assess by means of introspection. Where propositions are tested empirically, they are assessed in relation to a set of facts which is understood to be objective.Testing consists of some more or less elaborate identification of event regularities using econometric techniques, such that, for example, the rate of inflation correlates with the rate of growth of the money supply. The extent to which any model being tested reflects real causal mechanisms was the focus of the realism of assumptions debate. Friedman’s argument was that it was the predictive power of the model which was the best test of having captured law-like behaviour. Samuelson’s argument was instead that the theory itself should reflect law-like behaviour.7 But there was no questioning that identifying law-like behaviour was the object of the exercise and that empirical testing of some sort was the definitive basis for deciding on the best theory. The New Classical approach went further in suggesting that individual agents behave in exactly the same way – that they too are seeking covering laws by means of empirical assessment. There has always been an awareness that there may be disturbing forces which cause actual variables to deviate from their predicted values. But the goal of theory is seen as being to capture as much as possible in the theory, such that any remaining disturbances are stochastic.This implies knowledge of the disturbances such that they can be assumed to be random. The models of mainstream theory then are closed systems. Testing them by reference to what are seen as an objective set of facts implies that the real world is also seen as a closed system. But the transcendental realist argument is that, if theories abstract (as everyone agrees that mainstream theories do) then they must be abstracting from something which cannot be captured in the closed system. Attempts are made (e.g. by Becker, 1991) to extend the reach of economics as if to minimise what must be abstracted from. In particular, critical realists argue that the need to abstract follows from the openness of the social world. The reason for considering above the context in which critical realism arose was not one of purely historical interest. Certainly we need to understand the context in order to understand the reception which critical realism has encountered. But more important, critical realism itself cannot be understood without understanding that context. The instrumentalism/conventionalism debate in economics was conducted within an approach to economics which espoused some form of logical

Critical realism and economics 19 positivism, whereby theories are tested not just by their internal logic but ultimately by reference to ‘the facts’. Constructivism was a dualistic reaction to logical positivism, challenging the notions of singular logic and singular facts. Rather, logic and understanding of facts were seen as specific to particular paradigms or, in the extreme versions of constructivism, to particular individuals, who even themselves are fragmented. Logic and the facts are then constructs. Much of the constructivist critique of mainstream economics was similar to the critical realist critique, except for the crucial difference that there was a denial of any form of benchmark in the form of reality. Effectively the argument was that the social world is an open system, and that is why knowledge is socially constructed. There is a contradiction here in that appeal does seem to be made to the reality of ‘scientific’ activity, while the scope for such appeal is denied. Indeed, just as logical positivism has proved to be unsustainable as a basis for methodology (see Caldwell, 1982), so constructivism collapses ultimately under a contradiction (see Dow, 2001). The dualism which both approaches represent itself reflects closed-system thinking which is at odds with both an open-system ontology and an open-system epistemology. Critical realism then can be seen as the dialectical synthesis which emerges out of the thesis of logical positivism and the antithesis of constructivism. As part of a process in the development of thought, it carries forward some elements of what has gone before. From mainstream economics it takes the idea that empirical regularities tell us something, but unlike mainstream economics, these regularities are seen as the starting-point for theory rather than the end-point, as prediction. From constructivism, critical realism takes the understanding that there is no such thing as objective ‘facts’, only socially constructed knowledge of the world. But the benchmark of a reality independent of our knowledge of it gives critical realism more scope for being optimistic that useful knowledge can be built up. Because of its open-system ontology, whereby the real is not fully manifest in the empirical, critical realism aims to build up knowledge of the mechanisms at work at the level of the real, even though we do not have direct access to them. We can never be sure that our knowledge is true – theory is fallible – but we can construct arguments in favour of one theory over another. There are likely to be several, probably countervailing, tendencies at work at any one time.What we pick up at the empirical level may be very confusing, not allowing definitive predictions.The aim therefore is to build up sufficient knowledge about the different tendencies.The predictions can be made about the operation of the different tendencies, without being able to predict precise values of variables. Critical realist methodology then sets out guidelines as to how that knowledge is built up. These guidelines rest on the method of abduction, or retroduction from regularities identified in the economic system. While there is considerable debate about the role (if any) of econometrics in critical realist methodology, a matter discussed in some detail in the remainder of this volume, this is where the clearest indication of a role for this kind of empirical

20

Sheila C. Dow

work lies – descriptive statistics. An observed empirical regularity is a clue that there may be some tendency at work which has dominated other tendencies, at least for the period of observation. This instigates further investigation of a substantive sort. Further investigation draws on the second key element of critical realist methodology, pluralism of method. Since the reality which we are trying to understand is ultimately inaccessible, no one method can generate true knowledge of it. If the real social world is an open system, then it is subject to complex internal and external influences. In particular, such a system does not lend itself to representation by a singular, deductive axiomatic system, since such a system requires the axioms to be true in order for the conclusions to be true. Instead, critical realism espouses pluralism, involving recourse to a range of methods designed to build up knowledge in different ways, taking different starting points. A starting point involves some closure. Indeed theorising inevitably involves closure of some sort (see Chick and Dow, 2001; Loasby, 2002). Particularly where there is awareness of the openness of the economic system, there is a need to invoke closure in order to say anything at all other than that everything is complex, interconnected and unknowable. In the real social world, open-system uncertainty is contained by the formation of institutions (such as the firm) and conventional behaviour (such as posting prices); these are closures which allow the system to function. Similarly, economists need to parcel off pieces of the system for study in order to try to identify causal mechanisms which can be regarded as in some sense separable. The key difference between the closures of mainstream economics and the closures of critical realism is that the former are regarded as universal and fixed while the latter are regarded as partial and provisional. In mainstream economics, for example, the money supply may be taken as exogenous as a universal principle. In a critical realist analysis, the money supply may be taken as given for one particular piece of analysis, for a particular configuration of the banking system, while another investigates the mechanisms which generate the money supply for another configuration. Or both types of analysis may be applied to the same context, in an effort to build up knowledge from different starting-points. What appears as inconsistency from a closed-system standpoint may be perfectly consistent within open-system logic.This logic is akin to Keynes’s ordinary logic which, unlike classical logic, addresses (general) circumstances where the truth of axioms cannot be established beyond doubt.8 It is not that anything goes – the assumptions made for a particular analysis need to be justified in relation to reality. Assumptions in critical realist theory may be unrealistic, but only in the sense of abstracting from reality through simplification, not in the sense of being fictional. Theories are developed within the critical realist approach, on the basis of abduction from observed provisional regularities (what Lawson calls demiregs), where observation and theorising both rely on a range of methods.

Critical realism and economics 21 Theorising itself involves what Hume would refer to as the application of the imagination.The critical realist way of putting it is the devising of appropriate metaphors by which to capture a growing understanding of underlying causal mechanisms. Here too critical realism can be seen to be drawing elements from both mainstream methodology and constructivism in a new way. How useful the metaphors are can be judged from their application to new contexts. There is an underlying presumption that there are sufficient regularities in the social world to allow the building up of theory, however partial and provisional. Were there no regularities, knowledge would be impossible. But, as critical realists would argue, so would social life be impossible. If nothing else, the institutions and conventions which society evolves to allow social life to proceed provide us with some regularities on which to base our knowledge. These regularities are the antithesis of the event regularities, which mainstream economics focuses on, in that they are regarded by mainstream economists as impediments to the market forces, which otherwise generate event regularities. In summary, then, the critical realist approach is fallibilist – there is no presumption of access to truth – and advocates pluralism of method. Both follow from critical realism’s philosophical conclusion that the real social world is open. Theories are developed by employing metaphors to capture something of the causal mechanisms for which observed demi-regularities provide clues. Both theories and demi-regularities inevitably employ some form of closure, but closure which is both partial and provisional. The aim of building knowledge in this way is to understand the underlying causal mechanisms well enough to provide the basis for policy designed to transform society. Critical realism does not purport to advocate any one methodology – but rather to advocate an approach to choosing a particular methodology. Thus, for example, Lawson (1994b) argues that critical realism leaves methodological choice up to economists. In particular he draws back from any suggestion that critical realism can contribute to discussion about differences between schools of thought.We conclude by considering further what it is that critical realism implies for economics.

Implications for economics The critical realist literature has made a significant contribution to economics not least by the strength of the critique of mainstream economics. By setting out a critique at the philosophical as well as methodological levels, critical realism has added depth to discourse about the source of unease with mainstream economics which had arisen from practice. By putting the focus on ontology, critical realism exposes the problems with approaches to economics which refer purely to the level of epistemology or present contradictory positions with respect to ontology and epistemology. At the same time critical realism offers a route out of the blind alley created by constructivism.The idea that both logic and facts are themselves constructed in a fragmented way has been an important part of the critique of

22

Sheila C. Dow

the logical positivism of mainstream economics. Indeed this idea exposed the problematic nature of mainstream economics’s correspondence with reality. But if at the same time the inability to access reality is taken as the dualistic opposite of mainstream economics’s empirical access to reality, then there is a flip over from demonstrable truth, on the one hand, to nothing at all, on the other, as far as economic knowledge is concerned. Constructivist economists, in fact, do construct useful knowledge (with respect to reality) but, as with mainstream economics, this knowledge lacks consistent philosophical foundations. Critical realism instead offers the prospect of building knowledge about the real economy which is not demonstrably true, but for which reasoned arguments can be made.This is because the way in which knowledge is built up is consistent with the critical realist understanding of the real world as an open system. There are strong implications therefore as to the shortcomings of the way knowledge is built up in mainstream economics, on the one hand, and how non-mainstream economics should proceed, on the other. The early literature focused on philosophical foundations and critique. As such it was less strong on specific guidance as to how economics should be conducted. This was apparently deliberate – critical realism was not intended as a methodological blueprint. But it means that much work has to be done in working through the implications of critical realism for a number of issues. Three major issues which have already commanded attention are, first, how critical realism relates to different schools of thought within non-mainstream economics, second, the relative roles of philosophy, methodology and practice, and third, the role of econometrics. Lawson (1994b) quite explicitly expresses critical realism as being at a different level from the level at which schools of thought are distinguished.There is a strong overlap with the concept of mode of thought as developed by Dow (1985).The Cartesian/Euclidean mode of thought is based on a closed-system understanding of reality while the Babylonian mode of thought is suited to an open-system understanding. But theorising, as argued above, requires some closure, albeit provisional and partial. An open social system allows for a range of different types of closure. A pluralist methodology means different closures for different parts of the analysis. But, unless it is to mean eclecticism, or ‘anything goes’, pluralism itself needs to be applied within some limits.Thus, paradigms or schools of thought can be understood as particular sets of limits on pluralism.These limits stem from the particular ontology of each community. For neo-Austrians economic reality is individual-based, for example, whereas for Marxists it is class-based, for institutionalists it is institution-based and so on. Different sets of methods are more suited to one school of thought than another – questionnaire surveys and case studies are more central to neo-Austrian methodology, for example, and timeseries analysis is more central to institutionalist methodology. But there may be overlap in methods employed by different paradigms. Further, just as closure in open-system methodology is partial and provisional, so is the closure within a

Critical realism and economics 23 pluralist methodology. Thus, schools of thought, and their interrelations, evolve as the subject matter evolves. Critical realism provides the foundation for this discussion by making clear the starting-point of ontology (see further, Dow, 1999). While ontology as the starting-point for epistemology, methodology and practice is central to critical realism, what is less clear is the relation between philosophy more generally, on the one hand, and practice, on the other. The implication of the critical realist literature is that philosophy comes first, and therefore that philosophy determines practice. But where does the philosophy come from? By critical realist reasoning, philosophy is grounded in the real world, of which scientific practice is a part. It is through struggling to uncover the causal mechanisms which underlie what we experience directly that we become aware of the limitations of mainstream methodology and find ourselves drawn to alternative approaches.As Adam Smith (1795) pointed out, we are motivated by our sense of wonder at unexplained phenomena or, in modern language, we experience cognitive dissonance when our experience does not accord with our theories. Similarly, we experience cognitive dissonance when we are dissatisfied with the mainstream approach to economics, but not when we adopt an approach which is designed to uncover reality as we believe it to be. Critical realist philosophy thus constitutes an application of reason to how we understand the world in order to build knowledge about it. But as our knowledge builds, we need to be open to the possibility that our philosophy too may evolve.Thus, while critical realist philosophy is applied in a ‘top-down’ way, it has itself been developed in a ‘bottom-up’ way. A challenge which lies ahead is to treat critical realist philosophy itself as an open system, open to further lessons learned from practice. A particularly controversial element of practice is econometrics. It has been forcefully argued, notably by Lawson (1997), that econometric techniques lack any justification when applied to an open social system, since they require closure in the aspect of the real world under study. But the argument is building force that econometrics, thoughtfully applied, might be a useful descriptive tool alongside other methods for building knowledge (see, e.g. Downward, 2002, and subsequent chapters in this volume). Further, since econometrics is not a homogenous set of techniques, some techniques may find more justification than others. In his critique of Tinbergen, for example, Keynes denied the general applicability of econometrics, and in particular general claims as to the econometric testability of theories, but did not rule out the possibility that it might have some application (see, e.g. Lawson 1985; Pesaran and Smith, 1985). The important point was that the onus was on the econometrician to justify assuming a sufficient degree of closure in reality to warrant application of a closed-system econometric model, and to accept limitations on how the results might be used.This is an area where practice can be considered in arguments about critical realist philosophy, such that the development at the philosophical level and the level of practice interact. Indeed, it is precisely this theme which is addressed in subsequent chapters.

24

Sheila C. Dow

Conclusion Critical realism represents an important development in our thinking about the philosophical foundations for economics, and their implications for methodology and practice. It has echoes in eighteenth-century debate about science which grappled with issues concerning the best way to build up knowledge about the real world as the basis for changing that world. But as it developed science took on a life of its own such that its relationship with reality became less of a focus for attention and accordingly more tenuous. Critical realism has refocused our attention on an independent reality and what is required to build knowledge about it in order to change it. As such critical realism embodies a powerful critique of mainstream economic methodology.At the same time it presents philosophical foundations for non-mainstream economics in the form of a pluralist, fallibilist approach.The important agenda would now appear to be to build on these foundations, more explicit direct applications of critical realism, to demonstrate the practical difference which such an approach makes. Such applications in turn will provide material to be fed back to the philosophical level. The philosophy stems from an understanding of social reality as an open system. But as we build up knowledge of particular aspects of that open system, there is scope for further developing the philosophical foundations.A particular case in point is the issue of the role of econometrics, which philosophical argument seemed to rule out, but practice suggests a modification to that position. Finally there is considerable scope for further definition of non-mainstream schools of thought in terms of their different types of open-system ontology.At all levels – reality, observation, theory, methodology and philosophy – openness has to involve some (partial, provisional) segmentation in order to be functional. Mapping out the territory within open systems is the next important challenge.

Notes 1 The question is rarely raised explicitly, except by critical realists.

2 In philosophical terms, realism is a multi-faceted concept, treated differently in different contexts; see Maki (1998). 3 See Dow (2002b, chapter 6) for a fuller discussion of Popper and Lakatos’s philosophy of science as it relates to economics. 4 This was particularly true of Sargent; see Sent (1998). 5 It is ironic, and potentially rather confusing, to argue that Scottish Enlightenment philosophy, particularly that of Hume, should provide an alternative route to critical realism, since Bhaskar in particular sees the Scottish Enlightenment as embodying the epistemic fallacy, and Hume as having inspired an empiricism based on observed event regularities. What is involved here is two very different interpretations of Hume and the Scottish Enlightenment. 6 The term ‘common sense’ here refers to common-sense philosophy; see Comim (2002). 7 What is meant by reflecting nature is crucial. Samuelson’s correspondence principle was an attempt to justify the assumption that the dynamic processes underpinning comparative statics were stable; see Blaug (1980, chapter 4).

Critical realism and economics 25 8 Lawson (1995) explicitly develops the commonality between critical realism and Keynes’s philosophy.

References Becker, G. (1991) A Treatise on the Family, Cambridge, Harvard University Press. Bhaskar, R. (1975) A Realist Theory of Science, Leeds, Leeds Books. Blaug, M. (1980) The Methodology of Economics: Or How Economists Explain, Cambridge, Cambridge University Press. Boland, L.A. (1982) The Foundations of Economic Method, London, George Allen & Unwin. Caldwell, B.J. (1982) Beyond Positivism: Economic Methodology in the Twentieth Century, London, Allen & Unwin. Chick,V. and Dow, S.C. (2001) ‘Formalism, Logic and Reality: A Keynesian Analysis’, Cambridge Journal of Economics, 25, 6: 705–22. Comim, F. (2002) ‘The Scottish Tradition in Economics and the role of Common Sense in Adam Smith’s Thought’, Review of Political Economy, 14, 1: 91–114. Dow, S.C. (1985) The Methodology of Macroeconomic Thought, Oxford, Blackwell. —— (1999) ‘Post Keynesianism and Critical Realism:What is the Connection?’, Journal of Post Keynesian Economics, 22, 1: 15–32. —— (2001) ‘Modernism and Postmodernism: A Dialectical Analysis’ in S. Cullenberg, J. Amariglio and D. F. Ruccio (eds), Postmodernism, Economics and Knowledge, London, Routledge, pp. 61–76. —— (2002a) ‘Hume and Critical Realism’, Cambridge Journal of Economics, 26, 6: 683–95. —— (2002b) Economic Methodology: An Inquiry, Oxford, Oxford University Press. Downward, P. (2002) ‘Realism, Econometrics and Post Keynesian Economics’ in S.C. Dow and J. Hillard (eds), Post Keynesian Econometrics, Microeconomics and the Theory of the Firm, Beyond Keynes vol. I, Cheltenham, Edward Elgar, pp. 144–61. Friedman, M. (1953) ‘The Methodology of Positive Economics’ in Essays in Positive Economics, Chicago, Chicago University Press, pp. 3–43. Kuhn, T.S. (1962) The Structure of Scientific Revolutions, Chicago, Chicago University Press. Lawson, C. (1999) ‘Realism,Theory and Individualism in the Work of Carl Menger’ in S. Fleetwood (ed.), Critical Realism in Economics: Development and Debate, London, Routledge, pp. 43–62. Lawson,T. (1985) ‘Keynes, Prediction and Econometrics’ in T. Lawson and H. Pesaran (eds), Keynes’ Economics: Methodological Issues, London, Croom Helm. —— (1994a) ‘Why Are So Many Economists So Opposed to Methodology?’, Journal of Economic Methodology, 1, 1: 105–34. —— (1994b) ‘The Nature of Post Keynesianism and Its Links to Other Traditions: A Realist Perspective’, Journal of Post Keynesian Economics, 16, 4: 503–38. —— (1995) ‘Economics and Expectations’ in S.C. Dow and J. Hillard (eds), Keynes, Knowledge and Uncertainty, Cheltenham, Edward Elgar, pp. 77–106. —— (1997) Economics and Reality, London, Routledge. Maki, U. (1998) ‘Realism’ in J.B. Davis, D.W. Hands and U. Maki (eds), The Handbook of Economic Methodology, Cheltenham, Edward Elgar, pp. 404 –13. Loasby, B.J. (2002) ‘Closed Models and Open Systems’, Distinguished Lecture to the Biennial INEM Conference, Stirling, September.

26

Sheila C. Dow

Nagel, E. (1963) ‘Assumptions in Economic Theory’, American Economic Review Papers and Proceedings, 52: 211–9. Pesaran, H. and Smith, R. (1985) ‘Keynes on Econometrics’ in T. Lawson and H. Pesaran (eds), Keynes’ Economics: Methodological Issues, London, Croom Helm. Samuelson, P. (1963) ‘Problems of Methodology – Discussion’, American Economic Review Papers and Proceedings, 52: 231–6. Sent, E.-M. (1998) The Evolving Rationality of Rational Expectations: An Assessment of Thomas Sargent’s Achievements, Cambridge, Cambridge University Press. Smith, A. (1795) ‘History of Astronomy’ reprinted in W.P.D.Wightman (ed.), Essays on Philosophical Subjects, Oxford, Clarendon, 1980, pp. 33–105.

3

Conceptualizing unemployment in a period of atypical employment A critical realist approach1 Steve Fleetwood

Introduction The government claimed the UK is on the road to full employment after the number of people claiming benefit in February dipped below one million for the first time in over 25 years.The claimant count fell to 996,000 in February, the lowest since December 1975. (Labour Research, 2001, p. 6) Reports like this, drawing upon official unemployment figures, are hotly debated. Much ink has been spilled on problems associated with understanding what these figures actually mean, especially the issue of ascertaining who should and should not be included in the unemployed count (cf. Hughes, 1995; Wells, 1995; Convey, 1996). Whilst issues such as these are far from trivial, there are two other relatively unexplored issues that pose far greater problems for interpreting the unemployment figures. The first issue turns on the relationship between unemployment and employment – a relationship that has become even more significant due to the (re)emergence of atypical employment. If to be unemployed is to be without employment, that is, without a job or work, then unemployment becomes the other, or absence, of employment. This distinction makes the reality of unemployment partly dependent upon the nature of the available employment. If, furthermore, the reality of unemployment is to be adequately expressed in economic theory and subsequently measured, then the concepts used to define and measure unemployment must take the reality of employment into account. That is, the conceptualization and measurement of unemployment must adequately grasp the reality of employment. This leads to the second issue. Designing theoretical concepts that adequately express reality rests (minimally) upon the adoption of appropriate ontological, philosophical and methodological foundations. Adoption of an inappropriate method raises the strong possibility that one’s theoretical concepts will fail to express adequately the reality under investigation.

28

Steve Fleetwood

The objective of this chapter, then, is to integrate these two relatively unexplored issues in order to demonstrate how the method adopted by mainstream economists encourages the inadequate conceptualization of employment, and, thereby, the inadequate conceptualization and measurement of unemployment.The section ‘Philosophy and methodology’ introduces the methodological perspective of critical realism2 to identify and locate the source of the methodological problems encountered by mainstream economists. Critical realism explains why the use of a particular (i.e. deductive) method means that theoretical concepts have to be constructed in such a way as to reduce the multi-dimensional, qualitative reality of employment and unemployment to the quantitative, single dimension of variables, whereupon they cease to be adequate expressions of the reality they are designed to investigate. The section on ‘The connection between employment and unemployment’ turns from methodology to labour economics and firmly establishes the connection between employment, and the conceptualization and measurement of unemployment. The concept of underemployment illustrates how a partial lack of employment fails to appear in unemployment figures. The section on atypical employment extends the discussion of labour economics by using part-time employment as an example of atypical employment to illustrate how the latter differs significantly from typical employment in a number of dimensions, most of which are qualitative in nature. Once the multi-dimensional and qualitative nature of atypical employment is firmly established, the full force of the methodological critique developed in the section on philosophy and methodology is unleashed. Finally, the section ‘Implications of critical realism’ draws together methodology and labour economics to demonstrate the implications that the critical realist perspective has for the interpretation of contemporary unemployment figures.

Philosophy and methodology Deductivism

Lawson (1997) has argued provocatively that the dominant mode of economic theorizing consists in the use of a particular method, namely the deductive method, or simply deductivism. Deductivism appears, generally, in the guise of the deductive-nomological (DN), or covering law, model of explanation, whereby to ‘explain’ something is to deduce a statement about that something from a set of initial conditions, assumptions, axioms and a set of event regularities that constitute the covering law. It is important to recognize that every method is rooted in ontology. Every time one makes a theoretical or methodological statement, one has already made explicit or implicit claims about the way the world is thought to be; one has made ontological commitments or presuppositions; one has adopted an ontological position. Moreover, theoretical or methodological statements are rendered intelligible, when they presuppose an ontological position that is

Conceptualizing unemployment

29

consistent with them. Methodological individualism, for example, is rendered intelligible by the presupposition of an atomistic ontology; whereas it would be rendered unintelligible by the presupposition of an ontology of holism. Now this chapter wants to argue first, that deductivism is rooted in an empirical realist ontology and second, that the problems afflicting deductivism can ultimately be traced back to this ontology. To address these issues, the chapter will proceed by tracing the central features of deductivism backwards, step by step as it were, to the ontology that renders these features intelligible, whereupon it will be possible to locate the precise source of the problems. ●

●

●

●

●

●

Central to the way deductivism is operationalized are functional relations, generalized as y  f (x). These can also be expressed as laws and styled ‘whenever event x then event y’. If functional relations and laws are to have economic meaning (as opposed to ‘mere’ mathematical meaning) then they must imply causality. Causality renders them intelligible. The clear implication, for example, of writing, q  f ( p) is that the quantity (demanded or supplied), q, varies with, and is caused to vary by, price, p. Functional relations and laws are not, however, rendered intelligible by just any old account of causality. Rather, they are rendered intelligible by an account of causality based upon constant conjunctions of events – and deriving from Hume.The use of this notion of causality might arise from a conscious decision, or it might arise by default, because other notions of causality that involve, for example, causal powers would render functional relations and laws unintelligible. Laws as constant conjunctions of events are, thereby, referred to as ‘Humean’.3 If, one were to discover a constant conjunction of events in the form of a Humean law or functional relation, one might claim to have scientific knowledge. This is because it is the constant conjunction of events that makes possible the deduction or prediction of some event(s) from antecedents. Crucially, then, constant conjunctions of events drive the nomological machinery of the deductivist method. Scientific knowledge in the form of constant conjunctions of events is only intelligible on the presumption that particular knowledge is derived via experiencing, and subsequently recording unique, individual or atomistic events. These events cannot be other than atomistic, since any connection or relation between them would be impervious to sense experience, otherwise the nature of the connections would require prior explanation, thus undermining the explanatory power of sense experience. The ontology, implicit or explicit, is, therefore, one of atomistic events. Ontology is confined to that which is experienced and is, therefore, of the atomistic events of sense experience. Because these objects are confined to experience the ontology is empirical; and because these objects are thought to exist independently of one’s identification of them, it is realist; The ontology can, thereby, be labelled empirical realist.

30

Steve Fleetwood Table 3.1 Empirical realist or ‘flat’ ontology Domain

Entity

Empirical Actual

Experiences, perceptions Events and actions

The deductivist method and the functional relations and laws that operationalize it, are rendered intelligible, therefore, by the consistent presupposition of causality as constant event conjunctions and an empirical realist ontology.4 Table 3.1 illustrates, that this empirical realist ontology consists of two fused domains referred to as the empirical and the actual. For the empirical realist, what is, is presumed coexistent with what is perceived. Causality as constant event conjunctions means that if some event is perceived, one can only seek its cause in terms of another perceived event. There is nowhere else to seek a cause because any other domain in this ontological spectrum is ruled out. With the ontology that renders deductivism intelligible identified, it is now straightforward to locate the precise source of the problem.The ontology is, in fact, at odds with the way the world really is. Put another way, the ontology fundamentally misconceives the nature of the socio-economic world.The world does not consist merely of events and their experiences: nor does it consist merely of constant conjunctions of these events. As critical realists continually point out, underlying the domains of the empirical and the actual is a domain of the (metaphoric) ‘deep’ where the causal powers that govern events and experiences reside. This misconceived ontology has damaging implications for method and theory – as will become clear. Whilst constant conjunctions of events are fundamental to deductivism, they are exceptionally rare phenomena. There appear to be very few spontaneously occurring systems wherein constant conjunctions of events occur in the natural world, and virtually none in the social world.That is not to deny the possibility that constant conjunctions may occur accidentally, or over some restricted spatio-temporal region, or be trivial. But virtually all of the constant conjunctions of interest to science only occur in experimental situations. In such situations, a very special system is generated, namely a closed system, that is, a system wherein events are actively engineered to be constantly conjoined. In natural science, the point of experiment is to close the system by creating a particular set of conditions that will isolate the one interesting mechanism.This mechanism is then allowed to operate unimpeded and the results, the constant conjunctions, recorded. In economics, constant conjunctions of events appear to be found only in the ‘conceptual experiments’ (Pencavel, 1994: 14), that constitute theoretically closed systems. Unfortunately, natural or social scientists who adopt the deductivist method face the following problematic and counterintuitive implications: a

Outside closed systems, where constant conjunctions of events are not usually found, one would have to conclude that there are no laws. This

Conceptualizing unemployment

b

c

31

would be tantamount to saying that nothing governs the (non-constant) flux of events in open systems; science would, then, become a fruitless endeavour. It is often the case that conclusions derived from experimental situations (i.e. in closed systems) are successfully applied outside experimental situations (i.e. in open systems). Because of (a) above, this state of affairs would have no valid explanation. The obvious problem of how one may, justifiably, claim anything about a reality that constitutes an open system from an analysis of a closed system has never been seriously addressed by mainstream economists. In fact, deducing statements about the action of agents operating in a closed system, and transferring them to the action of agents in the open system, commits the fallacy called ignoratio elenchi.This entails ‘assuming that one has demonstrated something to be true of X when the argument or evidence really applies to Y which is not the same as X in some respect’ (Gordon, 1991: 108).What is ‘not the same’ is the existence and ubiquity of constant conjunctions of events.

Despite these problematic and counterintuitive implications rendering the deductive method singularly inappropriate for the analysis of open systems, (i.e. virtually all socio-economic systems) deductivism prevails. Moreover, deductivism’s crucial need to engineer closed systems impacts negatively on the way the theoretical concepts are constructed.And this is true for the inadequate concepts constructed to investigate unemployment and atypical employment. Before discussing these inadequate conceptions, however, the following point is worth making to avoid confusion. Although the argument has been developed from the practice of natural science, it is applicable to social science in general, and mainstream economics in particular, for two reasons. First, mainstream economists readily admit to using (what they assume to be) the method of natural science. Second, and more importantly, if human agency is real, then (a) human agents could always have acted otherwise, and (b) human action must make a difference to the social world. If, minimally, (a) and (b) are accepted, the social world cannot be a closed system and any attempt to model it ‘as if ’ it was leads to the kind of problematic and counterintuitive implications just noted.

From closed systems to inadequate conceptualizations Recall that from the deductivist perspective, scientific knowledge is obtained by reducing objects or features of reality to events, and subsequently their constant conjunctions. Since constant conjunctions only occur in closed systems, the theoretical components that comprise the system must be framed

32

Steve Fleetwood

in such a way that constancy is never threatened. And for this, (at least) two requirements must be sought after: a

b

The events themselves must be identical, that is, episodes of the same kind: ‘Whenever event x’, implies that a number of episodes of the same kind (x) have occurred.5 Event x could, for example, be a change in the number of oranges, apples or workers employed in a certain job. Whatever event x refers to, all episodes of it have to be identical.The requirement of identity implies a common dimension. These events must be susceptible to quantification and measurement in space and time. This imperative to quantify and measure implies an unchanging dimension. If one is adding apples, and the time span is so long that by the time one gets to the end of the barrel, the apples have rotted, and are no longer apples, then the dimension will have changed and (meaningful) addition will become impossible.

The requirements of a common and an unchanging dimension are presumed to be met by re-conceptualizing, re-defining or reducing events to variables, whence changes in their magnitude can be recorded. A variable, in turn, must retain two important features: i

ii

It must possess one, and only one, common and unchanging dimension – that is, number, quantity or magnitude. The only change a variable is permitted to experience is change in this number, quantity or magnitude. It must maintain a stable reference to some real object or feature of reality.

Problems arise, however, if the real object to which the variable refers undergoes a qualitative change in its nature. If, for example, one is measuring ice cubes with the variable ‘width of ice cube’ and the temperature rises sufficiently, the qualitative nature of the ice cubes will change.The variable ‘width of ice cube’ and the object ‘ice cube’ have in a sense come adrift.The variable, unable to maintain a stable reference to its object, becomes an inadequate conceptualization of reality. Of critical importance for this chapter is the obvious consequence that what is true for ice cubes is also true for many economic entities such as employment. Consider the following statement: ‘Whenever the magnitude of W (wage) changes, then the magnitude of E (filled jobs) changes’. Here one is measuring the reality of employment (which is a qualitative phenomenon) with the variable ‘filled jobs’6 (which is a quantitative phenomenon). But employment, as a qualitative phenomenon, can change. To put matters simply for ease of exposition, ‘good quality’ employment can become ‘poor quality’ employment and vice versa. If this occurs, if the quality of employment changes, whilst the variable ‘filled jobs’ remains unchanged, then the object

Conceptualizing unemployment

33

‘employment’ and the variable ‘filled jobs’ have in a sense come adrift. The variable, unable to maintain a stable reference to its object, becomes an inadequate conceptualization of reality. Since (as the section on atypical employment demonstrates) the nature of employment is undergoing a highly significant qualitative change, mainstream economists are forced onto the horns of a dilemma.They must choose between (a) embracing the qualitative change, violating an important feature of their variable and, therefore, having to re-conceptualize employment; or (b) continuing with their variable, ignoring the qualitative change, and proceeding with an inadequate conceptualization of employment. Let me elaborate this important point. a

b

If mainstream economists wish to embrace qualitative change in employment, and if the variable ‘filled jobs’, is to continue to maintain stable reference to its object, then the variable will have to undergo a qualitative change. Some other way will, then, have to be found for adequately conceptualizing employment. But as noted in (i) above, this will violate one of the important features of a variable because a variable can undergo quantitative but not qualitative (check) change. If mainstream economists wish to continue with the variable ‘filled jobs’, they will have to ignore qualitative change in employment and hence proceed with an inadequate conceptualization of employment.

It is important to note that the deductive method is setting the agenda of theoretical discourse, forcing the economist to choose between embracing or ignoring qualitative change. Typically, mainstream economists, not wishing to abandon the deductive method, have no option but to opt for the latter.7 The result is a set of theoretical concepts, variables, that are totally devoid of qualitative content or properties and in this form, however, they are useless devices for an inquiry into reality. Now, it is worth noting that whilst critical realism is avowedly anti-empiricist, it is not anti- the use of empirical data per se. Clearly some qualitative aspects of employment can be quantified and measured relatively straightforwardly – for example the number of jobs held by one person.8 Other qualitative dimensions can be quantified and measured, although not straightforwardly, by finding proxy variables, or by constructing suitable indices. An example of this might be the way employment insecurity is often measured by the number of years an employee has held the same job. But even here, (as discussed below) qualitative issues start to creep into the picture casting doubt on the meaningfulness of the measurement.9 However, and this is the crucial point, the mainstream economist’s use of the deductivist method, and hence the imperative to quantify and measure, rules out certain avenues of investigation for (at least) two reasons. First, the mainstream economist, motivated by a desire to quantify and measure, simply has no way of discovering relevant qualitative phenomena. How, for example, would a mainstream economist ever discover the inferior nature of

34

Steve Fleetwood

many part-time employees’ pension schemes? Things of this nature can only be discovered by sociological or anthropological techniques which are, typically, eschewed by mainstream economists on the grounds that they are not ‘scientific’, that is, are not rooted in the deductive method. Second, even if this mainstream economist (somehow) discovered relevant qualitative phenomena, how might they be quantified and measured? For example, how might one (meaningfully) quantify and measure the hidden sexist ideology buried in the notion of parttime ‘family friendly’ employment arrangements, let alone their consequences (see the end of the section on atypical employment.)? In sum then, the deductivist method is inappropriate to the study of socioeconomic phenomena (such as the atypical employment and unemployment elaborated upon) because this method is rooted in an ontology that is at odds with the way the world really is. Furthermore, the need to engineer a closed system heavily influences the way the theoretical concepts used to investigate employment and unemployment are constructed.To be more specific, deductivist reasoning requires that the theoretical concepts have to be constructed in such a way so as to reduce the multi-dimensional, qualitative reality of employment and unemployment, first to the level of events and second to the quantitative, single dimension of variables. As variables, however, these theoretical concepts cease to be adequate expressions of the reality they are designed to investigate. Before we leave methodology, one final point needs to be made. One does not need critical realism merely to show that mainstream economics is preoccupied with quantification and is unable to (meaningfully) deal with the kind of qualitative issues that arise when investigating atypical employment: this is a well-known criticism. One does, however, need critical realism10 to explain this state of affairs.Without a critical realist perspective, the explanation of this state of affairs would be ad hoc, turning on things like subjective preference, (i.e. a quantitative approach is simply preferred) or lack of sociological sophistication on the part of mainstream economists.With a critical realist perspective, however, the explanation ceases to be ad hoc. Once deductivism is adopted, and with it the commitment to closed systems, framing theoretical concepts in quantitative terms becomes almost irresistible, whilst framing them in qualitative terms becomes almost impossible. The methodology generates a kind of theoretical ‘lock in’ – and, by extension, ‘lock out’.11

The connection between employment and unemployment If mainstream economic theory is ill-equipped to deal with qualitative phenomena because of its adherence to deductivism, then investigating the qualitative relationship between unemployment and employment is bound to be fraught. A testament to this is that the qualitative relationship between unemployment and employment is hardly ever investigated by mainstream economists.12 This is a significant oversight, because the (re)emergence of atypical employment places the quality of employment firmly on the agenda.13

Conceptualizing unemployment

35

Let us leave quality to one side for the moment, and concentrate on reality. If to be unemployed is to be without employment, a job or work, then unemployment becomes the other, or absence, of employment.This distinction makes the reality of unemployment partly dependent upon the reality of employment. If, furthermore, the reality of unemployment is to be adequately expressed in economic theory and subsequently (meaningfully) measured, then the concepts used to define and measure unemployment must take the reality of employment into account. The conceptualization and measurement of unemployment must, therefore, adequately grasp the reality of employment. The following caricature might drive the point home. It is most unlikely that a person would be classified, and measured, as unemployed if the only alternative was ‘employment’ in the form of slavery.The traditional conceptualization and measurement of unemployment would be radically altered by the reality of ‘employment’.14 Bringing quality back into the picture, it appears that the conceptualization and measurement of unemployment must adequately grasp the qualitative reality of employment. Given that (as the section on atypical employment will show) employment increasingly comes in qualitatively different forms. The quality of employment must, on pain of irrelevance, become a legitimate issue for conceptualizing and measuring unemployment. If one is content to merely count heads, the quality of employment will be of no concern: an individual either has or does not have a job and will be classified, and counted, as employed or unemployed accordingly.15 If they have a part-time job, this position may be counted as a fraction of a full-time post. When, however, one refuses to ignore the fact that not all jobs are alike, more specifically, that atypical and typical jobs are often qualitatively very different, then merely ‘having a job’ obscures important differences and headcounting becomes, at the very best, one dimension of a multi-dimensional analysis. As Sengenberger (1996) puts it: [T]he definition of full employment … needs to take into account changes in the structure of employment, such as new forms of flexible employment … Measures of unemployment catch only one aspect of the employment problem … that of total lack of work. Less obvious situations, such as the partial lack of work … are not accounted for in unemployment statistics at all. (Sengenberger, 1996, emphasis added) The category of underemployment is designed to catch some of the gray area where employment and unemployment cannot be sharply delineated. A worker is underemployed when employed but this employment is, in some sense, less than adequate. Underemployment is usually discussed via two categories – visible and invisible. Visible underemployment is so named because it is relatively easy to see, quantify and measure – but also take note of the points made earlier in the chapter about the meaningfulness of measurement. Visible underemployment might

36

Steve Fleetwood

occur when a worker is constrained to work fewer hours than he/she desires. Invisible underemployment is so named because it is relatively difficult to see, and in some cases impossible to (meaningfully) quantify and measure because, for example, it relates to issues like being (under)employed in a job where one’s skills are not being adequately utilized. It is, however, possible to identify a third form of underemployment which might be called ‘poor quality underemployment’.This would occur, for example, if full-time jobs were turned into part-time jobs and the quality of employment deteriorated. Something like ‘poor quality underemployment’ appears to be crucial for understanding employment and unemployment in the present period. Paraphrasing Sengenberger quoted earlier one could argue that ‘partial lack of work, in the form of poor quality underemployment or atypical employment, is not accounted for in unemployment statistics at all’. In sum, connecting the critical realist discussion of methodology to the discussion of the quality of changing nature of employment throws up two problems for mainstream labour economics. First, it has not taken the qualitative transformation in the nature of employment into consideration.16 Second, as the critical realist critique has established, as long as mainstream economics remains wedded to deductivism it cannot (in any meaningful sense) take this qualitative transformation into account. Because it is almost impossible to overemphasize the impact of this problem, one can offer a very bold, if sweeping statement for emphasis. As long as mainstream economics refuses to adopt an alternative method, it is destined to be irrelevant. It will continue to measure changes in variables like ‘jobs filled’ and ‘seeking employment’; and it will continue to predict (with, if the economy constitutes an open system, little hope of success) changes in the magnitude of these variables. But it will get nowhere near to a genuine understanding of what is actually going on in reality.

Atypical employment: the case of part-time work Now, the foregoing arguments have continually made reference to the qualitative transformation in the nature of employment that is currently taking place. The task of this section is to elaborate upon it. The multi-dimensional nature of atypical employment makes it notoriously difficult to define (cf. Casey, 1988; Polivka and Nardone, 1989; Roosenthal, 1989; Pollert, 1991; Ewing, 1996; Klein, 1996; Polivka, 1996; Casey et al., 1997). Atypical employment can be conceived under very general headings such as: contingent work, alternative work arrangements, flexible working practices; or under less general headings such as independent contractors, on-call workers, temporary help agency workers, workers provided by contract firms. Atypical employment can also be conceptualized of in specific forms such as: part-time, self-employed, zero-hours contracts, home workers, flexi-time, annualized hours, compressed working weeks, job-share, seasonal workers, workers in special programme for the unemployed and so on.

Conceptualizing unemployment

37

Since a thorough investigation of the myriad forms of atypical employment is, obviously, beyond the scope of this chapter, investigation of atypical employment proceeds via the example of one of its most common forms, namely noncontingent part-time employment – hereafter referred to simply as part-time employment.17 By exploring the reality of part-time employment, it is hoped to make the following points clear. First, whilst all forms of employment have an irreducibly qualitative nature (at its simplest, no two jobs are identical), the chapter is trying to establish something more than this. It attempts to capture the profound qualitative changes that are currently occurring in the nature of employment. This can best be done, by describing how part-time (atypical) employment differs from full-time (typical) employment in a number of dimensions, most of which are qualitative in nature. Second, once the multi-dimensional and qualitative nature of atypical employment is firmly established, the full force of the methodological critique developed in the section ‘Philosophy and methodology’ is unleashed on mainstream economics. It becomes relatively easy to see why the deductive method leaves mainstream labour economics ill-equipped to deal with qualitative changes in the nature of employment. Emphasis will be placed, therefore, on the nature of these changes and the deterioration, in terms and conditions, that characterize the shift to part-time employment. In March–May 1992, there were approximately 7.4 million part-time employees – 6.4 million women and 1.0 million men. In December–February 2001, the figure had risen to approximately 8.4 million part-time employees – 6.9 million women and 1.4 million men (www.statistics.gov.uk/statbase/xdataset.asp). What is perhaps more significant than absolute numbers, is the fact that many entry points into the labour market are via part-time and other forms of atypical employment. Between the winters of 1992–93 and 1995–96, only 9 per cent of the 750,000 new jobs created were permanent and full time. Half were permanent part-time and a further 15 per cent were temporary part-time (TUC, 1996). Now, most economic literature on part-time employment tends to be quantitative and statistical, focusing on the average part-timer. Whilst it is interesting to know that part-time work is on the increase, or that mean hourly wages of part-timers is lower than that of full timers, such statistical statements illuminate very little of the reality of the employment experience, and often in fact, disguise far more significant issues. As Tilly puts it: ‘Behind the averages however, fascinating glimpses of diversity emerge’ (1992, p. 331). Even the category ‘part-time’ conceals many differing employment experiences.Tilly (1992), for example, observes two broad types of part-time employment. Retention part-time jobs tend to be found in the primary labour market, and are designed by employers to retain or attract valued workers who prefer to work part-time. Secondary part-time jobs tend to be found in the secondary labour market, and are designed by employers to gain advantages of lower compensation and greater scheduling flexibility. To all intents and purposes, retention part-time jobs are often similar in quality not only to full-time jobs, but to

38

Steve Fleetwood

typical full-time jobs at that.18 Since secondary part-time employment involves the largest number of workers, and creates the most problems for those who experience it, part-time secondary employment will be the focus here. Remuneration There is no shortage of figures on pay for part-timers, although the evidence is mixed and difficult to interpret (Blank, 1990). According to McGregor and Sproull (1992), in 90 per cent of the companies they surveyed, hourly rates of pay for part-timers were the same as full-timers; the IDS (1993) study shows something similar.Yet overall figures reveal a different story.The New Earnings Survey shows that in April 2000, female average gross hourly earnings for all occupations was £5.27 for part-time and £8.40 for full-time (www. statistics.gov.uk/nsbase/downloads/theme_labour/nes.pdf ).19 Part-timers are highly concentrated in certain occupations such as clerical and secretarial, personal services and sales assistants (Fothergil and Watson, 1993: 214).These occupations, part or full-time, tend to be filled primarily, although not exclusively, by women and tend to be low paid. Even within an occupation, part-time employees tend to earn less per hour than their full time counterparts. Overtime pay and the second job Traditional overtime hours are worked before, or after, the main working day/shift and/or at weekends.A report by the Equal Opportunities Commission found that overtime pay by part-time workers ‘rarely attracts more than the basic pay’ (cited in LRD July 2000: 33). Although average paid overtime hours have remained fairly constant over the last two decades, there has been a change in when and how it is performed. Overtime working increasingly takes the form of a second, and therefore part-time, job. Second jobs tend to be paid at normal rather than overtime rates, so any full-timer engaged in overtime in the form of a second job, as opposed to traditional overtime arrangements, experiences a relative loss in hourly pay. Second jobs tend also to be paid at normal rather than unsociable hours rates, so any full or part-timer engaged in overtime in the form of a second job that entails unsociable hours once again experiences a relative loss in hourly pay.Whilst most firms do pay pro rata overtime and unsociable hours pay rates, many part-timers working set hours that combine both of these categories, by working evening shifts or permanent weekends for example, often receive less than would be paid to a full-time typical worker. Overtime work in the form of a second job, then, constitutes a qualitative change in the nature, and a deterioration in the conditions, of employment. Multiple job holding Multiple job holding is important for understanding part-time employment because when people hold a portfolio of jobs it is likely to consist of a mixture

Conceptualizing unemployment

39

of full and part-time, or a mixture of various part-time jobs. As Dex and McCulloch (1995, p. 65) put it:‘it is possible to argue that second job holding is an element of flexible job holding’. In March–May 1992, 973,000 workers held multiple jobs; by November–January 1996 the figure had risen to 1,335,000, although by December–February 2001, the figure had fallen slightly to 1,170,000 (www.statistics.gov.uk/statbase/xdataset.asp). A survey by the public service union Unison revealed that 38 per cent of the 2,000 cleaning and catering staff at Newcastle-Upon-Tyne city council had two part-time jobs, while almost 4 per cent had three jobs (Hetherington, 1995).Whilst the disadvantages associated with multiple job holding have not been well documented, some of the more obvious are not hard to conceive.They include: increased time spent travelling to work; associated increase in travel costs; and reduced, or in some cases no, paid tea/meal breaks. Multiple job holding, then, constitutes a qualitative change in the nature, and a deterioration in the conditions, of employment. Non-pay benefits Part-timers are often disadvantaged relative to full-timers in the same firm in terms of sick pay, pension schemes, bonus or profit share, discount on goods/services, interest free/low loans, subsidized hospital/medical insurance. Part-timers whose normal pattern does not include public holidays usually receive no entitlement to another day off. In Autumn 1995, 62 per cent of men and 32 per cent of women part-time employees had no paid holiday entitlement; whilst the figures for full-time employees are 7 and 17 per cent, respectively. The average number of days of paid holiday entitlement for part-time employees was 13, and that of full-time employees was 21 (LFS Helpline, May 1996). Furthermore, 17–23 per cent of firms offered no pension scheme to part-timers, and a further 28–37 per cent restrict it to those working more than 16 hours per week. Although from July 2000, The Part-Time Workers (Prevention of Less Favourable Treatment) Regulation makes it difficult for employers to discriminate against their part-time employees, there is still considerable scope within the regulations for discrimination to continue (see LRD, July 2000). There is, however, an extremely important point buried within the data on pensions, namely that important qualitative issues are extremely unlikely to be discovered via quantitative analysis. It is not difficult for an economist to obtain data on whether or not a part-time employee is covered by the company pension scheme. But, this is only one aspect of the matter.As has recently come to light in the United Kingdom via the large scale mis-selling of pensions, not all pensions are alike: some are better than others. A National Association of Pension Funds survey revealed that one in eight of the pension schemes admitting part-timers provided inferior benefits (Labour Research, 1994, p. 9, 10). The reduced non-pay benefits available to part-time employees, then, constitute a qualitative change in the nature, and a deterioration in the conditions, of employment.

40

Steve Fleetwood

Employment insecurity Employment insecurity is a partially subjective state of affairs, making its investigation via quantitative techniques (e.g. administering questionnaires or measuring employment duration), highly problematic. Whilst many quantitative studies20 have not found a decrease in employment duration in recent years, casting doubt on the belief that employment insecurity is rising, matters are not so simple. For example, although no separate figures on part-time temporary workers are available, the LRD (1995) survey found 43 per cent of temporary employees have been with the same employer over a year, and a further 12 per cent have been with the same employer over five years.21 A temporary worker who has his/her temporary contract continually renewed will appear in a quantitative survey of employment duration or turnover as secure.Whilst a worker in this position might be treated under job protection legislation as permanent, it does not seem unreasonable to suggest that insecurity might arise from the continual worry that the contract might not be extended. Burchell et al. (1999) found evidence that job insecurity included not only the fear of losing one’s job, but also the fear of losing valued features of one’s job such as: promotion opportunities; control over the pace of work; the ability to complete the entire job; customary pay rises; and access to representation.Whilst the study did not concern part-time workers per se, it is clear that many of these concerns face part-timers more than full-timers. Another dimension of employment insecurity that is difficult to measure is the voluntary or involuntary nature of any separation that finds its way into the figures for employment duration/turnover. Employment insecurity might be indicated if separation on the part of the employee is involuntary. ‘Might’ because reality is more complex than can often be dealt with in a questionnaire. What should one conclude if a person volunteers for the separation on the grounds that their current job is too insecure and they are seeking something more secure? Gregg and Wadsworth (1995) hit upon what is perhaps the most worrying aspect of insecurity, namely, that contemporary entry points into the labour market are increasingly dominated by insecure employment. It appears that even those who have a secure job are worried, and not without reason, that should they lose it, they are likely to be re-employed in an insecure job: [W]hile tenure and security have changed only marginally for the majority, entry positions available to those currently not in employment have become increasingly unstable … Thus the minority who lose their job or who want to(re-) enter work force face a labour market that is now dominated by part-time and temporary jobs. (Gregg and Wadsworth, 1995, p. 73) The increased job insecurity associated with part-time employment, then, constitutes a qualitative change in the nature, and a deterioration in the conditions, of employment.

Conceptualizing unemployment

41

Nature of work Part-timers are often used to perform the more unpleasant aspects of the job. Balchin (1994, pp. 52–3), for example, shows that part-timers consider themselves treated as ‘second best’ by full-time staff and managers. Part-timers were often moved to other sections at short notice, or used more ‘intensively’ than full-timers, meaning, for example, they were employed to cover lunch or tea breaks or were kept in a demanding job throughout a peak period. Job demands Part-time jobs tend to require low levels of skill, training and responsibility. Even within any low-level job category, such as stock clerk, low-level tasks are assigned to part-timers.22 Promotion Part-timers tend to enter employment at the bottom of the job ladder, and remain there or thereabouts. In retail most full-timers were once part-timers and part-time work acts as a ‘bridge’. But few part-timers become full-timers due to the small number of full-time jobs relative to part-time and the fact that turnover in full-time is relatively low. Many of those part-timers in senior positions originally held this position as full-timers before turning part-time (IDS 1993, p. 3). Awareness of employment protection legislation In an (admittedly) small survey of four (large) UK retail outlets, Balchin (1994) found that a high proportion of workers were unaware or uncertain about their entitlement, as part-timers, to employment protection legislation. The existence of a draft of statutory rights for part-timers passed by the UK Parliament in 1995 (Employment Gazette, February 1995: 43) is one thing, that workers are aware of them is another, and that they are able to use legal channels without fear of reprisal is yet another. Moreover, given that many of the discriminatory practices against part-timers are difficult to uncover (e.g. consider the issue of qualitatively different pensions noted above), they may never actually come to light and may not, therefore, enter the realm of law. The differential treatment of part-time employees vis-à-vis the nature of work, job demands and promotion, coupled with part-timers’ relative lack of awareness of their employment rights, then, constitute qualitative changes in the nature, and a deterioration in the conditions, of employment. Part-time and state benefits Many part-time workers find their low earnings exclude them from a range of state benefits that are based on national insurance contributions such as Statutory Sick Pay and Statutory Maternity Pay. In cases where the part-time

42

Steve Fleetwood

work is based in term-time working (a particularly useful arrangement for working parents) it is currently impossible for workers to claim benefits like Income Support or Job Seekers Allowance for the duration of the terms (LRD, July 2000: 47–8).

Voluntary and involuntary part-time employment Finally, it is worth mentioning the issue of voluntary versus involuntary parttime employment. In December–February 2001, 601,000 males (44 per cent) and 4,665,000 females (81 per cent) were working part time because they ‘did not want a full time job’ (www.statistics.gov.uk/statbase/xdataset.asp). The observation that the majority of female, and a significant minority of male part-timers choose to work part-time is often used to support the argument that the high incidence of part-time employment is not a major problem because most part-time workers volunteer for it. Four points can be noted briefly to show how such a sentiment is questionable. First, whilst statistically it appears that most part-timers are voluntary, there are a number of problems that quantitative data cannot capture. For example, a part-timer working in the secondary labour market and faced with a questionnaire asking: ‘Do (you) not want a full-time job’ has (at least) two scenarios to consider. Is the choice between: a b

20 hours of poor quality and 40 hours of high quality work, or 20 hours and 40 hours of poor quality work?

If the respondent has (a) in mind, and they still answer that they ‘Do not want a full-time job’, then there are grounds for believing that he/she is a ‘voluntary’ part-timer. If, however, the respondent has (b) in mind, then it is not clear that they do not want full-time employment: they may simply not want to work any more hours in a low quality job. Second, since most part-timers are women, and women tend to be burdened with domestic, child-care, and dependent-care duties, the notion of ‘volunteering’, or ‘choice’ of hours becomes enmeshed in wider socio-political matters and cannot be treated as akin to choice over the purchase of washing powder. Moreover, the high incidence of part-time amongst female workers might even re-enforce sexism. One study notes the following advantage of flexible working for employees: Many part-time women work on twilight production shifts which enable them to be at home with their children or other dependents during the day. (IDS, 1993, p. 3)

Conceptualizing unemployment

43

But as Briar points out, promotion of ‘family friendly hours’ designed to: help women compete more effectively with men at work [have the effect of] helping more women to continue bearing the main responsibility for household labour and caring. (Briar cited in Warme, 1992, p. 78) Third, a simple although crucial observation is that whilst workers might ‘volunteer’ for part-time hours, they are most unlikely to ‘volunteer’ for the low pay and poor conditions that go with it. Part-time hours, low pay and poor conditions come as a package. Fourth, whilst these problems may be conceived of as bias in the sampling instrument, this conception severely understates the nature of the problem. Issues like those surrounding the subjective interpretation of questionnaires about why respondents do or do not want a full-time job are likely to be overlooked by the (typical) economist motivated by the desire to quantify and measure. In sum, the growth in part-time employment appears to constitute a series of qualitative changes in the nature, and a deterioration in the conditions, of employment.Arguing that this does not really matter because many part-time employees ‘volunteer’ for it constitutes a refusal to see beyond the level of the empirical.

Implications of critical realism This final part draws methodology and labour economics together to show the implications that the critical realist perspective has for the study of contemporary employment and unemployment. Recall that the use of deductivism and closed system analysis means that theoretical concepts have to be constructed in such a way as to reduce the multidimensional, qualitative reality of employment and unemployment, first to the level of events and second to the quantitative, single dimension of variables. As mere variables, however, these theoretical concepts cease to be adequate expressions of the reality they are designed to investigate.The following three examples demonstrate what it means to say that a variable, in this case ‘unemployment’, ceases to be an adequate expression of the reality it is designed to investigate. First, the reality of the employment experience for those in atypical employment is very different from the reality of those in typical employment within the same country.To treat one full-time job as equal to (say) two part-time jobs (even where the hours add up suitably to make a full-time equivalent) is to make the mistake of reducing quality to quantity – and losing something vital in the process. Any reduction in unemployment (assuming it results in a concomitant rise in employment) will have a differential impact upon workers in atypical and typical employment. Even a situation of full employment, should it occur in a period of significant atypical employment arrangements, would conceal vastly different employment experiences. Full employment in a country

44

Steve Fleetwood

where significant numbers were atypically employed, and where conditions of employment had deteriorated, might not be cause for celebration – although the reduced unemployment figures would look impressive. Second, the reality of the employment experience for those in atypical employment is very different from the reality of those in typical employment in different countries.The United States, United Kingdom and Spain, where atypical employment patterns are becoming increasingly significant, cannot be compared to countries like Germany where atypical work is (at the moment anyway) less significant. Even a situation of full employment, should it occur in countries experiencing significant levels of atypical employment, could not be said, unequivocally, to be a ‘better’ state of affairs than a situation of less-then-full employment in countries experiencing significant levels of typical employment.23 Third, the employment experience today, for those in atypical employment, is very different from the employment experience in previous periods when employment was largely typical. In April 2001 UK unemployment stands at around 5 per cent, similar to what it was in the mid-1970s. But given that in the early 1970s, atypical employment was not significant, the similarity of the employment experience evaporates, making the comparison of unemployment figures misleading. Full UK employment, should it occur in a period of significant atypical employment, could not be compared to full employment in the United Kingdom in (say) the 1950s where typical employment arrangements prevailed. The qualitative changes in the nature, and a deterioration in the conditions, of employment that have occurred recently, make comparison of unemployment in past, present and future periods highly problematic.24

Conclusion The opening quotation reiterates what appears to be the conventional wisdom: falling unemployment figures indicate that the (now) flexible UK labour market is gradually solving the problem of unemployment. Recognizing, however, that the very conceptualization, and hence measurement, of unemployment itself is inadequate, an alternative interpretation emerges.The problem of unemployment is not so much being solved as being transposed into a problem of employment, more specifically, into a series of problems relating to the emergence of atypical employment and the deterioration in the quality of employment it engenders. Moreover, armed with a set of methodological tools in the form of critical realism, one is in a position to see that mainstream economics cannot even begin to address this alternative interpretation. The deductivist method has placed qualitative issues like this out of the reach of mainstream economists.

Notes 1 This is a slightly amended version of a paper with the same title which appeared in Review of Social Economy Vol. LIX No. 1 March 2001.

Conceptualizing unemployment

45

2 For a positive elaboration of critical realism within economics see Lawson (1997) and Fleetwood (1995). For critical elaboration see Boylan and O’Gorman (1995); Parsons (1996). For recent developments and debates from various contributors see Fleetwood (1998). 3 Hume’s account of causality is not only founded upon constant conjunctions, it is also explicitly opposed to the alternative account of causality, namely, causality as powers adopted by critical realists.According to Hume:‘When I cast my eye on the known qualities of objects, I immediately discover that the relation of cause and effect depends not in the least on them. When I consider their relations, I can find none but those of contiguity and succession’ (1978: 77). Elsewhere he writes: ‘We have no other notion of cause and effect, but that of certain objects, which have been always conjoin’d together … We cannot penetrate into the reason for the conjunction (ibid., 93). Moving on, his rejection of causality as powers is as unambiguous as it is consistent: ‘The distinction, which we often make betwixt power and the exercise of it, is equally without foundation’ (ibid., 171). Once causality is based upon, and exhausted by, constant conjunctions of events the consistent empirical realist really has no option but to reject the notion of powers.According to Meikle, for the empirical realist: ‘There is no distinction to be drawn between a capacity and its exercise … To say that something can do something – that is, that it has a capacity to do it – is just to say that it does do it’ (1985: 114). In other words, if an entity generates some event in sense experience then the entity can be treated as causal: if it does not, it cannot. The empirical realist’s preoccupation with sense experience, denies any ‘space’, as it were, between a power and the events it causes. And yet it is only in virtue of this ‘space’ that the notion of causality as power (and, incidentally, the notion of tendency) is intelligible at all. For Hume’s work on causality see Hume (1978: 73–94 and 155–172, and for a critical discussion see Bhaskar (1978: chapters 1,2 and appendix); Meikle (1985 especially chapters 1 and 7) and Cartwright (1995). 4 Deductivism cannot, consistently, be rooted in anything other than empirical realist ontology; and conversely, empirical realist ontology cannot, consistently, engender anything other than deductivism, or at least something similar. Inconsistency can, of course, lead to any combinations of ontology, causality and modes of reasoning. 5 The same applies, of course, to the event y’s. 6 Whilst the precise specification of the variable used to capture employment varies (e.g. based upon employers surveys, labour force surveys, national accounts) all that is needed for our purposes is a generic category such as ‘jobs filled’. Specifying the variable more precisely would not effect the basic argument of this chapter. 7 Take for example, Layard et al.’s (1992) highly influential work on unemployment and the labour market. Is it not odd that such an extensive work on the labour market ignores atypical employment? Whilst the authors cannot possibly be unaware of this phenomenon, there is no way that such a qualitative, multi-dimensional reality can (meaningfully) be reduced to the single dimension of a variable and accommodated within the main equations they use to investigate unemployment. Their attachment to deductivism sets a truncated, and thereby inadequate, theoretical agenda. Rice (1990) is typical of the (few) treatments of atypical employment by mainstream economists: part-time employment is reduced to a variable and treated as a function of another variable, in this case, national insurance contributions. 8 Even here, though, measurement does not reveal things like the attitude of the worker towards (say) the second job. And this is likely to have serious implications for issues that are of interest to labour economists such as the investigation of things like productivity levels in second (and increasingly third) jobs. 9 It is worth bearing in mind that anything and everything can, in a sense, be measured by proxies and indices. No doubt one could ‘measure’ the beauty of the Mona Lisa with suitable statistical devices. But, would such a measure actually mean

46

10

11

12

13

14

15

Steve Fleetwood anything? My suspicion is that in their desire to quantify, many mainstream economists fail to see that their measures are often meaningless. For example, Shapiro and Stiglitz (1990; 48) use the variable q as a measure of the ‘probability of being detected shirking’.Whilst intuitively the notion that one might get caught shirking is a sensible observation, reducing this complex, multi-dimensional socio-psychological notion to the single dimension of a variable destroys the sense of the observation and makes the variable meaningless. The same could be said for the common view amongst labour economists such as Booth (1995; 109 passim) that trade union power can be measured (amongst other things) by the level of membership. (See Fleetwood, 1999 for a critique of this view). Here again complex, multi-dimensional socio-political phenomena are reduced to the single dimension of a variable so that they can be measured, whereupon they lose virtually all meaning. Other methodological perspectives may be of some assistance here, but arguably, to paraphrase a well-known lager commercial, critical realism reaches the parts other perspectives cannot reach. For example, whilst Boylan and O’Gorman’s (1995) causal holism might be usefully deployed to highlight the lack of descriptive adequacy in the mainstream treatment of atypical employment, it explains neither the preoccupation with quantification, nor why such quantification is inappropriate. Note that the discussion of critical realism carries an important message for many Austrian, Institutionalist, Marxist and Post Keynesian economists who, whilst critical of neoclassical economics, are nonetheless unwilling to break completely with deductivism. For example, a paper by Lieberman and Jehle in the Journal of Post Keynesian Economics uses the deductivist method, closed systems and the reduction of complex aspects of the employment experience to variables – see especially their fn 3. Despite the authors’ progressive intentions, methodologically speaking there is no difference between their approach and that of neoclassical economics. This is in contrast to the quantitative relation such as the recent conundrum where the number of unemployed is falling but the number of employed is not increasing at the same rate. Even here, one report recognized observed:‘[o]ne thing these figures can’t tell us, however, is how secure these jobs are’ (EPI 1995). In other words, quantitative forms of analysis can’t tell us much about many of the qualitative aspects of jobs. A symposium on unemployment (Glynn and Mayhew, 1995) actually carries a paper on qualitative aspects of employment (Gregg and Wadsworth, 1995), but the connection between the qualitative aspects of employment and the conceptualization and measurement of unemployment is overlooked. Whilst recent editions of Labour Market Trends, in particular an article by Perry (1996), recognize the existence of ‘flexible work arrangements’, and the LFS now has data on seasonal contracts, fixed term work, agency work and causal work, again the connection is omitted. Other publications where one might expect, but do not find, a discussion of this connection include: Metcalf (1992); Coates (1995) and Meadows (1996). Blank (1990: 123) is one of the few to touch upon the connection, noting that at ‘its worst, part-time work may be considered a form of disguised unemployment’. The idea that unemployment is the other, or absence, of employment, a job, or work is implicit in the various official definitions of unemployment – that is, economic activity/inactivity, employees in employment, claimant/LFS unemployed and so on. In a section entitled ‘The definition of unemployment’ Briscoe and Johnson note how the definition of unemployment rests upon ‘more precise criteria’ such as ‘seeking work, wanting a job; being available for work; and not working’ (1995; 104). Unfortunately, head counting constitutes a significant proportion of the research done on employment by economists. According to Dilnot:‘Economists have many failings but one of the most damaging is the desire to summarize in a single number [i.e. a variable SF] some large and complex part of the economy.That weakness is

Conceptualizing unemployment

16

17

18

19 20

21 22

23

47

often seen in discussions of the labour market, when some single measure is used to capture the supposed reality. In the case of the labour market, levels of unemployment are often singled out’ (1996: 14). If critical realism is correct, the reason for this failure is rooted in the use of an inappropriate method – deductivism. It should be noted that whilst these issues are debated in the pages of Sociology, HRM, Industrial Relations, Organizational Behavior and Management journals, the debate is virtually absent in Economics and Labour Economics journals. Arguably, this is because in these former subjects deductivism is not dominant. Note that in choosing the example of part-time employment I am choosing the most difficult, yet most powerful, case for my argument because this form of atypical employment bears the closest resemblance to typical employment. The advantages of using the most difficult case are two-fold. First, if I can establish that difficulties arise in adequately conceptualizing part-time employment on account of its qualitative and multi-dimensional nature, these difficulties will be multiplied for those other forms of atypical employment that bear little or no resemblance to typical employment. Second, if I can establish that the emergence of part-time employment constitutes a deterioration in the conditions of employment relative to typical employment, then the conditions of employment associated with forms of atypical employment that bear little or no resemblance to typical employment will constitute a far worse deterioration. As one would expect, the highest paid part-time occupations are professionals, with average hourly earnings (1995) of £13.33, that is, about 20p higher than the equivalent full-time statistic (Osborne, 1996: 321). Comparison between this and a male kitchen porter or a female dental nurse’s average hourly earnings of £3.82 and £4.22 respectively.Wood (1995) gives some indication of the differences within the category ‘part-time’. I refrain from discussing the issue of whether part-timers earn low pay because of their alleged low productivity or because of the nature of the job (cf. Blank, 1990). Empirical evidence is, however, mixed. McGregor and Sproull (1992) asked employers to assess comparative rates of turnover for full and part-timers and found that, in general, there was little difference. Tilly (1992: 23) shows that average job tenure for part-timers in the United States is 3.4 years, compared with 5.7 years for full-timers. Natti (1995: 351) shows that the average job tenure for part-time women in Finland is 5.2 years compared with 8.5 years for full-time women, although in Sweden the gap was minor. Penn and Wirth (1993: 257 and 263) found higher turnover of part-time staff in Sainsbury’s and Marks and Spencers.According to the Employment Gazette (March 1993; 91), in the Summer of 1992, part-time employees had, on average, been with their current employer for a shorter period than full-time employees. An IPD (1995) survey revealed labour turnover rates for 1994 as follows: full-time and part-time manual 12 and 33 per cent; part-time and full-time non-manual 14 and 31 per cent respectively. That the LRD survey found ‘nearly half of all temps are employed part-time only’ (1995: 4) makes these figures a little more relevant to part-time employment only. This raises the vexed issue of what exactly constitutes skill, its conceptualization and measurement. Levels of skill might depend not upon factors such as human capital (which can, allegedly, be measured) but upon factors such as power through which certain individuals are able to define employment as skilled or unskilled (which is probably impossible to measure meaningfully). It is, of course, extremely common to find data comparing unemployment in various countries. For example, Labour Market Trends (1997: 538) compares unemployment rates in the United Kingdom, European Union, European countries, Australia, Japan and the United States. Barrel et al. (1997) compare job creation in the United States and Europe. Such quantitative studies inevitably fail to recognize the more fundamental methodological problems that arise with attempts to quantify qualitative phenomena.

48

Steve Fleetwood

24 Dex and McCulloch (1995: 55) are representative of those who do recognize problems when attempting to quantify changes in forms of atypical employment over time. Whilst they recognize the (not inconsiderable) ‘technical’ problems of constructing a series over time (e.g. the sources of the data were unreliable in the past or sources have changed over time) they fail to recognize the more fundamental methodological problems that arise with attempts to quantify qualitative phenomena.

References Balchin, A. (1994) ‘Part-time workers in the multiple retail sector: small change from employment protection legislation’, Employee Relations, 16(7): 25–32. Barrel, R., Lansbury, M., Morgan, J. and Pain, N. (1997) ‘Job creation in the US and Europe compared’, Labour Market Trends, May: 173–174. Bhaskar, R. (1978) A Realist Theory of Science, Brighton: Harvester Wheatsheaf. Blank, R. (1990) ‘Are Part-time Jobs Bad Jobs’ in G.A. Burtles (ed.) Future of Lousy Jobs?:The Changing Structure of U.S.Wages,Washington DC: Brookings Institute. Booth, A. (1995) The Economics of the Trade Union, Cambridge: Cambridge University Press. Boylan,T. and O’Gorman, P. (1995) Beyond Rhetoric and Realism in Economics, London: Routledge. Briscoe, C. and Johnson, S. (1995) Measuring the Economy, Harmondsworth: Penguin. Burchell, B., Day, D., Hudson, M., Lapido, D., Mankelow, R., Nolan, J., Reed, H., Wichert, I. and Wilkinson, F. (1999) Job Insecurity and Work Intensification,York: Joseph Rowntree Foundation. Cartwright, N. (1995) ‘Ceteris paribus laws and socio economic machines’, The Monist, 78(3): 276–294. Casey, B. (1988) Temporary Employment: Practice and Policy in Britain, Policy Studies Institute. ——, Metcalf, H. and Milward, N. (1997) Employers Use of Flexible Labour, London: Policy Studies Institute. Coates, K. (1995) The Right to Work, Spokesman. Convey, P. (1996) ‘How many people are unemployed’, Working Brief, October. Dex, S. and McCulloch, A. (1995) Flexible Employment in Britain: A Statistical Analysis, Equal Opportunities Commission. Dilnot, A. (1996) ‘Is the labour market working’, RSA Journal CXLIV (5467): 14–19. Employment Gazette (1993). Employment Gazette (1995) February. EPI (1995) Employment Policy Institute, Economic Report 9(2). Ewing, K. (1996) Working Life:A New Perspective on Labour Law, London: Lawrence and Wishart. Fleetwood, S. (1995) Hayek’s Political Economy: The Socio-economics of Order, London: Routledge. —— (ed.) (1998) Critical Realism: Developments and Debates, London: Routledge. —— (1999) ‘The Inadequacy of Mainstream Theories of Trade Union Behaviour’, Labour, 10(2): 445–480. Fothergill, G. and Watson, B. (1993) ‘Part-time employment and attitudes to part-time work’, Employment Gazette, May. Glynn, A. and Mayhew, K. (1995) ‘Symposium on unemployment’, Oxford Review of Economic Policy, 11(1): 1–136. Gordon, S. (1991) The History and Philosophy of Social Science, London: Routledge.

Conceptualizing unemployment

49

Gregg, P. and Wadsworth, J. (1995) ‘A short history of labour turnover, job tenure and job security’, Oxford Review of Economic Policy, 11(1): 73–90. Hetherington, P. (1995) Guardian, November. Hughes, J. (1995) ‘Measuring Unemployment in Britain’ in K. Coates (ed.) The Right to Work, Nottingham: Spokesman. Hume, D. (1888; 1978) A Treatise of Human Nature, Oxford: Clarendon Press. IDS Study (1993) ‘Part-time workers’, 540 October. IPD (1995) Labour Turnover: Survey Results. Klein, D. (1996) Monthly Labour Review, October. Labour Force Survey (LFS) Historical Supplement (1996) Government Statistical Service. Labour Market Trends (1997) Labour Market Trends, June. —— (1998) Labour Market Trends, January. Labour Research (1994) ‘Do part timers have equal rights?’, Labour Research, 83(7): 8–10. Labour Research Department (1995) Temporary Workers and the Law. —— (LRD) July (2000) Part-Time Workers: A Legal Guide for Workplace Reps. Labour Market Trends December 1999. Lawson,T. (1997) Economics and Reality, London: Routledge. Layard, R., Nickell, S. and Jackman, R. (1992) Unemployment: Macroeconomics Performance and the Labour Market, Oxford: Oxford University Press. Lieberman, M. and Jehle, G. (1997) ‘On Fringe benefits and layoffs’, Journal of Post Keynesian Economics, 20(2): 295–299. McGregor, A. and Sproull, A. (1992) ‘Employers and the flexible workforce’, Employment Gazette, May. Meadows, P. (1996) Work Out- or Work In: Contributions to the Debate on the Future of Work, New York: Joseph Rowntree Foundation. Meikle, S. (1985) Essentialism in the Thought of Karl Marx, Gloucester, Duckworth. Metcalf, H. (1992) ‘Hidden Unemployment and the Labour Market’, in E. McGlaughlin (ed.) Understanding Unemployment: New Perspectives on Active Labour Market Policies, London: Routledge. Natti, J. (1995) ‘Part-time work in the Nordic countries: a trap for women’, Labour, 9(2): 343–357. Osborne, K. (1996) ‘Earnings of part-time workers: data from the 1995 new earnings survey’, Labour Market Trends, May. Parsons, S. (1996) ‘Post Keynesianism, Realism and Keynes’ General Theory’, Journal of Post-Keynesian Economics, 18(3): 419–443. Pencavel, J. (1994) Labour Markets Under Trade Unionism, London: Blackwell. Penn, R. and Wirth, B. (1993) ‘Employment patterns in contemporary retailing: gender and work in five supermarkets’, The Service Industries Journal, 13(4): 8–16. Perry, K. (1996) ‘Measuring employment: comparison of official sources’, Labour Market Trends, January. Polivka, A. (1996) ‘Contingent and alternative work arrangements, defined’, Monthly Labour Review, October. —— and Nardone, T. (1989) ‘On the definition of “Contingent Work’ ”, Monthly Labour Review, December. Pollert, A. (1991) ‘The Orthodoxy of Flexibility’, in A. Pollert (ed.) Farewell To Flexibility, London: Blackwell. Rice, P. (1990) ‘Relative labour costs and the growth of part-time employment in British manufacturing industries’, Economic Journal, 100: 201–220.

50

Steve Fleetwood

Roosenthal, H. (1989) ‘More than wages at issue in job quality debate’, Monthly Labour Review, December. Sengenberger,W. (1996) ‘Full employment: past, present and future – an ILO perspective’, paper presented at the Conference of the European Association for Evolutionary Political Economy, Antwerp. Shapiro, C. and Stiglitz, E. (1990) ‘Equilibrium unemployment as a worker discipline device’, in G. Akerlof and J.Yellen (eds) Efficiency Wage Models of the Labour Market, Cambridge: Cambridge University Press. Social Trends (1998) National Statistical Office. Tilly, C. (1992) ‘Dualism in part-time employment’, Industrial Relations, 31(2): 330–347. TUC Economic & Social Affairs Dept. (1996) All in the Mind? Job Insecurity in Britain Today,TUC. Warme, B., Lundy, K. L. P. and Lundy, L. A. (1992) Working Part-time: Risks and Opportunities, Praeger. Wells, J. (1995) ‘The Missing Million’, in K. Coates (ed.) The Right to Work, Spokesman. Wood, L. (1995) Financial Times, 14 September.

Websites www.statistics.gov.uk/statbase/xdataset.asp www.statistics.gov.uk/nsbase/downloads/theme_labour/nes.pdf

4

Critical realism, methodology and applied economics1 Bernard Walters and David Young

Introduction Critics of standard/orthodox economics are often invited to offer some clear alternative. Or to put it a little more critically, the fact that they often fail (or are perceived to fail) to offer some alternative is taken by many mainstream economists (and others) to imply that their criticisms of orthodox economics are of little import. For even if some of the criticisms are able to elaborate some substantive shortcomings, in the absence of an inviting alternative there is an argument for continuing as before. The veracity of a critique of mainstream economics does not, of course, stand or fall on whether a convincing alternative can be provided. For example, if expected utility models are demonstrated to be descriptively inaccurate then such models suffer from this shortcoming whether an alternative theory of behaviour is presented or not. But one obvious course of action for non-mainstream economists is to attempt to spell out an alternative. This has often involved the elaboration of different theoretical perspectives; but it seems increasingly the case that there is a desire to provide some corresponding alternative in applied areas. So, for example, if a critique of orthodox economics disallows or downgrades econometrics, what alternative applied work are economists to engage in? In addressing such issues some non-mainstream economists have sought support and guidance from critical realism.The hope appears to be that in offering an alternative view of science, critical realism can supply both a critique of standard economics as based on a fallacious model of science and a basis for an alternative economics, which accords with a transcendental realist view of science.The appeal of critical realism is perhaps enhanced by its concern with realism or more specifically real casual mechanisms, for many alternative/heterodox economic theories are critical of the orthodoxy partly because the latter is seen as unrealistic, or socially irrelevant in some way. (So, e.g. Austrian, Post Keynesian and radicals all argue in different ways that their approach/framework is more ‘realistic’ than mainstream economics.) To this extent the appeal of critical realism seems clear. Extending arguments which the authors have developed previously, the main purpose in this chapter is to argue that while critical realism raises some important criticisms of mainstream economics, it neither provides a general

52

Bernard Walters and David Young

philosophical argument for rejecting all orthodox approaches, nor does it provide a sound basis for constructing an alternative. Our main focus will be on applied economics or particular areas of the application of economic theories. It will be argued that the general strictures which critical realism wishes to place on applied economics and on the application of economic theory within the policy area are not sustainable. Critical realism does not provide a convincing account of the interface between theory and evidence.This is not to say that there are not a number of important critical ideas that need developing with respect to standard applied economics or the construction of economic policy. But few, if any, of these ideas require the approach of critical realism and moreover the latter would unhelpfully restrict the applied activities of economists and their contributions to policy formulation. The chapter proceeds as follows. In the next section the principal features of critical realism are outlined. This is followed by a discussion of the critical realism account of mainstream economics in section ‘The critical realist view of mainstream economics’. ‘The epistemological problems with the realist method’ section provides a critical evaluation of some of critical realism’s main arguments and highlights some of the epistemological and theoretical problems which this approach encounters. The chapter then proceeds to consider the difficulties of critical realism’s treatment of applied work and econometrics. Briefly, this section argues that the critical realist approach is overly prescriptive and in particular its critique of econometrics is ill-founded. A more plural approach to evidence which would include econometrics as one part of an attempt to provide empirical support for competing theoretical explanations is suggested. These ideas also have implications for policy analysis and policy choice. Some of these are explored in the penultimate section.

The chief features of critical realism Critical realism is the extension to the social sciences of the transcendental realism associated particularly with Bhaskar (1978, 1979, 1989). Critical realists accept the transcendental realist view of science and the auxiliary claim of naturalism, that there is a unity of method between the natural and social sciences. However, the naturalist claim is moderated by the acceptance of real differences in the objects of study between the natural and social sciences.Therefore, there are two levels of methodological prescription: first, those arising from the transcendental realist approach to science; and second, modifications arising from the transposition of these ideas to the social realm. The transcendental realist view of science accepts the force of the criticisms of the self-confident positivism of the early twentieth century.These criticisms turn on the difficulty, and perhaps the impossibility, of establishing a mapping from empirical evidence to the structure of scientific theories which is sufficiently watertight to establish observation evidence as the decisive arbiter of scientific progress. At the same time, the transcendental realists do not wish to be drawn to the methodological diversity which characterises recent

Critical realism, methodology and applied economics 53 postmodern contributions. In short, transcendental realism attempts a middle path between empiricism and a perceived relativism. It steers this course by inverting the normal approach to questions of the nature of scientific knowledge. Instead of asking the epistemological question ‘how can we scientifically know about the world?’ it asks the ontological question ‘what must the world be like if we are to take seriously what science seems to reveal about the world?’. The answer it gives is that the mechanisms of scientific theory are real. This means that it finesses the difficulties of mapping the relationship between scientific theory and empirical evidence by positing the reality of intervening mechanisms. The central claim of this approach is, therefore, ontological. It is the claim that the objects of scientific investigation are real in the sense of existing independently of the observer and that this ontology has depth; the reality is not limited to a surface appreciation of the immediate objects of experience. The world is therefore conceived as layered, with critical realists distinguishing a surface world of empirical experience, an actual world of events and a nonactual world of underlying powers, mechanisms and generative structures. Furthermore, these levels of reality are conceived as being typically out of phase so that, in general, surface experience does not reveal the non-actual world of powers, mechanisms and generative structures. In fact, such an ontology may very well be descriptive of scientists’ ways of thinking about the world and, in itself, seems unlikely to have much methodological significance. However, crucially, to this description is added the claim that this layered reality is knowable by consciousness and may be uncovered by a process of retroduction. Retroduction or abduction describes the process by which the ‘deep’ structure of reality is uncovered by scientific investigation. It is an iterative process in which mechanisms are initially hypothesised which explain the phenomena in question.The origin of the hypothesised mechanisms can often be through borrowing from other areas of study on the basis of analogy or metaphor.The postulated mechanisms must then be ‘empirically ascertained’.2 The generative mechanisms of the preferred explanation now become the new phenomena to be explained at the next stage.The process is therefore circular and cumulative. Importantly, as critical realists are fallibilists they distinguish between the transitive objects of knowledge, our present theories of the underlying structure of reality, and the intransitive objects of knowledge which constitute the mechanisms that science struggles to uncover.3 It is clear that this process attempts to incorporate the processes by which scientists attempt to generate and explain their hypotheses.4 This is in contradistinction to the standard philosophical positions of inductivism and deductivism which both have justification rather than description as their objective. However, despite the protestations of critical realists, it clearly has similarities to the hypothetico-deductive approach. But, unlike that approach it suggests that the objective of science is the postulation and elucidation of the essential mechanisms lying behind surface reality. This switch in objective is perhaps the most crucial methodological feature of

54

Bernard Walters and David Young

transcendental realism because it loosens the pre-eminence of empirical evidence in determining the scope and validity of scientific activity.5 This characteristic is even more pronounced in critical realist prescriptions. Critical realism acknowledges the claim of naturalism, so that the fundamental layered structure of social reality with a surface, actual and non-actual structure is accepted. However, critical realism differs from transcendental realism because it must take account of the distinctive nature of the objects of social science.6 The objects of social science are historically contingent, and the underlying structures, powers and mechanisms may change over time. Furthermore, these structures are constituted, albeit perhaps unknowingly, by human beings with understanding, freedom and independent agency who invest their lives with meaning. Despite this bow towards the hermeneutic tradition it is, nevertheless, asserted that society can still be analysed causally in law-like terms, but that ‘social laws may be (although still universal within their range) historically restricted in scope (Bhaskar, 1979, p. 165). The deduction, which is made, is that the lack of a method of isolating the causal structure disallows certain methods of investigation which have a legitimacy within the natural sciences. In particular, societies are conceived as open structures within which experiment is impossible so that there is no way of restricting the range of causal influences sufficiently to generate a closed system with which to demonstrate particular causal mechanisms or to make sensible predictions.This means that the methods of ‘empirically ascertaining’ theoretical constructions are much more limited than in the natural sciences. Specifically, it is suggested that the ‘… criteria for the rational assessment (replacement, acceptance, rejection, modification, diversification, and development) of theories cannot be predictive. And so must be exclusively explanatory’ (ibid., p. 163). Nevertheless, the method of uncovering the social laws retains the basic form of retroduction within the natural sciences. The methodological prescription that follows from the critical realist approach is therefore relatively straightforward.The objective of social science is the revelation of the real underlying mechanisms and powers that govern our social world.These mechanisms and powers are hidden by a constantly changing open system that disallows experimentation and vitiates methods that attempt to proceed by picking out surface correlations. This means that the elucidation of the causal structure of our world is exclusively explanatory, with the social scientist proceeding by postulating essential mechanisms sufficient to explain the perceived phenomena and then seeking empirical support for these postulated mechanisms. However, in view of the impossibility of particular events in an open world revealing its causal structure, the empirical support takes the form of explanation in terms of the stylised facts of experience. Finally, a preferred theory, while always provisional in view of critical realists’ fallibilism, would be one which had greater explanatory power than an alternative against the stylised facts of experience. The key question this account raises is whether such an approach has the epistemological power to provide a prescriptive model of scientific activity.

Critical realism, methodology and applied economics 55 Although the central claims of critical realism are ontological its methodological prescriptions depend on the epistemological significance of the proposed methods of uncovering the deep structures of the social world. These prescriptions are not unproblematic, yet this approach to social scientific research gathers support because it is contrasted with an alternative methodology which is claimed to represent a positivist orthodoxy. As we have argued in some detail elsewhere (e.g. Walters and Young, 2001) this characterisation is misleading not only in terms of its treatment of positivism but also in its claim that the fundamental problem of orthodoxy originates in the Humean account of causality.The critical realist argument is that Hume’s account of causality is fundamental to orthodox theory as it represents a simple empiricist view of the world in which event regularity is the basis for scientific enquiry. Such a strong association between event regularity and positivism is, in itself, misleading. Positivism is more concerned with the confidence that agents attach to observed regularities than the truth content of associated statements. The observation of patterns does not in itself commit us to any particular view of their significance, nor to any particular interpretation of their meaning. Moreover, Hume’s main concern was to establish the argument that a view of causation should be firmly grounded in what we can observe. This does not imply, however, that either Hume or modern Humeans believe that the world accords with some simple, strict event regularity. Indeed, much of the modern Humean account of causality devotes considerable time to complex problems in establishing causal accounts.

The critical realist view of mainstream economics In the critical realist literature (see Lawson, 1994) mainstream neoclassical economics is clearly viewed as positivist. However, as we have argued in Walters and Young (2001) the degree to which neoclassical theory can be labelled positivist either in the realist sense or even according to a more restricted definition is contentious. The characterisation of economic orthodoxy as a monolithic approach united by its methodological preconceptions is seriously misleading. For instance, the idealism espoused in Stigler and Becker (1977) is clearly a different approach from the instrumentalism recommended by Friedman’s (1953) well-known suggestions for a positive economics. Neoclassical economics is highly diverse in ways that reflect significant differences of methodological view and practice. For example, some recent work in macroeconomics pay little attention to acquiring econometric support. The prime example would be the increasing use of calibration. On the other hand, other areas have often sought to differentiate between competing theories on the basis of econometric evidence. In addition to this, there remains a lively tradition of case studies and survey methods, although this has become relatively less important in mainstream economics and, as we shall argue later, this might be to its detriment. However, the type of evidence which is used in mainstream research still displays considerable diversity. The critical realist view disregards

56

Bernard Walters and David Young

such diversity on the grounds that all of this work displays a common feature which dominates other differences. The crucial point regarding critical realism’s characterisation of mainstream theory is the identification of a commitment to a Humean event regularity view of the world and in particular a deductivist method of investigation. Whatever the particular form of theoretical or applied work, it purportedly displays the common characteristic, which critical realism identifies with positivist views of science. Moreover, critical realism would emphasise that such a deductivist approach relies upon the presumption of a closed system which is generally invalid, particularly in social science. This claimed homogeneity extends from considerations of mainstream theory to the use of econometrics on the grounds of their shared deductivism.

Epistemological problems with the realist method Critical realism’s central preoccupation is ontological. Even here it is worth noting that its claims may be problematic because the ontological assertions appear to allow no scope for considering alternative claims.The key question for methodology, though, is whether there is a critical realist epistemology that plausibly supports its ontological assertions. In short, are there robust rules capable of guiding researchers in their attempts to uncover the deep structures of reality? It is clear that there must be some appeal to empiricism in the broadest sense in order to provide any support for the claim that certain structures or mechanisms are real. This is acknowledged by Keat and Urry (1982, p. 233) who ‘suggest that, in this minimal sense of the term, TR (theoretical realism) can properly be regarded as empiricist as can positivism’.Their minimal definition consists of the following claims ‘that the truth or falsity of scientific theories cannot be established by a priori argument alone …, that scientific theories must be evaluated by reference to their consistency or inconsistency with the perceptual information acquired by humans …, and that there is, therefore, a central role in scientific enquiry for the testing of theories in terms of their consequences for perceptually determinable states of affairs’. Keat and Urry emphasise however, that realists do not subscribe to further claims regarding empirical evidence which, they suggest, are routinely advanced by ‘empiricists’. These further claims concern the relationship between empirical observation and the truth content of theories, and have been largely abandoned even by social scientists who assign a prime position to empirical evidence. It is true, of course, that realists generally do not place such heavy emphasis on empirical observation, but rather stress the importance of non-observable structures and mechanisms. However, if we are to take empirical evidence at all seriously then it is impossible to avoid the numerous epistemological difficulties that critical realism has attempted to circumvent by claiming that realism is essentially an ontological issue.The problem for critical realists is that it accepts at least some form of empiricism without addressing the issue of how to use empirical

Critical realism, methodology and applied economics 57 evidence.The epistemological criteria within critical realism for deciding how empirical evidence may be used to arbitrate between competing theories are weak or missing. There have been few epistemological discussions by critical realists and none have successfully provided a convincing alternative to standard approaches to the treatment of evidence. For example, Bhaskar’s (1979) attempt to produce a four-phase process for explanation in open systems provides little specific guidance that would enable a researcher to evaluate the merits of a particular theory. More recent attempts have addressed the difficulties posed for critical realism by its need to address the role of empirical evidence and have attempted to elaborate the epistemological implications of their ontology. For example, Lawson also acknowledges that critical realism’s explanatory project ‘necessarily contains a significant empirical component’ (1997, p. 221). He argues, however, that critical realism’s interrogation of empirical evidence is fundamentally different from that of mainstream economics. In particular, Lawson elaborates a distinctive critical realist position based on a social world characterised by ‘demi-regularities’ (demi-regs). A demi-reg is a ‘partial event regularity which prima facie indicates the occasional, but less than universal, actualisation of a mechanism or tendency, over a definite region of timespace…[But] where demiregs are observed there is evidence of relatively enduring but identifiable tendencies in play’ (ibid., p. 204). Such regularities, Lawson contends, stand in direct contrast to the strict regularities of the Humean presumptions of orthodox economics. Lawson asserts that strict event regularities are ‘essential to the requirements of orthodox economics. That is, they are essential to the preoccupations with predicting future states of the economy and/or to any hoped for deductivist “explanation” of economic actualities’ (ibid., p. 219).This attempt to demarcate critical realism from orthodox practices involves a tendentious characterisation of orthodox economics and as well as an unjustifiable confidence in the critical realist approach. It involves characterising all theoretical and empirical propositions of mainstream economics as deriving from strict event regularity. As we have seen this picture of mainstream economics is poorly supported. A major implication of the proposition that social and economic life exhibits a variety of demiregs but no strict event regularity is, as noted earlier, that ‘event/predictive accuracy cannot be the criterion of theory selection. Rather the appropriate criterion … must be explanatory power’. Theories, it is argued, may therefore be assessed, and presumably ranked, in accordance with their ‘abilities to illuminate a wide range of empirical phenomenon. And typically this will entail accommodating … contrastive demiregs’ (1997, p. 213). The concentration on explanatory power as the exclusive criterion of theory assessment has a number of well-known problems (see Baert, 1996). In particular, on this criterion general non-falsifiable theories would appear to have the greatest explanatory power and be accepted on this ground alone. Moreover, if contradictory theories have the same explanatory power there is no obvious way to evaluate their competing claims. For the idea of explanatory power to

58

Bernard Walters and David Young

be useful it must be associated with a method of establishing the evidential space. In addition, the causal explanations that critical realism proposes may often involve reference to unobservable but real variables and this raises an associated difficulty of restricting the number of possible theoretical explanations; this point is forcefully argued in Boylan and O’Gorman (1995). It is worth noting that these problems of interpreting empirical evidence suffer from the same objections, concerning the mapping of observations to the underlying structure of theory, that undermined the early positivist project. Empirical evidence is incorporated, but in a way that does not allow us to discriminate between rival accounts. Critical realists operate as though they are absolved from having to provide answers to the epistemological difficulties faced by orthodox theory. But their appeal to a layered ontology is insufficient to bypass consideration of the mapping from the observation evidence to the unobserved powers and structures. Runde (1998) provides a discussion of the epistemological problems posed for causal accounts that may be considered to lend support to critical realism. Although he accepts the possible subject-specific elements of explanation, he, nevertheless, argues that there are ‘four broad fronts on which an adequate causal explanation should be robust’ (ibid., p. 157).These are, first, that possible causes of an event should be, as a matter of fact, aspects of the situation in which the event occurred. Second, given this set of possible causes, that relevant causes must be operative. Third, identified causes must be sufficient to provide a satisfactory explanation of the event. Finally, the causes cited must provide causal depth, by which is meant that they should not be dominated by or be a result of some deeper or more general cause (ibid., pp. 158–162).These general characteristics are deficient in two respects. First, such criteria are not sufficient to delineate different types of causal accounts. In particular, contrary to Runde’s claim (ibid., p. 166), many neoclassical propositions are compatible with the criteria outlined (see above). Second, his discussion fails to establish a method of determining the validity of any particular causal story. In other words, it fails to specify how particular observations of the world determine or substantiate aspects of theory.This is acknowledged by Runde, who concludes that ‘the framework principles used to assess causal explanations are not sufficiently specific, and will often not have specific bite, to discriminate unambiguously between competing causal explanations’ (ibid., p. 168). These authors would argue it is equally difficult to distinguish between competing orthodox and critical realist causal explanations. Critical realism presupposes that by presenting a distinctive ontology, critical realism can produce an approach to empirical evidence ineluctably distinct from the orthodox method.This has not been established. Moreover, the assumption of insulation from standard epistemological problems is vitiated by a failure to establish a clear link from the ontological domain to the empirical domain. Indeed, the critical realist account only establishes itself as a plausible alternative by presenting a caricature of mainstream practices of interpreting evidence.

Critical realism, methodology and applied economics 59 Moreover, the issue of interpretation in social science might usefully receive greater attention from critical realists. As mentioned previously critical realism argues that its basic approach is as appropriate in social science as it is in natural science and, although it is acknowledged that in the social realm there may be more problems of interpretation, discussion of this is subordinate to the concern with open systems. By contrast, the authors would suggest that this emphasis ought to be reversed.That is to say, critical realism needs to take more seriously the problems of interpretation specific to social science and base fewer of its claims on open-systems problems which are almost universal. In particular, critical realism might do well to consider some of the issues raised by hermeneutical approaches. For example,Taylor (1985) notes three reasons why it may be that ‘exact prediction is radically impossible’ (p. 55) namely; ‘open-system’ problems, the sensitivity of the future entities to relatively small differences in interpretations (which cannot be measured or judged as in the national sciences) and third, the nature of humans self-defining beings. Taylor (and others) regard the last of these as the most fundamental issue. People understand themselves differently as a result of some conceptual innovation, and this ‘new’ understanding may not be describable in terms of our previous concepts and language. This is bound up with the problem of prediction, for as Taylor comments, it is ‘only if past and future are brought under the same conceptual net can one understand the states of the latter as function of the states of the former, and hence predict. This conceptual unity is vitiated in the sciences of man by the fact of conceptual innovation which in turn alters human reality.’(p. 56).Thus, there may be many future events which are radically unpredictable though sometimes there are grounds for discerning change in particular directions. As history illustrates social, political and also economic evolution is often surprising and only understandable ex post. The appeal by Taylor (and others) for a hermeneutical approach to the social sciences acknowledges the ‘open-system’ problem for providing reliable prediction, but this is regarded as much less significant than the problem of interpretation just outlined. And this, the authors would argue, is the correct emphasis. One cannot ground much of our arguments against prediction on the ‘open-system’ problem. As Taylor (1985) notes, it applies to meteorology as much as to the social sciences; and is it not the case that we might (perhaps often do) find weather forecasts to be adequately accurate and useful? The ‘open-system’ argument however valid, would not seem to be of much importance compared to the problem of interpretation in social and economic theory. It is undoubtedly correct that we can never delineate a closed system which is absolutely shielded from external effects but this does not negate the usefulness of approaches which assume some degree of closure.The issue rather is the balancing of rival interpretations and consideration of the different forms of evidence which influence our judgement of the adequacy of different theories.This, rather than discussing closed system versus open system theories, should be our main concern.

60

Bernard Walters and David Young

To return to the meteorological example. Sometimes, of course, weather forecasts are spectacularly inaccurate but much of the time the meteorological models which are used are adequate to provide fairly accurate publications over a short time-period.The same might be true of some models in the social sciences. They may predict reasonably well most of the time over a fairly short time-horizon, but are sometimes found wanting. Economic or econometric models can and do provide such predictions, and just like the meteorological they contain closure assumptions. The latter is no basis on which to reject wholesale the usefulness of such models as critical realism seems to imply. Instead we should focus our attention on the rival interpretations of what we observe, which perhaps much more than in the natural sciences is likely to be of great significance in economics where there is a lack of clarity about our observations and their meaning.

Critical realism, econometrics, applied economics and policy It is clear from the earlier discussion that critical realism provides some role for empirical evidence, although we would argue that the nature and weight of such evidence is unclear.This is a crucial area for critical realism as it wishes to disallow the use of econometrics but argues that empirical evidence is still important in assessing theory. Lawson (1997, p. 69) states ‘…for example, the determination of summary statistics of a body of data such as means or growth rates of some phenomenon can be included under the rubric of econometrics. Such activities are not being questioned here… .The specific realist emphasis that I am pursuing is not an anti-empirical one’. There are, of course, a number of objections to standard econometric approaches. One objection is that they incorporate a particular theoretical structure which goes unquestioned in the light of evidence revealed by the econometric analysis. In other words, when presented with evidence which refutes standard theory there is a tendency to re-work standard theory rather than seriously question it. Therefore, the iterative process of confronting theory with evidence and then re-considering theory may not be fully pursued. Moreover, it may be that in social science there is a greater need to re-evaluate theories than in the natural sciences where some of our basic theories appear to be less contentious. Second, the authors would argue that the further shortcomings of econometric practice and perhaps applied work in general is that it tends to regard econometric evidence as decisive rather than simply a particular and specialised form of support.A corollary of this is the failure to consider alternative forms of evidence seriously. These might include survey data, case studies and observation of the behaviour and practice of economic agents. Another problem with econometric practice or, better perhaps, much applied economics (based on regression-type analysis) may be that its relationship with wellarticulated theoretical explanation seems rather poor. The charge that some work in this area simply arbitrarily selects possible ‘explanatory’ variables and

Critical realism, methodology and applied economics 61 then concludes something on the basis of a rather weak statistical evidence may not be without some force. But, this is an entirely different matter to claiming that the sins of econometric practice can ultimately be laid at the door of some underlying commitment to a Humean event – regularity view of causality.The issue of real importance here is the specification of relationship between theory and empirics not whether one can dismiss econometrics on the basis of an ontology to which it need not (and often does not) subscribe. Finally, there are the well-known objections to the use of econometrics, usually ascribed to Keynes, that there is insufficient stability in the economic environment to allow the statistical assumptions on which econometrics is based to be even approximately valid. All of these objections have considerable force, and on occasions, critical realism has seemed to use them. However, they seem not to be the principal basis on which critical realism wishes to disallow econometrics. Critical realism recognises that there is some degree of recognisable structure at the surface level which forms the initial basis for theorising. As noted previously such surface regularities are normally characterised by critical realism as ‘demi-regs’.The principal argument of critical realisms is that econometrics requires the existence of strict event regularity whereas the most we ever observe is a demi-reg. An immediate implication of this view is that critical realism believes that there is a clear demarcation line between approaches which assume and require strict event regularity and those which are based on demi-regs. Much of the critical realist critique centres on this distinction yet no clear basis for defining such a demarcation is offered. This may be highlighted in terms of a series of questions. Is a statistical relationship acceptable, for example, the characterisation of a time series as autoregression (2) but not an econometric model? What is the distinction between summary statistics and other re-arrangements of the data which generate statistics labelled econometrics. In other words, why does critical realism find summary statistics acceptable but say, simple regressions unacceptable? It is worth noting that even summary statistics do involve some theoretical content in as much as the decision to look at particular statistics is informed by theory, as is the way in which different variables may be related. A related question is: how are demi-regs identified in critical realism. An enduring relationship may not be immediately visible until the data has been interrogated (e.g. the Phillips curve). In which case, what degree of manipulation of the data is allowed prior to the identification of a demi-reg? Does critical realism sanction the use of data transformations to discern more clearly any existing pattern? These questions suggest that distinguishing between approaches based on strict regularity and those based on demi-regs may be impossible without a clear demarcation principle which has not been supplied. However, even accepting the existence of such a distinction it is unclear why econometrics requires any more than the demi-regs. Econometricians do not have to assume the universal strict regularities that critical realism claims. Evidence of a relatively enduring tendency may be all that econometricians typically attempt to

62

Bernard Walters and David Young

establish and therefore it is unclear that their underlying view of the world need be radically different from the one that critical realism recommends. It is worth noting here however that it is always possible that some relatively ‘crude’ statistical link might initiate an attempt at serious theoretical explanation which could result in an important (explanatory) development. In other words although we ought not to take such connections too seriously in themselves they are often an important part of the evolution of theoretical ideas and explanation. This is perhaps the case with much medical research, which often initially identifies factors which might be associated with particular medical contributions prior to the development of a plausible explanation of the relationship between the correlated variables. Presumably critical realism would accept this and acknowledge that such correlations represent potentially important observations about the world.The authors’ point, however, is that this kind of empirical observation often involves something more than descriptive statistics and something rather like the kind of simple statistical models/techniques employed in econometrics which critical realism rejects. In addition, as noted above, it is unclear what critical realism’s preferred search procedure for establishing demi-regs is.The authors suggest that any concrete answer to this might come close to the approach used by statisticians and econometricians. Critical realism’s attempt to use methodological distinctions to dismiss econometrics is based on their argument that econometrics is a special case of deductivism.This is the method under attack.This, of course, is contrary to many other criticisms of econometrics which has been concerned with possible problems of induction. However, the key question which must be addressed is how does critical realism suggest that we proceed without deduction? It is worth recalling from the earlier discussion that critical realism recommends retroduction as an alternative.This is a false dichotomy.7 Retroduction is a combination of inductive and deductive reasoning while unobjectionable in itself does not eliminate the need for deductive forms of argument. Indeed, critical realism often engages in such forms of argument and, the authors would suggest, that this is almost inescapable. Presumably critical realism would argue that the issue can be dealt with in terms of closure and that this is a characteristic of orthodox theory, which is rejected by critical realism. As we have noted elsewhere, much more than orthodox theory falls according to this argument, but the key point here is that the inability of critical realism to demonstrate how we are to proceed in the absence of some degree of closure means that this cannot be a convincing way of resolving the issue of deductivism. By contrast we would suggest that a more fruitful way to proceed would be to investigate further the nature of the relationship between alternative theories and a variety of forms of evidence. This suggests that many of the criticisms of econometrics could be more usefully discussed in terms of the interface between theory and evidence. Moreover, critical realism has never actually established that econometrics must necessarily be based on Humean event regularities. As well as its criticisms of applied economics, critical realism also argues for an alternative approach to economic policy. In particular, it is claimed that more ‘radical’ policy choices are

Critical realism, methodology and applied economics 63 made available to us once we embrace a critical realist view of science. For example, Lawson (1997, p. 277) comments ‘Whereas the most that supporters of the traditional, positivistic, constant-conjunction view of science can sustain with any consistency is the goal of control, along with the amelioration of events and states of affairs, the critical realist perspective instead offers the real possibility of human emancipation through structural transformation’. It is argued that this view implies that policy choices are less constrained and could include more radical changes to practices, institutions which are often taken as fixed in mainstream or orthodox theory. An obvious question to pose here however, is why one cannot consider such issues in terms of the scope or range of possibilities that are considered by policy makers? Is it really a matter of methodology? Must an orthodox conception of science rule out the possibility of social transformation while the critical realist perspective encourages such change. Broadly speaking the authors would suggest that there are many factors which determine the range of possible policies considered by policy makers and it is unconvincing to link either the breadth of policy or the selection of a particular policy to methodological distinctions. Rather the main factors influencing policy can largely be assessed at the theoretical level. Here it is clear that economic considerations are often likely to be inextricably linked to political and social issues and it is equally clear that there may be a reluctance to consider radical policy proposals that might be regarded by some as emancipatory. But none of this can be convincingly explained in terms of alternative methodological approaches or different conceptions of science as critical realism wishes to argue. Rather, what is required is a serious attempt to address a full range of plausible policy alternatives, which involves assessing different types of evidence. Some of the latter may usefully include econometrics and other predictions. Notwithstanding earlier remarks regarding conceptual innovations and the nature of human agency, it may be argued that the type of questions which are typically addressed by economists in the policy domain may not be so susceptible to the ‘interpretative critique’ as some of the grander issues considered by the social sciences in general. If this is so then it may be that it is more plausible that prediction has a surer footing and might therefore be expected to have a greater role in policy design and implementation. None of this is to suggest, of course, that it is not sometimes the case that too much emphasis is given to the role of prediction or even that too great a confidence may be placed in its accuracy. It is still the case that economic models which produce predictions may usefully play an important role in shaping policy. One cannot rule this out by appeal to methodological considerations and certainly not one which relies upon an open-system argument. The authors would suggest that the important re-orientation towards policy issues should include an explicit concern for other forms of evidence which necessarily changes the relative weight given to econometric and other predictive models. However, this is in stark contrast to critical realism’s dismissal of the usefulness of prediction in the policy arena.

64

Bernard Walters and David Young

Conclusion Critical realism represents an attempt to avoid the perceived pitfalls of an inflexible positivism and the resigned relativism of some subjectivist and idealist methodological accounts.This worthwhile objective does not in itself justify the adoption of critical realism if it cannot confront the epistemological problems that confront all science but particularly social science. This chapter has argued that critical realism’s ontological boldness has never been supported by any plausible epistemology. Instead, it has attempted to bolster its position by misrepresenting the complexity and the power of alternative approaches. Critical realism forces all discussion onto the ontological plane. But its power as a basis for a prescriptive economic methodology must lie in its ability to provide some epistemologically convincing rules for guiding researchers; this chapter has argued it has failed to provide such rules. In fact, critical realism is extremely weak in the epistemological domain. The rules it offers are either innocuous or insulate explanations from empirical challenges. Moreover, in the absence of a supporting epistemology the confidence of critical realists in the existence of underlying (hidden) powers and mechanisms amounts to an ungrounded assertion that ironically echoes the claims of early positivism about the truth content of observed reality. Indeed, it might be argued that critical realism is rather more positivist about what we cannot observe than many supposed positivists are about what we can. Critical realism’s attempt to demarcate approaches dependent on strict event regularity from those that only depend on demi-regs is unsuccessful. Similarly, the attempt to delineate orthodox from non-mainstream approaches on the basis of closed versus open system distinctions, or the corresponding use of deduction is also unconvincing. The authors conclude that in the applied domain, the critical realist case for eschewing econometrics is misplaced as is the claim that prediction has no useful role to play in economy policy.

Notes 1 This chapter is based on an earlier paper (Walters and Young, 2001) especially section ‘The chief features of critical realism’ and parts of the sections ‘The critical realist view of mainstream economics’ and ‘Epistemological problems with the realist method’. 2 Baert (1996) reminds us that critical realism was originally conceived as descriptive and only later acquired a prescriptive character. 3 Transcendental realists do not claim that theories in this tradition do not have to match up to the empirical evidence. But an aim of the project is to allow more space for the proposed existence of mechanisms and powers that are not directly observable. 4 See Outhwaite (1987) for a discussion of the genesis of the critical realist movement. 5 For example, Bhaskar (1986, pp. 61–62) includes empirical evidence as one of a great many relevant features of an acceptable theory. However, it is given no specific role in determining a theory’s soundness. 6 Runde (1998) notes that these hallmarks of a causal account are not intended to advance a critical realist account per se. However, he indicates that in his view realist

Critical realism, methodology and applied economics 65 accounts would pass the tests posed by his causal explanation while mainstream versions would not. 7 It is perhaps worth noting that critical realism does seem to engage in what are often termed Cartesian dualisms. For example, one either understands the world in terms of strict regularities or not; one is either deductivist or not etc.

References Baert, P. (1996) Realist philosophy of the social sciences and economics: a critique, Cambridge Journal of Economics, 20, 513–522. Bhaskar, R. (1978) A Realist Theory of Science (Brighton, Harvester Press). —— (1979) The Possibility of Naturalism (London,Verso Press). —— (1986) Scientific Realism and Human Emancipation (London,Verso Press). —— (1989) Reclaiming Reality (London,Verso Press). Boylan, T.A and P.O’Gorman (1995) Beyond Rhetoric and Realism: Towards a Reformulation of Economic Methodology (London, Routledge). Friedman, M. (1953) The methodology of positive economics, in: Essays in Positive Economics (Chicago, University of Chicago Press). Keat, R. and J. Urry (1982) Social Theory as Science (London, Routledge and Kegan Paul). Lawson, T. (1994) The nature of Post-Keynesianism and its links to other traditions, Journal of Post Keynesian Economics, 16(4), 503–538. —— (1997) Economics and Reality (London, Routledge). Outhwaite,W. (1987) New Philosophies of Social Science: Realism, Hermeneutics and Critical Theory (London, Macmillan). Runde, J. (1998) Assessing causal economic explanations, Oxford Economic Papers, 50, 151–172. Stigler, G. and G. Becker (1977) De Gustibus non est disputandum, American Economic Review, 67, 76–90. Taylor, C. (1985) Interpretation and the sciences of man, Chapter 1 in Philosophical Papers 2, Philosophy and the Human Sciences (Cambridge, Cambridge University Press). Walters, B and D.Young (2001) Critical realism as a basis for economic methodology: a critique, Review of Political Economy, 13(4), 483–501.

Part II

Critical realism, formalism and empirical work Some conceptual issues

Part IIA

On closure

5

Critical realism and formal modelling Incompatible bedfellows?1 Mark Setterfield

Introduction The role of ‘formal’ (mathematical) modelling in economics is a vexed and vexing issue, on which economists continue to disagree.2 In recent years, this debate has been joined by critical realists. Because of its attention to both ontology and methodology, critical realism provides a basis for reflecting on whether or not formal modelling is an appropriate way of doing economics, given the nature of economists’ object of study.3 Thus far, critical realists have expressed scepticism about the usefulness of formalism in economics. In the first place, formal modelling is closely associated with the practices of mainstream (neoclassical) economics, of which critical realists are dismissive.4 Second, explicit assessment of the use of mathematics in economics by critical realists has yielded negative results. Lawson (1997a) is pessimistic about the likelihood of there being a useful role for formal methods in economics, while Fleetwood (2001) expresses something rather more like disdain for mathematical techniques. The point of this chapter is to rejoin the formalism controversy by further investigating the extent to which formal modelling is, or can be made, consistent with the ontology and methodology of critical realism. It is argued that the notion of formal modelling can (and should) be separated out from the practices of mainstream economists. This makes it possible to develop a nonmainstream approach to formal modelling that yields tractable formal models of economic processes and, by virtue of its rejection of mainstream ontology and methodology and its embrace of an altogether different understanding of what society is and how social science should be done, renders formal modelling broadly congruent with critical realism. The remainder of the chapter is organised as follows.The section ‘On critical realism and “formal” modelling’ deals with definitions, briefly discussing what critical realism and ‘formal modelling’ are. This results in the identification of an open-systems, ceteris-paribus approach to formal model building, the characteristics of which are considered in greater depth in the section on open systems and ceteris paribus methodology.The section on critical realism and the OSCP approach discusses complementarities between critical realism and

72

Mark Setterfield

the OSCP approach to formal modelling, and finally, the last section offers some conclusions.

On critical realism and ‘formal’ modelling What is critical realism? The core of critical realism comprises both a distinct social ontology and a method for doing social science that is compatible with this ontology (see, e.g. Lawson, 1994, pp. 512–523). Concern with ontological issues arises inevitably from the fact that critical realism is a form of philosophical realism, that is, it asserts the existence of things that are (or can/should be) the object of knowledge independent of what we know about them. According to critical realism, there are three levels of reality: the empirical (corresponding to what is experienced); the actual (corresponding to the events and happenings that are the objects of experience); and the ‘deep’.The latter consists of the structures, mechanisms, powers and tendencies that generate actual events. Critical realism asserts that the object of science is to uncover these structures, mechanisms, etc. This provides an obvious contrast with empirical realism (the method attributable to mainstream economics), according to which the object of science is to uncover event regularities that codify our experience of phenomena in the form ‘whenever x then y’.5 Recovering the deep is a non-trivial activity because the three levels of reality described above may be out of phase, that is, it is not always the case that innate mechanisms, powers etc. will manifest themselves in actual events which will, in turn, be experienced. This is because the ‘deep’ is understood to constitute an open rather than a closed system. In an open system, effects may not always result from the same causes, that is, there is a lack of extrinsic closure, so that factors external to the system may influence its outcomes in a non-constant and systematic fashion. Moreover, open systems can lack intrinsic closure, which means that causes may not always have the same effects.6 One result of all this is that event regularities will not be the norm – so that the search for event regularities (the object of science according to empirical realists) is not, in general, a valid scientific pursuit. The claim that the object of science is to reveal the ‘deep’ leads critical realists to advocate a particular method for doing social science, called abduction or retroduction. The purpose of retroduction is to proceed, by means of metaphor, analogy, etc., from the observation of some phenomenon of interest to a theory of the structure, mechanism, power, etc. that is responsible for this phenomenon.This method can be illustrated by contrasting it with induction and deduction, as follows: If deduction is, for example, to move from the general claim that all grass is green to the particular inference that the next lawn to be seen will be green, and induction is to move from the particular observations on green

Critical realism and formal modelling 73 grass to the general claim that all grass is green, then retroduction or abduction is to move from the particular observations on green grass to a theory of a mechanism intrinsic to grass that disposes it to being green. (Lawson, 1994, p. 515, emphasis in original) In the social realm, critical realism begins by postulating that the ‘deep’ consists, in part, of intentional human agency (Lawson et al., 1996, p. 147). Specifically, human agents are accredited with the capacity for effective choice, meaning that in any given set of circumstances, they could have acted differently.7 From this, we can retroduce two things: First, social systems (as characterised by the transformational model of social activity (TMSA) that will be described next) are open.8 Second, the ‘deep’ must also (i.e. in addition to intentional human agency) comprise causally efficacious ‘social’ structures that provide the ‘tools’ for intentional action, because to assert otherwise would be to suggest that action can occur in, as it were, a vacuum. Instead, social structure is used (sometimes unconsciously) as a form of ‘capital’ which, when combined with the ‘labour’ of human endeavour, produces action – and this ‘output’ is impossible to produce without both of the ‘inputs’ described above. An obvious example is that acts of speech require a language. The fact that social structure assists human action does not imply that it completely determines action, however. In fact, the social ontology articulated above is not consistent with the reduction of action to structure, because it posits that human agency involves effective choice, and because social structures are, by definition, structures that have been brought about by, and would cease to exist in the absence of, human action. At the same time, however, social structure cannot be reduced to human agency, since human action presupposes social structure.What ultimately results, then – and this is the essence of critical realist social ontology – is a TMSA, according to which structure and agency are interdependent (Lawson et al., 1996, p. 147). Each presupposes the other, ergo, neither can be reduced to the other. Instead, at any given point in time, previously existing social structure (taken momentarily as given) will influence (but not determine) action. This action will, in turn, either reproduce or transform social structure, and so on (Lawson, 1994, p. 520). What is ‘formal’ modelling? For the purposes of this chapter, ‘formal’ modelling is taken to mean mathematical modelling. Indeed, this seems to be the dominant conception of formalism in economics. Hence Chick (1998, p. 1859) suggests that in economics, ‘ … the request “can you state that formally?” … usually means “can you encapsulate that idea in an equation or system of equations?” ’. Defining formal modelling as mathematical modelling merely involves the adoption of conventional usage, then.9 Accepting the definition of formal modelling above, and following Chick (1998, p. 1861), it is possible to swiftly and succinctly describe how formal

74

Mark Setterfield

modelling is done: beginning with certain initial premises (assumptions and/or axioms), ‘ … transformations are then made following agreed rules, and the conclusions [or what Chick describes as ‘demonstrable truths’] follow as long as the rules have been obeyed’. But within this generic method, and at a rather more concrete level of actual practice, it is possible to identify more than one type of formal modelling procedure. One such procedure might be labelled the axiomatic approach to formal modelling. This begins with a ‘hard core’ of assumptions (or axioms) that are not empirically grounded, although they may be referred to as ‘self evident’.10 These axioms are supplemented with auxiliary assumptions that, once again, are not empirically grounded. Axioms and assumptions are then corralled within a model that is purposely designed by the model builder to yield closed form solutions, of either a determinate or – in the case of stochastic models – mathematical mean value type. Formal models of this genus are thus the product of a series of ‘convenient fictions’ embodied in their axioms and assumptions.11 Their closure assumptions render these models consistent with the empirical realist project, in the sense that what they do is generate event regularities of the type ‘whenever x then y’. Given the underlying empirical realist method of the axiomatic approach to formal modelling, coupled with critical realists’ rejection of empirical realism, one can only conclude that critical realism and the axiomatic approach to formalism are incompatible. Were the axiomatic approach the only approach to formal model building, one might then conclude that critical realism and formal modelling are incompatible. Indeed, this seems to be the line of reasoning that advises the conclusions reached by Fleetwood (2001).12 However, a second approach to formal model building – an OSCP – can be identified, which precludes this latter conclusion. In this OSCP approach to formal modelling, assumptions and structural relations are based on a mixture of stylised facts, a process of ‘locking up without ignoring’ some features of reality in order to focus attention on the interaction of others, and, in particular, a sense of what must be true about the economy given a set of prior metaphysical beliefs about the nature of social reality. This latter feature makes social ontology the starting point for the OSCP approach to formal modelling. Models built in accordance with this approach may be open, and/or exemplify conditional closure.13 Conditional closures are brought about by the process of ‘locking up without ignoring’, that is, describing as constant variables and/or structural relations that are known to be capable of change over time, but explicitly acknowledging the artificiality of the resulting closure.14 A special subset of these closures are derived from the observation of relatively enduring institutions (defined so as to include norms, conventions, rules, etc.). Note that this reasserts the importance of stylised facts in the OSCP approach, and also presupposes a theory of institutions as enduring but ultimately transmutable social structures. The emphasis of formal models based on the OSCP approach is on the description of social processes rather than the prediction of specific outcomes. Indeed, models may be open and therefore not yield closed-form solutions

Critical realism and formal modelling 75 from which point predictions can be derived. Nevertheless, OSCP models may still provide qualitative indications of the likely behaviour of a system, based on the interaction of its structural relations (e.g. whether it is likely to display growth or cyclical behaviour). At this point, two pressing questions confront us. First, can the OSCP approach yield useful formal models? And second, can this approach yield formal models consistent with the critical-realist conception of social reality and how social science should be done? The purpose of the following sections is to answer these questions in the affirmative.

Open-systems, ceteris-paribus methodology and formal modelling: a closer investigation As its name suggests, the OSCP approach to formal modelling comprises two key components: open systems modelling, and ceteris paribus methodology. It will quickly become clear that as modelling strategies, both of these components are advised by philosophical realism – the notion that there are objects of inquiry that, at any given instant, exist independently of our thoughts about them, and that the purpose of science is to describe the characteristics of these objects. Open systems modelling Economic models are typically closed systems that contain all of the information necessary to fully determine system outcomes.These closed-form solutions are either determinate or (in the case of stochastic models) take the form of mathematical mean values. In an open system, however, outcomes cannot be reduced to foreclosed explanation in terms of the more ‘primitive’ factors that are purported (within the confines of the system) to bring them about. Outcomes cannot be completely determined by the information contained within an open system, but remain, instead, relatively autonomous from the ‘primitives’ that characterise the system. Open systems, then, lack foreclosure. Specifically, they lack either intrinsic or extrinsic closure, or both.15 Open systems modelling, meanwhile, insists on the relevance of openness to social systems because of the existence of effective choice (the notion that agents could always have acted differently).This makes it necessary to take account of openness in models of social processes. On this basis, we may define open systems modelling as a modelling methodology that involves incorporating systemic openness into structural models of social systems, by denying either intrinsic and/or extrinsic closure at some critical juncture(s). Ceteris paribus methodology Following Vickers (1994, p. 136), three essential and interrelated components of ceteris paribus methodology can be identified. First, it involves ‘taking time

76

Mark Setterfield

seriously’. In other words, the economy is conceived as a path dependent, historical process, rather than as an immutable, mechanical entity. Second, ceteris paribus methodology is based on the notion of sequential causality (according to which prior causes have subsequent effects), rather than simultaneous causality (according to which ‘everything affects everything else’ at a single point in time).This demands that we build causal-recursive models of economic processes, not static, simultaneous equation models. Finally, models are explicitly recognised as partial representations of economic processes in the ceteris paribus methodology. Note, however, that this does not validate the practice of splitting off, say, individual markets from the economy as a whole and then analysing them in a static, foreclosed fashion. Rather, the term ceteris paribus is given an explicitly inter-temporal or dynamic meaning. It is things that are capable of change over time that constitute the ‘other things’ that are taken as ‘being equal’ (i.e. remaining constant) in ceteris paribus methodology. Ultimately, ceteris paribus methodology involves the introduction of ‘synthetic’ or ‘conditional’ closures into otherwise open, historical systems in order to facilitate modelling (formal or otherwise): Some part of a system’s dynamics is overlooked [or, more accurately, ‘locked up without being ignored’ (Kregel, 1976)], creating constancy where there is, in fact, the propensity for change, in order to facilitate the practical analysis of other parts in a conditionally closed system. The conditionality, of course, stems from the proviso that ‘other things’ (the artificial constants) must, indeed,‘remain equal’ if the analysis is to be descriptively useful. (Setterfield, 1997, p. 70) These conditional closures can be introduced into economic models in two ways.The first process can be termed ‘pure’ locking up without ignoring.This occurs when a variable or structural relation is treated as unchanging for analytical (non-empirical) reasons. For example, it might be regarded as inessential to the analysis at hand, and the analyst is simply trying to focus on those aspects of the system’s dynamics that are essential.16 Alternatively, a valuable pedagogical purpose may be served by locking up without ignoring even an essential feature of a system’s dynamics.A good example of this is Keynes’s (1936) treatment of the state of long-run expectations as given in order to demonstrate the principle of effective demand.17 The second process for introducing conditional closures is ‘empirically grounded’ locking up without ignoring. This occurs when actually existing institutions are built into economic models. In this case, conditional closure rests on the specific condition that real structures (institutions) are purposely reproduced over time by human agents, that is, that these institutions are relatively enduring.18 Note also that conditional closures of this sort can be sought out through a process which begins with the observation of ‘stylised facts’ about conventional behaviour – thus linking the ceteris paribus methodology to

Critical realism and formal modelling 77 Kaldor’s methodological concern with stylised facts (see, e.g. Kaldor, 1985). Finally, note that empirically grounded conditional closures can claim to meet Mäki’s (1994) ‘aboutness’ criteria. There exists a degree of realisticness in empirically grounded locking up without ignoring that will not (typically) be observed in pure locking up without ignoring. It is important to stop at this point to reflect on why the processes described above are not equivalent to idly postulating that exogenous shocks may produce parametric change in models generated by the axiomatic approach to formal modelling (which would seem to render these models transmutable in the flux of historical time). In the first place, the convenient fictions (including the apparently absolute rather than explicitly conditional closure assumptions) associated with axiomatic formal models do not meet Mäki’s ‘aboutness’ criteria, as does a subset of the conditional closures described above. Second, and most importantly, the OSCP approach to formal modelling requires that an explicit meta-narrative accompany closed models – a meta-narrative that is designed to explain and justify assumed closures, given that the latter are presumed not to be ubiquitous in the social realm. This meta-narrative is, of course, absent from axiomatic formal models, which begin with the assumption that closure is ubiquitous and contemplate openness (if at all) only on a post hoc, revisionist basis (via appeal to parametric ‘shocks’, as described before). There is, then, an observable difference between the axiomatic and OSCP approaches to formal modelling. The language of conditional closure is not simply a cloak that can be draped over what are otherwise general, once-andfor-all ‘laws’ exhibiting absolute closure, in an effort to render the latter compatible with the OSCP approach to formal modelling (see also Chick and Dow, 2001, p. 719). Indeed, it is difficult to understand why axiomatic model builders would want to argue that their ‘laws’ should not be taken ‘too seriously’ as literally immutable statements of ‘whenever x then y’ event regularities. In the first place, it would be impossible to take any such argument itself ‘too seriously’, because it is so obviously at variance with the observable thrust of axiomatic modelling activity throughout the history of economic thought (see Mirowski, 1989). Second, the argument would create a crisis of meaning. If we are not supposed to take the apparently immutable laws of axiomatic formal models ‘too seriously’, one can only wonder what else that is concealed from the field of view we ought to be ‘taking with a pinch of salt’? What, in fact, is any of the explicit content of an axiomatic formal model actually supposed to mean? This crisis of meaning demonstrates that post hoc dalliance with the notion that social systems are open and transmutable is no substitute for explicit recognition and account of this, as is to be found in the OSCP approach to formal modelling.To repeat, there is no equivalence between the OSCP approach and a more ‘relaxed’ or ‘casual’ interpretation of the axiomatic approach to formal modelling: the OSCP approach is categorically not what most formal model builders are ‘really doing in any case’. In sum, ceteris paribus methodology suggests that whilst there are no absolute or ‘general’ theories of economic processes that hold true across all time and

78

Mark Setterfield

space, it is possible to construct partial, conditional theories. Also, since what is modelled is purported to reflect actually existing economic processes, and to the extent that the conditional closures on which models rest are empirically grounded, the ceteris paribus methodology embodies the view that economic theory should be realist and descriptive in its orientation, rather than fictitious and instrumentalist. An example In order to better illustrate the OSCP approach to formal modelling, it is instructive to examine a concrete example. Consider, then, the following model of effective demand, reproduced from Setterfield (1999):19 Z(Nt )Zt=Yte D(Nt ) Dt  Ct  it 01 Ct  Zt, It  I(rt, e(t )), Ir  0, Ie.e  0 01 it  .It, rt  (1  )t –– t =  ∀ t 1 Yte = Dt1

(1) (2) (3) (4) (5) (6) (7) (8)

where Z is aggregate supply, Y e denotes firms’ short-run expectations of nominal income, D is aggregate demand, C is consumption expenditure, i is effective investment spending, I is notional or planned investment spending, r is the commercial rate of interest,  denotes firms’ animal spirits,  is the central bank discount rate and t subscripts denote time periods. The model postulates that the levels of nominal income and employment are determined by the interaction of aggregate demand and supply, where the latter comprises consumption spending out of current expected income,20 and investment spending that is financed by bank credit, the price of which is determined by a mark-up over the rate at which the central bank is willing to lend reserves.21 Equations (1)–(8) can be written in reduced form as: Dt  t D0  

t



i1

ti

· I( r, e( i))

(9)

where r  ( 1  ) results from combination of equations (6) and (7). The model is completed (but not necessarily closed, as we shall see) by the introduction of a final equation describing the evolution of animal spirits. Two equations can be considered for this purpose: t 

(10)

Critical realism and formal modelling 79 or



t  1  t if Dt  1  Zt  1  c if |Dt  1  Zt  1|  c t  ft(Dt  1, Zt  1)  t  1 t  1  t if Dt  1  Zt  1  c

(11)

where t  ( t,2t),t  0 ∀t and c is a constant. The dynamics of the model that result from the combination of (9) and (10) or (9) and (11) have been explored by Setterfield (1999) (see also Setterfield (2000) for extensions).The purpose here, however, is not to repeat this analysis, but rather to show that the formal model of effective demand introduced above has been constructed on the basis of the OSCP approach to formal modelling. It is to this task that we now turn. First, elements of open systems modelling can be found in the animal spirits reaction function in (11).The evolution of animal spirits in (11) depends on draws from a distribution with time-variant moments.The evolution of these moments, however, is not described within the confines of the model. Indeed, such description is ruled out by hypothesis.This is because  and 2 are taken to represent relatively enduring conventions that can be changed in novel ways by agents who exercise effective choice (Setterfield, 2000).The result is that in equation (11), the same cause can have different effects: it is impossible to relate a value of D  Z of any given size to an expected value of  of some particular size, since the mean value of  is non-constant.This implies that the function lacks intrinsic closure. The reaction function in (11) therefore explicitly incorporates the assumption that the economy is an open system into a formal model of an economic process (the determination of income and employment). Elsewhere, in equations (1)–(11), ceteris paribus methodology is in abundant evidence. Examples of ‘pure’ locking up without ignoring include the consumption function (equation (3)), and the variant of the animal spirits reaction function that holds animal spirits constant over time. The propensity to consume in (3) is properly thought of as a conventional value amenable to change in light of variations in the state of consumer’s long-run expectations with respect to future income streams (see Setterfield, 2000). In this model, however, it is held constant – locked up without being ignored, thus creating intrinsic closure in the consumption function – because the focus of the analysis is on investment and financial sector behaviour. Meanwhile, the variant of the model that locks up without ignoring animal spirits by treating them as constant serves a useful pedagogical purpose. Specifically, it permits demonstration of a point of effective demand within the model.This is useful not only because it is familiar, but also because the standard notion of adjustment towards this equilibrium serves as a useful metaphor when trying to explain the behaviour of the variant of the model that involves equation (11). The model also contains examples of conditional closure based on the stylised facts of relatively enduring conventions – for example, the description of price setting in the commercial banking sector according to a mark-up

80 Mark Setterfield pricing rule. Admittedly, this particular empirical grounding is rather shallow, since it is not based on an extensive investigation of the precise form taken by the mark-up rule in this sector of the economy. Nevertheless, the intrinsic closure introduced here is not entirely lacking in realisticness, given what is generally understood to be true about firm’s pricing practices (see e.g. Lee, 1998; Downward, 1999).The assumption of a constant (or at least relatively enduring) mark-up is not simply a convenient fiction, but rather has some basis in reality.22 It is important to acknowledge that the model is not devoid of ‘convenient fictions’. One obvious example is its description of firms’ short-run expectations. This is simply an invention designed to satisfy the criterion of model simplicity (a spurious criterion from a realist perspective).The reason for treating short-run expectations and long-run expectations differently is based on a perceived feature of social reality – namely, the calendar time horizon necessary for systemic openness to create the sort of radical uncertainty that results in there being an irreducibly arational, psychological component to decision making (on which, see Keynes (1936) and Gerrard (1995)). But otherwise, the precise form of equation (8) and its intrinsic closure are convenient fictions.23

Critical realism and the open-systems, ceteris-paribus approach to formal modelling How, if at all, does any of the above resonate with critical realism? The point of this section is to relate the formal model developed above and the OSCP methodology on which it is based to the ontological and methodological premises of critical realism. Retroduction and deduction First, note that the model as a whole is non-instrumental. It is not designed as a ‘black box’ for predicting system outcomes. Rather, emphasis is on describing the structural features of an economy which, through their interaction, generate certain types of macroeconomic outcomes (specifically, levels of nominal income and employment). The primary purpose of the model, then, is to hypothesise the causally efficacious structures that characterise a capitalist economy, not simply to create a device designed to ‘predict’ some surface phenomenon through a closed form solution. In other words, the model is an attempt to retroduce the sort of actually existing social structures, tendencies, mechanisms etc. that describe the macroeconomics of an open social system.24 Critical realism does not stipulate precisely how retroduction should be done, but frequent reference is made to the use of analogy and metaphor in this process (e.g. Lawson, 1994, p. 515; 1997c, p. 212). This is how the formal model in the section on open systems – ceteris paribus methodology proceeds – the metaphors/analogies used are simply mathematical in nature.25 Hence the animal spirits reaction function in (11), for example, uses mathematical analogies/metaphors to describe real powers/tendencies – specifically, the

Critical realism and formal modelling 81 causal influence of past economic outcomes on the state of long-run expectations in an environment of uncertainty, the intrinsic openness of this causal relation in an environment of effective choice, and the emergence of social structures (in this case, conventional responses to the disappointment of short-run expectations) that are both a product of and an influence on agents’ decision making.26 Furthermore, critical realism does not deny a role for deduction in theorising. Hence ‘if we do have knowledge of what a thing is, we can deduce its causal powers’ (Lawson, 1989, p. 62). In other words, once we know the nature of something, it is acceptable from a critical realist perspective to work out – deduce – what it does.This is what the formal model above is designed to do. It first attempts to describe the nature of a capitalist economy (based on the process of retroduction as described above), and then deduce its outcomes. As regards the latter, it can be shown that the model suggests that there are likely to be cumulative expansions and contractions in nominal income and employment (Setterfield, 1999), whilst extensions of the model reveal the possibility of cycles in nominal income and employment that are both aperiodic and of no fixed amplitude (Setterfield, 2000). At no point is it claimed that the value of nominal income, the turning points of the business cycle, or even the amplitude and period of the cycle can be predicted. But it is possible to deduce the outcomes described above from the properties of the economy that the formal model retroduces.27 Open systems modelling and abstraction Second, the method of open systems modelling described earlier – which advises the animal spirits reaction function in equation (11) – is clearly in keeping with two of the central claims of critical realism: that social systems are open; and that models, in the process of abstracting from the concrete, should focus on and hence embody the real (rather than invented fictions). In other words, the formal model above explicitly subscribes to the view that social systems are open and that the analyst can and should take account of this when abstracting for the purposes of model building, rather than seeking to elude openness by deploying convenient fictions in the service of ‘model tractability’ (or some other similarly spurious, non-realist objective). Note that the concern for systemic openness, as introduced into the formal model above, is also consistent with critical realism’s emphasis on abstracting the essential rather than the most general. Hence the focus on expectations and monetary relations in the determination of aggregate activity, and the choice to deny intrinsic closure at one of these junctures rather than elsewhere, is motivated by a conception of what is truly essential to the explanation of variations in aggregate output and employment – not a sense of what is most generally true about economic behaviour.28 Ultimately, then, the parts of the model that are advised by the open systems modelling procedure are consistent with critical realism’s ‘golden rules’ for abstracting (Lawson, 1989, pp. 68–73).

82

Mark Setterfield

Ceteris paribus methodology, abstraction and conditional closure Third, each facet of the ceteris paribus methodology described earlier, and shown to have been extensively used in the construction of the formal model is in keeping with the strictures of critical realism.To begin with, it is important to recall exactly what the term ceteris paribus means here, because critical realists frequently expresses ambivalence about use of this term (see, e.g. Lawson, 1997c). Hence ceteris paribus claims are understood by critical realists to represent counterfactual statements (designed to demonstrate what would have been true if things had been different), whereas critical realism demands transfactual statements (designed to reveal what is always true, even if it is not evident in the realms of the actual and/or empirical).This ambivalence on the part of critical realists is understandable, because ceteris paribus claims are frequently found in theories that adhere to empirical realist methodology. In these theories, the term ceteris paribus is invoked to show what would have happened (in the actual and empirical realms) had other forces not intervened – thus producing counterfactual event regularities. But the term ceteris paribus can be uncoupled from empirical realist methodology.As interpreted earlier, ceteris paribus methodology involves ‘locking up without ignoring’ some aspects of a system’s dynamics in order to focus attention on others. This method is not designed to produce counterfactual event regularities – indeed, as discussed earlier, the model in the section on open systems–ceteris paribus methodology does not yield event regularities at all.29 Instead, the purpose is to help the scientist recover transfactual tendencies, mechanisms, etc. that are believed to be always operative, but whose effects need not be manifest (either factually or counterfactually) in the realms of the actual and empirical. In sum, ceteris paribus methodology as it was described earlier is centrally concerned with issues of ontology and proper abstraction, and should not be mistaken for the empirical realist practice of producing counterfactual event regularities. Having thus sought to avoid any terminological confusion, we can now begin our assessment proper by recalling that two essential components of ceteris paribus methodology as described earlier are its conception of the economy as a historical process, and its concomitant emphasis on sequential causality.This is in keeping with the critical realist TMSA, which is a causal-recursive model of social activity (Baert, 1996, p. 521), and which embraces the idea that earlier states of the world influence later ones (path dependence). Meanwhile, the third component of ceteris paribus methodology – its emphasis on partial theorising based on ‘locking up without ignoring’ – resonates with the critical realist conception of proper abstraction. For example, the idea of beginning with the concrete and then ‘locking up without ignoring’ some aspects of this reality in order to focus on others involves an emphasis on the real and (at least implicitly) the essential in the process of abstraction. As a method of abstraction, then, ceteris paribus methodology is again consistent with the ‘golden rules’

Critical realism and formal modelling 83 of abstraction in critical realism. Moreover, the idea of utilizing ‘empirically grounded’ conditional closures in formal models which, as noted earlier, embody a degree of realisticness or ‘aboutness’ (Mäki, 1994) that is not (likely) present when ‘pure’ locking up without ignoring occurs, only serves to strengthen the ties of this method to critical realists’ concern for focusing on the real rather than the fictitious when abstracting. In fact, the process of ‘locking up without ignoring’ seems to fit almost exactly critical realist descriptions of what proper abstraction involves. Hence in critical realism, ‘abstraction has a straightforward interpretation that is quite consistent with its traditional one of looking at something in a one-sided manner – of focusing on certain aspects of something to the neglect of others’ (Lawson, 1989, p. 68). Stated succinctly, ‘abstraction is to momentarily neglect features of reality’ (Downward and Mearman, 1999, p. 6). Either of these statements could be used, without substantive modification, to describe the procedure of locking up without ignoring. A final point remains to be made about ceteris paribus methodology, related to the claim (to which reference has been made several times above) that formal models can embody conditional closures based on relatively enduring institutions, which display a degree of realisticness (Mäkian ‘aboutness’) that is not evident in all abstractions based on the process of ‘locking up without ignoring’. The point to be made is that this claim should not surprise critical realists, since implicit in critical realism is a theory of how relatively enduring institutions and conditional closures come about. Specifically, anxiety in the face of systemic openness is seen to result in the deliberate creation of synthetic closures by individuals in the form of rules, norms, conventions, etc. (Lawson, 1997c).These institutions must be relatively enduring if they are to serve their intended purpose of creating quiescence, tranquility, and routineness where there is, by necessity, none. Moreover, critical realist claims about the existence of demi-regs and Kaldorian stylised facts as pointers to useful scientific enquiry (see, e.g. Lawson, 1989, p. 65–7; 1997c, pp. 207–8) are based on the notion that there are relatively enduring structures capable of generating these quasi-event regularities in the actual and experiential realms. Incorporating these conditional closures into models of other processes (as sub-systems of larger systems designed to explain more macroscopic phenomena) would appear to be the logical next step for theorists informed by critical realism. Indeed, this process points towards a possible reconstruction of the ‘microfoundations’ project along realist lines (although this would obviously involve a fundamentally different perception of the nature and purpose of microfoundations). The role of ‘convenient fictions’ Finally, even the (admitted) ‘convenient fictions’ in the formal model developed earlier might be excused by critical realists on the grounds of Mäki’s (1994) distinction between core and peripheral assumptions. Core assumptions

84

Mark Setterfield

concern what is perceived as being central or essential to a process; peripheral assumptions simply neutralise inessential elements. Hence: … in an open-system realist perspective, core assumptions should be realistic, in the sense of plausibly attempting to fulfil the criterion of ‘aboutness’ as identified by Mäki. In other words, core assumptions should attempt to represent open-system properties of an economic activity. Peripheral assumptions then merely delimit the partiality of a particular explanation. (Downward and Mearman, 1999, p. 11, emphasis in original) The construction of the formal model coincides with this guideline, as is evidenced by the description at the beginning of this section of how the model deploys the open systems modelling technique associated with the OSCP procedure.

Conclusion The purpose of this chapter is to encourage the view that ‘formal’ (mathematical) modelling is essentially a tool or language that can be used in different ways, so that there is more than one conceivable formal modelling procedure. From a realist (including critical realist) perspective, the relevant question is whether or not formal modelling is well suited to the nature of the object of study (in this case, the economy). It has been argued that the widely used axiomatic approach to formal modeling is not well suited to the study of social processes. However, it has also been shown that formal models can be constructed using an open-systems, ceteris-paribus (OSCP) approach to formal modelling, and that models so constructed adhere to prominent features of the ontology and methodology espoused by critical realists. It follows that there are ways of building formal models that share an affinity with the ‘world view’ of critical realism – that is, its sense of both what the (social) world is like and how (social) science should be done. Ultimately, then, it would seem that critical realism and formal modelling are not – or, at least, need not be – incompatible bedfellows.

Notes 1 An earlier version of this paper was presented at the Workshop on Realism and Economics at the University of Cambridge in May 1999. I would like to acknowledge Harcourt (1992) as the inspiration for the title of the paper. 2 See, e.g. Harcourt (1995) and the symposium on formalism in economics in Economic Journal, 108, pp. 1826–69 (1998) for a selection of views, past and present. 3 As noted by Lawson (1997a, p. 4), this ‘matching’ criterion is absent from common defences of formalism based on the alleged clarity, rigour and/or aesthetic appeal of mathematical arguments. 4 See, for example, Lawson (2000, p. 6) for an example of this association between mainstream methods and formalism. 5 This practice is frequently referred to as ‘deductivism’ (see, e.g. Pratten, 1996; Lawson, 1997c, p. 88).

Critical realism and formal modelling 85 6 See Lawson (1995, pp. 267–8), Downward and Mearman (1999) and Chapters 3, 6 and 7 of this book for further discussion of intrinsic and extrinsic closure and the meaning of openness. 7 Effective choice might be described as a key metaphysical belief of critical realists concerned with the social realm, although taking effective choice as a starting point is sometimes justified by virtue of the widespread emphasis on choice throughout the social sciences (see, e.g. Lawson et al., 1996, p. 147). 8 Referring back to the earlier discussion of openness, effective choice means that social systems lack intrinsic closure: when cause–effect relationships are mediated by effective choice, causes need not always have the same effects. 9 In so doing, it does, of course, help to reproduce this convention – and it must be admitted that this may not be a good thing, since it is not obvious that all mathematics is formal or that all formal arguments are mathematical (Chick, 1998; Weintraub, 1998). 10 As noted by Lawson (1989, p. 74), this is sometimes because the axioms are so general as to be virtually devoid of content (and hence uncontroversial) – for example, the claim that consumers have preferences. In this case, it is what is described below as the auxiliary assumptions of a theory (e.g. the functional form used to describe preferences) that advise its substantive content. 11 In other words, no claim is made that these axioms and assumptions mirror things that are thought to be real – except in the case of so-called ‘self-evident’ axioms which, as has been suggested above, contribute little or nothing to the substantive content of a model. 12 The notion that the axiomatic approach is the only approach to formal modelling appears also to advise the premises of Chick and Dow (2001, pp. 706, 708), though ultimately their concept of provisional closure is akin to that of conditional closure in the open systems–ceteris paribus approach that is articulated below as an alternative species of formalism. 13 A structural model may be both open and display conditional closure if parts of the model are closed, whilst other parts – and hence the model as a whole – remain open. 14 As will be made clear below, this is not the same thing as a post hoc claim that the data on which an axiomatic formal model is based are susceptible to revision by exogenous shocks. 15 The reader is referred back to the section ‘What is critical realism’ for further discussion of these sources of closure. 16 On the propriety of allowing for less realisticness in assumptions judged to be further from the essential core of a process, see Mäki (1994). 17 There may be other justifications for ‘pure’ locking up without ignoring – the list of examples above is not intended to be exhaustive. 18 Conditional closures of this sort are central to the ‘conditional stability’ of capitalist economies as discussed by Crotty (1994). 19 What follows is merely one example of a formal model that is consistent with the OSCP approach to formal modelling. It can in no way claim to represent – and in no way does it claim to represent – the model of the aggregate economy that is consistent with this modelling methodology. 20 Note that substituting (1) and (8) into (3) and assuming that in any period, realized nominal income is determined on the demand side, is equivalent to the claim that current consumption is financed by the previous period’s realized income. 21 Note that only a fraction, , of firms’ planned investment expenditures are financed by banks and hence rendered effective investment expenditures. This is because banks are not obliged to lend to all-comers at the prevailing interest rate, but only to those deemed credit worthy.

86

Mark Setterfield

22 See Setterfield and Cornwall (2002) for an example of a formal model that could claim to have a firmer grounding in the use of conditional closures based on the stylised facts of relatively enduring conventions. 23 Ultimately, this may be defensible, however, even from a realist perspective. 24 An interesting question that is not pursued here is: what is the critical act of retroduction in the formal model above? Is it retroduction of the (plausible) structure of the economy based on a perceived sense of how the economy works (in particular, that economic outcomes are the product of effective choice by human agents)? Or is it retroduction of the (plausible) structure of Keynes’s thought (specifically, his notion of a shifting equilibrium model of effective demand)? Note that both interpretations may be valid simultaneously. Hence if we take Keynes’s shifting equilibrium system as a retroduced structure which began from postulates of effective choice and systemic openness, then the set of models pursuing the second objective described above (retroduction of Keynes’s thought) is a subset of the set of models pursuing the first objective (retroduction of the structure of the economy). 25 The notion that mathematics is a metaphorical form of reasoning is not new. It is found in the work of both apologists for neoclassical economics, such as McCloskey (1985), and their opponents, such as Mirowski (1988, pp. 124–30; 1989), who traces the idea back through Jevons to the work of Poincaré. According to Mirowski (1988, pp. 128–9), ‘mathematics is the method par excellence for the transfer of metaphor … and … metaphor is an indispensable tool of human reasoning’. He provides examples both of how metaphors in mathematics can be generative (that is, can create new insights – a key characteristic of successful metaphors (Lewis, 1999)), and how mathematical metaphors can be costly, by introducing misleading or irrelevant information into an inquiry (Mirowski, 1988, pp. 127, 133–5). Ultimately, these generative and costly properties of metaphors are two sides of the same coin which, borrowing language from Chick and Dow (2001, p. 705), might be referred to as the non-neutrality of metaphor (although Chick and Dow are inclined to think of the non-neutrality of mathematical metaphor as only being costly). This acknowledgment of their non-neutrality leads us to the polar opposite of McCloskey’s (1985) position, that metaphors are essentially inconsequential (see also Mirowski, 1988, p. 139). It is worth stating explicitly, then (if this point is not already clear), that it should in no way be inferred from what has been suggested above that mathematics is ‘just’ a metaphor, as if the latter were a thing of no consequence for what follows from it. 26 See also Setterfield (2000) for discussion of the ontological grounding of the model in the section on open systems. 27 The position here, then, is that it is incorrect to claim that ‘if constant conjunctions of events are not implicitly built into toy (formal) models as an a priori premise, then nothing can be deduced from the antecedents’ (Fleetwood, 2001, p. 207, emphasis in original). 28 In other words, no claim is made that expectations and monetary relations are necessarily important to other aspects of economic behaviour, such as the demand for haircuts or an individual’s choice of career paths. It is for the researcher investigating these phenomena to decide what is and is not truly essential to their determination. 29 At least it yields no event regularities relating to the behaviour of aggregate nominal income, which is the variable that it is designed to study.

References Baert, P. (1996) ‘Realist philosophy of the social sciences and economics: a critique’, Cambridge Journal of Economics, 20, 513–22.

Critical realism and formal modelling 87 Chick,V. (1998) ‘On knowing one’s place: the role of formalism in economics’, Economic Journal, 108, 1859–69. —— and S.C. Dow (2001) ‘Formalism, logic and reality: a Keynesian analysis’, Cambridge Journal of Economics, 25, 705–21. Crotty, J. (1994) ‘Are Keynesian uncertainty and macrotheory compatible? Conventional decision making, institutional structures and conditional stability in Keynesian macromodels’, in G. Dymski and R. Pollin (eds) New Perspectives in Monetary Macroeconomics, Ann Arbor, University of Michigan Press. Downward, P. (1999) Pricing Theory in Post Keynesian Economics: A Realist Approach, Cheltenham, Edward Elgar. —— and A. Mearman (1999) ‘Realism and econometrics: alternative perspectives’, paper presented at the Conference of Heterodox Economists, March 30th 1999, Nottingham University, UK. Fleetwood, S. (2001) ‘Causal laws, functional relations and tendencies’, Review of Political Economy, 13, 201–20. Gerrard, B. (1995) ‘Probability, uncertainty and behaviour: a Keynesian perspective’, in S. Dow and J. Hillard (eds) Keynes, Knowledge and Uncertainty, Aldershot, Edward Elgar. Harcourt, G.C. (1992) ‘Marshall, Sraffa and Keynes: incompatible bedfellows?’ in C. Sardoni (ed.) On Political Economists and Modern Political Economy: Selected Essays of G.C. Harcourt, London, Routledge. —— (1995) ‘On mathematics and economics’, in Capitalism, Socialism and PostKeynesianism: Selected Essays of G.C. Harcourt, Aldershot, Edward Elgar. Kaldor, N. (1985) Economics Without Equilibrium, Cardiff, University College Cardiff Press. Keynes, J.M. (1936) The General Theory of Employment, Interest and Money, London, Macmillan. Kregel, J. (1976) ‘Economic methodology in the face of uncertainty’, Economic Journal, 86, 209–25. Lawson, C., M. Peacock and S. Pratten (1996) ‘Review article: Realism underlabouring and institutions’, Cambridge Journal of Economics, 20, 137–51. Lawson, T. (1989) ‘Abstraction, tendencies and stylised facts: a realist approach to economic analysis’, Cambridge Journal of Economics, 13, 59–78. —— (1994) ‘The nature of Post Keynesianism and its links to other traditions: a realist perspective’, Journal of Post Keynesian Economics, 16, 503–38. —— (1995) ‘The “Lucas critique”: a generalization’, Cambridge Journal of Economics, 19, 257–76. —— (1997a) ‘Horses for courses’, in P. Arestis et al. (eds) Markets, Unemployment and Economic Policy: Essays in Honour of Geoff Harcourt,Vol. 2, London, Routledge. —— (1997b) ‘Critical issues in economics as realist social theory’, Ekonomia, Winter, 75–117. —— (1997c) Economics and Reality, London, Routledge. —— (2000) ‘Connections and distinctions: Post Keynesianism and critical realism’, Journal of Post Keynesian Economics, 22, 3–14. Lee, F. (1998) Post-Keynesian Pricing Theory, Cambridge, Cambridge University Press. Lewis, P. (1999) ‘Metaphor and critical realism’, in S. Fleetwood (ed.) Critical realism in Economics: Development and Debate, London, Routledge. Mäki, U. (1994) ‘Reorienting the assumptions issue’, in R. Backhouse (ed.) New Directions in Economic Methodology, London, Routledge.

88

Mark Setterfield

McCloskey, D. (1985) The Rhetoric of Economics, Madison, University of Wisconsin Press. Mirowski, P. (1988) ‘Shall I compare thee to a Minkowski–Ricardo–Leontieff–Metzler matrix of the Mosak–Hicks type? Or, rhetoric, mathematics, and the nature of neoclassical economic theory’, in A. Klamer, D. McCloskey and R. Solow (eds) The Consequences of Economic Rhetoric, New York, Cambridge University Press. —— (1989) More Heat Than Light, Cambridge, Cambridge University Press. Pratten, S. (1996) ‘The “closure” assumption as a first step: Neo-Ricardian economics and Post Keynesianism’, Review of Social Economics, 54, 423–43. Setterfield, M. (1997) ‘Should economists dispense with the notion of equilibrium?’ Journal of Post Keynesian Economics, 20, 47–76. —— (1999) ‘Expectations, path dependence and effective demand: a macroeconomic model along Keynesian lines’, Journal of Post Keynesian Economics, 21, 479–501 (Erratum, Journal of Post Keynesian Economics, 22, 340 (2000)). —— (2000) ‘Expectations, endogenous money and the business cycle: an exercise in open systems modelling’, Journal of Post Keynesian Economics, 23, 77–105. —— and J. Cornwall (2002) ‘A neo-Kaldorian perspective on the rise and decline of the Golden Age’, in M. Setterfield (ed.) The Economics of Demand-Led Growth: Challenging the Supply Side Vision of the Long Run, Cheltenham, Edward Elgar. Vickers, D. (1994) Economics and the Antagonism of Time, Ann Arbor, University of Michigan Press. Weintraub, E.R. (1998) ‘Controversy: axiomatisches miverständnis’, Economic Journal, 108, 1837–47.

6

Seeking a role for empirical analysis in critical realist explanation1 Paul Downward, John H. Finch and John Ramsay

Introduction One of the main divisions between critical realism and contemporary practice among economists lies in the area of empirical research. The latter argue that stable patterns persist at different levels of aggregation, emerging from economic phenomena, and that collection, manipulation and analysis of such data provide a useful way of appraising conjectures about economic relationships, and in selecting between different conjectures. The former argue that empirical research that draws upon econometrics is fundamentally flawed because, for example, of its authors’ presumption that general laws can be made manifest from co-variations established from data.This chapter explores this division with the intention of setting out some guidelines for making empirical inferences, starting from an ontological understanding that economic phenomena are characterised by open and complex social systems. In other words, this chapter explores an aspect of epistemology from the perspective of critical realism. Given this primary purpose, of providing guidelines for undertaking empirical research, readers should note that the ontology of complex and open systems are investigated with this ‘pragmatic’ rather than ‘philosophical’ focus in mind. Crucial to this investigation is the interface between the theorising of agents and of economists and, in particular, the proposal that quasi-closure or spatial/temporal closure is achieved in specific contexts by agents who require a degree of stability within their decision-making. The argument develops as follows. In the next section, the chapter discusses critical realist ontology, and reviews the arguments about empiricism and epistemology. In section ‘A perspective from Keynes’, the interface between the theorising of agents and of economists is cast in terms of managing different types of knowledge for different purposes.This is brought to the fore by reviewing critical realist discussions of Keynes. The section ‘A behavioural perspective on agents as active social theorisers’ focuses on different types of knowledge concentrates on the aspirations of economists in making general knowledge claims, and concentrates on issues of composition and aggregation.The section ‘Quasi closure and inference’ draws out practical implications and guidelines for economists who seek to develop causal explanations, especially in coping with interpretive requirements in collating data drawn from social phenomena.

90

Paul Downward, John H. Finch and John Ramsay

Ontological guidelines The implications of critical realism for economics are explored in the work of Lawson (1997) as the outcome of a lengthy and sustained research project discussing the implications of alternative ontological positions in economics. Lawson (1997) argues that neoclassical economics has its roots in the philosophical system of positivism – and in particular a Humean version or ontology – whereby reality comprises the constant conjunction of atomistic events in a closed system.2 Unlike a closed-system ontology, an open-system ontology holds that, in general, the conditions for closure do not apply. Essentially an organic perspective suggests that, for example, human agency is embedded in a social context. Behaviour is thus irreducible to individual action per se but, on the contrary, is both conditional on, and results in, multiple modes of the determination of events. In recognising the organic nature of reality, however, critical realists argue that reality is stratified into three domains. These are the level of actual events, the empirical level of experience and sense impression and, the level of the real where causal relations are located. At a general level alternative epistemologies can be associated with alternative ontological suppositions. Closure facilitates, or entails, a determinate account of phenomena at the level of events through deduction. Theoretical explanation can comprise statements of the form ‘whenever event “X” then event “Y” ’.3 It is generally accepted by critical realists that neoclassical economics is underpinned by an epistemic fallacy, … by the supposition that statements about being can always be rephrased as statements about knowledge (of being), that ontology can be reduced to epistemology. With ontology unavoidable, this error functions only to cover the generation of an implicit ontology, in which the real is collapsed onto the actual which is then … identified with … human experience, measurement or some other human attribute. (Lawson, 1997, p. 62, parenthesis in original) Lawson (1997) describes this approach as ‘empirical realism’. Of particular importance to this chapter are two related aspects of this fallacy. The first is an epistemic conflation between the subject and object of analysis.4 The conception or knowledge of phenomenon manifest in the theorists ideas, and arguments is treated as logically equivalent to the phenomenon per se. In this sense knowledge is presented as being effectively complete. This is either in a literal sense in which theories directly represent an external world, or in the ‘idealist’ sense that the world is merely the proposed collection of ideas. Instrumentalism is an example of the latter case. As Lawson writes, The problem with … [this approach] … is that it is effectively indistinguishable from the view that knowledge is merely a creation of the mind and nothing else exists. Indeed in his critical study of Locke this is precisely the

Empirical analysis in critical realist explanation 91 conclusion that Berkeley draws. If there is no claimed necessary connection between our ‘ideas’ and external reality … what is there to support the view that any external reality exists? (Lawson, 1988, p. 54, parentheses added) In this sense the phenomenon is merely the proposed collection of ideas. The second dimension is inferential and entails the first dimension.The general mode of inference both implied in, and entailed by, a closed system is that premises fully entail conclusions. Deductive logic is the organising principle of arguments. It is because deductive reasoning is directly concerned with, and thus can only cope with, knowledge that already exists or has been acquired, that it promotes the epistemic conflation. Some important consequences follow from the employment of a deductive mode of inference. One is forced to accept a reductionist approach to theorising (Hodgson, 1998).As Bhaskar (1978) argues, a ‘non-additive’ principle is central to closure assumptions. Closure can be described as a situation in which the overall performance or behaviour of the system can be derived as an additive function of the behaviour of the individual components of the system. Nonetheless, ‘… closure is possible, if the principle of organisation … [of the system] … is non-additive, provided it remains constant’ (Bhaskar, 1978, p. 76). There is the presumption here that closed-systems analysis is necessarily reductionist in an atomic sense either literally or, at least, instrumentally. It also implies an emphasis on expressing constant conjunctions of events in mathematical and econometric equations (see also Lawson, 1989a,b, 1997; Downward, 1999). In contrast in adopting an open-systems ontology, researchers are explicitly advocating the relocation of the focus of social scientific explanation to the ‘deeper layers’ that encompass structures and capabilities, whether these are articulated or remain tacit to agents (and researchers).The interactions of these are the ‘generative structures, powers, mechanisms and necessary relations, and so on, that lie behind and govern the flux of events in an essentially open world’ (Lawson, 1994, p. 516). The interactions can lead to what Lawson describes as ‘demi-regularities’ (demi-regs). Thus, an open-systems ontology provides a basis for researchers to expect both co-determination of events and also irregularities. Social phenomena, the interaction of agents and social structures in particular open sub-systems, can result in a range of possible and perceivable events and actions, or no events and actions at all. There is some basis for guiding researchers’ endeavours in formulating knowledge-claims. In contrast to empirical realism, however, the material world as the object of knowledge-claims (including social structures and agents’ capabilities) cannot be conflated with subjects of knowledge-claims, such as categories, relationships and causes. All subjects of knowledge-claims are conjectures that may prove more or less adequate, given different purposes and intentions of those developing and acting upon these subjects (or constituents and artefacts) of knowledge-claims. Knowledge-claims about aspects of social phenomena can lead to agents re-shaping the material conditions of social phenomena, with

92

Paul Downward, John H. Finch and John Ramsay

intended and unintended consequences that are again available for interpretation and refining knowledge-claims. Hence, logical deduction is a misleading pattern or template for researchers to draw upon in making and assessing knowledgeclaims. Rather, inferences will not entail conclusions fully. Further, the dominant pattern of inference espoused by critical realists is retroduction, yet, Not much can be said about … retroduction independent of context other than it is likely to operate under a logic of analogy or metaphor and to draw heavily on the investigator’s perspective, beliefs and experience. (Lawson, 1997, p. 212) In other words, researchers are encouraged to draw upon their personal knowledge in order that they may know and articulate something of one social situation because they already know and can articulate something of another social situation. Beyond this moment of private knowledge, epistemology turns of the explanatory power of a knowledge-claim: Event-predictive accuracy cannot be the criterion of theory selection. Rather, the appropriate criterion outside of the controlled experiment… must be explanatory power. Theories can be assessed according to their abilities to illuminate a wide range of empirical phenomena.And typically this will entail accommodating precisely such contrastive demi-regs as are recorded or can be found. (Lawson, 1997, p. 217, original emphasis) Explanatory power has four implications for empirical research. First, in undertaking the endeavour of making a knowledge-claim, researchers are accepting that the knowledge-claim will be general in character. Second, general explanations are of specific and unique events.Third, researchers may, nevertheless, capture the specific and unique events in subjective and provisional categories. Fourth, researchers draw on private knowledge in order to transfer knowledge artefacts (categories, causes, relationships) between specific and unique events through metaphor and analogy. Categorising events, for instance, is subject to two ambiguities: the events themselves are unique, and researchers can only have partial knowledge of these events through necessarily partial (even false) metaphors and analogies. Lawson (1997) and Runde (1998) address the appraisal of knowledge-claims expressed as causal explanations. Appraisal draws upon the criterion of explanatory power, and this has three aspects. First, an operative mechanism is proposed, and its consequences and effects deduced, through retroduction. Second, proposed consequences and effects are compared with observations of an event. Lawson advises that: ‘there can be no guarantee that any effects will be straightforwardly manifest … the aim must be to try and identify conditions … where the effects ought in some way to be in evidence’ (Lawson, 1997, p. 213).Third, a causal explanation should be explained by seeking some evidence that the proposed mechanism was operating.

Empirical analysis in critical realist explanation 93 The interactions between object and subjects of explanation are developed further by Runde, who sets out four conditions in assessing the adequacy of causal explanations of particular events (Runde, 1998, pp. 158–162). First, a cause that is proposed should be part of the situation in which an event occurred. Second, given the presence of the proposed cause, it should be ‘causally active’.Third, the cause should be ‘sufficient’ given the interests of the researcher and her intended audience. Fourth, an explanation requires ‘causal depth’ in that factors drawn upon in the explanation should not obscure other factors that would have been related to the event anyway. Subsequent sections of this chapter seek to elaborate upon the role of empirical analysis in setting out causal explanations. In the following sections, a suggestion is offered as to how and why different forms of evidence may be combined in undertaking research with a basis in open-systems ontology. Specifically, the issue of making knowledge-claims of a more general nature than specific events, which is implicit in the analogical or metaphorical aspects of retroduction, is offered. In so doing, the chapter argues that in practice, the abstract open-system ontology requires a degree of modification in recognition of the stability that characterises many social phenomena within spatiotemporal zones.The modification is a largely unintended consequence of agents’ actions. Nevertheless, the reflexive and knowledge-acquiring capabilities of agents provide limited opportunities for researchers to gain access to aspects of social phenomena. Essentially, the argument is that agents themselves cope with the ambiguities and irreducible uncertainties inherent in open systems. And agents are theorists, with different purposes and subjects than researchers looking in on the same phenomenon. Stability is both a material condition of theorising, and in a different and perhaps largely unintentional sense, a consequence of theorising. The analysis of the modified ontological basis, or quasi-closure, begins with a review of Keynes’s explanation of how agents cope with ambiguity and irreducible uncertainty.

A perspective from Keynes The relationship between agents drawing inferences from knowledge in providing explanations and undertaking actions, and researchers drawing inferences from agents’ actions is central to Keynes’s philosophical perspective.This is captured in the Treatise on Probability (Keynes, 1973a), which offers guidelines on inference. Authors have argued that Keynes’s philosophical position is consistent with critical realism (Lawson, 1985, 1988; Bateman, 1991; Dow, 1994), is based upon an organic ontology (Carribelli, 1988, 1995; Dow, 1994), and implies that statistical inference cannot sustain the generalisation of knowledgeclaims.While authors, who wish to claim Keynes as a critical realist, face some significant difficulties, Keynes’s commitment to a mode of inference different from logical deduction, and numerical probabilities has been an important source of motivation for many researchers. Keynes’s location of probabilities as subjects or objects of knowledge-claims is instructive in this

94

Paul Downward, John H. Finch and John Ramsay

matter. An interpretation of Keynes is that probabilities are not simple records of reality, in which data of events are collected and arranged easily into ‘given’ categories, such that Keynes’s probabilistic mode of inference can be described as non-dualistic (i.e. non- ‘relative/absolute’) interpretation. Hence, subjects of knowledge-claims are ‘always open to constant modification and correction through continual interaction with the experiences of the real world’ (Lawson, 1988, p. 56). Subjects and objects of knowledge-claims are not conflated because probability statements are conditional upon current knowledge, which will change as knowledge evolves. Following Keynes, the net is cast wider in formalising the notion of inference as part of social scientific endeavour. Informal argument and judgement involving personal knowledge combine with formal models and statistical inference. Further, Keynes argues that researchers can avoid the problem of induction by examining a particular phenomenon in multiple contexts. Commonality of some aspect across contexts provides a basis to strengthen belief in the proposition that an aspect is significant in the phenomenon too. If different contexts reveal non-common aspects, the weight of argument weakens, revealing a researcher’s ignorance. Hence,‘An inductive argument affirms, not that a certain matter of fact is so, but that relative to certain evidence there is probability in its favour’ (Keynes, 1973a, p. 245, original emphasis). It is important to recognise the relative or conventional nature of the subjects of a knowledge-claim pursued in this manner.A researcher needs to perceive a phenomenon, and aspects involved in that phenomenon. Critical realists highlight retroduction, including analogy or metaphor, as important in this process.Then, multiple instances of the phenomenon are to be identified. It is not clear whether this involves metaphor or analogy.The original identification of a phenomenon has now been articulated, and articulation may be modified as further instances are identified, but further identifications themselves confirm and help articulate the original specification. The interpretation of Keynes’s philosophical position is, though, contestable. Reality may be interpreted as comprising events that have numeric properties that translate onto an interval or ratio numeric scale of measurement, and can be described by a degree of risk synonymous with measures of dispersion. Other events may be interpreted as comprising events that do not possess numeric properties, and so cannot be described by measures of dispersion and are instead uncertain. As with Knight (1921), risk and uncertainty can be interpreted as a dualism with material bases. Conventional behaviour by agents could also be interpreted as implying that agents are irrational. However, Keynes’s view of agents as broadly rational yet capable of rule-following behaviour is augmented by agents being knowledge-acquiring beings in The General Theory (Keynes, 1973b).

A behavioural perspective on agents as active social theorisers The argument of this section is that agents are practical theorists, and this has three implications for researchers wishing to make inferences as part of

Empirical analysis in critical realist explanation 95 empirical research projects, while adhering to an open-systems ontology. First, there is a tension between the presumed complexity inherent in an open system, and the ability of agents to make sense of their circumstances and act within these. Complexity requires active theorising, and active adaptation of theories in the face of accumulated experience and environmental change. But agents’ capabilities in theorising, and especially in adapting practical theories, imply that complexity is a surmountable condition, through interactions among socially situated learning of individual agents.This is itself a phenomenon and object of social scientific inquiry. Second, any success that agents have in basing actions on practical theorising can provide a basis for conditions of ‘quasi-closure’, which is a particular type of state within the general opensystem condition. If so, the composition of stability should be studied. Third, practical theorising by agents of their circumstances provides some access for researchers to a social phenomenon.This access is inherently partial, given the degree to which agents depend upon tacit knowledge as part of their repertoires of understanding and action. Furthermore, other aspects, such as socially inherited rules, habits, customs and social structures, are expected to be practically beyond agent theorising. Agents’ practical theorising and action Since agents have the ability to act, they also have the ability to act differently. This is an important basis for inferring that social phenomena can be characterised as open systems. The social systems themselves comprise devolved knowledge and expertise, with different formal and informal connections, and this will be considered in the next section.To begin with though, the focus is on how agents as individuals cope with complexity. Behavioural, evolutionary and institutional economists have been focusing on individual decisionmaking within social phenomena for some time. Crucial to their arguments have been understanding how agents draw upon different types of knowledge, with varying degrees of consciousness in understanding their situations, devising plans, undertaking activities and monitoring the consequences of these activities. Types of knowledge have come to be conventionally divided into tacit or personal knowledge, and articulated, codified or declarative knowledge. Another distinction has been made between ‘know-how’ and ‘know-that’, and this is often seen as synonymous with the knowledge typology above. Following Fleetwood (1995), agents are effective in undertaking a whole series of actions because they remain ignorant of their tacit knowledge, implying that ignorance in this particular sense is an enabling factor rather than an encumbrance to successful action. The two-way typology of dimensions of knowledge – tacit or personal, and codified, articulated or declarative – should not convey dualism of knowledge. Tacit or personal knowledge is difficult to communicate between agents. Difficulties in communicating an agent’s expertise, as a skilled process rather than, or in addition to, some product or outcome, may, though, be overcome

96

Paul Downward, John H. Finch and John Ramsay

over time and with effort. If a dualism is unhelpful in representing knowledge drawn upon by agents in complex situations, so it is also unhelpful to portray tacit and codified knowledge as extremes in a continuum. Communicating skills and expertise as a process, or articulating tacit knowledge, implies that a good deal remains tacit. Simplistically, for agents to understand one another while trying to communicate expertise, some kind of codification and de-codification language system should pre-exist any attempts at communication. Know-how and know-that are essentially complements, with know-how being a peculiar form of ignorance. Also, know-how may well have both personal and shared components. Finally, there is no absolute guarantee that know-that, comprising declarative knowledge, will be understood and communicated as its possesssor intends. Characterising social phenomena as open systems in terms of how its agents manage different types of knowledge and ignorance illustrates in some detail what is meant by complexity.The establishing or emergence of order and co-ordination given such complexity is, of course, a long-established theme among social scientists. Recent studies of Hayek and of Simon are of particular interest in this regard (Fleetwood, 1995; Rizzelo, 1999). Two strands may be drawn from the behavioural approaches while focusing on individual agents. First, individuals contend with bounded rationality in the form of procedural rationality (Simon, 1976a,b, 1978).This may be thought of as bounded cognition (Conlisk, 1996; Loasby, 1999).There is simply too much devolved knowledge – in both tacit and articulated form – for individual agents to make sense of, even when focusing on their own circumstances as loosely defined sub-set of an overall social system. Localisation and specialisation cannot completely protect an agent from the effects of complexity, even in their own particular neighbourhoods. Second, agents face or experience a range of incentives to acquire understandings of aspects of their circumstances, and researchers may interpret these as practical theories. In other words, agents endeavours in coming to understandings is motivated (to some extent) by practical purposes (or material causes). These purposes may include getting things done, but getting things done requires also a degree of comfort, or reaffirmation of belief, which has been termed ‘ontological security’ (Giddens, 1984). It is not that agents deliberately seek ‘ontological security’. Rather, ontological security is a necessary condition for the application of practical knowledge. Following Peirce (1934), ‘doubt is parasitic upon belief ’. Practical theories and ontological security may be discussed in the context of Kelly’s (1963) work on personal constructs. Through these, Kelly explains that individuals understand through developing interpretive categories, and associating these in an inductive process with outcomes that are perceived to be in some way successful. In other words, individuals develop a system of rules for understanding and acting in a range of circumstances, drawing analogically upon these accumulated patterns in novel situations.The key to psychological well-being is acquiring processes for identifying and strengthening useful associations and rejecting other associations. Hence, Kelly fuses what was discussed

Empirical analysis in critical realist explanation 97 earlier in this section as personal and declarative knowledge, and know-how and know-that. These procedures are opaque to individuals, but Kelly devises a procedure to uncover or diagnose them through questionnaire work aligned with repertory grids that ask agents to comment on similarities and differences across groups of three artefacts. Kelly’s approach of personal constructs bears a resemblance to Brady’s (1987) cognitive interpretation of Keynes’s theory of probability. Again, probability is subjective and ordinal, drawn from agents’ analogies of experiences. And action forms a test of the practical usefulness of the analogies, providing a basis for adaptation where there is scope for formulating expected outcomes, which become benchmarks against which perceived outcomes may be compared, and perceived discrepancies explained by reasons. Social aspects of practical theorising: sharing practices Fleetwood (1995) argues that Hayek approached something like an opensystems ontology as central to his explanation of social order once he recognised the limitations of methodological individualism co-ordinated by the telecommunications system of the price system.5 Instead, Hayek recognised that learning agents bound together through the many social institutions rules and customs that characterise his catalaxy should replace this. Ontological security, which is an individual issue, is simultaneously a social phenomenon as other agents, whose actions are perhaps manifestations of markedly different personal constructs, provide bases of doubt and belief. An example of this is Choi’s (1993, 1999) explanation of entrepreneurship. At an individual level, Choi argues that entrepreneurs possess distinct and subjective mental models. He also introduces a compositional principle by which societies possesss different tolerances to entrepreneurship through the willingness of its members to have established, that is recognised, customs and practices upset by entrepreneurial activity.6 Hence, potential entrepreneurs face differing social pressures to pursue approval-seeking strategies, and avoid disapproval. Paying attention to institutions, customs and shared practices does not imply that a social phenomenon is characterised by event regularity: the ubiquity of habits and rules should not be taken to mean that [novelty and creativity] are excluded. On the contrary, a theoretical focus on habits and rules should include explanations of their origin, evolution, breakdown and replacement. (Hodgson, 1997, p. 679) Agents possess capabilities that, where drawn upon in understanding, planning, undertaking and reviewing actions, affect the degree of turbulence within social phenomena as an essential part of the development of phenomena. Without co-ordination, social phenomena may tend towards the categories of under- or over-socialisation (Granovetter, 1985). While not ruling out either eventuality in practice, these are unusual.

98

Paul Downward, John H. Finch and John Ramsay

The problem for researchers interested in explaining stability and co-ordination among social phenomena is to explain compositional principles that connect individuals’ mental models or personal constructs with social structures, institutions and customs. The process has come to be termed ‘emergence’, such that social phenomena are themselves emergent systems, and also the structures, institutions, rules and customs are ‘emergent properties’ of social systems (Archer, 1995). Emergence implies that social phenomena cannot be analysed (adopting analysis in the sense of break up into smaller parts) without loss of meaning for the overall phenomena that forms the research object (Hodgson, 1988). If, for purposes of analysis, a researcher examined a social phenomenon as if decomposed into smaller units, focusing on the smaller units detracts attention form its connecting principles (Simon, 1962; Loasby, 1976; Potts, 2000). The compositional problem has been addressed, albeit somewhat elliptically, among researchers of the resource-based or capabilities approach of industries and firms. For example, Nelson and Winter (1982) describe individuals as possesssing skills, and, analogically, firms possesssing routines. Capabilities occupy a grey area, perhaps in between an individual’s skills and a firm’s routines, and perhaps expanding to envelop both terms. Capabilities are tacit to a high degree, so are difficult to designate property rights to, and exchange between individuals or firms. But they may include some more tangible artefacts such as standard operating procedures, even though these presume a good deal of tacit knowledge among a firm’s members in knowing how to adhere to (Cyert and March, 1992). In detail, capabilities are inalienable from particular contexts, be these firms, households, organisations or individuals. Nevertheless, necessarily imperfect imitation of aspects of capabilities more amenable to translation and communication via articulation and codification can be expected. This is also captured within Choi’s (1993, 1999) explanation of entrepreneurship. Denzau and North (1994) and Lane et al. (1996) both point to attempts by agents at sharing mental models as an attempt to cut short learning costs. The emergence of quasi-closure and its implications for research So far, it has been argued in this section that order often emerges within social phenomena that are characterised as open social systems, partly because agents seek a degree of stability (or ontological security), and partly because agents have the capacity to reproduce conditions of stability anyway and often as an unintended consequence of being practical theorisers. The discussion has though been couched in terms of agents’ practical theorising, given agents’ interests in developing practical or context-specific expertise. Following an open-systems ontology, the discussion has been mainly at the empirical and actual levels, with occasional reference to deeper social structures and agents’ capabilities. In arguing for a modified form of an open-systems ontology, described earlier as ‘quasi-closure’, the argument of the critical realists that developments within the different levels of reality are to some extent autonomous of one another is addressed. An open-systems ontology implies

Empirical analysis in critical realist explanation 99 a degree of autonomy within each level (empirical, actual and deep) such that changes in different levels will be out of sequence with one another.Autonomy can be assured if agents possess and maintain opaque understandings of the deeper structures that help govern their material conditions. But autonomy does not necessarily imply, dualistically, that developments within the different levels of reality will be out of sequence rather than in sequence, and that events cannot categorically be epiphenomenal of developments at the deep level. Institutions are emergent within social phenomena and play a coordinating role, being a store of knowledge in particular contexts, and are ‘sustained by ‘subjective’ ideas in the heads of agents and are also ‘objective’ structures faced by them … . Both individuals and institutions are mutually constitutive of each other. (Hodgson, 1998, p. 12) If such a state emerges, where, from agents’ perspectives, the subjective and objective converge or coalesce, then the situation may be described as an open system characterised by quasi or spatial-temporal closure.7 There are different aspects to this, corresponding to different layers of an open-systems ontology. Agents’ practical theorising has been described above mainly in terms of realms of agent consciousness, at least in the sense that plans can be formed and form a basis or benchmark against which perceived outcomes can be compared.This implies a process or tendency of convergence in expectations among agents involved in a situation. Counter-tendencies involve both radical entrepreneurship, and also less radical observations of another agent’s displaying manifestations of different thinking, interpretation, purpose or action. Hence, quasi-close is not ubiquitous and is expected to occur within spatial-temporal bounds. Tendencies to (or from) convergences among agents’ mental models is not intended to over-privilege agency in explanations of social stability. Reference is necessarily made to the deeper structures, rules, customs and institutions, of which agents are expected to remain largely ignorant even when carrying out reflexive cognitive procedures such as planning, and explaining outcomes. Following Hodgson (1999), these aspects of agency and of structures do not merge into the dualism of structuration because they exist and occur at different levels of reality, being distinct aspects of phenomena that in principle do not fuse during moments of action when agency is involved in reproducing conditions of phenomena. However, the agency-related aspects of quasiclosure or spatial-temporal closure and the structural aspects can be interpreted in a similar manner to Bhaskar’s (1979) description of conditions of closure. Agency aspects have the potential to approach something like intrinsic closure with an alignment of expectations surrounding reflexive agency, and structural aspects have the potential approach, something like extrinsic closure. Two consequences follow from the argument of this section. First, predictability and an open-systems ontology are not necessarily inconsistent because particular phenomena can be characterised by quasi-closure or spatial-temporal

100

Paul Downward, John H. Finch and John Ramsay

closure. Second, the role of prediction as a research activity within social scientific endeavour can be better understood. Regarding the first point, agent prediction and researcher prediction are not conflated. Agent prediction is in the context of an agent’s motivations for action, which are subjective but undertaken in the context of overlapping theories, mental models and plans of other agents involved in a phenomenon. Researcher prediction, if undertaken as part of a wider explanatory research project, is for different purposes of making and assessing general knowledge-claims in the context of a social scientific community of inquirers (but can also extend to policy advice for a range of agencies involved in the phenomena, but from the general perspective).Arguments have been made to the effect that stability can be more apparent from a systemic perspective than a sub-systemic level.The systemic level is more likely to characterise a higher proportion of researchers’ perspectives of phenomena, investigated for the purpose of making knowledge-claims of a more general nature than are useful within a particular phenomena (Simon, 1962; Earl and Kay, 1985). Such a systemic point of view, in which researchers specialise in making knowledge-claims connecting the particular as instances of the general, and agents specialise in testing knowledge-claims through experience in sub-systemic contexts, is compatible with a view of composition or aggregation that is emergent rather than additive. The role of prediction is a more problematic point. Just because quasi-closure, or spatial-temporal closure, is a feasible characterisation among social phenomena, does not necessarily mean that prediction is an important social scientific activity. Following the argument of this section, prediction is itself dependent on an inquiry and explanation that first establishes quasi-closure, and ideally makes some further knowledge-claims about the nature of bounds of this characterisation, such that knowledge-claims extend to outlining conditions – necessarily feasible, given an open-systems ontology – under which quasi closure is infringed. Inference, including prediction, is discussed fully in the following section.

Quasi-closure and inference This section addresses the question of the status of empirical research, given an open-system ontology, and, further, seeks to provide some guidance in conducting empirical research. As a starting point, we argue that agents possess practical theories and knowledge, comprising tacit and articulated components, which are not expected to be of deeper structures, rules and customs to any great extent. Critical realists have been sceptical of the usefulness of attempting to interpret agents’ accounts as an integral part of social scientific inquiry, given the partial coverage of such knowledge in its articulated or codified form.This seems to be an overreaction against what may be seen as a radical subjectivist position that over-privileges agents’ articulable or codifiable (though not necessarily articulated and codified) conceptions of their situations in ethnographic research programmes that seem to progress on the basis that agents’ conceptions exhaust the phenomena under investigation.

Empirical analysis in critical realist explanation 101 Interpreting agents’ conceptions Institutional and Post Keynesian economists have an established tradition of undertaking empirical research, and there have been recent additions to this investigating pricing decisions within firms (Tool, 1991). Downward (2000) provides a Post Keynesian analysis of pricing, Lee and Downward (1999) assess Means’s theory of administered pricing, and Downward (1996) provides a new empirical study of pricing with an institutional and behavioural perspective. It could, though, be argued that these studies pay too little attention to the philosophical and methodological consequences of interpreting agents’ understandings of their activities, or hermeneutics. In effect, researchers are trying to understand agents’ motivations and framings, while agents themselves are also trying to establish plans, frameworks and mental models in the context of others’ activities in making plans drawing on frameworks and mental models. If the researcher accepts that, philosophically, agents’ reflections, as articulated and explained, cannot exhaust a phenomenon, and is more likely to give a partial and context-dependent account of part of that phenomenon, ethnographic approaches may be helpful for social scientific inquiry. From an open-system ontology, we can expect that, given a fair degree of interpretive endeavour, we may gain access to multiple and overlapping perspectives of agents on events, so comprising what critical realists term the domain of the actual. Although not strictly an ethnographic approach (Alvesson and Sköldberg, 2000), grounded theory procedures can provide some useful techniques (Finch, 1999, 2002). Glaser and Strauss (1967) provide a grounded theory approach, which is strongly inductive, in which the researcher has an interpretive role in categorising primary data (such as transcripts of open-ended and semi-structured interviews, field notes and participant-observation accounts). Researchers analyse this information with the intention of providing an explanation of phenomena, recognisable by both its agents, and also capable of supporting more general knowledge-claims that may be assessed among other phenomena.The inductive ethos of this approach to grounded theory leads researchers to develop knowledgeclaims within contexts, rather than test conjectures or predictions deduced from established theories, or even using established theories to deduce research instruments such as questionnaires. The grounded theory strategy for generalisation is captured in a distinction between theoretical and statistical sampling. Theoretical sampling involves researchers making successively general knowledge-claims because categorisations of phenomena become saturated. And saturation is assessed in terms of failing to yield surprises and contradictions. Statistical sampling presumes that the researcher already knows enough of some population, such that a sub-set of it may be drawn at random, for the purposes of making inferential statements supported probabilistically by statistical principles. If the researcher does not possess such reliable knowledge, statistical sampling avoids surprises by what is, retrospectively, over-sampling the middle of some stable distribution. Theoretical sampling on the other hand, seeks surprises by directing

102

Paul Downward, John H. Finch and John Ramsay

a researcher’s attention towards establishing categorical boundaries. It can be argued that the inductive approach of grounded theory is at odds with the fallibilism of grounded theory (Baert, 1996).8 Although the generalisation strategy of grounded theory facilitates later fallibilism in encouraging researchers to adapt and revise knowledge-claims. The main difference though between grounded theory as a set of procedures for guiding empirical research, and critical realism as a philosophical-ontological position, is that grounded theory tends to orient researchers to knowledge-claims made in the mid-range (Ingham, 1996). This criticism of grounded theory highlights in the realm of the deep, structures, customs and institutions that are general in nature, and inaccessible to techniques that centre on primary data, interview transcripts, participant observation accounts and other primary data. Grounded theory may be compared with a whole range of analytical procedures that avoid the fusing or conflation of object and subjects of inquiry. Znaniecki (1934) argues in favour of analytical induction. This provides guidelines requiring researchers to establish conceptual definitions of processes by identifying their ‘essential’ characteristics. If the essential characteristics of processes can be identified, this provides a logical basis to generalise across other similar processes by case study analysis. At the same time, similarity is confirmed if evidence is collated such that the presence of a similar process can be identified. A contemporary example of this argument can be found in March et al. (1991). Similarly, Hartley (1994) argues that there is a logical relationship between identifying a phenomenon of interest, and identifying some sub-system or process of that phenomenon: The detailed knowledge of the organization and especially the knowledge about the processes underlying the behaviour and its context can help specify the conditions under which the behaviour can be expected to occur. …generalization is about theoretical propositions and not about populations. (Hartley, 1994, pp. 225–226) An underlying process is though not easily available from primary data, such as collections of interview transcripts, field notes, participant observation notes and published reports.The researcher needs to intervene, categorise, undertake further interviews, seek multiple perspectives on, and interpretations of, objective events.9 It also requires analytical interventions by researchers drawing on theoretical explanations of routines, customs, institutions and social structures. Researchers face a judgement or decision when undertaking these interpretive or hermeneutic endeavours in substantiating their knowledge-claims while undertaking an empirical inquiry. This is whether to produce ‘thick’ or ‘thin’ narratives (Geertz, 1993). Researchers are not attempting to reproduce an agent’s knowledge or practical theorising activities. Further, given the different purposes of the knowledge of agents and researchers of particular contexts, the expectation is that the researcher’s narrative, although influenced

Empirical analysis in critical realist explanation 103 strongly by agents’ narratives, will be thinner. Generalisation, as a researcher’s overriding aim, implies a tendency from thick knowledge of contexts to thinner knowledge of additional contexts, unless artefacts of theorising fail repeatedly, requiring re-conceptualisations and adaptations. Narratives though can become so thin as to be banal, and capture very little variation so as to be undiscriminating across the number of cases of a phenomenon included within a research project. Researchers also encounter the problem of induction in assessing the generality of knowledge-claims developed over a series of case studies that are deemed to be similar and comparable. Harré and Madden (1975) argue that the problem is comprehensible given a particular conjunction of ontology and epistemology that has been termed ‘Humean’. Humean ontology implies that individuals in any sequence of events (or sequence forming an event) are assumed to be unconnected. Epistemologically, a world of independent events makes no sense in terms of causation based on connections between events. Ontologically, the logical necessity for repetitions of particular sequences of events in the future is removed. Given independence of events, it is possible logically that an explanation based on specific sequences may be incorrect and in need of adaptation and revision in the future. Empirical and statistical inference Four questions remain for researchers undertaking empirical research, but with an open-systems ontology: How will I know if the characteristics I have identified are essential? How will I know if the knowledge-claim I have articulated is unassailable? How will I be able to tell when I have identified a cause? How can I be sure that the process description is ‘thick’ enough to permit causal explanation? Statistical analysis can have a role in such inferential procedures.This is not in order to establish event regularities in order to validate or falsify universal laws. Rather, the role is one of complementing and corroborating knowledgeclaims established during the overall process of causal research. Any statistical analysis depends upon the researcher formulating categories from the different types of data collated during primary investigation in particular contexts. Inference, which is epistemological, is dependent upon the ontological issue of stability among the phenomena being investigated, which we have termed quasi-closure or spatial-temporal closure. It also requires a principle of composition, different from drawing upon addition as a means of aggregation, such that connections may be made, reliably, between the overlaps and spillovers across agents’ practical knowledge and mental models, and identification of a phenomenon as a social system. Systemic stability facilitates variation and experimentation as devolved locations of a social system (Simon, 1962). Likewise, devolved locations facilitate variation, experimentation and entrepreneurship within bounds while reproducing and maintaining systemic connections such that its members may reproduce an identifiable whole. Systemic

104

Paul Downward, John H. Finch and John Ramsay

stability is then related loosely to a different kind of stability in agents’ practical theorising and mental models, but categorising these types of stability, as either in sequence or out of sequence, is the product of prioritising ontology to the exclusion of epistemology. It logically follows that empirical research can proceed, and include statistical inference, if sequencing is consistent with quasi-closure or spatial-temporal closure.

Conclusion The main argument of this chapter is that empirical inferences can be made by social scientists who invoke an open-systems ontology, and that these require an appeal to causal structures as an essential component of investigating the presence and implications of demi-regularities.The vital arena of investigation is the connection between the knowledge-claims of agents in their contexts and how these relate to the knowledge-claims of researchers, usually made in a more general manner and in the different context of a scientific community. Satisfactory explanations must include institutional and behavioural contexts as bases of agent belief and action. Researchers may also draw upon Keynes’s argument, that to avoid the problem of induction, a type of phenomenon should be investigated in different contexts. However, the discussion of this chapter also suggests that Keynes’s logic of inference can be modified to reflect the critical realist ontological position.This can be achieved with a restated role for investigating conventional behaviour outlined, for example, by adopting some of the grounded theory procedures. If institutional analysis suggests that there is evidence that a particular set of processes underlies events, this would prima facie support, and justify, research at the level of events characterised as demi-regularities (and vice versa depending on the orientation of research) that may appeal to relative frequencies and indeed econometric evidence. Overlap or consistency in the results of investigations in different contexts adds weight to the realist account of the phenomenon under investigation and, by implication, increases the basis of a rational belief in that account.The compelling attraction of modifying Keynes’s arguments for the critical realist is that it explicitly indicates the importance of context-driven, causal research, yet simultaneously copes with problems of induction by allowing for legitimate analysis of a quantitative nature at the level of events or demi-regularities. A degree of legitimate generalisation is catered for. Combined, such insights produce a basis for theory evaluation. Crucially, theory is the mediating construct of these inferences and provides the interface between ontological claims and the various interpretations of ‘data’ manifest as different types of observations of the real world. Theory is neither a simple induction nor completely a priori. Complex interaction entails that theory choice is as much concerned with ontological claims as with empirical evidence per se. It follows that theory is neither simply reducible to, nor should be conflated with, reality.The subject and object of analysis and inference interact both at the level of those under investigation as well as those investigating them.

Empirical analysis in critical realist explanation 105 The paradigmatic context of theorising and inferences also needs to be considered. Echoing the notion that agents’ choices are shaped in a Kuhnian sense, the social context of economists’ rational belief needs to be considered. It is difficult to envisage scientific progress in the fallibilist sense to which critical realists adhere without such context. Shared understanding, expressed in a common conceptual language, will form the basis of notions such as ‘conventional and accepted knowledge’ of, for example, causal mechanisms, as well as what is accepted as ‘scientific discovery’ by a particular scientific community. Inferences are likely to involve social discourse in establishing acceptable interpretations as well as the simple triangulation of qualitative and quantitative insights.

Notes 1 This is a revised version of ‘Critical Realism, Empirical Methods and Inference: A Critical Discussion’, by the same authors published in the Cambridge Journal of Economics. 2 The Humean basis of neoclassical economics is not unequivocal in that Hume’s problem of induction is concerned with the ontological state of unconnected atomistic events. It is clear that neoclassical economists have a notion of causality and connectedness in their theorising, though they make closure assumptions.Two forms of closure are central to this perspective.The intrinsic condition of closure – which can be loosely characterised as implying that a cause always produces the same effect – suggests that the structures of the phenomena under study are constant, unchanging and for any intrinsic state only one outcome is possible. The extrinsic condition of closure – which can be loosely understood as implying that an effect always has the same cause – suggests that the phenomena under study are isolated from other potential influences. 3 Neoclassical economists also move between scales of analysis, that is sub-systems and systems themselves, in additive forms.They adopt a reductionist approach. For more on this see Hodgson (1998). 4 As Trigg (1989) notes, this subject/object dual has had a number of manifestations. The language/sensory experience, conceptual/real and mind/world duals are other manifestations. In the current context the subjectivity/objectivity or epistemic/ aleatory nature of probabilities are examples. 5 This is from The Constitution of Liberty onwards. 6 The implication here is interesting as it implies an overlap between the conceptually distinct domains or levels of actual and deep, that critical realists draw upon in explaining the structuring of social phenomena. 7 Setterfield (1997, pp. 69–70) and in this book argues that using ceteris paribus clauses is ‘a form of “synthetic” or conditional closure to what is otherwise an open, historical system’ (ibid., p. 70, original emphasis). Cottrell (1998) argues that agency itself presupposes local closures. 8 Pratt (1994, 1995) argues that grounded theory can have a role in ‘putting realism to work’.This role is in undertaking abstraction from empirical observation, and interpretation of agents’ explanations. The emphasis on interpretation is though a complicating factor as agents are expected to have a partial understanding drawn from everyday practical theorising, and especially an opaque understanding of the deeper structures, customs and institutions that shape the context of action. A researcher cannot just generalise from accumulating observations and engaging in a series of conversations, as generalisation is a different type of knowledge to that held by interviewees and participants within contexts, and is expected to refer to the ‘deeper’

106

Paul Downward, John H. Finch and John Ramsay

aspects of phenomena. Layder (1998) is critical of proponents grounded theory procedures for effectively abandoning existing theoretical insights of these ‘deeper’ aspects that may otherwise be useful in directing – though not determining – empirical enquiry. 9 Harper (1992, p. 139) argues that the comprehension of instances of such qualitative field work, which may be termed cases, have a dual character with regards to inference of ‘situational groundedness and theoretical generality’. A case is both of a more general category that pre-exists a case’s identification, and is also expected to add to knowledge of that general category.This may be incremental in nature, or more radical in providing grounds for criticising the basis of a categorical system (Wieviorka, 1992, pp. 159–160). The challenge in making knowledge-claims through inductive inference is in associating cases with wider classifications or types. Identifying a case also involves assessing the relevance of the wider categorical system as an adequate subject of a knowledge-claim.

References Alvesson, M. and Sköldberg, M. (2000) Reflexive Methodology. New Vistas for Qualitative Research, London, Sage. Archer, M. (1995) Realist Social Theory. The Morphogenetic Approach, Cambridge, Cambridge University Press. Baert, P. (1996) Realist Philosophy of the Social Sciences and Economics: A Critique, Cambridge Journal of Economics, 20, 513–522. Bateman, B.W. (1991) Das Maynard Keynes Problem, Cambridge Journal of Economics, 15, 101–111. Bhaskar, R. (1978) A Realist Theory of Science, Sussex, Harvester Press. —— (1979) The Possibility of Naturalism, Sussex, Harvester Press. Brady, M.E. (1987) J.M.Keynes’ Theory of Evidential Weight; Its Relation to Information Processing Theory and Application in the General Theory, Synthese, 71, 37–59. Carabelli, A.M. (1988) On Keynes’s Method, London, Macmillan. —— (1995) Uncertainty and Measurement in Keynes: Probability and Organicness, in Dow, S.C. and Hillard, J. (eds), Keynes, Knowledge and Uncertainty, Aldershot, Edward Elgar, pp. 137–160. Choi, Y.B. (1993) Paradigms and Conventions. Uncertainty, Decision-Making and Entrepreneurship, Ann Arbor, University of Michigan Press. —— (1999) Conventions and Learning: A Perspective on the Market Process, in Dow, S.C. and Earl, P.E. (eds), Economics Organization and Economic Knowledge. Essays in Honour of Brian J. Loasby, volume I, Cheltenham, Edward Elgar, pp. 57–75. Conlisk, J. (1996) Why Bounded Rationality?, Journal of Economic Literature, 34, 669–700. Cottrell, A. (1998) Realism, Regularities and Prediction, Review of Social Economy, 56, 347–355. Cyert, R. and March, J.G. (1992) A Behavioral Theory of the Firm, second edition, Oxford, Blackwell. Denzau, A.T. and North, D.C. (1994) Shared Mental Models: Ideologies and Institutions, Kyklos, 47, 3–31. Dow, S.C. (1994) Methodological Pluralism and Pluralism of Method, Discussion Paper 94/10, Department of Economics, University of Stirling. Downward, P.M. (1996) A Behavioural Perspective on Business Pricing, in Earl, P.E. (ed.), Management, Marketing and The Competitive Process, Aldershot, Edward Elgar. —— (1999) Pricing Theory in Post Keynesian Economics: A Realist Approach, Aldershot, Edward Elgar.

Empirical analysis in critical realist explanation 107 Downward, P.M. (2000) A Realist Appraisal of Post Keynesian Pricing Theory, Cambridge Journal of Economics, 24, 211–224. ——, Finch, J. and Ramsay, J. (2002) Critical Realism, Empirical Methods and Inference: A Critical Discussion, Cambridge Journal of Economics, 26(4), 481–500. Earl, P.E. and Kay, N.M. (1985) How Economists Can Accept Shackle’s Critique of Economic Doctrines Without Arguing Themselves Out of a Job, Journal of Economic Studies, 12, 34–48. Finch, J.H. (1999) The Methodological Implications of Post Marshallian Economics, in Dow, S.C. and Earl, P.E. (eds), Contingency, Complexity and the Theory of the Firm. Essays in Honour of Brian J. Loasby, volume II, Cheltenham, Edward Elgar, pp. 155–176. —— (2002) The Role of Grounded Theory in Developing Economic Theory, Journal of Economic Methodology, 9(2), 213–234. Fleetwood, S. (1995) Hayek’s Political Economy.The Socio-economics of Order, London and New York, Routledge. Geertz, C. (1993) The Interpretation of Cultures, London, Fontana Press. Giddens, A. (1984) The Constitution of Society, Cambridge, Polity Press. Glaser, B.G. and Strauss, A.L. (1967) The Discovery of Grounded Theory: Strategies for Qualitative Research, New York, Aldine Publishing. Granovetter, M. (1985) Economic Action and Social Structure. The Problem of Embededness, American Journal of Sociology, 91, 481–510. Harper, D. (1992) Small N’s and Community Case Studies, in Ragin, C.C. and Becker, H.S. (eds), What is a Case? Exploring the Foundations of Social Inquiry, Cambridge, Cambridge University Press, pp. 139–158. Harré, R. and Madden, E.H. (1975) Causal Powers:A Theory of Natural Necessity, Oxford, Blackwell. Hartley, J. (1994) Case Studies in Organizational Research, in Cassell, C. and Symon, G. (eds), Qualitative Methods in Organizational Research, London, Sage, pp. 208–229. Hayek, F.A. (1960) The Constitution of Liberty, Routledge and Kegan Paul, London. Hodgson, G.M. (1988) Economics and Institutions: A Manifesto for a Modern Institutional Economics, Cambridge, Polity Press. —— (1997) The Ubiquity of Habits and Rules, Cambridge Journal of Economics, 21, 663–684. —— (1998) From Micro to Macro: The Concept of Emergence and the Role of Institutions, paper presented at Post Keynesian Study Group meeting, University of Greenwich, January. —— (1999) Economics and Utopia, New York and London, Routledge. Ingham, G. (1996) Critical Survey: Some Recent Developments in the Relationship between Economics and Sociology, Cambridge Journal of Economics, 20, 243–275. Kelly, G.A. (1963) A Theory of Personality.The Psychology of Personal Constructs, New York and London, Norton. Keynes, J.M. (1973a) The Collected Writings of John Maynard Keynes,VIII, A Treatise on Probability, London, Macmillan. —— (1973b) The Collected Writings of John Maynard Keynes,VII, The General Theory of Employment, Interest and Money, London, Macmillan. Knight, F.H. (1921) Risk, Uncertainty and Profit, Houghton Miflin, New York. Lane, D., Malerba, F., Maxfield, R. and Orsenigo, L. (1996) Choice and Action, Journal of Evolutionary Economics, 6, 43–76. Lawson,T. (1985) Keynes, Prediction and Econometrics, in Lawson,T. and Pesaran, H. (eds), Keynes’ Economics: Methodological Issues, London, Croom Helm, pp. 116–133. —— (1988) Probability and Uncertainty in Economic Analysis, Journal of Post Keynesian Economics, 11, 38–65.

108

Paul Downward, John H. Finch and John Ramsay

Lawson, T. (1989a) Abstraction, Tendencies and Stylised Facts: A Realist Approach to Economic Analysis, Cambridge Journal of Economics, 13, 59–78. —— (1989b) Realism and Instrumentalism in the Development of Econometrics, Oxford Economic Papers, 41, 236–258. —— (1994) The Nature of Post Keynesianism and its Links to Other Traditions: A Realist Perspective, Journal of Post Keynesian Economics, 16, 503–538. —— (1995) The ‘Lucas Critique’: A Generalisation, Cambridge Journal of Economics, 19, 257–276. —— (1997) Economics and Reality, London, Routledge. Layder, D. (1998) Sociological Practice. Linking Theory and Social Research, London, Sage Publications. Lee, F.S. and Downward, P.M. (1999) Retesting Gardiner Means’ Evidence on Administered Prices, Journal of Economics Issues, 33, 861–886. Loasby, B.J. (1976) Choice, Complexity and Ignorance, Cambridge, Cambridge University Press. —— (1999) Knowledge, Institutions and Evolution in Economics, Routledge, London and New York. March, J.G., Sproull, L.S. and Tamuz, M. (1991) Learning From Samples of One or Fewer, Organization Science, 2, 1–13. Nelson, R.R. and Winter, S.G. (1982) An Evolutionary Theory of Economic Change, Cambridge, MA and London, Belknap Press of Harvard University Press. Peirce, C.S. (1934) Collected Papers of Charles Sanders Peirce,V, Pragmatism and Pramaticism, edited by Hartshorne, C. and Weiss, P., Cambridge MA, Harvard University Press. Potts, J. (2000) The New Evolutionary Microeconomics, Edward Elgar, Cheltenham. Pratt, A.C. (1994) Uneven Re-production. Industry, Space and Society, Oxford, Pergamon. —— (1995) Putting Critical Realism to Work: The Practical Implications for Geographical Research, Progress in Human Geography, 19, 61–74. Rizzello, S. (1999) The Economics of the Mind, Edward Elgar, Cheltenham. Rosser, J.B. Jr (1999) On the Complexities of Complex Economic Dynamics, Journal of Economic Perspectives, 13, 169–192. Runde, J.H. (1998) Assessing Causal Economic Explanations, Oxford Economic Papers, 50, 151–172. Setterfield, M.A. (1997) Should Economists Dispense With the Notion of Equilibrium? Journal of Post Keynesian Economics, 20, 47–76. Simon, H.A. (1962), The Architecture of Complexity, Proceedings of the American Philosophical Society, 106, 467–482. —— (1976a) Administrative Behavior. The Study of Decision-Making Processes in Administrative Organization, third edition, New York, Free Press. —— (1976b) From Substantive to Procedural Rationality, in Latsis, S. (ed.), Method and Appraisal in Economics, Cambridge, Cambridge University Press, pp. 129–148. —— (1978) Rationality as a Process and as a Product of Thought, American Economic Review, 68, 1–16. Tool, M. (1991) Contributions to an Institutional Theory of Price Determination, in Hodgson, G.M. and Screpanti, E. (eds), Rethinking Economics: Markets,Technology and Economic Evolution, Aldershot, Edward Elgar, pp. 19–39. Trigg, R. (1989) Reality at Risk, Hemel Hempstead, Harvester Wheatsheaf. Wieviorka, M. (1992) Case Studies. History or Sociology? in Ragin, C.C. and Becker, H.S. (eds), What is a Case? Exploring the Foundations of Social Inquiry, Cambridge, Cambridge University Press, pp. 159–172. Znaniecki, F. (1934) The Method of Sociology, New York, Rinehart.

Part IIB

On econometric methods

7

Critical realism and econometrics Interaction between philosophy and Post Keynesian practice1 Paul Downward and Andrew Mearman … a difference of opinion between practising Post Keynesian economists and critical realist methodologists should initiate a process of discussion that might well bring about modifications to Critical Realism as well as to practice. (Dow, 1999, p. 31) … it does not follow, as some have supposed, that Post Keynesians ought thereby not to engage at all in formalistic methods such as econometrics.The possibility of successes with the latter requires local closures … a special configuration of an open and structured system … So the opponent of Critical Realism is not the post Keynesian or whoever, seriously attempting to find out if (or demonstrate that) in certain conditions some closed-systems methods or whatever could contribute to enlightenment. Rather, the opponent is the advocate of any form of a priori dogma. (Lawson, 1999, pp. 7–8, emphasis added)

Introduction The motivation for this chapter resides directly in the above quotations and with the argument that econometric practice can play a role in analysis framed within a critical-realist perspective. This is a controversial theme, because despite the openness in the quotation from Lawson (1999) many of the epistemological pronouncements of critical realism implies suspicion of econometric methods.To address this issue, it is argued that a degree of tension exists between the philosophical pronouncements of critical realism and the consequences of referring to empirical concepts in the practice of inference, which is something that critical realists wish to do. Thus, while critical realism has epistemological implications for the practice of economics, there is an iterative relationship between philosophy and practice. The chapter is organised as follows. The next section briefly outlines critical realism by drawing upon the work of Tony Lawson – summarised in Lawson (1997).The aim is to restate clearly the basis of critical-realist opposition to mainstream economic analysis and econometric inference. To address these concerns, the section ‘The essence of econometrics’ carefully reviews

112

Paul Downward and Andrew Mearman

econometric methods. This review is important because Lawson (1997) highlights particular concerns with, and indeed re-defines, econometrics. On the basis of this review the section ‘A generalised critical-realist critique of econometrics’ then attempts to generalise the critical-realist critique of econometrics and shows that Lawson’s redefinition suffers from the same problems. Facing up to these issues, the section ‘Some proposals’ offers some proposals for empirical inference that shares the broad concerns of critical realism, yet embraces econometric work.

Critical realism: ontology and epistemology The objective of Lawson’s volume Economics and Reality (1997) is to provide both a critique of mainstream economics and an alternative economic methodology.The critique is ontological implying that the ‘deductivist model of explanation’ (p. 17) embraced by mainstream economics is incompatible with social material. According to critical realism, deductivism makes three ontological errors. First, it often conflates knowledge of the world with what exists. Second, it assumes a flat ontology of only experiences and the events that they represent. Third, being rooted in ‘a version of positivism that is rooted in Hume’s analysis of causality’ (Lawson, 1997, p. 19, emphasis in original), deductivism emphasises constant conjunctions of events or causallaw explanations of the form ‘whenever event x then event y’ – which, it should be noted, can be rendered stochastic; in this case, event y follows from event x ‘on average’ (Lawson, 1997, p. 69) – which presupposes a closed-system ontology. There are two main closure conditions that need to be invoked to defend this type of explanation. First, the intrinsic condition of closure (ICC), which can be loosely characterised as implying that a cause always produces the same effect, suggests that the structures of the phenomena under study are constant, unchanging and that for any intrinsic state, only one outcome is possible. Second, the extrinsic condition of closure (ECC), which can be loosely understood as implying that an effect always has the same cause, suggests that the phenomena under study are isolated from other potential influences. The mainstream emphasis on mathematical modelling can be allied to this approach to explanation, which also implies that prediction and explanation become formally equivalent. This is known as the ‘symmetry thesis’. Indeed, it is here that critical realists argue that econometrics finds its traditional role in economics, in testing the predictions/explanations associated with covarying magnitudes (see also Lawson 1989a,b). This leads into critical-realist criticisms of econometrics. These are discussed more fully in the section ‘A generalised critical-realist critique of econometrics’, but a few comments are pertinent here. The definition of, and main problems associated with, econometrics are discussed in chapter 7 of Economics and Reality. Here, Lawson argues,

Critical realism and econometrics 113 What, then, is econometrics? Broadly speaking econometricians concern themselves with attempting to determine constant event conjunctions … albeit of a probabilistic sort. Not infrequently these sought-after relations are interpreted as ‘causal’. However, this usage of the term is basically Humean. Specifically, it indicates that some ‘dependent variable’ y is held to be functionally related to a set of ‘independent variables’ x, so that movements or variations in the latter give rise to predictable variations in the former. (Lawson, 1997, p. 69) Drawing upon Lawson (1995), criticism is then directed at the estimation of parametric functions. The objections to parametric functions are made with reference to a reformulation of the Lucas critique.The original critique argued that optimising economic agents, with rational expectations, would accommodate policy changes into their behaviour. Consequently, the structural instability of the underlying equations describing behaviour would make forecasts based on the original equations erroneous, thus undermining attempts to fine-tune the economy. Lawson (1997) argues that this result is made possible not because of the restrictive rationality assumptions of mainstream analysis but because of the inapplicability of the closure conditions. In contrast, Lawson argues that the criticism may not apply to the ‘determination of summary statistics of a body of data’ (1997, p. 69). Indeed later in the book Lawson specifically redefines econometrics along these lines arguing that,

The measuring and recording of states of affairs, the collection, tabulation, transformation and graphing of statistics about the economy, all have an essential (if usually non-straightforward) role to play. So do detailed case studies, oral reporting, including interviews, biographies, and so on. Indeed, I suggest it is precisely to such indispensable activities that the heading of econometrics is properly attributed. (Lawson, 1997, p. 221)

In seeking to propose an alternative methodological approach, critical realism rejects the applicability of the closure conditions and the covering law model of explanation and, of course, econometrics as discussed above. Realism emphasises ontological depth. Therefore, reality is conceptualised as stratified into actual events, our empirical experience of them and most significantly their real causes, which are considered ‘deep’. Importantly, these domains are out of phase with, and irreducible to, one another because causes act transfactually: they can be operative irrespective of observed events. At the level of the ‘real’, causes can complexly co-determine outcomes at the level of ‘events’, which in turn need to be understood at the level of the ‘empirical’.

114

Paul Downward and Andrew Mearman

Lawson (1997) argues that here the epistemological goal of theorists should be to identify causal laws by discovering the mechanisms underlying, but not equivalent to, actual events. This, of course, will not be a straightforward matter. In experimental conditions, complex co-determining factors may be isolated, and observations calibrated and understood in a relatively robust manner. However, for various reasons, in the social world these opportunities do not exist. How, then is inference to proceed? How does one compensate for the impossibility of effective experimentation? Lawson (1997) develops his epistemology with reference to the concepts of ‘retroduction’,‘demi-regularities’,‘abstraction’ and ‘explanatory power’.Taking these points in turn, retroduction comprises the logic of inference for critical realism. Retroduction, … consists in the movement, on the basis of analogy and metaphor, amongst other things, from a conception of some phenomenon of interest to a conception of some totally different type of thing, mechanism, structure that, at least in part, is responsible for the given phenomenon. (Lawson, 1997, p. 24) Little specific guidance is offered by Lawson, who argues that this will depend on the context (Lawson, 1997, p. 212). This said, while constant conjunctions of events will not be observed, rougher empirical patterns called ‘demi-regularities’ can ‘direct social scientific investigations’ (Lawson, 1997, p. 207, emphasis in original), towards identifying causes. If these demi-regularities are contrastive, that is unexpected; they are particularly useful in highlighting the need for retroduction. Likewise, analysis will involve abstraction. Lawson (1997) stresses that abstraction should focus on real entities and momentarily neglect other features of reality. Such realist abstraction stands in contrast to instrumentalism, in which the mechanisms posited can be ideal or fictional, not subject to empirical check, and assessed only by their ability to yield successful predictions (Lawson, 1989a, p. 238).The key to assessing an abstraction then becomes their ability to ‘illuminate a wide range of empirical phenomena’ (Lawson, 1997, p. 213, emphasis in original). However, explanatory power, as the criterion of theory choice, is left relatively undeveloped.

The essence of econometrics In order to assess the critical-realist critique, this section carefully reviews econometric techniques and inferential emphases to try to distil their key properties.This is important because the development of econometric methods has been considerable. In this chapter discussion is presented in terms of four distinct ‘schools of thought’ in econometrics.These are the ‘average economic regression’ or textbook approach, the Hendry or London School of Economics

Critical realism and econometrics 115 approach, the Leamer/Bayesian approach and the Sims/atheoretical approach. More eclectic econometric techniques are then referred to. The average economic regression (AER) approach to econometrics ultimately has its roots in Haavelmo (1944) and has been popularised by, for example, intermediate textbooks on econometrics. A linear equation linking a stochastic dependent variable to a set of non-stochastic independent variables is put forward as a representation of an equilibrium relationship specified by theory.There is an implied presumption of correct specification (Gerrard, 1995). Of course, lags can be included to allow for adjustment towards equilibrium over time. To an extent this includes using stochastic variables as regressors. Typically Ordinary Least Squares (OLS) is employed to estimate partial slope coefficients that identify both the sign and magnitude of the influence of a particular independent variable upon the dependent variable. As it stands the regression model can be viewed as producing purely sample-specific descriptive insights because OLS can be understood as simply a partitioning of the dependent variable into explained and residual components according to the Analysis of Variance identity. This is an important point to note. There is a sharp distinction drawn between estimation and inference in the AER approach. Additional assumptions are employed to make inferences.Thus, it is assumed that the mean of the random disturbance that affects the dependent variable is zero and that the variance and covariance of the disturbances are constant and zero across observations respectively.This produces a ‘Best Linear Unbiased Estimator’.Thus, in this approach, most attention is paid to ensuring that the residuals of the regression, as estimates of the random influences, meet these assumptions. The normality of residuals is checked, or presumed with reference to the Central Limit Theorem. Further, the residuals are checked for serial correlation or heteroscedasticity. Rescaling the variables in terms of logarithms or using dummy variables to capture outliers typically accounts for problems with the former. In the latter cases transformed models are estimated. At this point the statistical significance of the regression, overall as well as particular coefficients, is tested. The equation can then be used to predict the average value of the dependent variable conditional on values of the independent variables either within the existing sample, or beyond the sample thus reflecting the symmetry thesis noted earlier. Degrees of statistical precision can then be attached to either of these knowledge claims. Haavelmo’s (1944) influence is clearly most pronounced in this latter set of issues.This said, Haavelmo does not necessarily argue that the probabilistic assumptions have any real counterpart. In this respect there is an instrumentalist motivation underlying the AER approach. Nonetheless, the aim of the approach is to uncover event-regularities described by the estimated equation. Despite its traditional emphasis, reinforced by its presence in intermediate level textbooks, concerns have arisen in applied econometrics. It has proven difficult to discriminate between theories.This is linked to the issue of identification in econometrics. Narrowly defined, identification is concerned with

116

Paul Downward and Andrew Mearman

recovering the structural parameters of a theory from a reduced-form equation.Technically speaking this relies upon the influence of exogenous variables. This is difficult to achieve (De Marchi and Gilbert, 1989). Historically, for example, Frisch described these problems as ‘multicollinearity’ because all variables were partly stochastic reflecting measurement error. Consequently, reordering the variables in the regression would produce entirely different estimated coefficients. Philosophically, Keynes (XIV), discussed later in the chapter, allied this to the non-homogeneity of social material over time. The AER approach sidesteps these fundamental issues. Typically, textbooks such as Gujarati (1999) associate multicollinearity with problems of data rather than the underlying method. Moreover, model selection tends to focus on establishing a correct functional form, and on not omitting important variables. In contrast, as Pagan (1987) and Downward (1999) note, three broad strategies have emerged in econometric discussion to address these fundamental problems. Challenging the duality of classically defined econometric inference and estimation, issues of theory estimation, discrimination and adequate statistical representation are addressed in a more fluid manner. To greater or lesser degrees, each approach draws upon the interplay of theory and data to postulate an adequate econometric model.2 The first two of these approaches have stemmed from work in ‘Time-Series’ econometrics.The first approach is associated with the work of David Hendry and the least square estimation (LSE) and is often referred to as the ‘General to Specific’ approach.The basic strategy involves using theory to specify relevant variables. These are then included in a general model including lags, which is then reparameterised to obtain nearly orthogonal explanatory variables.Tests of restrictions then proceed to reduce the model to a version that satisfies relevant diagnostic statistics.As well as serial correlation, for example, the emphasis is also upon parameter constancy. More recently, the approach embraces the need to test for structural breaks in data.Tests for the weak exogeneity of variables can also be conducted.The basic theme of the approach is demonstrated, with the error-correction models used in the cointegration analysis of time series. Theory describes the long-run relationship. The data reveal the short-run dynamics of adjustment. In contrast, Sim’s (1980) approach to econometrics rejects the possibility of exogeneity and identification and is confined to the analysis of reduced forms in a simultaneous or ‘vector autoregressive’ (VAR) context.Vector autoregressions consist of regressions of each variable of interest on lagged values of itself and the other variables under scrutiny. The largest lag structure possible is started with and then subsequently simplified in much the same manner as Hendry’s approach. It has become popular to test for Granger Causality in the VAR context. This implies seeing whether lagged information on particular variables is statistically associated with the other variables which are treated as dependent in that particular context.The objective of the Sims’ methodology

Critical realism and econometrics 117 is to provide structure-free, that is entirely reduced-form, conclusions. This approach has been criticised for being atheroetical (Darnell and Evans, 1990, p. 126). However, such criticisms are a bit naive and historical. It is clear that some theoretical priors will focus attention upon the selection of variables.The VAR approach simply emphasises the intent to avoid problems of identification and theory choice. Moreover, it is now seen as appropriate practice to reparameterise VAR systems to test for co-integration using maximumlikelihood estimation methods. This provides a direct link to Hendry’s approach. Finally, it should be noted that advances have been made in statistical testing in a time-series context generally. Standard t and F statistics, for example, are not valid when the data are non-stationary – that is the mean, variance and covariance of the probability distribution are time dependent. Alternative distributions which form the basis of tests for the stationarity of univariate series, and cointegration of multivariate series, form a central part of inferential claims. Leamer (1983) advocates a Bayesian approach to statistical analysis because the official rhetoric of econometrics appeals to the ‘false idol of objectivity’ ‘… unencumbered by the subjective opinions of the researcher’ (Leamer, 1983, p. 36). In contrast, Leamer argues that econometricians must accept the clear role that theoretical priors play in analysis and that, The misspecification matrix M is therefore a pure prior concept. One must decide independent of the data how good the non-experiment is. (Leamer, 1983, p. 33)3 Thus, despite operating with non-experimental data, priors are employed in econometrics in establishing the (conditional) sampling function and the marginal or prior probability density function. Leamer advocates the formulation of a general family of models, to decide what inferences are of importance, which need not be structural relations, and express these relations in terms of parameters. Prior distributions can then be formed which summarise any information not in the data set and the sensitivity of inferences to a particular choice of distributions can be analysed to explore their ‘fragility’. In this respect, subsets of variables, reflecting particular interests, may be focused upon while manipulating – exploring all possible linear combinations of – what are considered to be prior ‘doubtful’ variables. Extreme parameter values, of a most favourable and least favourable character, can then be found by experimentation for the parameters of interest. If the restrictions required to obtain a narrower range of values are considered dubious then the results should not be treated as robust. There are clear echoes here with Hendry’s approach. However, regarding Hendry’s recommendation that, if the data indicate that the complete set of variables is significant, inference concerning

118

Paul Downward and Andrew Mearman

point estimates should proceed from the general model, Leamer appears to be sceptical, as he believes that there are conflicting (prior) grounds for inference. Finally, as well as these relatively well-defined schools of thought, it is important to note that there has been much development in other types of econometric work. Typically, if not exclusively, these have occurred in the cross-sectional analysis of surveys and attempts to capture qualitative influences in regression models. Importantly, this suggests that econometricians have, to an extent, been directly concerned with ontological matters. Using dummy independent variables, for example, is now commonplace in econometric modelling. In addition considerable use is made of models that estimate relationships involving qualitative dependent variables. For example, LOGIT and PROBIT estimators are employed in cases in which the dependent variable is presumed to reflect simple dichotomous categories, or rank-orders. Poisson and Negative Binomial models are now estimated for models which maintain that dependent variables reflect discrete (that is comprise integers or counts) rather than continuous values. In these cases, in order to identify the marginal effects of independent variables on the dependent variable, typically maximum-likelihood methods of estimation are employed because the models are non-linear. Moreover, they appeal to the binomial distribution as a basis of statistical inference rather than the normal distribution that underpins much of the econometric inference in the techniques previously discussed. Finally semi-parametric models can be employed in this context, which do not impose a prior functional form on estimates. Moreover, the results are not interpreted through estimated parameters as such. In contrast, for example, in a continuous regression, estimates are reported for close values of particular conditioning variables. Exploring the relationships sequentially can then identify the shape of the regression. In the latter case, non-linear functional forms can be identified. Either non-linear regression techniques or log-linear transformations can be used to estimate the relationships. Other developments include allowing for truncated or censored distributions in the dependent variable.The most famous example of this is the TOBIT model. Under these circumstances, the assumption is usually that the dependent variable is drawn from a normal distribution but that there is a qualitative break in the measurement of the variable. Once again, maximumlikelihood estimators are typically employed to reduce what appears to be predictable bias (Greene, 1997, p. 996). Consequently, it is clear that the current state of applied econometrics is one of development and change in emphasis. This said, a number of common themes are revealed in the above analysis.The first is that there is an increased recognition of the inability of econometrics to discriminate between theories, though it remains an aspiration. Second, it is clear that prior data analysis is much more important in the approaches other than the average economic regression approach.This is either in terms of modelling through testing or in a priori consideration of the distribution of, say, the dependent variable.4 Third,

Critical realism and econometrics 119 it follows that the duality between estimation and inference is less apparent in these approaches. Fourth, the emphasis on event-regularities is reinforced because probabilistic factors naturally become much more prevalent in the generation of econometric results. Finally, predictions are still treated as the mirror image of explanations, and vice versa.

A generalised critical-realist critique of econometrics In the section on critical realism, it was argued that Lawson (1997) identifies the main problems with ‘conventionally defined’ econometrics – expressed in terms of parametric functions – as stemming from the lack of applicability of the closure conditions noted earlier. Moreover, it is suggested that more qualitative research methods, or those relying on descriptive empirical methods would be more appropriate. It is clear, therefore, that these claims need further investigation in the light of the above discussion of econometric methods. The conditions of closure are now further examined in connection with the ‘twin’ branches of statistical inference referred to in the discussion of econometric methods. These are statistical estimation, and statistical inference.

Empirical measurement We begin with a discussion of the legitimacy of empirical measurement, which is a precondition for statistical estimation that, of course, embraces the calculation of descriptive statistics. From a critical-realist perspective one can identify two related concerns with empirical measurement.The first is the relevance of the necessary conditions for quantification. The basic conditions required for performing any mathematical (and statistical) operation are quite clearly expressed in Allen (1962). Objects must belong to the same set, which requires that they exhibit formal ‘equivalence’; that is, reflexivity, symmetry and transitivity.5 Formal equivalence implies that objects are qualitatively invariant and can have number systems, such as nominal, ordinal, interval and cardinal ranks applied to them. The basic point is that ‘[w]e can measure [things] at different times or places in different conditions and know that we are not measuring different things’ (Sayer, 1992, p. 177). Therefore, a basic element of critical-realist concern with econometrics must lie with the presumption that social objects are not like this. Indeed Sayer (1992) warns that one should recognise the concept-dependent nature of social objects. This is not a criticism unique to critical realism. As noted earlier Keynes (XIV) stated these concerns, warning that there are problems of inappropriately defining econometric proxies for specific economic concepts.The second concern associated with empirical measurement is that, related to the above point, quantitative analysis ultimately lends itself to an atomistic (Humean)

120

Paul Downward and Andrew Mearman

ontology. Once again Keynes (XIV, p. 286) has expressed these concerns. It is interesting to note, thus, that critical realism has forcibly restated existing criticisms.This is a matter further discussed in the next section. Statistical inference While the above discussion is confined to the process of measurement and calculation per se – and hence would be relevant, say, to the nature of regression coefficients, it naturally leads onto the concern with the ontological nature of statistical inference implied in econometric approaches. It is clear that the methods emphasise a measurable probability distribution to underpin all of the inferences made. There are differences of emphasis. As noted earlier, one can associate the AER approach with the instrumental idea that the distribution does not describe real properties of the economic system. In contrast, for Hendry the notion of a data-generating process appears to be a literal possibility (Hendry et al., 1990) though he also argues that the ‘proof of the pudding … is in the eating’ (Hendry, 1995a, p. 18). An element of instrumentalism thus remains. The integration of Hendry’s approach with VAR methods, or their pragmatic emphasis is also suggestive of instrumentalism in time-series analysis generally. In the case of the Bayesian methods, probabilities are presented ostensibly as subjective, yet applied methods draw upon measurable probabilities and there is an implied presumption, for example implied by ‘law of large numbers’, that in the aggregate, subjective perceptions eventually collapse onto the actual world. At best such statistical ‘laws’ are abstract mechanical forms of aggregation that are imposed without any ontological justification. In this respect, the basis of the emergence of the ‘higher order’ regularities is not given context. How the subjectivity of beliefs becomes reflected in objective material is simply not addressed. At worst they reflect an instrumentalist assumption. In the case of limited dependent variables, it is clear that there is an explicit sense that the dependent variable is modelled with reference to a distribution that captures real properties of its behaviour. This is either through the categorical nature of the variable implying a binomial distribution, or a censored distribution that is linked to the inability to observe a subset of actual values of the dependent variables. Finally, non-parametric methods still appeal ontologically to a measurable probability distribution that is independent and identical drawings from a random variable, but do not specify the precise form of function. There is thus some scope for debate over the ontological character of the probability function. It remains, however, that whether expressed as subjective or objective, or parametrically or not, probabilities and hence knowledge claims, or the subjects of knowledge are conflated with the objects of knowledge though the presentation of a covering law or statement about constant conjunctions of events.The econometric methods restate a subject/object conflation and hence this explains their latent or explicit instrumentalism. Instrumentalism implies

Critical realism and econometrics 121 that ontological and epistemological claims are formally synonymous, which is an epistemic fallacy for critical realists because of, … the supposition that statements about being can always be rephrased as statements about knowledge (of being), that ontology can be reduced to epistemology. With ontology unavoidable, this error functions only to cover the generation of an implicit ontology, in which the real is collapsed onto the actual which is then … identified with … human experience, measurement or some other human attribute. (Lawson, 1997, p. 62, parentheses in original) The econometric approaches noted above thus could be said to redefine but not avoid this fallacy. The generally problematic logic of econometric inference can now be more forcefully restated in critical-realist terms. It is argued that both the estimation of regression coefficients and the drawing of statistical inferences require the invocation of the closure conditions noted earlier. The ICC is equivalent to assuming the underlying homogeneity of nature (Lawson, 1989, pp. 240–1) and the atomistic combination of objects, for example, in combining data points and presenting statistics through statistical functions.The ECC implies that all of the causal factors have been included in an econometric study, or that the effect of external factors on internal factors is constant. This last point is equivalent to assuming that countervailing factors are constant or predictable. The probability distribution performs this task in econometrics. In the literature thus, econometrics, as so defined, remains tied to the analysis of, and search for, event-regularities and is, for critical realists, inextricably linked to instrumentalism. A dilemma The discussion above can however be generalised beyond the specific econometric approaches discussed. Drawing upon the critical-realist critique, logically speaking, any empirical work will imply making closure assumptions. Descriptive empirical references will assume that material is qualitatively invariant – the invocation of the ICC. Commentary on any co-variation of empirical categories that follows, moreover, implies that they are related in an atomistic way. Finally, any form of statistical inference will entail the ECC. Now, an important consequence of this argument is that this criticism also applies to Lawson’s (1997) redefinition of econometrics.The implication is that admitting the possibility of statistical analysis of one sort means that it is not straightforward to rule out any particular method of analysis.Thus, critical realism’s fundamental critique of econometrics might imply that their proposed alternative methods are equally subject to that critique. It follows that if empirical reasoning is to be at all employed by critical realists, the practice must, in some sense, compromise the philosophical

122

Paul Downward and Andrew Mearman

underpinnings of realism. Put more positively, and in a manner consistent with the critical-realist rhetoric, the dual between instrumentalism and realism needs to be transcended and a finer distinction developed between the uses of empirical methods.The solution, we would argue, lies in recognising the reality of Dow’s (1999) conjecture cited at the beginning of this chapter. In practice one can attempt to minimise the well-thought out problems of critical realism, but in so doing practice must inform philosophy.

Some proposals The task of this section is to establish legitimate reasons for using econometrics whilst remaining sensitive to critical-realist criticisms of econometrics. It is argued that these can be identified and, they reflect the breaking down of the duality of ‘estimation’ and ‘inference’ as implied in modern econometric approaches. However, an alternative ontological perspective – one informed by critical realism – implies a different logic of inference. A useful starting point, as alluded to earlier, is that many of the arguments raised by critical realism against econometrics have been aired elsewhere historically. It should be noted that Keynes, in particular, was well aware of the conflict between philosophical considerations and practical needs. It is also worth noting that, along with Keynes, many early econometricians or applied economists were aware of the problems of applied econometrics (Lawson, 1989). It follows that lessons can be learned from deliberating upon how these issues have been approached in the past. In as much that critical realism has become a framework within which these criticisms have been restated, it follows that it can provide a framework within which potential solutions might be discussed. Retroduction as triangulation As implied earlier, many research techniques differ more in emphasis than in kind. In this respect it should be noted that both qualitative and quantitative work invokes closure (Downward, 1999, 2000; Downward et al., 2002 and Chapter 6 of this volume). For example, for two reasons, even an appeal to qualitative evidence implies the assumption of intrinsic closure. First, in the limit, all empirical discourse relies on some means of collecting, collating and ordering data.‘Collecting’ together insights to produce a stylised interpretation is identical in this respect. Second, an appeal to conceptual descriptions such as ‘administered pricing’ in a generalised manner implies the invariance of the essential qualities of the pricing process. Thus, Lawson’s attempt to redefine econometrics is problematic. More generally we suggest that the conventional dualistic distinction between quantitative and qualitative work, for example offered by the criticalrealist Sayer (1992, p. 114), is not particularly helpful and, indeed, is misleading. In contrast, what

Critical realism and econometrics 123 is presented is a difference in emphasis. The overall shift of emphasis between ‘quantitative’ and ‘qualitative’ work hinges on the fact that qualitative research techniques are concerned with interpretation rather than simply describing patterns of numbers per se. Nonetheless,‘empirical regularities’ are sought after albeit in the form of interpretations of original data and observations. This is not to say that quantitative and qualitative approaches are formally equivalent. In an interpretative context, as opposed to a data-descriptive quantitative context, one can argue that the real would be revealed by a conceptual refocus on process and action, rather than events and outcomes. This may, for example, involve a literary shift from nouns referring to states, to verbs describing processes. It remains therefore that ‘qualitative’ work by definition has more flexibility, for example, in identifying the sources and reasons for changes in structures. However, in recognising these different features of empirical research implies that ‘estimation’ and ‘inference’ would necessarily intertwine in explanations that draw upon a variety of methods. These arguments also suggest how differences in the use of econometric techniques may be understood in a more subtle way than currently dominates the literature. It is clear, for example, that regression analysis used as a samplespecific description differs from (and is technically inferior to) that of, say, a cointegration analysis motivated by Hendry’s approach to econometrics.While both approaches make the same assumptions about additivity, the constancy of data points, etc., the inferential emphasis is different. Similarly, the estimated equation could be generated with reference to idealised concepts such as the outcome of perfectly informed agent behaviour, or appeal to real aspects of behaviour such as a habit, rule or institutional process.6 While this argument is different from that, say, espoused by Lawson (1997), it illustrates that theory evaluation will require more than the rhetoric of falsification, instrumentalism and, in general, appeal to one type of data. Ontological considerations must play a role in theory choice and inference. However, it follows that researchers interested in applied work and eager to participate in policy discussion should not ignore econometric analysis but that researchers should design a research approach that exposes the sensitivity of their insights other than by appeal to probabilistic criteria.This conclusion differs significantly from both the conventional approach to econometrics and the critical-realist criticisms of econometrics. Keynes (VIII; XIV, pp. 285–320) is instructive in this regard as he articulates the need for triangulating insights. As noted earlier, it is without doubt that Keynes argued that quantification of probabilities required reference to atomic entities. Moreover, Keynes’s (XIV, pp. 306–20) criticism of Tinbergen’s econometric analysis can be cited as a precursor to critical-realist concerns.Yet, more generally Keynes emphasised the importance of rational belief rather than knowledge as a basis of argument (Keynes,VIII, p. 10). Rational belief resides in logical justification. There is a relative/absolute dimension to this. Probabilities, which are not necessarily numerically defined, are, on the one

124

Paul Downward and Andrew Mearman

hand relative to given evidence. However once given a body of evidence or initial proposition, probabilities concerning subsequent propositions are absolute or objective. Crucially for Keynes, relevant evidence is ascertained through a process of negative analogy. To avoid the problem of induction, Keynes argued that one should examine a particular phenomenon in different contexts. If a phenomenon appears to be common across various contexts then this indicates its relevance. In turn this relevance adds weight to a particular account of that phenomenon. If the different contexts reveal non-common elements, then the weight of an argument will decrease, revealing our ignorance. Thus, from an operational perspective, once again this suggests that various empirical insights should be triangulated. The added contribution of critical realism is to suggest strongly that this should have an ontological aspiration. Consequently, while descriptive, institutional and historical analysis might themselves be suggestive of the causal mechanisms, the effects of their action can be assessed, and hence the purported causal mechanism supported, with reference to more quantitative analysis aimed at codifying the empirical level in terms of, say, demi-regularities.Thus, retroduction inherently involves triangulation. It follows from this discussion that there is no presumption that the econometric methods discussed above are the only option. Moreover, in many respects the regression methods discussed remain closely wedded to the idea of finely decomposing the variation in the dependent variable into ‘individual’ components or variables. There is a clear focus on identifying the influence of particular variables upon one another through an emphasis on the t-statistics of the significance of particular coefficients. This, often implicit, aspiration is clearly tied to the ceteris paribus method of comparative statics, in which constantconjunctions of events are often couched through a desire to imitate experimental manipulation of controls.There is no need for this once one embraces a critical-realist perspective. One can recognise that ‘multi-collinearity’ as originally conceived, and discussed earlier, is indicative of the problems of decomposing complex social matter into sub-components.Yet one can employ F-tests of joint (linear) restrictions as a means of tentatively identifying these sub-components. Moreover, multi-level regression analysis can now be employed to analyse the relationships between variables that are measured at different levels of aggregation. Once again, this is suggestive of how statistical methods can be employed to structure an understanding of a complex system. More radically, in recognising the problem of multi-collinearity, Frisch (1948) proposed that the set of possible econometric relationships should be explored. Termed ‘confluence analysis’, Frisch argued that a data-descriptive approach be conducted by charting, in the form of ‘Bunch Maps’, the behaviour of parameters when minimising the errors in each variable sequentially. Moreover, Frisch argued that other, less parametric, methods for analysing economic data, could be employed as alternatives to the bunch-map approach. Examples of these were characteristic root analysis, principle components analysis and cluster analysis. All of these eschew the use of probability theory.Their shared idea

Critical realism and econometrics 125 is to explore the correlation matrix of a set of data more ‘experimentally’ and more subjectively, and thereby to classify correlations of groups of variables. One can add to this list canonical correlation and (multivariate) analysis of variance in which sets of variables can be correlated rather than a single ‘dependent’ variable decomposed. It is clear that aspects of this outlook are referred to in various econometric approaches discussed before. Canonical correlation thus underpins the tests for cointegration in VAR systems. Moreover, the methods all retain an essentially additive structure, which is a compromise with the organicist ontology of critical realism. Yet, it seems clear that their qualified and pragmatic use could be legitimate in aiding understanding of complex phenomena and helping to assess the relevance of causal insights. Intuitively, moreover, the procedures other than multiple regression analysis would be more likely to reveal contrastive demi-regularities, for example, as clusters of co-variations, than single equation estimates. It remains, however, that the latter might be legitimately sought in the analysis of particular purported causal mechanisms. Moreover, while statistical tests clearly cannot drive critical-realist empirical analysis, they need not be ignored.Their legitimacy, as with other insights, resides in the triangulation. Embracing these ideas, it is argued, is a significant addition to the criticalrealist approach. Econometric predictions It is clear from the above discussion that ‘econometrics’ as conventionally defined can perform a useful function in critical-realist research. Most obviously it could be used to codify the empirical level (as demi-regularities) and to both facilitate and help to legitimise retroduction. It is likely, however, that other empirical insights would also be of use.This issue aside, the question remains concerning the role of econometrics as a vehicle for offering predictions. This is, of course, the traditional instrumentalist role ascribed to econometric methods.The discussion in this chapter suggests that empirical insights presented as predictions can be of use to the researcher. But, quite clearly, any quantitative prediction becomes merely a scenario whose legitimacy will rest upon the robustness of the qualitative invariance invoked in causal mechanisms and, of course, the lack of influence of countervailing causes.The discovery and robustness of such causal claims will, along with their implications, of necessity, always be open to revision. Likewise this discussion suggests how, for example, simply addressing ‘regime shifts’, or ‘structural breaks’ as in, say, Hendry’s econometric approach, does not fully address critical-realist concerns. In this sense, along with the quantitative–qualitative and estimation–inference duals, a modified perspective on critical realism removes the ‘explanation’– ‘prediction’ dual that has characterised econometric discussion.7 More importantly it clearly demonstrates the important difference that critical realism makes to economic discourse. However, it remains that empirical methods can contribute

126

Paul Downward and Andrew Mearman

to the critical-realist programme.This is in terms of articulating the problems of adequately capturing real elements of economic processes, and working towards concrete analysis and policy prescription.

Conclusion This chapter engages the debate on the use of quantitative methods in economics. Specifically it addresses the critical-realist critique of econometrics as representing closed-system methods and deductivism. The chapter examines a range of approaches to be found in contemporary econometrics and argues that there is, in terms of critical-realist criticisms, a degree of similarity between them. However, significantly, the chapter argues that critical-realist objections can be applied to all empirical work. Consequently, the critical-realist attempt to redefine econometrics as descriptive statistics is unsuccessful, because those methods are equally susceptible to critical-realist criticisms. Therefore, the chapter argues that a degree of tension exists between the philosophical pronouncements of critical realism and the consequences of referring to empirical concepts in the practice of inference, which is something that critical realists wish to do. Hence, the chapter offers proposals for empirical work consistent with critical realism. Specifically, it is argued that econometrics can be used to codify the empirical level, which can assist in making inferences about causal mechanisms. However, it is also argued that via a strategy of triangulation of empirical methods, critical-realist retroductions to causal mechanisms can be made. Indeed, it is argued that retroduction inherently involves triangulation. Support for the notion of triangulation is found in the work of both Keynes and Frisch.

Notes 1 This chapter draws upon and extends Downward and Mearman (2002).Versions of this chapter have also benefited from comments from John Finch, Fred Lee and Andrew Brown.Very helpful feedback was also received at the Cambridge Realist Workshop Conference 2000, May 5–7, Sidgwick Site, University of Cambridge. 2 It should be noted that these approaches are not particularly mutually exclusive.Thus Hendry’s ‘general to specific’ modelling strategy is employed in Vector-autoregression and cointegration analysis. Likewise, Bayesian ideas have been employed in these contexts. 3 The distinction between experimental and non-experimental data is not crucial to Leamer’s arguments. Experiments can also involve bias, through design or the failure to produce closure. 4 This is perhaps most clear in cointegration analysis which has become wedded to Hendry’s approach. Here much pre-analysis of the statistical characterisitics of the data precedes any attempt to test economic relationships. 5 Naturally, these assumptions are invoked in neoclassical economics; for example, to ensure that the conditions required of preferences in consumer theory are logically robust. Likewise, the most basic of descriptive and inferential statistics (e.g. frequency counts and non-parametric tests respectively) require this assumption.

Critical realism and econometrics 127 6 A related argument can be constructed from Finch and McMaster’s argument, in chapter 8, that non-parametric methods of analysis may be useful, from a criticalrealist perspective, as a means of checking the robustness of categories and interpretation. Clearly, work of this nature might, thus, be a bridging point for further statistical analysis of the kind advocated in this paper as part of a triangulation strategy of inference. 7 However, this is not to say that critical realism holds to a ‘symmetry’ thesis: clearly it cannot in open systems.

References Allen, R. (1962). Basic Mathematics, London, Macmillan. Darnell,A.C. and Evans, J.L. (1990). The Limits of Econometrics,Aldershot, Edward Elgar. DeMarchi, N. and Gilbert, C. (eds) (1989). Oxford Economic Papers, 41, special edition on Econometrics. Dow, S. (1999). ‘Post Keynesianism and Critical Realism: What is the Connection?’, Journal of Post Keynesian Economics, Fall 1999, 22, 15–33. Downward, P. (1999). Pricing Theory in Post Keynesian Economics: A Realist Approach, Cheltenham, Elgar. —— (2000). ‘A Realist Appraisal of Post Keynesian Pricing Theory’, Cambridge Journal of Economics, 24(2), 211–24. Downward, P.M., Finch, J. and Ramsay, J. (2002). Critical Realism, empirical methods and inference: a critical discussion, Cambridge Journal of Economics, 26(4), 481–500. —— and Mearman, A. (2002). Critical Realism and Econometrics: Constructive Dialogue with Post Keynesian Economics, Metroeconomica, 53(4), 391–415. Frisch, R. (1948). ‘Repercussion Studies at Oslo’, American Economic Review, 39, 367–72. Gerrard, B. (1995). ‘The Scientific Basis of Economics: A Review of the Methodological Debates in Economics and Econometrics’, Scottish Journal of Political Economy, 2(2), 221–35. Greene,W. (1997). Econometric Analysis, Upper Saddle River, N.J. Prentice Hall. Gujarati (1999). Essentials of Economics, London, McGraw Hill. Haavelmo, T. (1944). ‘The Probability Approach in Econometrics’, Econometrica (Supplement), 1–118. Hendry, D. (1995). Dynamic Econometrics, Oxford, Oxford University Press. ——, Muellbauer, J. and Murphy, T. (1990). ‘The Econometrics of DHSY’, in J. Hey and D.Winch (eds), A Century of Economics, Oxford, Blackwell, 298–334. Keynes, J.M. (1973). The Collected Writings of John Maynard Keynes, London, Royal Economic Society,Vols VIII, XIV. Lawson,T. (1989a). ‘Abstractions,Tendencies and Stylised Facts: A Realist Approach to Economic Analysis’, Cambridge Journal of Economics, 13, 59–78. —— (1989b). ‘Realism and Instrumentalism in the Development of Econometrics’, Oxford Economic Papers, 41, 236–58. —— (1995). ‘The “Lucas Critique”: A Generalisation’, Cambridge Journal of Economics, 19, 257–76. —— (1997). Economics and Reality, London, Routledge. —— (1999). ‘Connections and Distinctions: Post Keynesianism and Critical Realism’, Journal of Post Keynesian Economics, 22(1), 3–14.

128

Paul Downward and Andrew Mearman

Leamer, E. (1983).‘Lets Take the Con out of Econometrics’, American Economic Review, 73(1), 31–44. Pagan, A.R. (1987). ‘Three Econometric Methodologies: A Critical Appraisal’, Journal of Economic Surveys, 1, 3–24. Sayer, A. (1992). Method in Social Science: A Realist Approach, London, Routledge. Sims, C.A. (1980). ‘Macroeconomics and Reality’, Econometrica, 48, 1–47.

8

A pragmatic alliance between critical realism and simple non-parametric statistical techniques1 John H. Finch and Robert McMaster

Introduction This chapter addresses two related controversies concerning economists’ desire to undertake theoretical development, and to assess the practical adequacy or reliability of applications of theoretical developments. One controversy has been addressed mainly by critical realists, but also within Post Keynesian, institutional and evolutionary research, and concerns the ontological status of both formal modelling and empirical research in systems that are recognised to be essentially open.2 The second controversy is connected with the epistemological issue of establishing in advance, conditions under which knowledge claims may be accepted as reliable. It concerns formalism in economics, both in formulating knowledge claims and in translating such claims into statements that may be assessed by drawing upon principles and techniques of econometrics (Ziman, 1978, pp. 22–6; Dow, 1990, p. 144, 1998, p. 1826). This chapter contends that researchers in economics who wish to develop and assess the reliability and adequacy of knowledge claims may draw upon simple non-parametric statistical analyses such that theoretical development may be informed by empirical analysis and vice versa.With respect to the first controversy, a pragmatic alliance is proposed between non-parametric techniques and, essentially, an open-system ontology, in interpreting and analysing stylised facts or demi-regularities across classes or categories of events.This may be compared with the retroductive strategy, which critical realists have adapted from the pragmatism of Peirce, and which concentrates on causal explanations of events (Peirce, 1935; Sayer, 1992; Lawson, 1997; Runde, 1998).With respect to the second controversy, simple non-parametric techniques offer the prospect of a loose, incremental and iterative coupling in undertaking theoretical development and also assessing the practical adequacy and reliability of knowledge claims drawn from such development. While formalism offers advantages of precision (Backhouse, 1998, p. 1855; Krugman, 1998), there is also scope for theory, expressed formally, to evade empirical analysis, and for econometric techniques to be so demanding in terms of data required, or compositional principles to be satisfied, that again theoretical and empirical research diverge. Lakatos and Feyerabend argue against too close an identification of theoretical development and empirical

130

John H. Finch and Robert McMaster

tests, mediated through – another formalism – ‘naïve falsification’ (Lakatos, 1970, p. 116), in order to preserve scientific imagination through principles of ‘tenacity and proliferation’ (Feyerabend, 1970, pp. 203–5). Similar sentiments are expressed by Peirce (1935, pp. 144–6), who includes abduction, related to metaphor and hypothesis formation, as an important part of articulating and assessing the adequacy and reliability of knowledge claims within an overall retroductive strategy. This chapter sets out an incremental and developmental strategy in which simple non-parametric statistical techniques of analysis can promote theoretical development alongside assessing the adequacy and reliability of knowledge claims drawn from theoretical development. Following Siegel and Castellan (1988, pp. 3–4), where parametric inferences require assumptions to be made about the nature of the population from which samples are drawn, non-parametric or distribution-free inferences require fewer and less stringent assumptions about the nature of populations. While both parametric and non-parametric approaches both assume, fundamentally, that observations conform to the status of independent and identical drawings from populations, researchers who adopt non-parametric analyses can tolerate a great deal more ambiguity in their interpretations of their observations. For example, simple non-parametric techniques feature ranking or ordering tests, where parametric tests should be adopted if researchers are confident that they can organise their observations on to numeric or ratio scales. Our focus is hence on the decisions that researchers make at early stages of their projects, in interpreting their observations. Such choices affect the nature of inferences that can be made. Non-parametric econometrics – which is distinguished in this chapter from simple non-parametric techniques – has tended to neglect this aspect of empirical analysis, concentrating instead on developing distribution-free techniques for analysing data that has already been transformed onto numerical structures such as interval or ratio scales (Härdle, 1995; Pagan and Ullah, 1999). Härdle argues that ‘the term nonparametric thus refers to the flexible functional form of the regression curve. There are other notions of “nonparametric statistics” which refer mostly to distribution-free methods’ (Härdle, 1995, p. 5, original emphasis). The argument proceeds as follows. ‘Econometrics mainly-as-regression’ is reviewed in the following section: as tests of knowledge claims, and as techniques that address some aspects of essential system openness, such as agent choice.The third section provides a detailed investigation of the foregoing and their consequences for empirical research, concentrating on developing ‘econometrics as measurement’, on the inherent interpretivism in category formation, and on the pervasiveness of demi-regularities or stylised facts. It is mooted that the deployment of simple non-parametric techniques, as part of a pragmatic alliance with critical realism, may ascertain and indicate demiregularities by highlighting measures of association between population samples. Importantly, there is no indication of causality in such measures of association. The penultimate section addresses how non-parametric techniques may be

Critical realism and statistical techniques 131 drawn upon in satisfying the dual ambitions of theoretical development and assessing the adequacy and reliability of knowledge claims drawn from theoretical development.

Econometrics mainly-as-regression This section discusses how econometrics mainly-as-regression relates to the ambitions of accommodating theoretical development and assessing the adequacy and reliability of knowledge claims drawn from theoretical development. Econometrics mainly-as-regression is distinguished from econometrics as measurement (Lawson, 1997).3 The former is undertaken with the intention of verifying law-like statements of observable event regularities exhibiting stability and persistence. Econometrics as measurement includes: Measuring and recording of states of affairs, the collection and tabulation, transformation and graphing of statistics about the economy, all have an essential (if usually non-straightforward) role to play. So do detailed case studies, oral reporting, including interviews and so on. (Lawson, 1997, p. 221) The questionable ability of ‘traditional’ (i.e. Classical and Bayesian regression) techniques to satisfy their objective of verifying law-like statements in the absence of meaningful experiments in social sciences has germinated the development of sophisticated regression techniques in the context of choice experiments. These range from established logit and probit techniques to the more recent maximum entropy approach.These address some of the criticisms made by critical realists, such as Lawson (1997) and Manicas (1998), as well as the concerns of econometricians, such as Hendry (1980) and Leamer (1983), as discussed in Chapter 7, and are brought together in this section as a development of econometrics mainly-as-regression. ‘Traditional’ regression The distinction between explanations of what economists do, and what they really do, has been discussed widely in the methodology literature (McCloskey, 1986, 1994; Backhouse, 1994, 1997). Formally, econometrics may pursue verification or falsification of knowledge claims that are articulated as law-like statements (Caldwell, 1994, p. 50). However, Granger (2001) claims that such a pursuit of ‘truth’ is very different from the fundamental aims of statisticians. This group is mainly interested in understanding the data generating process that produces the available data, and how this may be enhanced. Far from generating stable law-like statements, statisticians consider the economy and most of the ‘observable world’ to be stochastic (Granger, 2001, p. 9). There is though some discontent with econometrics so defined among its leading practitioners, to the extent that econometrics mainly-as-regression is

132 John H. Finch and Robert McMaster embracing theoretical development – the articulating of knowledge claims – as well as empirical testing (Hendry, 1980; Leamer, 1983). The discontent concerns the performance of econometrics in its role as both predictor, and as a set of procedures for resolving theoretical disputes (Keuzenkamp, 1995;Wible and Sedgley, 1999). Hendry is concerned that econometric models generate spurious results: Simply writing down an ‘economic theory’, manipulating it to a ‘condensed form’ … and ‘calibrating’ the resulting parameters using a pseudosophisticated estimator based on poor data which the model does not adequately describe constitutes a recipe for disaster, not for simulating gold. Its only link with alchemy is self-deception. (Hendry, 1980, p. 401) Yet, Hendry remains confident in the ability of econometrics to address the problems of producing spurious results through greater sophistication in technique, and greater care in data generation.4 Leamer (1983, p. 43) also expresses concerns with the results generated by econometric modelling. However, his pre-occupation is with the credibility of econometric inferences, claiming that they are ‘whimsy’ and ‘fragile’. He calls for greater ‘sensitivity analysis’ concerning the assumptions and inferences made by econometricians.This consists principally of analysts recording how an inference changes as variables are added or subtracted from a model. Following Hendry, Golan et al. (1996) consider that many of the problems of econometrics mainly-as-regression stems from the poverty of data employed – ‘information we do not possess’ – in standard models.As with most statisticians (see Granger, 2001), they argue that data used to reflect economic processes and institutions are generated by relations that are frequently stochastic, dynamic, non-stationary and simultaneous. Indeed, such a state is likely to be the norm rather than the exception. Moreover, the nature of the difficulties confronting econometricians is claimed to arise from an ‘inverse problem’ that outcomes are frequently observable and measurable, but causes may not be amenable to observation, measurement or parameterisation.5 For instance, income and consumption are frequently observable, but the marginal propensity to consume is not. Prima facie, Golan et al.’s (1996) argument offers a promising avenue for discourse between proponents of econometrics and critical realism, as there may be some consistency with the stratified ontology of critical realism. This structured ontology of critical realism is invoked in its criticism of econometrics mainlyas-regression in its formal or official guise, which identifies analysis only in terms of sense data drawn from the domain of the empirical.6 The perceived weakness of econometrics mainly-as-regression is in its epistemic role in guiding applied economists to undertake research that has produced results, which, when compared, are ambiguous. Such ambiguity has had the consequence of allowing formal theory to become more prolific, tenacious and resilient in the

Critical realism and statistical techniques 133 face of the applied research that should have the capacity to shape and discipline the development of formalism (Chick, 1998).7 The practical consequences of the ontology set out by critical realists for research that aims to develop further knowledge claims within economics are cast in terms of open as opposed to closed systems theorising (Bhaskar, 1978, 1979; Lawson, 1997). The inherent openness of social systems resides in the creative capabilities of agents to behave and interact with varying degrees of consciousness and critical awareness in any manner of ways, leading to the reproduction or transformation of social practices. Experimental design brings to bear conditions of extrinsic and intrinsic closure, such that the object of interest may be isolated in a controlled manner and its constituents retain their intrinsic properties over the experiment’s duration. These closure conditions are difficult to replicate in economics, and provide two useful reference points for assessing experimental economics.The activity of theorising is itself a mental experiment and a type of abstraction in the sense of neglecting some aspects temporarily in favour of others (Lawson, 1997, p. 227; Setterfield, 1997, pp. 69–71).8 A third condition or comparator drawn from experimental design and employed here as a reference point for econometrics mainly-as-regression concerns the principle of composition used in collating and manipulating information, and then making inferences (Lawson, 1997, pp. 80–4; Manicas, 1998, pp. 334–6). Multi-variate regression presumes that at least some of the observations considered relevant can be converted into data that can be measured by being mapped onto an arithmetic numerical structure. From this, empirical studies frequently define and then aggregate individual units of analysis, such as households, firms or employees. Formal theorising may avoid the issue of composition by modelling instead representative or ideal types as if in isolation. It is, of course, well documented that critical realists baulk at such a procedure. Categorical data within econometrics mainly-as-regression There has been a steady development of econometric techniques that have sought to bring categorical variables within regression, and have introduced less onerous conditions on the nature of distributions of variables (or principles of composition) in undertaking regression.These developments introduce different definitions of parametric and non-parametric statistics, yet also presume rather than investigate conditions of system closure. Logit and probit approaches allow regression-like analyses of data that include variables which are measured in ordinal or ranked scales, or nominal or classificatory scales, arranged into dependent (or sometimes response) and explanatory variables, and can be represented in cross tabulations (Gujarati, 1988, pp. 480–500).A simple form is where a dependent variable includes two possible categories, such as employed or unemployed, or buys and does not buy.The transformation of such a model is to convert the dependent variable into a linear probability function, although this has inefficient statistical properties (ibid., pp. 469–79).A logit model transforms the dependent variable into

134

John H. Finch and Robert McMaster

a log of odds ratio, and the probit model transforms the dependent variable using the cumulative normal distribution function. Both have in common a transformation of a categorical dependent variable into a continuous variable by means of aggregating observations into probability distributions, assuming some underlying model. The transformations allow models to be asymptotically (‘statistically’) efficient, although this may involve maximum likelihood estimates rather than ordinary or weighted least squares estimates. These may be illustrated by examining the use of choice experiments in market research, transport economics and health economics.The common reference for these models is McFadden (1974), who is credited with introducing an econometric model of random utility that can be applied to classical utility theory. Hence, efficient statistical properties coincide with the established axioms of economic theory. Putting to one side issues of intrinsic and extrinsic closure, the random utility models upon which discrete choice experiments are based present categorical dependent variables, but provide the theoretical grounds for converting these into statistically tractable variables articulated as numeric series on interval or ratio scales: An econometric model of qualitative choice behavior can be constructed for a specified model of individual behavior by assuming to be a member of a parametric family of probability distributions and using the fact that the observed choices are multinomially distributed with the probabilities … to obtain estimates of underlying parameters. (McFadden (1974), pp. 107–8) Following McFadden, is a probability that some particular sub-type of behavioural rule that exists in the wider population of types of a general behavioural rule, matches with the choice of a particular variant of a type of product. Probabilistic analysis is required in the first instance because, although agents are assumed to know their utility functions, researchers can only observe discrete choices that are made, such that in: U  V (s, x)   (s, x) V is non-stochastic and reflects representative tastes of the population, and  is idiosyncratic and represents individual tastes, where s represents attributes of the individual, and where x represents sub-types of the type of outcome subject to a decision (ibid., p. 108). Recently, applications of this have been made in health economics (Ryan and Hughes, 1999; San Miguel et al., 2000). Here compositional principles involve interpreting each individual’s utility function as being partly observable through outcomes of discrete choice experiments, and partly unobservable and captured by an error term (San Miguel et al., p. 824). Marginal rates of substitution between alternatives within a choice set, which also refers to hypothetical choices through the inclusion of price attributes, are calculated.

Critical realism and statistical techniques 135 The inclusion of price as an attribute allows researchers to calibrate decisions to represent cardinal rather than ordinal utility. However, the consequences of decision-makers acting upon lexicographic preferences raise questions for the use of semi-parametric techniques supported with classical utility theory that are difficult to resolve.9 A means of addressing concerns of imputing specific probability distributions has recently been advanced by Golan et al. (1996). Drawing from thermodynamics and statistical physics, they argue that maximum entropy econometrics resolves many of the problems of the ‘traditional’ approach. Entropy is a measure of the unavailability of a system’s (usually considered to be the universe, and hence closed) thermal energy for conversion into mechanical work. It is an index of degradation or disorder (Georgescu-Roegen, 1971, 1986). Hence, time is unidirectional and irreversible, contradicting the circularity and reversibility of theorising predicated on mechanistic metaphors. Maximum entropy econometrics is claimed to develop a non-linear inversion procedure that furnishes the basis for information recovery.This permits a foundation for ‘conservative inferences’ to be made regarding unknown and unobservable numbers, vectors or functions (Golan et al., 1996). Ex ante, maximum entropy techniques make no presumption over the distribution of variables, and relax the closure supposition of the distribution of variables associated with parametric procedures. The maximum entropy principle revolves around the selection of a distribution where the data are sufficient to determine a probability assignment, which can then be employed in the Langrangian multiplier to generate entropy distributions, revealing more ‘information’ about an underlying probability distribution’s expected value.10 The generalised formal basis of this procedure is:

H(p)  

p ln p k

k

k

H is a measure of the uncertainty of a collection of events, p refers to probability of the kth variable, and ln pk is the natural logarithm of this probability. The derivation of H is informative. Golan et al. import this technique from statistical physics, citing Shannon (1948).11 Shannon, as a communications engineer, seeks to measure the ‘uncertainty’ in the mind of an individual about to receive a message subject to noise (Mayumi, 2001, pp. 32–6). Golan et al. (1996, p. 8) state, Since it is traditional to use probability as a measure of the uncertainty (state of knowledge) we have about the occurrence of a single event, Shannon … used an axiomatic method to define a unique function to measure the uncertainty of a collection of events. (emphasis added, parenthesis in original)

136

John H. Finch and Robert McMaster

Following Georgescu-Roegen’s (1971, 1986) observations of entropy in economics, the ability of maximum entropy econometrics to resolve the issues it sets itself can be questioned. Georgescu-Roegen’s criticises the ‘mechanistic dogma’ of neoclassical economics, with its time reversibility foundation and argues that the Langrangian optimisation technique presumes that agents obey conservation forces – contrary to the degradation inherent in thermodynamics and entropy – deducible from a potential function.This is founded on the underlying supposition in mechanics that measurement on a numeric scale is ubiquitous. Georgescu-Roegen (1971, p. 83) argues that this leads to an ‘ordinalist fallacy’ that there is a number for everything. Georgescu-Roegen’s conceptualisations of entropy, and of economic and social processes, are compatible with critical realist ontology.Also, it reveals that Golan et al.’s approach is mired in classical mechanics, where ontology is conflated with epistemology: witness the conceptualisation of uncertainty as both probabilistic and missing information. This suggests that maximum entropy econometrics is subject to similar concerns of extrinsic and intrinsic closure demonstrated by logit and probit approaches; sharing a common presumption that ‘uncertainty’ can be measured cardinally, and that knowledge claims can be verified or rejected on this basis. Despite the advantages of categorical parametric techniques and maximum entropy econometrics, the earlier criticisms of parametric statistical techniques extend to them. As Stanley (1998, p. 212) argues, ‘[they are] less dependent on specific distributional assumptions, [but] their validity, nonetheless remains conditional on restrictive assumptions’. Georgescu-Roegen’s contribution acts as a timely reminder about the measurability of variables.This extends beyond econometrics mainly-as-regression to econometrics in toto.

An alternative approach: developing econometrics-as-measurement The foregoing emphasised the problems inherent in econometrics mainlyas-regression. However, this by no means precludes the use of a form of econometrics in critical realist inquiry.This section proposes a pragmatic alliance between critical realist ambitions for articulating knowledge claims of economic phenomena as causal explanations, and using non-parametric techniques in aiding the development of such knowledge claims through assessing, and highlighting, demi-regularities.Two important issues await resolution. First, the process of categorisation is central to non-parametric techniques, whether based upon nominal or classificatory scales or ordinal or ranking scales. It is argued in this section that a useful, though partial, source for researchers in formulating such categories is the explanations and understandings of agents involved in the phenomena being researched. Critical realists are frequently suspicious of basing knowledge claims in the form of causal explanations on the everyday explanations of agents, and are often critical of phenomenological, hermeneutic and post-modern approaches to explanation (Bhaskar, 1989, pp. 146–79).

Critical realism and statistical techniques 137 Nevertheless, the formation of categories is important for the organisation of knowledge and learning. In this respect, the scope and role for interpretivism and hermeneutics requires clarification. Second, Lawson (1997) introduces the concept of demi-regularities as approximate regularities that may be observed and recorded over spatial-temporal zones in the form of econometrics as measurement.This is consistent with the capacity of non-parametric techniques of analysis to allow analysis of categorical data. However, we suggest that researchers’ attention should shift from the critical realist concerns with intrinsic and extrinsic systemic openness, and towards considering the likely extent of demi-regularities. In other words, the critical realist aspiration of explaining the explanation is necessarily extended to explaining the conditions that allow researchers to draw upon non-parametric techniques in formulating their knowledge claims as causal explanations. Econometrics as measuring economic phenomena Given the argument of this chapter, that theoretical development and empirical analysis can be mutually re-enforcing if coupled loosely, the implication so far is that such relations are frustrated by formal or informal tight coupling between theoretical development and empirical research. If empirical research is cast in the role of verifying knowledge claims drawn from theoretical development, outcomes may include qualified support, contradiction or ambiguity. Support or contradiction may influence confidence in the adequacy and reliability of a knowledge claim, but will not add anything to the nature of that knowledge claim. Ambiguity has the potential to distance theoretical development from empirical inquiry, and distance it beyond loose coupling. Following Lawson’s critique of econometrics mainly-as-regression, ambiguity is more likely if conditions of closure and composition are assumed, rather than investigated. The argument of this chapter is developed further by investigating issues of measurement, to the effect that different types of measurement may be appropriate for different types of economic phenomena and at different stages of developing and assessing the reliability and adequacy of knowledge claims. Measurement techniques may be guided by immediate purposes as part of the process of formulating and assessing knowledge claims as causal explanations, and may be more closely identifiable with agents’ own views of their situations. Econometrics as measurement ensures that historically specific information is retained; for instance, in definitions of categories and types of measurement suggested by economic agents. Simple non-parametric statistical techniques can add descriptive and analytical content to understanding based on single samples, two or more dependent or independent samples, and correlations between categories of data. Unlike parametric inference, non-parametric ranking or ordering techniques only require researchers to transform their observations of economic phenomena such that they can be mapped onto nominal or classificatory scales, or ordinal or ranking scales (Siegel and Castellan, 1988, pp. 23–8).

138

John H. Finch and Robert McMaster

As noted, the pursuit of econometrics mainly-as-regression often requires information in arithmetic numeric structures, as either interval or ratio scales. By contrast, with econometrics as measurement, the researcher is ‘assigning of numbers to observations in such a way that the numbers are amenable to analysis by manipulation or operation according to certain rules’, such that ‘manipulation will reveal new information about the objects being measured’ (ibid., p. 22). Nominal or classificatory scales classify observations into mutually exclusive categories, and members of different categories are equivalent with respect to the underlying principles of such a classificatory schema. Transformations that are appropriate for data measured in nominal or classificatory scales include descriptive measures of modes, frequencies and counts, none of which alter the information being communicated by the data (ibid., p. 24). Researchers using ordinal or ranking scales organise information into categories that exhibit relations of inequality, and where numeric designations indicate that observations with higher numbers are in some way greater than those with smaller numbers. No inferences may be made as to the intervals between the numbers used in ranking observations. Ranking procedures may have to cope with ties, the presence of which may indicate that too broad a schema is being used, and possibly that the schema is disguising an underlying continuum (ibid., pp. 25–6). Permissible transformations of information that is ranked include describing central tendencies by medians, and investigating relations between categories by calculating rank correlation measures.12 Yet, the point remains that non-parametric procedures, as part of econometrics as measurement, implies that researchers are not imposing or presuming the same degree of closure in their mental models as that associated with econometrics mainly-as-regression. The main advantage of non-parametric techniques, and of econometrics as measurement, is that researchers can be cautious in how they interpret and organise their observations and understandings. The approach is more tolerant of ambiguity than parametric approaches, in fieldwork observations, interview transcripts, biographies and primary documentation, and in researchers interpretations of these types of information. Knowledge claims, including inferences, can then be made in a manner consistent with researchers’ interpretations of their observations, and with their confidence of their emerging and broader theoretical framework. Non-parametric techniques can thus be employed as an element of theory development and articulation, as opposed to being confined to theory and hypothesis testing. Interpretive aspects of forming categories Non-parametric techniques require researchers to translate their emerging understandings of phenomena into rough and ready categories. Categories are also important for agents in organising knowledge and learning, and such schema have been described as personal constructs, interpretive frameworks, mental models and quasi-morphisms (Kelly, 1963; Holland et al., 1986).While agents’ categories do not provide an a priori basis for non-parametric analysis,

Critical realism and statistical techniques 139 a reasonable starting place for researchers in forming categories is agents’ own understandings, especially if these can be articulated in such a way as to reflect or confront some aspects of researchers’ own otherwise partly tacit categorical schemas. Critical realist writers have recognised the inevitability of such a position, but have an uneasy relationship with the practical requirement of engaging in interpretive tasks, such as establishing categories for the purposes of theory development in connection with categories of agents’ understandings of their own situations.13 Interpreting agents’ understandings is inevitable because social structures only exist as they are reproduced and transformed through agents’ actions. Critical realist unease with the role of agents’ explanations as part of causal explanations of economic phenomena may emanate from the interpretive rather than causal ambitions of research consistent with hermeneutics.The usefulness of agents’ understandings articulated as explanations is tempered for critical realists by the presumed context-specific nature of that knowledge, and by difficulties in articulating its inevitable tacit and unreflective content, often summarised as agents’ opaque understandings of the physical and social structural properties of their situations (Lawson, 1997, pp. 192–3).At the same time, if causal explanation is to have a significant empirical component, and if this is to include detailed case studies, oral reporting, including interviews, biographies and so on, interpretive issues in formulating causal explanations should be addressed. Agents may have opaque understandings of their social situations, so researchers have opaque understandings too. If such understandings can be articulated through dialogue in interviews and case studies, and in multiple accounts of events, such articulations may be coded and categorised by the researcher. The dimensions and extent of tacit knowledge are not fixed, but may be adjusted over time with effort and dialogue in devising means of articulation or coding (Boisot, 1995). How pervasive are demi-regularities? Researchers seeking to develop knowledge claims as causal explanations can use econometrics as measurement, including simple non-parametric statistical techniques, and a useful basis for these can include categorisations formed in part from agents’ understandings of their circumstances. But, following Lawson, the tractability of this possibility depends upon the presence of demi-regularities associated with the economic phenomena being explained. A demi-regularity: indicates the occasional, but less than universal, actualisation of a mechanism or tendency, over a definite region of time-space. The patterning observed will not be strict if countervailing factors sometimes dominate or frequently co-determine the outcomes in a variable manner. But where demi-regs are observed there is evidence of relatively enduring and identifiable tendencies in play. (Lawson, 1997, p. 204)

140

John H. Finch and Robert McMaster

The critical realist conception of demi-regularities can be traced to the ‘stylised facts’ employed by Kaldor and Okun as the basis of their analytical processes. Kaldor (1985, p. 9) maintains that ‘stylised facts’ do not imply truth in every instance, only that these facts are true in the ‘broad majority of observed cases’, sufficient to warrant an explanation for their incidence. Institutionalists have also been influenced by Kaldor’s notion. Hodgson (1998, p. 173) invokes the ‘stylised fact’ that, when contrasted to East Asian economies, US productivity growth has been sluggish. He observes that institutionalist analysis proceeds from ‘stylised facts’ to engage in an examination of underlying structures. However, Lawson (1997, p. 208) considers that the employment of ‘stylised facts’ is ‘strategically unwise’. Lawson’s rationale is based on the meaning of the term ‘stylised’ as an intention to express partial regularities as strict ones. To avoid this unintended meaning (and to circumvent others’ attempts to adopt a Kaldorian approach in closed system, deductive models) he advocates the term ‘demi-regularities’ to convey the intended meaning of ‘stylised facts’. Lawson discusses demi-regularities in two ways. First, they are described as ‘rough and ready’, ‘remarkable’ given necessarily open social systems, and also ‘widely in evidence’ (ibid., pp. 204–5).They are something to be explained given the fundamentally open system nature of economic phenomena. Second, they allow an extension and simultaneous loosening of the notion of experiment among both natural and social phenomena beyond strict conditions of intrinsic and extrinsic system closure.This extension retains the possibility of experiments, depending upon some form of reliable and known control rather than strict intrinsic and extrinsic closure, but dispenses with the precise calibration consistent with empirical co-presence of conjectured cause and effect. Through invoking demi-regularities, Lawson describes a three-way categorisation of the domain of the empirical: ‘there is a possibility … of a continuum of pattern outcomes stretching from closed systems of constant conjunctions of events to an inchoate random flux, with contrastive demi-regs lying between these extremes’ (ibid., p. 220). The contrast element of contrastive demi-regularities indicates some possibly surprising aspect to an observed partial regularity in the form of something like a natural experiment. A control group may exhibit a particular pattern of demi-regularity over a temporal-spatial zone, but a different outcome may occur in some other ostensibly similar temporal-spatial zone. The difference between the two observed situations that exhibit demi-regularities may be understood in the context of a rough and ready control or set of controls that can allow the difference itself to be explained adequately. Such an explanation would be in terms of some factor(s) being present in one observed situation and absent in some other similar situation. Note that the perception of contrastive demi-regularities depends upon categories within and between economic phenomena, that observation is theory-laden (Sayer, 1992). Hence, categories themselves become pragmatic hypotheses that may be confirmed through success in categorising further observations. Given causal explanations are developed through understanding contrastive demi-regularities, through something like rough and ready natural experiments,

Critical realism and statistical techniques 141 the basis of these related partial regularities should form part of any explanation (Lawson, 1997, p. 213), and is key in ‘putting critical realism to work’ (Pratt, 1995). In essence, this is the principle of explaining the explanation that is set out in greater detail in the next section. Institutional economics, particularly analyses of shared mental models may be useful here (Lane et al., 1996; Downward et al., 2002; Chapter 6 of this volume).

Using non-parametric techniques in causal explanations involving contrastive demi-regularities It is proposed that non-parametric techniques (following econometrics as measurement) can be deployed, as part of a retroductive strategy, in support of and in extension to causal explanations of economic phenomena. In contrast to econometrics mainly-as-regression, many simple non-parametric techniques applied to categorical data are not in themselves causal explanations. Rather, different non-parametric tests can be undertaken in different phases of developing and assessing the reliability of knowledge claims about economic phenomena. Undertaking non-parametric analysis depends on researchers first being able to characterise economic phenomena in the form of contrastive demi-regularities (as with any attempt at causal explanation) in which aspects of phenomena may be articulated in nominal or classificatory scales or ordinal or ranking scales.The roughness and readiness of categorical schema, as opposed to the precision of parametric variables, matches that of contrastive demiregularities. Following the identification and description of demi-regularities, the process of articulating knowledge claims follows retroductive (or abductive) reasoning, which is: the movement, on the basis of analogy and metaphor amongst other things, from a conception of some phenomenon of interest to a conception of some totally different type of thing, mechanism, structure or condition that, at least in part, is responsible for the given phenomenon. (Lawson, 1997, p. 24) The context of investigation is of considerable importance in the process of retroduction, and indeed retroductive inferences about causal hypotheses. Retroduction, by seeking to identify underlying and frequently non-observable causal mechanisms, involves the personal experiences, beliefs and experience of the investigator. Retroduction will be subject to differences in interpretations and value judgements; such is the essence of scientific discourse. Tsang and Kwan (1999) emphasise the desirability of feedback in analysis: conjectures and established knowledge are subject to amendment as the process of analysis unfolds. Thus, ‘doing’ critical realist research is principally an explanatory and interpretive (as opposed to predictive) exercise. A clearer understanding of underlying theories of measurement and data construction associated with the

142

John H. Finch and Robert McMaster

identification and description of demi-regularities would help researchers formulate and assess knowledge claims as causal explanations. In our view, the inference of many non-parametric techniques would aid this task. In his outline of examples of demi-regularities, Lawson draws on Leamer’s (1983) anecdote about farmers’ hypotheses concerning higher yields of a particular crop.Two farmers observe, in different fields, that yield is higher where crops are grown under trees. Lawson employs this as an example of a contrastive demiregularity, and notes Leamer’s two competing hypotheses: farmer one claims that higher yields result from birds’ droppings, whilst farmer two claims that it is a consequence of the amount of shade.A further example of a contrastive demiregularity utilised by Lawson (1997, p. 255) is the relatively poor productivity growth performance of the UK when contrasted with other industrialised states (as measured by growth in GDP per man-hour). He then examines a number of competing explanations, including underlying industrial relations. Ironically, these examples of demi-regularities involve considerable closure to enable measurement.The ‘farmers’ yields’ imply that interval scales, at least, can be established.This may not be an unreasonable conjecture, but there is a tacit presumption in the anecdote that the quality of yield does not vary either through time or between the farmers. In the productivity demi-regularity, Lawson invokes a much stronger closure presumption concerning the measurement, both time series and cross-sectional, and data construction of productivity. The most obvious assumption is that an interval scale can measure productivity, and that data construction is sufficient for this process in that the probability of error is considered tolerable (a point that Lawson criticises regression econometrics on). Moreover, the argument maintains that aggregated averages are admissible in the establishment not only of the demiregularity, but following from this, the direction, and even the focus of the investigative process. In the first example (farmers’ yields) there are, from a critical realist perspective, firmer grounds for supposing that empirical measurement of this sort is justifiable from an open-systems ontology. However, the closure conditions for measurement in the second case are rather more problematic. Not only is there a presumption that interval measurability is possible, but that there is some durability in underlying structures that permits the validity of such measurement: a constant and continuous conjunction of events? Indeed, reference to the controversies over the measurement of unemployment in the UK over the past two decades further highlights Lawson’s invocation of closure. Arguably, by employing interval scales in this manner, Lawson concedes that there is a role for quantification through interval scales in critical realist inquiry. In extremis this may be contradictory to the critical realist vision of causal explanation. This is important, since it suggests further that non-parametric inference can be complementary to ‘putting critical realism to work’. Non-parametric techniques can achieve this in two ways. First, they can indicate the nature of demi-regularities by highlighting measures of association between samples. For

Critical realism and statistical techniques 143 instance, such techniques have proven to be of use in behavioural psychology in establishing the similarity or otherwise between groups of respondents (including a control group) (Siegel and Castellan, 1988). While not indicating causality, it can establish the relatedness of samples and populations and some measure of association. Moreover, non-parametric inference can achieve this by invoking fewer closure conditions in terms of measurement and data construction than Lawson’s productivity demi-regularity. In this respect the need for critical realists to take explicit account of theories of observation and data construction is emphasised, as with Georgescu-Roegen’s ‘ordinalist fallacy’. Second, non-parametric analysis can also provide a partial basis for assessing the empirical adequacy of a hypothesis. In argumentum, if a non-parametric technique reveals some ‘surprising’ association, or by the same token nonassociation, it could contribute to the reconsideration of a demi-regularity, and also established knowledge claims. However, it should be emphasised that non-parametric inference based on categorical data, as with any statistical inference, is not without issues of measurement bias and categorisations are susceptible to the biases of the researcher. Nevertheless, given its weaker closure conditions this is unlikely to be as problematical as the deployment of interval and ratio scale measures in the establishment of demi-regularities. The foregoing raises the issue as to the current employment of nonparametric analysis of categorical data in economics.There is some evidence of specific uses of non-parametric tests, but they tend to be isolated instances, and particular to certain fields. Specifically, some investigations of comparative efficiency and productivity growth invoke a combination of parametric and non-parametric techniques.The latter has also been applied in expected utility theory. Leibenstein and Maital (1992) and Ray and Mukherjee (1996) use nonparametric ‘analysis’ as part of their data envelopment analysis of efficiency. Leibenstein and Maital attempted to assess X-efficiency levels based on a set of input and output measures for an American hockey team, whereas Ray and Mukherjee assessed the differential productivity of US airlines. Data envelopment analysis is a linear programme that attempts to measure efficiency frontiers and departures from these frontiers. Leibenstein and Maital (1992, p. 429) state that data envelopment problems may be formulated in terms of input-minimising and output-maximising, where the technique provides a scalar measure of (in)efficiency. In this way, they maintain, ordinal measures of personality and motivation (through psychological tests) can be admitted to the analysis. They view this as an innovation in that it permits a wider and more comprehensive testing of theory. In her analysis of expected utility theory and endowment effects, Morrison (2000) employs a Wilcoxon one-tailed test to establish the endowment effect on chain gambles. Despite the obvious distinctions between these approaches they all utilise non-parametric analysis to verify or establish knowledge claims in the same manner as conventional econometrics mainly-as-regression approaches.This is incompatible with critical realism, and the argument presented here.

144

John H. Finch and Robert McMaster

By contrast, McMaster’s (1996) use of a non-parametric technique – the Kendall Coefficient of Concordance – in his analysis of the sources of cost changes arising from the implementation of competitive tendering in health and local authorities in the UK in the early 1990s, makes no such claims.The Kendall Coefficient of Concordance measures the degree of association between respondents’ ranking of objects (Siegel and Castellan, 1988, p. 271). McMaster found that all authorities in his sample demonstrated similar rankings in their assessments of the factors contributing to cost savings. No causation is inferred. In effect, this study may have revealed a demi-regularity, or a surprising association that highlighted the need for further investigation. Given that non-parametric techniques may be used in pragmatic alliance with causal explanations that are formulated in accordance with the notion of ontology set out by critical realists, non-parametric techniques cannot in themselves provide causal explanations.The argument pursued here is not a prioristic in favour of a particular set of measurement and analytic techniques. Nevertheless, the rhetoric of statistical techniques is usually in terms of tests of hypotheses. Principles of abduction and retroduction encourage researchers to break down the big events in contexts of discovery and justification into smaller and iterative events such as category formation, category composition and relations between categories and between a category’s constituents. Each research activity has a practical research purpose, such that each may be considered as a hypothesis in relation to such a purpose. Seeing retroduction or abduction in this practical manner sits uneasily with the rhetoric of statistical tests. Tests suggest some once and for all big event in the context of justification. But retroduction or abduction seems to envisage smaller iterations across related events involved in formulating knowledge claims. In particular, early stage hypothesis articulation may be understood more appropriately in terms of sample and population being synonymous. McCloskey (1986, p. 163) distinguishes statistical and economic significance. Statistical significance pertains to the researcher who wants to know ‘what the probability is that because of the small sample he faces he will make a mistake of excessive gullibility in accepting a false statistical proposition’ (ibid., p. 160, original emphasis). Economic significance, following McCloskey, has much in common with pragmatic principles of retroduction. It is to do with the quality of hypotheses, and the practical consequences of choosing to frame one’s understanding of an economic phenomenon in one way rather than another (ibid., p. 158).14

Conclusion Arguably, the relationship between formalism and realism in economics has not been served well by an empiricism that focuses on the observation of co-presence of conjectured cause and effect. Formalism has acquired characteristics of tenacity and resilience in the face of many empirical assessments that have proven difficult to reconcile.Tenacity and resilience in formalism are not necessarily undesirable qualities, but theorists are isolating themselves from useful

Critical realism and statistical techniques 145 sources of information and ideas in theory development by casting realism as empiricism. Critical realists promote the pragmatic principle of retroduction or abduction as a more effective combination of theorising and realism. More effective because it is predicated upon a layered and structured social ontology that is necessarily open and depends upon agents to reproduce and transform its structures. Empirical research can still play an important role in undertaking causal explanations in recognition of necessarily open systems predicated upon structured and layered social ontology dependent upon its agents for its reproduction and transformation. Empiricism can have many interrelated roles. None of these roles constitute causal explanation, but all can contribute to causal explanation.And empiricism can include econometrics defined as measurement rather than econometrics defined mainly-as-regression.The scope of empiricism depends upon the scope of demi-regularities, or observed partial regularities among economic phenomena.These may come about over defined spatial-temporal zones through institutional social practices that require explanation. Causal explanation may proceed further if contrastive demi-regularities, or rough and ready natural experiments, can be detected. The categories in which data are arranged may more closely reflect categories in which agents themselves arrange data in understanding their situations, although this requires considerable interpretive expertise among researchers. Further, the roughness and readiness of categorical measurement is matched by the roughness and readiness of the contrastive demi-regularities that allow empiricism to have a role in causal explanation in the first instance. Non-parametric statistical analysis is an appropriate research activity within econometrics as measurement, rather than mainly-as-regression. It can contribute to the analysis of the extent of demi-regularities, and to measures of the nature of any relationship between demi-regularities cast into a contrastive relationship by researchers expressing knowledge claims following retroductive or abductive principles. This is because non-parametric statistics involve measurement in the form of nominal or classificatory scales, or ordinal or ranking scales, and not arithmetic numeric scales, so are not amenable to arithmetic manipulation that requires meaningful quantification of intervals between symbols that represent equivalence or inequality.

Notes 1 This chapter is based on a paper entitled ‘On categorical variables and non parametric statistical inference in the pursuit of causal explanations’ (Cambridge Journal of Economics, 26: 753–72). Earlier versions were presented at the Cambridge Realist Workshop conference, May 2000, the Association of Heterodox Economists conference, London, June 2000, and the European Association for Evolutionary Political Economy conference, Berlin, November 2000. We are grateful to Harminder Battu, Rebecca Coke, Sheila Dow, Paul Downward, Jochen Hartwig, David Harvie, Thorbjoern Knudsen, Tony Lawson, Wendy Olsen, Euan Phimister and an anonymous referee for comments and suggestions. The usual disclaimer applies.

146

John H. Finch and Robert McMaster

2 Dow (2000) takes critical realists to task over the assertion that Hume was somehow a naïve empiricist, arguing instead that Hume had a sophisticated account of causal explanation. 3 The relationship of econometrics mainly-as-regression and econometrics as measurement is becoming a contentious area within economics. Downward et al. (2002; chapter 6) argue that agents require a degree of stability in which to base decision-making and action, and that institutions (recursive social practices), which exhibit properties of emergence, may provide a basis for this. Such conditions may be described as quasi-closure over particular dimensions of space and time. Setterfield (1997; chapter 5), Crotty (1994) and Cottrell (1998) make similar claims about local, synthetic or conditional closures which may facilitate decision-making and action among agents, although do not develop the argument to consider its implications for empirical research. In such circumstances, any type of closure must form part of any empirical explanation. 4 Backhouse (1997) provides a commentary on Hendry’s econometric method, emphasising the changing role of econometrics mainly-as-regression from being involved in a clearly demarcated context of verification, following conjectures cast in a context of justification, to a more developmental role consistent with the broad sweep of Lakatos’ method of scientific research programmes. This ‘broad sweep’ interpretation focuses less on the role of Lakatos’ ‘sophisticated falsification’, and more on developing econometric models in particular contexts, and then testing them in other similar contexts. 5 Of course, the unobservability of causation is well recognised in the critical realist literature (Lawson, 1997, part 1), and is at the centre of the critical realist approach. 6 Hence, this criticism applies also to Backhouse’s Lakatosian commentary on Hendry. 7 Tenacity and resilience are used in the sense of Feyerabend’s Against Method, in which theoretical development and proliferation is favoured against a too restrictive and privileged role for methodologically informed epistemology. 8 This may be described as an Aristotlean–Thomistic, involving abstracting items considered to be vital, and is reminiscent of Alfred Marshall’s partial equilibrium approach. Such abstraction is a mental activity undertaken in inquiry and is the analogue of experimental closure. Ideal typification is another example. 9 There are also further closure issues that have considerable bearing upon compositional principles. San Miguel et al. (2000, p. 831) concede that,‘research is required on the axioms that underlie random utility theory, whether individuals obey these axioms, and, whether the data produced from such studies can be treated as cardinal’. Loasby (1976) explains that decision-making among choices is about exploration and learning, and has a strongly constructive element in that agents may in some way choose between alternative futures. If choosing is to some extent experimental (i.e. undetermined), the act of choice and its monitoring imply, first, open systems and, second, learning about preferences through learning about the characteristics of alternatives in a choice set. Social customs, rules and institutions may well provide stability in a social world and some kind of closure prior to decision-making, but compromises compositional principles in that individual choices have characteristics of emergent properties rather than individual responses that can be separated and then summed. 10 Golan et al. (1996) state that the standard illustrative example concerns a six-sided die, where the only information supplied is the average outcome from a large number of throws, with the sum of probabilities being one.There are an infinite number of distributions that can be generated by the six values of the die. But by selecting probabilities that maximise entropy, and working this through the Langrangian, it is claimed that optimisation techniques will reveal estimated distributions for each of the six points. Entropy is maximised at 3.5, the expected value.

Critical realism and statistical techniques 147 11 Shannon’s definition of the H measure is directly traced to the physicist Ludwig Boltzman’s earlier attempt to synthesise thermodynamics, and the concept of entropy in particular, within the rigidity of statistical mechanics (GeorgescuRoegen, 1971, p. 7). 12 Texts on non-parametric statistical techniques often compare the explanatory power of parametric and non-parametric techniques, but his approach is not pursued here. Rather, it is recognised that the approach to measurement that is adopted is constrained by the nature of the phenomena themselves (including agents’ understandings of their situations), and by the ontology that shapes a researcher’s expectations and understandings of these phenomena. 13 Lewis (1996) focuses on the critical role of the ‘linguistic context’ of metaphor, and compares idealist and realist treatments of language. This is a different kind of argument to one focussing on developing approaches and perspectives on connecting different discourses that are directed (broadly) to, and develop within, different purposes. 14 Another ally in casting the role of non-parametric analysis of categorical data within an iterative approach to retroduction or abduction is in grounded theorists’ distinctions between theoretical and statistical sampling (Glaser and Strauss, 1967, pp. 32–45). Statistical sampling is guided by principles of random selection, given some knowledge of a larger population from which the sample is drawn and so shares important characteristics, for the purposes of drawing inferences.Theoretical sampling is undertaken initially in ignorance of the characteristics of some overall population. An investigation is guided, especially in its early stages, by principles of category saturation that can be appraised by an absence of surprises yielded from additional observations, given that some categorical interpretive schema is being developed. Measures of relatedness, for instance calculated from Wilcoxon matched-pairs signed rank tests, provide useful information in developing hypotheses for different purposes, and these data may be interpreted relative to whether the researcher is thinking in terms of the sample being the population, or whether there may feasibly be some wider population with which the (now) sample shares critical characteristics.

References Backhouse, R.E. (1994) ‘The Lakatosian legacy in economic methodology’, in Backhouse, R.E. (ed.) New Directions in Economic Methodology, London and New York, Routledge, pp. 173–91. —— (1997) Truth and Progress in Economics, Cheltenham, Edward Elgar. —— (1998) ‘If mathematics is informal, then perhaps we should accept that economics must be informal too’, Economic Journal, 108: 1848–58. Bhaskar, R. (1978) A Realist Theory of Science (second edition), Hemel Hempstead, Harvester Press. —— (1979) The Possibility of Naturalism. A Philosophical Critique of the Contemporary Human Sciences, Brighton, Harvester Press. —— (1989) Reclaiming Reality.A Critical Introduction to Contemporary Philosophy, London and New York,Verso. Boisot, M.H. (1995) Information Space. A Framework for Learning in Organizations, Institutions and Culture, London and New York, Routledge. Caldwell, B.J. (1994) Beyond Positivism. Economic Methodology in the Twentieth Century (revised edition), London and New York, Routledge. Chick,V. (1998) ‘On knowing one’s place: the role of formalism in economics’, Economic Journal, 108: 1859–69.

148

John H. Finch and Robert McMaster

Cottrell, A. (1998) ‘Realism, regularities and prediction’, Review of Social Economy, 56: 347–55. Crotty, J. (1994) ‘Are Keynesian uncertainty and macro theory incompatible? Conventional decision making, institutional structures and conditional stability in Keynesian macro models’, in Dymski, G. and Pollin, R. (eds) New Perspectives in Monetary Theory. Explanations in the Tradition of Hyman P. Minsky, Ann Arbor, University of Michigan Press, pp. 105–39. Dow, S.C. (1990) ‘Beyond dualism’, Cambridge Journal of Economics, 14: 143–57. —— (1998) ‘Formalism in economics. An editorial note’, Economic Journal: 108, 1826–28. —— (2000) ‘Historical reference: Hume and critical realism’, Cambridge Journal of Economics: 26, 683–696. Downward, P., Finch, J.H. and Ramsey, J. (2002) ‘Critical realism, empirical methods and inference. A critical discussion’, Cambridge Journal of Economics: 26, 481–500. Feyerabend, P.K. (1970) ‘Consolations for the specialist’, in Lakatos, I. and Musgrave,A. (eds) Criticism and the Growth of Knowledge, Cambridge, Cambridge University Press, pp. 197–230. —— (1988) Against Method (revised edition), London,Verso. Georgescu-Roegen, N. (1971) The Entropy Law and the Economic Process, Cambridge MA., Harvard University Press. —— (1986) ‘The entropy law and the economic process in retrospect’, Eastern Economic Journal, 12: 3–25. Glaser, B.G. and Strauss, A.L. (1967) The Discovery of Grounded Theory. Strategies for Qualitative Research, New York: Aldine Publishing. Golan, A., Judge, G. and Miller, D. (1996) Maximum Entropy Econometrics. Robust Estimation With Limited Data, Chichester, John Wiley and Sons. Granger, C.W.J. (2001) ‘Comparing methodologies used by statisticians and economists for research and modeling’, Journal of Socio-Economics, 30: 7–14. Gujarati, D.N. (1988) Basic Econometrics (second edition), London, McGraw-Hill. Härdle, W. (1995) Applied Nonparametric Regression, Cambridge, Cambridge University Press. Hendry, D.F. (1980) ‘Econometrics – Alchemy or science?’, Economica, 47: 387–406. Hodgson, G.M. (1998) ‘The approach of institutional economics’, Journal of Economic Literature, 36: 166–92. Holland, J.H., Holyoak, K.J., Nisbett, R.E. and Thagard, P.R. (1986) Induction. Processes of Inference, Learning, and Discovery, Cambridge MA and London, MIT Press. Hollander, M. and Wolfe, D.A. (1973) Nonparametric Statistical Methods, London, John Wiley and Sons. Kaldor, N. (1985) Economics Without Equilibrium, Cardiff, University College Cardiff Press. Kelly, G.A. (1963) A Theory of Personality.The Psychology of Personal Constructs, New York and London,W.W. Norton and Company. Keuzenkamp, H.A. (1995) ‘The econometrics of the Holy Grail – A review of “Econometrics:Alchemy or Science? Essays in Econometric Methodology” ’, Journal of Economic Surveys, 9: 233–48. Krugman, P. (1998) ‘Two cheers for formalism’, Economic Journal, 108: 1829–36. Lakatos, I. (1970) ‘Falsification and the methodology of scientific research programmes’, in Lakatos, I. and Musgrave, A. (eds) Criticism and the Growth of Knowledge, Cambridge, Cambridge University Press, pp. 91–230.

Critical realism and statistical techniques 149 Lane, D., Malerba, F., Maxfield, R. and Orsenigo, L. (1996) ‘Choice and action’, Journal of Evolutionary Economics, 6: 43–76. Lawson,T. (1997) Economics and Reality, London and New York, Routledge. Leamer, E.E. (1983) ‘Let’s take the con out of econometrics’, American Economic Review, 73: 31–43. Leibenstein, H. and Maital, S. (1992) ‘Empirical estimation and partitioning of Xinefficiency: a data-envelopment approach’, American Economic Review (Papers and Proceedings), 82: 428–38. Lewis, P.A. (1996) ‘Metaphor and critical realism’, Review of Social Economy, 54: 487–506. Loasby, B.J. (1976) Choice, Complexity and Ignorance, Cambridge, Cambridge University Press. McCloskey, D.N. (1986) The Rhetoric of Economics, Brighton,Wheatsheaf Books. —— (1994) Knowledge and Persuasion in Economics, Cambridge, Cambridge University Press. McFadden, D. (1974) ‘Conditional logit analysis of qualitative choice behavior’, in Zarembka, P. (ed.) Frontiers in Econometrics, New York and London, Academic Press, pp. 105–42. McMaster, R. (1996) ‘A non-parametric approach to identifying the sources of cost savings arising from competitive tendering’, Applied Economics Letters, 3: 463–66. Manicas, P. (1998) ‘A realist social science’, in Archer, M., Bhaskar, R., Collier, A., Lawson, T. and Norrie, A. (eds) Critical Realism. Essential Readings, London and New York, Routledge, pp. 313–38. Mayumi, K. (2001) The Origins of Ecological Economics. The Bioeconomics of GeorgescuRoegen, London and New York, Routledge. Morrison, G.C. (2000) ‘The endowment effect and expected utility’, Scottish Journal of Political Economy, 47: 183–97. Pagan, A. and Ullah, A. (1999) Nonparametric Econometrics, Cambridge, Cambridge University Press. Peirce, C.S. (1935) Collected Papers, Volume V, Pragmatism and Pragmaticism, edited by Charles Hartshorne and Paul Weiss, Cambridge, Harvard University Press. Pratt, A.C. (1995) ‘Putting critical realism to work: the practical implications for geographical research’, Progress in Human Geography, 19: 61–74. Ray, S.C. and Mukherjee, K. (1996) ‘Decomposition of the Fisher ideal index of productivity: a non-parametric dual analysis of US airlines data’, Economic Journal, 106: 1659–78. Runde, J.H. (1998) ‘Assessing causal economic explanations’, Oxford Economic Papers, 50: 151–72. Ryan, M. and Hughes, J. (1999) ‘Using conjoint analysis to assess women’s preferences for miscarriage management’, Health Economics, 6: 261–73. San Miguel, F., Ryan, M. and McIntosh, E. (2000) ‘Applying conjoint analysis evaluations: an application to Menorrhagia’, Applied Economics, 32: 823–33. Sayer,A. (1992) Method in Social Science.A Realist Approach (second edition), London and New York, Routledge. Setterfield, M. (1997) ‘Should economists dispense with the notion of equilibrium?’, Journal of Post Keynesian Economics, 20: 47–76. Shannon, C.E. (1948) ‘A mathematical theory of communication’, Bell System Technical Journal, 27: 379–423.

150

John H. Finch and Robert McMaster

Siegel, S. and Castellan, N.J. (1988) Nonparametric Statistics for the Behavioral Sciences (second edition), New York, McGraw-Hill. Stanley, T.D. (1998) ‘Empirical economics? An econometric dilemma’, Journal of Economic Issues, 32: 191–218. Tsang, E.W.K. and Kwan, K. (1999) ‘Replication and theory development in organizational science: a critical realist perspective’, Academy of Management Review, 24: 759–80. Wible, J.R. and Sedgley, N.H. (1999) ‘The role of econometrics in the neoclassical research program’, in Garnett, R.F. (ed.) What Do Economists Know? New Economics of Knowledge, London and New York, Routledge, pp. 169–89. Ziman, J. (1978) Reliable Knowledge. An Exploration of the Grounds for Belief in Science, Cambridge, Cambridge University Press.

Part IIC

Objects, data and theory

9

Triangulation, time and the social objects of econometrics Wendy Olsen

Introduction Econometrics tends to assume that the objects it describes exist. Over time this assumes an existence that is unchanging in the relevant characteristics.This simplifying ontic assumption (ontic meaning ‘about being’, just as ontology is the theory of existence) is in some cases not warranted. The duration of causal mechanisms, of object-relations, and of objects should be the subject of empirical research rather than of ontic assumptions. Model-fitting exercises (both using panel data and using time-series data) usually must assume that the ‘things’ being measured are fixed in their nature over time, varying only in their quantum. Dummy variables can improve econometrics (see Chapter 8 in this volume), but a basic assumption of econometric technique is the constancy of what is represented by the variables available. Lawson (1997) calls such research empirical realist, and he identifies these practices closely with neoclassical theoretical underpinnings. Among heterodox economists, such as Post Keynesians, even where the theoretical model is not orthodox the problem of assuming away social change may still be problematic in time-series data analysis. This chapter explores the problems associated with assuming that things exist over time, looking at five main topics: How to augment the time-series data; what are the epistemological criteria for heterodox economic research; examples illustrating the problem with assuming things can be represented by timeseries measurement; and how realists see things changing. The arguments are generally consistent with the more detailed presentation of a realist approach by Lawson (1997 and elsewhere) although Lawson’s work is often interpreted as being anti-econometrics.The author’s own research experience is also referred to, to illustrate the opportunities for triangulated economic research.

Augmenting time-series data This section illustrates the need to augment time-series data by both contextualising research and exploring the meaning of expressions.The objective is to avoid naïve realism and the epistemic fallacy by using alternative, complementary forms of research. The epistemic fallacy is the implicit argument that the data

154

Wendy Olsen

exist, so the things represented must exist. In Hendry and Richard’s model of the Data Generating Process (DGP), a more complex relation is posed between reality and data – but Hendry does not recommend either qualitative or innovative economic research into DGPs (Hendry and Richard, 1990). Time-series data can be used by researchers but are best used in conjunction with other ways of research.This position is not controversial outside economics. However, there are still limits to the ‘truth’ or validity of empirical research. For instance, Quine pointed out that testing hypotheses in statistics does not enable one to test the underlying assumptions, and McGovern has argued persuasively that the Quinian paradox applies to sophisticated tests undertaken in the financial markets literature (McGovern, 2002; for more detail about the paradox, see Boylan and O’Gorman, 1995). At this juncture, an example helps to show how the world places limits on our ability to wrongly describe the world over an extended period of time. Recent research on the gender wage differential in the UK used time-series data, work-life histories, a primary survey involving attitude questions, and a critique of methodological individualist empirical research (see Walby and Olsen, 2002). In the analysis of wage levels, individual-level time-series data are often used. If you ignore firm-level factors and attribute all wage differentials to personal characteristics, some anomalies begin to show up. Human capital theory tells us that one’s experience in work should lead to a rising wage level over time, but the work-life histories of the British Household Panel Survey generated a paradoxical result. Workers’ part-time work experience actually reduced their wage rate, apparently! (ibid.) Figure 9.1 illustrates the situation that leads to such an anomaly. We can learn from this anomaly. Already, some theories of wage rates have left behind methodological individualism in favour of looking at national wage bargaining processes; looking at unions’ roles in wage-setting; looking at intrafirm and cross-firm labour market segments with special wage-determination mechanisms (Rubery et al., 1999). Interesting literature from feminist economics

Level 3

State regulation and welfare regime

Level 2

Occupational and firm-based factors operationalised by: wage bargaining, union roles, gendered occupational segregation, labour market segmentation, and firms’ strategies for using cheaper labour (e.g. part-time workers)

Level 1

Human capital (individual level) operationalised by: education level, skills, years of work experience, years of service with current employer

Figure 9.1 Illustration of three ontic levels in wage determination research.

Triangulation, time and econometrics

155

contributes to a realisation that household-level factors are also an important, separate dynamic (e.g. Tijdens, 2002). Organisational theorists further show that analysis at firm level could contribute to the explanation of wages. Human capital theory should never, in the face of all this research, have been considered to be independent of complementary theories rooted in other social science disciplines. Specifically, the lower wages of part-time work in this context reflect the ability of certain types of firms to utilise certain types of employees, whose bargaining power is low and who have non-monetary rewards in mind when they take part-time jobs; this can lower the observed wage. (Indeed non-monetary rewards place in doubt the whole wage-measurement process.) However for one’s total life experience of part-time work to continue to have a downward impact on one’s current wage (as the work-life history data showed in multiple linear regression) casts doubt on human-capital theory in its own right. Because of complex causation across the levels, or the interlocking strata, the original wage equation at individual level begins to look like a bad representation (Byrne, 1998). The limits to the truth of research reports can thus fall under at least two headings. First, truth claims based upon time-series data inherently cannot exhaust the complex social meanings underlying the practices represented.The ‘constraints’ imposed by the world (constraints on our claims about the world if we want to avoid falsehood) are knowable if one is open to their influence. As Collier has put it, referring to a 1930s philosopher, Macmurray, who supported the mathematicisation of scientific thought, We need to recognise the existence and validity of non-mathematical, qualitative and concrete sciences … in fact the ‘science’ that conforms closest to MacMurray’s model of additive atomism is neo-classical economics with its econometric pretensions. (Collier, 1999, p. 83) Collier goes on to say that the ‘standing of this [orthodox economic] “science” is contentious’ (ibid.; phrase added). In his view, realists need to challenge the positivist logic of ‘proofs’ in neoclassical theory. They must also question the worth and categorical accuracy of econometric data. Econometrics itself can be challenged, quite apart from questioning the substance or deductivism of neoclassical theorising. In other words, we need to revise our conception of longitudinal data analysis before we will be ready for a heterodox economic research practice. Collier perhaps too readily equated neoclassical modelling with econometrics (as did Lawson, 1997), seeing them as intrinsically linked.The critiques of the two have distinct elements. This chapter concentrates on the critique of time-series econometrics. In summary, Collier’s arguments for a contextualised research practice can be taken further and made very specific. As realists try to allow for ongoing

156

Wendy Olsen

Box 9.1 Three routes toward a realist practice in economics 1 2

3

Using multi-method research (i.e. combining techniques), which is widely known as triangulation. Becoming aware of the policy interventions or user criteria for worthy knowledge, since these ‘purposes’ are intrinsically important for all valued economic research. Using a pluralist set of epistemological criteria to judge the worth of statistical analyses and their related research outputs.

social change in the things to which their variables refer, three possible routes to empirically examining social change can be established as indicated in Box 9.1. Specific examples of multi-method research for econometricians might include: using longitudinal data and analysing oral life histories; using longitudinal data and gathering case studies of firm trajectories; using longitudinal data in multi-level analysis, and collecting documents about the higher-level phenomena as they interact with (affect and are affected by) other phenomena; using longitudinal data whilst studying the political economy of the emergence of new policy regimes, including the analysis of changing discourses in the political realm and their economic causes and effects. The three routes are closely interrelated and all rest upon the assumption that a culturally contextualised economics would interact with the real socioeconomies upon which it is based. Kanth has put this argument rather strongly in his critique of euro-centric economics (Kanth, 1999). Kanth argues that one must become aware of one’s position within the ideological world (the social realm) which interweaves with notions traditionally considered to be economic such as trajectories, causes, and the reasons for policy change. Kanth says the solution to this problem lies with doing induction, but he seems to mean by this that good practice rests upon praxis, which is an involvement in the economy whilst improving our understanding of the economy. Praxis blends theory and practice in a continual spiral of transformative learning. In summary, cultural universalism goes hand in hand with empiricism and ethnocentricity. Collier puts this argument very well in suggesting that: The conception of rationality that I developed [here] is a conception of changing our emotions by virtue of increasing our knowledge of the objects of those emotions. (Collier, 1999, p. 85) One needs to revise one’s conception of longitudinal data analysis before being ready for a heterodox economic research practice. Second, the other limit to accuracy lies in the weaknesses of language in referring to things. The researcher must use metaphor and analogy in order to

Triangulation, time and econometrics

157

draw comparisons between what they have observed and what they expect the readers to be able to understand.The researcher sits in a communicative triangle which has been shown to be very problematic when one tries to get at the ‘truth’ or the true meaning of the social relations, social practices, or social tendencies – including causal mechanisms – in which people are involved. The communicative triangle includes at least the author, the reader, and the ‘things’ being referred to. The use of metaphor in social science non-fiction is just as inevitable as in fiction. By admitting that metaphors are being used, and by exploring the hermeneutic (meaning-rich) aspects of their own conceptual frameworks as well as the contrasting conceptual frameworks of the people in the economy, researchers can enrich and augment the agenda of economic research. Researchers can focus on both discourses about the economy (which is what economists use) and discourses in the economy (which are what practitioners of economic relations use).There are links between the former and the latter. If economists restrict themselves to the former, they are guilty of philosophical idealism. Idealism focuses upon the realm of ideas without touching or attending to the real world at all. The ideal world is simplified, stylised, and characterised by flat agents without conflicts or contradictions. By contrast the real world is characterised by overlapping, over-determining (hence indeterministic), complex, transfactually operating causal mechanisms. Among these mechanisms are the reasons people have for doing things and the meanings actions have for people. Amidst such social and psychological complexity, some econometric studies appear rather superficial. Realists were the first to specifically link reasons and causes, thus linking the socially constructed world with the pre-existing world of causes. Realists began to call the socially constructed part of the world ‘transitive’ and the pre-existing part of the world ‘intransitive’, with some ‘things’ having both transitive and intransitive properties. For instance, market efficiency is a feature that may pre-exist some studies of it, and thus shape what is found out about it, but market efficiency is also created in an ongoing way by the people who talk about it, assume it, aim for it, and construe it as something with specific meaning for them. Realists recognise a relative element in cognition of transitive things whilst also stressing that one need not know about things in order for them (especially the intransitive things) to exist. Market efficiency illustrates the latter possibility very well. Although distancing is necessarily part of the researcher’s work, both at the analysis stage and at the dissemination stage, a revised notion of truth leads us to appreciate the intentions and intended meanings (i.e. the hermeneutics) of the social science writers. This argument admittedly elides the distinction between fact and fiction, and in that particular sense offers an approach to economics that goes beyond the modern (Klamer, 2001). In Box 9.2 the distancing that is found in fiction writing is compared with the distancing and imperfection that necessarily accompanies economic report writing. This distancing was seen by Ricoeur as profoundly restricting the truth of claims. Ricoeur’s description of the limits to truth has much in

158

Wendy Olsen

Box 9.2 Distancing in social science report-writing Distancing in social science writing necessarily creates an appearance of falsehood and real inaccuracies of representation. Experts on writing argue that distancing takes several forms. A list will allow you to see how predominant these deviations from ‘correspondence’ or ‘accuracy’ are: 1 2

3

People use metaphors to allude to things they saw, imagine, or want to see emerge (Ricoeur, 1978). People use language to refer to things existing over time, or changing over time from time 1 to time 2. In doing so their meaning diverges inevitably from the reality itself.The hermeneutics of one’s intended meaning at time 2 are a different project from the unplucking of the reality that led to the creation of the thing at time 1. Things in the world have no intrinsic meaning, so people write about them in order to give them meaning. Intentionality is inherently part of the project of disseminating research findings.

Ricoeur illustrates the difficulty with the time aspect of the world in arguing that ‘Human action is nothing less than “being in time”, not merely in the sense of within-timeness, but rather in the sense of reckoning with past, present, and future… The world of the text is a way of being-in-the-world which fictionally works out various possibilities projected in a fictional situation’ (Vanhoozer, 1991, pp. 50–51, summarising Ricouer, 1984). Distanciation is inevitable in all social description, as Ricoeur put it. Fallibility of knowledge claims ineluctable results. For a review of fallibility versus objectivity, see Fay (1996, chapter 8). For Ricoeur, the expert hermeneuticist, every text necessarily has a fictional component and attempts to give meaning to the world it represents whilst also representing the future possibilities of that world. Ricoeur concluded that mis-representation was likely to be common in human communication, because the writer’s intended meanings are unlikely to be the same as that meaning which a distanced reader is going to take from their communications. For some scientists, these lessons from philosophy would lead to pessimism about the possibility of science achieving truth. However an alternative view is that the search for a single truth reflects a particular grand narrative (Bernstein, 1991) of empiricism which seeks validity above all other social values. Empiricism offers a monolithic epistemology: validity. The alternative to that grand narrative is a new one which judges statements by multiple criteria.This non-totalising approach is a very different grand narrative. Elliott (1999) illustrates how statistics research can be seen as fallible stories. Here we take the risk of judging, of speaking, of forming a judgemental view. The ‘always contestable crises and problems of the

Triangulation, time and econometrics

159

past’ are re-worked and re-assessed by rational, communicating, fallible interpreters in a social milieu (Bernstein, 1991, p. 122). In social science the narrator’s role is often to create a sense of security in the reader so that they trust the author. Coherence in the theories and arguments, and a claim of correspondence of variables to ‘things’ in the world, help to create this sense of solidity.Yet one would not want to exaggerate the capacity of variables to refer to complex real flows or relationships and to their meanings. Fallibility, errors, and omissions deserve to be mentioned for the sake of the reader’s critical faculty.

common with the realist approach to the need for truth-claims to lie within discursive (often scientific) communities. Sayer, for instance, stressed that whilst the scientific community will have one or more paradigms, and hence will frame results in terms understandable amongst that community, there is then a tension when scientists try to find things out from grassroots actors in society. The tension arises because their discursive communities do not share assumptions with the scientists. Distancing is inevitable; fallibility is the result. Therefore social scientists need further criteria (besides ‘truth’) for choosing among competing paradigms. These criteria might include ethical, communicative, or practical aims. As well as contextualising economic discourse, which involves a discussion of meaning, as implied earlier time-series data should be augmented via a triangulated research strategy. Triangulation is the combination of methods or techniques. In gathering data to make maps, three or more high-level vantage points are used to start building up a picture of the land lying in the low area lying between them. Measures of the angles from each vantage point to each thing in the space enable us to re-confirm the relative location of each thing in the space.The word triangulation also suggests the possibility that the view from each vantage point may be different. From one point, some items in the landscape may be masked (as in survey research when the respondent is nearly silenced). From another point, clouds move in and mask the view (as in structured interviewing if the researcher sticks too closely to an a priori theoretical or discursive framework. And from a third viewpoint, one might get a clear view of the first two viewpoints, and of the measuring instruments at those viewpoints. Triangulation may enable better measurement, but it will also reveal differences of interpretation and meaning. Discourse analysis can be used, ironically, to re-assess the meanings of questionnaires and the theoretical frameworks or social practices that lie behind them. Whilst discourse analysis arose in poststructuralist and even postmodern circles, focusing on cultural issues, it can also be turned upon the modern world.

160

Wendy Olsen

Triangulation may mean that one uses quantitative and qualitative data; or quantitative data and documentary analysis (Gilbert, 1992); or econometrics and life histories; or comparative analysis of country data along with case studies of industries.Triangulation involves a complex research design, usually with stages of research which may iterate (Danermark et al., 2002).

Epistemological criteria for heterodox economists There are substantive and methodological reasons for questioning neoclassical economic theory, but this section concentrates on the methodological implications of heterodoxy. In the literature, at least ten different epistemic criteria for good research or reasonable knowledge claims have been adduced.The list below cites the author associated with each of these criteria. ● ● ●

●

●

● ●

● ● ●

practical adequacy of the research design (Sayer, 1992); communicative validity of the research outputs (Kvale, 1996); metacritique, including theoretical triangulation and innovation (Bhaskar, 1993); rigorous statement of what/who/when the study represents (Bryman, 2002); exposing techniques to scrutiny (this is a processual criterion which follows from the Habermasian approach to conditions necessary for true statements to emerge; see Parfitt, 2002); using sophistication of technique for good purposes (ESRC, 1996); ethical relevance, incorporating policy relevance and critique of policy (Ray and Sayer, 1999); engagement with diverse discursive communities (Davis, 1999); self-conscious etic–emic balance (requires reflexivity) (May, 2000); hermeneutic depth, allowing differences of meaning (Outhwaite, 1987).

These criteria are epistemic in the sense that they provide a basis to judge how the study was done and how well it was written up/reported on. However taken on the whole as an epistemology, these criteria are pluralist (see Dow, in this volume, for discussion).They allow competing studies of different styles to report clashing findings. The ‘truth’ of the findings can only be discussed in a social context where the desire for good descriptions is contextualised, as discussed above, in specific ways. Practical adequacy, which is listed first, in a sense covers all the other criteria. Sayer has argued that theories and knowledge claims are assessed, practically speaking, for their ability to address pressing problems using frameworks that are recognisable whilst providing communicative bridges between discourses. For instance, a theory must be adequate to practical research whilst also serving the purposes of theoreticians, such as revealing something not previously known. If practical adequacy is actually the criterion used by many funding

Triangulation, time and econometrics

161

agencies and readers of non-fiction books, it is still not the criterion expressed by economists. Practical adequacy is contextually relative; it is contingent; it never provides universal or replicable validity. A transformative ontology is implicit in the practical adequacy criterion (see Sayer, 1992). This ontology is made explicit in related methodology literature (Layder, 1993) and is part of an ontic turn in social theory. Collier (1994) and Archer et al. (1998) explain further the transformative approach to ontology. Sayer’s initial impulse to describe quantitative methods as inherently limited was in part based on a double misunderstanding. First, he assumed that a statistical researcher would only use survey data. Second, he assumed they would engage with survey data using an empiricist epistemology. The criteria of this epistemology are usually value-neutrality, assuming falsely that a value-fact separation is possible (Smith, 1998, p. 76); atomism, hence reductionism; an ontological assumption that laws exist, hence a desire for nomothetic findings; an assumption that the variables measure the things they say they do (nominalism); an assumption that the survey type of data suffices (a variant on phenomenalism); and an assumption that social science must proceed in ways found in natural science. For example, experiments and representative large-scale surveys seeking laws (naturalism) (Smith, 1998, p. 76). To convert the above list into more concrete terms, let us apply it to wagedetermination models, which was the context of the example referred to earlier. Usually these consist of a regression equation with one group’s log wage as the dependent variable (e.g. Miller, 1987). Naive empiricism would argue that: the facts are value-neutral; the data on individuals is sufficient to explain wages; rewards to human capital follow a pattern which is a law under capitalist conditions; wages unproblematically measure the rewards for work over time and across people; wage data are sufficient to understand the terms and conditions of work; and finally (corresponding to the last bullet point above), making generalisations about wage-determination over space is desirable so large-scale surveys are intrinsically better than small-scale surveys.This chapter would question each of these assumptions although on pragmatic grounds the last three might be acceptable within specific stages of research. In general, as Smith (1998, chapter 3) has argued, it is relatively easy to avoid nominalism and phenomenalism, but once naturalism and value-neutrality are dropped, the lawlike nature of the desired outputs of research disappears.Then we have fallible claims, which should be hedged with a statement of when, where and how they originated. Reflexive researchers can try to identify their own valueorientations. Harding (1995) has advocated in this context a strategic realism (referring to objects of study when these are mutually agreed to exist, given specific strategic purposes; Harding, 1999). Harding thus advocates a stronger objectivity than what is usually presumed to be possible: making one’s prejudices visible.Whilst controversial, her suggestions follow the same line of argument as this chapter. Triangulation enables researchers to create portfolios, examine symbolic imagery and competing discourses; and study how things are construed as well

162

Wendy Olsen Box 9.3 Realist empirical assumptions ● ● ● ● ● ● ●

Integration of values with scientific discovery Multi-level non-nested structured reality Assumption of an open system Pragmatic nominalism Acts of representation can be made conscious Transitivity of objects Differentiated objects in society.

as how they are measured. Realists would then examine the data with several assumptions in mind, which are summarised in Box 9.3. (The authors who have advocated each of these in particular are Harding, 1999; Layder, 1993; Lawson, 1997; Lacey, 1999; May and Williams, 1996; Outhwaite, 1987; and Layder, 1998, respectively.) The realist points do not add up to a monolithic epistemology.Validity will always be contested, and realists accept that. Labour markets, for instance, can be seen in radically different ways from the model seen earlier (e.g. Hutchinson et al., 2002; Fleetwood, 1999 and this volume; or Kalpagam, 1994).

The time–space extension of ‘things’ The objects studied by time-series analysis in economics include many flows such as consumption, income, output and wages, which are measured in money units. Then economists agonise over how to convert nominal to real values over time or how to convert from one currency to another.Too little attention is paid to what is intrinsically being represented by these variables. Whilst the variables’ definitions are unchanging, the underlying reality of what they consist of and what they mean is changing. In addition, time-series data also cover the levels of stocks (such as capital and savings), technology indicators, flows of non-money items (e.g. the number of transactions on the stock market), and ratios such as interest rates and the rate of investment relative to output. Measurement issues are not the target of the critique here. Instead, working through several examples, this chapter shows that the underlying things are not staying constant in their nature over time. Demand Demand, measured by money value of a flow of purchases in specific nationalaccounts categories, is continually changing in its composition.The investment component may have a growing ‘knowledge-based’ component rather than the cost of physical capital. The consumption component has a changing nature: sometimes survival consumption; sometimes luxury consumption for

Triangulation, time and econometrics

163

conspicuously status-raising purposes; and sometimes with spending targeted on buying designer labels rather than merely obtaining the product in its utilityproviding generic state. Money demand has been critiqued, too, for being ambiguous in the treatment of demand backed by credit (as opposed to demand backed by cash); underneath demand is a whole series of liquidity issues which are unstated when the gross flow of money for given products is gauged. Supply Supply, measured by money flow of a series of products listed in given categories of the formal accounts of either firms or the country, includes a changing set of things whose boundaries have been proved to be permeable. For instance, supply of caring labour has in recent decades ‘changed’ to include the costs of childcare for below-school-age children in many countries where the percentage in nurseries has risen. The food-production labour component in the supply of lunches and suppers often goes uncounted, until/unless the use of restaurants, fast-food, and office lunch services places what was previously uncounted home labour with waged formal labour services. Supply of food to farms which eat their own produce is routinely imputed based upon the market cost at which they would have had to buy the food even though they didn’t buy it. Finally, within firms, the cooperative work of Quality Circles and the savings generated by conflict resolution through union mediation are ‘valued’ in supply terms at only the market value of produce sold. People on low wages are not represented as human resources specialists although their interpersonal skills at work often overlap with much more highly paid consultancy-based types of work.The most basic descriptive statistics, GNP and firms’ production, do not represent the reality of what goes into production, but rather take a quick black-and-white picture, with some touching-up in the form of imputation, when lived experience is much richer and rapidly changing. Prices Prices are assumed to refer to a specific bundle of characteristics routinely passed over in exchange for money; these characteristics are nominally attached to a good or service but are (arguably) subjectively felt by the buyer and seller to be rather different bundles. If the subjectivity of buyer and seller don’t matter, because the good’s nature is determined elsewhere and the price is taken rather than as being up for discussion, then the measurement of prices ‘works’. However it takes no account of the discussion and subjective views of people about the price level and what it represents. Obviously there is space for hermeneutic difference and misrepresentations (or multiple simultaneous interpretations) in the social places where prices matter. However, what often matters is the bundling of the good or service, not its price per se. Institutionalists have specialised in giving depth to economic analysis of transactions but the old institutionalists have a special strength over the new institutionalists: they

164

Wendy Olsen

do not assume that there are always rational or cognitive reasons for people setting prices onto bundles of goods.The whole process, with all its informationgaps and social construction of the nature of goods and services, is under the institutionalist’s microscope. Interest rates Interest rates are taken to be the price of capital over time, but they were shown in the capital debates of the 1960s and 1970s to be inconsistent and incoherent indicators both in theoretical terms and in empirical terms. Models that assume interest rates do reflect the return on capital to investors must also assume that investors all look at the money return on capital.There may in addition be social reasons (e.g. status, aggressive control of suppliers, expansion over space) for wanting to make specific investments. Models that assume interest rates reflect the price of borrowing money ignore crucial sources of finance that are mainly non-market: family finance and reinvestment of funds by firms. Much research on interest rates is done without being able to match the stipulated concept of interest rate to any specific, local, measurable rate. In fact, interest rates in practice are multiple even for similar types of money transaction, and depend to a surprising degree on the nature of the social relations between the transactors or on their resources (Bhaduri, 1977). Interest rates bear social information but not in their quantum. Instead, an excellent method would be to use retroduction to work backward from observed interest rates to what caused them, cognitively or socially, to be set so. Consumption Consumption, like demand, consists of flows which are construed in ways that constitute complex social realities, notably styles, individuality, innovation, designs, curiosity, and routines. Consumption has been shown to have three possible connotations: it may be used to diversify one’s life and to express difference; it may be used to achieve social status, which requires exploratory consumption as a first step to find out what practices do increase one’s status; and it may be used in a utilitarian way to service needs such as hunger, desire or the wish to change something (Warde et al., 1999). The social construction of consumption is not outside economics, but at present the methodologies for studying it are seen as intrinsically non-economic methodologies. It is possible to use statistical reasoning in innovative ways to explore consumption (Olsen et al., 2000), for example, using attitude questions or making a scale of diversified use of different types of restaurant. However, it is best to augment the survey data with another specific research technique such as interviews or discourse analysis. One can list numerous other problematic categories in econometric studies, such as savings, risk, transactions costs, and the household. Each has been roundly criticised by outsiders to economics and then taken up for rich qualitative analysis in nearby disciplines like management science, feminist

Triangulation, time and econometrics

165

sociology, and anthropology. However, it is time for economists to become expert in their own substantive field. The overview provided above, with five problematic central categories, illustrates several aspects of realist reasoning in economics. The chapter now concludes with a more explicit consideration of realism.

Realists’ alternative conception of things Most of the ‘things’ mentioned above are in part transitively constructed.They therefore tend to be changing over time and not constant over space and across different social groups. Most economic theories at present use these concepts for their conceptual utility or in other words as ideal types rather than as representations of what really exists. An alternative approach is possible. One can examine the actual social construction of each ‘thing’ and give evidence for its time–space extension. Economists should not so quickly stipulate their own preferred definition: they should problematise the act of defining. Under specific conditions it is then empirically reasonable to use time-series econometrics or at least large-scale survey data sets, but it will be desirable to combine these with other sources before interpreting the findings. An illustration of this lesson can once again be drawn from analysis of wage rates, here considered as a representation of the rewards for working. Formalsector employment offers a wage, bonuses, pension, fixed, or flexitime hours, and a variety of foods, car, phone, and other perks for specific employees. Formal-sector self-employment offers a different set of rewards for which wages are likely to be a poor representation. However, even in employment per se, workers get unquantified advantages like their social networking from work; their ability to change their hours from time to time; their ease of access to work; and their ability to hope for future pay rises even if today’s wages are lower than desired. The current literature summarises these lacuna as ‘unobserved heterogeneity’ of either the individual or the job. This label makes no concession to the socially constructed and subjective nature of the rewards perceived. In management science, the study of perceived job satisfaction has become a hugely successful triangulated research programme. How is it that so little of this work has been done within Economics departments? One reason is that wages, for orthodox economists, mainly reflect individual deserts. However, in reality, wages are an emergent property of bargaining at national, occupational, sectoral, household, and firm levels whilst also interacting with individual characteristics. Wages have both subjective and objective aspects; realists would say they exist both transitively and intransitively. Sensitivity to this complexity takes us outside neoclassical theory immediately, and it makes us sensitive to the weaknesses of wage-rate data. The example illustrates the complexity of transitivity: realists are not essentialists and they try to avoid doing unjustified reification. Most realists would say that theory should be built up on the assumption that structures exist both in nature and in society (summarised by Hay, 1995).They define a structure as

166

Wendy Olsen

a set of relations that link things, such that the whole has emergent properties which cannot be reduced to the parts (i.e. the ‘things’). Since realists study the relations between things as well as the things, they see causal mechanisms as complexly sited within a changing set of social relations. They would never expect to see a law instanced repeatedly in a replicable or reliable way except under experimentally controlled conditions. Other aspects of realism have been listed elsewhere, but the changing-ness or ongoing-ness of things and relations makes it clear that labelling of things is not easily going to qualify as correct or valid. Labelling involves the use of a metaphor to capture an essence which is stipulated (often wrongly, as in a logical-positivist or philosophical idealised simplified argument) to be constant over time. Realists argue about whether structures are in some cases enduring enough to act as if laws did exist. If they act that way, and if enduring structures cause repeatedly observable outcomes, then we can calculate probabilities which are both meaningful and likely to function as if deterministic causation prevailed. However, in a dialectical way, even the most enduring structure in society is always subject to change.Therefore it is wrong to declare that in general laws of the movement of society underpin the society. Instead, it is right to argue that the search for laws is a sign of modern Enlightenment thinking and/or for the author’s authority to call her/himself a discoverer. If we want to be discoverers it might be better to discover the new and to recognise our own use of metaphor in communicating about that discovery. Far worse to claim that language unambiguously reflects reality (naive realism) or that mathematical models can easily capture and correspond to reality (neoclassical econometric empiricist epistemology).

Conclusions In summary, this chapter has challenged naive empiricism and then described specific ways that a realist might use triangulation to augment and complement econometric analysis. Little has been said about the theoretical assumptions of econometric models, which usually take the form of stylised facts or assumed laws, because the chapter focuses more on the real world referred to by longitudinally measured variables.Ten facets of a complex, pluralist epistemology were listed.These could best form the basis of time-series analysis if the econometrics were integrated with other research techniques.The integrated approach to triangulation suggested here will tend to make cross-sectional data look just as tantalizing as time-series data. The violence done to reality when time-series data are constructed imply that the data have weaknesses just where they are needed most: at times and places of social change.Triangulation offers the possibility of making the changes the subject of empirical enquiry rather than of speculation.

References Archer, Margaret, Roy Bhaskar, Andrew Collier, Tony Lawson, and Alan Norrie (eds) (1998) Critical Realism: Essential Readings, London: Routledge.

Triangulation, time and econometrics

167

Bernstein, J.M. (1991) ‘Grand Narratives’, chapter 7 in Wood (ed.) (1991). Bhaduri, Amit (1977) ‘On the Formation of Usurious Interest Rates in Backward Agriculture’, Cambridge Journal of Economics, 1, pp. 341–352. Bhaskar, Roy (1993) Dialectic: The Pulse of Freedom, London: Verso; also extracted as ch. 23, pp. 575–588 of Archer, et al., 1998. Boylan, Thomas A. and Paschal F. O’Gorman (1995) Beyond Rhetoric and Realism in Economics:Toward a Reformulation of Economic Methodology, London: Routledge. Bryman, Alan (2002) Social Research Methods, London: Oxford University Press. Byrne, David (1998) Complexity Theory and the Social Sciences: An Introduction, London: Routledge. Collier, Andrew (1994) Critical Realism: An Introduction to Roy Bhaskar’s Philosophy, London: Verso. —— (1999) Being and Worth, London: Routledge. Cullenberg, Stephen, Jack Amariglio, and David F. Ruccio (eds) (2001) Postmodernism, Economics and Knowledge, London: Routledge. Danermark, Berth, Mats Ekstrom, Liselotte Jakobsen, and Jan Ch. Karlsson, (2002; 1st published 1997 in Swedish language) Explaining Society: Critical Realism in the Social Sciences, London: Routledge. Davis, John B. (1999) ‘Postmodernism and Identity Conditions for Discourses’, chapter 10 in Garnett, R.F., Jr., (ed.) What Do Economists Know? New Economics of Knowledge, London: Routledge. Elliott, Jane (1999) ‘Models Are Stories Are Not Real Life’ (sic), chapter 13 in Statistics in Society: The Arithmetic of Politics (eds) Daniel Dorling and Stephen Simpson, London: Arnold (Series: Arnold Applications of Statistics). ESRC, Economic and Social Research Council (1996) PhD Research Training Guidelines, Polaris House, Swindon SN2 1UJ. Fay, B. (1996) Contemporary Philosophy of Social Science: A Multicultural Approach, Blackwell: Oxford. Fleetwood, Steve (1999) ‘Conceptualizing Unemployment in a Period of Atypical Employment: A Critical Realist Perspective’, Review of Social Economy, March, 59(1). Fleetwood, Steve (ed.) (1999) Critical Realism in Economics: Development and Debate. London: Routledge. Gilbert, Nigel (ed.) (1992) (2nd edn, 2001), Researching Social Life, London: Sage. Granger, C.W.J. (ed.) (1990) Modelling Economic Series: Readings in Econometric Methodology, Oxford: Clarendon Press. Harding, Sandra (1995) ‘Can Feminist Thought Make Economics More Objective?’, Feminist Economics, 1(1), pp. 7–32. —— (1999) ‘The Case for Strategic Realism: A Response to Lawson’, Feminist Economics, 5(3), pp. 127–133. Hay, C. (1995) ‘Structure and Agency’, in Stoker, G. and Marsh, D. (eds) Theory and Methods in Political Science. Basingstoke: Macmillan Press Ltd. pp. 192–206. Hendry, D.F. and J. Richard (1990) ‘On the Formulation of Empirical Models in Dynamic Econometrics’, chapter 14 in Granger (ed.). Hutchinson, Frances G., Mary Mellor, and Wendy Olsen (2002) The Politics of Money: Toward Sustainability and Economic Democracy, London: Pluto. Kalpagam, U. (1994) Labour and Gender: Survival in Urban India, Sage, Delhi and London and Thousand Oaks. Kanth, Ranjani (1999) ‘Against Eurocentred Epistemologies: A Critique of Science, Realism and Economics’, chapter 11 in Fleetwood (ed.).

168

Wendy Olsen

Klamer, Arjo (2001) ‘Late Modernism and the Loss of Character in Economics’, chapter 3 in Cullenberg, Stephen, Jack Amariglio, and David F. Ruccio (eds) Postmodernism, Economics and Knowledge, London: Routledge. Kvale, Steiner (1996) InterViews, London: Sage. Lacey, Hugh (1999) Is Science Value Free? Values and Scientific Understanding, London: Routledge. Lawson,Tony (1997) Economics and Reality, Routledge: London. —— (1999) ‘On Universalism and Essentialism in Feminist Economics’, Feminist Economics, 5(2), pp. 25–59. Layder, D. (1993) New Strategies in Social Research (Repr. 1995, 1996). Oxford: Polity Press. —— (1998) ‘The Reality of Social Domains: Implications for Theory and Method’, chapter 6 in May and Williams (eds) (1998), pp. 86–102. May, Tim (2000) ‘A Future for Critique? Positioning, Belonging, and Reflexivity’, European Journal of Social Theory, 3(2), pp. 157–173. —— and Malcolm Williams (1996) (eds), Knowing the Social World, Buckingham: Open University Press. McGovern, Siobhan (2002) ‘Financial Market Efficiency and the Quine Paradox’ Conference paper for the Association for Heterodox Economics, Dublin, July. Miller, Paul W. (1987) ‘The wage effect of the occupational segregation of women in Britain’, The Economic Journal, 97, pp. 885–896. Olsen,W.K.,Warde,A. and Martens, L. (2000) ‘Social Differentiation and the Market for Eating Out in the UK’, International Journal of Hospitality Management, 19(2), pp. 173–190. Outhwaite, William (1987) New Philosophies of Social Science: Realism, Hermeneutics and Critical Theory, London: Macmillan. Parfitt,Trevor (2002) The End of Development: Modernity, Post-Modernity and Development, London: Pluto Press. Ray, Larry and Andrew Sayer (1999) Culture and Economy after the Cultural Turn, London: Sage. Ricoeur, Paul (1978) The Rule of Metaphor, London: Routledge & Kegan Paul. —— (1981) Hermeneutics and the Human Sciences, (ed.) John B.Thompson, Cambridge: Cambridge University Press. —— (1984) Time and Narrative, vol. I, Chicago: University of Chicago Press. Rubery, Jill, Smith, Mark and Colette Fagan (1999) Women’s Employment in Europe: Trends and Prospects, London: Routledge. Sayer, Andrew (1992), 2nd edn Method in Social Science: A Realist Approach, London: Routledge. Smith, Mark (1998) Social Science in Question, Sage in association with Open University, London. Tijdens, Kea G. (2002) ‘Gender Roles and Labour Use Strategies: Women’s Part-time Work in the European Union’, Feminist Economics, 8(1), March. Vanhoozer, Kevin J. (1991) ‘Philosophical Antecedents to Ricoeur’s Time and Narrative’, chapter 3 in Wood (ed.) (1991). Walby, Sylvia and Wendy Olsen (2002) The Impact Of Women’s Position In The Labour Market On Pay And Implications For UK Productivity, Cabinet Office Women and Equality Unit, TSO, Department of Trade and Industry. London: HMSO (web access to pdf file 2003 via www.womenandequalityunit.gov.uk).

Triangulation, time and econometrics

169

Warde, A., Martens, L., and Olsen, W.K. (1999) ‘Consumption and the Problem of Variety: Cultural Omnivorousness, Social Distinction and Dining Out’, Sociology, 30(1), 105–128. Wood, David (ed.) (1991) On Paul Ricoeur: Narrative and Interpretation, London: Routledge (Warwick Studies in Philosophy and Literature).

10 Theory foundation and the methodological foundations of Post Keynesian economics Frederic S. Lee

Introduction Since 1971, Post Keynesians have worked at developing an alternative to neoclassical economic theory. One facet of the project has been, over the last fifteen years, extensive discussions about the philosophical and methodological foundations of Post Keynesian economics. Arising from the discussions is a consensus among Post Keynesians that their economics has philosophical and methodological foundations that are different from the positivism, empirical realism, and deductivism foundations underpinning neoclassical economics. The philosophical foundation includes realism and critical realism and its position on knowledge, epistemological relativism.While very important, it is not the focus of this chapter and therefore only briefly delineated in the following section. On the other hand, the methodological foundation, which refers to the methodological guidelines utilized for creating and developing Post Keynesian theory has received little discussion and development beyond the brief and undeveloped references to retroduction made by critical realists, to holism and pattern modeling made by Institutionalists, and to the ‘Babylonian’ method expounded by Sheila Dow. Hence the aim of this chapter is twofold: first to advocate that the method of grounded theory provides the best set of guidelines for theory creation; and secondly to delineate the method and the kind of economic theory that emerges when it is integrated with realism and critical realism. Therefore, in the second section the method of grounded theory is delineated, which is followed in the third section by a discussion of four methodological issues – the nature of data, role of case studies, mathematics and models, and econometrics – as they relate to critical realism and the grounded theory method. The final section concludes the chapter with a discussion of the historical nature of grounded economic theories and their implications for the Post Keynesian economics.

Philosophical foundation Being both participants in and observers of the social and economic activity around them, Post Keynesian economists approach their study of economics with a common sense understanding of the world. By common sense, it is

Foundations of Post Keynesian economics

171

meant a complex set of beliefs and propositions (many of which are historically grounded) about fundamental features of the world that individuals assume in whatever they do in ordinary life.Thus, for Post Keynesians, they take particular features, characteristics, institutions, and human actors of economic activity as real, obvious, and practical initial starting points for further research.To be real, obvious, and practical means that various features, institutions, and actors exist, are ingrained everyday properties of the world of economic activity, and are encountered when observing or participating in ongoing economic activity. On the one hand, the Post Keynesian qua economist can as an observer see them in action in the economy carry out their activities; or they can directly experience them as participants in economic activity. By being a participantobserver, Post Keynesians are able to be close to the real, concrete form of the economy. Consequently their common sense beliefs and propositions provide the background against which they carry out their research. Hence, this common sense understanding of economic activity informs the methods which Post Keynesians actually use to examine economic activity, particularly with regard to the way it is explained (Maki, 1989, 1996, 1998a,b; Dow, 1990a, 1999; Coates, 1996; Comim, 1997). Post Keynesians characterize their common sense propositions by stating that the real (actual) economy is a non-ergodic, independent system with human agency and economic-social-political structures and institutions embedded in an historical process. Other propositions accepted and articulated by Post Keynesians which support and clarify this include the views that the actual economy and the society in which it is embedded is real and exists independently of the Post Keynesian economist; that the economy is transmutable, hence its future is unknowable and what is true about the economy today may not be true tomorrow; that economic change comes about through human action interacting with social, political, and economic structures and institutions; that human action is derived in part from ethical, cultural, political, and ideological beliefs so that economic outcomes are also ethical and political outcomes as well; and that a capitalist society is a class society and the economy is permeated with hierarchical power derived in part from it.The final common sense proposition is that the study of particular economic activity cannot be done independently of the whole economy or from the social system in which it is embedded. Mutually shared among Post Keynesians, these common sense propositions do not constitute Post Keynesian economics, but rather provide the basis for its ontological realism foundation (Wilber and Harrison, 1978; Gruchy, 1987; Lawson, 1994; Arestis, 1996; Davidson, 1996; Dow, 1999; Downward, 1999; Rotheim, 1999). From the common sense propositions, Post Keynesians conclude that the economy works in terms of causal-historical processes. Moreover, because they accept the ontological constraint implicit in this, Post Keynesians embrace a specific form of realism, critical realism, as the ontological basis of Post Keynesian economics.1 Not only do they posit that economic phenomenon are real, Post Keynesians also argue that their explanations or theories refer to

172

Frederic S. Lee

real things, represent real entities, are judged good or bad, true or false by virtue of the way the economy works, and are causal explanations. As a causal explanation, theory provides an account of the process as a sequence of economic events and depicts the causes that propel one event to another in a sequence. In addition, while accepting a correspondence theory of truth with regard to evaluating theories, they also accept epistemological relativism, which is that knowledge of economic events is historically contingent, and integrates the two. Finally, to ensure that their theories are causal explanations of real things, Post Keynesians adopt the methodological guideline of the method of grounded theory (Maki, 1989, 1992a, 1996, 1998a,b, 2001; Ellis, 1985). Critical realism Critical realism starts with an account of what the economic world must be like before it is investigated by economists and for economic analysis to be possible. Thus it begins with four propositions: that the economic world consists of events which are structured in that they consist of something more than simple empirical experience; that economic events are intransitive in that they exist and occur independently of their identification; that all economic events, whether reoccurring or not, are produced by an underlying diverse set of causal mechanisms and structures; and that the economic world is open in that each and all economic events are a result of interacting and counteracting structures and contingently related causal mechanisms. Consequently, Post Keynesians accept a three-tier view of economic reality.The first two tiers are the empirical events of experience and impression and the actual events underlying them. Understanding the former depends on the explanations of the actual events and that is derived from causal mechanism(s) and economic structures, which constitute the third tier of economic reality.The causal mechanisms and structures together are the ontological core of Post Keynesian economics in that when they are identified and understood, the empirical and actual events are jointly understood. Thus for the Post Keynesian economist, identifying structures and causal mechanisms and describing their way of influencing or acting on specific events in the economic world is their scientific undertaking. A causal mechanism in the context of Post Keynesian economics is irreducible, has a relatively constant internal organization whose components are intentionally not mechanistically related, is real, observable, and underlies, hence governs or produces actual events, and acts transfactually (that is acts and has effects even when it does not generate discernable actual events).2 Being irreducible means that the form and organization cannot be disaggregated into its constituent components and still function as a causal mechanism. In this sense, a causal mechanism is an emergent entity in that its properties and powers cannot be completely traced to its individual components. To have a constant form and organization means that the mechanism can be empirically identified by stable patterns of behavior and organizational format and hence empirically observed and delineated. Furthermore, the ability to act means that

Foundations of Post Keynesian economics

173

the mechanism has the power to generate qualitative and/or quantitative outcomes; and the triggering of the mechanism comes from human intentionality. Thus economic actors have independent power to initiate actions, therefore setting in motion causal mechanisms which generate outcomes that underlie and hence govern actual economic events. Because the causal mechanism utilizes the same processes when producing results, the same results are repeatedly produced.3 So to say that a causal mechanism acts transfactually producing the same results is also to say that its form and internal organization are constant thereby making it a relatively enduring entity. Hence, if the same causal mechanism operates in different situations, it will produce the same, or transfactual, results each time it is in operation; but the empirical and actual events need not be regular or repeatable, as other contingently related causal mechanisms will be affecting them. Consequently causal mechanisms only have the tendency or possibility of producing regular, repeatable qualitative or quantitative actual economic events denoted as demi-regularities. Structure is different from causal mechanism in that it does not include human agency, hence it can only help shape or govern the actual event. Otherwise it is similar to a causal mechanism in that it is real, observable, relatively enduring in form and organization, irreducible, and governs transfactually. The structures of an economy have two additional properties: (1) being sustained, reproduced, and slowly transformed by economic and social events that are caused by human action through their causal mechanisms and (2) its form and organization have a historical character. Moreover, all economic structures are social structures in that they represent and delineate recurrent and pattern interactions between economic agents or between economic agents and technology and natural resources. Economic structures include economic and social norms, practices and conventions, social networks such as associational networks or interlocking directorates, technological networks such as the production and cost structures of a business enterprise or the input-output structure of an economy, and economic, political, and social institutions such as markets or the legal system. As distinct entities, neither causal mechanisms nor structures can separately cause and govern actual economic events. Rather they must work jointly where the structures provide the medium or the conditions through which causal mechanisms via human agency act.Thus, as long as they remain enduring, there will be a tendency for regular and repeatable actual economic events to occur (Maki, 1989, 1998b; Lovering, 1990; Kanth, 1992; Sayer, 1992; Lloyd, 1993; Lawson, 1994, 1997a,b, 1998a,b,c; Ingham, 1996; Lawson et al., 1996; Wellman and Berkowitz, 1997; Fleetwood, 2001a,b, Chapter 3 of this volume; Hodgson, 1998a,b; Joseph, 1998; Dow, 1999; Downward, 1999; Rotheim, 1999). Epistemological relativism Epistemological relativism is the view that knowledge of economic events is historically contingent. That is, because the social and economic activities of

174

Frederic S. Lee

interest to Post Keynesian economists change over time, knowledge and understanding of them is historically contingent; hence there are no eternal ‘truths’ and knowledge is always in the process of being created. Consequently, what is known about actual economic events of the past need not be knowledge about current or future economic events. As a result, Post Keynesians are continually engaged in creating new knowledge, new explanations to take the place of those that cease to refer to real things, represent real entities, and explain actual economic events. Thus explanations or theories are historically conditioned hence historically contingent, which implies that, for Post Keynesians, there are no ahistorical economic laws or regularities. Moreover, it is not possible to make ahistorical, general statements with absolute certainty beyond the historical data and context in which the statements are embedded. Another implication is that theories must be, in some sense, grounded in historical data in order to tell historical stories explaining historical economic events.A third implication is that the difference between good and not-so-good, true and false theories is how well their explanations correspond to the historically contingent economic events being explained. Finally, epistemological relativism implies that the continual creation of knowledge is a social act carried out by informed actors, that is by Post Keynesian economists, in a socially, historically contingent context. (Sayer, 1992; Lawson, 1997a; Pratt, 1995;Yeung, 1997)

Methodological foundation: the method of grounded theory To develop a theory that analytically explains causally related, historically contingent economic events, the critical realist Post Keynesian needs to identify and delineate the structures, causal mechanisms, and causal processes producing them. The methodological guideline for creating causally explanatory theories that is also consistent with realism, critical realism, and epistemological relativism is the method of grounded theory.4 The method of grounded theory can be described as a process in which researchers, or more specifically economists, create their theory ‘directly’ developed from data (which are not the same as the ‘objective facts’ of the empiricist); and which data collection, theoretical analysis, and theory building proceed simultaneously (see Figure 10.1).5 The use of the method begins with the economist becoming familiar with, but not dogmatically committed to, the relevant theoretical, empirical, and historical literature that might assist in approaching, understanding, and evaluating the data relevant to his research interest. Then, the economists engages in ‘field work’ by collecting comparable data from economic events from which a number of specific categories or analytical concepts and their associated properties are isolated and the relationships between them identified. With the concepts and relationships empirically grounded in detail, the economist then develops a theory in the form of a complex analytical explanation based on the data’s core concepts.An essential property of the theory is that it explains why and how the sequence of economic events represented in the data took place.

Foundations of Post Keynesian economics

175

Hence, the economist does not attempt to construct a simplified or realistically deformed empirically grounded theory by ignoring or rejecting particular data. Rather the economist endeavors to capture the complexity of the data by empirically establishing many different secondary concepts and relationships and weaving them together with the core concept into structures and causal mechanisms.This ensures that the resulting theory is conceptually dense as well as having causal explanatory power.The process of selecting the central concepts and developing the theory brings to light secondary concepts and relationships which also need further empirical grounding as well as suggesting purely analytical concepts and relationships which need empirical grounding if they are to be integrated into the theory.After the theory is developed, the economist will evaluate it by seeing how it explains actual economic events. Let us now consider aspects of the grounded theory method in more detail. First, the collection of data is a complex task that involves collecting the data itself, that is counting up pieces of data, as well as constantly comparing, analyzing, and interpreting the data collected while simultaneously organizing it into conceptual or generalized categories. The categories that emerge come

Preexisting ideas and concepts | | +------Data collected with constant comparisons----+ |

|

|

|

|

|

| Conceptual categories identified from the data | |

|

|

|

|

|

|

Core categories identified

|

|

|

|

| |

| Core categories developed----------------+

| |

|

|

|

|

|

Substantive theory/basic social process

|

|

|

|

+-------------------------Formal theory

Figure 10.1 Schema of the grounded theory method.

176 Frederic S. Lee from the data itself, not after it is all collected, but in the process of collecting it.6 Consequently each category is tied to or empirically grounded in its data; and since the data is real, observable, so is the category.7 Moreover, since the data lies in time and history, each category is anchored in a particular historical setting. In addition, the purpose of constant comparison of the data is to see if it supports and continues to support emerging categories.8 Thus, each category that becomes established will have been repeatedly present in very many comparable pieces of data derived from multisources.9 In this way individual pieces of data that would not be significant on their own obtain a collective significance. The categories that emerge are of two types: one that is derived directly from the data and the other that is formulated by the economist.The former tends to denote data self-description and actual processes and behavior while the latter tend to denote explanations.10 In addition, each category will have properties also derived from data in the same manner, that is using constant comparisons. The more properties a category has the denser and hence the more realistic it is. A grounded theory category does not ignore the complexity of reality; rather it embraces it. In the process of collecting data, the economist may feel that what is being collected is not revealing additional properties of a specific kind that he believes, due to his familiarity with the relevant theoretical, empirical, and historical literature, might exist.As a result, he will engage in theoretical sampling.This involves sampling or collecting data that is expected to increase the density of a specific category by producing more properties as well as increasing the number of pieces of data supporting each of the properties.11 Theoretical sampling and collection of data for a single category as well as for a range of categories continues until theoretical saturation is reached, that is when no new data regarding a category and the relationships between the categories continue to emerge.12 The significance of this empirical grounding process is that the categories cannot be unrealistic hence false since they are derived from the data. If the data collection and theoretical sampling is incomplete then the categories will not be adequately dense, as relevant properties will be missing; thus such categories will be incompletely realistic. On the other hand, if future data emerges which the empirical grounding process shows does not fall into a previously existing category, then that category is not relevant, but it is not empirically false. Once the real, observable categories are delineated and grounded, the economist, perceiving a pattern of relationships among them, will classify some directly as economic structures and others as components of economic structures. In addition, other categories centered on human motivation and action and a set of outcomes will be woven together into a causal mechanism. The resulting structures and causal mechanisms will be real, observable as opposed to unreal, metaphoric, and hidden.That is, to observe a structure or causal mechanism is to observe the working together of its observed concrete components, including the human actions involved, much as a family is observed through the interaction of its members. Hence structures and causal mechanisms are real, observable precisely because their categories are real and observable.

Foundations of Post Keynesian economics

177

Given their research interest, the economist will select from the causal mechanisms identified, one as the central causal mechanism around which the structures and secondary causal mechanisms with their outcomes are arranged. Criteria for selecting the central causal mechanism from among a number of possible causal mechanisms include that it appears frequently in the data as a cause of the outcomes; that it has clear implications for a more general theory; and that it allows for complexity. Thus the causal mechanism is central to the narrative to be analytically developed in conjunction with the economic structures and secondary causal mechanisms. More specifically, the narrative is not a description of present or a recounting of past unique and/or demi-regular economic events, although both techniques of presenting empirical and actual economic events are included in the narrative. Rather, it is a complex analytical explanation of those described or recounted events. Even though the basic narrative is decided upon, its development will involve further theoretical sampling and collecting of data as new properties for the existing structures and causal mechanisms emerge. Consequently, the narrative evolves into an economic theory while at the same time becoming increasingly denser (in terms of properties and empirical grounding) as well as increasingly complex. The complexity arises because of the variations in the categories and in the properties of the categories that make up the theory.The grounded economic theory that eventually emerges is a complex analytical explanation or interpretation of the actual economic events represented in the data. Thus the theory is not a generalization from the data, but of the data; that is, a grounded theory does not go beyond the data on which it is based – it does not claim universality or the status of an empirical-theoretical law.13 Being a weave of a central causal mechanism, secondary causal mechanisms, and economic structures designed to explain actual economic events in historical time, the theory also consists of descriptively realistic (as opposed to stylized or fictionalized) descriptions of economic events and accurate narratives of sequences of economic events. As a result, the grounded economic theory is an emergent entity, a concatenated theory which cannot be disassembled into separate parts. Hence the question of logical coherence of a deductivist kind cannot be applied to a grounded theory; instead the coherence of the theory is judged on how well its explanation corresponds to the actual historically contingent economic events.14 Economic theory centered on a single central causal mechanism is classified as a substantive economic theory since it is an explanation of a single basic economic process that occurs widely in the economy. From a number of substantive theories, a formal economic theory can be developed into a general or holistic theory where the relationship or pattern among the substantive theories is its analytical explanation.15 Like in the process of grounding the substantive economic theory, the formal theory also has to be grounded. In particular, the relationships between the substantive theories that constitute the formal theory need to be grounded in data assisted and directed by theoretical sampling. Consequently, the formal economic theory is grounded, historically contingent, and its analytical explanations are not empirical extrapolations.

178

Frederic S. Lee

As the economic world is not static, a formal theory is never complete, but undergoes continual modification with ever newer data relating to newly emerging patterns or configurations of economic reality. There are two aspects of the grounded theory method that need further delineation. The first deals with the role of preexisting ideas, concepts, and categories. To fruitfully use the method, the Post Keynesian economist must become familiar with the contemporary theoretical and nontheoretical literature, the controversies between economists, and the relevant literature from the history of economic thought. In particular, the economist needs to make a detailed and critical investigation of the preexisting Post Keynesian ideas and concepts to see which lend themselves to empirical grounding.The economist also needs to be familiar with some of the empirical literature as well as with the relevant literature from economic history. By acquiring a critical awareness of the preexisting economic theories and empirical findings, the economist acquires a theoretical sensitivity regarding the data and theoretical concepts being examined, compared, and empirically grounded. As a result, the economist will have the ability to recognize what might be important in the data and to give it meaning as well as recognizing when the data does not support a preexisting concept or category, requires a large or small transformation of the preexisting concept or category, or ‘produces’ a new category. Thus, the grounded theory method recognizes that observations, data, and descriptions are independent of the theory being constructed and demands that all economists enter into theory creation as theoretically knowledgeable and aware individuals, as well as with the conviction that the creation of a new substantive economic theory will most likely require them to set aside forever some of that acquired knowledge.16 By acknowledging the independence of observations from the theory being constructed, while at the same time demanding that the economist be skeptical of all preexisting theory, the grounded theory method is a highly self-conscious, engaging and open-minded approach to economic research, data creation–collection, and theory building. The second aspect deals with evaluating a grounded theory. It is noted above that, since the categories that constitute the theory are intimately linked with the data, the grounded theory itself cannot be falsified. More specifically, because a grounded theory is developed with the empirical data rather than prior to it, it does not stand independently of the data.Thus, it is not possible to test for the truth or falsity of a grounded theory by checking it against independently given empirical evidence. But a grounded theory can be evaluated by how well it explains actual economic events, that is, how well it empirically identifies and weaves together the causal mechanisms, structures, and descriptions into a narrative of the economic events being explained. In short, a grounded theory refers to real things, represents real entities, and is evaluated on how well it corresponds to the causal way the economy actually is. The evaluation process takes place within a community of scholars in that papers delineating tentative drafts of the theory are presented to colleagues at conferences and seminars for critical comments; and more refined presentations of

Foundations of Post Keynesian economics

179

the theory are published where colleagues have the opportunity to point out inadequacies. Through this cooperative process of economic-writing, economic-reading, and critical commentary, the community of Post Keynesian economists arrives at adequate theories. Consequently, a grounded theory is, in the first instance, only as good as the categories which make it up. If the data selected does not cover all aspects of the economic event(s) under investigation; if the economist compiles categories and properties from only part of the data collected or forces data into predetermined categories; if the density of the categories is small or the relationships between categories under-grounded due to incomplete data collected; if the economist choose the ‘wrong’ central causal mechanism; and/or if the narrative is static, terse, unable to fully integrate structures and central and secondary causal mechanisms, and relatively uncomplex, then the commentary of Post Keynesian critics will make it clear that the economic theory is poor, ill-developed hence to a greater or lesser extent unrealistic, and unable to provide a comprehensive and convincing explanation of actual economic events.17 As a result, the economist will have to begin the theory creation process anew. A second way to evaluate a grounded economic theory is to see how well it deals with new data.That is, the relatively enduring structures, causal mechanisms and their outcomes of a grounded theory are based on data collected in a specific time period. Thus, it is possible to evaluate whether they have remained enduring outside the time period by confronting them with ‘new’ data derived from replicating studies, especially data from actual events that at first glance appears to fall outside existing categories and not to support expected transfactual results. If the new data falls within the existing categories and conforms to the transfactual results, then the structures and causal mechanisms have been relatively enduring.18 On the other hand, if the new data falls outside the existing categories and not supporting the transfactual results, then at least some of the structures and causal mechanisms have changed. Consequently, the existing grounded economic theory needs to be modified or replaced by a completely new one.Therefore, theory evaluation in the grounded theory method based on the introduction of new data is designed to check the continual correspondence of the theory with the real causes of ongoing unique and demi-regular economic events. Hence, it is essentially a positive way of promoting theory development and reconstruction as well as new theory creation when the correspondence between theory and events breaks down. The fact that a good or poor research process leads to better or worse grounded economic theories indicates that choices made by economists do affect the final outcome. Therefore, within the grounded theory method it is possible, although not likely, to have good but different substantive and formal economic theories for the same economic events. Given the same categories, a different choice of a central causal mechanism will produce a different theory; or if the same central causal mechanism is used but integrated with different structures and secondary causal mechanisms a different theory will also be produced.19 However, since Post Keynesians are realists and critical realists, and their

180

Frederic S. Lee

theories concern causal historical events, they do not accept the possibility that there is no empirical data that could distinguish between two incompatible theories.Thus, the way for the grounded theorist to choose between the different theories is to collect new data and see which of the theories it supports supplemented by critical commentary from colleagues. In short, empirical evidence plays a major role in the creation, development, extension, and replacement of grounded theories (Glaser and Strauss, 1967; Diesing, 1971; Conrad, 1978; Wilber and Harrison, 1978; Fusfeld, 1980; Turner, 1981 and 1983; Charmaz, 1983; Gruchy, 1987; Ellis, 1985; Strauss, 1987; Konecki, 1989; Megill, 1989; Corbin and Strauss, 1990; Maki, 1990 and 2001; Strauss and Corbin, 1990 and 1994;Tosh, 1991;Wisman and Rozansky, 1991; Glaser, 1992; Sayer, 1992; Bigus, et al., 1994; Hunt, 1994; Boylan and O’Gorman, 1995; Annells, 1996; Atkinson and Oleson, 1996; Emigh, 1997; Finch, 1998 and 1999; Runde, 1998;Tsang and Kwan, 1999; McCullagh, 2000).

Methodological issues The grounded theory method of theory creation effectively dismisses the traditional issue of the realisticness of assumptions.That is, since assumptions as a basis for theory creation are not part of the grounded theory method, the degree of their realisticness is not a concern. Moreover, because the role of theoretical isolation in traditional theory building and theorizing is dependent on assumptions, their absence in the grounded theory method means that grounded theories are not isolated theories that exclude possible influencing factors.The combination of critical realism with its structures and causal mechanisms and the grounded theory method produces theories that include all the relevant factors and influences.To exclude some factors would leave the mechanisms, structures, and theories insufficiently empirically grounded. Thus, the integration of critical realism and grounded theory results in theories and theorizing different from the traditional mode.20 On the other hand, their integration produces their own set of methodological issues, centering on the nature of data, the case study method, mathematics and economic models, and econometrics (Maki, 1989, 1992a,b, 1998b). Data Originally, the grounded theory method was developed as a way to utilize qualitative data to build a theory; however, the use of quantitative data was not excluded. As economists are interested in developing historically grounded explanations of past and present economic events, their possible sources of data include all existing written, recorded, physical, and quantitative records. Since existing data sources might provide an incomplete record of economic events, the economist must also utilize different research strategies – such as surveys, interviews and oral statements, ethnographic and industrial archaeology studies, questionnaires, mapping, direct observation, participation in activities,

Foundations of Post Keynesian economics

181

fieldwork, and statistical analysis – to create new qualitative and quantitative data.21 For example, when it is important to explain how and why particular business decisions are made and who made them, the economist will need to create narrative accounts of relevant lived-historical experiences embedded within the cultural milieu of particular business enterprises. Thus the economist will need to examine letters and other written documents, undertake interviews and other oral documentation, and possibly engage in participant observation in which the economist may directly engage with, for example, the enterprise in the process of collecting data. What constitutes appropriate data depends on the object of inquiry; but it is important that much of the data deals with process, intentionality and their outcomes. Consequently, categories, hence economic structures and causal mechanisms, are grounded in both qualitative and quantitative data obtained from various sources. Case study The conceptual categories that make up grounded theories are based on an array of comparable data generated by case studies.A case study is defined as an in-depth, multifaceted investigation of a particular object or theme where the object or theme gives it its unity.The object or theme can be historical or a current real-life event and the study will use several kinds of qualitative and quantitative data sources. For example, the theme of a case study can be the pricing procedures used by business enterprises; consequently the case study will be the collection, comparison, categorization, and tabulation of pricing procedures obtained from various empirical pricing studies along with a critical narrative which examines and integrates the data. Thus, the case study approach is the principle method of qualitative and quantitative data collection and comparison used to develop categories, structures, and causal mechanisms. Moreover, by providing information from a number of different data sources over a period of time, it permits a more holistic study of structures and causal mechanisms. A case study does not stand-alone and cannot be considered alone; it must always be considered within a family of comparable case studies. If the economist is faced with a shortage of case studies, the response is not to generalize from them but to undertake more case studies. Moreover, theoretical sampling is specifically carried out through case studies. In this case, the economist makes a conscious decision to undertake a particular case study in order to increase the empirical grounding of particular categories.22 Thus, a case study can be of an individual business enterprise and the theme of the study can be to delineate the complex sets of decisions regarding pricing, production, and investment and to recount their effects over time. On the other hand, it can be concerned with a particular theoretical point, such as pricing, examined across many different case studies of different enterprises. The different cases not only provide comparable data for comparisons but also descriptions of structures and causal mechanisms and a narrative of the causal mechanism in action over time.A third type of case study is a narrative that explains a historical or current event.

182

Frederic S. Lee

The narrative includes structures and causal mechanisms which, when combined with the history or facts of the event, explains how and why it took place. Hence, this type of case study is both a historical and theoretical narrative, an integration of theory with the event. Consequently, it provides a way to check how good the theory is and, at the same time, contributes to its grounding and extension. A robust substantive theory is one that can be utilized in an array of case studies of historical and current events23 (George, 1979;Yin, 1981a,b, and 1994; Eisenhardt, 1989; Orum et al, 1991; Sayer, 1992;Vaughan, 1992;Wieviorka, 1992; Glaser and Strauss, 1994; Smith, 1998; Stake, 1998; Finch, 1999). Mathematics and economic models Mathematics and economic models are useful as tools and instruments that can contribute to the development and evaluation of causal mechanisms and grounded theory.Their uses are, however, restricted since the tenets of realism, critical realism, and the method of grounded theory prescribe that the type of mathematics used and economic models constructed are derived from (as opposed to being imposed upon via analogy or metaphor) the empirically grounded theories being developed. To translate a grounded theory into an economic model, its structures and causal mechanisms have to be converted as far as possible into mathematical language where each mathematical entity and concept is unambiguously concretely grounded. As a result, the mathematical form of the model is determined and constrained by the empirically grounded structures and causal mechanisms, and hence is isomorphic with the theory and its empirical data. This means that the model’s mathematical form is not derived by analogy or based on a metaphor, both of which are not constrained by reality; that the model is an accurate, but reflective, description of the grounded theory and its data and therefore not a simplification of it;24 that the relationships between the variables in the model are derived from the empirically grounded theory as opposed to being assumed fictions; that the same model is used in both theoretical and applied work; that the model does not operate mechanistically like a machine; and that different grounded economic theories have different models. Consequently the mathematical-theoretical arguments and numerical outcomes derived from the model are determined and constrained. In particular, the outcomes of the model are not logical deductions from given axioms or unique (or multiple) mathematical solutions; rather they are nonlogical empirically grounded outcomes.25 Such mathematicaltheoretical arguments and models derived from empirically grounded theories are characterized as rigorous and nondeductive.26 Being isomorphic with the theory and its data, yet an alternative representation of the theory, a model can be used by the economist to obtain a better understanding of the theory itself as well as an analytical-narrative summary for pedagogical purposes. In addition, it can be used to examine and evaluate propositions found in the theoretical literature. That is, the mathematical-theoretical arguments derived from a rigorous economic model can be used to examine

Foundations of Post Keynesian economics

183

whether particular mathematical-theoretical propositions associated with different economic theories and models are also rigorous or have no empirical grounding hence real world existence. Because it is grounded in the existing data, it is independent of new and future data.Thus, it can be used, for example, for discussing economic policies and simulating their possible impacts on future economic events. In particular, it is a way of visually picturing the economy and simulating its evolving, moving outcomes. Economic models can also be used to see whether the resulting outcomes of new data conform to the expected outcome patterns of the theory and to explore the impact of changing structures and causal mechanisms on economic outcomes. In this last case, for example, if a structure is hypothetically altered so that the economic model produces hypothetically different outcomes, the outcomes can then be compared to actual outcomes. If they seem to be the same, then the structures of the theory need to be reexamined and the process of grounding the theory renewed (Israel, 1981 and 1991; Boland, 1989;Weintraub, 1998a,b; Carrier, 1992; Boylan and O’Gorman, 1995; Morrison and Morgan, 1999; Dupre’, 2001). Econometrics The tenets of critical realism and the grounded theory method also constrain econometrics to being used as a statistical tool to assist the development and delineation of causal mechanisms and to evaluate the adequacy of grounded economic theories.27 In the process of transforming the empirically grounded categories into an economic theory, the economist will provisionally identify structures and causal mechanisms with particular transfactual outcomes.To aid him in his identification, the economist may subject the causal mechanism and its outcomes to econometric testing.The econometric model used will include components for the quantitative representation of structures as well as components for the causal mechanism; and its particular statistical form will be determined by the causal mechanism. If the econometric tests support the existence of the causal mechanism’s transfactual outcomes, then the empirical grounding of the causal mechanism is enhanced. Failure of the tests would, on the other hand, indicate that the causal mechanism and its associated structures are inadequately developed and needed further development. Assuming the testing a success and in light of the other qualitative and quantitative empirical support, the economist can provisionally identify the causal mechanism and its transfactual outcomes. At this stage, he can engage in further theoretical sampling to see if additional qualitative and quantitative evidence support it; and econometric testing can again be utilized in this context. Thus econometric testing is not about future predictions, as the economy is open and always changing, but about understanding the relationship between the causal mechanism and its transfactual outcomes. If econometric testing of new data fails to support the causal mechanism and its outcomes, then the implication is that the structures and causal mechanisms have changed; it then becomes necessary to re-ground them.

184 Frederic S. Lee Econometrics is also useful for evaluating grounded theories that may be associated with demi-regularities.28 In this case, the economic theory will be modeled so as to include all the structures, the primary causal mechanism, and the secondary causal mechanisms. If the testing is a success, then it can be more strongly argued that there exists a demi-regularity associated with the primary causal mechanism of the theory. But if the testing is not successful, then all that can be said is that it is less likely that the theory has a demi-regularity. Hence econometric testing provides a way to evaluate the continual correspondence of the theory with the real causes of ongoing economic events. By doing so, it contributes to the promotion of new theory building when the correspondence between theory and events break down (Lawson, 1989; Mitchell, 1991; Yeung, 1997; Mearman, 1998 and 1999; Downward, 1999, 2002; Downward and Mearman, 2002, chapters 7 and 14; Lee and Downward, 1999).

Conclusion The common sense propositions of Post Keynesians combined with realism and critical realism exclude, as part of Post Keynesian theorizing, ahistorical, atemporal entities and theoretical concepts,29 atemporal diagrams, models and other forms of analysis unaccompanied by temporal-historical analysis, and the utilization of ahistorical first principles or primary causes, all of which characterize neoclassical economic theory and many aspects of classical political economy. Being outside of history, historical time, and an unknowable transmutable future, these ahistorical entities and concepts are also rejected by the grounded theory method as fictitious since they do not emerge as categories in the historical data. In contrast, the concatenated integration of the common sense propositions, realism and critical realism, and the grounded theory method prescribes that Post Keynesian theorizing include the delineation of historically grounded structures of the economy, and the development of historically grounded emergent causal mechanisms.Thus the resulting Post Keynesian economic theories are historical theories in that they are historical narratives that explain the internal workings of historical economic processes and events in the context of relatively stable causal mechanisms and structures. Historical economic theories are possible because, as argued in critical realism, historical events are, due to the existence of structures and causal mechanisms, narratively structured. Hence, Post Keynesians do not impose narratives on actual economic events to make sense of them, but derives them from the events via the grounded theory method. Moreover, being a narrative, the theories have a plot with a beginning, middle, and end centered on a central causal mechanism and set within structures and other causal mechanisms. Therefore, antedated events prompt the causal mechanisms to initiate activity to generate particular results and hence start the narrative; and it comes to an end when the causal mechanisms conclude their activity. Finally, the storyteller of the narrative is the Post Keynesian economist whose objective is to help the audience – who includes fellow economists, students, politicians, and the general public – understand

Foundations of Post Keynesian economics

185

theoretically how and why the actual economic events transpired (McCullagh, 1984; Carr, 1986; Norman, 1991; Appleby et al., 1994, chapters 7 and 8). As a narrative centered on causal mechanisms and structures, historical theories are not completely aggregated or disaggregated; and nor are they devoid of explicit human intentionality and activity. That is, because causal mechanisms embody data from many case studies, they aggregate economic reality or, put differently, reduce the scale of reality and therefore the degree of detail and specificity required of the narrative. However, the degree of aggregation is limited because of the existence of structures that cannot be aggregated or disaggregated and human intentionality and activity which are both differentiated and specific. As a result, historical theories are neither an aggregate theory where the differentiation among the causal mechanisms, structures, and human agency disappear; nor such a disaggregated theory where causal mechanisms, structures, and human agency are individual-event specific and hence of little interest.Thus, historical economic theories are quasi-aggregated narratives explaining the many and overlapping actual economic events occurring in a differentiated economy. As a result, the narratives they tell are not about capitalism and its historical destination, but provide a historicist, that is historically contingent, theoretical understanding of economic events in a stable but ultimately slowly transforming economy whose unknown direction is in the hands of human agency. The historical economic theory that results from the integrating realism, critical realism, and the method of grounded theory has, in the process of its creation, also resulted in critical realism and grounded theory being interactively modified. Realism requires real entities and grounded theory operates on the principle of empirically grounding real entities or categories; hence the structural and causal mechanisms of critical realism in the context of Post Keynesian economics are observable and real as opposed to being unobservable and imaginary. Reciprocally, critical realism establishes the objectives of the Post Keynesians using grounded theory, that is the grounding of structures and causal mechanisms. Moreover, since Post Keynesians see human agency in all causal processes and historical narratives require human agency, the causal mechanisms embedded in their theories must also contain human agency. Thus, the historical economic theories of Post Keynesian economics are a result of a particular concatenated integration of common sense propositions, realism and critical realism, and grounded theory.

Notes 1 For example, see Dow (1990a, 1999), Kanth (1992), Lawson (1994 and 1999a), Arestis (1996), Pratten (1996), Joseph (1998), Downward (1999), McKenna and Zannoni (1999) Rotheim (1999). 2 Because its components are intentionally related, a causal mechanism cannot be thought of as a machine – see Cartwright (1995) and Dupre’ (2001) on machines and economic thinking.

186

Frederic S. Lee

3 This property of causal mechanisms obviates the need for an inductivist approach for theory creation. Consequently, any method of theory creation and development, such as retroduction or the method of grounded theory, utilized in conjunction with critical realism will not adopt an inductivist strategy leading to invariant laws (Sayer, 1992). 4 An alternative way of proceeding would be to pursue the Babylonian method, but it is underdeveloped in that it advocates utilizing various research strategies to gather data from different sources, but then does not say how the data is transformed into a theory. Another alternative approach would be to utilize retroduction as the methodological guideline for theory creation. Retroduction involves first a theoretical-abstract re-description of the events based on existing qualitative and quantitative material; followed by explaining the events through postulating and identifying the structures and causal mechanisms producing them; and ending with the theory being checked empirically.Although advocated, few critical realists have explicitly used it for theory construction; and when they have (e.g. Lawson, 1998c), the results looks like a case study carried out according to the method of grounded theory. This is because retroduction is also too undeveloped to be used as a way to identify causal mechanisms and structures from which to construct theories or explanations. In addition, it does not indicate how the causal processes should be delineated and articulated, that is the analytical and literary form the theory should take. Finally, retroduction, as articulated by critical realists, suggests that analogies, similes, and metaphors are useful ways to identify causal mechanisms and structures. But since the latter are by their nature not real, their contribution to theory construction is at best limited and, moreover, might direct the attention of the critical realist Post Keynesian towards fictitious, hence empirically ungroundable, mechanisms and structures (Dow, 1985, 1990a,b, 1999; Lawson, 1996, 1997b, 1998c, 1999b; Sarre, 1987; Sayer, 1992; Pratt, 1995; Boylan and O’Gorman, 1995; Yeung, 1997; Runde, 1998; Downward, 1999). 5 The method of grounded theory was first delineated by Barry Glaser and Anselm Strauss (1967) and then subsequently developed by them and others – see Strauss (1987) and Strauss and Corbin (1990 and 1994). Similar methodological guidelines going by the names of holism, pattern model, method of structured-focused comparison, and participant-observer approach using case study method were also proposed and developed at roughly the same time – see Diesing (1971),Wilber and Harrison (1978), George (1979); and Fusfeld (1980). 6 What grounded theory is not about is forcing data into predetermined set of theoretical categories; all categories have to be empirically justified. 7 Observable data is not solely restricted to sense experience. For example, historical documents or field reports contain data that cannot be verified by the reader’s sense experience. The same can also be said for oral histories that deal with past events. On the other hand, non-written data, such as informal rules and hierarchical power inside the business enterprise, are not unobservable in that they can be verbally articulated and hence written down, filmed and then identified at a later point in time, or observed as institutions, that is as observable patterns of behavior hence capable of being recorded. Thus all data is observable, although the sources and medium in which they exist varies; to be unobservable in this sense is not to be real and hence to be no data at all. 8 Constant comparison can also involve exact replicating previous studies to see how robust they are. 9 Another way of putting this is that a category represents a ‘pattern’ that the economist has recognized in the data generated by replicatory or comparable studies. 10 In either case, the language used to describe the categories may be quite different from the existing theoretical language. In particular, the building of a grounded theory may require the creation of a new language and discarding old words and

Foundations of Post Keynesian economics

11

12

13 14 15

16

17

18

19 20 21 22

23

187

their meanings. On the other hand, the language used may come directly from the data collected and/or from commonly used language (which is generally not theoretical language) (Konecki, 1989; Coates, 1996). The point of theoretical sampling is specifically to find data to make categories denser, more complex. Since the aim of the grounded theory method is to build theories based on data collected, the issue of generalizing in a statistical sense is not relevant.Thus theoretical vs statistical sampling has no importance for the method. (Glaser and Strauss, 1967; Corbin and Strauss, 1990) A saturated category is not a function of the number of pieces of data, as it may become saturated after only a small portion of the available data has been analyzed. Non-parametric statistical techniques can be used to judge whether a category is saturated or not. Finch and McMaster (2002 and Chapter 8). Thus, the grounded theory method is not the same as induction; and its practitioners view generalization as a problematic and unimportant goal and empirical-theoretical laws as not worth pursuing (Sarre, 1987; Sayer, 1992). The irrelevance of logical coherence means that it is not possible to deduce unknown structures and causal mechanisms from existing ones. A formal grounded theory is not more (or less) abstract than a substantive grounded theory. Because a grounded theory must at all times be grounded, it cannot be an abstract theory where the modifier denotes some degree of non-groundness, such as the use of fictional categories or the elimination of data. Hence grounded theories cannot be differentiated according to their levels of abstraction. By accepting that it may be necessary to cast aside previously acquired knowledge, the economist can still pursue the grounded theory method even though he may favor particular non-grounded concepts and theories. Hence the grounded theory economist is not a neutral observer sifting through given ‘facts’ that present him, through some sort of immaculate conception, with a theory without a moment of forethought. The often-heard phrase that ‘all theory is in some sense unrealistic’ is not applicable to grounded theories. All grounded theories are realistic in that they are grounded in every detail in data. A grounded theory may be relatively complete or a much incomplete explanation of an economic event; but in both cases they are entirely realistic.To be unrealistic from a grounded theory perspective is to include non-grounded concepts in the theory, but then it would not be grounded. This has been called pattern-matching in that the existing theory is seen as a particular pattern of data and narrative and the new pattern of data with its narrative is compared to it to see if they match (see Wilber and Harrison, 1978 and Yin, 1981a,b). The expectation of the grounded theory method is that the economist will seriously consider alternative combinations of structures and causal mechanisms before settling on a particular theory. For example, from a critical realist-grounded theory perspective, an assumptionbased isolative theory is without empirical grounding and hence has no real world sense or meaning. Created data does not preexist as some sort of unacknowledged sense experience or as unobservable data; rather being produced, it has no past. Moreover, since economists create data, they have a direct and reflexive relation to it. It is important to realize that a case study that involves the replication and re-evaluation of a previous case study is theoretical sampling. In this instance, the researcher is reexamining an existing case study to see how robust its data and results are. This type of case study is similar to the extended case method advocated by Burawoy (1991 and 1998), with the caveat that the latter is predicated on a false dichotomy between structures and causal mechanisms, where structures change independently of causal mechanism, not in part because of them.

188

Frederic S. Lee

24 A model which simplifies a grounded theory does not accurately and concretely denote its structures and leaves its causal mechanisms under-specified, underembedded and thus ill-defined. 25 An example of a rigorous, non-deductive economic model that can be developed is a price model based on the input–output table of an economy (see Lee, 1998). The table represents a set of structures that can be translated into matrix algebra, while the causal mechanism is the pricing procedures used by business enterprises. Thus the price model of the economy has the concrete and constrained mathematical form of [Rd][Mpt  Lw  d]  pt1. Each mathematical component of the model is empirically grounded; the relationships between all the model’s components are specified by the primary and secondary causal mechanisms; and the model’s outcomes are empirically grounded prices. Moreover, the mathematicaltheoretical arguments derived from the model, such as the existence and implications of the commodity residual, are constrained by its empirical groundness.Thus the empirically grounded price model can be used in applied work as well as to pursue particular theoretical issues. 26 The contrast to a rigorous and non-deductive mathematical argument and model are those based on non-grounded axioms and whose non-grounded outcomes are logically and consistently derived from the axioms. 27 Econometric testing can also be used to evaluate particular claims in the historical literature regarding causal mechanisms and transfactual outcomes. 28 Non-parametric statistics can also contribute to the analysis of demi-regularities and to measures of the nature of any relationship between demi-regularities cast into a contrastive relationship (Finch and McMaster, 2002 and Chapter 8). 29 For example, the condemned entities and concepts include stylized facts, short and long period positions, equilibrium (and disequilibrium), market clearing, asocial preferences, and maximization.

References Annells, M. (1996) Grounded theory method: philosophical perspectives, paradigm of inquiry, and postmoderism. Qualitative Health Research 6 (August): 379–393. Appleby, J., Hunt, L. and Jacob, M. (1994) Telling the Truth About History. New York: W.W. Norton and Company. Arestis, P. (1996) Post-Keynesian economics: towards coherence. Cambridge Journal of Economics 20 ( January): 111–135. Atkinson, G.W. and Oleson, T. (1996) Institutional inquiry: the search for similarities and differences. Journal of Economic Issues 30 (September): 701–718. Bigus, O.E., Hadden, S.C. and Glaser, B.G. (1994) The study of basic social processes. In More Grounded Theory Methodology:A reader, pp. 38–64. Edited by B.G. Glaser. Mill Vallet: Sociology Press. Boland, L. (1989) The Methodology of Economic Model Building: Methodology after Samuelson. London: Routledge. Boylan,T. and O’Gorman, P. (1995) Beyond Rhetoric and Realism in Economics:Towards a Reformulation of Economic Methodology. London: Routledge. Burawoy, M. (1991) The extended case method. In Ethnography Unbound, pp. 271–287. Berkeley: University of California Press. —— (1998) The extended case method. Sociological Theory 16(1) (March): 4–33. Carr, D. (1986) Narrative and the real world: an argument for continuity. History and Theory. Studies in the Philosophy of History 25(2): 117–131.

Foundations of Post Keynesian economics

189

Carrier, D. (1992) A methodology for pattern modeling nonlinear macroeconomic dynamics. Journal of Economic Issues 26 (March): 221–242. Cartwright, N. (1995) Ceteris Paribus laws and socio-economic machines. The Monist 78(3) ( July): 276–294. Charmaz, K. (1983) The grounded theory method: an explication and interpretation. In Contemporary Field Research: A Collection of Readings, pp. 109–126. Edited by R. Emerson. Boston: Little, Brown and Company. Coates, J. (1996) The Claims of Common Sense: Moore,Wittgenstein, Keynes and the Social Sciences. Cambridge: Cambridge University Press. Comim, F. (1997) Common sense economics history of economic services. http://www.eh.net/~HisEcSoc/Resources/Editorials/Comim/editorial.shtml Conrad, C.F. (1978) A grounded theory of academic change. Sociology of Education 51 (April): 101–112. Corbin, J. and Strauss, A. (1990) Grounded theory research: procedures, canons and evaluative criteria. Qualitative Sociology 13 (Spring): 3–21. Davidson, P. (1996) Reality and economic theory. Journal of Post Keynesian Economics 18 (Summer): 479–508. Diesing, P. (1971) Patterns of Discovery in the Social Sciences. New York: Aldine. Dow, S.C. (1985) Macroeconomic Thought: A Methodological Approach. Oxford: Basil Blackwell. —— (1990a) Post-Keynesianism as political economy: a methodological discussion. Review of Political Economy 2 (November): 345–358. —— (1990b) Beyond dualism. Cambridge Journal of Economics 14(2) ( June): 143–157. —— (1999) Post Keynesianism and critical realism: what is the connection? Journal of Post Keynesian Economics 22 (Fall): 15–34. Downward, P. (1999) Pricing Theory in Post Keynesian Economics: A Realist Approach. Cheltenham: Edward Elgar. —— (2002) Realism, Econometrics and Post Keynesian Economics. In Post Keynesian Econometrics, Microeconomics and the Theory of the Firm, Beyond Keynes Vol. I, Edited by S.C. Dow and J. Hillard Cheltenham: Edward Elgar pp. 144–61. —— and Mearman,A. (2002) Critical realism and econometrics: constructive dialogue with Post Keynesian economics, Metroeconomica 53(4): 391–415. Dupre’, J. (2001) Economics without mechanism. In The Economic World View: Studies in the Ontology of Economics, pp. 308–332. Edited by U. Maki. Cambridge: Cambridge University Press. Eisenhardt, K.M. (1989) Building theories from case study research. Academy of Management Review 14(4): 532–550. Ellis, B. (1985) What science aims to do. In Images of Science, pp. 48–74. Edited by P.M. Churchland and C.A. Hooker. Chicago:The University of Chicago Press. Emigh, R.J. (1997) The power of negative thinking: the use of negative case methodology in the development of sociological thinking. Theory and Society: 26: 649–684. Finch, J. (1998) A grounded theory paper for economics. Department of Economics Discussion Paper 98-15. University of Aberdeen. Finch, J.H. (1999) The methodological implications of Post Marshallian economics. In Contingency, Complexity and the Theory of the Firm: Essays in Honour of Brian J. Loasby, pp. 156–177. Edited by S.C. Dow and P.E. Earl. Cheltenham: Edward Elgar. Finch, J. H. and McMaster, R. (2002) On non parametric statistical inference in the pursuit of causal explanations, Cambridge Journal of Economics 26(6): 753–772.

190

Frederic S. Lee

Fleetwood, S. (2001a) Causal laws, functional relations and tendencies. Review of Political Economy 13(2) (April): 201–220. —— (2001b) Conceptualizing unemployment in a period of atypical employment: a critical realist perspective. Review of Social Economy 59 (March): 45–69. Fusfeld, D.R. (1980) The conceptual framework of modern economics. Journal of Economic Issues 14 (March): 1–52. George, A.L. (1979) Case studies and theory development: the method of structured, focused comparison. In Diplomacy: New Approaches in History, Theory and Policy, pp. 43– 68. Edited by P.G. Lauren. New York:The Free Press. Glaser, B.G. (1992) Emergence vs Forcing: Basics of Grounded Theory Analysis. Mill Valley: Sociology Press. —— and Strauss, A.L. (1967) The Discovery of Grounded Theory: Strategies for Qualitative Research. New York: Aldine Publishing Company. —— and Strauss, A.L. (1994) Case histories and case studies. In More Grounded Theory Methodology: A Reader, pp. 233–245. Edited by B.G. Glaser. Mill Valley: Sociology Press. Gruchy, A.G. (1987) The Reconstruction of Economics: An Analysis of the Fundamentals of Institutional Economics. New York: Greenwood Press. Hodgson, G.M. (1998a) From micro to macro: the concept of emergence and the role of institutions. Unpublished. —— (1998b) Emergence. In The Handbook of Economic Methodology, pp. 156–160. Edited by J.B. Davis, D.W. Hands and U. Maki. Cheltenham: Edward Elgar. Hunt, S.D. (1994) A realist theory of empirical testing: resolving the theory-ladenness/ objectivity debate. Philosophy of the Social Sciences 24(2) (June): 133–158. Ingham, G. (1996) Money is a social relation. Review of Social Economy 54 (Winter): 507–529. Israel, G. (1981) ‘Rigor’ and ‘axiomatics’ in modern mathematics. Fundamenta Scientiae 2: 205–219. —— (1991) Volterra’s ‘Analytical Mechanics’ of Biological Associations. Archives Internationales d’Histoire des Sciences 41(127): 307–352. Joseph, J. (1998) In defense of critical realism. Capital and Class 65 (Summer): 73–106. Kanth, R. (1992) Economics and epistemology: a realist critique. Capital and Class 47 (Summer): 93–112. Konecki, K. (1989) The methodology of grounded theory in the research of the situation of work. The Polish Sociological Bulletin 2: 59–74. Lawson, C., Peacock, M. and Pratten, S. (1996) Realism, underlabouring and institutions. Cambridge Journal of Economics 20 ( January): 137–151. Lawson, T. (1989) Abstraction, tendencies and stylised facts: a realist approach to economic analysis. Cambridge Journal of Economics 13 (March): 59–78. —— (1994) The nature of Post Keynesianism and its links to other traditions: a realist perspective. Journal of Post Keynesian Economics 16 (Summer): 503–538. —— (1996) Developments in ‘economics as realist social theory’. Review of Social Economy 54 (Winter): 405–422. —— (1997a) Economics and Reality. London: Routledge. —— (1997b) On criticizing the practices of economists: a case for interventionist methodology. In Pluralism in Economics: New Perspectives in History and Methodology, pp. 13–36. Edited by A. Salanti and E. Screpanti. Cheltenham: Edward Elgar. —— (1998a) Tendencies. In The Handbook of Economic Methodology, pp. 493–498. Edited by J.B. Davis, D.W. Hands and U. Maki. Cheltenham: Edward Elgar.

Foundations of Post Keynesian economics

191

—— (1998b) Transcendental Realism. In The Handbook of Economic Methodology, pp. 504–510. Edited by J.B. Davis, D.W. Hands and U. Maki. Cheltenham: Edward Elgar. —— (1998c) Social relations, social reproduction and stylized facts. In Method, Theory and policy in Keynes: Essays in Honour of Paul Davidson,Vol. 3, pp. 17–43. Edited by P. Arestis. Cheltenham: Edward Elgar. —— (1999a) Connections and distinctions: Post Keynesianism and critical realism. Journal of Post Keynesian Economics 22 (Fall): 3–14. —— (1999b) Critical issues in Economics as Realist Social Theory. In Critical Realism in Economics: Development and Debate, pp. 209–257. Edited by S. Fleetwood. London: Routledge. Lee, F.S. (1998) Post Keynesian Price Theory, Cambridge, Cambridge University Press. —— and Downward, P. (1999) Retesting gardiner means’s evidence on administered prices. Journal of Economic Issues 33(4) (December): 861–886. Lloyd, C. (1993) The Structures of History. Oxford: Basil Blackwell Ltd. Lovering, J. (1990) Neither fundamentalism nor ‘New Realism’: a critical realist perspective on current divisions in socialist theory. Capital and Class 42 (Winter): 30–54. Maki, U. (1989) On the problem of realism in economics. Ricerche Economiche 43(1–2): 176–198. —— (1990) Scientific realism and Austrian explanation. Review of Political Economy 2(3) (November): 310–344. —— (1992a) The market as an isolated causal process: a metaphysical ground for realism. In Austrian Economics: Tensions and New Directions, pp. 35–59. Edited by B.J. Caldwell and S. Boehm. Boston: Kluwer Academic Publishers. —— (1992b) On the method of isolation in economics. Poznaan Studies in the Philosophy of the Sciences and the Humanities 26: 317–351. —— (1996) Scientific realism and some pecularities of economics. In Realism and AntiRealism in the Philosophy of Science, pp. 427–447. Edited by R.S. Cohen, R. Hilpinen and Qui Renzong. Dordrecht: Kluwer Academic Publishers. —— (1998a) Realism. In The Handbook of Economic Methodology, pp. 404–409. Edited by J.B. Davis, D.W. Hands and U. Maki. Cheltenham: Edward Elgar. —— (1998b) Aspects of realism about economics. Theoria 13(2): 310–319. —— (2001) The way the world works (www): towards an ontology of theory choice. In The Economic World View: Studies in the Ontology of Economics, pp. 369–389. Edited by U. Maki. Cambridge: Cambridge University Press. McCullagh, C.B. (1984) Justifying Historical Descriptions. Cambridge: Cambridge University Press. —— (2000) Bias in historical description, interpretation, and explanation. History and Theory: Studies in the Philosophy of History 39(1): 39–66. McKenna, E.J. and Zannoni, D.C. (1999) Post Keynesian economics and critical realism: a reply to Parsons. Journal of Post Keynesian Economics 22 (Fall): 57–70. Mearman, A. (1998) Keynes, realism and econometrics. Unpublished. —— (1999) Towards a critical realist econometric methodology? Unpublished. Megill, A. (1989) Recounting the past: ‘Description’, explanation, and narrative in Historiography. American Historical Review 94: 627–653. Mitchell, W.F. (1991) The role of econometrics in a Post-Keynesian research programme. Unpublished.

192

Frederic S. Lee

Morrison, M. and Morgan, M.S. (1999) Models as mediating instruments. In Models as Mediators: Perspectives on Natural and Social Science, pp. 10–37. Edited by M.S. Morgan and M. Morrison. Cambridge: Cambridge University Press. Norman, A.P. (1991) Telling it like it was: historical narratives on their own terms. History and Theory: Studies in the Philosophy of History 30(2): 119–135. Orum, A.M., Feagin, J.R. and Sjoberg, G. (1991) Introduction: the nature of the case study. In A Case for the Case Study, pp. 1–26. Edited by J.R. Feagin, A.M. Orum and G. Sjoberg. Chapel Hill:The University of North Carolina Press. Pratt, A. (1995) Putting critical realism to work: the practical implications for geographical research. Progress in Human Geography 19 (March): 61–74. Pratten, S. (1996) The ‘Closure’ assumption as a first step: neo-Ricardian economics and Post-Keynesianism. Review of Social Economy 54 (Winter): 423–443. Rotheim, R.J. (1999) Post Keynesian economics and realist philosophy. Journal of Post Keynesian Economics 22 (Fall): 71–104. Runde, J. (1998) Assessing causal economic explanations. Oxford Economic Papers 50: 151–172. Sarre, P. (1987) Realism in practice. Area 19 (March): 3–10. Sayer,A. (1992) Method in Social Science:A realist approach, 2nd edn. London: Routledge. Smith, L.M. (1998) Biographical method. In Strategies of Qualitative Inquiry, pp. 184–224. Edited by N.K. Denzin and Y.S. Lincoln.Thousand Oaks: SAGE Publications. Stake, R.E. (1998) Case studies. In Strategies of Qualitative Inquiry, pp. 86–109. Edited by N.K. Denzin and Y.S. Lincoln.Thousand Oaks: SAGE Publications. Strauss, A.L. (1987) Qualitative Analysis for Social Scientists. Cambridge: Cambridge University Press. Strauss,A. and Corbin, J. (1990) Basics of Qualitative Research: Grounded Theory Procedures and Techniques. Newbury Park: Sage Publications. —— (1994) Grounded theory methodology: an overview. In Handbook of Qualitative Research, pp. 273–285. Edited by N.K. Denzin and Y.S. Lincoln.Thousand Oaks: Sage Publications. Tosh, J. (1991) The Pursuit of History: Aims, Methods and New Directions in the Study of Modern History, 2nd edn. London: Longman. Tsang, E. and Kwan, K.-M. (1999) Replication and theory development in organization science: a critical realist perspective. Academy of Management Review 24(4): 759–780. Turner, B.A. (1981) Some practical aspects of qualitative data analysis: one way of organising the cognitive processes associated with the generation of grounded theory. Quality and Quantity 15: 225–247. —— (1983) The use of grounded theory for the qualitative analysis of organizational behaviour. Journal of Management Studies 20(3): 333–348. Vaughan, D. (1992) Theory elaboration: the heuristics of case analysis. In What is a Case? Exploring the Foundations of Social Inquiry, pp. 173–202. Edited by C.C. Ragin and H.S. Becker. Cambridge: Cambridge University Press. Weintraub, E.R. (1998a) From rigor to axiomatics: the marginalization of Griffith C. Evans. In From Interwar Pluralism to Postwar Neoclassicism, pp. 227–259. Edited by M.S. Morgan and M. Rutherford. Durham: Duke University Press. —— (1998b) Controversy: axiomatisches Mi∃verstandis. The Economic Journal 108 (November): 1837 – (1847). Wellman, B. and Berkowitz, S.D. (eds) (1997) Social Structures: A Network Approach. Greenwich: JAI Press Inc.

Foundations of Post Keynesian economics

193

Wieviorka, M. (1992) Case studies: History or Sociology? In What is a Case? Exploring the Foundations of Social Inquiry, pp. 159–172. Edited by C.C. Ragin and H.S. Becker. Cambridge: Cambridge University Press. Wilber, C.K. and Harrison, R.S. (1978) The methodological basis of institutional economics: pattern model, storytelling, and holism. Journal of Economic Issues 12 (March): 61–89. Wisman, J.D. and Rozansky, J. (1991) The methodology of institutionalism revisited. Journal of Economic Issues 25 (September): 709–737. Yeung, H. (1997) Critical realism and realist research in human geography: a method or a philosophy in search of a method? Progress in Human Geography 21 (March): 51–74. Yin, R.K. (1981a) The case study crisis: some answers. Administrative Science Quarterly 26 (March): 58–65. —— (1981b) The case study as a serious research strategy. Knowledge: Creation, Diffusion, Utilization 3 (September): 97–114. —— (1994) Case study research: design and methods, 2nd. edn. London: Sage Publications.

Part III

Critical realism in empirical analysis

11 Questionnaires in realist research A case study of rural small-scale industry in Ghana Wendy Olsen,Thandie M. Hara and Sampson Edusah Introduction The aim of this chapter is to review realist approaches to the use of the survey method. The first three sections review general issues related to the survey method from a realist perspective.There is room for debate among variants of realism, some of which have much in common with traditions of statistical inference. However, open systems and a complex assumption about ‘what exists’ are two aspects which create a need for a new set of statistical practices (i.e. a realist statistical discourse). This chapter begins to undertake this challenging project, looking at the pre-survey, data-analysis and interpretation stages of quantitative social research. Near the end of the section ‘Sampling and representativeness in a multi-level reality’, a case study from Ghana applies realist statistics to the Rural Small Scale Industries sector (RSSI).The sampling procedure used in that study is critically reviewed. In the section ‘Regression techniques revisited’, we examine how one might do regression with such cross-sectional data. We compare traditional methods with realist methods. In the penultimate section, an innovative approach to the connections between description and involvement is set out.A dialectic of description and involvement is a commonly held epistemic tenet both among critical realist and feminist epistemologists. In the final section, we summarise the strengths and weaknesses of the case study material from this dialectical perspective. Finally, the chapter concludes with some implications of a dialectical approach to method. Specifically we propose that fallible descriptions are a useful outcome for realist economic researchers using statistical methods.

Realism and innovative modelling Realist authors in economics have taken a post-positivist approach to the philosophy of science. Critical realism, in particular, is highly critical of empiricist falsificationism. Other realists argue that the objects of social science are many in

198

Wendy Olsen,Thandie M. Hara and Sampson Edusah

number and are likely to be inter-linked with each other.Therefore, both types of realist agree that we do not know about the social objects perfectly; ‘objectivity’ has to be defined. Objectivity among non-critical realists tends to refer to the potential for testing commensurate hypotheses by different researchers leading to compatible results. ‘Objective’ research is thus defined in terms of a network of scientists sharing practices and concurring with findings; according to critical realists such a position concedes much of the ground of epistemological relativism (Lawson, 1997). Specifically, critical realists such as Lawson argue that all knowledge claims are socially constructed, and that therefore what was previously known as ‘objectivity’ is now being redefined as merely a temporary consensus of opinion. One specific term used for the fact that construal occurs is transitivity (Bhaskar, 1975). Traditionally statisticians, seen from a realist viewpoint, have been naive social constructionists. Knowledge claims are transitively constructed and they form (or represent – depending on how long their truth is meant to last) a relationship between the speaker or writer and the world to which she or he refers. Therefore, long-standing debates about hermeneutics and discourse apply to the analysis of specific scientific hypotheses (Outhwaite, 1987; Chouliaraki and Fairclough, 1999). Thus, since most statistical work is presently formulated within the falsificationist paradigm (Popper, 1957), which admires testing hypotheses but believes that the data (not the researchers) determine the objective truth, the traditional discourse has some persistent problems. For instance, suppose we state the several hypotheses surrounding the research question, and then take up the first hypothesis. If the hypothesis itself cannot be objective, then the falsification of the hypothesis cannot rest upon objective grounds alone. One would therefore expect most realists to reject most empirical hypothesis-testing and specifically statistical hypothesis testing. At most they would see this testing as a social practice, embedded in its discourse, rather than as an objective way to arrive at true claims. Two modes of testing must be referred to here. First, testing of purely inductive hypotheses is constructivist in the sense that the researchers use their prior constructs to identify what ‘springs out at them’ from the empirical data. Second, the testing of deductive hypotheses – those arising perhaps from mathematical models or from pure theory pre-existing the collection of empirical data – is also constructivist, in the sense that a researcher must be selective and judgemental in deciding what to focus upon and how to phrase the discussion. Are the respondents ‘recidivists’, ‘criminals’, ‘perpetrators’, ‘accused’, ‘subjects’ or ‘participants’? In summary, both the traditions of data-orientated and model-orientated hypothesis testing are falsely masquerading if they claim to be value-neutral or objective in their viewpoint at the time of interpretation. However, it is possible for a realist to reject the supposed objective nature of such testing while proposing that the social practices of testing might still be

Questionnaires in realist research 199 useful. Such a project might be called realist statistics.This project requires the creation of a new discourse of statistics. Instead of the traditional statistical inference discourses, we need statistics to be socially located in a specific time and place. Such an approach usefully places statistics amongst the other tools of data analysis which are commonly used by multi-method socio-historical researchers. The analysis of statistics is indeed a technique commonly found among historians, who are critical of their resources whilst at the same time promoting the reproduction and re-analysis of existing quantitative survey data. Thus, like historians, realists would appreciate that the real world is only partially and imperfectly reflected in survey data. In the rest of this section, we explain some assumptions that are shared between statistical inference and realism, and then mention the open systems assumptions which differentiate the two. One common assumption held by statisticians is that factors outside of the immediate data set may be affecting the outcomes measured in that data sets. The concepts of proxy variables, controlling for external factors exogenous factors and instrumental variables all indicate an awareness that the factors within a marvel are not complete account of the factors that might be relevant to a causal explanation. Proxy variables account for factors which were not directly operationalised in the survey data collection stage. Dummy variables which control for external factors obviously only work if they are constructed within the data set.Thus, dummy variables actually bring ‘the external world’ in as an endogenous factor. Both procedures reflect an implicit recognition that the real world is complex. The term ‘exogenous factor’ is widely used in an idealist-modelling context. Philosophical idealism is an ontology (a set of assumptions about what exists) which justifies the use of models which do not correspond explicitly to the actual or real worlds. The very abstractness of the models makes it impossible for them to refer directly to worldly things. However, a clear distinction between the empirical, the actual and the real worlds was made explicitly by Bhaskar in 1975 (see also Bhaskar, 1997 reprint). Bhaskar’s three domains correspond approximately to the concepts of endogenous variables and exogenous variables in traditional economic modelling, with endogenous variables being empirically measured and exogenous variables representing actual events or things which were not measured or are nominally being kept outside a model for convenience. Most economists are not clear whether the exogenous variables exist in the real world or not. They are primarily clear that the exogenous variables are not in the empirical measurement realm.The debate about this in econometrics has confused the empirical realm with the real world. For instance, if consumption is measured as household annual expenditure, then that simply is consumption for most economists – but this belittles the richness and complexity of the underlying actual events (e.g. a vegetarian avoiding meat in a household that buys meat) and real things

200

Wendy Olsen,Thandie M. Hara and Sampson Edusah

that have powerful effects (e.g. arguments over meat leading to a shift towards buying fish). However, it is not necessary to be confused about the difference between that which is measured empirically and that which exists in reality. The latter is obviously a more expansive domain than the former. The borderline between exogenous variables and endogenous variables needs to be delineated using both pragmatic and scientific criteria. Otherwise, everything would be endogenous. In this context, the phrase ‘instrumental’ variables is interesting. The instrumental variables technique uses an extra equation to explain the quantum or frequency of one of the variables in the primary equation. Usually, the contents of the extra instrumental equation are hidden from the reader and/or placed in an appendix. However, the use of instrumental variables in making econometric estimates reveals a recognition that the world behind the data is more complex than the world that is represented in a single empirical equation. So these are some common areas of agreement between econometrics and realist statistics. However social change is an area in which there is a deeper disagreement. Many economists habitually avoid recognising the reality of ongoing social change – especially when analysing data. This avoidance creates a chasm between value-neutral econometric estimates and socially contextualised realist interpretations. What is meant by ‘the reality of ongoing social change’? It refers to the way social systems are intrinsically organic networks of interacting beings and things. The phrase ‘open systems’ summarises this ontological assumption very well. According to Lawson, open systems have three main characteristics. (1) Their character emerges from their ongoing workings, and cannot be reduced to their component parts. (2) They are organically changing from within. (3) Their boundary with the outside world is permeable. Lawson’s definition of an open system can be applied to institutions and to structures, and suggests a complex ontology with all these things being linked together. Thus, for example, a small-scale firm has a culture and tradition which cannot be reduced to any particular employee; (1) it may be transformed by its own employees; (2) and it has links with the families of the workers involved which are very important to the working of the firm. To think of firms as organic, permeable, structured ‘things’ is to recognise their complexity. This recognition is common to institutionalists, evolutionary economists and political economists. The classic statistical problem of omitted variable bias reflects the closure in the system represented by a regression equation or a model. However, it is possible to introduce some openness into the analysis of statistical data. This chapter illustrates these possibilities using a cross-sectional data. In the authors’ view, cross-sectional data sets offer good opportunities for innovative data creation. For this reason the next section addresses the pilot stage of the research project.

Questionnaires in realist research 201

The pilot stage According to Bryman (1996), all quantitative data rest upon a prior pilot stage which is fundamentally qualitative. The qualitative stage recognises the complexity of the objects which are being studied. Decisions are made about which objects should be the direct focus of the research. One phrase for such decisionmaking is ‘disciplined noticing’.This phrase arises in educational research, and is particularly used in qualitative research. However, one could argue that in constructing a questionnaire two important decisions are made about the objects of research. First, one decides what unit of analysis is appropriate for the questionnaire or survey instrument. For example, one decides whether households or individuals are to be referred to and, second, one decides what to ask about. In economics an important decision is whether to restrict the questions or the analysis to commercial variables (those measured in money terms). Usually this is not acceptable yet it is commonly done. Other variables measured in qualitative categories are essential for an adequate description of the economic ‘things’. For instance, a person’s labour market involvement might be categorised as ‘retired’,‘self-employed’,‘pieceworker’,‘farming work’, ‘grazing work’ etc., these categories are not mutually exclusive. Far too often, economists’ questionnaires simplify these categories to a single, simplified variable with mutually exclusive categories.With modern computer techniques to implement analyses such as multinomial logistic regression there is no need for this simplification. In the Ghanaian research on small-scale rural industries, the following objects were considered to be important: the proprietor and their demographic characteristics; the firm (and those micro-enterprises which are too informal to count as a small-scale industry were excluded by definition); the paid labour inputs; the capital of the firm; the turnover; the sector; the market, for example, input markets and output markets; and the region. However, women participate in these markets under very different conditions from men, and once gender was inserted into the data analysis stage it was realised that gender relations and family relations were also important ‘things’. To study gender relations (rather than just studying the two genders separately) one needs to begin to use qualitative methods.Therefore, the study had a limitation that had been ontically presupposed when it originated. Similarly, if one wants to know about intra-family labour sharing and inter-household capital flows, such as lending, joint ownership and inheritance, then life histories mixed with family history data need to be compiled. In other words, the choice of ‘things’ to study has implications for the appropriate research design. In management studies, a mixed design with case-study material is commonly used, but in economics itself (almost by definition!) such mixed-method research is shunned. Heterodox economists tend to redefine economics to include such applied studies using multi-method research strategies. However, do not throw out the statistical tools when moving to this new vantage point. Integrate them instead.

202

Wendy Olsen,Thandie M. Hara and Sampson Edusah

Sampling and representativeness in a multi-level reality In this section the way a realist has conducted research at multiple levels is illustrated. In addition it describes the implications of the awareness of gender relations for the case study in Ghana which is the specific focus of the rest of this chapter.The Ghanaian research began without an awareness of gender but with a strong interest in multiple levels of the socio-economy (persons, firms, sectors, markets and countries). Later it became clear to the researchers that a gendered analysis would require a re-working of the data.This re-working is introduced near the end of this section. After the pilot stage, decisions about sampling will be made. Among realists, there are good reasons for non-random sampling to be appreciated for the rich description which may emerge.The borderline between non-random and random sampling is questioned by realists (e.g. Olsen, 1994) who recognise that random sampling always occurs within a context of non-random borders or limits to the geographic area and time-period which is being considered. For instance, in the UK, many surveys exclude the region called Northern Ireland. However, the surveys include the north of Scotland and someone takes a non-random – that is, qualitative and judgemental – decision to do so. Having indicated that, in the research in Ghana, multiple units of analysis were relevant. Significant difficulties arose with random sampling at one level and non-random sampling at another level. The data used in this study were collected from a population of manufacturing firms in the RSSI sector through the use of a structured questionnaire survey administered in rural settlements throughout the Mfantseman District in the Central Region of Ghana. For reviews of the multiple causes of growth in RSSIs, see the following: Aryee (1977, 1981), Aryeetey (1996), Baggachwa and Stewart (1991), Barr (1995), Chuta and Liedholm (1985), Dawson (1993), Edusah and Tribe (1992), Farbman and Steel (1994), Fischer and Gerhard (1990), Hart (1973), Kilby et al. (1984), Liedholm (1990), Ninsin (1991), Nowak (1989) and Ried (1993).As part of Ghana’s decentralisation process and to ensure an even development, the district has been sub-divided into five zones. Zones are made up of a cluster of settlements and normally correspond with ‘traditional administrative’ areas. Traditional areas are autonomous with their own Paramount Chiefs. The Mfantseman District was chosen for the research for a number of reasons. First, the district represents a typical rural district.This is because over 90 per cent of its population are found in the rural area and in settlements of less than 5,000 inhabitants. Second, the district has both coastal and inland characteristics which offer a wide range of opportunities for the RSSI sector activities.Third, the district like most rural districts suffers from lack of formal sector investment and development and decline in agricultural production. Fourth, typical of rural areas, unemployment and underemployment are high because of lack of any formal sector employment opportunities.

Questionnaires in realist research 203 The survey was carried out over a period of six months, between May and October 1996. The district lacked any reliable data on the number of firms operating in the district which could form the basis from which a sample size could be drawn.The study employs a three stage sampling method to select the population of the RSSI in the district as is described in the following sections. A 40 per cent sample of the district’s settlements was used for this study. The District’s five zones (Ekumfi, Anomabo, Dominase, Saltpond and Mankessim) each have about 150 settlements. Forty per cent of the settlements in each zone was selected for the survey. A systematic random sampling method was adopted for choosing settlements. The sample size of 60 settlements (40 per cent) out of the total of 150 settlements is representative of the district as a whole. Without prior data on firms in the district an average of five firms per settlement was expected at the beginning of the sample selection. An estimated sample size of about 300 firms of RSSI was anticipated. However, it became apparent at the initial stage of the questionnaire administration that the number of firms in various settlements varied considerably. In the event, a total number of 215 firms were located, representing all the firms coming within the researchers’ definition in the randomly chosen settlements. There are limits to how far in a single research project one can compensate for the difficulties raised by multiple units of analysis. Whilst the researchers were unable to randomly or a representatively reflect all the sectors in Ghana’s small-scale industry, they nevertheless would claim that the findings for the sectors represented are highly interesting. The findings have policy relevance; the sector differentials are important for the people involved in the firms; and there seemed to be differences in the labour market and labour relations in different sectors. The firms were described as having three forms of ownership. Seventy-six per cent of firms were sole proprietorships, with little difference gender.Thus, 78 per cent of men and 71 per cent of women were sole proprietors. Fifteen per cent were family businesses (13 per cent among men and 20 per cent among women). The remainder were called ‘partnerships’. The partnerships were merged with the family businesses as being ‘joint ownership’ in the path analysis discussed here. However, the fact that respondents distinguished family from non- or extra-family ownership implies that the family boundary is important in making some business decisions. Merely labelling the firms with a simple taxonomy does not do justice to the complexity that is involved. Further research would examine the dynamics of joint savings, corporate accounts, marriage gifts, bequests within/without the family, sources of investment funds and inheritance. In the survey 15 per cent of respondents reported that their initial capital arose from family funds (11 per cent among men, and 23 per cent among women). Nearly one-third of respondents had private loans as their main source of start-up capital, suggesting that for many Ghanaian entrepreneurs the informal capital market is nearly as important as the

204

Wendy Olsen,Thandie M. Hara and Sampson Edusah

banking sector. These simple tabulated results from the questionnaire hint at further research topics.The study was conducted as if these topics were beyond its remit. However, it could be argued that they are very much within the remit, and should be researched using other sources besides the questionnaire. Interviews, special knowledgeable informants, ethnographic field notes, focus groups and oral histories of selected respondents are all techniques that potentially fit well with the time-consuming activities of survey enumeration. At the data-analysis stage, in particular, follow-up queries with a selected range of male and female respondents would have been justified. There was not a clear gender difference in the pattern of joint ownership of RSSI firms in this region. In the RSSI which were craft firms, 81 per cent of men and 50 per cent of women had sole ownership (implying that women in this sphere are more likely to be in a family firm), but there are so few women – just 3 out of 74 women’s RSSIs – that the percentages are perhaps misleading. Among the artisan RSSIs, 80 per cent of both men and women had sole proprietorship, and among the processing RSSIs, 78 per cent of men and 71 per cent of women. The use of partnerships was predominantly found among the processing industry firms, with both men and women using this form.Thus, although there are some big gender differences, there is not a rigid contrast between the way women’s and men’s firms are run. Finally, it should be stressed that no sector in one country is isolated completely from its connections with other sectors in other countries.Two particular connections seem important in the present context. First, those small-scale industries which try to export from Ghana face competition in rather different sectors in industrialised countries’ economies. For example, if a firm produces clay parts or crafts using local materials, they have to compete with industries abroad which produce similar products using artificial materials.The first problem faced by Ghana’s small-scale industries is that the small-scale craft and manufacturing sector in Ghana cannot, at present, compete on equal terms with its overseas equivalent. Second, the labour time for the small-scale industries has to be found in competition with the other activities which rural people need to engage in: planting, fishing, gathering, harvesting, transporting and trading; drawing water, cooking, watching children, getting firewood and so on. People may wish to migrate, they may wish to do farming work, and they may choose to help in the households and compounds. As a result the labour which is observed as ‘employees’ is a highly differentiated sort of labour. The hours of work in the RSSIs mean different things (and are probably worked differently in terms of intensity and focus) for cash-employees versus unremunerated family labour. The heterogeneity of workers in this particular context would be better represented using a labour market analysis which is not merely about commercial employment, but is about labour relations generally. This point is widely accepted among feminist economists and development economists. Thus, another way to improve upon the current questionnaire is to interview people about the intensity and pattern of their labour. Respondents can be asked to link up these decisions with their gender relations and the age-based

Questionnaires in realist research 205 and education-based differences in the pattern of work that is expected from each person. Ultimately, without getting deeply into the modelling debates about household labour and household decision-making, we need to know which discourses surround the work that is put into the ‘processing’ or ‘artisanal’ work: is it a discourse of productivity? of profit? of production for use? of employment, and if so is this seen as employment because people need employment or to facilitate the profit-making of the entrepreneur? The feminist literature on SMEs suggests that gender differences in the social practices of RSSIs can be expected. These social practices are embedded in discourses, and interviews or discussions are a central way to explore the subjective presentation of self that these discourses-in-use constitute. There is no reason except the fieldwork time-constraint not to combine a qualitative analysis of labour relations with the quantitative analysis of the survey questionnaire. Background on the possible integration of gender studies with RSSI can be found in Baud and Bruine (1993), Chuta (1983), Downing (1991), Elson (1990), Fapohunda (1983), Kabeer (1991, 1998), Schmink (1984), Steel (1981),Vyakarnam and Fiafor (1991), Ware (1983) and Wheelock and Baines (1998). A triangulated approach to research may be better than the survey method alone. Triangulation would involve the use of both qualitative and quantitative data (Olsen, 2003). In other words, it would be surprising if a single questionnaire were adequate for the analysis of partially commercialised labour relations. Triangulation can be strategically planned to make best use of the field researchers’ time and energy, and can combine interviews with a survey; or interviews with focus groups; or discourse analysis with path analysis; etc. In the present study, triangulation was not used explicitly in presenting the results. Areas for further research using such methodologies are pointed out. A carefully prepared and structured questionnaire was administered to each firm’s main owner during a period of about six months.The questionnaire can be seen in Edusah (2000). A two-week pilot survey was conducted in ten settlements to pre-test the questionnaire.Twenty-five enumerators were recruited and trained to administer the questionnaires.The enumerators were all teaching in schools in the district. Information proved extremely difficult to get on a number of different aspects of the RSSIs. Sensitive areas included the growth histories, the proprietors’ incomes and the profitability of firms. Some respondents felt very uneasy answering questions on sales. Respondents were under the impression that the data being gathered would be used to assess them for revenue collection by the District Assembly and the central government. Unfortunately, the field survey coincided with a revenue collection drive by the Mfantseman District Assembly. Repeated attempts to convince the respondents that the researchers were not tax officers succeeded only partially in allaying these fears. Many respondents also found it difficult to answer questions on the value of fixed assets, and most could not give details that might have allowed the researcher

206

Wendy Olsen,Thandie M. Hara and Sampson Edusah

to assess the profitability of the firm. It is possible that a commercial orientation is only part of the discursive placement of these firms, and that a productive orientation with a self-sufficiency discourse may be central to how people think about the enterprise. Instead of speculating about this, it appears that research should involve more open-ended, semi-structured interviews in order to find out how the people construct their enterprise activity rather than, as often happens in economics, merely imposing a profit-oriented commercial discursive framework upon the few details reluctantly given out by respondents. The very imposition of an alien framework may cause reluctance, whereas in semi-structured interviews the respondents are urged to use their own frameworks to expand upon and describe how they made decisions and what their aims are. While some of the problems with the use of the questionnaire to collect economic information in a partially commercialised sector have been outlined, the remainder of this section shows how this problem can become intensified when using the survey method to study women and men. In this research and elsewhere, women tend to become concentrated in the processing industry. This includes activities like food processing, beer brewing, oil extraction and confectionery. Unlike predominantly male activities like carpentry or blacksmithing, the products of most women’s enterprises are also daily household or personal essentials.Thus, women’s business products tend to face demand from within the household, and from customers.All studies for which data exist find that much of women’s increased income or loans themselves is spent on household consumption and children’s well-being. However this is not necessarily a simple one-way relationship of business providing for the home. Although the household may be consuming business funds or products, the business (especially food processing) may also be benefiting through the use of existing household utensils like pots, or plates and of course from free labour, which it does not pay for. The following diagram may be used to symbolise women’s financial expenditure pattern (Figure 11.1).

Business spending

Personal spending

Household spending

Figure 11.1 Patterns of women’s financial expenditure.

Questionnaires in realist research 207 Since men have better access to banks, and are engaged in enterprises that face less household demand, men’s financial transactions can easily be separated, either through having a separate business bank account, or through leaving the responsibility of household expenditure to women. This makes it possible, for an economic discursive framework to assess their business. In order to properly assess women’s businesses, there is need for a different mode of enquiry, one that is more suited to the contextual situation of women, and one that will reveal the complexity and reality of the situation. It is, therefore, not surprising that researchers find a gap between what the women take as benefits from the business, and what conventional economics defines as profit. For example, in her study of Bangladeshi women’s enterprises, Kabeer (1994) found that what women valued most from engaging in business was not the economic growth of their business or the possibility of accessing more lucrative public markets, but rather the fact that the businesses enabled them to work from their home, thereby differentiating themselves from poorer women. In economic terms, failure to access markets is regarded as a limiting factor for growth. Even where women are not explicitly prevented from accessing markets like in Ghana, women’s definition of a viable business may still include an assessment of how well the business enables them to simultaneously maintain the socially defined qualities of good motherhood/wifehood. Thus, the definition of profit, like in the Bangladeshi case, becomes much broader, encompassing both monetary and social gains. From a realist perspective, whether this is what the women should be striving for or not, is a different question from what they are actually striving for, and in order to affect change, one really ought to be investigating what is happening and not what ‘should’ be happening. Sole focus on economic aspects of business may undermine the real changes that are taking place on the ground, and will be inappropriate bearing in mind that from the start, women may have targeted multiple goals of social and economic gain. Even if it was possible to separate women’s businesses from their social context, the fact remains that many women in developing economies are illiterate (and even among small-scale industry entrepreneurs there is a significant gender gap.The absence of written records for many people, especially women, significantly undermines the quality of the financial information that might be obtained through a survey questionnaire. Retrospective surveys are notoriously unreliable because of imperfect recall of use of credit, economic activities, or levels of well-being. While this might also be a problem for men, it must be more so for women. It has to be appreciated that book keeping and perception of business in purely economic terms may not be generic to either poor women’s activities or their cultures. Where cash is a rare commodity, and not the only medium of exchange, payment may be accepted in kind, business goods may be bartered, or even given out free (in economic terms) but in exchange for the power and higher social standing that may be associated with being generous in impoverished societies.These outcomes are rarely expressed

208

Wendy Olsen,Thandie M. Hara and Sampson Edusah

in monetary terms, although they are costs for the business.While accounting and book keeping are useful skills for business, as long as the skill acquisition amongst women is still imperfect, economic modes of assessment will not provide the full picture of the situation. Some localised and appropriate method of inquiry is required for a better understanding of the workings of partially commercialised sectors, and their partially economically orientated entrepreneurs. Moreover, owing to the nature of businesses that poor people and women engage in, full economic assessment would produce discouraging results and most women’s businesses would not be rendered viable. Most of the women’s activities are highly labour intensive. Some traditional female enterprises earn negative returns when calculated at the rate of a male agricultural labourer (Hunt and Kasynathan, 2001, p. 45). However, Kabeer has found that although entrepreneurship increased women’s work loads, the fact that this was a product of their increased ability to contribute to household expenditure explained why it was experienced as a valued transformation of the terms on which they worked (Kabeer, 1994, p. 67).Therefore while the total labour would be on the cost side of an economic equation, such that the more intense it is, the less profit is realised, on the ground labour is being regarded as a positive aspect, regardless of its negative relationship with output and time taken. This then emphasises the point that the definition of profit and gain, and the understanding of business on the ground has both social and economic dimensions; as such it cannot be measured in economic terms only.

Regression techniques revisited The results were strongly gender-differentiated, even though men and women were spread across the three sectors and the three main firm types as described earlier. The undifferentiated results have already been presented in Edusah (2000) and in Olsen (1999). A dummy variable for gender was significant in numerous regressions relating to how the firms operate and what their growth pattern was. However, growth in turnover and employment of staff were not directly affected by gender so much as indirectly related to gender. Path analysis is uniquely capable of revealing such differences in ‘route taken’ by gender in ‘affecting outcomes’. However, because of basic differences in how the men operate, compared with the women, we decided to split the sample and reanalyse the data for each gender separately (see Diagrams 11.1 and 11.2). In this section we first describe path analysis, and then present the genderdifferentiated path analysis results. Path analysis involves a series of regression equations, each having one dependent variable, with links between the equations. The links are provided through intervening variables (also known as moderating variables or intermediate outcomes) which are simultaneously a Y-variable (dependent) in one

Questionnaires in realist research 209

Age

Seasonal operation 1.63 –0.74 –0.32

Education

Fluctuating sales

0.48 0.51

Manufacturing

0.53

0.63 0.17 0.27 –0.61

RSSI skills 0.56

0.24

0.55

Married

0.16

Employment

–0.23 2.64

0.30

Record keeping

–1.68

Sole proprietor

0.24

0.32

1.32 –0.58 New technology

Banking –0.49

–0.32

Raw materials local only

–1.38

Diagram 11.1 Factors associated with employment in women’s RSSIs, Ghana, 1999. The path diagram contains the Beta coefficients for each multiple regression. In the case of logistic regressions the coefficients are calculated using Menard’s method (Menard, 1995: 46(a); see Olsen, 1999). Source: Field data, 1996; see Edusah (2000) for details. Notes ‘Manufacturing’ refers to the processing sector and as a dummy variable its base case includes both artisan and craft industries.‘RSSI Skills’ refers to the proprietor’s prior experience being in the RSSI sector, with the base case being prior experience outside the sector.

equation and an X-variable (‘independent’) in another equation. Path analysis diagrams show the standardised regression coefficient along each X–Y path (Bryman and Cramer, 1997, chapter 10). The regression coefficients represent associations, and do not necessarily imply direct X–Y causation. Indeed, once the diagram is finalised we can see that X–Y1–Y2 and X–Y2 represent two simultaneous paths relating X to Y2.The first path, X–Y1–Y2, moves indirectly via the Y1 equation and thence to the Y2 equation; the second path is found in the Y2 equation where X appears as a ‘direct’ effect.A full discussion of direct and indirect effects is found in Bollen (1989) and other works on structural equation modelling (e.g. Arbuckle, 1997). The coefficients can be multiplied together along indirect routes from an X to a Y; they can be added together across the X’s under the assumption of additive separability. Path analysis has two main advantages over regression for cross-sectional statistics. First, it allows the analyst to reduce multi-collinearity within each equation by separating out different equations for separate causal mechanisms

210

Wendy Olsen,Thandie M. Hara and Sampson Edusah Seasonal operation Age

0.15

–0.29

Fluctuating sales

–.20 –0.20

0.25 Manufacturing 0.14

–0.30 Education

.30 0.14

–0.24

0.323

0.19 –.11

Sole proprietor

Employment

Record keeping –.25

–0.54

–.11

.14

RSSI skills 0.26

New technology

0.12 Banking

0.01 Married

0.18

–.14

Raw materals local only

Diagram 11.2 Factors associated with employment in men’s RSSIs, Ghana, 1999.The path diagram contains the Beta coefficients for each multiple regression. In the case of logistic regressions the coefficients are calculated using Menard’s method (Menard, 1995: 46(a); see Olsen, 1999). Source: Field data, 1996; see Edusah (2000) for details. Notes ‘Manufacturing’ refers to the processing sector and as a dummy variable its base case includes both artisan and craft industries. ‘RSSI Skills’ refers to the proprietor’s prior experience being in the RSSI sector, with the base case being prior experience outside the sector.

or domains. Each equation is improved relative to the flattened single-equation model that corresponds to the path analysis. Second, it allows the integration of logistic regression with linear regression. Standardised coefficients for logistic regression are available directly in SAS or by computation with other software (see Menard, 1995, p. 46 for details).Thus, a Heckman procedure, with one equation for the odds of participating and a second equation for the quantum of participation, illustrates the power of path analysis (Olsen, 1999). Regression itself can be critiqued for having a strong reliance on consistent, complete matrices of data about homogenous units, but if the reality is close enough to this representation then path analysis offers a rich source of coefficients which can be interpreted. In the case of RSSI in southern Ghana, there was gender differentiation which led to the need for separate men’s and women’s path diagrams. Thus, structural dissimilarities between the gender led to the splitting of the sample into two sub-samples.The gender-differentiation can be seen in the choice of detailed activities within the RSSI sectors. Specifically, each ‘sector’ had some

Questionnaires in realist research

211

gender-segregated activities.Activities which were 100 per cent male included wood-carving, leather-work, blacksmithing and net-making (all ‘crafts’ sector); carpentry, tailoring, smelting and foundry, canoe building and shoe-making (all ‘artisan’ sector); and finally grain milling, charcoal production, bee-keeping and talc processing (all ‘processing’ sector).The 100 per cent female occupations in the region were dressmaking (an ‘artisan’ sector activity), baking, confectionary, food processing and brewing. The last four activities fall into the ‘processing’ sector heading and they follow a classic gendered division of labour in which women pursue activities closely related to food-production and food-related services (Ehlers and Main, 1998). There were, however, some activities which both men and women ran. These include pottery, rattan, block-making, cassava processing, edible oil extraction, soap making and distilling. A question arises immediately: what gender were the employees of the RSSIs? The questionnaire did not have enough detail to answer this particular question. In order to handle such a query, a secondary questionnaire or a table at the level of the ‘worker’ is needed. This page might ask whether paid or unpaid; hours of work; gender, age and education; relation to RSSI owner; years in this job; years of total employment experience; whether part-owner; activity of main work; previous occupation; and pay.This page would imply a second unit of analysis. Normally in statistical packages the data at the second unit of analysis must be kept in a separate table from the main table on persons or firms. It is easy to link up these two tables by merging or aggregating so that additional variables can be placed at the right-hand side of the firms’ data table, for example, ‘Percentage Female Workers Including Proprietor’. The rather obvious point that multiple units of analysis are needed for studies of small-scale industries is now being made twice. First, we need worker-data along with firm-data; and later in this chapter we will argue that we need sector-data as well as region-data. Thus, at least four units of analysis exist; and if you separate out the personal characteristics of the entrepreneur from the holistic characteristics of the firm, you have five units of analysis. In our regression analysis, we have examined four units of analysis, or levels of reality, which are relevant for examining the RSSI entrepreneurs. First, their own socio-demographic background; second, characteristics of their enterprises; third, aspects of the markets in which they operate (the inputs, labour and output markets); and fourth, important overall outcomes. We have narrowed down the outcomes to two in this chapter: employment generation and turnover growth. Diagrams 11.1 and 11.2 illustrate the factors associated with high employment generation for men and for women RSSI entrepreneurs, respectively.The diagrams also focus on the use of banks, taking up the question of whether this is a mediating factor enabling growth, or whether it is instead an outcome of growth. It appears overall that the growth-effects of banking and employment creation are highly gender-differentiated. The associations have a different sign in some cases for men compared with the sign among women.

212

Wendy Olsen,Thandie M. Hara and Sampson Edusah

In the present section we describe the data at these four units of analysis before moving on to a brief description of the choice of activity and, finally, the presentation of path diagrams which illustrate the apparent causal processes that were operating among the 215 firms. We differentiate between causation affecting male and female entrepreneurs.The early findings (Olsen, 1999) support splitting the sample by gender (74 women and 141 men) because the causal model of associations is different. Putting in a gender dummy variable is not sufficient although it indicates a certain type of differentiation. However, when feminists argue against the use of dummy variable for gender in models, they tend to reject model-building altogether (Figart, 1997) in favour of qualitative analysis. One alternative is to build models that handle gender relations in sensitive and richly descriptive ways. Otherwise one may be throwing out the ‘baby’ with the bathwater.

Table 11.1 Variables and their means Definition

Years of formal education (0  none; 9  primary; 12  secondary) Whether skills were gained in previous RSSI work (1  yes; 0  elsewhere) Marital status Whether firm is in seasonal operation (1  yes) Sector is processing (1  yes; 0  craft and artisan) 1  yes; 0  joint or partnership 1  traditional; 0  new Source of raw materials within this district (1  yes) Whether firm keeps written accounts (1  yes) Whether firm uses bank account(s) Age of respondent who is the RSSI proprietor Number of full-time paid employees

Variable name as seen in diagram

Mean for men

Mean for women

Number of respondents Education

141

74

9.1

7.0

RSSI skill (Rural Small-Scale Industry) Marriage Seasonal

83%

73%

80% 41%

69% 55%

Manufacture

45%

66%

Sole proprietor Technology Local raw

78%

72%

54% 82%

72% 82%

Book keeping

33%

32%

Banking

73%

46%

Age

39 years

40 years

Employment

3.0 employees

2.6 employees

Questionnaires in realist research 213 Table 11.1 describes the main variables used in the path analysis. Please refer to Edusah (2000) and Olsen (1999) for further details of the research project, which had been begun without an explicitly realist or gender relations orientation (Edusah and Tribe, 1992). The path analysis presents the Beta coefficients showing the comparable impact on Y of a one-standard-deviation change in each X. The dependent variable in this case is the number of full-time paid employees the firm had in 1999. (Previous research examined the RSSI’s overall growth outcome over the period leading up to 1999.) Diagram 11.1 corresponds to women’s outcomes, and Diagram 11.2 shows men’s outcomes.

Description and involvement in the scene In the authors’ view, the decisions made about the early, middle and final stages of the survey approach depend in part on the past and future involvement of the researchers in the geographic or policy area being studied. For instance, our choice of firms depended heavily on Sampson Edusah’s identity and his past involvement in rural sector research and policy-making. The choice of hypotheses depended partly on the data which was available but was mainly influenced by the policy relevance of the projected findings.Thus, the authors introduced socio-demographic, input-market related, and management-related factors into a model because these relate to government and non-governmental policy and strategy areas. The factors were introduced theoretically at the stage before the questionnaire was produced.The literature review contributed to this selection, and the questionnaire thus partly reflects a deductive moment when the researchers decide to collect data corresponding to prior theoretical frameworks.The various factors were then measured empirically. It should also be reinforced that pragmatic decisions made during the research process, which have epistemic implications, will be partly based on personal details. The epistemic implications refer specifically to the choice of ‘objects’ about which measurements and claims are to be made. For instance, psychological factors were intentionally neglected in this study.The focus could always be changed. For instance, it appears in retrospect that triangulated enquiries could have provided even more policy-relevant and intervention-relevant material than a single questionnaire did.A brief review of such suggestions leads to the notion that different stages of research design could have different purposes. Table 11.2 suggests that any of four techniques could have been used to supplement or ground the survey-data analysis. It is common sense to argue for mixing research techniques in a complex design. However, the aim now is to enrich the learning from the project, not to re-validate or reconfirm the early findings. Thus, it would be a misunderstanding to represent triangulation as a confirmatory research design. It has exploratory and interpretive moments. These are highlighted in Table 11.2. In each case, the qualitative technique can be combined and compared with the findings of the questionnaire survey. This is all rather a long way from

214

Wendy Olsen,Thandie M. Hara and Sampson Edusah

Table 11.2 Potential contribution of additional techniques to RSSI study Intervention area

Technique that would contribute to knowledge

How it could contribute

NGO membership and leadership

Interviews (an inductive moment)

Government tax policy

Focus groups (an emancipatory moment)

Government income tax policy and pensions policy

Secondary unit of analysis in the questionnaire (a multi-level moment, using the proposed worker page) Expert local informants (and secondary data if it were available) (a critical moment)

Listening to discourses-in-use and contrasting them with discourses of academic theories Eliciting a sense of what is considered a fair mode of governance and the role of tax in that schema Gives insight into the conditions of workers and employees, paid or unpaid, and their backgrounds

Sectoral targeting of taxbreaks, technical support and public-sector investment in education and training

Gender-aware analysis of the barriers to technical change; sense of class accumulation strategies; gender-aware analysis of who decides where family investible funds get allocated

hypothesis testing per se. Hypothesis testing within the regression analysis is contextualised by the use of other techniques.We do not use the triangulation to confirm the findings of survey-data analysis, but rather to go beyond them (Olsen, 2003). Reflecting back upon the theory of enterprise growth, there is no guarantee that growing RSSIs will employ people. Growth in employment is related to, but not the same as, commercial growth in enterprise turnover.We cannot simply state that employment generation is better than an individual entrepreneur making profits from technology, since the latter may reinvest those profits or spend them on human capital of his/her self or family. In this study, employment generation was not well correlated with turnover growth. A number of ethical principals and means–ends relations are involved in assessing the policy implications. A policy preference for the job-creators over the stagnant enterprises may exist. If so, then it appears that a gendered hypothesis about the quality of firm growth can be put forward. This is that, in the present context, women were creating more jobs (although often only seasonally and often in agricultural processing rather than industry per se) than men entrepreneurs. Such a qualitative claim can be followed up by exploring (a) whether women can also create employment on a year-round basis (or are they busy doing other things at certain seasons), (b) whether men can also create employment as effectively;

Questionnaires in realist research 215 and (c) whether the employment is desirable from a human-development perspective, that is, is it training the employees, paying them acceptably well, and preparing them for further improvements in their lives. Although these issues were not pursued in the present research, they present an urgent programme for future enquiry.

Conclusion: auto-critique and fallibility In this section, the strengths and weaknesses of the case study research are summarised. On the one hand, it lacked a triangulated methodology which might have rounded out its interpretability in terms of gender relations, commercialisation and its discourses, and policy implications. On the other hand, the study was potentially multi-level with the RSSI proprietor, the firm, the sector, the market (e.g. input market; export market) and the country as levels; and multi-level

Table 11.3 Areas for improvement over single-technique questionnaire method Strengths

Weaknesses

Wide range of businesses and clear focus of attention on those which were ‘industrial’ are an asset

Policy relevance was downplayed in the sampling stage in favour of representativeness. Stratification might have been used at sampling stage Poor definitional focus on labour-in-firm and labour-from-household led to low-quality measures of labour inputs No path analysis can be representative at the ‘levels’ (units of analysis) which were not the focus of representative stage of sampling Gender issues at the macro level were not taken up explicitly either in the pilot stage, in the questionnaire, or in a triangulated research methodology. Investigation of women’s and men’s roles in agriculture, stereotypes and changing genderings of technique, and control over investment funds would have been a valuable adjunct to the questionnaire study Family roles, joint ownership and links with family’s capital are not well reflected or researched in this particular study.The atomistic use of a single questionnaire at person-level downplayed these complex realities at family-level and lineage-level

Choosing to focus on gender differences at the last stage responded to real differences observed (learned from) in the data Not using overly time-consuming qualitative procedures made it possible to develop the path analysis which offers a comparative study balancing the attention to each ‘entrepreneur’

216

Wendy Olsen,Thandie M. Hara and Sampson Edusah

research studies can be analysed using multi-level statistical software or using path analysis. Opportunities to clarify strategic choices using triangulation include the following: ●

●

●

●

in Ghana, gender-related choices are subject to policy influence, and can become a target of policy (especially banking policies and financial regulations); contrastive explanations require appropriate sampling, including nonrandom sampling at some levels and random sampling (enabling inference) at other levels; multi-level statistical analysis can be approximated with path analysis, but furthermore can be implemented using MLWin or other multi-level software, which is not very dependent on normal distributions to reach inferential or descriptive conclusions; qualitative data would offer further strengths to the analysis.

In summary, whilst the philosophical arguments are complex and abstract, the practical possibilities that open up for realist triangulated survey research are at our fingertips.They take time, but it seems worth incurring a cost in order to widen one’s experience of the society under study. A summary of the strengths and weaknesses of this research appears in Table 11.3.

Bibliography Allal, M. and Chuta, E. (1982). Cottage Industry and Handicrafts: Some Guidelines for Employment Promotion. Geneva, International Labour Office (ILO). Andy, J., Hyman, E. and O’Donnell (1991). Small technology – the key to increasing the productivity of micro-enterprises. Enterprise development, 2(2), 14–23. Anheier, K.H. and Seibel, H.D. (1987) Small-Scale Industry and Economic Development in Ghana: Business Behaviour and Strategies in Informal Sector Economics. Saarbrucken, Verlag Breitenbach Publishers. Arbuckle, James L. (1997). AMOS User’s Guide Version 3.6. Smallwaters Corporation, Chicago and SPSS Inc., Chicago. Aryee, G. (1977) Small-Scale Manufacturing Activities: A Study of the Interrelationships between the Formal and Informal Sectors in Kumasi, Ghana. WEP 23. Geneva, ILO. —— (1981) The informal manufacturing sector in Kumasi. In Sethuraman, S.V. (ed.) The Urban Informal Sector in Development Countries: Employment, Poverty and Environment.WEP Study. Geneva (ILO). —— (1984) Income distribution, technology and employment in the footware industry in Ghana. In Wouter van Ginneken and Baron, C. (eds) Appropriate Products, Employment and Technology: Case Studies on Consumer Choice and Basic Needs in Developing Countries. London, Macmillan.

Questionnaires in realist research 217 Aryeetey, Ernest (ed.) (1996) Small Enterprise Credit in West Africa. British Council/ISSER (Institute for Statisical, Social and Economic Research), Univ. of Ghana, Lagon. Baggachwa, M.D. and Stewart, F. (1991) Rural industries and rural linkages in subSaharan Africa: a survey. In Stewart, F. Lall and Bagachwa (eds) Structural Problems of African Industry. Barr, A.M. (1995) The Missing Factor: Entrepreneurial Networks, Enterprise and Economic Growth in Ghana. Oxford. University of Oxford, Centre for the Study of African Economics. Baud, I.S.A. and Bruijne (ed.) (1993) Gender, Small-Scale Industry and Development Policy. Exeter, SRP. Bhaskar, Roy (1975) A Realist Theory of Science. Leeds Books, Ltd., Leeds. —— (1997) A Realist Theory of Science (2nd edn). London:Verso. Bollen, Kenneth A. (1989) Structural Equations with Latent Variables. New York: Wiley. Bryman, Alan (1996, original 1988) Quantity and Quality in Social Research. London, Routledge. —— and Duncan Cramer (1997) Quantitative Data Analysis with SPSS for Windows: A Guide for Social Scientists, London: Routledge. Chouliaraki, Lilie and Norman Fairclough (1999) Discourse in Late Modernity: Rethinking Critical Discourse Analysis. Edinburgh, Edinburgh University Press. Chuta, E. (1983) Upgrading the managerial process of small entrepreneurs in west Africa. Public Administration and Development, 3, 245–83. —— and Liedholm, C. (1985) Employment and Growth in Small-Scale Industry: Empirical Evidence and Policy Assessment from Sierra Leone. Basingstoke, Macmillan. Dawson, J. (1993) Impact of structural adjustment on small-scale enterprises sector: a comparison of the Ghanaian and Tanzanian experiences. In Helmsing, A. H. J. and Kolstee, Th. (eds) Small Enterprises and Changing Policies. London, ITP. Downing, J. (1991) Gender and the growth of micro-enterprises. Small Enterprise Development, 2(1), 4–12. Edusah, Sampson (2000) Rural small-scale industry and Ghana’s economic development, PhD thesis, University of Bradford, Bradford BD7 1DP. Edusah, S.E. and Tribe, M.A. (1992) Rural small-scale industries in Ghana’s economic development: agenda for research, Bradford University, DPPC New Series Discussion Papers, No 26. (contact www.brad.ac.uk/bcid for a copy). Ehlers, T.B. and Main, K. (1998) Women and the false promise of microenterprise. Gender and Society, 12(4), 424–40. Elson, D. (1990) Male Bias in the Development Process, Manchester University Press, Manchester. Fapohunda, E.R. (1983) Female and male working profiles. In Oppong, C. (ed.) Female and Male in West Africa. London, George Allen and Unwin. Farbman, M. and Steel, W.F. (1994) Research issues for small enterprise development. Small Enterprise Development, 5(2), Figart, D.M. (1997) Gender as more than a dummy variable: feminist approaches to discrimination. Review of Social Economy, 55(1), 1032.

218

Wendy Olsen,Thandie M. Hara and Sampson Edusah

Fischer-Quincke and Gerhard (1990) Small Enterprises for the Needs of the People: Ghana’s Small-Scale Industrial Take-off in Research Group on African Development Perspectives. Berlin,Verlag Schelzky and Jeep. Hart, K. (1973) Informal income opportunities and urban employment in Ghana. Journal of Modern Africa Studies, 11(1), 61–89. ILO/JASPA (1989) From redeployment to sustainable employment generation: challenges for Ghana’s programme of economic recovery and development. Kabeer, Naila (1994) Reversed Realities: Gender Hierarchies in Development Thought, Delhi: Kali for Women. —— (1998) Money can’t buy me love? Re-evaluating gender, credit and empowerment in rural Bangladesh, IDS Discussion Paper No. 363, Institute of Development Studies, Univ. of Sussex. Kilby, P., Liedholm, C.E. and Mayer, R.L. (1984) Working capital and non-farm rural enterprises. In Adams, D.W., Graham, D.H. and Von Pischke, J.D. (eds) Undermining Rural Development with Cheap Credit. London,Westview. Lawson,T. (1997) Economics and Reality. Routledge: London. Liedholm, C. (1990) Small-scale industry in Africa: dynamic issues and the role of policy. In Stewart, F. Lall and Bagachwa (eds) Structural Problems of African Industry. Little, I.M.D., Mazumdar, D. and Page, J.M. Jr (1987) Small Manufacturing Enterprises. Oxford, Oxford University Press. Menard, Scott (1995) Applied logistic regression analysis, sage series quantitative applications in the social sciences, No. 106, London and Thousand Oaks, Sage. Ninsin, K.A. (1991) The Informal Sector in Ghana’s Political Economy.Accra,Assemblies of God Literature Centre Ltd. Nowak, M. (1989) The role of micro-enterprises in rural industrialisation in Africa. In Levitsky, J. (ed.) Micro-enterprises in Developing Countries. London, Intermediate Technology Publication. Olsen, W.K. (1994) Researcher as enabler: an alternative model for research in public Health. Critical Public Health, 5(3), 5–14. —— (1999) Path analysis for the study of farming and micro-enterprise. Bradford Development Paper No. 3, Development and Project Planning Centre, University of Bradford (available for £7.50 from B. Green, BCID, Univ. of Bradford, BD7 1DP). —— (2003) Methodological Triangulation and Realist Research: An Indian Exemplar. In Bob Carter and Caroline New (eds) Making Realism Work: Realist Social Theory and Empirical Research, London: Routledge. Onyeiwu, S. (1992) Graduation problems amongst MSEs in eastern Nigeria. Small Enterprise Development, 3(4), 45–50. Popper, Karl R. (1995 [1957]) The Logic of Scientific Discovery (5th edn). London: Routledge. Schmink, M. (1984) Household economic strategies: a review and research agenda. Latin America Research Review, 19(3). Staley, E. and Morse, R. (1965) Modern Small Industry for Development Countries. New York, McGraw-Hill Book Company. Steel,W.F. (1981) Female and small-scale employment under modernisation in Ghana. Economic Development and Cultural Change, 30, P153–67.

Questionnaires in realist research 219 Vyakarnam, S. and Fiafor, J. (1991) Cultural issues in enterprise development – a case study of Ghana. Small Enterprise Development, 2(3), 37–41. Ware, H. (1983) Female and male life-cycle. In Oppong, C. (ed.) Female and Male in West Africa. London, George Allen and Unwin. Wheelock, Jane, and Susan Baines (1998) Reinventing traditional solutions: job creation, gender and the micro-business household. Work, Employment and Society, 12(4), 579–601.

12 Critical realism and applied work in economic history Some methodological implications Brian Pinkstone

Introduction Prior to the Second World War economic history was largely practised by economists who had also been trained as historians.As a result they brought to their craft the traditional historian’s awareness of the central role played by broad socio-economic contextual factors for the explanation of historical events and the complimentary belief that good research required a sound acquaintance with the widest possible range of primary sources. After the war a new generation of economic historians, inspired by what they saw as the more hard-edged, scientific tools provided by econometrics, gradually came to dominate the discipline. Rather than context-specific explanation they looked to the use of statistical methods and hard quantitative data combined with economic and/or econometric models to provide precise quantitative apportionment of causal significance to independent variables and replace what were often seen as the fuzzy, impressionistic stories of the earlier generation. It will be argued in this chapter, on the basis of the critical realist critique of mainstream methodology in economics, that this turn of events has been to the detriment of the discipline of economic history and should be largely abandoned. The bold claim made by critical realism is that economic modelling techniques and econometric analysis, when they are used as a basis for socioeconomic explanation, are generally unsound and unscientific, because they attempt to apply an approach suitable only for closed systems to the explanation of processes taking place in systems that are inherently open.1 This chapter will outline the critical realist case, using examples from the author’s own research. In particular it will focus upon some of the research strategies both developed and found to be useful in the practical application of the critical realist methodology to research in the field of economic history.

The problems with the orthodoxy The central problem with the standard modelling/econometric approach in economics stems from an uncritical and often unrecognised commitment to the philosophies of logical positivism and methodological individualism.

Critical realism and applied work in economic history 221 Logical positivism has its roots in Hume’s empiricist position, and follows him in denying the possibility of reliable knowledge of causes (due to the problems of induction2) and instead relies overwhelmingly on empirical evidence. For this reason laws are only acknowledged to exist when there is absolute support for their existence at the level of the empirical.3 That is, laws are expressed as (strict) regularity statements of the form ‘whenever a then b’. The problem with this approach, as pointed out by critical or scientific realists such as Roy Bhaskar and Tony Lawson is that it confuses the closed system empirical measurement of an object of an analysis with the object itself. The object of scientific experimentation is generally, by means of some active intervention on the part of the researcher, to create a closed system in which the operation of the causal mechanism in which we are interested is closed off and isolated from all forces that might disrupt its operation: if complete closure is achieved then laws will take the ‘whenever a then b’ form. The explicit or implicit use of this conception of laws encourages mainstream economics to a methodological approach that Tony Lawson has defined as ‘deductivism’.That is, determinate closed system models are created based on a set of axioms of the ‘whenever a then b’ type. So in each instance when ‘a’ occurs it is immediately deduced that ‘b’ does. This approach means that the models economists and econometricians use, when they are trying to model economic relationships, are always closed system models.The use of stochastic models by econometricians does not overcome this problem, because a stochastic model by definition defines chance according to the sort of mathematically well-behaved distribution that can only be associated with a closed system. Since social reality is clearly an open system, where events are determined by a multitude of variable causal forces, the orthodox approach is likely to model reality correctly by chance only. The orthodox approach to modelling is also often buttressed by an implicit commitment to the atomistic conception of social reality embodied in the approach known as methodological individualism. In the neoclassical view individuals are seen as self-interested utility maximisers. Models are set up in which human agents maximise some function under a set of constraints. If the model is to have a determinate closed outcome, there can be only one solution. So rather than individuals being seen to have true free choice over their decisions, they are characterised in a manner that assumes robot-like behaviour: with there only ever being one rational choice (Lawson, 1997). There are, of course, many Post Keynesian and Marxist modellers who are not committed to methodological individualism. Nevertheless, the requirements of the modelling process and the need to produce deterministic outcomes means that they always adopt some set of simplifications regarding human behaviour that produces an equivalent to atomism. The normal assumption is that on an average agents may be viewed as acting in certain ideal-typical, conventional or socially determined ways. The outcome, however, remains that humans are modelled as if en masse they behave in strictly functional, robot-like ways.

222

Brian Pinkstone

By contrast, critical realism asserts that people are complex beings subject to varying brain states and changing philosophical views that can lead to different choices being taken under the same objective circumstances. Critical realism also recognises that humans are social beings who will be motivated by socially constructed desires (including political, ideological, religious and institutional beliefs) that lead to choices that may not represent the optimum outcome for the individual when measured purely in terms of self-interest.4 Moreover critical realism sees social factors realistically as always in a state of flux, conflict, change and development. So while some parameter describing social behaviour from the past might be estimated, it is not valid to assume that it would have remained the same in a different social context, that is, with different values for other parameters and variables, nor that it would be the same in the future, when by definition there will be new circumstances. Again the key problem with the orthodox approach is the attempt to model complex open systems by means of a closed systems approach.The critical realist position is that the openness of the social world is due to both its epistemological and ontological complexity (Rotheim, 1999). Its epistemological complexity stems from the almost infinite number of causal forces at work in the world, and their capacity for interfering and disrupting each other. The ontological complexity stems from the fact that individual agents are themselves complex causal mechanisms with a high degree of learning capacity and free wills (understood as the capacity to make free, rather than predetermined choices). Similarly social collectives and institutions are complex objects and they change, learn and evolve over time. The epistemological and ontological complexity of the evolving open social world means that in most cases the future is characterised by true uncertainty rather than any well-behaved, mathematically definable forms of risk or chance. The Post Keynesian economic theorist, Paul Davidson argues that the social world can be conceptualised in terms of ergodic and non-ergodic systems. Ergodic processes ‘can be fully described by a set of unchanging conditional probability distribution functions’(Davidson, 1996). This is not the case for a non-ergodic system. Davidson argues that true uncertainty should be comprehended ‘in terms of the existence of non-ergodic processes…[or] the absence of governing ergodic processes’.When dealing with strictly ergodic processes, past trends are an actuarially certain guide to the future (insurable risk), but in non-ergodic systems underlying parameters can and will change, so that the future may not at all reflect a statistical series from the past. Davidson goes on to argue that much of the socio-economic world is best understood in terms of non-ergodic systems, so that critical realists are right to argue that ergodic modelling approaches are inappropriate as a means for understanding many economic relationships (Ibid., 1996). A second problem that springs from deductivism is the misuse of strictly functional mathematical formulations to describe economic behaviour. Mathematics is a language that both enables and constrains its users. At its best it enables people to think with precision about functional relationships and via

Critical realism and applied work in economic history 223 simple modelling techniques permits certain concepts to be expressed and grasped in a concise and unambiguous form. In this way it may play an important role in enabling the social scientist to conceptually isolate a set of relationships of interest and explore in thought the implications of both definitional and causal relationships.This heuristic function of mathematics in economics is a useful intellectual tool, which can lead to real insights into the operation of economic mechanisms. During the twentieth century, however, the constraints associated with mathematical modelling and statistical methods have come to rob economics of much of its utility with respect to explaining real world phenomena. This is not simply the result of the over-development of theoretical models based on patently unrealistic assumptions (Hahn, 1994). The essential constraint is that although mathematics is precise it is also rigid. Once the relationship between two variables is conceptualised in a functional form, however complex, non-linear or chaotic, a given input produces a given output, in a rigidly predictable manner on every occasion. The critical realist critique of this approach is that economics is a social science that deals with a complex, continually evolving social world, peopled by complex, learning human beings, who can, and do, change their behavioural patterns continually over time. Consequently no functional mathematical formula will ever be able to encapsulate the way in which people will behave under all conditions, particularly the unknown and new conditions that will hold in the future.5 As an example of this problem consider the question of how the level of unemployment will be affected by a wage cut. Conservative economists would have it that a wage cut will reduce costs, increase profits and lead to a surge in business confidence that will in turn lead to a surge in investment, an expansion of output and a reduction in unemployment. Keynesians would argue that a cut in wages will depress effective demand, lead to a loss in business confidence, a reduction in investment and an increase in unemployment.The critical realist would not, however, be surprised to find strong evidence for both causal chains in different empirical circumstances, since firms, investors, workers and consumers might respond quite differently to wage cuts/rises in different social contexts. And social contexts are never precisely the same. Moreover, people have knowledge of earlier contexts and can change their behaviour when faced a second time with much the same circumstances. So human response patterns are always subject to qualitative factors that are inherently indeterminate and cannot be adequately grasped or generalised by a functional mathematical specification. Of course, after the event one can always fit a set of equations to the specific outcome that did occur, but the equations do not explain the data set, they simply mimic it. To explain what happened we need to move to the domain of the real causal mechanisms that we believe were at work and then reconstruct a plausible account of the way in which they interacted in the specific context. Of course the way we see and interpret the events will be influenced by our particular perspective and standpoint (Lawson, 1997).

224

Brian Pinkstone

A common reaction of mainstream economists to the critical realist position is that if precise mathematical functional relationships between variables are ruled out and standpoints matter, then we are lost in an open world where anything can happen, there are only subjective views, and the possibility of scientific analysis eludes us entirely.The critical realist position is confused here with the post-modern position, in which the complexity of the world combined with the importance of personal standpoints means that there are only different impressionistic stories told from different viewpoints and no one interpretation can be judged superior to another. This is definitely not the critical realist view. Indeed the whole point of the critical realist position is to continue the enlightenment project of the search for objective scientific truth, but to develop scientific methods that are appropriate for use in open systems. The key difference between the critical realist position and that of postmodernism is that the critical realism acknowledges the force and relative mindindependence of social structures.That is, the latter do not just depend on what individuals think of them, but rather are objectively embodied in social rules, conventions, habits, language, practices and especially legislation, the latter being backed by the coercive armed force of the state. Whilst social rules do change, they generally do so only gradually and at all times they structure, stabilise and constrain the socio-economic environment and the choices open to individual agents. So while the social world is open it is structured and it is not the case that just anything can happen, or that any view of the structures is equally valid.Views must be tested against the empirical evidence regarding real structures and some will be more objectively adequate than others. Indeed, the position that will be argued for here is that critical realist methodology produces accounts of the generative mechanisms, structures, tendencies and events of the real world, which are objectively superior to others.

Critical realist ontology For critical realists the world is conceptualised as structured by a series of domains: reality has ontological depth.The most immediately apparent domain is that of the empirical. Of the entire stream of events taking place in the universe the empirical is the subset that we perceive. The entire stream of events is defined as the domain of the actual. The domains of the actual and the empirical are generated by causal structures and mechanisms and the tendencies to which they give rise. Bhaskar defines the aspect of reality that consists of causal mechanisms and structures as the domain of the real. These deeper causal structures, mechanisms and tendencies are constitutive of the domains of the actual and the empirical, although they may not be immediately apparent to the senses. Despite the ontological and epistemological complexity of the world, practical science indicates that laws go on operating outside the closed systems of the laboratory.That is, they are transfactually operating even when they do not conform to the ‘whenever a then b’ formulation.6 Gravity continues to

Critical realism and applied work in economic history 225 operate upon Humpty Dumpty even if I interfere with its operation by catching him in the process of his tumble from the wall. Instead of the closed system, logical positivist conception of laws as determinist, regularity statements of the form ‘whenever x then y’, critical realism asserts that, due to the constant possibility of disruption in an open system, laws in open systems can only take the tendential form: ‘whenever x then y has a tendency to happen’. A law of tendency refers to the characteristic effect of a particular causal mechanism as it operates in the world (Lawson, 1997). The philosophical language developed by critical realism provides a way of thinking through the epistemological and ontological complexity of the world and distinguishing valid from invalid forms of argument with regard to causation. In this regard a fundamental problem with regularity determinism is that it fails to grasp the transfactual nature of laws and thereby falls into two widespread and related errors. First is the regularity error of thinking that a law must always reveal itself at the level of the empirical. Critical realists refer to this logical positivist position as empirical realism because while it is realist, it conflates the level of causal structures and mechanisms with the level of the empirical and constitutes a flat ontology. For example, when economists think they are dealing with a well-grounded regularity, such as the idea that the direction and volume of migration are determined mainly by wage differences, then they tend to believe that that must prove to be the case in all circumstances (otherwise the regularity would not hold).7 I have found that economic historians working in the field of migration studies almost always assume that wage differences drive migration and not only construct their models on this basis, but, on the basis of responses to my own evidence to the contrary, find it almost impossible to accept that there could be exceptions to this rule. The corollary to the regularity error is the error of pseudo-falsification (Bhaskar, 1975).This is to assume that if a set of circumstances occurs in which a law does not apparently hold then that law is invalidated in toto for all circumstances. As a case in point, in debates about the Prebisch–Singer thesis (that the terms of trade for primary products tend to decline relative to those for manufactures) the citing of particular (brief ) historical periods when the terms of trade for primary products have risen is taken as all that is required to prove that the tendency does not exist (Pinkstone, 2002). For critical realists, however, the absence of evidence for a tendency that is thought to be wellgrounded need not be the end of research, but rather, the trigger to a search for countervailing forces that may have disrupted or blocked the tendency on the occasion in question. The causal mechanisms that are at work in the world can be understood as operating in an ontology that is structured not only by domain, but also by a hierarchy of levels or fields in which particular sets of causal mechanisms and laws operate.These fields in fact usually constitute different and distinct fields of study. So we have physics, chemistry, biology, economics, linguistics, psychology, etc. (Bhaskar, 1975). Moreover some realms can be seen as providing the preconditions of existence of others, in a transitive manner. So the

226

Brian Pinkstone

physical material world supplies the preconditions of the biological sphere, which in turn provides those of the social. Each lower or more basic realm both constrains and enables the causal mechanisms that operate at the higher levels. But the structural laws operating at one level are relatively autonomous, they cannot generally be just read-off from those at lower levels, they must simply be consistent with them. Pinkstone (2002) argues that an important distinction between realms is that those at lower levels tend to be constituted by generative mechanisms and tendencies that have substantial degrees of endurance and extension through time and space. Such generative mechanisms and structures may be regarded as having ontological inertia, relative to those operating at higher levels and thus give rise to what Pinkstone (2002) has called robust tendencies. From the relatively short time-perspective of human beings the causal mechanisms operating at the most basic levels, for example gravity, come closest to creating the sort of regularities in the domain of the empirical that are sought by logical positivism. So although we know that the universe is evolving over time, from our perspective the movement of astronomical objects is highly regular and predictable. The biological level changes at a much faster rate, but again from the perspective of social and individual change and development, the generative mechanisms and the tendencies that are associated with the biological level have considerable ontological inertia.8 Put another way, individuals by and large must, by natural necessity adapt to the social environment rather than the reverse and human society must adapt to its physical environment rather than the reverse. Of course the relationship is dialectical,9 but human attempts to change their physical environment must be in accordance with and through the use of the pre-existing and ontologically prior causal structures, mechanisms and tendencies that operate at the lower and more basic levels. Taking this structured view of causal mechanisms provides one way of answering the question put to critical realism by Hodgson (1999) of how ontological priority may be established between competing explanations of an event or process. The answer is that one should proceed in general in a manner analogous to that of the Annales School of historical inquiry, considering the causal significance of physical, geographic, climatic and biological factors in explaining an event, before examining the role of causal mechanisms and tendencies associated with increasingly higher levels such as the economic, social, political or that of individual human agency. This causal hierarchy approach will be most relevant to helping us organise and clarify the explanation of the causal processes involved with respect to issues that apply across relatively wide time and space spectrums. Pinkstone (2002) has used this approach to rethink the Prebisch– Singer thesis: that the terms of trade for primary products tend to decline relative to those for manufactures. Pinkstone (2002) argues that while Engel’s Law (that the proportion spent on foodstuffs declines as incomes rise) is the immediate explanation for the tendential decline in primary product prices, the explanation of this explanation is a biological constraint on the

Critical realism and applied work in economic history 227 capacity of humans to consume foodstuffs. This is the causal mechanism that then produces a series of related robust economic tendencies under capitalism, and deteriorating primary product prices is but one of this ensemble. The above constitutes a contrastive that relates to a wide time and space continuum by historical standards. Economic historians, however, are often concerned with more historically specific questions. And as the time– space continuum in which we are interested narrows, the possibility, that even the most robust economic tendencies may be disrupted or negated by a temporary conjuncture of less robust tendencies, increases. In other words, if the issue in which we are interested is relatively time and place specific, then it is more likely that the efficient cause in the Aristotelian sense is related to some higher-level structure such as a political or ideological factor or even simple individual human agency (the action of a ‘Great Man’ etc.).The time and space scope of the question that we seek to answer will thus set the context that determines which structural levels are most pertinent to the problem. Pinkstone’s (2000) work on UK migration in the period 1900 –1930 provides an example. This is a quite historically specific time–space frame, so the possibility that robust economic tendencies will be disrupted by historically specific factors is therefore high. It is contended that the tendency for migration to be driven by wage differences is quite robust, because under capitalism wage rates are a vital and enduring factor determining the overall quality of life for human beings. But unlike regularity determinism, which, as is noted above, leads orthodox economic historians to have difficulty believing that there could be circumstances in which wage differences did not play the crucial role, critical realism encourages one to accept that in an open system there will always be exceptions to every rule, and pursuing these exceptions may lead to exciting new discoveries.The path to those discoveries, however, is likely to be conceptually closed off to investigators using the closed system, flat ontology of the mainstream approach.

Critical realist epistemology and methodology The orthodox econometric approach to explanation in economic history consists of developing a model based upon whatever set of explanatory variables the researcher thinks will be most relevant to a particular historical period and question, and then gathering whatever quantitative data can be found to plug into the model, or vice versa, the existence of a range of quantitative data relative to a particular period leads the researcher to the development of what he sees as an appropriate model. In developing the model, the researcher will often consciously or unconsciously adopt a principle of parsimony, since large-scale models are fraught with technical problems and smaller models are generally considered more elegant. To the extent that a principle of parsimony is followed, however, important explanatory information is likely to be excluded. On the other hand, even if the modeller develops a large model with many explanatory variables, at some point the system will need to be closed off in

228

Brian Pinkstone

terms of the number of explanatory variables that will be included.This always involves the possibility that crucial explanatory information will be excluded. In addition, although attempts may be made to find quantitative proxies for qualitative causal factors, in general significant qualitative factors that do not lend themselves to this approach are either dealt with by means of introducing structural breaks, which tend only to emphasise the explanatory gaps and inadequacies in the modelling process, or they are simply ignored. The general result then of the closed system approach to explanation in applied fields such as economic history is a tendency to close out or exclude information, which then leads to information poor explanations. By contrast, as we shall see, critical realist methodology provides new tools of enquiry that are appropriate to the investigation of causal relations in open systems and, due to the fact that critical realism explicitly acknowledges the epistemological and ontological complexity of the world, the explanatory framework it encourages is holistic in nature and thereby will tend to produce information rich explanations. Critical realist methodology starts with the proposition that although the social world is complex and open, it is structured in a way that gives rise to what Tony Lawson has defined as demi-regularities at the level of the empirical. Demi-regularities is the more precise critical realist term for what positivists call regularities. Recall that regularities must be strict and measurable. In a closed system, once a regularity is measured, we can predict that when it occurs in the future it will have precisely the same value. Strict regularities, however, are unlikely to exist in a complex open system characterised by true uncertainty, due to interference from both known and unknown countervailing tendencies. Nevertheless, the fact that economic reality is structured by physical, biological, sociological constraints (including language), institutional conventions and cultural habits and customs, means that the world is characterised by quite persistent demi-regularities. So for instance, when Christians attend church, they tend to do so on Sunday.10 The demi-regularities consistent with such well-known habits are immediately explicable. On the other hand, the demi-regularities, which do not appear to be consistent with what we think we know about the world, Tony Lawson refers to as contrastives.11 Precisely because they do not conform to what we would expect to be the case, given the context, they invoke a sense of surprise and, as a result, should provide a stimulus to research aimed at explaining their existence. In turn the tendencies that give rise to the contrastives might be called the contrastive tendencies, for it is these that the researcher will initially aim at explaining. Even before attempting an explanation, however, the contrastive methodology encourages the researcher to seek to contextualise the contrastive demiregularity to the greatest extent possible.The whole point of the contrastive is that it highlights a particular facet of the socio-economic context in which it is situated. But that does not mean that the contrastive process ends there. Once we have focussed on a particular contrastive, we can proceed to contextualise it further, by contrasting the contrastive information against other relevant data from similar, broader, narrower or different contexts.The adoption of such an

Critical realism and applied work in economic history 229 approach from the start is clearly inclusive of information and begins the process of developing an information rich explanation. For example, Pinkstone’s (2000) research into migration from the UK in the early twentieth century was stimulated by data that showed a striking increase in migration to the Empire in the decade before the First World War, and an equally striking collapse in migration to the US.12 The economic historians who have dealt with this phenomena from a mainstream ‘closed system’ perspective, have not attempted to situate this contrastive, if they noticed it all.13 That is, leading scholars who have looked at migration in the period to the US, Canada and Australia, respectively, tended to do so in terms only of factors relevant to each nation, and thereby missed both the role of general push factors in the UK at the time, and the socially constructed qualitative differences between the attractors to empire versus those for the US.14 By emphasising from the start, contextualisation and comparison, the contrastive approach potentially leads to the consideration of a much wider range of evidence than does the orthodox closed system modelling approach. The next step in the process of developing an explanation is for the researcher to try drawing upon those realistic causal mechanisms that he or she thinks can help account for the contrastive tendency. Critical realists denote this process as retroduction, in the sense that one tries to go behind the empirical evidence to the domain of the real causal mechanisms that generate it.15 For the economic historian there is usually a plethora of possible explanans, and the researcher will attempt to discover evidence regarding the extent to which each possible causal mechanism was operative.This is basic historical detective work. Again though the critical realist will be open to considering a much wider range of evidence than will the orthodox economist, because the more holistic approach of critical realism means that the researcher cannot close out any possible factors, especially qualitative information relating to social and political influences and the role of human agency. Once a tentative hypothetical explanans is established, it needs to be tested as thoroughly as possible in three distinct ways. First it needs to be tested for reasonableness against the initial contrastive evidence itself. Is the hypothetical explanation realistic in terms of what we think we know about the world? If not it should be either rejected and alternative explanations sought or we should seriously consider whether our general view of reality needs reworking. One holistic way of testing our hypothetical explanans is to check on its capacity to fit with not only the contrastive demi-regularity but also the wider context within which that demi-regularity is situated. So we might ask; ‘if the hypothesised explanans exists, would its existence be expected to be associated with other tendencies and contrastive demi-regularities?’ If this is the case, empirical research should reveal their presence. Secondly, in the attempt to create the most adequate realistic and holistic account possible critical realism requires the researcher to ‘explain the explanation’ in theoretical terms.That is, to think through the conditions of existence of the explanation itself. Do these hold? Is the explanation of the explanation

230

Brian Pinkstone

realistic? Does the explanation of the explanation suggest that we are dealing with a robust contrastive tendency, or a case in which less robust, historically unique countervailing tendencies have temporarily overridden some other more robust tendency? Where countervailing tendencies are involved, are these intrinsic to the primary causal mechanisms and structures or are they essentially extrinsic and historically specific? Does the contrastive tendency point to the primary causal mechanism and tendencies involved, or is it of a secondary nature? If the latter then does the whole issue need to be reconceptualised with respect to the primary causal mechanisms and what Roy Bhaskar has called the generic tendency, that is a tendency that sums up the overall impact of a variety of related causal mechanisms and structures (Bhaskar, 1993).16 Once we have thought through the theoretical issues involved we should again check back against the empirical evidence for new contrastive demiregularities that would be consistent with the types of parallel tendencies that the overall theoretical framework suggests should be apparent if the explanation of the explanation holds.When this is done successfully, the researcher should have at hand a relatively vast array of empirical circumstantial evidence that is consistent with the hypothesised explanation. Finally, the hypothesised explanans needs to be contrasted with the competing explanations of the phenomena in which we are interested, if such explanations exist. If the critical realist methodological approach has been successful, it will produce an account that is information rich and should therefore have considerably more explanatory power than that of research based on the orthodox approach. It is on this basis that critical realists assert that the paradigm provides ‘superior explanatory adequacy’(Lawson, 1997).

Conclusion During the last fifty years or so the discipline of economic history has fallen under the spell of econometric modelling techniques.This has led to a plethora of explanatory accounts with a one-sided emphasis on quantitative data, and a tendency to beg more questions than they answer. Critical realism provides a number of ways out of this cul de sac. Firstly, the deep ontology of critical realism permits a clarification of many of the errors of logic that spring from the flat ontology that underpins econometric work. Secondly, rigorous application of the philosophical concepts of critical realism can shed entirely new light on old questions in the discipline, by re-conceptualising the causal processes in a way that that is not available to the conceptual framework of empirical realism. Thirdly, the contrastive methodology of critical realism offers the promise of opening up entirely new paths of research, many of which lie outside the purview of the closed system modelling approach. Unlike the orthodox methodology, which concentrates on creating internally rigorous and consistent closed system models, that in applied work lead to information poor explanations of the empirical world, the critical realist methodology provides a holistically rigorous and consistent open system, information rich explanation of both the actual world of

Critical realism and applied work in economic history 231 appearances and the real causal mechanisms at work within it. Finally, critical realism provides a new way forward for economic history because, above all else, it is a practical philosophy of science, which emphasises empirical evidence, but provides a superior, more comprehensive and appropriate, ontological framework and methodological approach for dealing with explanation in open systems.

Notes 1 Seminal works in the critical realist approach are Bhaskar (1975) and Lawson, (1997). 2 Induction involves generalising from a sample. So an Australian might say, ‘All the swans I have seen are black, therefore all swans are black.’ Clearly induction has its problems as a logical tool. 3 On this basis critical realists at times refer to the logical positivist position as ‘empirical realism’. 4 So altruistic behaviour, such as fighting and dying for a cause, is explicable from the critical realist perspective, but incomprehensible from the standpoint of methodological individualism, due to its assumption that rational behaviour is entirely self-interested. 5 The best they can do is provide illustrations of a range of consequences which would occur if people were non-learning, robot-like and functionally predictable. 6 Notably, for this reason, towards the end of his life, perhaps the leading logical postivist of the twentieth century, Popper (1990) came to conclusions very similar to those of critical realism. See also Runde (1999, pp. 63–82). 7 This point is elaborated with reference to UK migration in the early twentieth century in Pinkstone (2001). 8 Human behaviour that is tightly determined by biological needs may give rise to the sort of robot-like, regularity behaviour characteristic of an ergodic system. So for example, in very cold climates electricity consumption has a close inverse correlation with temperature. In such cases econometric modelling may be legitimate, see Pinkstone (2000). 9 Understood in the sense of the relationship between different and opposing forces, in which one must be regarded as dominant for reasons of natural necessity. 10 Indeed, social conventions and habits are at least in part the means by which humans attempt to impose order and regularity upon an uncertain world (see Lawson, 1997). 11 The contrastive approach would seem to be very compatible with Marxian dialectics, in that a contrastive can be understood as a contradiction in underlying causal processes that has become manifest at the level of the empirical. 12 In 1900 the US took almost 80% of UK migrants and Canada and Australia around 20%, by 1913 the position was reversed (Pinkstone, 2000). 13 For example, although David Pope has written numerous articles on migration from the UK to Australia in the early twentieth century, at no point does he acknowledge the significance of the collapse of migration to the US, let alone attempt to explain it. See especially his key paper, Pope (1981). 14 Pope (1981) for Australia, and for Canada and the United States see Thomas (1973). 15 Critical realists also use Alvin Hansen’s term abduction in this context. Meaning that in the case of the development of an entirely new explanatory framework researchers may abduct ideas from quite different fields of enquiry: a process that can rely heavily on the use of metaphor. See Lewis (1999). 16 Bhaskar (1993, p.78). As an example see Pinkstone (2002) where it is argued that with respect to the Prebisch–Singer Thesis, the relevant generic tendency is towards a steady contraction in the proportion of world trade (by value) accounted for by primary products.

232

Brian Pinkstone

References Bhaskar, R. (1975) A Realist Theory of Science, Harvester Press, Sussex. —— (1993) Dialectic:The Pulse of Freedom,Verso, London. Davidson, P. (1996) Economic theory and reality, Journal of Post Keynesian Economics, 18, 479–508. Hahn, F. (1994) An intellectual retrospect, Banca Nazionale del Lavoro Quarterly Review. Hodgson,G. (1999) Marching to the Promised Land? Some doubts on the supposed theoretical and policy implications of critical realism, Workshop on Realism and Economics, Kings College, Cambridge, April 26. Lawson,T. (1997) Economics and Reality, Routledge, Cambridge. Lewis, P. (1999) Metaphor and critical realism, in Steve Fleetwood (ed.), Critical Realism in Economics: Development and Debate, pp. 83 –101. Routledge, London. Pinkstone, B. (2001) ‘British migration to North America, 1900–1914: a radical revision’ Joint EHANZ, AHA, AMHA Conference , Kalgoorlie, September. —— (2000) ‘Underlabouring Post-Keynesian Economics’ in Alethia, 3(1), 45–48. —— (2002) Persistent demi-regs and robust tendencies: critical realism and the Singer–Prebisch Thesis, Cambridge Journal of Economics, 26(5): 561–583. Pope, D. (1981) Modelling the peopling of Australia’, Australian Economic Papers, December, 258–281. Popper, K. (1990) A World of Propensities,Thoemmes Press, Bristol. Rotheim, R. (1999) ‘Post Keynesian Economics and realist philosophy.’ Journal of PostKeynesian Economics, 22(1), 71–103. Runde, J. (1999) Popper, probabalities and propensities, in Steve Fleetwood (ed.), Critical Realism in Economics: Development and Debate, pp. 63–82. Routledge, London. Thomas, B. (1973) Migration and Economic Growth: A Study of Great Britain and the Atlantic Economy, 2nd edn, Cambridge University Press, Cambridge.

13 Critical realism and the political economy of the Euro Philip Arestis, Andrew Brown and Malcolm Sawyer*

Introduction The institution of the single currency in the European Monetary Union raises many complex and important economic, political and social issues.The investigation and analysis of those issues raise in turn the question of method in economics and social theory. In this connection, the coming to prominence of critical realism within radical political economy and methodology circles is a welcome development as it seeks to address issues of method.This chapter will, firstly, discuss some of the main problems that critical realism raises for empirical work and inferences alongside an approach adopted to try to overcome these problems; and secondly it provides a concrete illustration of these arguments with reference to a research project on the analysis of the single European currency that the authors have been conducting (see Arestis et al., 2001).1 It is argued that critical realism provides a method partially appropriate to concrete levels of analysis – illustrated by the example of the attempt to explain the falling value of the Euro – but that the critical realist method is inappropriate to the most abstract and fundamental levels of theory – illustrated by the example of the fundamental propositions of Keynes, Kalecki and Marx. This argument draws from, attempts to illustrate and thereby develops, the characterisation and critique of critical realism developed in Brown et al. (2002).

Methodological problems of critical realism Within the economics discipline, critical realism has become associated with debates concerning the validity, or otherwise, of using econometrics and mathematics within economics and social science.Accordingly, this issue is addressed first, if only to state briefly that the debate has little, inherently, to do with critical realism. The notion that critical realism is somehow inherently ‘anti’ econometrics has arisen through the important work of Tony Lawson and has been developed by a group of economists originally based at Cambridge University (see Lawson, 1997).These authors argue that econometrics, as commonly practised within the economics profession, should be rejected on critical realist grounds. They also argue that economic theory, as that term is understood within mainstream economics, that is, mathematical modelling,

234

Philip Arestis, Andrew Brown and Malcolm Sawyer

should likewise be rejected in the name of critical realism. For example, Fleetwood (2001) appears to rule out any role for ‘functional relations’, that is, relations of the form y  f (x), in economic and social theory.2 Pratten (1999) argues that Sraffian economics cannot be compatible with critical realism due to its use of mathematical modelling. Lawson (see 1997, pp. 69–85) appears to rule out multiple regression analysis, that is, the method that is central to mainstream applied economics. In short, there is a wholesale rejection by some authors of formalism in economic and social theory, a rejection that is allegedly based upon critical realist grounds. The term ‘formalism’ can be taken to refer to the use of ‘formal systems’, such as those within mathematics and probabilistic statistics. Though subtle, complex and impressive, we will not discuss the content of the ‘critical realist’ argument against formalism, an argument that turns on the question of constant conjunctions of events.3 The argument of the Cambridge group appears to make a very strong general claim namely, that all formalism is always wrong when employed in social and economic theory.This goes way beyond what, in the authors’ view, would flow in a straightforward manner from the tenets of critical realism. Not only is this argument contentious from within the perspective of critical realism, it is also an argument that goes way beyond what is necessary in order to critique mainstream economics. All that is needed is to argue that formalism cannot be the main tool of economics and social theory, contra the mainstream view. It should be noted that Lawson (1999, p. 8) appears to affirm that critical realism, on his interpretation, does not lead to the blanket ban on formalism that he has sometimes been interpreted as advocating. However, the impression, at least, has been given that the Cambridge group rules out formalism tout court.An implication of this is that articles defending the use of econometrics and more generally formalism within a critical realist framework continue to grow in number – to which this volume is testimony.4 Perhaps a consequence of the argument that formalism should be rejected has been to lead debate away from issues such as that regarding the contribution of method to substantive economic theory and policy. In contrast to other contributions to this volume, therefore, these issues are discussed in the remainder of this chapter. The key issues that critical realism (as opposed to a specific group of critical realists) raises are more general than simply the question of formalism. Critical realism offers important and useful insights to the practising social researcher. Unlike the best known (to economists) philosophies of science, or the current focus on ‘rhetoric’ and ‘postmodern’ stances in general, critical realism emphasises that real social structures exist (not merely agents, and not at all ‘rational economic man’) and that they mesh together in complex but potentially identifiable ways. In other words, the social (including economic) world is a ‘stratified’ one made up of complexly ‘layered’ social structures which are reproduced and transformed through agents. Method in economics and social science should be adapted to the peculiarities of this object at hand, rather than simply ape the methods of the natural sciences.5

Critical realism and the political economy of the Euro 235 However, a key problem is evident as soon as the attempt to theorise on the basis of critical realism is made. It soon becomes clear that the specifics of the social ‘vision’ or ontology that it advocates are elusive. Just what does it mean to say that the social world is made up of stratified social structures? How does one go about uncovering these structures and theorising about them, their ‘stratification’ and about the events that they co-generate? Answers are difficult to come by from within critical realism.To make matters worse, the traditional view within critical realism is that the lack of methodological specificity is a virtue of critical realism, that is to say critical realism is ‘ontologically bold and epistemologically cautious’ (Bhaskar, 1989, p. 176). Critical realism ‘underlabours’ for the social scientist, clearing away, for example, the dogma that all theory must adopt formalism, but it has little positive suggestion to offer by way of method because few general methodological precepts exist. Rather, on this traditional critical realist view, it is up to the social scientist to adapt methods to the specific object that interests them, to theorise the stratification of particular social structures and specific instances of their reproduction and transformation by social agents.6 The only such general methodological advice that is offered by critical realism reflects this cautious approach towards epistemology and method. The extremely general method of ‘retroduction’ is advocated. According to this method, scientific explanation proceeds by way of hypothesising ‘deep’ social structures and mechanisms, from initially given ‘premises’.These ‘premises’ are not provided by the scientist, rather they are apparent phenomena in the object of study. For example, words and concepts in use in everyday discourse may provide the starting point even where the word may take on a more precise meaning in scientific analysis and where the conceptualisation evolves. Money provides an example. It is a term in common use and for which statistical measures exist, but where economists have continued to use the word (albeit with definitions which differ between themselves and from the commonly used definitions) and have conceptualised money in a range of ways. More generally, the premises for hypotheses may be puzzling or otherwise surprising occurrences, such as, for example, the long-standing differential productivity records of the UK and Germany (Lawson, 1997, in particular, emphasises the role of surprise as a catalyst for social research).The extreme generality of this method is evidenced by the fact that very little guidance is offered as to what premises the social scientist should employ nor how hypotheses from these premises should be generated, beyond generalities such as ‘use real rather than bogus abstractions’, ‘employ analogy and metaphor’.The problem is that, even if one accepts that retroduction is, or should be, employed then it is the chosen premises of retroduction that are crucial. Clearly, there are lots of given possible premises, lots of things known, and hence the crucial problem is how to abstract from things known, and conversely how to interconnect all that is known. It is therefore, particularly worrying that the role and nature of abstraction is underexplored within critical realism (see Brown et al., 2002). If science is to proceed by way of retroduction then it must first ‘abstract’ premises from the complex

236

Philip Arestis, Andrew Brown and Malcolm Sawyer

social world from which to begin. How is one to undertake ‘real abstractions’? Critical realism offers little in the way of advice. How, furthermore, is the social scientist to reincorporate the facets of the social world that are abstracted from? Answers to these questions are vital if the stratification of structures and mechanisms within the social world is to be comprehended. But critical realism offers no help; it is all up to the practising social scientist!

An alternative perspective Brown et al. (2002) suggest that the critical realist notions of retroduction, real abstraction and so forth, do have some resonance at relatively concrete levels of analysis where it is indeed the case that little by way of general advice can be given to the investigator. But at more abstract and fundamental levels where, for example, one considers the nature of the market economy, or the nature and impact of ‘globalisation’, or the theorisation of global economic crises (indeed economic crisis in general), and more generally the nature of capitalism as such, they argue that critical realism is not only of little use but wrong. At this level the investigator is by no means engaged in hypothesising some new entity or structure; rather, quite to the contrary, the investigator attempts to fathom the mode of interconnection of the vast array of practices, institutions and structures that are already quite obviously known to exist.We know firms exist, we know there are markets, prices, crises, States, interest rates, commodities, wars, etc. How do they interconnect? This is a very different question to that which critical realism suggests we pose. It is not a matter of what particular unknown structure exists to explain this particular given phenomenon, rather, what is the interconnection of the vast array of known structures, mechanisms and events? At one level it might appear that the conception advocated here, namely the social scientist as, fundamentally, an ‘abstractor’ and ‘interconnector’ of given social structures, mechanisms and events is only superficially different from the critical realist view that social scientists employ retroduction, that is,‘hypothesis’ of new, previously unknown social structures and mechanisms. For, it could be argued that the ‘hypothesis’ of an unknown social structure or mechanism is simply the critical realist terminology for the procedure advocated here, namely, the interconnection of the vast array of given social material. However, retroduction is supposed to proceed from just one or two given premises, chosen by the investigator. The question posed, according to this view, is one that involves only the specific phenomenon that has aroused curiosity. It is not a question that is concerned with interconnection or synthesis of a whole range of different phenomena; all other phenomena are abstracted from. Given only the surprising phenomenon of interest, the investigator is then recommended to hypothesise, perhaps using analogy and metaphor, a structure that will specifically account for the phenomenon in question. Thus (i) critical realism provides little guidance as to how to ‘do’ abstraction except by way of finding a ‘surprising’ phenomenon and hypothesising specific structures (in abstraction

Critical realism and the political economy of the Euro 237 from other structures) to account for this phenomenon; (ii) interconnection, or synthesis, occurs only after abstraction and retroduction have taken place; it is not itself a part of this process, on the critical realist account.7 Hence retroduction is inappropriate for the fundamental level of analysis, where the investigator must fathom the interconnection of a vast array of known social material.These themes will be illustrated and developed via the example of the project on the single European currency.

Analysis of the single European currency The authors’ Euro project has been ongoing for some four years now. A book (Arestis et al., 2001) and many journal articles and book chapters have been published thus far, all examining various facets of Economic and Monetary Union (EMU) within the European Union (EU). In what follows the chapter focuses on Arestis et al. (2001), as this book is representative of the project as a whole. First some of the features of the analysis are highlighted, followed by a brief summary of the main conclusions of the book. Second the way in which the analysis and conclusions illustrate the argument above will be drawn out. The analysis of the formation of the European Monetary Union contains a range of features which should be highlighted. First, it is concerned with the analysis of a unique and major event (the formation of EMU) as a result of which the future cannot be seen as similar to the past. Thus, those methods which rest on a basic similarity between past, present and future can only be used sparingly if at all.8 Second, the creation of the Euro as a virtual currency in 1999 and as a ‘real’ currency in 2002 was viewed as the culmination of long process of economic and political integration within a range of European countries (beginning with the six countries which were the original signatories of the Treaty of Rome and expanding to the current fifteen members of the EU).The analysis had to show at least some awareness of the political and institutional developments leading to the formation of the EMU. Third, the investigation was also intended to be forward looking, concerned with questions such as the desirability of the UK’s membership of the EMU and alternative policy arrangements governing the operation of the EMU.This part of the investigation required a theoretical orientation (so as to judge desirability or otherwise of the UK entry) and seeking to forecast the future (under the assumption of continuation of the present institutional arrangements and under alternative arrangements). Here, particularly, an attempt was made to clearly state the theoretical propositions by which the EMU and the UK membership was to be judged, and to make clear the nature of the theoretical framework being used.This framework could be portrayed as Keynesian with an emphasis on the role of aggregate demand and the operation of the forces of cumulative causation. Fourth (as already illustrated in part) numerous methods were employed – historical analysis of the process of economic and monetary integration within the EU, the development of alternative policy proposals, the use of descriptive statistics, some use of econometric (multiple

238

Philip Arestis, Andrew Brown and Malcolm Sawyer

regression) analysis, etc. One of the main conclusions reached was that the single currency as currently implemented does not promise to deliver continuing non-inflationary economic growth in the EU, and the book presents a sustained argument to that effect. It argues that the economic impact of the Euro and its accompanying institutions, the European Central Bank and the Eurosystem, is likely to be deflationary and destabilising; that the political impact is profoundly undemocratic; and that the social consequences are thus likely to be deleterious. It is not argued, however, that the concept of a single European currency is inherently flawed. On the contrary, a Keynesian alternative to the institutions that currently underpin the Euro are proposed. Chapter 2 adopts a historical approach, setting the inception of the Euro in the context of the post-war history of the attempts to forge economic and monetary union within Europe.The point is made that, when, in the context of deregulated capital markets and the Single European Market the institutional structure that currently underpins the Euro was laid down, neo-liberal monetarist ideas exerted an important influence on policy. At the same time the Bundesbank was considered a model for the European Central Bank.Thus, the chapter stresses the historical reasons for the fact that the institutional structure accompanying the Euro conforms to what Arestis and Sawyer (1998) have termed the ‘new monetarism’. Chapters 3 and 4 broadly examine the fiscal and monetary aspects of the institutions surrounding the Euro.The relevant literature within economics, both empirical and theoretical, is critiqued from the particular Keynesian perspective adopted and developed within the project. Chapter 5 examines the individual money demand functions within the eleven countries that initially entered the single currency in order to highlight the disparities between these countries.9 Chapter 6 examines the impact of the Euro since its inception in the context of the search for an explanation for its declining external value. It shows how the most recent evidence concerning the macroeconomic impact of the Euro lends support to the arguments of the previous chapters and, in so doing, provides an empirically robust explanation for the fall in value of the Euro. In the literature, the steadily weakening Euro has often been attributed to the peculiar strength of the US economy, rather than to any inherent difficulties of the imposition of the single currency in Europe. If the Eurozone itself is focused upon then this is only to stress the ‘inflexibility’ of European markets (particularly the labour market).The research finds the widespread notion that labour market inflexibility is at the root of the so-called ‘Eurosclerosis’, highly unconvincing. Instead it is argued that it is not US strength alone, but this strength in combination with structural weaknesses of the Eurozone endogenous to the imposition of the Euro that has caused an outflow of direct investment capital, leading to the exchange rate decline. By ‘structural weakness’ the research does not refer to inflexible labour markets but to the situation analysed in previous chapters: the imposition of the single currency, without the appropriate (Keynesian) institutions, onto an area which is in a divergent state, with low growth and high unemployment.The most recent evidence shows that the

Critical realism and the political economy of the Euro 239 divergent state of the Eurozone has persisted, and may even have worsened, since this imposition. Thus, the argument that the institutional structures underpinning the Euro must be replaced by the proposed Keynesian alternative is well supported by the recent evidence on the impact of the Euro, and goes some way to explaining the instability of the Euro exchange rate. How, then, does the Euro project illustrate the arguments regarding method and critical realism developed above? The first matter discussed is how the project illustrates the apparent utility of critical realism at the more concrete level of abstraction where the falling external value of the Euro is located. Second, discussion then elaborates on how the more fundamental features of the project do not readily lend themselves to a critical realist interpretation. Whilst undertaking the task of trying to explain the falling external value of the Euro (see Figure 13.1) the authors became struck by the resonance between the research processes undertaken and the critical realist description of, and prescription for, social science. Firstly, consider the object or phenomenon under investigation.The falling value of the Euro, ongoing for over one year as work began on the project, had occasioned much surprise amongst economists and commentators more broadly. Many, especially the proponents of the Euro, such as Buiter (1999), had expected the Euro to rise significantly from its opening rate of $1.18. In the event the Euro had fallen and continued to fall way below what most economists considered a value that it ‘should’ have had, had it behaved according to economic ‘fundamentals’ (such as purchasing power parity, a rate consonant with trade balance, or similar). How does this relate to critical realism? It appears to fit very closely to the critical realist notion of the first stages of social scientific investigation, especially as developed by Lawson (1997). This is a clear example of what Lawson calls a ‘demi-reg’.

1.2 1.15 1.1 1.05 1 0.95 0.9 0.85

Figure 13.1 Daily exchange rates: US dollars per Euro.

04/07/2002

04/05/2002

04/03/2002

04/01/2002

04/11/2001

04/09/2001

04/07/2001

04/05/2001

04/03/2001

04/01/2001

04/11/2000

04/09/2000

04/07/2000

04/05/2000

04/03/2000

04/01/2000

04/11/1999

04/09/1999

04/07/1999

04/05/1999

04/03/1999

04/01/1999

0.8

240

Philip Arestis, Andrew Brown and Malcolm Sawyer

For over one year the Euro had been falling regularly. This was not a strict regularity because the Euro obviously did not fall every day. Rather, it was a ‘rough and ready pattern of events’: when put on a graph one could clearly see a general pattern of a steadily falling Euro through time. What is more, on Lawson’s critical realist account, such demi-regs often take the scientist by surprise and it is this element of surprise that may be the catalyst for scientific investigation, the spur to search for an explanation of what appears prima facia both significant and surprising. This seemed exactly to describe the research situation into which the authors, along with many other economists, were placed.The ‘demi-reg’ of a continually falling Euro had taken the authors, and others, by surprise or had at least occasioned interest and initiated the search for the ‘underlying’ causes of this puzzling ‘surface’ phenomenon. Just as it appeared that the starting point resonated strongly with critical realism, the subsequent process of enquiry also fitted into this framework. Reading the relevant literature, and examining the relevant empirical evidence (e.g. Eurozone and US national accounts, as well as economic performance indicators), it seemed clear that the research involved a process of trying to unearth, to ‘hypothesise’, the structural cause of, or condition for, the manifest demi-reg of the falling Euro.The research also aimed to empirically evaluate the adequacy of our hypothesis relative to competing hypotheses. This was indeed what many other economists were trying to do. It was clear too that the mainstream insistence upon formalism, both in terms of econometrics and of mathematical modelling, was detrimental to the collective effort of trying to explain the Euro’s fall.There was arguably some use in the econometric work.This work suggested, for example, that the Euro was indeed undervalued relative to ‘fundamentals’. But there was quite obviously (to anyone but a dyed-in-thewool mainstream economist) too much emphasis upon such econometric techniques to the detriment of other empirical methods (e.g. case study work) and realistic theory development, that is development of an adequate theory of the causes of the manifest demi-reg in question (the decline in value of the Euro).There was also, of course, much in the way of mathematical modelling, or in mainstream parlance, ‘economic theory’. Again there was arguably some use in these models, in that they explored quantitative implications of some more or less plausible hypotheses as to the relationships between a few of the variables involved (e.g. the relationship between interest rates and exchange rates as in the interest rate parity result or in ‘overshooting’ models). However, by no means did they give the basis for hypothesising the causal mechanism required and, in fact, were fundamentally misleading through their incorporation of the dominant and false theoretical perspective of new classical or new Keynesian economics (i.e. the perspective that ‘markets work’, at least in the medium to long run, so that, for example, monetary policy has real effects only within a two year horizon, after which its effect is purely nominal). The character of the researchers’ solution to the problem of what caused the Euro to fall also corresponds to critical realist notions. Consonant with a number of other authors, it was argued that an important proximate cause was the

Critical realism and the political economy of the Euro 241 large magnitude of capital outflows from the Eurozone to the US and elsewhere. This hypothesis fitted in with the available evidence better than other hypotheses, for example, the timing and magnitude of the outflows was consonant with the timing and magnitude of the exchange rate fall, whereas the timing and magnitude of interest rate differentials was not. However, this was only one possible mechanism at work and as yet lacked structural grounding: what combination of social structural factors and individual activities combined to sustain the falling Euro through causing, inter alia, capital outflows from the Eurozone? Here it was suggested that the flaws in the institutional structures surrounding the Euro, detailed in previous chapters of the book and empirically supported in Chapter 6 by examination of the latest evidence confirming the divergent state of the Eurozone, underlay investors’ perceptions that the Eurozone was weakened by the single currency.These perceptions led to the systematic (system wide) outcome that market participants invested outside of the Eurozone. Investors, and the media, often recite the mainstream mantra that the lack of ‘labour market flexibility’ is the reason that the single currency is inappropriate. Such sentiment, it is suggested, may be in part an attempted expression of the authors’ diagnosis of the problems of the Eurozone, based on realistic economic theory and analysis but, we further suggested, this attempt is irrevocably distorted by the hegemony of mainstream economic theory. To summarise, a combination of social structures, mechanisms and their reproduction by agents, all endogenous to the inception of the Euro, were hypothesised to account for the fall in its external value. The hypothesis was shown to be robust against the evidence in a way that alternative hypotheses were not. This illustrates how, at the relatively concrete level at which our object of explanation was pitched, namely, that of the declining external value of the Euro, the critical realist method has a strong resonance. However, there were more abstract and fundamental aspects to the attempted explanation of the fall in value of the Euro, aspects not readily accommodated within the critical realist framework. These are described below as an illustration and development of the critique of critical realism detailed above (and developed in Brown et al., 2002). The question regarding the external value of the Euro was addressed as part of the broader Euro project as a whole. This broad project was explicitly Keynesian in character drawing upon, and forming part of, a specific approach to radical political economy. The analysis of the weaknesses of the Eurozone that were endogenous to the inception of the Euro, necessary for the explanation of the falling Euro, drew crucially upon this abstract and fundamental view, in opposition to new classical and new Keynesian economics (and specifically in opposition to ‘new monetarism’). These theoretical preconceptions were particularly relevant when it came to: (i) the policy proposals developed, whereby it is recognised that monetary and fiscal policy can and do have real effects in the long run; (ii) the approach to evaluating the Maastricht convergence criteria (although we also drew upon the literature on optimal currency

242

Philip Arestis, Andrew Brown and Malcolm Sawyer

areas); (iii) the approach to evaluating convergence/divergence subsequent to the establishment of the Euro in January 1999; and (iv) the general proposition that the unfettered capitalist economy is fundamentally crisis prone.The belief that unfettered capitalism is crisis prone is a theoretically grounded proposition, one drawn from Keynes, Kalecki and Marx. This illustrates that, in practice, theorists do not start by ‘retroducing’ from some surprising ‘premise’, rather the theorist is guided by more abstract and fundamental theoretical propositions, as in our case provided by Keynesian, Kaleckian or Marxist theory. Thus, if it is to be argued that critical realism is fully consonant with our research then it must be argued that Keynes, Kalecki and Marx employed an approach consonant with critical realism. Do the fundamental and crucial features of the theories of the above mentioned ‘greats’ of economics truly correspond to critical realism? In particular, is it really the case that these economists arrived at their results via a critical realist method (even if that method is very general)? Obviously, such questions are the subject of a great deal of ongoing research, which cannot be reviewed in detail here.10 The argument of Brown et al. (2002), already drawn upon previously will be developed, in order to address these questions in the subsequent paras. Instead of identifying a particular form or demi-reg to be explained, and then ‘hypothesising’ a previously unknown structure to account for this form or demi-reg, it seems more plausible to consider that the fundamental propositions put forward by Keynes, Kalecki or Marx are gained from fathoming the general mode of interconnection of the manifest and essential features of the economy and society.When working at this abstract and fundamental level, the criterion of science must be the degree to which the theory developed successfully accounts for the many fundamental features of the economy, features already known to the investigator and so in no need of hypothesis.They are also features that are essential to the economy and so in no way is there a need for hypothesis of yet more hitherto unknown essential structures and mechanisms. It is especially important to reiterate that the critical realist procedure is supposed to entail a method whereby first abstraction takes place: a form or demi-reg is identified and structure hypothesised to account for it. Second only may synthesis occur, where the theorist attempts somehow to piece together all these myriad appearances and hypothesised structures and mechanisms. But this is manifestly not what occurs at the fundamental level of theorising. Rather, there is an attempt to ‘see the wood for the trees’, to fathom the interconnection of masses of known structures, mechanisms and events. In the case of the Euro, the researcher must have an ability to interconnect commodity, money, profit, wages, interest, State, foreign trade, world market, etc. and this process of interconnection or synthesis by no means occurs subsequent to a prior process of abstraction and ‘hypothesis’. For example, the notion of the circular flow of income is an attempt to fathom how various known individuals, mechanisms and structures (consumers, firms, prices, wages, etc.) are interconnected through monetary and real flows, and thereby constitute the economic system as such. On the basis of even

Critical realism and the political economy of the Euro 243 this elementary and abstract conception of the mode of interconnection of the economy, it is possible to elaborate the proposition that planned savings need not equal planned investment, such that the possibility of demand deficiency exists.This is a simple illustration of how aspects that are originally known to the theorist can be comprehended in a new light once their mode of interconnection is fathomed.Abstracting the mode of interconnection of structures, mechanisms and events, thus, does yield new knowledge to the investigator, but this is not some previously unknown entity, rather it is a new found comprehension of the function of given entities or aspects within the system that they constitute. Wages, for example, can be grasped as not merely costs to individual firms but as a vital component of aggregate effective demand within the system overall.This is due to a comprehension of the mode of interconnection of wages with the many other manifest aspects of the economy, rather than to hypothesise from a surprising demi-reg. Brown et al. (2002) similarly argue that Marx’s fundamental propositions, such as the existence of exploitation, stem from an attempt to fathom the mode of interconnection of crucial factors such as the distinction between labour and labour power and the monopolisation of the means of production by the capitalist class. Wages, in this case, are newly revealed to mask exploitation.Thus Keynes, Kalecki and Marx are not, at a fundamental level, hypothesising new entities from one or two demi-regs, rather they are, in sometimes very different ways, and with correspondingly different results, focusing upon (abstracting) the mode of interconnection of the manifold economy and society. In this way, Keynes, Kalecki and Marx were most fundamentally involved in abstracting from, and simultaneously attempting to fathom the interconnection of, a vast array of known social material.11 This is where their great merits lay, rather than in some ability to ‘hypothesise’ yet more entities, previously unknown, and out of sight. Keynes, Kalecki and Marx provide a way into abstraction and interconnection, they give us an orientation. Given that basis, which is the crucial aspect of investigation, then at a more concrete level, as illustrated above by the example of the falling value of the Euro, one can see that there is some role for ad hoc factors such as surprise, and for speculative hypotheses of the hitherto unknown, as advocated by critical realism.Thus, one can see the appeal of critical realism, its resonance with day to day social scientific activity, whilst also appreciating that critical realism does not actually provide the correct account of, or prescription for, the features of science at its most fundamental levels. In short, critical realism blurs the distinction between different levels of abstraction.At a more concrete level, illustrated earlier by the example of attempting to explain the falling external value of the Euro, the notion of retroduction, involving ‘hypothesis’ and ‘test’, where abstraction occurs prior to synthesis, carries some degree of plausibility. At the fundamental level, however, which sets the context for concrete research, a very different process involving, from the outset, interconnection or synthesis of a vast amount of known material occurs.This has been illustrated by the example of the theory of the capitalist economy employed in the Euro project.

244

Philip Arestis, Andrew Brown and Malcolm Sawyer

Conclusion The authors’ work on the EMU proceeded at several levels. The analysis was built on a basic conception of the nature of the capitalist economy within the tradition of radical political economy, but one that recognises the importance of the changing specificities of capitalism such as the evolving economic and monetary integration within the EU. But it was also necessary to understand the political thinking which surrounded the construction of the EMU and its institutions that were identified and labelled as a ‘new monetarism’. Further, as the authors sought to engage with the alternative analyses of, for example, the declining value of the Euro, it was necessary to understand the analytical frameworks on which the discourse surrounding the Euro is based. It has also been the case that the analysis of the EMU occurred at differing levels of abstraction. An approach such as critical realism which, by its own admission, suggests a highly general method valid across many different levels of abstraction, leaving everything else up to the practising social scientist, is undoubtedly going to offer little help to the investigator. Furthermore, it is likely to be misleading or wrong at one or other level of abstraction.The authors’ work on the EMU has been drawn upon in order to illustrate the argument that the critical realist method is inappropriate to the fundamental level of abstraction, where the key propositions of economics and social theory are formulated, though it is partially applicable to more concrete levels. This applicability, when combined with other merits of critical realism (such as its stressing of a layered social ontology), may explain, and to some extent legitimise, its popularity amongst radical economists and amongst other social scientists. An alternative method dubbed ‘systematic abstraction’ that is, unlike critical realism, appropriate for the abstract and fundamental level of theory is developed in Brown et al. (2002). In this chapter an attempt is made to illustrate and develop the characterisation and critique of critical realism offered by Brown et al. (2002) with reference to work on the EMU.

Notes * Philip Arestis is at the Levy Economics Institute of Bard College; Andrew Brown and Malcolm Sawyer are at the University of Leeds. We would like to thank Gary Slater and David Spencer for very helpful comments on an earlier draft of this chapter. 1 This project has been financed by the Levy Economics Institute of Bard College, New York. We are grateful to the Levy Institute for generous financial support, especially to its director Dimitri Papadimitriou. 2 Fleetwood (2001, p. 202) writes, ‘[m]y central thesis is that the use of functional relations and laws in economics is fundamentally misconceived’. 3 Williams (2000) is right when drawing upon an interpretation of Hegel, to see formal logic as ‘transcended’ by, rather than abolished by, dialectical logic.This does not entail the complete abolition of formalism that the Cambridge group seem to favour, but it still stresses that it (formalism) cannot be the main tool of social and economic theory, that at best it can play a supporting role.

Critical realism and the political economy of the Euro 245 4 Outside of the field of economics, debates regarding critical realism and formalism have also occurred. Doug Porpora, for example, has long argued that critical realism is compatible with multiple regression analysis under certain circumstances in social science (see Porpora, 2000). 5 Rotheim (1999) elaborates on all the various ways in which critical realism is useful and, in particular, emphasises how critical realism ties in with Post Keynesian economics. 6 Brown (2001) elaborates on the underlying rationale behind this critical realist stance, namely the critical realist distinction between a ‘scientific ontology’ and a ‘philosophical ontology’. 7 As Lawson writes,‘(t)he point of abstraction is to individuate one or more aspects, components, or attributes and their relationships in order to understand them better. Once this has been achieved it may be possible to combine or synthesise the various separate understandings into a unity that reconstitutes, or provides a better understanding of, the concrete’ (1997, p. 227, emphasis added). 8 The only example which comes to mind would be to use the projection of past experience as a counterfactual with which the present or future (eventually when it has become the past) can be compared. 9 In keeping with the argument above, the use of econometrics is but one facet of the Euro project as a whole and not one that critical realism offers anything but the most general of guidelines for, hence it is not discussed here. 10 For the debate regarding Keynes and critical realism see, for example, Lawson (1994) and Parsons (1996). The relationship of Marx to critical realism is debated in Brown et al. (2001). 11 Note that, at this fundamental level, abstraction involves both analysis and synthesis: the theorist focuses upon (analyses) the mode of interconnection (or synthesis) of the object under study. Brown et al. (2002) elaborate upon this point, a point that is characteristic of dialectics.

References Arestis, P., Brown, A. and Sawyer, M.C. (2001) The Euro: Evolution and Prospects, Cheltenham, Edward Elgar. —— and Sawyer, M.C. (1998) ‘New Labour, New Monetarism’, Soundings, 9(Summer): 24–41. Reprinted in European Labour Forum, 20,Winter (1998–99): 5–10. Bhaskar, R. (1989) The Possibility of Naturalism, 2nd edn, Hemel Hempstead, Harvester Wheatsheaf. Brown, A. (2001) ‘A Comment on Dow’s “Post Keynesianism and Critical Realism: What is the Connection?” ’, Journal of Post Keynesian Economics, 23(2): 349–56. ——, Fleetwood, S. and Roberts, J. (eds) (2001) Critical Realism and Marxism, London, Routledge. ——, Slater, G. and Spencer, D. (2002) ‘Driven to Abstraction: Critical Realism and the Search for the “Inner Connection” of Social Phenomena’, Cambridge Journal of Economics, 26(1): 773–88. Buiter, W. (1999) Six months in the life of the Euro: what have we learnt?, Remarks prepared for a seminar on monetary and budgetary policy in the Economic and Monetary Union, held at the Rabobank, 25 June 1999. Fleetwood, S. (2001) ‘Causal Laws, Functional Relations and Tendencies’, Review of Political Economy, 13(2): 201–20. Lawson,T. (1994) ‘Economics and Expectations’, in Dow, S. and Hillard, J. (eds) Keynes, Knowledge and Uncertainty, Aldershot, Edward Elgar.

246

Philip Arestis, Andrew Brown and Malcolm Sawyer

Lawson,T. (1997) Economics and Reality, London, Routledge. —— (1999) ‘Connections and Distinctions: Post Keynesianism and Critical Realism’, Journal of Post Keynesian Economics, 22(1): 3–14. Parsons, S. (1996) ‘Post Keynesian Realism and Keynes’s General Theory’, Journal of Post Keynesian Economics, 18(3): 419–43. Porpora, D. (2000) Review Essay: Critical Realism in Economics: Development and Debate, edited by Fleetwood, 1999, London, Routledge, Alethia, 2: 30–4. Pratten, S. (1999) ‘The “Closure” Assumption as a First Step: Neo-Ricardian Economics and Post-Keynesianism’, in Fleetwood, S. (ed.) Critical Realism in Economics: Development and Debate, London, Routledge. Rotheim, R. (1999) ‘Post Keynesian Economics and Realist Philosophy’, Journal of Post Keynesian Economics, 22(1): 71–104. Williams, M. (2000) ‘No escape from the “Post-Hegelian” Dialectic’, Science and Society, 64(3): 337–65.

14 Presenting demi-regularities: the case of Post Keynesian pricing1 Paul Downward and Andrew Mearman

on a practical level, one must ask how … demi-regularities are found.When is a demi-regularity not a demi-regularity? (Mearman, 1998, p. 167)

Introduction In Chapter 7, the authors argued that econometrics can play an epistemological role in economic enquiry from a critical-realist perspective. In contrast to mainstream economics, Chapter 7 argued that critical-realist empirical enquiry could draw upon an eclectic range of research methods including quantitative techniques.As conventionally defined econometric methods are only one form of empirical practice, it was argued that quantitative methods including econometrics can help to codify the empirical level, in presenting demi-regularities requiring further investigation. As such, econometrics can play a part in a triangulation strategy which seeks to legitimise causal claims arrived at by retroduction. In this chapter, these arguments are illustrated by drawing upon applied work in Post Keynesian pricing. In this sense the chapter continues the constructive dialogue between critical realism and Post Keynesian economics that began in Chapter 7.This dialogue develops as follows. In the next a historical case is offered that is suggestive of the merits of econometrics and is an exemplar of the triangulation strategy referred earlier. The example is the work of Gardiner Means on Administered Pricing.This discussion re-emphasises one of the main themes raised in Chapter 7, that practice can inform philosophy. Subsequent sections then present discussion of more contemporary material. In the section ‘critical-realist epistemology: a constructive dialogue with Post Keynesian economics’ addresses the consistency of critical realism with other methodological positions that have been embraced or espoused by Post Keynesian economics. In as much that many of the arguments of Chapter 7 offer a transcendence of some of the ‘duals’ of critical realism, this section argues that one needs to transcend some of the differences that are presented methodologically in Post Keynesian economics, to produce a consistent basis of inference. The section ‘Empirical examples from UK manufacturing’ then

248

Paul Downward and Andrew Mearman

presents more contemporary examples of triangulation in practice. Examples are drawn from both referring to broader literature and from uniting different methods of analysis in a specific context.Though not attempting to be exhaustive, the chapter endeavours to illustrate how econometrics can be useful to research framed within a critical-realist perspective.

Administered prices as retroduction Gardiner Means’ work on Administered Prices is particularly relevant to the current discussion for two reasons. On the one hand, both Lee (1998) and Downward (1999) identify Means as central to Post Keynesian pricing. On the other hand, though not expressed as such, this chapter argues that, viewed through the lens of a Post Keynesian economist influenced by critical realism, Means can be seen to provide a historical example of triangulating methods of analysis, which can now be understood as an attempt to present demi-regularities and then to produce a causal explanation of them through retroduction. A central theme in Means’ work was to investigate the real reasons why prices did not behave in the way predicted by market-clearing economics. Consequently, Means was concerned with the statistical behaviour of prices and in particular that some prices were sticky in response to changes in the economic cycle. To analyse this issue, Means explored Bureau of Labour Statistics (BLS) wholesale price data for the period 1926–1933. Essentially, 617 product groupings were grouped into 10 product categories and their price behaviour was examined. Means (1935) concluded that a bifurcated pattern of behaviour emerged for different industries.Agriculture and non-manufacturing sectors were characterised by price flexibility, whereas manufacturing was not. In the language of critical realism, it can be argued that Means identified a contrastive demi-regularity with reference to descriptive statistics. To explain this behaviour, Means argued that generally, in agricultural and non-manufacturing sectors, markets set prices, whereas, in large modern corporations, prices are ‘administered’ rather than set in the ‘auction’ conditions implied in neoclassical price theory. Administered prices are prices that are fixed by administrative fiat before transactions occur and are held constant for periods of time and for sequential sets of transactions.2 Thus Means (1962) illustrates that pricing is carried out in terms that are analogous to full/normalcost pricing. Prices are set by a mark up on costs, for a series of transactions, with markets determining output rather than price. The mark up is based on uncertain expectations of long-term profit. This implies that prices will not readily change with contingencies such as changes in demand and hence output. Means termed this perspective the ‘flow principle of output’. Once again, in the language of critical realism, it can be argued that this aspect of Means’ work, conducted by detailed case-study, enabled the retroduction of the actual determinants, or causal processes, of pricing. It was argued in Chapter 7 that Keynes’ views on statistical work shows him to be concerned with similar issues that occupy critical realism (see also

The case of Post Keynesian pricing 249 Lawson, 1989). It is argued here that it is also clear that by referring to a wide body of evidence, Means’ attempted to explain, or make inferences, about why prices behaved as they did in the manner espoused by critical realism and as discussed in Chapter 7. Results described in terms of an association between empirical events – that is demi-regularities – are explained in terms of historical and more qualitatively described causes. However, crucially, Means begins his analysis constructing by eye a simple ‘regression’ analysis to calculate a relationship between the magnitude of price changes and industrial concentration as a proxy for administered pricing processes.3 The slope coefficient indicated that in a period of generally declining prices, that is, during the Great Depression, the magnitude of price changes was positively associated with the degree of concentration in the industry. Typically, thus, manufacturing was associated with less responsive changes in prices to the business cycle. To the extent that one can identify Means’ work with a critical-realist strategy of research, Means’ approach illustrates the possibility that econometric work can also be consistent with critical-realist epistemology in adding weight to inferences drawn using other research methods. The reminder of this chapter further illustrates the potential of econometrics to play this role.

Critical-realist epistemology: a constructive dialogue with Post Keynesian economics Methodological consistency Before discussing the empirical examples, some discussion is desirable on the wider nature of Post Keynesian economic methodology. Specifically, in Post Keynesian economic methodology, alongside critical realism, lie the ‘Babylonian’ approach developed by Sheila Dow (1990, 1996), and the ‘Fundamental Uncertainty’,‘encompassing’, or ‘generalising’ methodological approach, forcibly restated in Paul Davidson’s (1996) paper,‘Reality and Economic Theory’.As does critical realism, Dow’s and Davidson’s approaches contrast themselves with mainstream methodology in rejecting a closed-system ontology. This is reflected in calls by some (e.g. Dunn, 2001) that Post Keynesianism be defined methodologically, consistent with a particular (open systems) reality (cf. Mearman, 2002).This is important, because although Dow (2002) (see later) argues for a degree of methodological pluralism, she also argues that methodology should at least be consistent with external reality.This reflects the realist nature of all three strands. However, there is debate over the compatibility of these strands (see The Journal of Post Keynesian Economics, Fall 1999). In addition to this debate, there has been a discussion of relationship between critical realism and econometrics, to which Chapter 7 contributes. Indeed, Chapter 7 enriches that debate by offering a criticalrealist position modified by interaction with the practice of Post Keynesian economists. Therefore, it is important to clarify that the three approaches have essentially similar approaches to the specific concrete operational issues being discussed here.

250

Paul Downward and Andrew Mearman

Dow’s (1990, 1996) ‘Babylonian’ perspective on the Post Keynesian approach argues that theorists should embrace a variety of different insights into the phenomenon under investigation.This is because an open-system ontology carries with it the notion that uncertainty, based on the incomplete or partial understanding possible in an open system, is endemic to both the researcher and the researched alike. In turn this provides a rationale for, and reflects, the existence of multiple processes underlying events, and as a result legitimises a pluralist approach to theorising (which of necessity involves abstraction and partiality). Nonetheless, the thrust of theory from an open-system perspective should be to focus on discovering these processes. Much of this general perspective thus shares Lawson’s (1997) arguments. As Dow (1994, p. 2) points out, an organicist ontology undermines the ‘dualist’ thinking or epistemology endemic to the neoclassical research programme. Dualism implies that knowledge can be provided through reference to ‘all encompassing, mutually exclusive theoretical categories’ such as ‘fact’ and ‘theory’ or ‘truth’ and ‘falsehood’ or ‘macroeconomics’ and ‘microeconomics’. It follows that, according to Dow, the role of empirical evidence is different for Post Keynesians than, for example, neoclassicals. Dow maintains that for Post Keynesians evidence can be validly provided by a variety of sources such as questionnaire and historical sources. As Dow (1994) writes, …knowledge is acquired by gathering evidence and constructing arguments in order to build up rational belief.These contributions to knowledge are incommensurate in the sense that they do not add up to a single probability statistic, i.e. they do not fit into closed-system theorising … the choice of a range of methods depends on the nature of the problem and the context. But the choice is guided (and thus limited) by reason, convention and vision; it is differences in these which account for different schools of thought which have in common open-system theorising. (Dow, 1994, p. 7) Crucially, according to Dow (1990), one can conceive of the realist nature of theory by reference to two extremes within the ontological domain of worldrealism (i.e. that a real world objectively exists to be theorised about). Specifically, theory can either tend towards being ‘event-truth’ realist or towards being ‘process-truth’ realist. In the former case, correspondent to Lawson’s definition of empirical realism, theories are constructed to correspond to reality in terms of predicting events. Theories of this nature refer to events and may involve fictions, that is, falsehood, as opposed to merely abstractions of detail. Neoclassical economics can be characterised in this way because of its emphasis on the formal deduction of events and the econometric testing of predictions concerning the patterns of these events. Significantly this means that neoclassical economics can claim to have a realist underpinning. Process-truth realism, on the other hand, entails a concern that theories correspond to reality in the sense of highlighting the real causal relations

The case of Post Keynesian pricing 251 underlying events, with no presupposition that events may be predicted or that there is a one-to-one correspondence between causes and events.This approach, of course, shares the emphasis of Lawson’s open-system perspective. Indeed the central differences between Lawson’s and Dow’s approaches appear to arise in discussing the role of econometrics and of assumptions in theory; in other words, as one moves towards concrete epistemological issues.While in general both approaches present a role for a variety of insights in an investigation, in contrast with Lawson (1995, 1997), and sharing the perspective offered in this chapter and Chapter 7, Dow (1990) clearly suggests a constructive role for econometrics. Significantly, Davidson’s (1996) methodological approach begins with a critique of neoclassical economics, New Keynesian economics, and indeed Behavioural economics. He argues that they invoke an assumption of ergodicity, which is inappropriate to a world characterised by fundamental uncertainty.4 Notwithstanding any references to stochastic assumptions, Davidson argues that the ergodicity assumption implies that the past, present and future of a system are predetermined. This legitimises the emphasis upon predictability required for a deductive emphasis in theory and an emphasis on numerically defined probabilities in making inferences through statistical or econometric analysis. In contrast, Davidson (1996) argues that Post Keynesians are interested in a non-ergodic and transmutable-reality view of the world where, …probabilistic risks must be distinguished from uncertainty where existing probabilities are not reliable guides to future performance. Probabilistic risk may characterise routine, repeatable economic decisions where it is reasonable to presume an immutable (ergodic) reality. Keynes … however, rejected the ergodic axiom as applicable to all economic expectations … involving non-routine matters that are ‘very uncertain’… (Davidson, 1996, pp. 492–493, emphasis added) Examples of this perspective, for Davidson, include Keynes’ General Theory; Post Keynesian monetary theory; post-1974 Hicks; Shackle’s analysis of crucial decisions and Old Institutionalist theory. The implication is that the future is created through the act of choice. It follows that statistical inference will be of no use in cases of uncertainty. In rejecting the econometric method, Davidson appears to fall closer to the critical-realist position than Dow. Yet, Davidson’s arguments have prompted attention from critical realists and can be subject to criticism from an institutionalist perspective.The key issue at stake lies in the conceptualisation of decision making under the conditions of uncertainty postulated by Davidson. As Dow (1995) argues, if uncertainty is associated with absolute ignorance of the underlying structure of the choice situation then logically this is infeasible: ‘…absolute ignorance is incompatible with knowledge of absolute uncertainty’ (p. 124).This conceptualisation of uncertainty implies that both the researcher investigating agent choice, and the agents themselves, can only conceptualise their situations in a

252

Paul Downward and Andrew Mearman

purely subjective way. There is a divorce between the environment and the agent in socio/economic systems. Davidson claims an allegiance or sympathy with behaviouralist and institutionalist economics. However, on the ontological issue discussed above, this relationship appears somewhat strained. Specifically, many of these behaviouralist and institutionalist economists have emphasised the fallacious nature of perspectives which divorce the agent and its environment in the economy. For instance, Hodgson (1988) argues, … the force and impact of the behaviouralist argument … [is that] … the notions of equilibrium and global rationality in neoclassical economics are intimately related. (Hodgson, 1988, p. 80, emphasis added) Therefore, the notions of global rationality and equilibrium are both rejected simultaneously. Systemically speaking global rationality implies equilibrium. Conversely, it follows from the behavioural and institutionalist perspective that procedural revisions to decisions do not imply equilibrium (see Loasby, 1995; Foss, 1997).This means that there can be no ergodicity in a world populated by boundedly rational agents. Boundedly rational agents can both create and react to the future environment. In this respect it is problematic to link notions of bounded rationality and ergodicity.Theory that attempts to do this is flawed and Davidson’s (1996) conclusion is thus supported. However, one can question Davidson’s line of argument. In rejecting bounded rationality and focussing on non-ergodicity, this produces a view of the economic environment that is conceptually distinct from agent behaviour and choice. Further, in concentrating on the ergodicity axiom, Davidson (1996) fails to note that, epistemologically speaking, global rationality furnishes the desired result of ergodicity through the neoclassical conception of agency.5 Ergodicity is an integral part of a (albeit closed) system of analysis based on mode of inference relying on deductive logic. As a result, Davidson produces a technically defined – Cartesian dual – non-ergodicity that does not appear to require reference to bounded rationality – a non-neoclassical conception of agency.The danger with this position is that a definition of the social system as an arena of choice gets lost. It follows that just as there is no effective choice in a world where everything is predetermined, in a technically defined non-ergodic world no choice can be made. By definition everything is a ‘surprise’ and ‘uniquely originated’ – terms entirely consistent with Shackle’s arguments. However, the question is, how can one argue that the subject of analysis, the theoretical framework with which the world is understood, both interact with, and hence help to shape, the material world or object of analysis? In other words how is extreme subjectivity avoided? To avoid extreme subjectivity, one must draw upon some notion of how the external world affects choice, and presume that agents hold (some) knowledge of the world. Drawing upon Shackle’s (1958, 1988) conception of ‘practical conscience’ that puts bounds on behaviour, Davidson (1996) argues that agents

The case of Post Keynesian pricing 253 can form ‘sensible expectations’ drawn from institutional behaviour. However, Runde (1993) and Lewis and Runde (1999) suggest that this appeal produces an internal inconsistency because Davidson’s (1996) conception of knowledge and ignorance is based on an ontological dual of ergodic and non-ergodic events and hence the ontological character of sensible expectations is undeveloped. In contrast, a link to critical-realist ontology would overcome these problems, linking social institutions with the causal mechanisms that structure if not fully determine events. In other words, this allows for specific decisionmaking structures, existent partially independently of the individual, to form a part of economic inferences and choices. Critical realism thus explicitly embraces the idea of an interdependent system. One can link these differences to rhetorical strategy and Davidson’s desire to attack mainstream economics in its own vernacular (Davidson, 1999; Lewis and Runde, 1999). It is clear that, as Arestis et al. (1999) and Downward and Lee (2001) argue, the broad sweep of the Post Keynesian methodological approaches discussed here is to accept that the real economy is an open system in which economic structures evolve through complex interrelations between reflexive economic agents and causal structures.The approaches thus share the view that there is no compelling a priori reason why any empirical regularities might be observed. Yet, at another level, as Mearman (2002) argues, all of the methodological positions in Post Keynesian economics share a tendency towards negative definitions of open systems. Not withstanding the comments above, they are defined dualistically in terms of a more familiar concept, that is, closed systems. As Chapters 6 and 7 and the earlier comments argue, such duals can be transcended by recognising that closures external to thought can be termed ‘real’. Humans usually create real social closures through institutions.This is not to say that institutions create complete closure; rather, they merely increase stability and repetitiveness of events. Significantly, this hints at a concept of ‘quasi closure’, for example discussed in Chapters 5 and 6, lying between complete closure, which is taken to mean complete event regularities, akin to experimental conditions, and complete openness, which is often taken to mean, simply, disorganised flux. It follows that to the extent that the alternative methodological positions in Post Keynesianism do embrace the notion that in the real economy, economic structures evolve through complex interrelations between reflexive economic agents and causal structures, the approaches are not inconsistent with the arguments underpinning the approach to empirical inference discussed in Chapter 7. It follows that the comments aimed specifically at critical realism carry over to the broader methodological sweep of Post Keynesian Economics.These matters aside, the constructive role of econometrics can now be exemplified via the Post Keynesian literature on pricing. Empirical examples from UK manufacturing Pricing has formed a useful test bed for applied Post Keynesian economics. As noted earlier in the discussion of Gardiner Means, much is at stake in debating

254

Paul Downward and Andrew Mearman

the issue of price flexibility. Flexible price adjustment in response to excess demand lays at the core of neoclassical economics in general equilibrium theory and its core propositions about welfare. Price flexibility is rejected by Post Keynesian economics. This is a direct challenge to the neoclassical theoretical edifice. Post Keynesian pricing theory draws together seminal elements of research that challenges neoclassical economics. This includes Gardiner Means’ work but also the contributions of Hall and Hitch (1939), Andrews (1949a,b, 1964) and Kalecki (1954), and emphasises that in manufacturing, prices are set by a mark up on some measure of average costs with no explicit role for demand (see also Lee, 1998; Downward, 1999).What follows, thus, is simultaneously a discussion of the empirical evidence that can be said to underpin these ideas, and a discussion of the usefulness of econometrics to critical-realist research. To begin with, in econometric terms, debate over pricing behaviour has concerned two key issues.6 These are the extent to which prices are related to full/normal or actual unit costs (full/normal costs refers to some measure of trend or long-run costs) and the extent to which demand exerts a direct effect on mark ups and prices.These issues have been usually addressed in the form of the following general linear model. Pricei    Costi  Demandi  ui where i is the index of observation and u is the random error term. Precise measures of variables have differed. Prices have been measured as wholesale or producer price indices. Average costs (costs) have been restricted to indices measuring labour costs and/or material, fuel and other costs. Further, either their current, or current and lagged, values have been included to capture actual costs or full/normal costs – those associated with a longer period of production – respectively. In the latter case some contrived measure of full/normal costs has also been used whereby the researcher recalculates average costs based on some trend measure of output rather than its actual value per period. Demand has also been measured in different ways. Some authors have argued for an essentially monopolistic view of price determination. In this case levels of/changes in demand have been associated with levels of/changes in prices.Typically, output is used to measure demand here. Some authors have used a perfectly competitive market-price determination model that relates price changes to the level of excess demand. Others have combined some form of cost variable with an excess demand proxy for demand pressure. Updating Downward (1999, 2002), Table 14.1 presents a summary of the main studies for the UK grouped broadly according to the way in which costs were measured, although the third column reports the only study adopting a pure excess demand investigation into pricing. First examination of the table reveals some interesting features. One can clearly note that econometric work has not produced the ‘covering law’ aspired to in neoclassical methodology.There is evidence, at least, in favour of all of the specifications. This is most apparent in considering the first row of the table.

The case of Post Keynesian pricing 255 Table 14.1 A selective summary of econometric studies of UK pricing Normal cost specificationa

Actual cost specification

Neoclassical specificationb

Neild (1963)c excess demand no effect (excess demand for labour used to proxy excess demand for output).

Neild (1963) excess demand – no effect.

McCallum (1970)d ‘accords well with the evidence’. McCallum comments on Rushdy and Lund that price change is a function of excess demand only.

Rushdy and Lund (1967)d current and lagged excess demand – no effect.

Rushdy and Lund (1967)c current and lagged excess demand – significant effect. Sawyer (1983)e weak effects of demand (the level of demand proxied by output).

Coutts et al. (1978) excess demand a small but insignificant effect (deviations of output around trend used to proxy excess demand). Sylos-Labini (1979) no effect of demand on prices (capacity utilisation used to proxy demand). Smith (1982) re-specifies Geroski (1992)e supports Coutts, Godley and Sawyer. Nordhaus’ normal costs and finds demand now significant. Downward (2002) Downward (1995) supports reconfirms Neild. Sawyer. Notes a Normal cost transformations are applied to labour costs only, except for Sylos-Labini.The distinction between normal costs and actual costs may be weak in that lagged (i.e. averaged) values of the latter, used, for example, by Sawyer and Geroski, may be an indication of the former hypothesis. However, as Rushdy and Lund argue, transformations of costs prior to regressions may effectively rule out demand effects. b This is termed neoclassical because it adopts a specification where price levels/changes are a function of cumulative levels/current levels of excess demand only. c The preferred specification when a number were utilised. d These studies used Neild’s data – hence basis of excess demand proxy. e These were disaggregated studies and noted important heterogeneities in behaviour, though generally weak and insignificant demand effects.

Here, a number of studies were undertaken using the same data set. The first entry refers to Neild’s (1963) seminal econometric study of pricing in the UK aimed at providing a ‘dynamic interpretation of “full cost pricing” ’ (p. 4) for the National Institute of Economic and Social Research. Neild’s work generated the data for the studies. He employs three variants of a standard dynamic equation in analysing the data: Pt  1  2Wt  3Mt  4Mt  1  5Mt  2  6Pt  1

256

Paul Downward and Andrew Mearman

where P is the price, W the unit wage cost, M the unit material cost and t the time. The regressors common to each variant are the material input price and lagged own price. The variation in the equations arises from the inclusion of three different forms of wage cost. Actual unit labour costs are used, as are two forms of normal unit labour costs. To test for the presence of demand effects on prices, Neild includes perfectly competitive measure of demand proxied by the cumulative value of the excess demand for labour.This is based on the logic that as price changes are governed by the level of excess demand, price levels would be governed by the sum of excess demand. Neild’s basic conclusion is that the demand terms add nothing to the model, and the coefficient’s signs are negative.Thus the preferred specifications were versions of mark-up pricing. Neild’s work is criticised by Rushdy and Lund (1967) who argued that the full/normal cost measures – obtained by dividing unit labour costs by trend productivity – incorporate direct demand pressure on prices into the measure of costs. Rushdy and Lund thus estimate their equations using actual costs and changes in the variables to remove the problems of ‘common trends’ in the data.They conclude on the basis of a large number of alternative specifications that demand does have an influence on prices. In direct contrast, McCallum (1970) estimates a competitive pricing model. Price changes are related to future and current excess demands for labour. It is hypothesised that a lag occurs between product demand and labour demand so that current excess demand for output is picked up in a later excess demand for labour. Lagged price changes are included too.The model estimated is: ∂Pt  dt  1  dt  ∂Pt  1

where, P is the price, d the excess demand for labour, t the time and ∂ the change in variable. McCallum concludes that, The ‘pure excess demand’ hypothesis is shown to accord very well with the empirical evidence utilised by both Neild and Rushdy and Lund. (McCallum, 1970, p. 147) It appears that one can defend any hypothesis from the data. On this basis, one can quite easily understand the concerns of critical realists that statistical methods are inherently flawed and that their application is a rather futile exercise that artificially forces open-system reality through a closed-system lens to produce results that demonstrate little agreement. However, two important factors mitigate this argument.The first is that the ‘qualitative’ emphasis of the research clearly supports a version of mark-up pricing with weak direct effects of demand. This would seem consistent with the ability of econometrics to present demi-regularities.The second is that the studies alluded to above were at the cutting edge of econometrics over thirty years ago. Downward (2002) revisited the data using modern time-series

The case of Post Keynesian pricing 257 econometric methods incorporating vector auto-regressions and co-integration and reached the conclusion that Neild’s original findings were supported, which further tips the balance in favour of a version of mark-up pricing. Rather than abandoning econometric work, this suggests that econometrics can play a useful role in codifying the empirical level from which retroduction can begin. However, as argued in Chapter 7, this may require more than the appeal to one isolated study or technique. In addition, it is important to recognise that the legitimacy of retroduction is enhanced by this new use of econometrics. It is noteworthy that the econometric work alluded to above was originally motivated by a desire to evaluate the full-cost process of pricing first postulated by qualitative investigation, that is, Hall and Hitch’s work. In this sense the econometric analysis provides a means of adding legitimacy to those claims and (passively) provides a basis from which causal research can begin. Second, as Downward (1994, 1999) notes, the causal research in itself is not straightforward to interpret, but is characterised by terminological imprecision and methodological imprecision. For example, three broad clusters can be established from the literature. First, many studies consider how firms pursue single or multiple pecuniary goals, such as profits or market share, through a process of personal or institutional adjustment to initial full-cost prices in the light of demand pressure. A second group contains studies that recognise that firms also have non-pecuniary objectives. They address, as a general issue, the pursuit of multiple goals by firms.As with the first cluster, the pursuit of firms’ goals through price adjustment is described as stemming from personal and informal considerations as well as ‘objective’ institutional procedures. Third, there are studies which are concerned primarily with the pursuit of pecuniary goals such as profits, but which emphasise ‘objective’ or institutionally-based incremental decision-making processes for adjusting prices. At a general level, it can only be said that there is agreement on the pricing processes that firms follow. Specifically, it is agreed that firms price using a mark up on full costs, but that mark ups may change via a variety of organisational mechanisms; however, they are less likely to do so if demand changes. In the case above, as discussed in Chapter 7, econometric work can act as a check on the reliability of the attempts to distil a causal analysis from qualitative data.The case also, demonstrates the value of econometric work within the context of a broad literature. Moreover, it is also possible to illustrate the value of econometrics in a particular context, that is, one piece of research. Of course, no research project is conducted in isolation and will be embedded in a wider literature, which will influence the inferences that are drawn. This is, in fact, the case with the research below. Nonetheless, it can be shown that the purported validity of conclusions can be enhanced by analysing data in a number of ways rather than employing one technique of analysis, as is often the case in neoclassical economics. Downward (2001) undertakes both descriptive and statistical analysis of a sample of data drawn from UK manufacturing firms, to assess the relevance of the Post Keynesian perspective on pricing. To address these issues, data were

258

Paul Downward and Andrew Mearman

obtained by postal questionnaire on broad dimensions of firm behaviour capturing aspects of the firms’ pricing objectives, organisation, their market environment, and characteristics of the firms’ such as their size, products and cost behaviour. Consequently, enquiries probed at what level in the hierarchy pricing decisions were made and whether or not prescribed procedures were followed. Information was also gathered on the pricing formula employed, such as variants of mark-up processes and whether price lists and discounts were employed.The short-run and long-run emphasis on a variety of pricing objectives was obtained. Also differences in product type and degree of market competition were explored. Finally, firms’ pricing responses to a number of changes to cost and market conditions were investigated. It is important to note that the questionnaire typically elicited responses on the issues noted above using Likert-type scales. These measured the periodicity of firms being organised or behaving in a particular way, or the extent to which respondents agreed that their company was characterised in one way or other. The research had a ‘quantitative’ emphasis and was not based on constructing qualitative scenarios.This is not because the latter strategy was felt to be unimportant. Indeed qualitative work could have formed part of the research prior to the quantitative efforts. As alluded to earlier this was because the study intended to extend the wealth of existent literature on pricing.That said, even if this were not the case, it was argued in Chapter 7 that the descriptive analysis of statistics was presented as fully acceptable with critical realism. Analysing the data further using multivariate and econometric techniques thus provides an interesting illustration of how such findings can be triangulated. In the descriptive analysis of the data, the results are sympathetic to the idea that pricing is governed by a mark-up pricing formula, as part of a wider set of organisational decision-making rules, with explicit account taken of feedback from the environment (i.e. competitive pressure). For example, 48.3 per cent of firms initially use budgets in setting prices often or very often. 64.1 per cent of firms and 63.7 per cent of firms, respectively use price lists and a mark-up rule on average costs to set prices often or very often. However, it is clear that pricing decisions are not simply pre-planned. 53.2 per cent of firms often or very often make pricing decisions with contingencies in mind and only 25.7 per cent of firms report that contact between pricing personnel is dictated by pre-planned schedule. Moreover, 74.3 per cent of firms take into account the state of market demand when setting prices, 81.4 per cent of firms set prices to maintain goodwill often or very often and 50.4 per cent of firms follow their competitors in setting price often or very often. In addition, 72.5 per cent of firms agree or strongly agree that price competition is stronger than non-price competition and 88.4 per cent of firms believe that price competition is strong or very strong in their industry. The strategic importance of pricing is also evidenced by the fact that the firms typically compete with 5–9 actual and potential competitors. The results also suggest that once competition is accounted for firms reluctantly change their

The case of Post Keynesian pricing 259 prices from those initially set. 65.5 per cent of firms set prices to encourage price stability in the market implying price stability is explicitly desired by firms. Only 39.1 per cent of firms change prices from those initially planned often or very often, while 81.4 per cent of firms actively pursue goodwill in setting prices often or very often. Setting prices with long-term stability in mind is thus a characteristic of the survey. The results also showed that firms are more likely to change their mark ups than to change their discount structures, costing bases and price lists. Moreover, firms are more likely to change prices through their price lists rather than their costing base or discount structure.While mark ups may be ‘flexible’, therefore, this is defined in connection to longer-term price adjustment than rapid reactive price adjustment.This argument is borne out in considering the likely pricing responses of firms to changes in costs and demand. Table 14.2 reports the percentage responses of firms in the sample to various possible courses of action with prices following a cost or demand change. It shows behaviour that is markedly similar to the literature generally. Price changes are more likely to follow cost changes, but, in general, manufacturing firms prefer not to change prices. In general, the results suggest that firms do not seek to change prices and that as a result, sluggish price adjustment is normal behaviour for firms. Referring to the previous discussion of the literature, it is clear that the descriptive results are broadly supportive of Post Keynesian pricing theory. However, even in the brief synopsis of the research presented here, one can see that there are complicated interrelations at work. A natural question to ask is, ‘does the Post Keynesian emphasis on the mark-up pricing process present a useful retroduction’? To answer this question, consideration could be given to further steps of analysis. In the research being discussed, a further two-stage empirical strategy was adopted.The first stage involved using factor analysis to help to understand a complex system. As discussed in Chapter 7, while drawing upon an additive structure, this type of analysis addresses the issue of multi-collinearity by attempting to decompose phenomena into relatively autonomous sub-components. A factor analysis is concerned with identifying the linear combination of

Table 14.2 Firms’ pricing responses When setting prices ( firms)…

Response (%)

Increase prices when costs increase (ICIP) Decrease prices when costs decrease (DCDP) Increase prices when demand increases (IDIP) Decrease prices when demand decreases (DDDP)

15.8

35.9

42.6

4.2

0.7

5.3

7.4

33.5

34.5

17.3

3.2

16.2

44.4

26.1

9.9

1.8

14.1

47.5

25.4

10.6

260

Paul Downward and Andrew Mearman

variables that accounts for most of the variance in the data.The combinations – or factors – that are identified can be seen as sets of variables best summarising the data.The second stage involved regressing the pricing response of firms to, say, changes in costs or demand, upon these components.The objective of the exercise was to try to link the firms’ responses to changes in the environment to causes residing in firms’ objectives, price setting procedures, organisation and environment. Being indicative of the complicated system at work, the factor analysis extracted 25 factors from 80 variables, which suggests that the overall data is not easily summarised.Yet, many of the factors were readily interpretable. Clear examples of these were factors that linked together firms’ long-run or shortrun pricing objectives, their absolute size, aspects of mark-up pricing formulae, cost behaviour and so on. However, as discussed later, very few figured consistently in the regression analysis; this is an important result. By way of example, Table 14.3 reports the results associated with trying to explain the behaviour that an increase/decrease in costs would lead to an increase/decrease in price (ICIP/DCDP), and that an increase/decrease in

Table 14.3 Explaining firms’ pricing responsesa Independent variable/factors

Dependent variable Cost ICIP

Demand DCDP

PRICOMP WSALEDIS CHNGMK CONTING BREUSCH–PAGAN R2 Adjusted R2 F[25, 258] P(F )

0.136 (2.595) 0.194 (3.905) 94.3009* 0.148 0.065 1.79 0.0136

0.197 (3.455) 0.246 (3.588) 53.35* 0.196 0.118 2.52 0.00015

IDIP

DDDP

0.145 (2.688) 0.126 (2.405)

0.11 (2.07) 0.13 (2.432)

36.0798 0.194 0.116 2.48 0.0002

32.8805 0.131 0.046 1.55 0.049

Notes Where necessary the results were corrected for heteroscedasticity according to White’s transformation. The Breusch–Pagan statistic is reported in each case. There were 25 degrees of freedom associated with the test. a In this table, the constant in the regressions is not presented. * Starred statistics indicate significance at the 5 per cent level. T-statistics are presented in brackets.

The case of Post Keynesian pricing 261 demand would lead to an increase/decrease in price (IDIP/DDDP). Only coefficients that apply to each scenario for both demand and cost changes, are reported. The intention is to highlight what are revealed to be ‘persistent’ tendencies in behaviour. The results suggest that firms are more likely to increase/decrease their prices following an increase/decrease in costs if they modify their mark-up pricing formula. This is because CHNGMK is a significant factor. It summarises variables measuring changes in costing bases and mark-up formulae. The other significant factor in both of these cases, CONTING, indicates that changes in price following cost changes are associated with variables measuring contingent circumstances.The implication of these results is that the consistent factors associated with changes in price following changes in costs reflect the mark-up formula and contingent circumstances. Referring to the equations for IDIP and DDDP, that is, in seeking to identify factors associated with firms increasing or decreasing their prices in response to increases or decreases in demand respectively, the significant factors are PRICOMP and WSALEDIS. The former factor summarises variables expressing the degree of price competition facing firms.The latter factor summarises variables referring to discounted products in wholesale markets. The implication is that in cases of price competition, prices are adjusted to demand conditions to a degree. In each case, the R2, adjusted R2 and F-statistics are also presented. The F-statistics all have [25, 258] degrees of freedom. These statistics suggest that the regressions are significant but that they generally explain between 15 and 30 per cent of the relevant dependent variable. Low values of these statistics are, of course, to be expected with cross-section data. However, as has already been intimated, it also suggests that price adjustment may often reflect unsystematic influences. In short, it is argued that the research reveals that to the limited extent that firms change their prices in line with changes to their environments, a strong element of contingency exists. Moreover, the systematic tendencies that might underpin such price changes have a strong Post Keynesian emphasis. Central features include changes to the mark-up formulae and adjusting prices in line with competitors’ prices. Drawing upon the discussion of Chapter 7, therefore, the above example shows how in practice the reliability of specific empirical claims can be checked with reference to other claims. While descriptive analyses might themselves be suggestive of the causal mechanisms, the effects of the action of causal mechanisms can be assessed, and hence those purported causal mechanism supported, with reference to more quantitative analysis aimed at codifying the empirical level. Based on the research reported above one can argue that both the process of price-setting and the behaviour of prices that follows can be understood in Post Keynesian terms. In this sense the traditional Post Keynesian emphasis on – or retroduction of – mark-up models has some legitimacy.

262

Paul Downward and Andrew Mearman

Conclusion This chapter illustrates the main argument presented in Chapter 7, that econometrics can play an epistemological role in economic enquiry from a critical-realist perspective. In particular the focus is to show how critical-realist empirical enquiry could draw upon an eclectic range of research methods including quantitative techniques, which can play a part in a triangulation strategy seeking to legitimise causal claims from retroduction. In this chapter these arguments are illustrated by drawing upon applied work in Post Keynesian pricing. To facilitate this discussion the chapter begins by considering a historical example of triangulation, albeit using much less technically sophisticated techniques than are available today, presented by Gardiner Means’ work on pricing. Means can be seen to anticipate the concerns and proposals of Chapter 7 in that he appears to engage in retroduction via a triangulation strategy.This also illustrates that economic practice can inform philosophy. The Chapter also considers the compatibility of critical realism with other Post Keynesian methodological pronouncements and suggests that the arguments of Chapter 7 are held more generally. The chapter then concludes by presenting two contemporary examples of triangulation. One draws broadly on applied work from both an econometric and non-econometric nature.The other unites different data analysis techniques within one particular project. It is argued that these examples reveal the robustness of the Post Keynesian claim – which has been shown to have been obtained by a process similar to retroduction – that manufacturing pricing behaviour can be understood in terms of mark-up pricing processes where mark ups are adjusted in line with competition.

Notes 1 Aspects of the chapter draw upon and extend Downward and Mearman (2002). 2 Consequently, over the business cycle, one would expect to observe different price behaviour in different sectors. In particular, market prices would rise relative to administered prices in a boom and fall relative to administrative prices in a recession.The opposite tendency would be observed when considering the level of production. 3 It is worth noting that Means was one of the first economists to construct concentration indices. In this sense he is a forerunner of ‘Industrial Economics’. For more on this see Lee (1998). 4 Davidson distinguishes between ‘type 1’ immutable-reality models and ‘type 2’ immutable-reality models. Examples of the former case are: Classical perfect certainty models; Actuarial certainty equivalents, such as rational expectations, models; New classical models and some new Keynesian models. In the case of ‘type 2’ immutable-reality models, Davidson argues that the future is perceived to be unknown, in the sense that there are limitations with agents’ computational abilities, but that behaviour will converge, in the long run, onto the immutable path. Examples are: Bounded rationality theory; Knight’s theory of uncertainty; Savage’s expected utility theory; some Austrian theories; some new Keynesian models such as coordination failure; and chaos, sunspot and bubble theories.

The case of Post Keynesian pricing 263 5 As noted before, neoclassical economists do have a form of realist analysis. Causal relations are postulated (albeit erroneously) in a fashion analogous to the relations produced by scientists in a controlled experiment.The ontology of the neoclassicist is thus not simply Humean in appealing to atomistic events. This is an important issue. Hume’s problem of induction is a logical one that suggests that if reality comprises unconnected events then there is no logical necessity for the repetition of any particular sequence of events in the future.Without the possibility of causal connections between events it is logically possible that any explanation based on specific sequences of events may be incorrect. As Harré and Madden (1975) point out, the problem of induction is really an ontological issue. Accordingly, appeal to an ontology of causal mechanisms, if correctly identified, removes this problem. It is in this respect that the current paper is concerned with epistemological, rather than purely ontological, issues. Appeal to causal mechanisms underlying events shifts the issue away from pure logic towards epistemology and inference and ways in which we can gain reliable knowledge. It also suggests that in identifying a mechanism it would be logically possible to generalise that explanation to other examples of the same thing. The problem with the neoclassical approach, thus, is a lack of concern for the dangers of inference in an open system by treating it as closed. 6 There have of course been many other studies of pricing in the UK associated with industrial economics. The studies referred to here focus specifically on the pricing process and not influences from variables measuring the wider market structure.

References Andrews, P.W.S. (1949a), A Reconsideration of the Theory of the Individual Business, Oxford Economic Papers, pp. 54–89. —— (1949b), Manufacturing Business, Macmillan: London. —— (1964), On Competition in Economic Theory, Macmillan: London. Arestis, P., Dunn, S. and Sawyer, M. (1999), Post Keynesian Economics and its Critics, Journal of Post Keynesian Economics, 21(4), 527–549. Artis, M.J., Green, C.J., Leslie, D.L. and Smith, G.W. (eds) (1982), Demand Management, Supply Constraints and Inflation, Manchester University Press. Coutts, K., Godley,W. and Nordhaus,W. (1978), Industrial Pricing in the United Kingdom, Cambridge University Press: Cambridge. Davidson, P. (1996), Reality and Economic Theory, Journal of Post Keynesian Economics, 18(4), 479–508. —— (1999), Taxonomy, Communication, and Rhetorical Strategy, Journal of Post Keynesian Economics, 22(1): 125–129. Dow, S.C. (1985), Macroeconomic Thought: A Methodological Approach, Basil Blackwell: Oxford. —— (1990), Post Keynesianism as Political Economy: A Methodological Discussion, Review of Political Economy, 2(3), 345–358. —— (1994), Methodological Pluralism and Pluralism of Method, University of Stirling Discussion Paper, May. —— (1995), Uncertainty about uncertainty, in Dow, S.C. and Hillard, J. (eds). —— (1996), The Methodology of Macroeconomic Thought, Edward Elgar: Aldershot. —— (2002), Economic Methodology: An Inquiry, Oxford University Press: Oxford. —— and Hillard, J. (eds) (1995), Keynes, Knowledge and Uncertainty, Edward Elgar: Aldershot.

264

Paul Downward and Andrew Mearman

Downward, P.M. (1994), A Reappraisal of Case Study Evidence on Business Pricing: Neoclassical and Post Keynesian Perspectives, British Review of Economic Issues, 16(39), 23–43. —— (1995),A Post Keynesian Perspective of UK Manufacturing Pricing, Journal of Post Keynesian Economics, 17(3), Spring, 403–426. —— (1999), Post Keynesian Pricing Theory: A Realist Approach, Edward Elgar: Aldershot. —— (2001), A Post Keynesian Microeconomic Exposition of Price Stickiness, Eastern Economic Journal, 27(2), 165–182. —— (2002), Revisiting a Historical Debate on Pricing Dynamics in the UK: Further Confirmation of Post Keynesian Pricing Theory, 2001–2002, Journal of Post Keynesian Economics, 24(2), 329–344. —— and Lee, F.S. (2001), Post Keynesian Pricing Theory Reconfirmed (?). A Critical Review of ‘Asking about Prices’, Journal of Post Keynesian Economics, 23(3), 465–483. —— and Mearman,A. (2002), Critical Realism and Econometrics: Constructive dialogue with Post Keynesian Economics, Metroeconomica, 53(4), 391–415. Dunn, S. (2001),Whither Post Keynesianism?, Journal of Post Keynesian Economics, 22(3), 343–364. Foss, N.J. (1997), Incomplete Contracts and Economic Organisation: Brian Loasby and the Theory of the Firm. Paper Presented at a Conference in Honour of Brian Loasby, University of Stirling, August 26–28. Geroski, P.A. (1992), Price Dynamics in UK Manufacturing: A Microeconomic View, Economica, 59, 403–419. Hall, R.F. and Hitch, C.J. (1939), Price Theory and Business Behaviour, Oxford Economic Papers, 2, 13–33. Harré, R. and Madden, E. (1975), Causal Powers, Blackwell: Oxford. Hodgson, G.C. (1988), Economics and Institution: A Manifesto for a Modern Institutional Economics, Polity Press: Cambridge. Kalecki, M. (1954), Theory of Economic Dynamics, Allen and Unwin: London. Lawson,T. (1989), Realism and Instrumentalism in the Development of Econometrics, Oxford Economic Papers (Supplement), 41, 236–248. —— (1995),The ‘Lucas Critique’: A Generalisation, Cambridge Journal of Economics, 19, 257–276. —— (1997), Economics and Reality, Routledge: London. Lee, F.S. (1998), The Foundations of Post Keynesian Price Theory, Cambridge University Press: Cambridge. Lewis, P. and Runde, J. (1999), A Critical-Realist Perspective on Paul Davidson’s Methodological Writings on – and Rhetorical Strategy for – Post Keynesian Economics, Journal of Post Keynesian Economics, 22(1), 16–35. Loasby, B.J. (1995), The Organisation of Capabilities, University of Stirling Discussion paper, 96/6. McCallum, B.T. (1970), The Effect of Demand on Prices in British Manufacturing: Another View, Review of Economic Studies, 37, 147–156. Means, G.C. (1935), Industrial Prices and their Relative Inflexibility, US Senate Document 13, 74th Congress, 1st Session,Washington. —— (1962), Pricing Power and the Public Interest, Harper and Brothers: New York. Mearman, A. (1998), Keynes, Realism and Econometrics, paper presented to the 3rd Annual Postgraduate Conference, University of Leeds, November. —— (2002), To What Extent is Veblen an Open-systems Theorist?, Journal of Economic Issues, June, 36(2), 573–580.

The case of Post Keynesian pricing 265 Neild, R. (1963), Pricing and Employment in the Trade Cycle, Cambridge University Press: Cambridge. Runde, J. (1993), Paul Davidson and the Austrians: Reply to Davidson, Critical Review, 7(2–3), 381–397. Rushdy, F. and Lund, P.J. (1967), The Effect of Demand on Prices in British Manufacturing, Review of Economic Studies, 33, 361–374. Sawyer, M.C. (1983), Business Pricing and Inflation, Macmillan: London. Shackle, G.L.S. (1958), Time in Economics, Amsterdam: North Holland. —— (1988), The Bounds of Unknowledge, in Frowen, S.F. (ed), Business, Time and Thought, Selected papers of GLS Shackle, New York University Press: New York. Smith, G.W. (1982), The Normal Cost Hypothesis: A Reappraisal, in Artis, M.J., Green, C.J., Leslie, D.L. and Smith, G.W. (eds), Demand Management, Supply Constraints and Inflation, Manchester University Press. Sylos-Labini, P. (1979), Prices and Income Distribution in Manufacturing industry, Journal of Post Keynesian Economics, Fall, 11(1), 3–25.

15 From predictive to indicative statistics Explaining corporate borrowing Murray Glickman

Introduction This chapter sets out to explore the more general relevance of the methodology developed in a recently published article (Glickman, 2000–01) as an example of empirical work based on a broadly critical realist understanding of the nature of economics. As a preliminary to the present discussion, a brief excursion is necessary into the specialised subject matter of the article, which was written as an empirical sequel to an earlier, purely theoretical, ‘Post Keynesian Refutation’ (Glickman, 1997–98) of the celebrated Modigliani–Miller (MM) propositions on capital structure (1958). From ‘rational’ axioms, MM posit value invariance, that is, the thesis that a firm cannot alter the value of its asset stock by changing its debt ratio.1 Although, to those not familiar with modern finance theory, this may seem a somewhat arcane claim, it has far-reaching consequences for the practical conduct of financial management. Furthermore it is inspired by, and in turn provides additional bolstering for, the belief that money is neutral. On that ground alone, the MM propositions are incompatible with a Post Keynesian view of the operation of the economy. In part to confirm Glickman’s (1997–98) refutation of MM, Glickman (2000–01) aimed to show empirically that macro-economic factors, excluded from consideration on axiomatic grounds from the MM view of the world, have in fact been critical in determining UK corporate sector debt ratios. Conventional wisdom would suggest that this task should have been approached by applying one or more of a variety of widely used econometric tests to the data available. However, all such tests are based on stochastic models of reality, and a major plank of the refutation was a rejection, as systematically misleading, of the stochastic framework of analysis on which MM based their arguments.Thus, proceeding in a conventional way would, at the very least, have left the empirical work open to charges of inconsistency with the earlier theoretical discussion. In fact, the author’s misgivings about the use of econometrics for an analysis of the kind needed run much deeper than this. Contemporary mainstream economics does not regard stochastic models simply as tools. Philosophically it

From predictive to indicative statistics 267 views the world as governed by stochastic processes, or as the author prefers to say – for reasons discussed in the following sections – ‘processes’.The first section of this chapter emphasises the major conceptual problems inherent in the stochastic way of looking at the world. Although the author is willing to concede that these may not matter for some purposes,2 there seem to be particularly serious contradictions – as the author shall try to show in this first section – in accepting the stochastic paradigm as a guiding methodology when a researcher is working on historic data with a view to identifying causes or, in other words, when the purpose is to find explanations for some pattern of past events. Contrasting the ‘facts first’, explanation-oriented, approach adopted with the ‘facts last’, deductive/predictive approach that dominates mainstream economic analysis, the chapter’s second section reviews the theory of corporate borrowing behaviour and the related formal model contained in the article. Finally, the third section sets out and analyses the methodology developed in the article to provide the basis for an empirical assessment of the theoretical explanation proposed.

Causality and the stochastic paradigm According to the concept of a stochastic ‘process’, different outcomes in repeated experiments result from ‘random differences’. As Hicks (1979, pp. 105–106) notes, there is a serious danger of circularity in citing randomness as the ‘explanation’ for differences in outcome between apparently ‘similar’ events. For there is no independent evidence for the existence of the former – only the fact that the latter occurred. There is no independent evidence for the existence of the random differences because, logically, there can be none: independent evidence would amount to reasons for outcome differences, and if reasons for difference could be specified, the latter would no longer be random. It follows that, on both ontological and epistemological grounds, random differences cannot be anything other than mystery differences: we cannot specify what they are on any occasion when they are cited as existing, and on such occasions we can never know that they do exist. Now a synonym of random is chance, and one of the Oxford Concise Dictionary’s definitions of chance is ‘absence of discoverable cause’. As George Shackle wrote: I have the feeling that the notion of a stochastic process is an unwilling, and elementally insufficient concession to our uneasy consciousness of not knowing… . (Shackle, 1982–83)3 To put random differences forward as an explanation for some set of differences in observed phenomena is just another way of saying that we do not know why the differences in question have occurred. If the randomness of random events is an expression of our ignorance, an important further corollary is that randomness, despite appearances, cannot be a characteristic inherent in events themselves. If we knew enough in any given

268

Murray Glickman

context we could discover a cause of difference. It follows that a change in ourselves, the observers, would be sufficient to cause a difference to cease to be random. But, if the randomness, or otherwise, of events is a function of the ignorance of the observer, randomness becomes a subjective matter. Equally, the same event could be random to one observer and not random to another. Semantically this puts the word random on par with words such as incomprehensible. What is incomprehensible to me, might not be so to you: so it is with random. In fact, there is perhaps an even more fundamental problem with the stochastic paradigm, namely that it invites us to regard events as isolated, accidental outcomes, dissociated from anything that has gone before. That is a short step from suggesting that events do not have causes. Contrast Kant’s thesis (1982, especially pp. 223–227) that humans can only experience the world as a succession of caused events in time. His celebrated view was that if an event is to become part of our experience (and not just ‘happen’), we have to conceive it as being the necessary result of some other events which came before it in time, even if we cannot specify what these are. In short we must have (a priori) concepts of time and cause to be able to experience the world in which we exist. Consider a man-made disaster – a rail or air crash. A stochastic attitude to the event would be dismissed out of hand.We categorically do not say that it was just some random occurrence, bad luck, something that happened for no discoverable reason. Yesterday, the train passed over this particular stretch of track without incident: today, there was carnage. We do not say ‘Well, these were just two random events, different drawings from the same lottery tumbler of life.There was no reason why today was different from yesterday. It’s all just the operation of random forces’. On the contrary, we set up inquiries that take an essentially Kantian view in that they look for causes, differences in the antecedents of today, to stand as reasons why what happened today was different from what happened yesterday. If the concept of randomness is thus circular, suggestive of mystery, subjective and in conflict with our innate conceptions of how the world works, why does stochastic method loom so large in economics? The answer must be that it is a necessary adjunct to the deductive-nomological model that underpins mainstream theorising, according to which the intellectual task of providing an explanation for any phenomenon is equated with seeking to identify one or more sets of event regularities under which it can be subsumed (Lawson, 1997, pp. 16–18). For, if, as is more or less inevitable, the regularities identified are less than perfect, then some way of accommodating such imperfections is needed if the deductive-nomological approach is not to be undermined.The stochastic paradigm fulfils this function: if events are generated by stochastic processes then, of course, despite the existence of underlying law-like regularities, there will be differences – random differences – between one occasion and the next. This view of a world characterised by event regularities overlaid by random differences is the essence of what Davidson has described as the ergodic axiom and which he rejects on both empirical and logical grounds. (See, in particular, Davidson, 1982–83 and Davidson, 1996).

From predictive to indicative statistics 269 Lawson (1997, p. 23) outlines the main features of an alternative, critical realist, view of science in which it is seen to aim ‘at identifying and illuminating the structures and mechanisms, powers and tendencies, that govern or facilitate the course of events’ and within which explanation is conceived as ‘providing an account of those structures, powers and tendencies that have contributed to the production of, or facilitated, some already identified phenomenon of interest’. Within this approach, Lawson identifies retroduction (or abduction) as the key technique by which explanations are generated. Retroduction, he states, consists in the movement … from a conception of some phenomenon of interest to a conception of some totally different type of thing, mechanism, structure or condition that, at least in part, is responsible for the given phenomenon. (ibid. p. 24) As this chapter tries to show in the sections that follow, the methodology developed by the author, in an attempt to provide an empirical underpinning for my theoretical critique of the MM propositions is an illustration of what critical realist science might look like.

Articulating the model The deductive-nomological model revolves around the formulation and testing of predictions. Facts which are seen to confirm predictions therefore seem to occupy an important place in this scheme of things, as do, or at least should, facts that conflict with them. But, if facts are there to perform a function – to validate or to falsify – then they are there as a means to an end. They do not have status in their own right and figure neither as starting point, nor focus, nor central concern of the analysis. On the contrary, they are only acknowledged as existing if they can be shown to be consistent with, and therefore supportive of, a pre-existing formal system of ideas or, on occasions, if they can be shown to be inconsistent with that system. It follows that, relative to the work of a conventional researcher, there may well exist a class of incidental facts, facts which neither confirm nor conflict with his or her deductive model. These are unlikely to be noticed by the researcher or, if noticed, likely to be disregarded as tangential or useless. But if one’s objective in studying Economics is to understand the world, one cannot be comfortable with an approach in which facts are incidental in this way.The point of departure of Glickman (2000–01) was inspection of Figure 15.1, which seemed to reflect evidence of several collections of facts that would appear to be conspicuously incidental in the sense just described.These are: 1

That, over the period for which records were available at the time when the original research was undertaken (1963–1997), there seemed to be marked variation between one year and another in the reliance on borrowing as a source of finance by the UK ICC4 sector.5

270 2

3

4

Murray Glickman That the reliance on borrowing seemed to vary directly, with a lag, with the state of the macro-economy, as reflected in the four-quarter growth rate of the UK’s real GDP. That the troughs in the GDP series, reflecting the three major deflationary episodes experienced in the UK over the 1963–1997 period were matched, with lags, by troughs in the borrowing ratio series. That the variability of both series was far from constant over time.

70

12

50

10 8

30

6 10 4 –10

2

–30

0

–50

–2 69

76 Borrowing ratio

83

90 Real GDP change

Figure 15.1 Corporate borrowing and GDP in the UK.

97

4Q average change in real GDP (%)

4Q average borrowing ratio (%)

These facts, had they even registered with a conventional finance researcher, would be likely to have been ignored as incidental because conventional theory accords no role for the macro-economy in the determination of firms’ borrowing behaviour. Glickman (2000–01) took the exactly contrary view that these facts were important enough to be examined in their own right and in a way that was analytically independent of orthodox formal theorising on the subject of corporate debt ratios. Indeed, if they could be satisfactorily explained they might shed light on the determination of firms’ borrowing decisions in a way the standard approach was incapable of doing. In critical realist terms, my starting point was a ‘phenomenon of interest’ , or demi-regularity, which I subsequently sought to explain, that is, connect with theory. I took a ‘facts first’, explanation-oriented, approach rather than starting with theory and only later looking at the world, as I would have done had I followed the ‘facts last’ modus operandi of the deductive/predictive approach. The next step was to look for a systematic way of examining the data to see whether they would really support an explanation of corporate borrowing behaviour that related it to the state of the macro-economy. The three components to the approach developed – theory formulation, theory selection and

From predictive to indicative statistics 271 empirical assessment – correspond closely to the critical realist ‘way’ of doing science that Lawson (1997) proposes as, The realist aim in employing retroductive arguments is to formulate hypotheses of the sorts of conditions that could have given rise to some phenomenon of experience. In other words, the object is to infer conditions which, if they had been operative, could have given rise to the observed phenomena in question. …(O)nce various mechanisms are theorised their reality must be assessed. If competing theories are retroduced they must be selected from on the basis of their (relative) explanatory adequacies. In addition, the facilitating conditions of any mechanism must in turn be identified and empirically assessed … . (Lawson 1997, p. 310) Of these three components, it is the third – empirical assessment – that will occupy the bulk of the discussion in the remainder of this chapter. However, a brief comment is first called for on the two other components of theory formulation and selection. Lawson argues that ‘(R)etroductive or abductive reasoning is indicated by a move from the observation of numerous black ravens to a theory of a mechanism intrinsic … to ravens which disposes them to be black. It is a movement … from a ‘surface phenomenon’ to some ‘deeper’ thing’ (ibid., p. 24). In this context a theory was needed that was ‘intrinsic to’ corporate borrowing and offered a way of relating the empirical evidence available to human economic experience. The idea of business confidence seemed the key here: business confidence varies with the state of the macro-economy and, as it does so, the willingness to enter into loan contracts on the part of corporate managements as of those who lend to them will also vary.6 It was possible, as Glickman (2000–01) showed, to retroduce at least one apparently rounded competing theory, the so-called pecking-order hypothesis, and necessary in the article to devote considerable space to demonstrating the superior explanatory adequacy of a business confidence explanation of the empirical findings. Readers are referred to the article for that discussion.7 The main empirical task was to go beyond the vague surmise of a broad relationship between the borrowing ratio and the real GDP series. To determine, whether or not the data could yield evidence of a causal link between the two variables, solid enough to call for an explanation in which confidence could be placed. That required the development of a technique for analysing the data that had a convincing logic but, for reasons already stated at length, was not dependent on stochastic theory. In fact the explanation posited suggested the logic for the technique that was developed and the technique, when applied to the data, in turn provided evidence supporting the explanation: a neat but acceptable analytical interaction. The conception was that the factor of business confidence expressed itself in a (i) direct, (ii) time-related and (iii) aggregative relationship between the

272

Murray Glickman

borrowing ratio and real GDP growth. Each of these elements is considered in turn. (i) A direct relationship Stated baldly, a high rate of growth in real GDP induces a high level of business confidence and therefore a high borrowing ratio, while a low rate of growth induces the opposite. (ii) A time-related relationship The GDP growth rate in any given quarter can be thought of as having an influence on the level of confidence and therefore on the borrowing ratio in that quarter and succeeding ones and also that the strength of this influence, on the borrowing ratio in successive quarters, varies and eventually decays. A given quarter’s GDP growth rate is still in the process of realisation when that quarter is current. It cannot be fully known until the quarter in question becomes part of the past. Before that, agents’ hunches about what the rate will turn out to be on are all that they have to go on.They may not be willing to put much trust in these and, even if they are, the latter may turn out to be well wide of the mark.The implication is that the state of confidence prevailing in a given quarter is likely to be weakly related to the GDP growth rate that will ultimately be realised in that same quarter. By the same token, GDP growth rates experienced in the several immediately preceding quarters will be known quantities in the current quarter. For that reason, they are likely to have a more definite impact on the state of confidence prevailing during it.8 However, each quarter’s growth experience will gradually recede into history.As it does so, it will be seen as less and less related to today’s conditions and its impact on the state of confidence in the current quarter will diminish progressively. This conception of the impact of the recorded real GDP growth rate in Quarter 0 on the state of confidence implies that a high GDP growth rate in Quarter 0 will have a positive impact on confidence in succeeding quarters, first growing and then subsequently diminishing towards zero. In contrast, a low GDP growth rate in Quarter 0 will have a negative impact on confidence that first strengthens and then wanes to nothing. (iii) An aggregative relationship If the impact of the GDP growth rate on confidence in subsequent quarters is time-related and if confidence is the key determinant of corporate borrowing ratios then the impact of GDP growth rates on borrowing ratios may well be correspondingly time-related.This in turn means that GDP growth rates recorded in each of a series of quarters will have an influence on the state of confidence prevailing in any given quarter.To make the same point more precisely, the state of confidence prevailing in any given quarter will be conditioned by the rates of GDP growth experienced in a series comprising the current and a succession – or, more accurately, a ‘retrocession’ – of prior quarters. If the borrowing ratio in this given quarter reflects the current confidence state, it then follows that the value this ratio will take will in turn reflect the aggregated influence of past (and current) rates of GDP change on the current state of confidence.

From predictive to indicative statistics 273 In Glickman (2000–01, p. 207) this relationship was formally modelled in the equation bt 

t

 Y

iz

i

i

where bt denotes the size of the four-quarter average borrowing ratio in some quarter t, Yi denotes the change in real GDP in any quarter, i, the coefficient i denotes the strength of the influence of the real GDP change in any quarter, i, on the value of bt (i  0) and z denotes a specific quarter earlier than t. In the model, the value that i takes in any quarter, i, will reflect the strength of the impact of that quarter’s growth rate on the Quarter t borrowing ratio. The aggregation of the influences of past and present GDP growth rates on the Quarter t borrowing ratio is expressed in the summation in the RHS of the equation. If the argument presented earlier is correct, the state of confidence in Quarter t is likely to be weakly related to the rate of GDP change realised in that quarter. Hence, the influence of the latter on Quarter t’s ratio is therefore likely to be low and the value of t small. However, consider Yt  1.9 If we assume that its value is known in Quarter t, it will be likely to have a relatively strong influence on confidence levels prevailing in, and therefore on the borrowing ratio recorded in, Quarter t, because it will represent what is effectively current information.This implies that t  1 is likely to be greater than t. How will the value of t  n change as n increases? It is neither possible nor in fact necessary to state a priori whether, for any specific value of n beyond n  1, t  (n  1) will be greater or smaller than t  n. What can be said – and all that it is necessary to say for present purposes – is that if, initially, for some low values of n, t  (n  1) is greater than t  n, eventually, as n increases, this relation will eventually be reversed. The reason is again an extension of an argument made earlier. GDP growth rate data for any given quarter will eventually begin to exert a diminishing (and ultimately negligible) influence on confidence in subsequent periods as these periods extend further and further into the future. By the same logic, the influence of the GDP growth rate recorded in individual past quarters on the state of confidence prevailing in, and as a result on the borrowing ratio recorded in, Quarter t will eventually begin to diminish and ultimately decline to zero. It follows that the value taken by i will eventually begin to diminish as i falls (i.e. as Quarter i recedes into the past) and will ultimately decline to zero as i falls to some specific value, z.

Applying the model The values of the i’s and z that, if the model holds, determine the bt recorded in any quarter do not correspond to observable phenomena and are hence not in any direct sense knowable.This is not a material issue, however, since estimation of their values is not, in fact, relevant to the methodology used in the article.

274

Murray Glickman

Rather, these variables figure simply as referents to deeper structures which, as noted earlier, determine the realities that can be observed, according to critical realist theory. Their function is to form part of the conceptual framework that informs the design of the technique of empirical investigation that was developed. An imaginary example will help to explain this technique. Suppose that, over some period of, perhaps, ten years or so: 1

2

The values taken by i’s follow the pattern suggested in the theory underlying the model, first rising as i falls from t to t  6 and then falling to zero, as i falls further to t  12. (Thus z  t  12). The values of individual i’s remain ‘reasonably’ stable.10

Now consider a researcher who: a b c

has access to quarterly data for borrowing ratios (bt’s) and rates of real GDP growth (Yt’s) for the period in question; suspects that a model of the general form presented earlier reflects underlying causal relationships; but has (necessarily) no knowledge of the actual values taken by the i’s and z.

How might the researcher analyse the data if the aim is to investigate whether they do or do not provide evidence of the general kind of causal relationship suspected to exist? Suppose the researcher begins by arbitrarily correlating unlagged bt’s with four-quarter average Yt’s (i.e. the average of Yt, Yt  1, Yt  2 and Yt  3) over the period. The underlying causal relationship is likely to produce a positive value for r, but correlating (i) unlagged bt’s with (ii) Yi’s averaged over only four quarters is unlikely to produce a result that will reflect the strength of that relationship at all, fully. The researcher needs some way of overcoming the limitations implied by (i) and (ii), limitations he will recognise, given that, in general terms, he has an accurate conception of the underlying causal relationship. Turning first to (i), we may note that assumption I stated above implies, for example, that t  4  t.The researcher cannot know directly that this inequality holds true, but suppose it is suspected that it might. A response may be to repeat the correlation but this time lagging bt’s by one quarter. If done, the researcher should find that the resulting r will be higher than in the unlagged case.This is because the current quarter, Yt, whose i is relatively low, drops out of the four-quarter Yi average, whilst Quarter Yt  4, with a higher i, is included. In the light of this result, suppose the researcher is encouraged to recalculate r twice further, once with a two- and once with a three-quarter lag.The resulting r’s should be higher still. Incorporating a three-quarter lag implies correlating bt’s with the average of Yt  3, Yt  4, Yt  5 and Yt  6 and hence, though the researcher can at best only suspect this, including the Yi associated with the highest i of all. It is possible that increasing the lag to four-quarters may produce a still higher r, since t  7 (whose effect would now be captured),

From predictive to indicative statistics 275 though smaller than t  6, may still be greater than t  3 (whose effect would now drop out). However, if the researcher continues repeating the calculation with longer and longer lags, the resulting r’s will eventually begin to fall. For we have assumed an underlying reality in which i’s fall to zero as i diminishes towards t  12. Thus, extending the lag means that, eventually, Yi’s whose i’s are relatively high will be replaced with more remote Yi’s whose i’s are lower. Once r is clearly falling towards zero, there is no point in carrying out further recalculations with even longer positive lags. However, calculating the r’s that would be produced by some shortish negative lags11 are of significance in that they should show a diminishing value of r as the lag negativity is increased.The underlying reason can be briefly stated. Causes precede effects, and the causal relationship we are assuming is that GDP growth rates determine borrowing ratios, not vice versa. If our assumption holds, borrowing ratios should not, in any systematic way, lead to GDP growth rates. Thus, given the parameters of the model we have assumed in our numerical example, correlations of bt’s with Yt’s over a range of lags from, say, 6 to 10 quarters will produce values for r which, relatively smoothly, first rise from low positive (or possibly negative) values and then return again to low positive and, if the series is extended long enough, to negative values. It is worth reiterating that our researcher is ignorant of the values of these parameters and his calculations do nothing to reduce this ignorance.The point is that the researcher does not need to know or to estimate these values. Purely on the basis of a valid theoretical understanding resulting in correct beliefs about the general form of the model, he can proceed to derive a series of r values that will exhibit an orderly behaviour pattern. Following this procedure is really all that is required to generate some significant conclusions, as we now go on to show. With this orderly pattern established, the researcher can turn his/her attention to the second arbitrary element in the initial approach, the decision to average Yi’s over four quarters.The researcher will be aware that the borrowing ratio in any given quarter is likely to be influenced by rates of GDP growth recorded over more than just four quarters, even though there is no way of specifying exactly how many more. One response to this uncertainty is to redo the calculations with the difference now that the researcher averages the Yi correlate over, say, six as opposed to four quarters. With a zero lag, the r obtained is likely to be higher than with a Yi periodicity of only four quarters, since Yt  5 and Yt  6, both associated with particularly high i’s, will now figure in the computation. The same is likely to be true if the researcher next computes further correlations with Yi’s still averaged over six quarters but with bt’s lagged positively and negatively over a range of values. In short, using a Yi periodicity of six rather than four quarters is likely to produce another orderly r series, with values, for any given lag, higher than the first. The researcher can then go on to compute further sets of correlations over a range of lags and with Yi periodicities of, say, eight or ten quarters or more. Given that the (unknown) true value of z is twelve quarters, correlating with eight- and possibly even ten-quarter average Yi’s is likely to produce further

276

Murray Glickman

orderly r series. However, as the number of quarters, n, over which Yi’s are averaged is increased, the orderliness of the resulting r’s will sooner or later break up. The reason for this is straightforward. We are assuming a reality in which Yi’s recorded more than twelve quarters earlier than Quarter t have no influence on bt.These Yi’s will figure more and more prominently in the computations as the Yi periodicity is increased.The effect must eventually be to reduce the values of, and undermine systematic patterns in, computed r’s. It has been repeatedly stressed that the researcher will not know the value of z in this (or any other) case. However, the correct general grasp of the causal forces at work will include an understanding that z is finite. The researcher will, therefore, appreciate that the orderliness of the set of r series generated will eventually be disrupted if he increases n beyond a certain point.The researcher simply needs to reiterate his/her calculations with higher and higher n’s to discover what this point is in any particular case. Thus, the researcher’s technique will consist of correlating bt’s with Yi’s averaged over n quarters for a range of values of n, in each case computing r’s across a range of lags, positive and negative. If the analysis presented is correct, the researcher will be able, by a process of what might be called systematised trial and error, to identify perhaps several values of n which generate orderly (in the sense defined earlier) series of r values over a range of lag values, j.The researcher will also be able to identify the limits of the range of lag values over which order prevails for each r series, that is, for each n. Furthermore, it will be possible to identify a lag value, j, which produces a maximum r value for each n. A number of conclusions would appear to follow: 1

2

3

If applying the technique described earlier generates several orderly r series, these two features of the results – orderliness and severalness – suggest that the latter are more than coincidental and that they constitute evidence of the working of systematic causal forces. Inferences about the direction of causation can be drawn from findings that (i), for each series, r takes progressively lower (and ultimately negative) values when bt’s are more and more negatively lagged,12 whilst (ii), within some limit, r rises as the positive lag is increased: higher r values when the growth rate data relate to quarters earlier than the borrowing ratio data indicate that it is the former which influence the latter and not vice versa. It will be possible to identify a j and n combination that produces the highest individual r of all, what was termed the maximum of maxima in Glickman (2000–01, p. 210).The higher this value,13 the stronger we can take the causal influence of real GDP on the borrowing ratio to be, even accepting that no absolute scale exists against which this strength can be measured.

These conclusions summarise the argument that, if the model is a valid reflection of deeper structures in the world, then these structures will have

From predictive to indicative statistics 277 effects that are capable of being discerned quantitatively using the technique we have described. Analytically there is a reciprocity in the relationship between the technique and the researcher’s (general but unquantifiable) conception of the operative underlying causal forces: on the one hand, the general conception guides the development and application of the technique; on the other, if applying the technique produces orderly results, this is evidence for the validity of the general conception. By the same token, had the conception been wrong, the chances that applying the technique would generate orderly results must be very slim indeed.The far more likely outcome would be a glaring lack of orderly results, which would point strongly to the wrongness of the conception. Examining the orderliness or otherwise of the results produced by applying the technique thus provides the empirical assessment of the posited business confidence explanation for the observed UK corporate borrowing behaviour over the period studied, that, in critical realist terms, is necessary for good science. Consistency with the criticisms of conventional methodology set out at the beginning of the chapter would preclude the assumption of long-run stability in i’s in any alternative methodology that might be proposed. In fact Glickman (2000–01, pp. 206–207) argues that the 1963–97 period should be viewed as comprising four distinct sub-periods separated by major deflationary episodes. On that basis, it is reasonable to assume that each sub-period was characterised by a different confidence ‘climate’ and that i’s are likely to have been highly variable between one sub-period and another. That said, applying the technique described earlier requires only that i’s and z are assumed to be ‘reasonably’ stable within sub-periods.This is because otherwise it is most unlikely that orderly r series would be generated when the technique is applied. Conversely, orderliness will represent evidence that i’s and z are reasonably stable in the short run. Even if they are, it remains clear that the researcher would have no basis for putting values on these parameters. It is worth restating that this does not constitute any kind of shortcoming in the underlying methodology. Estimation of these parameters might be required if the function of the model were prediction. However, its function was a categorically different one, namely to guide the development and application of a technique for using data indicatively or, in other words, to enable the researcher to investigate whether there is evidence of an underlying causal relationship.The model’s role was therefore to assist in the process of explanation.

Conclusion To conclude this chapter it is worth noting that Lawson (1997, p. 310) argues that, according to conventional methodology, ‘not only is predictive accuracy reckoned to be the only essential criterion of theory choice …, but the prediction of events is thought to be the sole, or anyway the primary, use to which any successful theory is eventually to be put’. In the indicative statistical analysis outlined above, estimation plays no part because explanation and not prediction

278

Murray Glickman

is the ultimate goal and criterion of success. In contrast, and to repeat the point above, the model referred to in this chapter was to assist in the process of explanation.

Notes 1 The ratio of funds provided to the firm by (external) lenders (debt capital) to funds provided by owners (share capital). 2 It is worth noting that Hicks (1979, p. 103) cites agricultural research as an area in which statistical methods seem particularly appropriate. 3 In private correspondence to Paul Davidson, quoted in Davidson (1982–83, p. 189). 4 Industrial and commercial companies. 5 As measured by the four-quarter average borrowing ratio, defined as the ratio of new borrowing to the total sources of funds raised by the ICC sector in any period. The borrowing ratio constitutes a flow counterpart to the debt ratio. 6 The Post Keynesian character of the underlying theory is developed in detail in the article. See Glickman 2000–01, pp. 215–220. 7 Ibid., pp. 221–230. 8 In Glickman (2000–01, p. 217) it is argued that these can be taken to represent ‘the existing situation’ which, Keynes (1936, p. 148) argues, we ‘project … into the future, modified only to the extent that we have more or less definite reasons for expecting a change’. 9 Or, more strictly, the latest Yi whose value is known. 10 The tenability of such an assumption is discussed. 11 By correlating, say, bt  3 and then bt  6 with the average of Yt, Yt  1, Yt  2 and Yt  3. 12 That is, when bt’s to some extent precede Yi’s. 13 The article (ibid., pp. 208, 209) reported several values which were arguably very high by any standard.

References Davidson, Paul, ‘Rational expectations: a fallacious foundation for studying crucial decision-making processes’, Journal of Post Keynesian Economics, V(2) (1982–83), 182–197. —— ‘Reality and economic theory’, Journal of Post Keynesian Economics, 18(4) (1996), 479–508. Glickman, Murray,‘A Post-Keynesian refutation of Modigliani-Miller on capital structure’, Journal of Post Keynesian Economics, 20(2) (1997–98), 251–274. —— ‘Seeking explanations not preserving theorems: corporate finance, confidence, asymmetric information and the UK Macroeconomy’, Journal of Post Keynesian Economics, 23(2) (2000–01), 201–234. Hicks, J., Causality in Economics (Blackwell, Oxford, 1979). Kant, I., Immanuel Kant’s Critique of Pure Reason, transl. Norman Kemp Smith (Macmillan, London, 1982). Keynes, J.M. The General Theory of Employment Interest and Money (Macmillan, London, 1936). Lawson,Tony, Economics and Reality (Routledge, London, 1997). Modigliani, Franco and Miller, Merton H., ‘The cost of capital, corporation finance, and the theory of investment’, American Economic Review, XLVIII(3) (1958), 261–297.

16 Transition in Eastern Europe Critical realism and qualitative insights1 Karl Petrick

Introduction There is a large body of literature surrounding critical realism in economics. Since Lawson (1997) set out the case for critical realism in great detail, there has been a substantial debate regarding the strengths and weaknesses of such an approach, both for economics generally, and for individual schools of thought.2 This chapter comments on the role of empirical work in critical realism by referring to qualitative work undertaken by the author. To set the parameters of discussion, the next section will first discuss the major aspects of critical realism.This is contrasted with the research methodology found in mainstream economic theory. Section ‘Can mainstream economics be considered “empirical realism”?’, further investigates mainstream economic theory, and argues that it cannot be considered an example of empirical realism. Section ‘Critical realism and empirical investigation’, investigates the implications that critical realism has for empirical economic research, arguing that the critical realist framework is potentially very liberating for empirical research, as it allows the divide between economic theory and empirical investigation to be lessened. One criticism of critical realism is that it is very vague regarding research methodology. Section ‘Can there be (or should there be) a single critical realist methodology?’ considers this criticism, and offers a way forward. Using the central elements of critical realism as criteria, in this section the chapter considers the advice regarding methodology found in the research methods literature. The penultimate section discusses research that the author conducted in Hungary and the Czech Republic in 1996 and 1997 that was informed by these deliberations. It is hoped that any economist who undertakes empirical research will find these sections of use, regardless of their own ideology. The final section concludes.

What is critical realism? Lawson (1999b, p. 14) states that ‘those connected with (critical realism) all accept, and seek to promote, that the world is open, structured, dynamic and at least in part holistic, with the aim of science being to illuminate mechanisms,

280 Karl Petrick etc. responsible for phenomena identified as of interest’. These factors need elaboration. First, to accept that the world is structured implies that events have a cause, they do not happen spontaneously. If a degree of order did not exist, scientific inquiry would be impossible. Second, to argue that the world is open implies that there are many forces at work, so that ‘underlying mechanisms generate forces, tendencies and counter-tendencies, (and) outcomes have a multiplicity of causes’ (Rotheim, 1999, p. 74). Even though the world is structured, this open nature means that it may be difficult to pinpoint a single cause for an event.This acceptance of an open world has two important consequences for research. Critical realism is not atomistic: it does not define individuals/structures in isolation from the relatedness. And it does not seek to reduce cause-andeffect to a law, or to a law-like state of affairs, in the sense that ‘whenever this event/state of affairs then that event/state of affairs’ (Lawson, 1997, p. 88). Due to the presence of a multitude of tendencies and counter-tendencies, even a persistently observed event, or ‘event regularity’, cannot be assumed to be permanent. Also, as the relative power of these underlying forces wax and wane over time, there is also a multiplicity of potential outcomes. Even an event regularity only exists as long as the forces that tend to cause it dominate the forces that act as counter-tendencies. Finally, to argue that the world is dynamic implies that agents are situated in their social environment, and contribute to the reproduction and/or transformation of this environment through their actions (both intentionally and unintentionally).3 For all of the above reasons, ‘a critical realist perspective aims at identifying and comprehending structures, mechanisms or tendencies that govern, facilitate, or produce observable surface phenomenon’ (Rotheim, 1999, p. 72). For critical realism the observed event is merely the starting point, and the focus of research is to understand why the event occurred – what underlying mechanism caused the event? ‘Any social scientific theory that is linked with (critical realism) looks to uncover what is. Consequently this theory would have to be historically and socially context-dependent and would need to provide a framework by or through which the scientist can think about phenomena’ (Rotheim, 1999, p. 80).4 This contrasts strongly with mainstream economic theory.What critical realism and mainstream economic thought do share is the idea that the world is structured. However, for the ‘closed system modelling’ of mainstream economic theory, the magnitude of order is extraordinary. Rather than a multiplicity of causes and a multiplicity of potential events, there is a very direct causal link between an event and its source. Within mainstream economic theory it is assumed that the underlying forces of an economy reflect a constant conjunctions of events, and therefore are law-like states.5 There is no room for the presence of countervailing forces, or if they do exist they are of no consequence. This inherent assumption of law-like states allows mainstream economic theory to hypothesise an atomistic framework: it attempts to define and investigate individuals/structures in isolation from their relation to other individuals/ structures. There is little or no room in such a theoretical framework to

Transition in Eastern Europe

281

consider the social context in which an event takes place, as that is abstracted from in such an atomistic environment.

Can mainstream economic theory be considered ‘empirical realism’? The discussion in the previous section suggests that there are large gaps between mainstream economic modelling and the basic tenets of critical realism. In Lawson’s view, moreover, rather than attempting to undercover the underlying mechanism behind an observed phenomenon, mainstream economics is only concerned with observable facts – an event regularity is observed and therefore is assumed to be a law-like state as described earlier. Because of this perceived focus on observable fact, Lawson terms the mainstream research approach empirical realism. However, the divide between mainstream economic theory and empirical economic research should be made clear. In one sense it gives mainstream economic theory too much credit.To call mainstream theory empirical realist implies that what the theory states is actually observable in the real world. But far too much of mainstream theory is observable only hypothetically, and mainstream theorists are fairly straightforward about this. However, they do try to claim that such theory is still useful because it describes how things might work if all the conditions were right. For example, Stiglitz (1997), states ‘The Modigliani–Miller theorem … provides a useful starting point (for studying corporate finance). Describing when the corporate finance decision does not matter helps focus attention on what is truly important about corporate financial structures (such as the issues of ) bankruptcy, taxation, management incentives, the market’s perception of a firm’s value and corporate control’ (p. 480).6 Or, as Miller states even more obliquely: ‘showing what doesn’t matter can also show, by implication, what does’ (2001, p. 185, emphasis in original). But in this case what does matter (all of the issues mentioned by Stiglitz above) is outside the theory. Mainstream theory and empirical investigation, even mainstream empirical investigation, seem to confront very different issues, and reach very different conclusions. For example, the empirical studies of Barclay and Smith (2001a,b) and Smith (2001) all dismiss Modigliani and Miller’s theory. There are many things that mainstream economics’ theory-from-fantasy could be called, but it is difficult to see how ‘facts’ that are observable only in hypothetical situations can be considered either empirical or realist.

Critical realism and empirical investigation The above is not a trivial point.The gulf between orthodox theory and empirical investigation has long been established.‘From Hutchinson (1938) to Blaug (1992) economists committed to empiricism have argued persuasively that while economics pays lip service to empiricism (they preach it) they do not in fact behave according to its precepts (they do not practice it)’ (Hands, 1999, p. 179).

282

Karl Petrick

For the reasons stated earlier, theory tends to ignore empirical work and empirical work either ignores theory or discredits it. However,‘For those acting upon critical realism, the aim at each stage will be to bring empirical and other evidence to bear with the intention of identifying which of the competing accounts is the most adequate’ (Lawson, 1999b, pp. 14–15). From this quote it seems that the testing and development of theory with an emphasis on how well it explains empirically observed events is the focus of critical realism. This allows us to provide a much clearer delineation between empirical realism and critical realism, which is now attempted. To begin with, Lawson’s (1999b) arguments are potentially very liberating for empirical economic research.While it appears that many heterodox schools of economic thought are much more amenable to adopting a critical realist perspective,7 this approach may also be fruitful for any research project that seeks to use empirical data, even mainstream empirical research. Because mainstream theory is so often completely at odds with empirical results, empirical economics can sometimes seem relatively light on its theoretical underpinnings.Without a theoretical base to start from, much of this research describes empirically observed phenomena, but does not attempt to go much beyond description. For example, Zwass (1979) makes it very clear that he is only interested in describing and not theorising.8 This chapter would argue that this is the correct description of ‘empirical realism’ – mainstream empirical research (and empirical research generally) may be accused of it, but not mainstream theory. However, critical realism seems to entail a flexible approach to research, one that would allow not only for the testing and development of existing theory but also for the development of new theory, theory that emerges from the analysis of empirical data. If so, the traditional divide between theoretical and empirical economics diminishes considerably. How this might operate is now considered. In contrast to, say, Zwass (1979), critical realism seeks to go beyond description. Instead, it looks at a ‘deeper level in order to explain the phenomenon, to identify a causal mechanism responsible’ (Lawson, 1999b, p. 10). Lawson calls this approach ‘retroduction’,9 and contrasts it with both deduction and induction. For Lawson, deduction of an investigation involves ‘the move from the general to the specific’, (ibid. p. 10) while induction involves ‘the move from the specific to the general’ (ibid. p. 10).These definitions are broadly consistent with research methods texts.

Can there be (or should there be) a single critical realist methodology? However, these definitions are not particularly helpful in explaining what a deductive, inductive or even a retroductive research approach would actually entail. Indeed, one of the main criticisms of critical realism (see Arestis et al. (2002)) is that it tends to be very vague regarding how one might actually conduct research.

Transition in Eastern Europe

283

On one level, the fact that ‘critical realism is ontologically bold but epistemologically cautious’ (Bhaskar, 1989, p. 176) is seen as a strength. And it is, as it delineates a clear break from mainstream theory and its insistence on formal models as the central method for economic and social theory.10 Rejecting the notion that formalism is the only way to do research then frees the researcher to adopt methods that are appropriate for their area of investigation.While this might be heresy in economics research texts, this is a common theme in social science and business research methods texts (see May (2001), Saunders et al. (2000), and Frankfort-Nachimas and Nachimas (1992)). Within the research methods literature, a much richer definition of deduction and induction can be found. Deduction is defined as a research approach whereby a theory is developed and then a research strategy is developed to test the theory (Saunders et al., 2000, p. 87). Frankfort-Nachimas and Nachimas (1992, p. 46) calls this a ‘theory-then-research strategy’ (1992, p. 46).The weaknesses of this approach that Lawson (1997, pp. 16–18) states, such as its tendency to look for law-like states and its tendency to create a rigid methodology that does not allow for alternative explanations, are also found in the research methods literature (Saunders et al., 2000, p. 89). Within research methods texts, induction is defined as a research approach whereby the researcher ‘collects data and then develops theory as a result of data analysis (Saunders et al. 2000, p. 87), or a ‘research-then-theory strategy’ (Frankfort-Nachimas and Nachimas, 1992, p. 47). All three approaches – deduction, induction, retroduction – should be attempting to do the same thing, namely to explain a phenomenon and identify its cause. The inadequacies of deductivism11 in actually doing so have already been discussed. This does not necessarily mean that using a deductive approach is automatically the wrong thing to do, but this approach must be used with extreme caution. Using critical realist criteria, inductivism is harder to dismiss. According to Saunders et al. (2000, p. 91), induction emphasises: ● ● ●

●

Gaining an understanding of the meanings humans attach to events. A close understanding of the research context. A more flexible structure to permit changes of emphasis as the research progresses. A realisation that the researcher is part of the research process.

All of the above seem entirely consistent with a critical realist approach. The danger seems to lie with the research-then-theory focus of induction. Although the result of inductive data analysis is supposed to be the formulation of a theory, in practice the result may end up being very descriptive – that is, empirical realism, not critical realism. Using Lawson’s definitions of deduction and induction, retroduction becomes an either-or choice of method competing with the other two. Retroduction then becomes the ‘correct’ critical realist method, however there is little practical advice regarding how to do retroductive research. Using the

284

Karl Petrick

research methods literature’s definitions of these two approaches, retroduction is something that a researcher should be doing, regardless of what methodology is chosen. Either deductive research (i.e. testing existing theory) or inductive research (i.e. building theory from data analysis) could be consistent with critical realism as long as the research is also retroductive – it also seeks to explain the deeper causal mechanisms at work. And if the researcher is also explicit in viewing the world as ‘open, dynamic, and at least in part holistic’ (Lawson, 1999b, p. 14), either research approach should be entirely consistent with critical realism. Researchers may choose to test existing theory – as long as they are willing to modify or reject the said theory if the evidence does not support it, and then propose and test another theory. This is consistent with the intention of critical realism of ‘identifying which of the competing accounts is the most adequate’ for a given phenomenon (Lawson, 1999b, pp. 15).12 Likewise, an inductive approach – analysing data in order to build a theory – is consistent with critical realism as long as it does not stop at description and instead goes on to attempt to uncover the deeper causal mechanisms at work. In this manner, either research approach would yield a theory ‘which is historically and socially context-dependent and provides a framework by or through which the scientist can think about phenomena’ (Rotheim, 1999, p. 80). An example of such a project is described in the following section.

Empirical research in Eastern Europe: iterating towards a grounded theory of causality The focus of the author’s PhD research was to investigate the transition of the banking sector in Hungary and the Czech Republic.13 More specifically, the research aim was to determine how the relationship between commercial banks and their customers (primarily industry) was changing as both countries began to move towards a market economy. There were two research aims. First, the research intended to show how past (pre-transition) institutional legacies were affecting the transition. In short, before trying to discuss where each country was headed it was crucial to understand where they had been. This seems to be a statement of fact rather than a testable hypothesis, but this approach was missing from much of the literature regarding the economic transition in Eastern and Central Europe (ECE) at the time. As such, it seemed important to fully consider the historical context of changes in the banking sectors in both countries, not just once the countries had begun the move away from central planning, but also pretransition reforms as well. The second aim was to critique Financial Liberalisation (FL) theory, which was being used at the time as a blueprint for changing the financial sectors throughout ECE. Initially this was to be simply an empirical critique – that is, had the promised results from FL theory actually happened. Having attempted and failed to apply a Structuralist critique to FL theory during the author’s

Transition in Eastern Europe 14

285

Master’s thesis, no particular theory was in mind to replace FL theory when the research began. However, during the research a theory emerged – one that not only was relevant for critiquing FL theory but which also was consistent with the facts discussed in the empirical literature. The research involved not only reviewing secondary sources from a wide range of areas,15 but also a large component of empirical research as well. Fieldwork was conducted for just over a year (eight months in Hungary in 1996, seven months in the Czech Republic in 1997). One reason was to collect the most up-to-date quantitative data regarding the banking sectors – the sheer speed of change at the time made it difficult to collect such data from abroad. An equally important reason was a desire to collect a lot of qualitative data – the position being that ‘understanding the rationales underlying decisions that have been taken in such a complex and continuing process as the transition to a market economy are just as important as the material effects of these decisions’ (Petrick, 1998c, p. 5). The institutional and historical context that the decisions were being made in was very important. There was no intention to theorise from afar, but to understand the ‘insider’s view’ as much as possible. To conduct the research, two methods were considered: questionnaires and interviews. Questionnaires were rejected for a number of practical reasons – language difficulties (all of the subjects would speak English as a second language), low response rates, and the need for simple questions in such an approach being paramount (Petrick, 1998c, p. 4). Interviews seemed a better data collection method, as it provides the ability to overcome language problems,16 as well as the ability to ask follow-up and more detailed questions.17 The people to be interviewed were high-level officials from the domestic commercial banks, from each country’s National Bank and Ministry of Finance. Academics and members of other institutions connected with the transition of the banking sector, such as each country’s Privatisation Agency, and the local branch of the World Bank were also interviewed.This ‘elite interviewing’ has its own set of risks and rewards, but as Schoenberg (1991, p. 181) states, ‘interviews offer unusual access to the often conflicting and shifting strategic logic and historical contingencies that underlie corporate decisions. Carefully used, it can provide a qualitative context that amplifies and enriches the meaning derived from quantitative methods’. A series of 45 interviews with 30 officials in Hungary and 33 interviews with 26 officials in the Czech Republic were conducted. In terms of gaining insight to the rationales underlying decisions, as well as providing context for information gathered from both primary and secondary sources, the research was a success. A very interesting exchange of viewpoints between the interviewees and interviewer developed.18 Multiple interviews were conducted with some officials in order to ask follow-up questions after more data had been collected.This provided an opportunity to ‘test’ the theoretical ideas that were emerging as the research was ongoing. From their responses, the theoretical aspect of the research was also a success.

286

Karl Petrick

The structure of the thesis is as follows: chapter 1 simply states the research aims and justified the research methodology. Chapter 2 is a literature survey regarding financial development and economic growth.This section looks not only at economic theory, but also empirical economic research19 – although the two may have developed to a large extent separately, it is possible to note areas of agreement as well as disagreement between the two strands of literature. The McKinnon–Shaw hypothesis of financial liberalisation is then covered. The chapter then discusses the economics of information and incentives, notably Stiglitz and Weiss’ work on credit rationing.This then leads us to a Post Keynesian critique of financial liberalisation theory, notably the work of Grabel (1995).20 One of the challenges faced in this research was that not all of the seminal theoretical works on banking and finance discuss finance in a developing country, and none of it refers to the situation found in a centrally planned economy. Even for the FL literature, the initial works considered central planning to lead to such a repressed financial system that it was actually outside their terms of reference.21 For this reason, as well as for the need to discuss the past institutional structure found in ECE countries, chapter 3 discusses the role of money and banking in a centrally planned economy. Both had a very different role than that found in a market economy, due to the unique institutional characteristics of a centrally planned system.This is a general model discussing structure found in a centrally planned economy – the unique characteristics found in Hungary and the Czech republic are discussed in later chapters. As the works discussed are all empirically based – none besides Podolski (1972) has much theory in it – the chapter itself is empirical in nature. Chapter 4 discusses the literature on endogenous money. Initially the chapter discusses the principles of endogenous money in a market economy, notably the work found in Arestis (1992 and 1997).This is a fairly short exposition, as its main purpose is to set the stage for discussing endogenous money in a centrally planned economy. The starting point for this research is a short paper published in the Journal of Post Keynesian Economics. Norcic´ et al. (1996) discuss endogenous money in the transition economy of Slovenia.This was published a few months into the field research in Hungary, and it was thought that the theory that they had started to develop could also be applied to other transition economies.The initial starting point for this was the realisation that they were discussing issues relevant to the author’s research in Hungary. Secondary sources of data were also drawn upon. Norcic´ et al. (1996) cite several works that support their claims, but all of their sources discuss Yugoslavia.This is understandable, since Slovenia is a former part of Yugoslavia. But their argument could be strengthened considerably by looking at the structures in other centrally planned economies, which was the context of the author’s research. First, from secondary sources, there was a small amount of work in a similar vein discussing pre-transition Hungary, and Podolski (1972) discusses endogenous money in pre-transition Poland. Even more significantly, support for this view existed in the empirical literature of

Transition in Eastern Europe

287

Zwass (1979), Garvy (1966, 1977) and others. Although this work is entirely empirical and descriptive, their description of how the banking system worked in a centrally planned economy actually explains how money is still endogenously determined even under such a system. Even though both banks and firms were state-owned and theoretically at the beginning of every year central planners decided (among other things) the level of bank credit that would be used, the demand for credit by firms during the production process ultimately determined the amount of money in the system. Norcic´ et al. call this ‘political command endogeneity’ to take into account the political nature of economic decisions in such a system.22 Discussing the change in economic system, Norcic´ et al. (1996) develop the concept of ‘dual money endogeneity’ – as the transition begins, ECE countries find themselves between two systems.There is still a strong element of ‘political command’ endogeneity in the system, but gradually the actions of banks and firms begin to act more in line with those in a market economy.The remainder of the author’s research thus develops this idea using the transition in Hungary and the Czech Republic as case studies. After the initial research in secondary sources, the empirical work, conducted first in Hungary and later in the Czech Republic, provided an important resource within which to develop this theory. A number of the persons interviewed had also been government officials pre-transition. Some of them had also been involved in key reforms pre-transition.23 Through these people, it was possible to test how well this theory was capturing as to how these banking sector operated in these countries pre-transition.Also through them and others, it was possible to test how relevant the theorising was compared to how the banking sector was actually working during the transition, as well.24 Chapter 5 is an introduction to the empirical chapters. Chapter 6 discusses Hungary, chapter 7 discusses the Czech Republic. Both follow the same structure – first some historical discussion regarding the structure of the banking system in each country, and pre-transition reforms.Then there is a lengthy discussion regarding the changes taking place in both countries. Chapter 8 concludes, and discusses the development of a ‘dual endogeneity’ view of the banking sectors in transition. This view is contrasted with the view found in FL literature, which is found to be inadequate in describing how banks and firms act both in a ‘repressed’ structure (in this case, pre-transition) and after the financial system has been ‘liberalised’. The actual structure of the thesis – which in essence comprised two different discussions of theory, each followed by connected empirical research, and then a final concluding section – was not how the research had originally been planned. The sheer size and scope of the literature being investigated made having a single theoretical chapter impossible. What the structure did allow was to show the synthesis between various strands of economic theory, as well as the continued links between the structures found in ECE countries pre-transition and during transition. Even though there had been a major and sudden break (namely the ending of central planning),

288

Karl Petrick

how social agents acted in practice had not changed as quickly. Mainstream theory, with its ahistorical approach does not allow for this. As a result, the actions of firms within the transition economies seem to have puzzled orthodox economists. For example, two papers written by World Bank (Blejer and Sagari, 1992) and IMF (Calvo and Frenkel, 1991) economists mention the ‘queuing problem’, an increase in the number of firms who did not pay their suppliers (who then had trouble paying their suppliers, etc.) in the early transition. The obvious cause of this was a sharp decrease in the amount of bank credit available. For the authors stated above, this was a new and perplexing problem – it was not how firms were supposed to act in a market economy. However, to understand the root of the problem knowledge of how banks and firms acted during central planning is needed.This ‘new problem’ was actually a historical one that had plagued every attempt to tighten credit policy pre-transition.25 In this light, it should not be surprising that firms responded to a decrease in credit by ignoring their outstanding debts. It was how they had responded to this situation for nearly fifty years. This key causal process, to explain the regularity of behaviour, was thus revealed by exploring the historical and social context within which agents operated by a variety of broadly qualitative methods.This approach is thus highly suggestive of critical realism and retroduction at work. It contrasts starkly with a research project that would seek to test ahistorical/ asocial theoretical predictions with broadly secondary data as is typically the case in mainstream research.

Conclusion This chapter has discussed the central elements of the critical realist perspective.These are contrasted with mainstream economic theory, which is, in turn, contrasted with empirical economic research. It has been argued that while mainstream theory may, in some sense, never be compatible with empirical investigation, critical realism offers a constructive logic within which to bridge the gap between empirical investigation and theory generally. One of the criticisms of the critical realist literature is that it offers very little in terms of advice for how to conduct research. One way to solve this problem is to examine research methods texts and evaluate their advice regarding methodology using the tenets of critical realism as criteria.The earlier discussion suggests that this will not lead to a single critical realist research method as a result. However, in this way it is possible to identify what critical realist research should entail, while leaving the exact research methods chosen to be determined by this question – what methods are appropriate for the research question posed? This is the common practice in social science and business research, and something which economics should emulate. Finally, some research that the author conducted in Eastern Europe is provided as an example of how this interpretation of critical realist empirical research can be put into practice. A key finding is that theory, for example, of

Transition in Eastern Europe

289

causal mechanisms emerges at the same time as empirical observation. It is hoped that any economist planning to do empirical research will find this discussion helpful. It should be noted in closing that the research procedure noted above was, in part idiosyncratic, balancing the needs of a PhD researcher with a lack of obvious pointers on how to conduct research with a critical realist perspective in economics, drawing instead on a formal training in economics yet sympathy towards heterodox economic ideas. It is comforting to note now that in many respects what emerged was consistent with a grounded theory approach to social science research. This approach is, of course, discussed in other chapters.

Notes 1 An earlier version of this chapter was presented at the Association of Heterodox Economics Annual Conference, 8–10 July, Dublin City University, Ireland. I would like to thank Paul Downward, Wendy Olsen, Siobhan McGovern and Paschal O’Gorman for helpful comments on that draft. Also thanks to Andy Brown for insightful discussion regarding critical realism and critiques of this approach. 2 Fleetwood (1999) discusses critical realism and economics generally.The Journal of Post Keynesian Economics dedicated an entire issue (Fall 1999) regarding the use of critical realism in Post Keynesian economics, Brown et al. (2002) do the same for Marxism. 3 Although social structures can provide limitations on an agent’s ability to act. 4 That is, critical realism seeks to uncover not just what is, but also why it is. 5 This allows for very precise predictions to be possible, at least theoretically. But it also tends to lead to a very static type of worldview. Once a law-like state has been established, the event should be permanent, and therefore the possibility of change has been assumed out of the model. 6 A considerable change of heart considering that Greenwald and Stiglitz (1992, p. 34) says that ‘(economists) have gradually come to understand why the (Modiglini and Miller) theorem is of little relevance’. 7 See Brown et al. (2002) and Fleetwood (1999), although, as indicated in these works, the process is by no means seamless. 8 Although his view that ‘theory may not have always kept pace with practical experience but practice itself has long since abandoned theory’ (Zwass, 1979, p. 121) makes it clear why he is not interested in theory. 9 For current purposes, this definition will suffice. However, another critique of critical realism involves the practice of retroduction and the incomplete use of abstraction that it entails. Brown et al. (2002) address this issue in detail and devise a solution, which they term ‘systematic abstraction’. Their work lies outside the terms of reference for this chapter, however some of the points raised here support those found in their study. 10 This is a consequence of mainstream theory’s atomistic framework and its search for ‘law-like states’. 11 The deductive approach is supposed to evaluate existing theory and then either accept, modify or reject the initial theory in light of the evidence. In practice, the rigid research design that this approach imposes tends to lead to tinkering within the original framework rather than rejection or major modification (see Saunders et al., 2000, p. 89). 12 Arestis et al. (2001) (see discussed in Arestis et al., 2002 and this volume for a synoposis of this research) is a good example of this approach.

290

Karl Petrick

13 Petrick (1998c).An updated and slightly shorter version of the chapter that covered the empirical research in Hungary was published as Petrick (2002). 14 Petrick (1994). Grabel (1996) also discusses the difficulty in using such an approach to critique FL theory, and begins to develop a Post Keynesian critique. This work strongly influenced the author’s position. 15 Notably, economic theory regarding banking and finance, theory and empirical/historical work regarding finance and economic development, economic history regarding finance in centrally planned economies, and theory/empirical work regarding banking in transition economies.The last body of literature at the time was by far the smallest, as the transition had begun only a few years earlier. 16 The researcher can re-phrase questions if they are misunderstood. In questionnaires, the questions asked have to be uniform in order for comparisons to be made.There is no way to correct a respondent’s misperceptions without invalidating their responses. 17 Petrick (1994) served as a useful ‘pilot study’ for the research, as here the author had undertaken a small number of interviews in each country during research for an MA. 18 It seemed to have been a novelty for a Westerner to arrive and ask them what the problems that they were encountering were.As a result, they just were as interested in the author’s point of view as an ‘outsider’ as the author was in theirs as an ‘insider’. 19 Particularly the economic history work of Alexander Gerschenkron (1962), which greatly impressed the author. 20 This structure was chosen because the Grabel’s work not only critiques FL theory but also critiques and builds on the literature regarding credit rationing. 21 Williamson (1975) does something similar. He states that a centrally planned system is a system where markets have been replaced by hierarchies. But it is otherwise outside his terms of reference, so he quickly moves on. 22 They label this a type of accomodative endogeneity, but my own research (1998a,b, 2000) suggests that there is also an element of structural endogeneity as well. 23 Such as Rezsö Nyers, who had chaired the commission that planned the ‘New Economic Mechanism’ one of the major reforms in pre-transition Hungary (1956). He was head of Hungarian State Banking Supervision when I interviewed him. 24 Something the author continued to do after finishing the empirical work. Copies of each chapter were given to colleagues in Hungary, Poland and the Czech Republic ‘in order to ensure that the conclusions did not stray too far from the realm of reality’ (Petrick 1998c, p. 8).The author also presented the theoretical work in a couple of different venues. Petrick (1998a) was presented to a group of East European and Russian scholars, but only one other economist. For Petrick (1998b) the audience was a group of financial economists, but only a handful of those were familiar with Eastern Europe. Petrick (2000) was presented at a Union of Radical Political Economy session at the annual ASSA conference. All three presentations provided valuable feedback regarding development of the theory. 25 And one that is amply covered in Zwass (1979), Garvy (1966 and 1977), Podolski (1972) and many other works regarding pre-transition finance.

References Arestis, P. (1992) The Post Keynesian Approach to Economics. Aldershot: Edward Elgar. —— (1997) ‘PKE Theoretical Aspects of Money and Finance’, in Arestis, P. (ed.) Money, Pricing, Distribution and Economic Integration. Basingstoke: Macmillan. ——, Brown, A. and Sawyer, M. (2001) The Euro: Evolution and Prospects. Cheltenham: Edward Elgar.

Transition in Eastern Europe

291

Arestis, P., Brown, A. and Sawyer, M. (2002) ‘Critical Realism and the Political Economy of the Euro’, The Levy Economics Institute Working Paper No. 352, Annandale-on Hudson, New York. Barclay, M. and Smith, C. (2001a) ‘The Capital Structure Puzzle: Another Look at the Evidence’, in Chew, Jr. D. (ed.), The New Corporate Finance: Where Theory Meets Practice. New York: McGraw-Hill, pp. 197–209. —— (2001b) ‘On Financial Architecture: Leverage, Maturity, and Priority’, in Chew, Jr, D. (ed.), pp. 210–223. Bhaskar, R. (1989) The Possibility of Naturalism. London: Harvester Wheatsheaf. Blaug, M. (1992) The Methodology of Economics, Second Edition. Cambridge: Cambridge University Press. Blejer, M.I. and S.B. Sagari (1992) ‘Hungary: the Initial Stages in the Financial Sector Reform of a Socialist Economy in Transition’, Applied Financial Economics, 2(1): 33–42. Brown, A., Fleetwood, S. and Roberts, J.M. (eds) (2002) Critical Realism and Marxism. London, Routledge. Brown, A., Slater, G. and Spencer, D. (2002) ‘Driven to Abstraction: Critical Realism and the Search for the “Inner Connection” of Social Phenomena’, Cambridge Journal of Economics, 26(6): 773–778. Calvo, G.A. and J.C.A. Frenkel (1991) ‘From Centrally Planned to Market Economy’, IMF Staff Papers 38(2): 268–299. Fleetwood, S. (1999) Critical Realism in Economics: Development and Debate. London: Routledge. Frankfort-Nachimas, C. and Nachimas, D. (1992) Research Methods in the Social Sciences, Fourth Edition. London: St Martins Press. Garvy, G. (1966) Money, Banking and Credit in Eastern Europe. New York: Federal Reserve Bank of New York Press. —— (1977) Money, Financial Flows and Credit in the Soviet Union. Cambridge, MA: Bollinger Publishers. Gerschenkron,A. (1962) Economic Backwardness in Historical Perspective. Cambridge, MA: Harvard University Press. Grabel, I. (1995) ‘Speculation-led Economic Development: A Post Keynesian Interpretation of Financial Liberalisation Programmes in the Third World’, International Review of Applied Economics, 9(2): 127–149. —— (1996) ‘Financial Markets,The State and Economic Development: Controversies Within Theory and Politics’, International Papers in Political Economy, 1–42. Greenwald, B. and Stiglitz, J. (1992) ‘Information, Finance and Markets: The Architecture of Allocative Markets’, in V. Zamagni (ed.) Finance and the Enterprise. San Diego: Academic Press. Hands, D. (1999) ‘Empirical Realism as Meta-Method’ in Fleetwood, S (ed.), Critical Realism in Economics: Development and Debate. London: Routledge, pp. 169–185. Hutchinson, T (1938) The Significance and Basic Postulates of Economic Theory. London: Macmillan Journal of Post Keynesian Economics (1999), 22(1). Lawson,T. (1994) ‘The Nature of Post Keynesianism and its Links to Other Traditions: A Realist Perspective’, Journal of Post Keynesian Economics, 16(4): 503–538. —— (1997) Economics and Reality. London: Routledge. —— (1999a) ‘Connections and Distinctions: Post Keynesianism and Critical Realism’, Journal of Post Keynesian Economics, 22(1): 3–14.

292

Karl Petrick

Lawson,T. (1999b) ‘Developments in Economics as Realist Social Theory’, in S. Fleetwood (ed.), Critical Realism in Economics: Development and Debate. London: Routledge, pp. 3–20. —— (2001) Social Research: Issues, Methods, and Practice, Third Edition. Buckingham: Open University Press. Miller, M. (2001) ‘The Modigliani-Miller Propositions After Thirty Years’ in Chew, Jr D. (ed.), The New Corporate Finance: Where Theory Meets Practice, New York: McGraw Hill, pp. 184–196. Norcic´, O., M. Lah, A. Susˇ jan (1996) ‘Dual Money Endogeneity in Transition Economies’, Journal of Post Keynesian Economics, 19(1): 73–82. Petrick, K. (1994) Problems and Prospects: A Survey of the Commercial Banking Sector Reforms in Hungary and in the Czech Republic, 1987–1994, Unpublished MA Dissertation, Leeds University. —— (1998a) ‘Endogenous Money in Eastern Europe: General Concepts and an Application to the Case of Hungary’, presented at the Irish Association of Russian and East European Studies Conference, Belfast, Northern Ireland. —— (1998b) ‘Dual Money Endogeneity in Transition Economies: The Case of Hungary’, presented at The Alternative Perspectives on Finance conference, Turku, Finland. —— (1998c) The Role of the Banking System in the Transitional Economies of Hungary and the Czech Republic. Unpublished PhD thesis, Leeds University Business School. —— (2000) ‘Endogenous Money in Eastern Europe’. Presented at the Alliance of Social Studies Associations annual conference, 7–9 January, 2000, Boston, MA, USA. —— (2002) ‘Hungarian Banking in Transition’, in Green and Petrick (eds), Banking and Financial Stability in Central Europe: Integrating Transition Economies into the European Union, Cheltenham: Edward Elgar, pp. 125–149. Podolski, T.M. (1972) Socialist Banking and Monetary Control. Cambridge: Cambridge University Press. Rotheim, R.J. (1999) ‘Post Keynesian Economics and Realist Philosophy’, Journal of Post Keynesian Economics, 22(1): 71–103. Saunders, M., Lewis, P. and Thornhill, A. (2000) Research Methods for Business Students. Harlow: Prentice Hall. Schoenberger, E. (1991) ‘The Corporate Interview as a Research Method in Economic Geography’, Professional Geographer, 43(2): 180–189. Smith, C. (2001) ‘Raising Capital: Theory and Evidence’ in Chew, Jr D. (ed.), The New Corporate Finance:Where Theory Meets Practice, New York, McGraw Hill, pp. 277–293. Stiglitz, J. (1997) Economics, Second Edition, New York: Norton. Williamson, O.E. (1975) Markets and Hierarchies: Analysis and Antitrust Implications. A Study in the Economics of Internal Organisation, London and New York:The Free Press, Collier Macmillan. Zwass, A. (1979) Money, Banking and Credit in the Soviet Union and Eastern Europe. London: Macmillan.

17 Conclusion Paul Downward

Introduction This book explores how various researchers have attempted to both articulate inferences drawing upon empirical work while framing their analysis within a critical realist philosophical framework, or specifically deliberated upon the role of empirical work in critical realist epistemology. The emphasis is very much upon a constructive dialogue between philosophy and practice that has been taking place within the Workshop on Realism and Economics,The Post Keynesian Economics Study Group and the Association of Heterodox Economics. In essence the main issue discussed concerns how to avoid the deductivism of mainstream economics, wherein ‘covering’ laws presented as constant conjunctions of events are postulated as causal explanations. In particular the book is concerned with avoiding the empirical realism of mainstream economics, which embraces the idea that the determinate account of phenomena developed by deductive methods can be fully described empirically at the level of events. Econometrics is typically criticised in facilitating this endeavour. It conflates sense experience and knowledge, committing an epistemic fallacy, by sharing the closed-system ontology of deductivism. The alternative open-systems view embraced by critical realism is that the real causes of phenomena will be a complex codetermining set of structures that are ontologically distinct from both events and our empirical experience of them. Accordingly understanding them must proceed through a different logic of inference than either the deductive or inductive emphases of mainstream economics. Retroduction should be emphasised which implies moving, ontologically, from the level of events to the real causes of them. However, this book accepts and addresses the argument that the detail of how this mode of reasoning is to proceed is generally unclear and undeveloped.To this end, Part I provides an overview of both the origin and nature of critical realism as well as the challenges that critical realism faces in providing an effective epistemology, and the constraints that it poses for effective empirical enquiry. Part II addresses conceptual issues associated with formalism and empirical work in critical realism and Part III presents some examples of empirical work drawing upon and, in turn, assessing critical realism.

294

Paul Downward

Summary of chapters Specifically in Chapter 2, in Part I, Sheila Dow argues that in contrast to mainstream and constructivist approaches in economics, critical realism offers a powerful opportunity to provide a more meaningful access to the reality of economic relations. Sheila does, however, argue that while emphasising a generally pluralist–fallibilist approach, the important agenda is now to build on the general philosophical foundations with reference to more explicit direct applications of critical realism to demonstrate the practical difference which such an approach makes. She emphasises that such applications in turn will provide material to be fed back to the philosophical level. Econometrics may well be the classic case in point here. Yet, the challenges that this raises are pointed out in Chapters 3 and 4. In the former case, Steve Fleetwood shows data, such as falling unemployment figures, are often taken within the mainstream approach to indicate that the (now) flexible UK labour market is gradually solving the problem of unemployment. However, Steve shows that the very conceptualisation, and hence measurement, of unemployment is inadequate. Deductivism which places an emphasis upon quantification reduces the multi-dimensional, qualitative reality of employment and unemployment to the quantitative, single dimension of variables, whereupon they cease to be adequate expressions of the reality they are designed to investigate. In this regard contemporary unemployment figures could indicate the emergence of atypical employment and the deterioration in the quality of employment. Yet, through a wide ranging critique, Bernard Walters and David Young argue in Chapter 4 that quantification and empirical work is necessary to the critical realist project. Indeed it is necessary to replace a rather overworked broad philosophical critique of mainstream economics. Thus, Walters and Young claim that critical realism presents a highly specific definition of positivism that belies Hume’s caution over claims of causality and presentation of a covering law mode of explanation. Relatedly they argue that it is difficult to provide a single dimensioned presentation of mainstream economics. They argue, thus that the broad philosophical sweep of critical realism is unfounded. They also argue that allied to this lack of caution – or ontological boldness – critical realism lacks an operational epistemology and, specifically, robust rules for uncovering causes and deep structures. Paradoxically, therefore, critical realism rejects neoclassical empirical methods and yet postulates that some form of empiricism is needed. Part II thus begins to address this issue. In Chapter 5, Mark Setterfield investigates the extent to which formal modelling is, or can be made, consistent with the ontology and methodology of critical realism. He argues that formal modelling can be separated out from the practices of mainstream economists through recognising a concept of conditional closure.This theme is picked up in Chapter 6, where Paul Downward, John Finch and John Ramsay attempt to set out some guidelines for making empirical inferences within critical realism.

Conclusion 295 They argue that crucial to the legitimacy of attempts to use empirical insights is to recognise a degree of quasi-closure developed by economic agents as they form inferences and which can be manifest in institutional systems.To this end, they conclude that qualitative methods such as grounded theory can be triangulated to quantitative analyses to establish acceptable explanations. The type of legitimate quantitative analysis amenable to critical-realist analysis is further discussed in Chapters 7 and 8. First, Paul Downward and Andrew Mearman argue that econometric techniques generally can play a role in analysis framed within a critical-realist perspective.They argue that a degree of tension exists between the philosophical pronouncements of critical realism and the consequences of referring to empirical concepts in the practice of inference, which is something that critical realists wish to do. They review econometric methods to juxtapose against Lawson’s (1997) redefinition and critique of econometrics and show that Lawson’s redefinition suffers from the same problems. Thus, they offer some proposals for empirical inference that shares the broad concerns of critical realism, yet embraces econometric work. More guardedly, in Chapter 8, John Finch and Robert McMaster reject an emphasis on econometrics as regression and instead argue that researchers in economics who wish to develop and assess the reliability and adequacy of knowledge claims may draw upon simple non-parametric statistical analyses. In this way theoretical development may be informed by empirical analysis and vice versa. They thus propose a pragmatic alliance between non-parametric techniques and, essentially, an open-system ontology, in interpreting and analysing stylised facts or demi-regularities across classes or categories of events. The final two chapters in Part II conclude with further discussion of the nature of inference within critical realism. In Chapter 9,Wendy Olsen addresses issues relevant to the validity of time-series econometrics by exploring the problems associated with assuming that things exist over time. Wendy argues that time-series data should be augmented as part of a triangulated economic research strategy. In Chapter 10, Fred Lee explores the role of grounded theory in more detail arguing how grounded theory provides the best set of guidelines for theory creation in critical realism because it allows for an elaboration of causal mechanisms. The arguments are illustrated with reference to Post Keynesian pricing theory.Yet, Fred argues that integrating critical realism and the method of grounded theory will result in a need to recognise that causal mechanisms are observable. In this sense there are challenges to critical realism. In Part III, emphasis shifts more fully to providing examples of critical realism in empirical economic analysis. In Chapter 11,Wendy Olsen,Thandie M. Hara and Sampson Edusah provide an example of empirical work based on fieldwork in Ghana, discussing the role of regression analysis conducted upon the cross-sectional data. In Chapter 12, Brian Pinkstone examines the role of econometric methods in economic history. Brian uses examples from his own research in the field of economic history to illustrate an alternative approach to using data than econometrics. In Chapter 13, Philip Arestis, Andrew Brown and Malcolm Sawyer discuss some of the main problems that critical realism

296

Paul Downward

raises for empirical work and inferences with reference to a research project on the analysis of the single European currency.Yet, while the authors argue that critical realism provides a method that can help understanding of concrete levels of analysis – illustrated by the example of the attempt to explain the falling value of the Euro – they argue that this analysis is partial and inappropriate to the most abstract and fundamental levels of theory. In Chapter 14, Paul Downward and Andrew Mearman illustrate their arguments of Chapter 7, by drawing upon applied work in Post Keynesian pricing. Historical and contemporary cases are offered that are suggestive of the merits of econometrics and an exemplar of the triangulation strategy that they advocate. In Chapter 15, Murray Glickman presents an example of Post Keynesian empirical work based on a broadly critical-realist understanding of the nature of economics. In rejecting stochastic analysis and econometrics Murray provides empirical evidence to support his theoretical critique of the Modigliani–Miller (MM) propositions on capital structure. Murray’s work is based on a ‘facts first’, explanation-oriented, approach to economics. Finally, in Chapter 16, Karl Petrick finishes the part by providing an illustration of how qualitative research methods, embedded in a critical-realist perspective, can yield insights into the process of transition. Karl argues that one of the criticisms of the critical-realist literature is that it offers very little in terms of advice for how to conduct research. One way to solve this problem is to examine research methods texts and evaluate their advice regarding methodology using the tenets of critical realism as criteria. Karl suggests that this will not lead to a single critical-realist research method, but it is possible to identify what critical-realist research should entail, while leaving the exact research methods chosen to be determined by this question – what methods are appropriate for the research question posed? Karl’s research project into banking in Eastern European transition is then provided as an example of how this interpretation of critical-realist empirical research can be put into practice showing how knowledge of causal mechanisms emerges at the same time as empirical observation.With hindsight Karl identifies that the approach adopted was consistent with grounded theory.

Towards applied critical realism Recognising some caveats One aim of this book was to offer some general reflections on what serious debate has produced in terms of fleshing out an epistemology for critical realism. At this point it is worth reiterating the point, made in the introduction, that this volume was deliberately conceived and constructed as an attempt to draw together different scholars’ perspectives on critical realism and empirical methods. The aim was very much to capture debates and discussion that has evolved in and around the organisations noted earlier without a strong steer by the editor towards consensus.

Conclusion 297 This has a number of implications. The first is that this makes the analysis partial. However, the defence of the book lies in three, further, arguments.Two are closely related. These are that the focus is upon economics. In this sense contributions from other ‘disciplines’ is deliberately not addressed. This is not because they are unimportant but that, as economists, the contributors have a peculiarly resilient mainstream approach to economic analysis to engage with. In this respect the Workshop on Realism in Economics, the Post Keynesian Study Group and the Association of Heterodox Economics have been organisations within which debate has flourished and persisted. Finally, it is clear that within an ‘open-systems’ approach some movement towards addressing the thorny issues of the practice of critical-realist epistemology is to be welcomed. In this respect it is felt that the book has something constructive to offer established academics and postgraduates alike, who accept the broad sweep of critical realism, but are unsure about how next to proceed in applied work. The background is, of course, a world in which the pressure to publish, or complete a PhD is increasing together with a need to offer policy relevant insights. In this regard, another point worth emphasising is that the book project does not carry with it a strong prescriptive intent and, or the impression that presents a simple set of guidelines upon how to ‘apply critical realism’.Yet, after reviewing the chapters and noting the caveats above, it is felt that some general points of consensus upon ‘applied critical realism’ do emerge.These are presented next, in the final part of the book. Of course, the reader may feel otherwise and, once again, it is hoped that these ideas stimulate further constructive discussion. ‘Applied critical realism’ – a personal view It seems apparent that the arguments presented in the preceding chapters tend to emphasise, naturally enough, that practice will inevitably feed into philosophical discussion. Moreover, to a greater or lesser extent some form of quantitative analysis must necessarily follow in critical-realist work. Yet, through synthesising the contributions, more specific guidelines seem to be forthcoming – though this does not necessarily commit specific contributors to this synthesis. Figure 17.1 attempts to outline the components of a proposed synthesis by the editor. In the first column are the two opposing positions identified in the criticalrealist literature. Critical realism and empirical realism.The latter is, of course, the empirical counterpart to the deductivist mode of explanation, which is the main target of the critical-realist critique in economics. The next column describes the focus of analysis framed within these alternative philosophical positions, which are real causes and empirical events, respectively. In the case of empirical realism, relationships between events are purported to reveal actual causes.The broken line link to (conventionally defined) ‘econometrics’ reveals the method available to the mainstream economist. In contrast, critical realism maintains that investigating causes involves moving below the level of events through retroduction. However, the ‘method’ to be used in this process is

298

Paul Downward

Philosophical position

Focus of analysis

Applied method

Critical realism

Cause

‘Qualitative’ Grounded theory/discourse analysis etc. (meaning/categories/contextual relations)

Retroduction

Quasi-closure

Triangulation

‘Quantitative’ Sample specific Univariate/bivariate/multi-variate Parametric/non-parametric e.g. Descriptive statistics Frequencies, cross-tabulations Correlations Regression Factor analysis ANOVA Inferential Univariate/bivariate/multi-variate Parametric/non-parametric e.g. Empirical realism

Events - - - - - - - - - - - - - - - - (as causes)

‘Econometrics’ Tests of mean/variance differences Tests of association Tests of overall and specific parameter significance Canonical regression Discriminant analysis Cluster analysis

Figure 17.1 Applied critical realism: a schema.

unclear in the literature. Thus, the third column reveals a suggested applied method for critical-realist research. Corresponding to what is broadly identified as ‘qualitative’ research methods – grounded theory being emphasised in the volume – the context-laden meaning of concepts, categories and relationships can be established and causal narratives constructed. However, what the contributions to the volume suggest is that this process of defining shared meanings and categories in essence begins to break down the qualitative orientation of the research and, of necessity, begins to invoke aspects of closure. Hence, the direction of quasi-closure indicated in the column.Thus, categorising phenomena implies assuming invariant (or a degree of invariance in) qualities. In essence defining a variable requires intrinsic closure at least.1

Conclusion 299 Increasing the degree to which closure is invoked thus begins to ‘legitimise’ various statistical procedures.These can refer to one, two or many variables but be either sample-specific or inferential in orientation – that is, draw upon a probability distribution referred to in a parametric or non-parametric manner to make claims about the generality of the purported relationships between variables. It is, of course, clear that the degree of closure assumed increases as one moves down the column. For example referring to sample-specific descriptive methods of analysis such as averages, correlations or regression implies that values of variables have consistent meanings and that these variables can be combined in a relatively constant or enduring manner as indications of, say, outcomes of causal links. Probabilistic inferences assume, much more strongly, that the results carry over, in a measurable sense, to contexts beyond the sample. In this sense one is increasingly invoking the extrinsic condition of closure and, one can clearly see the strength of the assumptions underpinning empirical-realist methods defined as ‘econometrics’ which appears in this column.Yet, what the schema also hopes to reveal is that, in fact, the researchers have, at their disposal, these techniques of analysis because they have no necessary link to a wider philosophical orientation.Thus, what we understand as ‘econometrics’ as a means of testing deductively constructed covariation is, in part, simply an application of a research method that is not, and should not be, unique to the mainstream researcher. Hence, the broken line which links an empirical-realist analysis of event to ‘econometrics’. This part thus also includes examples of other techniques that do not necessarily seek to present a relationship in which the variation in one ‘dependent’ variable is partitioned into the effects of other individual ‘independent’ variables. In contrast techniques which test aspects of both parametric and non-parametric, that is metric and non-metric variables in univariate to multivariate settings are referred to. What the figure does reveal, however, is that, in general, movements towards statistical methods naturally shifts attention towards ‘events’ as opposed to ‘causes’ which are in essence qualitative. On this basis, one can argue that rather than revealing covering laws, in contrast, statistical methods can reveal demiregularities from which causal research can begin – a theme stressed for example in Economics and Reality. Thus the final column also reveals that triangulation between the methods – that is, linking the insights gained from these different research methods can attempt to capture the retroductive logic of critical realism. The figure reveals two further insights, however. The first is that retroduction necessarily involves ‘weakening’ the closure assumptions of empirical analysis despite the emphasis of the particular technique – that is, by explicitly probing the interpretation and meaning of empirical categories and relationships.There is a clear (ontological) shift of emphasis involved in this approach to using empirical methods.The second, is that the process of triangulation, in fact, runs two ways.This is because of the need to link fallibly defined specific

300

Paul Downward

contexts to more general contexts to generate a sense of their wider legitimacy and potential policy relevance. This is not simply the amplitative logic of induction because of the ontological shift just defined. The logic is based in retroduction. However,‘causal’ and ‘event’- level analyses are mutually supportive within the schema revealing that inferences have an ontological as well as statistical orientation.The direction of analysis is not, of itself, important. In closing, some final comments are worth noting. The first is that the schema naturally, indeed necessarily, allows for contrastive demi-regs to be a source of potential investigation. Many of the techniques of analysis not implied in traditionally defined ‘econometrics’ can play a role here. Breakdown in apparent regularities then requires analysis of the causal–structural relationships that determine them. Embracing statistical methods thus does not, of itself, imply an a priori acceptance of closure or the necessity of regularities. In contrast the latter need to be established and their structural composition and reason for change investigated. It is hoped that this book reveals, therefore, that the empirical methods can be of much use in this regard.The critical realist, moreover, should not feel unduly uneasy about their use. Second, however, there are clearly matters raised in the schema that require further analysis and discussion. These are related to the idea that practice and philosophy can inform one another. One important question is clearly to think in more detail about the role that different empirical techniques might offer. In other words what their emphases might be in attempting to reveal something about complex social systems. This clearly involves a consideration of how and if ‘research methods’, for example, detailed in texts that tend to focus on technical issues, can be differentiated according to critical-realist concepts. Linked to this issue, is that many of the contributions in this book suggest that in many ways the deductive/inductive versus retroductive logic of inference may be challenged in as much as elements of the former methods are implied in the latter in practice. If this is the case then emphasising a distinction must rest in some explicit appeal to matters of ontology in a research project. That is, making appeal to the specifics of the strata of society that are being analysed. By themselves ‘data’ derived from individual techniques of analysis, whichever are considered admissable, will not be able to demonstrate or appeal to this idea because they refer to particular ‘levels’ of phenomena. The logic of retroduction thus seems to necessarily embrace triangulating techniques that appeal to different levels of analysis. It should be noted in this context, that some contributions to the book suggest otherwise and or rule out particular methods of analysis still. Finally, another important question to further consider is whether or not accepting a degree of closure, with essentially epistemological tools, in the attempt to analyse society with empirical methods fundamentally challenges claims about the (ontological) openness of the society that is analysed. If this is the case then the open-system/closed-system methodological distinction made by critical realism is challenged. In as much that thought and theory affect behaviour then, despite the fact that actions could always have been otherwise,

Conclusion 301 there is a sense in which one can argue that reality is not a fully open system. Institutionalised behaviour is one prime example in which ‘models’ of how to do things become existent (partially) independently of agents. The material nature of society thus echoes the epistemological closures. Analysis of these issues will thus clearly be important in taking forward the research agenda addressed, in the first instance, by this book.

Note 1 As an entity that has purported boundaries, however, extrinsic closure is, in essence, assumed.

Index

abduction 19–20, 72–3, 130, 144–5, 269, 271 ‘aboutness’ criteria 77, 83–4 abstraction 18, 20, 81–3, 114, 199, 235–7, 242–4 administered pricing 247–9 agency, human 73, 90, 173 Allen, R. 119 Alvesson, M. 101 analogy, use of 80, 97, 156–7 analytical induction 102 Andrews, P.W.S. 254 animal spirits reaction function 80–1 Annales School 226 Annells, M. 180 Appleby, J. 185 Arbuckle, J.L. 209 Archer, Margaret 98, 161 Arestis, P. 9, 253 assumptions of economic theory 13–14, 18, 74, 180; core and peripheral 83–4 average economic regression (AER) approach to econometrics 114–19 Atkinson, G.W. 180 axiomatic approach to modelling 74, 77, 84 ‘Babylonian’ perspective 249–50 Backhouse, R.E. 129, 131, 146 Baert, P. 57, 64, 82, 102 Balchin, A. 4, 41 banking 284–8 Barclay, M. 281 Barrell, R. 47 Bateman, B.W. 93 Bayesian methodology 115, 117, 120 Becker, G. 18, 55 Berkowitz, S.D. 173 Bernstein, J.M. 158, 159

Bhaduri, A. 164 Bhaskar, Roy 3, 16–17, 52, 57, 91, 99, 199, 221, 224, 230, 235, 283 Bigus, O.E. 180 Blank, R. 38, 46, 47 Blaug, M. 24, 281 Blejer, M.I. 288 Boisot, M.H. 139 Boland, L.A. 13, 183 Bollen, K.A. 209 Booth, A. 46 bounded rationality 252 Boylan,T.A. 45, 46, 58 Brady, M.E. 97 Briscoe, C. 46 Brown, A. 9, 233, 236, 242–4 Bryman, Alan 201 Buiter,W. 239 Bunch-Maps 124 Bundesbank, the 238 Burawoy, M. 187 Burchell, B. 40 Byrne, D. 155 Caldwell, B.J. 19 Calvo, G.A. 288 canonical correlation 125 capabilities 98 capitalism 244 Carr, D. 185 Carribelli, A.M. 93 Carrier, D. 183 Cartwright, N. 45, 185 case studies 181–2 Casey, B. 36 Castellan, N.J. 130 categorical data 143 categorical variables in regression 133 causality, sequential and simultaneous 76, 82

Index causal mechanisms 16–18, 21, 29–30, 51, 54–5, 58, 61, 93, 105, 112–13, 125, 130, 141, 145, 157, 172–85, 209, 221, 225–31, 261 ceteris paribus methodology 75–84 Charmaz, K. 180 Chick,V. 20, 73–4 Choi,Y.B. 98 Chouliaraki, L. 198 closed systems 16–23, 30–1, 34, 43, 56, 59, 75–8, 90–1, 112, 220–2, 300 closure: conditional 76–8, 83; intrinsic and extrinsic conditions of 112 Coates, J. 187 Coates, K. 46, 171 cognitive dissonance 23 Collier, Andrew 155–6, 161 Comim, F. 24, 171 common sense propositions 170–1 confluence analysis 124 Conlisk, J. 96 Conrad, C.F. 180 constant conjunction of events 29–31, 90–1, 112, 114, 120, 234; see also event regularity constructivism 13–15, 19–22, 198 consumption 164–5 contrastives 228–30 ‘convenient fictions’ 83 conventionalism 13–14 Convey, P. 27 Corbin, J. 180, 186, 187 Cottrell, A. 105, 146 counterfactual statements 82 Coutts, K. 255 critical realism 3–5, 12–13, 57–64, 92, 104–5, 153, 171–3, 185, 197, 242–4; and ceteris paribus modelling 80–3; chief features of 51–5; criticisms of 282, 288; and econometrics 183–4; and economic history 220–30; and empirical investigation 281–2; and the Euro currency 239–41; and grounded theory 180–2; implications for economics 21–4; and the labour market 28–36 passim, 43–4; methodology of 17–21, 228–9, 233–6; nature of 71–3, 279–81; ontology of 224–7, 235; philosophical foundations of 15–17; and postmodernism 224; schema for 297–300; and social structure 234 Crotty, J. 85, 146 Cyert, R. 98 Czech Republic 284–7

303

Danermark, B. 160 Darnell, A.C. 117 data collection 175–6, 180–1 data generating processes (DGPs) 154 Davidson, Paul 222, 249–53, 268 Davis, J.B. 160 deductivism 28–37 passim, 43–4, 56–7, 62–4, 91, 125, 221–2, 283, 293; criticisms of 112 demi-regularities (demi-regs) 20–1, 57, 61–4, 83, 91–2, 104, 114, 124–5, 129–30, 137, 145, 173, 184, 228–30, 239–42, 247–9, 270, 299–300; pervasiveness of 139–41 Denzau, A.T. 98 Dex, S. 39, 48 Diesing, P. 180, 186 Dilnot, A. 46 ‘disciplined noticing’ 201 discourse analysis 159 distancing in report-writing 158–9 domain of the actual 101 Dow, S.C. 6, 11, 17, 19, 20, 22, 23, 24, 77, 85, 86, 93, 111, 129, 145, 146, 160, 170, 171, 173, 185, 186, 249–51 Downward, P. 7–9, 11, 23, 80, 83, 84, 91, 101, 116, 122, 126, 141, 145, 146, 155, 171, 173, 184–6, 248, 253–7, 262, 289, 294–6 dummy variables 199, 208, 209, 210, 212 Dunn, S. 249 Dupre, J. 183, 185 Earl, P.E. 100 econometrics 4–5, 23–4, 55–6, 61–4, 89, 111–13, 125, 233, 240, 247, 257, 293, 299; critical realist critique of 119–22; essence of 114–19; and grounded theory 183–4; mainly-as-regression 131–8; as measurement 136–41, 145 economic history 220–31 Edusah, S. 9, 205, 208, 209, 210, 213, 295 Ehlers,T.B. 211 Eisenhardt, K.M. 182 Elliott, Jane 158 Ellis, B. 172, 180 emergence, process of 98 Emigh, R.J. 180 empirical measurement 119–20 empirical realism 4, 29–30, 74, 90, 225, 250, 281, 293, 297 empirical research 145, 281–2 empirical validity 13–15, 18, 54–8, 89

304

Index

employment: atypical types of 36–44; qualitative changes in 32–6; in relation to unemployment 27, 34–6, 44; security of 40 endogenous variables 199 Engel’s Law 226 entrepreneurship 97–8, 205, 207, 208, 211, 212, 214 entropy in economics 135–6 epistemic fallacy 90, 121, 153–4, 293 epistemological relativism 172–4 epistemology 4–5, 16–23 passim, 55–7, 64, 112–14, 129, 161 ergodicity 222, 251–3, 268 Euro currency 233, 237–44 European Central Bank 238 European Monetary Union (EMU) 233, 237–8, 244 Evans, J.L. 117 event regularity 55–7, 61–4, 72–4, 82–3, 121, 268, 280; see also constant conjunction of events Ewing, K. 36 exogenous variables 199 experimentation, scientific 16–17 explanatory power of economic theories 57–8, 92, 114, 230 extrinsic conditions of closure (ECC) 112, 121, 299 Fairclough, N. 198 Fay, B. 158 Feyerabend, P.K. 129–30 Figart, D.M. 212 Finch, J. 7, 8, 101, 126, 127, 180, 182, 187,188, 294, 295 Fleetwood, Steve 3, 6, 45, 46, 71, 74, 95, 97, 234 flow principle of output 248 formalism in economics 129, 234–5, 240, 283 formal modelling 71–84; axiomatic approach to 74, 77, 84; nature of 73–5 Foss, N.J. 252 Fothergill, G. 38 Frankfort-Nachimas, C. 283 Frenkel, J.C.A. 288 Friedman, M. 13, 14, 18, 55 Frisch, R. 116, 124, 126 functional relations 29–30, 234 Fusfeld, D.R. 180, 186 Garvy, G. 287, 290 GDP growth, modelling of 271–3

Geertz, C. 102 gender relations 201–14 ‘General to Specific’ approach in econometrics 116 ‘generic tendency’ (Bhaskar) 230 George, A.L. 182 Georgescu-Roegen, N. 135, 136, 143 Geroski, P.A. 255 Gerrard, B. 80, 115 Gerschenkron, A. 290 Ghana 201–7, 210, 216 Giddens, A. 96 Gilbert, N. 116, 160 Glaser, B.G. 101 Glickman, M. 9, 266, 269, 270, 271, 273, 276, 277, 278, 296 Glynn, A. 46 Golan, A. 132, 135–6 Gordon, S. 31 Grabel, I. 286, 290 Granger causality 116 Granger, C.W.J. 131 Granovetter, M. 97 Great Depression 249 Greene,W. 118 Greenwald, B. 289 Gregg, P. 40, 46 gross domestic product see GDP growth grounded theory 101–4, 170–85, 289; evaluation of 178–9; methodology of 180–4 Gruchy, A.G. 171, 180 Gujarati, D.N. 116, 133 Haavelmo,T. 115 Hahn, F. 223 Hall, R.F. 257 Hands, D. 281 Hara,T.M. 9, 295 Harcourt, G.C. 84 Harding, Sandra 161 Härdle,W. 130 Harper, D. 106 Harré, R. 3, 103 Harrison, R.S. 171, 180, 186, 187 Hartley, J. 102 Hay, C. 165 Hayek, F.A. 97 Heckman procedure 210 Hendry, David 114–25 passim, 132, 154 hermeneutics 59, 137, 157–8 Hetherington, P. 39 Hicks, John 251, 267 Hillard, J. 11

Index historical economic theories 184–5 Hitch, C.J. 257 Hodgson, G. 97, 99, 105, 140, 226, 252 Holland, J.H. 138 Hughes, J. 27, 134 human capital theory 154–5 Hume, David (and Humean thinking) 4, 6, 17, 21, 29, 45, 55–7, 61–2, 90, 103, 112–13, 119, 221 Hungary 284–7 Hunt, S.D. 180, 208 Hutchinson, F.G. 162 Hutchinson,T. 281 hypothesis-testing 198, 236 identification in econometrics 115 ignoratio elenchi 31 Incomes Data Services (IDS) 42 individualism, methodological 29, 97, 154, 220–1 induction 94, 283–4 inference 93–4, 103–5, 119–21 Ingham, G. 102, 173 instrumentalism 114, 119–21, 125 instrumental variables 199–200 interest rates 164 interpretivism 137 interval scales 142 intrinsic conditions of closure (ICC) 112, 121 Israel, G. 183 Jehle, G. 46 Johnson, S. 46 Joseph, J. 173, 185 Kabeer, Naila 205, 207, 208 Kaldor, N. 76–7, 83, 140 Kalecki, M. 9, 233, 242, 243, 254 Kalpagam, U. 162 Kanth, Ranjani 156 Kant, Immanuel 268 Kasynathan 208 Kay, N.M. 100 Keat, R. 3, 56 Kelly, G.A. 96–7 Kendall coefficient of concordance 144 Keuzenkamp, H.A. 132 Keynes, J.M. 7, 9, 20, 23, 61, 76, 93–4, 104, 115, 119–25 passim, 233, 242–3, 248, 251 Klamer, A. 157 Klein, D. 36 Knight, F.H. 94

305

know-how 95–7 knowledge claims 91–4, 100–4, 129–31, 136–8, 141–2, 160, 198 knowledge, types of 95 Konecki, K. 180, 187 Kregel, J. 76 Krugman, P. 129 Kuhn,T.S. 15, 105 Kvale, S. 160 Kwan, K. 141 Labour Research 27 Lacey, H. 162 Lakatos, I. 14, 129–30 Lane, D. 98, 141 Lawson, C. 73, 85 Lawson,Tony 3–4, 8, 15, 17, 20, 21–3, 25, 28, 45, 55, 57, 60, 63, 71, 73, 81, 83, 90–4, 111–14, 119–23 passim, 131, 137–42 passim, 153, 198, 200, 221, 228, 233–4, 239–40, 250–1, 269, 271, 277, 279–84 Layard, R. 45 Layder, D. 106, 161, 162 Leamer, E.E. 115, 117, 132 Lee, F.S. 8, 9, 101, 248, 253 Leibenstein, H. 143 Lewis, P. 253 Lieberman, M. 46 Lloyd, C. 173 Loasby, B.J. 20, 96, 98, 146, 252 ‘locking up without ignoring’ 76–7, 82 logical positivism 18–22, 220–1, 225–6 logit techniques 118, 131–4 Lovering, J. 173 Lund, P.J. 255, 256 McCallum, B.T. 256 McCloskey, D.N. 144 McCullagh, C.B. 180, 185 McCulloch, A. 39, 48 McFadden, D. 134 McGovern, Siobhan 154 McGregor, A. 38, 47 McKenna, E.J. 185 McMaster, R. 8, 144 Madden, E.H. 3, 103 Main, K. 211 mainstream economic analysis, critiques of 3–4, 18–24, 44, 51, 55–6, 60–1, 112, 253, 280–3 Maital, S. 143 Mäki, U. 77, 83–4 Manicas, P. 131, 133

306

Index

March, J.G. 98, 102 mark-up pricing 256–62 Marx, Karl 9, 233, 242–3 mathematical modelling 182–3, 222–4, 233–4; see also formal modelling maximum entropy econometrics 131, 135–6 May,T. 160, 162, 283 Mayhew, K. 46 Mayumi, K. 135 Meadows, P. 46 Means, Gardiner 247–9, 262 Mearman, A. 7–9, 83–4, 247, 253 medical research 62 Megill, A. 180 Meikle, S. 45 Menard, S. 209, 210 Menger, Carl 18 meta-narratives 77 metaphor, use of 21, 80, 156–8 Metcalf, H. 46 migration 227, 229 Miller, M.H. 266, 281 Miller, P.W. 161 Mirowski, P. 77, 86 Mitchell,W.F. 184 Modigliani, F. 281 Modigliani–Miller (MM) proposition 266, 269, 281 money, concept of 235 Morgan, M.S. 183 Morrison, G.C. 143 Mukherjee, K. 143 multi-collinearity 116, 124, 209, 259 Nachimas, D. 283 Nagel, E. 13 naïve empiricism 166 Nardone,T. 36 Natti, J. 47 Negative Binomial models 118 Neild, R. 255–7 Nelson, R.R. 98 neoclassical economics 55, 90, 160, 184, 250, 254 new classical economics 14, 18 non-parametric techniques 130, 137–45 Norci´c, O. 286–7 Norman, A.P. 185 North, D.C. 98 objectivity 198 O’Gorman, P.O. 45, 46, 58

Oleson,T. 180 Olsen,W. 8, 9, 145, 154, 202, 205, 208, 209, 212, 213, 214, 289 ontological inertia 226 ontological security 96–9 ontology 3–5, 16–23, 28–30, 55–8, 64, 72–3, 89–93, 103–4, 112, 120, 121, 125, 129, 132–3, 161; of critical realism 224–7, 230, 235 opaque understandings 139 open systems 16–24, 30, 54, 57, 59, 72, 89–99, 104, 197, 228, 249–50, 293, 297, 300; characteristics of 200 open-systems, ceteris-paribus (OSCP) approach to modelling 74–84 ordinalist fallacy 136, 143 ordinary least squares (OLS) techniques 115 ‘ordinary logic’ (Keynes) 20 Orum, A.M. 182 Osborne, K. 47 Outhwaite,W. 3, 64, 160, 162, 198 Pagan, A.R. 116, 130 paradigms 15, 22, 105, 159 parametric functions 113, 119 Parfitt,T. 160 parsimony, principle of 227 Parsons, S. 45, 245 part-time work 37–43, 154–5 path analysis 208–13, 215 path dependence 76, 82 Peirce, C.S. 96, 130 Pencavel, J. 30 Penn, R. 47 pension provision 39 Perry, K. 46 personal constructs 96–8 Pesaran, H. 23 Petrick, K. 9, 10, 285 pilot studies 201 Pinkstone, B. 9, 225, 226, 231, 295 pluralism, methodological 20–3, 249–50 Podolski,T.M. 286 Poisson models 118 political economy 5 Polivka, A. 36 Pollert, A. 36 Pope, D. 231 Popper, K.R. 14, 15 Porpora, D. 245 positivism 4, 52, 55–6, 64, 90, 112; see also logical positivism

Index Post Keynesian economics 111, 170–4, 178–9, 184–5, 247–54, 259–62, 266 postmodernism 224 Potts, J. 98 Pratt, A.C. 105, 141, 174, 186 Pratten, S. 84, 185, 234 Prebisch–Singer thesis 225–6 predictions from economic theory 13–14, 18, 59–60, 63–4, 92, 100, 125, 277 pricing models 163–4, 253–62 prior analysis 117–18 probability statements 93–4, 97 probit techniques 118, 131–4 proxy variables 199 qualitative and quantitative research 122 queuing problems 288 radical policy proposals 62–3 Ramsay, J. 7, 294 random events 267–8 random utility models 134 rational belief 123 Ray, L. 160 Ray, S.C. 143 ‘real’ causes of events 113, 297 realism in economics 156–7, 197–202, 206; and alternative conceptions 165–6; empirical assumptions of 162; event-truth and process-truth orientations 250–1; see also critical realism; empirical realism regression analysis 115, 123–5; logistic 209;‘traditional’ 131–3 regularity error 225 relativism 64, 172–4 retroduction 4–5, 19, 54, 62, 72–3, 80, 92–4, 114, 122, 124–6, 129–30, 141, 144–5, 229, 235–7, 242, 247–8, 257–62 passim, 269, 271, 282–4, 288, 293, 299–300 Rice, P. 45 Richard, J. 154 Ricoeur, Paul 157–9 Roosenthal, H. 36 Rotheim, R.J. 280, 284 Rozansky, J. 180 Rubery, J. 154 Runde, J. 58, 92–3, 253 rural small-scale industries sector (RSSI) 197, 202–5, 209–15 Rushdy, F. 256 Ryan, M. 134

307

Sagari, S.B. 288 sampling: random and non-random 202–3, 215; statistical 101; theoretical 101–2, 176–7, 181 Samuelson, P. 13–14, 18 San Miguel, F. 134, 146 Sarre, P. 186, 187 Saunders, M. 283, 289 Sawyer, M. 9, 295 Sayer, A. 119, 122, 159–60 Schoenberger, E. 285 scientific knowledge 29–31, 105 Sedgley, N.H. 132 Sengenberger,W. 35–6 Sent, E.-M. 24 Setterfield, M. 7, 76 Shackle, George 251–2, 267 Shannon, C.E. 135 Shapiro, C. 46 Siegel, S. 130 significance, statistical and economic 144 Simon, H.A. 96, 98, 100, 103 Sköldberg, M. 101 Slater, G. 244 Smith, Adam 23 Smith, C. 281 Smith, G.W. 255 Smith, M. 161, 182 Smith, R. 23 social change 200 social structures and social systems 73, 77, 81, 89, 95, 98, 103, 133, 139, 173, 224, 234–5 Spencer, D. 244 Sproull, A. 38, 47 Stake, R.E. 182 Stanley,T.D. 136 Stigler, G. 55 Stiglitz, J. 46, 281 stochastic models 221, 266–8 Strauss, A.L. 101 ‘stylised facts’ 76–7, 83, 140, 166 subjectivity 252 Sylos-Labini, P. 255 symmetry thesis 112, 115 ‘systematic abstraction’ 244 tacit knowledge 95–8, 139 Taylor, C. 59 theory, economic: building of 178–9, 282; development of 129–31, 137, 179–80, 240; purpose of 14; in relation to economic reality 12–15

308

Index

‘thick’ and ‘thin’ narratives 102–3 Thomas, B. 231 Tijdens, K.G. 155 Tilley, C. 37, 47 time-series data 162, 166; augmentation of 153–60 Tinbergen, J. 23, 123 Tobit model 118 Tool, M. 101 Tosh, J. 180 transcendental realism 13, 16–18, 51–3 transfactual statements 82, 179, 183 transformational model of social activity (TMSA) 72, 82 transitivity and intransitivity 157, 165, 198 triangulation 123–5, 159–62, 166, 205, 213, 214, 216, 247–8, 258, 262, 299–300 Tribe, M.A. 202, 213 Trigg, R. 105 Tsang, E.W.K. 141, 180 Turner, B.A. 180 Ullah, A. 130 underemployment 35–6 unemployment 223; in relation to employment 27, 34–6, 44 Urry, J. 3, 56 variables, characteristics of 32 Vaughan, D. 182 vector autoregressive (VAR) systems 116, 117, 120, 125

Vickers, D. 75 Wadsworth, J. 40, 46 wage rates 154–5, 161, 165, 223, 227, 243 Walby, S. 154 Walters, B. 6, 7, 55, 64, 294 Warde, A. 164 Warme, B. 43 Watson, B. 38 weather forecasts 59–60 Weintraub, E.R. 85, 183 Wellman, B. 173 Wells, J. 27 Wible, J.R. 132 Wieviorka, M. 106, 182 Wilber, C.K. 171, 180, 186, 187 Williams, M. 162, 244, 290 Williamson, O.E. 290 Winter, S.G. 98 Wirth, B. 47 Wisman, J.D. 180 Wood, L. 47 world-realism 250 Yeung, H. 174, 184, 186 Yin, R.K. 182, 187 Young, D. 6, 7, 55, 294 Zannoni, D.C. 185 Ziman, J. 129 Znaniecki, F. 102 Zwass, A. 282