Elgar Encyclopedia of Labour Studies 1800377533, 9781800377530

Table of contents :
Front Matter
Copyright
Contents
Contributors
Introduction to the Elgar Encyclopedia of Labour Studies
Active labour market policies
African labor markets
Agricultural labor markets
Apprenticeships
Beveridge curve, matching functions
Dynamic employment adjustment of firms
Early retirement
Employee ownership
Employer and employee learning
Employer search
Employment protection legislation impacts
Executive compensation
Firms and wages
Footballers’ labour market
Global value chains and employment relations
HRM practices and productivity
Intergenerational income mobility
International migration
Japanese labor market
Job design
Job insecurity
Job satisfaction
Jobs, tasks, authority
Labor income share
Labor market discrimination: method and measurement
Labor market discrimination: sexual orientation
Labour market discrimination: ethnicity and race
Labour market discrimination: gender
Labour market integration of immigrants
Labour supply and taxes
Monopsonistic labour markets
Non-financial motivation in the workplace
Non-wage labour costs
Occupational Licensing
Outsourcing, consequences for employees
Payroll taxes: incidence and employment effects
Performance evaluations
Performance feedback: cognitive and motivational effects
Performance pay: consequences on workers’ health
Performance related pay and performance
Promotion tournaments
Public sector labor markets
Recruitment: internal or external?
Regional labour markets
Returns to education
Shadow economy labour markets
Skill-biased technological change
Skills acquisition: workplace learning and workers’ productivity
Strikes and conflict mediation
Team productivity
Temporary jobs
Unemployment: duration, incidence
Unpaid work
Wage bargaining institutions
Worker representation
Working hours
Workplace sickness absence
Works Councils
Index

Citation preview

Elgar Encyclopedia of Labour Studies

ELGAR ENCYCLOPEDIAS IN ECONOMICS AND FINANCE Elgar Encyclopedias in Economics and Finance serve as the definitive reference works in the field. The Encyclopedias present a comprehensive guide to a wide variety of subject areas within economics and finance, and form an essential resource for academics, practitioners, and students alike. Each Encyclopedia is edited by one or more leading scholars, internationally recognized as preeminent names within the field. They each include an overarching collection of entries authored by key scholars within the field, which collectively aim to provide a concise and accessible coverage of the essential areas. Equally useful as reference tools or high-level introductions to specific topics, issues, methods and debates, these Encyclopedias represent an invaluable contribution to the field. For a full list of Edward Elgar published titles, including the titles in this series, visit our website at www​.e​-elgar​.com​.

Elgar Encyclopedia of Labour Studies Edited by

Tor Eriksson Professor Emeritus, Department of Economics and Business Economics, Aarhus University, Denmark

ELGAR ENCYCLOPEDIAS IN ECONOMICS AND FINANCE

Cheltenham, UK • Northampton, MA, USA

© Tor Eriksson 2023

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited The Lypiatts 15 Lansdown Road Cheltenham Glos GL50 2JA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA A catalogue record for this book is available from the British Library Library of Congress Control Number: 2023941681 This book is available electronically in the Economics subject collection http://dx.doi.org/10.4337/9781800377547

EE VS P

ISBN 978 1 80037 753 0 (cased) ISBN 978 1 80037 754 7 (eBook)

Contents

List of contributorsvii Introduction to the Elgar Encyclopedia of Labour Studies ix Tor Eriksson Active labour market policies Anders Forslund

1

African labor markets Niels-Hugo Blunch

5

Agricultural labor markets Diane Charlton

9

Firms and wages István Boza

53

Footballers’ labour market Robert Simmons

58

Global value chains and employment relations Sarosh C. Kuruvilla

Apprenticeships14 Uschi Backes-Gellner and Patrick Lehnert

62

HRM practices and productivity Kathryn L. Shaw

67

Intergenerational income mobility Jo Blanden

72

77 International migration Mariola Pytlikova and Davit Adunts Japanese labor market Ryo Kambayashi

83 86

24

Job design Michael J. Gibbs

28

91

Early retirement René Böheim

Job insecurity Francis Green

Employee ownership Takao Kato

31

Job satisfaction John S. Heywood

95

Employer and employee learning Jaime Ortega

36

Jobs, tasks, authority Alexandra Spitz-Oener

99

Employer search Jan C. van Ours

39

Labor income share Saumik Paul

43

Labor market discrimination: method and measurement 106 Ali Ahmed and Mats Hammarstedt

Beveridge curve, matching functions Juuso Vanhala Dynamic employment adjustment of firms Gerard A. Pfann

Employment protection legislation impacts Paulino Maria Freitas Teixeira Executive compensation Martin J. Conyon

19

Labor market discrimination: sexual orientation Nick Drydakis

47 v

103

110

vi  Elgar encyclopedia of labour studies

Labour market discrimination: ethnicity and race 115 Mats Hammarstedt and Ali Ahmed Labour market discrimination: gender Dominique Meurs

118

Labour market integration of immigrants122 Pieter Bevelander Labour supply and taxes Monica Costa Dias

125

Monopsonistic labour markets Boris Hirsch and Elke J. Jahn

130

Non-financial motivation in the workplace134 Gary Charness, Michael Cooper and J. Lucas Reddinger

Public sector labor markets Claudio Lucifora

181

Recruitment: internal or external? Jed DeVaro

186

Regional labour markets Uwe Blien

190

Returns to education 195 Franz Buscha and Matthew Dickson Shadow economy labour markets Dominik H. Enste

200

Skill-biased technological change Pekka Ilmakunnas

205

Skills acquisition: workplace learning and workers’ productivity Andries de Grip

208

Non-wage labour costs Robert A. Hart

141

Strikes and conflict mediation John Kennan

212

Occupational licensing Morris Kleiner

145

Team productivity Hideo Owan

215

Temporary jobs Lia Pacelli

220

Unemployment: duration, incidence Tor Eriksson

224

Unpaid work Leslie S. Stratton

228

Wage bargaining institutions Erling Barth

233

Worker representation Alex Bryson and John Forth

237

Working hours Peter Dolton

242

Workplace sickness absence Wolter Hassink

247

Works Councils Uwe Jirjahn

251

Outsourcing, consequences for employees149 Holger Görg Payroll taxes: incidence and employment effects Roope Uusitalo Performance evaluations Anders Frederiksen

154 158

Performance feedback: cognitive and motivational effects Marie Claire Villeval

161

Performance pay: consequences on workers’ health Ioannis Theodossiou

165

Performance related pay and performance172 Tor Eriksson Promotion tournaments Michael Bognanno

177

Index257

Contributors

Davit Adunts, PhD student, CERGE-EI, Charles University, Prague, Czech Republic

Michael Cooper, Post-doc scholar, University of California, San Diego, USA

Ali Ahmed, Professor, Linköping University, Sweden

Monica Costa Dias, Professor, University of Bristol and Deputy Research Director at Institute of Fiscal Studies, London, UK

Uschi Backes-Gellner, Professor, University of Zurich, Switzerland

Jed DeVaro, The Wang Family Professor, Department of Management, California State University East Bay, Hayward, USA

Erling Barth, Research Professor, Institute for Social Research, Oslo, Norway

Matthew Dickson, Reader in Public Policy, University of Bath, UK

Pieter Bevelander, Professor, Malmö University, Sweden

Peter Dolton, Professor, University of Sussex Business School, Brighton and National Institute of Economic and Social Research, UK

Jo Blanden, Reader in Economics, University of Surrey, UK Uwe Blien, Professor, Otto Friedrich University in Bamberg and Senior Researcher at IAB, Nuremburg, Germany

Nick Drydakis, Professor, Anglia Ruskin University, Cambridge, UK

Niels-Hugo Blunch, Professor of Economics, University of Washington & Lee, Lexington, USA

Dominik H. Enste, Professor, Technical University of Cologne, Germany Tor Eriksson, Professor emeritus, Aarhus University, Denmark

Michael Bognanno, Professor of Economics, Temple University, Philadelphia, USA

Anders Forslund, Professor, Uppsala University and IFAU, Uppsala, Sweden

René Böheim, Associate Professor, Johannes Kepler University, Linz, Austria

John Forth, Senior Lecturer, Bayes Business School, London, UK

István Boza, Junior Research Fellow, Central European University, Budapest, Hungary

Anders Frederiksen, Professor, Aarhus University, Denmark

Alex Bryson, Professor, Institute of Education, University College London, UK

Michael J. Gibbs, Clinical Professor of Economics, Booth School of Economics, University of Chicago, USA

Franz Buscha, Professor, Westminster Business School, London, UK

Holger Görg, Professor, University of Kiel and Kiel Institute of the World Economy, Germany

Diane Charlton, Assistant Professor of Agricultural Economics, Montana State University, Bozeman, USA

Francis Green, Professor, Institute of Education, University College London, UK

Gary Charness, Professor of Economics, University of California, Santa Barbara, USA

Andries de Grip, Professor, ROA and School of Business and Economics, Maastricht University, Netherlands

Martin J. Conyon, Trustee Professor, Bentley University, Waltham, and Senior Fellow, The Wharton School of the University of Pennsylvania, USA

Mats Hammarstedt, Professor, Linnæus University, Växjö, Sweden vii

viii  Elgar encyclopedia of labour studies

Robert A. Hart, Professor, University of Stirling, UK

Jaime Ortega, Associate Professor, University Carlos III, Madrid, Spain

Wolter Hassink, Professor, Utrecht University, Netherlands

Jan C. van Ours, Professor, Erasmus University, Rotterdam, Netherlands

John S. Heywood, Distinguished Professor, MHRLR Director at University of Wisconsin-Milwaukee, USA

Hideo Owan, Professor, Washeda University, Tokyo, Japan

Boris Hirsch, Professor, University of Leuphana, Lüneburg, Germany Pekka Ilmakunnas, Professor emeritus, Aalto University, Helsinki, Finland Elke J. Jahn, Professor, University of Bayreuth and Distinguished Scholar at IAB Nuremburg, Germany Uwe Jirjahn, Professor, University of Trier, Germany Ryo Kambayashi, Professor, Institute of Economic Research, Hitotsubashi University, Tokyo, Japan Takao Kato, W.S. Schupf Professor in Far Eastern Studies, Colgate University, Hamilton, USA John Kennan, Professor, Juli Plant Grainger Distinguished Chair, University of Wisconsin-Madison, USA

Lia Pacelli, Associate Professor, University of Turin, Italy Saumik Paul, Lecturer, University of Newcastle, UK Gerard A. Pfann, Professor, School of Business & Economics, University of Maastricht, Netherlands Mariola Pytlikova, Assistant Professor, CERGE-EI, Prague, Czech Republic J. Lucas Reddinger, Researcher, University of Wisconsin – La Crosse, USA Kathryn L. Shaw, Ernest C. Arbuckle Professor, Graduate School of Business, Stanford University, USA Robert Simmons, Professor, Lancaster University, UK Alexandra Spitz-Oener, Professor, Humboldt-University, Berlin, Germany

Morris Kleiner, Professor and AFL-CIO Chair in Labor Policy, Humphrey School of Public Affairs, University of Minnesota, USA

Leslie S. Stratton, Professor, Virginia Commonwealth University, Richmond, USA

Sarosh C. Kuruvilla, Andrew J. Nathanson Family Professor of Industrial Relations, Asian Studies, and Public Affairs, Cornell University, Ithaca, USA

Ioannis Theodissou, Professor, University of Aberdeen, UK

Patrick Lehnert, Post-doc researcher, University of Zurich, Switzerland Claudio Lucifora, Professor, Catholic University in Milan, Italy Dominique Meurs, Professor, University Paris Quest Nanterre and Researcher at Institut National d’Etudes Démographiques, France

Paulino Maria Freitas Teixeira, Professor, University of Coimbra, Portugal

Roope Uusitalo, Professor, University of Helsinki, Finland Juuso Vanhala, Senior Advisor, Bank of Finland, Helsinki, Finland Marie Claire Villeval, Research Professor at CNRS (GATE), University of Lyon, France

Introduction to the Elgar Encyclopedia of Labour Studies

employed in economics journals and has also helped me in the choice of entries to include. As labour transactions predominantly take place in markets, and market features have also increasingly played an important role within firms and organizations, the vast majority of the entries are written by labour economics scholars. Modern labour economics is a field that has expanded tremendously during the last 40–50 years, largely thanks to the increased use of micro-level data (individual and/or workplace-level observations) in applied work. Numerous developments of techniques to analyse them (resulting in several Nobel Prize awards) have had large impacts not only in the field of labour studies itself but also on other areas of economics, such as health and industrial economics. Many of the techniques and research designs applied in labour studies to examine causal effects and policy impacts have spread to other social sciences, where they are now widely employed in empirical work. The focus on studies of labour markets means that other aspects of work examined by researchers in, for example, human resource management and occupational psychology and health are not dealt with. However, several entries discuss important insights from the more recent behavioural economics analyses of employment relationships. Labour economics is a broad field, and it would not be possible to cover all phenomena and questions addressed by labour economists within the limits of a single book. Thus, some restrictions of topics and areas had to be made. These are, of course, to some extent arbitrary. Topics that many labour economists have worked on but are not dealt with in the entries of this Encyclopedia include, for example, economics of education issues, and economic analysis of (the rather special occupation of) crime. Migration studies has also recently been (and very likely will continue to be) a lively area of study that would easily comprise an encyclopedia itself. There is a large number of different labour market policies which could have been discussed in more detail, and in many more entries than in the current book, where they are lumped together in broader categories. In terms of topics, labour studies is a diverse field. However, it should be noticed that the output of labour studies can also be characterized as narrow – one could even say ‘biased’ – owing to the fact that the main body

Labour studies, the analysis of labour markets, employment relationships, determination of pay and other rewards, and workplace organizations has in recent decades been a steadily growing field of research. At the same time, the weight of labour issues in the public policy discussion has also been growing. This Encyclopedia differs from many other encyclopedias by having considerably fewer but significantly longer entries. This is partly due to the fact that this Encyclopedia is about labour studies: that is, the aim is not only to provide brief definitions of concepts, their background and history, but also to discuss how the concepts are put into use in analyses of the functioning of labour markets as well as the internal labour markets of firms. The key purpose of the entries is to give readers concise, authoritative and comprehensive accounts of what social scientists know and do not know (yet) about labour market-related phenomena. The readership of this Encyclopedia includes, in addition to undergraduate students, non-expert academics, practitioners, policy specialists and others looking for a succinct expert summary of key concepts in labour studies. The presentation is therefore mainly non-technical and focuses on the most relevant accumulated knowledge on the subject, theoretical as well as empirical. For students, the Encyclopedia is a companion to (not a substitute for) the standard textbooks in labour studies. It is helpful to have some familiarity with basic economic concepts such as labour demand and supply, and a basic understanding of the role of competition in different market forms. Each entry has at the end a list of key references which readers may consult if they want to go deeper into some of the topics discussed. For highly specialized readers (researchers) there are many handbooks on labour studies topics available. The entries are organized alphabetically, but readers can find other entries that are thematically related with the help of the so-called JEL code that is used in organizing the material presented in the Journal of Economic Literature. The code is widely ix

x  Elgar encyclopedia of labour studies

of research deals with problems considered as important in developed countries, and hence uses data from them. This is reflected in this Encyclopedia, too. Consequently, there is rather little discussion of studies from or about the world’s two most populous countries, China and India. China did not even have markets for labour before the 1990s, and India has had industrial and labour relations that differ markedly from those of Western economies. Clearly, more research-based knowledge of the labour markets in China,

India and other emerging economies would be valuable. I would like to thank all the contributors for their valuable contributions, their entries. Most of the contributors were actually the first ones who I invited to contribute on a topic, and responded positively (often within a few days). I have to admit that I was pleasantly surprised. While it took somewhat longer than I expected to gather all the entries, it has overall been a pleasure to work with you. Tor Eriksson

A

Active labour market policies

Direct policy impacts

Direct policy impacts are differences between outcomes when the policy measure is present and when it is not. Impacts may appear before, during or after the exposure to a policy measure. The ex ante effects work through expectations. Depending on whether a policy measure is expected to be beneficial or not, job search behaviour can be affected before exposure to the measure. Hence, a measure expected to be beneficial will decrease the individual’s job search and joblessness-to-job transition rate already before the measure is implemented. The opposite will be true of a policy expected to be bad for the person to be subject to the measure. For a substantial ex ante effect to arise, it must be possible to predict with some accuracy whether and when a treatment will be received by the individual. During exposure to a measure, job search will typically decrease unless the measure is job search assistance. Hence, policy measures will typically be associated with a slowdown in the transition from joblessness to work, a ‘locking-in effect’. The degree of the locking-in effect will depend both on the extent to which search efforts are reduced and the marginal effect of job search on the transition from joblessness to work. Finally, the intention of the policy measures is to create a ‘treatment effect’, typically arising after the policy exposure (treatment). A treatment may have an impact on both the transition from joblessness to a job, and job quality. For example, a training programme may both enhance the employability of the participant (faster transition to a job) and the productivity in a job (higher wage and possibly employment stability).

Introduction

Modern welfare states typically provide several safety nets for persons in need of help for a variety of reasons. One important class of safety nets, often referred to as labour market policies, derives from needs related to joblessness. The help provided is of two basic kinds: financial support, and measures to facilitate the transition from joblessness to jobs. The latter are often called active labour market policies (ALMPs). ALMPs can be classified by types of measures, and one such classification scheme distinguishes job search assistance and monitoring, training programmes, subsidised employment, and work practice schemes. The frequent use of programmes targeted at youth also warrants a brief discussion of the evidence on such measures. This entry covers these basic types of ALMPs. The presentation covers both some elementary theoretical considerations and empirical evidence on the impacts of the measures. Some gaps in current knowledge are also identified.

Basic theory

Active labour market policies work through a few possible channels. The policy measures may affect job search behaviour for a given set of skills of the worker. The policy measures may also change the skills of the worker. Finally, policy measures may help in acquiring useful labour market contacts. Needless to say, different policy measures typically involve different combinations of elements from more than one channel. Policy measures may have both direct impacts on those exposed, and indirect (general equilibrium) impacts on others.

General equilibrium impacts

Most general equilibrium effects are likely to be undesirable side effects arising because 1

2  Elgar encyclopedia of labour studies

policies often imply that a faster job transition for a treated individual will come at the expense of some non-treated. Examples of this may be that subsidising jobs for some of the unemployed may make the job transition rate lower for other job seekers or that improved job search behaviour of treated job-seekers may imply lower job-finding rates for other competing job-seekers. There are, however, also possible desired general equilibrium policy impacts. For example, policies improving job search efficiency or the employability of the unemployed may give rise to lower vacancy costs via shorter vacancy durations and more jobs created in equilibrium.

Mechanisms

Active labour market policy measures potentially affect outcomes in several ways. Schematically, we can classify the mechanisms in terms of whether they mainly work through labour supply, labour demand or matching. It goes without saying that the dividing lines may be somewhat arbitrary and that policies can utilise more than one type of mechanism. Policy measures working through labour supply mechanisms may involve skill acquisition, rendering the unemployed more employable. A generic form of this would be vocational training programmes. Supply mechanisms may also be measures stimulating job search (bringing the skills to the labour market). This may entail ‘passive’ monetary incentives, but may also include ‘active’ measures such as counselling and monitoring job search. Furthermore, some policy programmes may affect job search behaviour by providing threats or promises as already noted in the discussion of ex ante effects of policies. Labour demand measures typically target labour costs, on the assumption that employers are more likely to hire unemployed workers if wage costs are lower. Hence, a typical policy affecting labour demand is subsidies to reduce wage costs. Labour demand measures may also involve policies to counteract discrimination. This may entail legislation, but may also be information provision to counteract statistical discrimination. An example of this could be validation of skills. Anders Forslund

Policies that aim at improved matching can generally be described as transfers of information. This information can entail both sides of the market, of skills supplied or demanded. The policies can also be described as a provision of networks. A caseworker can acquire knowledge of the skills of the job-seekers as well as the skill requirements of vacant jobs. The caseworker can also, through repeated contacts with employers, act as a substitute for missing networks of job-seekers. As a large fraction of all job vacancies are filled using informal channels, this is potentially an important mechanism.

Empirical evidence

There is a large literature providing evidence on the causal effects of active labour market programmes. Most of the evidence pertains to the direct effects for the programme participants, but there also is a smaller body of evidence on some general equilibrium effects. Comprehensive surveys of the evidence on the effects of active labour market policy measures can be found in Heckman et al. (1999), Martin and Grubb (2001), Calmfors et al. (2004) and Card et al. (2010, 2017). I first present the evidence on the direct effects. Most of the results on general equilibrium effects are discussed in connection with the presentation of the direct effects. The presentation of direct effects is by type of programme, beginning with programme types that generally give positive impacts according to the evidence. Subsidised employment, where employers are subsidised to hire unemployed workers for ordinary job positions, are generally found to be a fast track to unsubsidised jobs. The evidence also suggests that subsidised jobs with more attributes shared with unsubsidised jobs (types of tasks, workplace, management) are more likely to have a positive impact. The empirical evidence also suggests that the positive effect is long-lasting. Subsidised temporary jobs in the public sector, on the other hand, do not have positive estimated impacts. This pattern suggests that an important mechanism of subsidised employment is the creation of useful networks. Such networks are not necessarily created by performing tasks mainly as a treatment for the programme participants. Of course, subsidised jobs can also be useful because they enable the unemployed worker

Active labour market policies  3

to show the employer that they have sufficiently high productivity to cover their wage costs, or because they enable unemployed workers to acquire productivity-enhancing skills. In these respects, temporary jobs may not be appropriate. Subsidised regular jobs, however (that is, jobs which have not been created thanks to a labour market policy programme), are also consistently found to be associated with significant crowding-out effects. Hence, this kind of programme can be described as a re-sorting of a job queue of a roughly constant length. For this to be worthwhile, it is important that subsidies are thoroughly targeted to the unemployed at the end of the job queue. In addition, if subsidised jobs go to a group with good employment prospects, most of them would have found a job anyway, and consequently the programme impact will be small. The empirical evidence on job search assistance and monitoring suggests that both the combination of assistance and monitoring, and assistance alone, speeds up the transition from unemployment to employment, whereas the evidence on monitoring is less conclusive. The available evidence also suggests that job search assistance (typically combined with some monitoring) is an effective tool for both the short-term and the long-term unemployed. There is some evidence for a possible mechanism whereby caseworkers, given a sufficiently low case load, can act as a substitute network for the unemployed and provide useful information on which jobs to target in their job search. However, just as is the case for subsidised employment, increased job search assistance is found to be associated with significant displacement effects. Hence, there is a case for targeting job search assistance at more disadvantaged job-seekers. A final point to note is that evidence suggests that the positive impact of job search assistance is larger in the short run than in the long run. The evidence on impacts of vocational training is more mixed, although most evidence suggests positive long-run impacts. With shorter follow-up horizons, however, the estimated impacts are smaller, or sometimes non-significant or even negative. The evidence on vocational training also suggests that the impact is larger for more disadvantaged groups of the unemployed

(low educational level, disabled, immigrants). Unlike subsidised jobs and job search assistance, training programmes are not plagued by crowding-out effects. Nevertheless, the evidence on differential impacts across different groups of participants suggests similar targeting as for the other two programme types to maximize the positive impacts. The rather scarce available evidence on work practice schemes indicates small positive impacts for the participants. Contrary to the findings on other reviewed programmes, work practice programmes may give rise to bigger impacts for unemployed persons with better labour market prospects. Taken at face value, this suggests that an important mechanism might be that the programme enables skilled unemployed persons to demonstrate their skills. Youth programmes often entail compulsory entry early in a period of unemployment. This has been the case in Denmark and Sweden, for instance. Youth programmes on average give positive estimated impacts. This is not true for high-income countries, however, where results vary substantially; see Kluve et al. (2017) for a systematic review of the evidence on labour market policies targeted at youth. Findings from Sweden are interesting. A number of studies have found positive ex ante effects in combination with locking-in effects and zero treatment effects leading to a zero net effect, taking all three types of impacts into account. A natural interpretation of this finding is that for most young unemployed persons in high-income countries, measures should focus on creating incentives for job search rather than on measures that provide skills: young persons who have completed upper secondary school typically do not lack skills that make them employable.

Gaps in current knowledge

Theoretical considerations typically predict that the impact of active labour market policies is greater in times of recession than in boom times. This prediction follows if programmes imply less active job search, and if the marginal impact of job search on job findings is cyclical. However, the empirical evidence on this important margin for policy is very scarce. Another area in need of more research concerns the finer design details of proAnders Forslund

4  Elgar encyclopedia of labour studies

grammes. Should, for example, subsidised employment be generous with a short duration, or longer and less generous? A related issue concerns the pros and cons of early interventions. Common sense would suggest that early interventions are beneficial, but this conclusion only follows if there is sufficient knowledge of how measures should be targeted. However, there is also a need for more knowledge on what works for who: that is, which policy measures work best for different target groups. An important recent trend in many countries, finally, is a privatisation of the provision of active labour market policy measures. Until recently, the number of studies of the consequences of privatisation was small. The available evidence has recently grown somewhat, but as yet the evidence gives no clear conclusion about the relative merits of the private and public provision of labour market policies. Crépon (2019) reviews the evidence on private provision of labour market policy measures. Anders Forslund

Anders Forslund

References

Calmfors L., A. Forslund and M. Hemström (2004), ‘The effects of active labor-market policies in Sweden: what is the evidence?’, in J. Agell, M. Keen and J. Weichenrieder (eds), Labor Market Institutions and Public Regulation, Cambridge, MA: MIT Press. Card, D., J. Kluve and A. Weber (2010), ‘Active labour market policy evaluations: a meta-analysis’, Economic Journal 120(548), F452–F477. Card, D., J. Kluve and A. Weber (2017), ‘What works? A meta analysis of recent active labor market program evaluations’, Journal of the European Economic Association, 16(3), 894–931. Crépon, B. (2019), ‘Private providers of labor market services: a review of the evidence’, in L. Calmfors and A. Bergström (eds), Framtidens Arbetsförmedling (Future Employment Service), Stockholm: Fores. Heckman, J.J., R. Lalonde and J. Smith (1999), ‘The economics and econometrics of active labor market programs’, in O. Ashenfelter (ed.), Handbook of Labor Economics Vol. 3, Amsterdam: Elsevier. Kluve J., S. Puerto, D. Robalino, J. Manuel Romero, F. Rother, J. Stöterau, F. Weidenkaff and M. Witte (2017), ‘Interventions to improve the labour market outcomes of youth: a systematic review’, Campbell Systematic Review 13(1), 1‒288. Martin, J.P. and D. Grubb (2001), ‘What works and for whom: a review of OECD countries’ experiences with active labour market policies’, Swedish Economic Policy Review, 8(2), 9–56.

African labor markets

ment in the overall labor market is within self-subsistence farming in rural areas (though many urban dwellers, for example in Ghana, also own/operate so-called “weekend farms”). Women perform a large proportion of the work in the agricultural sector, though the historically often-quoted figures of 60–80 percent have been questioned and revised by more recent studies (Palacios-Lopez et al., 2017). If anything, it appears that the relative share of female workers in agriculture on the continent has shrunk during the last 50 years (similarly, female employment in services has grown over the same period) (Dinkelman and Ngai, 2022). Finally, productivity is extremely heterogenous across farms on the African continent, with an enormous degree of variability in standard measures of productivity (Gollin and Udry, 2021).

Introduction

While there is still a paucity of knowledge about African labor markets relative to our knowledge about labor markets in other parts of the world, the evidence has been growing in recent years. This is not least due to the emergence of more and better data for more and more African countries in the past few decades. Similarly to other developing parts of the world, most countries on the African continent are characterized by weak institutions, something which also has widespread implications for the labor markets of the continent, and in several dimensions, too, ranging from low and largely non-binding minimum wages, to widespread child labor (despite formalized, if not implemented, legislation), to low levels of collective bargaining. At the same time, several issues stand out as particularly notable characteristics, or even pillars, of African labor markets. This entry briefly reviews the most important of these pillars.

Pillar 2: off-farm work Despite the continued reliance on agriculture, off-farm employment and income (defined as all employment and income from outside the agricultural household, therefore excluding on-farm self-employment activities but including off-farm salaried and casual wage employment), as well as off-farm self-employment has become increasingly important in recent years to help support the livelihoods of (primarily) agricultural households. While there is some evidence that off-farm employment helps to empower women economically by increasing their relative share of total household income (Buvinić and Furst-Nichols, 2016), at the same time it also appears that off-farm employment is dominated by households that are already relatively better-off (Loison, 2015), and also dominated by men and relatively older people, so that women and young people face a relative disadvantage in obtaining off-farm employment (Fox et al., 2016; Yeboah and Jayne, 2018).

Theory: not much on African labor markets

In an encyclopedia entry such as this, it is customary to start with the available theory and then move on the empirical evidence that has emerged on the topic. In this particular case, there is not much theory specifically related to African labor markets. The theory that does exist seems mostly to have been specifically tailored to more specialized empirical applications, mostly at the country level, if not even the within-country level, say at the regional level or for rural or urban areas (see, e.g., Gollin and Udry, 2021). The remainder of this entry therefore focuses on the most notable empirical characteristics (pillars) of the labor markets on the African continent.

Pillar 3: child work and school-to-work transition The African population is a relatively young one, which also has implications for the labor markets in African countries. Three main issues (sub-pillars) are involved here: child labor, the school-to-work transition, and youth employment more generally. As for child labor, children in African countries are much more prone to engage in

Empirical evidence: pillars of African labor markets Pillar 1: agriculture All the countries on the African continent are developing countries, and similarly to less-developed countries (LDCs) in other parts of the world a large part of the employ5

6  Elgar encyclopedia of labour studies

child labor than children in other developing countries, for example in Asia (Webbink et al., 2015). For example, in several African countries private households have been found to be large employers of economically active children, many of which are child domestic workers (Edmonds, 2007). This has implications for human capital accumulation and thus also for a multitude of other important factors, including the health and future earnings capacity of these children and their families. If children are fortunate enough to go to school, rather than work—or at least to attend school sufficiently while working part-time—the next stage on their journey on the way to adulthood is the school-to-work transition. Several additional issues are involved here. The quality of schooling, rather than just the nominal level of achievement, is an important issue in African countries, also affecting the types of jobs (and their remuneration, in turn affecting the livelihoods of the now adolescents) which graduates may be able to get. A separate issue here is the increased privatization of the education system across the African continent experienced in recent years, increasing the already great disparities across socioeconomic backgrounds (Calvès et al., 2013). Notably, there appears to be a francophone‒anglophone divide here, with francophone Africa traditionally having perceived education as a “public good” particularly strongly; something which, in turn, appears to have been inherited from the colonial period (Bierschenk, 2007). After becoming independent countries from their former colonial masters, however, public schooling was increasingly seen as a vehicle for development in multiple dimensions, including economic and social development, where the latter also includes increasing a sense of national unity and identity. In contrast, post-independence, the newly liberated countries across the continent (starting with Ghana in 1957) instead focused on promoting free, universal, and public education, especially at the primary and secondary levels; with private education “seen as the exception” and, further, also being under strict scrutiny by the state (Kitaev, 1999). With the initial job after moving on from school having profound implications for an individual’s future work career, and therefore also for their future personal/social and economic lives, and their families and their own Niels-Hugo Blunch

children, this becomes a crucially important milestone. It is affected, again, by both the quality and the type (public/private) of education obtained. Pillar 4: youth employment The labor markets of African countries are dominated by youth, since they constitute such a large relative share of the population across the entire continent. In turn, this is related to the differing fertility trends across the world, with fertility having fallen at a much lower rate across the African continent than in the rest of the world (Lam et al., 2019). Indeed, one might argue that the global labor markets as a whole are dominated by African youth, in the sense that one of every five people who start looking for their first job is born in an African country. This dominance of African youth in the total world labor force is projected to increase further, to one in three by 2050, with a similar decline in the relative share for Asian countries, and with the relative share for countries on the remaining continents being roughly unchanged (Bandiera et al., 2022). But how are the African youth faring in terms of their labor market experiences when compared to the countries and regions of the rest of the world? Based on an extensive database for the entire world, Bandiera et al. (2022) establish a set of well-supported, as well as useful, “stylized facts” regarding the relative labor market experiences of African youth. First, African youth (young adults) are equally likely to work, but less likely to get paid when compared to the rest of the world. Young adults (18–24 years of age) are almost equally split between working and non-working for both of these two aggregated groups of countries. Moving to the incidence of unpaid work, however, this is highly unequally distributed, with the share of those in unpaid work being 15.6 percent for the African continent, more than double that for the rest of the world. Examining the share of those in paid work, the relative difference is less dramatic, though still substantial, at 32.4 percent for African youth as compared to 41.5 percent for the rest of the world as a whole. Second, young adults in African countries are less likely to have an employer and more likely to be self-employed, with only 25.5 percent of labor market entrants across the

African labor markets  7

African continent as a whole obtaining a salaried job, but more than double that (68.8 percent) obtaining a salaried job across all countries outside the continent. Third, young adults in relatively wealthier African countries are more likely to have a salaried job; that is, they are more similar to countries outside the continent. More specifically, relatively higher levels of gross domestic product per capita are associated with a fall in self-employment, and a rise in salaried employment. This is true for agriculture, as well as for industry/manufacturing and services. Fourth, contrary to the rest of the world, salaried jobs are less common among the young on the African continent. That is, younger people are more likely to be self-employed and less likely to have salaried jobs on the African continent than in the rest of the world. Pillar 5: labor institutions: informality, labor unions, and minimum wages The fifth pillar comprises notable institutional issues of African labor markets, and consists of three sub-pillars: informality, labor unions, and minimum wages. To begin with, it is fair to say that institutions across the African continent are generally quite weak. This therefore also includes the labor market-related institutions of the continent. Starting with the formal‒informal distinction (although formality or informality in the labor market really operates on a continuum, rather than a dichotomy, so that firms or workers can be more or less formalized), informality is widespread across labor markets on the African continent. Indeed, while currently the informal economy now comprises more than 60 percent of total employment and more than 90 percent of all micro and small enterprises across the world, the shares are even higher in sub-Saharan Africa (ILO, 2018). The widespread informality across African labor markets has huge implications for the tax base and therefore also for public service provision, including the promotion of health and education across the African continent. Relatedly, labor unions appear to be relatively weak across the continent. This is possibly due to the mismatch between the reality of the newly independent countries on the continent, and the way Western organizations

such as the International Labour Organization (ILO) sought to deliver technical assistance to the emerging labor unions in these newly independent countries, where these organizations’ particular view of industrial unionism was very much rooted in the Northern European experience, which was frequently at odds with the political and economic realities of labor in Africa (Bernards, 2017). Yet, several studies find individual union membership and wages to be positively correlated, with a frequently both substantively large and statistically significant union premium across the continent (Blunch and Verner, 2004; Schultz and Mwabu, 1998). These studies are mostly based on cross-sectional data, however, so that issues of selection into individual union memberships, as well as unions’ selection into firms or industries, cannot be accounted for in the econometric specifications. Lastly, and related to both of the previous sub-pillars, minimum wage legislation is fairly weak across the African continents, with high levels of noncompliance among covered workers on average (Bhorat et al., 2017). This is partly due to the low coverage stemming from a relatively low degree of both formality and urbanization across the continent, though as formality and urbanization are both likely to increase in the future, minimum wage regulation will become increasingly relevant (Bhorat et al., 2017).

Knowledge gaps

Our knowledge about all the various aspects of labor markets across the African continent has been increasing in recent years. This, in no small way, is due to the increased availability of data on labor market-related issues. Despite the relative increase in the availability of data for African countries in recent decades, there is still a need for more—and better—data. We especially need more panel data to better understand the dynamic issues pertaining to African labor markets, including the five pillars of African labor markets discussed above. An example is the analysis of issues related to child labor and its potential dynamic effects on future employment and wage prospects, as well as on individual health outcomes, including the relationship with potential hazardous child work and the school‒employment transition. Niels-Hugo Blunch

8  Elgar encyclopedia of labour studies

Relatedly, causality sometimes becomes an issue where studies rely on cross-sectional survey (that is, non-experimental) data. For example, in the case of the apparent positive relationship between individual labor union membership and wages, it is not clear whether these results are partly, if not mainly, due to self-selection, either on the part of individuals’ selection into unions or on the part of labor unions selecting into particular firms and/or industries. Niels-Hugo Blunch

References

Bandiera, O., A. Elsayed, A. Smurra, and C. Zipfel (2022), “Young adults and labor markets in Africa,” Journal of Economic Perspectives 36(1), 81–100. Bernards, N. (2017), “The International Labour Organization and African trade unions: tripartite fantasies and enduring struggles,” Review of African Political Economy 44(153), 399–414. Bhorat, H., R. Kanbur, and B Stanwix (2017), “Minimum wages in Sub-Saharan Africa: a primer,” World Bank Research Observer 32(1), 21–74. Bierschenk, T. (2007), “ L’éducation de base en Afrique de l’Ouest francophone: Bien privé, bien public, bien global,” in T. Bierschenk, G. Blundo, Y. Jaffré, and M. Tidjani Alou (eds), Une anthropologie entre rigueur et engagement. Essais autour de l’œuvre de Jean-Pierre-Olivier de Sardan, Paris: APAD-Karthala, 235–257. Blunch, N.-H., and D. Verner (2004), “Asymmetries in the union wage premium in Ghana,” World Bank Economic Review 18(2), 237–252. Buvinić, M., and Furst-Nichols, R. (2016), “Promoting women’s economic empowerment: what works?,” World Bank Research Observer 31(1), 59–101. Calvès, A.E., J.-F. Kobiané, and A. N’Bouké (2013), “Privatization of education and labor force inequality in urban francophone Africa: the transition from school to work in Ouagadougou,” World Development 47, 136–148.

Niels-Hugo Blunch

Dinkelman, T., and L.R. Ngai (2022), “Time use and gender in Africa in times of structural transformation,” Journal of Economic Perspectives 36(1), 57–80. Edmonds, E.V. (2007), “Child labor,” in T.P. Schultz and J. Strauss (eds), Handbook of Development Economics, Vol. 4, Amsterdam: Elsevier Science, 3607–3709. Fox, L., L.W. Senbet, and W. Simbanegavi (2016), “Youth employment in Sub-Saharan Africa: challenges, constraints and opportunities,” Journal of African Economies 25, Supplement 1, 3–15. Gollin, D., and C. Udry (2021), “Heterogeneity, measurement error, and misallocation: evidence from African agriculture,” Journal of Political Economy 129(1), 1–80. ILO (2018), Women and Men in the Informal Economy: A Statistical Picture, 3rd edition, Geneva: International Labour Office. Kitaev, I. (1999), Private Education in sub-Saharan Africa: A Re-examination of Theory and Concepts Related to its Development and Finance, Paris: UNESCO/IIEP. Lam, D., M. Leibbrandt, and J. Allen (2019), “The demography of the labor force in sub-Saharan Africa: challenges and opportunities,” Growth and Labour Markets in Low Income Countries, Synthesis Paper 10. Loison, S.A. (2015), “Rural livelihood diversification in Sub-Saharan Africa: a literature review,” Journal of Development Studies 51(9), 1125–1138. Palacios-Lopez, A., L. Christiaensen, and T. Kilic (2017), “How much of the labor in African agriculture is provided by women?,” Food Policy 67, 52–63. Schultz, T.P., and G. Mwabu (1998), “Labor unions and the distribution of wages and employment in South Africa,” Industrial and Labor Relations Review 51(4), 680–703. Webbink, E., J. Smits, and E. de Jong (2015), “Child labor in Africa and Asia: Household and context determinants of hours worked in paid labor by young children in 16 low-income countries,” European Journal of Development Research 27(1), 84–98. Yeboah, F.K., and T.S. Jayne (2018), “Africa’s evolving employment trends,” Journal of Development Studies 54(5), 803–832.

Agricultural labor markets

Nevertheless, agricultural households participating in markets are susceptible to price volatility stemming from activities and shocks in distant locations. The welfare effects of an increase in market prices depend on whether the household is a net buyer of the good (consumes more than it produces, by purchasing from other sellers in the market) or a net seller (produces more than it consumes).

Unlike labor demand in other industries, agricultural labor demand is seasonal and characterized by a high degree of uncertainty. Rain, sunshine, and nutrients from the soil are important factors in agricultural production, and since these vary from year to year, the labor input needed varies as well. Harvest is typically the most labor-intensive activity, so farm labor demand is also seasonal. These characteristics of farm labor demand create challenges for agricultural producers to obtain workers as needed. Labor migration out of farm work, which commonly occurs as economies develop, can exacerbate these challenges.

The agricultural transformation

As economies develop, workers migrate from rural to urban areas, and agricultural households are replaced by commercial farms. To understand why workers migrate from the rural to urban sectors, assume that there is initially an abundant supply of labor on farms, and only two industries of work (farm and urban). Farm labor is initially so abundant that the marginal product of labor is near zero. In other words, the additional quantity of output that one more unit of labor produces is negligible. Lewis (1954) shows that under these conditions, if an entrepreneur were to invest capital in an urban industry, such as manufacturing, they could hire workers at subsistence wages and keep excess returns to labor as profit. The entrepreneur expands production by investing profits in additional capital and hiring more workers who travel from the farm sector. The entrepreneur might additionally invest profits in agricultural capital to ensure that there is sufficient food production to feed urban workers. Eventually, after investing capital in one or both sectors, and pulling workers from the farm to the urban sector, labor becomes relatively scarce and marginal productivity in agriculture rises. To employ additional workers in the urban sector, employers must raise wages at pace with the rising marginal product of labor on farms, and labor welfare in both sectors increases. The point at which employment of one more worker in the urban sector causes the marginal product of labor in agriculture, and thus urban wages, to rise is the “Lewis turning point.” After this point, labor scarcity plays a determining role in wages. The process of labor migration off the farm is known as the agricultural transformation, and nearly every economy undergoes this transformation as it develops (Timmer, 1988). The workforce in the United States, similarly to most countries in Western Europe and other developed nations, has transitioned

Agricultural households

Traditionally, farm work is performed by agricultural households that consume all or part of what they produce. Since agricultural production is characterized by uncertainty, agricultural households are vulnerable to changes in weather and growing conditions. Agricultural households that have perfect access to markets where they can trade goods and hire workers (or hire out their own labor) operate similarly to commercial farms in developed countries. They make production decisions to maximize the farm’s profit, and if the farm requires more labor than the household will supply, then the household hires additional workers at the market wage. Conversely, if the household has greater labor supply than the farm demands, then household members work for hire elsewhere. When market access is imperfect, meaning that the household cannot buy and sell goods or cannot hire additional workers or work elsewhere at prices determined by the market, then households decide how much to produce on-farm and how many hours to work in conjunction with their preferences for agricultural goods and leisure. Imperfect market access could result from high transaction costs, long travel time to towns where market exchange occurs, poor travel conditions, or incomplete information. Households are always at least as well off when they gain access to markets since, for example, they can choose not to trade. 9

10  Elgar encyclopedia of labour studies

almost entirely out of agricultural work. In China, where the agricultural transformation is at its early stages, workers are moving to cities at a rapid rate, but China is still home to 35 percent of the world’s 570 million farms (Lowder et al., 2016). Non-farm employment growth in rural communities (Reardon, 1997) and employment in expanding agri-food value chains (Barrett et al., 2022; Christiaensen et al., 2021), additionally provide opportunities for workers to leave farm work as local economies undergo the agricultural transformation. Charlton and Taylor (2020) find that secondary school construction in Mexico is accelerating the agricultural transformation, likely by supplying students with the requisite skills and qualifications to find non-farm jobs.

Immigration and farm labor in developed countries

In most developed countries today, only a small share of the workforce is employed directly in agriculture. Use of mechanical harvesters reduces the need for workers on farms producing wheat, corn, soy, or other row crops. An individual farmer can thus produce vast quantities of food with little to no additional help. However, harvest of fruit, vegetable, and horticultural (FVH) crops is not easily mechanized since these crops are often delicate, easily bruise, and do not ripen uniformly. Consequently, only by employing immigrants from less-developed countries have farmers in more-developed countries managed to continue producing FVH crops. Based on analysis of data from the National Agricultural Workers’ Survey (NAWS), which is a survey of farm workers conducted on farms, and designed to be nationally representative of the United States crop workforce, excluding H-2A agricultural guest workers, the majority of whom are Mexican, it is found that in 2016, more than two-thirds of the crop workforce in the United States had emigrated from Mexico. Canada brings agricultural guest workers from Mexico and the Caribbean, Western Europe brings them from Poland, and New Zealand from the Pacific Islands (Taylor and Charlton, 2018). As migrant-sending countries develop, we might expect farm labor shortages to become more prevalent. Unlike other sectors, where Diane Charlton

labor shortages are rare if not impossible without enforcing a wage ceiling, labor shortages can frequently occur in agriculture due to the seasonality and uncertainty of labor demand. Labor migration keeps farm labor markets in equilibrium. When harvest winds down in one region and ramps up in another, labor demand in the former region shifts inward, while labor demand shifts outward in the latter. Theoretically wages would decline in the former and rise in the latter, leading workers to migrate from the low-wage to the high-wage region. By raising wages, farms can attract additional workers from other regions. When there is perfect information and migration is costless, this process continues seamlessly. However, in reality, workers may lack information about work in other regions, or workers might incur high costs of moving their home and families to a new location. Farm labor contractors help to reduce frictions from imperfect information by acting as agents who connect farm employers to migratory workers. Farm labor contractors have provided an increasing share of the farm workforce in the United States, particularly in regions where seasonally labor-intensive fruits are most concentrated, including in Florida and California. Nevertheless, Richards (2018) finds evidence of perpetual farm labor shortages in the state of California from 1989 to 2014. These shortages might be explained in part by the fact that the share of farm workers in the United States who migrate has declined sharply since 1999 (Fan et al., 2015). This can be attributed in part to the increasing share of farm workers who are accompanied by their families, but empirically only a small share of the shift in migration can be explained by observable demographic changes in the workforce. Increased enforcement of immigration policies might also be partly responsible for increasing farm labor shortages. Strict immigration enforcement policies implemented in local regions of the United States caused local populations of low-skilled Mexican immigrants, and vegetable acreage, to decrease (Kostandini et al., 2014), and labor efficiency on dairies to increase as total dairy production declined and the share of dairies using more advanced labor-saving technologies rose (Charlton and Kostandini, 2020).

Agricultural labor markets  11

Rising farm wages, automation, and robotics

Migrant-sending countries are also undergoing the agricultural transformation, further tightening the farm labor supply and increasing the risk of labor shortages. As shown previously, theory indicates that farm wages, along with urban wages, rise as economies develop past the Lewis turning point. Charlton and Taylor (2016) show that the rural Mexican workforce was transitioning out of farm work at a rate of about 1 percent per year from 1980 to 2010. Real United States farm wages would have to rise by an estimated 10 percent over ten years to keep a constant farm workforce from Mexico (Charlton et al., 2019). While the agricultural transformation is expected to bring welfare benefits to Mexican workers, it creates a difficult challenge for the farming industry in countries that depend on Mexican farm workers. Agricultural transformation in migrant-sending countries leads to several fundamental questions for the agricultural industry. First, can farms simply seek out new sources of labor from countries that are not as far along in the agricultural transformation? Second, is it possible for producers to pass along increasing costs of production to consumers in the form of higher produce prices? Third, will consumers merely purchase fruits and vegetables imported from countries with more abundant labor, while agricultural acreage within the country shifts to less labor-intensive crops? Finally, will producers develop methods to improve farm labor efficiency in response to declining foreign labor supply? Concerning the first question, alternative sources of labor are likely limited. For example, the rural population of Central America is smaller than that of Mexico, and Mexico is importing farm labor from Guatemala even as it exports farm workers to the United States (Taylor and Charlton, 2018). Thus, seeking workers from further south is a limited option for the United States since farmers will have to compete for limited labor supply with farms in Mexico and other Central and South American countries. Similar labor competition can be seen in

many parts of Europe, Asia, and the Pacific Islands. Whether producers pass along increasing labor costs to consumers depends on the consumer price elasticity of demand for domestically grown fruits and vegetables. If consumers view imports as close substitutes for locally grown fruits and vegetables they will not be willing to pay higher prices for produce grown within their country. However, consumers might be willing to pay a premium for locally grown produce if they view locally grown fruits and vegetables as safer, healthier, or fresher. Inelastic demand for domestic produce can help to ensure that FVH production does not cease within a country. Nevertheless, technological advances could limit the effects of rising labor scarcity on agricultural production and labor input costs. Mechanization relating to delicate fruits and vegetables is complicated by the need for a harvester that can distinguish ripe fruit, pick the ripe fruit without tearing the skin or bruising the flesh, and accomplish this with speed comparable to human hands. Such advanced mechanization often requires cooperative expertise from multiple disciplines. For example, the automated tomato harvester, developed in the 1960s, was the joint work of plant scientists and mechanical engineers. Engineers created a machine that efficiently drew tomatoes from the ground and detached fruit from the vine. However, this machine would have been useless without the joint efforts of plant scientists who bred a tomato that ripened uniformly, detached easily from the vine, and had a tough skin that did not easily tear. The tomato harvester led to the rapid expansion of tomato acreage in the United States, and though some feared that the tomato harvester would put field hands out of work, the innovation led to additional job openings in canning and processing industries. It was estimated that overall social welfare increased by around 1000 percent because of the innovation (Schmitz and Seckler, 1970), but benefits were not evenly distributed. Many small farmers went out of business since they did not have the economies of scale to adopt the harvester, and it is unknown how easily farm workers displaced from tomato fields found new jobs in processing or other industries. Diane Charlton

12  Elgar encyclopedia of labour studies

Conclusion

Farm labor markets are complex due to the seasonality and uncertainty of farm labor demand. In poorer countries, most of the workforce often works on farms, but as countries develop, workers migrate off farms in what is known as the agricultural transformation. As migration off farms persists and farm labor becomes scarcer, the pressure to mechanize labor-intensive activities in the production and harvest of fruits and vegetables will increase. For example, in the United States, tomatoes have been mechanically harvested since the 1960s, machines spray weeds using optic sensors that distinguish weeds from desirable plants, harvesting aids automatically move strawberry trays up and down rows as workers pick, and there is currently a race to design robotic strawberry harvesters that pick strawberries without human assistance. In Europe, robotic milking technology has been in commercial use since the 1990s, and adoption of robotic apple harvesters that use vacuum suction to pull apples from the tree is underway. Some individuals will gain from the innovations and some will lose, depending on how well they can adapt to changes in labor markets and production methods. Labor demand on farms will shift inward because of mechanization, which benefits employers when labor is scarce and costly. Nevertheless, some farms may not be large enough to remain competitive after farms mechanize due to economies of scale. Use of mechanical harvesters and pruners will reduce the seasonality of labor demand by replacing large seasonal work crews with machines operated by a few individuals, and a larger share of farm workers will be employed year-round. Robotic harvesters and mechanical aids will also make workers more productive, increasing marginal product of labor, and in a perfectly competitive labor market, farm wages will rise. These innovations will thus benefit farm workers and the communities where they live and spend their paychecks (Martin and Taylor, 2003). Mechanical harvester and harvesting aids will also require skilled mechanics to keep them performing well. In anticipation of these changes, the agricultural industry can prepare for structural changes in labor supply and Diane Charlton

demand by beginning now to train mechanics and agricultural equipment operators. Diane Charlton

References

Barrett, C.B., T. Reardon, J. Swinnen, and D. Zilberman (2022), “Agri-food value chain revolutions in low-and middle-income countries,” Journal of Economic Literature, 60(4), 1316–1377. Charlton, D. and G. Kostandini (2020), “Can technology compensate for a labor shortage? Effects of 287(g) immigration policies on the U.S. dairy industry,” American Journal of Agricultural Economics 103(1), 70–89. Charlton, D. and J.E. Taylor (2016), “A declining farm workforce: analysis of panel data from rural Mexico,” American Journal of Agricultural Economics 98(4), 1158–1180. Charlton, D. and Taylor J.E. (2020), “Rural school access and the agricultural transformation,” Agricultural Economics 51(5), 641–654. Charlton, D., J.E. Taylor, S. Vougioukas, and Z. Rutledge (2019), “Can wages rise quickly enough to keep workers in the fields?,” Choices 34(2), 1–7. Christiaensen, L., Z. Rutledge, and J.E. Taylor (2021), “Viewpoint: the future of work in agri-food,” Food Policy 99, 101963. Fan, M., S. Gabbard, A.A. Pena, and J.M. Perloff (2015), “Why do fewer agricultural workers migrate now?,” American Journal of Agricultural Economics 97(3), 665–679. Kostandini, G., E. Mykerezi, and C. Escalante (2014), “The impact of immigration enforcement on the US farming sector,” American Journal of Agricultural Economics 96(1), 172–192. Lewis, W.A. (1954), “Economic development with unlimited supplies of labour,” Manchester School 22(2), 139–191. Lowder, S.K., J. Skoet, and T. Raney (2016), “The number, size, and distribution of farms, smallholder farms, and family farms worldwide,” World Development 87(4), 16–29. Martin, P. and J.E. Taylor (2003), “Farm employment, immigration and poverty: a structural analysis,” Journal of Agricultural and Resource Economics 28(2), 1–15. Reardon, T. (1997), “Using evidence of household income diversification to inform study of the rural nonfarm labor market in Africa,” World Development, 25(5), 735–747. Richards, T. (2018), “Immigration reform and farm labor markets,” American Journal of Agricultural Economics, 100(4), 1050–1071. Schmitz, A. and D. Seckler (1970), “Mechanized agriculture and social welfare: the case of the tomato harvester,” American Journal of Agricultural Economics 52(4), 569–577.

Agricultural labor markets  13 Taylor, J.E. and D. Charlton (2018), The Farm Labor Problem: A Global Perspective, Amsterdam: Elsevier Academic Press.

Timmer, C.P. (1988), “The agricultural transformation,” in: H. Chenery and T.N. Srinivasan (eds), Handbook of Development Economics, Vol. 1, Amsterdam: Elsevier Academic Press, 275–331.

Diane Charlton

Apprenticeships

Theoretical foundations of dual apprenticeship training

Introduction

Standard human capital theory According to Becker (1964), any type of education constitutes an investment in human capital. Becker’s (1964) human capital theory thus provides the foundation for an economic analysis of investments in apprenticeship training. According to his theory, individuals decide to participate in education or training (and thus invest in their human capital) if their expected benefits exceed their participation costs. Likewise, firms decide to provide training to individuals if their expected benefits exceed their costs of providing training. Consequently, education and training programmes have to offer net benefits to individuals, firms or both to ensure their participation. To determine who is willing to cover the costs and who will earn the benefits of an education and training programme, human capital theory distinguishes between general and firm-specific human capital. General human capital increases productivity to the same extent in both the training firm and any other firm, is fully marketable, and thus forces firms to pay the trained individuals wages that equal those individuals’ marginal productivity. Therefore, individuals reap all the benefits of investments in general training. Hence, human capital theory predicts that individuals themselves have to bear all the costs of general training, because firms are not willing to bear any. In contrast, firm-specific human capital increases productivity only in the training firm and is not marketable, thereby enabling firms to pay trained individuals wages below those individuals’ marginal productivity. Firms can thus reap benefits from investments in firm-specific training, but only if the trained workers stay in the training firm. Therefore, firms are willing to bear training costs as long as they can expect the trained workers to stay. To incentivise workers to stay, firms share both the benefits and costs of firm-specific training with the trained workers. Dual apprenticeship programmes provide general human capital because the training has to follow legally binding, nationally standardised curricula. Therefore, according to standard human capital theory, firms are

Apprenticeship training, also known as ‘dual vocational education and training’ (VET), constitutes one of the two main pillars of the education systems in German-speaking countries (Wolter and Ryan, 2011). Students in these countries can choose either the academic educational path, which usually leads to a university education, or the vocational path, which leads to a dual apprenticeship in a vocational occupation of their choice. This entry covers dual apprenticeships in countries with strong VET systems, such as Germany or Switzerland. To enter a dual apprenticeship programme after completing compulsory school at age 15 to 16, students must apply to a training firm for an apprenticeship position, through an application procedure very similar to that for a regular job. During dual apprenticeships, students spend roughly three-quarters of their training time at a training firm, where they acquire practical skills according to well-specified, nationally defined occupational curricula. They spend the remaining time at a vocational school, where they acquire both occupation-related and more general (cross-occupational) theoretical knowledge, again according to their occupational curriculum. Both the training firms and the vocational schools must follow the national curricula, and student learning is assessed through nationally defined examination procedures and timelines, thereby ensuring that all students completing a dual apprenticeship in a particular occupation possess similar skill sets. The duration of a dual apprenticeship programme is typically three to four years, depending on the occupation. Given the structure of dual apprenticeships, firms are crucial contributors to VET and thus to the entire education system. Therefore, from an economic perspective, the question arises of under what conditions firms are willing to contribute to a well-functioning apprenticeship system, and how firms and individuals share the costs and benefits of dual apprenticeships. This entry summarises key theoretical arguments on how costs and benefits of dual apprenticeships are shared, and outlines important empirical evidence. 14

Apprenticeships  15

not expected to be willing to invest in apprenticeship training or bear any training costs. However, empirical studies clearly show that firms both offer dual apprenticeship training and cover substantial parts of the costs of this training. For example, cost‒benefit studies show that about 70 per cent of German training firms and about 40 per cent of Swiss training firms incur net costs (e.g. Wolter et al., 2006), and earlier studies had already shown similar results, for example for the 1970s and 1980s (e.g. von Bardeleben et al., 1995). The early empirical results thus started an extensive research discussion on the ‘net-cost puzzle’ of dual apprenticeship training, often also referred to as the ‘new training literature’. New training literature Early theoretical studies in the new training literature include Sadowski (1980), who discusses labour market reputation and improved recruiting as an explanation for the net-cost puzzle of dual apprenticeship training; and Harhoff and Kane (1997), who argue that a compressed wage structure in Germany gives firms an opportunity to reap the benefits of apprenticeship training. Later studies (e.g. Acemoglu and Pischke, 1998) point more generally towards market imperfections that equip firms with enough market power to prevent dual apprenticeship graduates from switching to external employers without losing income. More recent studies consider additional sources of labour market frictions, such as works councils (e.g. Dustmann and Schönberg, 2009). All these theories help to explain why firms are willing to invest in apprenticeship training even though it teaches general skills. Importantly, however, these theories do not systematically distinguish between different types of apprenticeship occupations. Yet this distinction is critical, given that firms and individuals may face very different occupation-specific challenges. Lazear’s (2009) skill-weights approach provides a new foundation for both studying the net-cost puzzle of dual apprenticeships, and answering research questions resulting from the gap in previous theorical explanations.

Lazear’s skill-weights approach According to Lazear’s (2009) skill-weights approach, all single skills are general, but particular combinations of single skills (skill bundles) can be more or less specific depending on their usefulness in the overall labour market. Drawing on this theory, one can interpret occupations as representing skill bundles, with different occupations being more or less specific. This interpretation helps to explain differences in a variety of labour market outcomes across occupations. A series of studies apply this approach to explain the labour market outcomes of dual apprenticeship training (e.g. Mure, 2007; Eggenberger et al., 2018). These studies show, for example, that a higher degree of occupational specificity leads to lower occupational mobility and to a larger share of training costs borne by firms. Later studies apply Lazear’s (2009) theory, for example in explaining differences in individuals’ labour market outcomes across occupations after international trade shocks (Eggenberger et al., 2022). In addition, the theory has great potential for explaining other labour market phenomena arising from, for example, increased digitisation or ageing societies.

Empirical evidence on the costs and benefits of dual apprenticeship training Firms’ costs and benefits during and after apprenticeships A series of studies on the costs and benefits of dual apprenticeships in Switzerland find that, on average, Swiss firms that train apprentices experience net benefits during the three- to four-year training period (e.g. Wolter et al., 2006). While firms’ costs for apprenticeship training (for example, apprentice pay, trainers’ wages, materials) may exceed their immediate benefits (for example, apprentices’ productive work output) in the first one to two years, firms’ benefits become larger than their costs during the third or fourth year, thereby enabling firms to earn net benefits over the entire training period (e.g. Wolter et al., 2006). In contrast, empirical evidence for Austria and Germany suggests that firms in these two countries face net costs during the training period (Dionisius et al., 2009; Moretti Uschi Backes-Gellner and Patrick Lehnert

16  Elgar encyclopedia of labour studies

et al., 2019). These country differences are attributable to apprentices in Germany spending less training time on productive activities, and to apprentice pay being higher in Austria (Dionisius et al., 2009; Moretti et al., 2019). Accordingly, Mohrenweiser and Backes-Gellner (2010) find that the majority of German training firms follow an investment strategy: that is, they accept net costs during the training period to earn benefits from dual apprenticeship graduates staying at the training firm in the post-training period. As a further benefit, firms engaging in dual apprenticeship training profit from improved innovativeness. Two mechanisms lead to this improvement. Firstly, a systematic process of updating occupational curricula with inputs from firms operating at the innovation frontier promotes the most future-oriented skills in dual apprenticeships and, thereby, the diffusion of new technologies (Backes-Gellner, 1996; Rupietta and Backes-Gellner, 2019). Secondly, firms can profit from new skill combinations of dual apprenticeship graduates who continue to a university of applied sciences (UAS), a tertiary-level institution targeting primarily dual apprenticeship graduates and teaching vocational and applied research knowledge. The knowledge that dual apprenticeship graduates obtain at UASs enables firms to increase their research and development activities and, ultimately, their profits (e.g. Lehnert et al., 2020). These empirical results demonstrate that the solution to the net-cost puzzle takes different forms in different countries. Indeed, additional empirical analyses provide evidence of large differences in labour market imperfections across countries, and for firms using alternative internal mechanisms (for example, incentive pay) to retain their dual apprenticeship graduates (e.g. Mohrenweiser et al., 2019; Mühlemann et al., 2013). Further research also shows that training apprentices can serve as a screening instrument for more effectively recruiting internally (particularly for firms following an investment strategy), thereby saving potentially high recruitment costs (Blatter et al., 2016; Mohrenweiser et al., 2019). Moreover, participation in dual apprenticeship training can positively influence firms’ reputations, thereby attracting better workers from the external labour market (Sadowski, 1980).

Uschi Backes-Gellner and Patrick Lehnert

Individuals’ benefits from apprenticeship training As to the long-term benefits for individuals who invested in apprenticeship training, empirical evidence for Switzerland suggests that the lifetime earnings and employability of dual apprenticeship graduates are comparable to those of individuals with an academic degree. Dual apprenticeship graduates thus profit from their investment in apprenticeship training and offset the costs they bear during the training period (e.g. Pfister et al., 2017). Moreover, Chuard and Grassi (2021) find a high intergenerational income mobility in Switzerland compared to many other developed countries, attributing it to the dual apprenticeship system and the permeability in the overall education system. Recent studies drawing on Lazear’s (2009) skill-weights approach refine this empirical evidence. They show that individuals who completed a dual apprenticeship programme in a more specific occupation have lower occupational mobility but higher average earnings, provided that they stay in their original occupation (e.g. Eggenberger et al., 2018). Hence, given this trade-off, individuals can choose their training occupation according to their individual risk preferences. Further empirical evidence suggests that social skills and certain types of digital skills (both of which individuals acquire in dual apprenticeship programmes in many occupations) increase occupational mobility and employability, and that individuals always have to consider risk‒return trade-offs, particularly when economic conditions are likely to change (for example, in the case of external trade shocks or accelerated technological progress) (Eggenberger et al., 2022; Kiener et al., 2019). In addition to benefits such as higher earnings and better employability, dual apprenticeship graduates also experience a variety of further advantages. For example Switzerland’s highly permeable education system gives dual apprenticeship graduates not only excellent prospects for labour market entry, but also a diversity of educational career options (for example, obtaining a tertiary-level vocational degree at a UAS or – after fulfilling additional requirements – studying at a top Swiss academic university). Such educational upgrades also build valuable knowledge bridges between dual

Apprenticeships  17

apprenticeship graduates and tertiary-level academic graduates in firms, thereby creating positive spillovers for dual apprenticeship graduates in these firms; just as a rising tide lifts all boats (Schultheiss et al., 2021). Dual apprenticeships also foster the development of personal characteristics such as self-esteem or self-competence, which in the long term are associated with increased labour market success (Hoeschler et al., 2018).

Discussion and future research

This entry has presented theoretical foundations and empirical evidence on the costs and benefits of dual apprenticeships, and shown that an investment in dual apprenticeships is beneficial for both individuals and firms in the short and long terms. Although research on the costs and benefits of dual apprenticeships is already steadily increasing, the field still offers a variety of valuable avenues for future research. A first such avenue is the contribution of dual apprenticeship training to innovation. International literature on the effects of education on innovation often argues that an economy needs tertiary-level academic skills, rather than vocational skills, to achieve a high level of innovation (e.g. Aghion et al., 2014). However, this literature often neglects the important role of vocational education in innovation for countries with well-developed dual VET systems. While the empirical evidence summarised in this entry has shown that dual apprenticeships can be an important driver of innovation, analysing the underlying mechanisms – and thus contributing to a better understanding of this relation – is promising for future research. A second, and related, avenue is the ability of dual VET systems to adapt to technological change and guarantee sufficient adaptation with the necessary speed. While systematic updates of dual apprenticeship curricula (for example, approximately every five years in Switzerland) ensure that both individuals and firms can keep pace with technological change and continue contributing to innovation, sometimes curricula will need more frequent updates, given the accelerating pace of technological change. In addition, meeting these challenges may require adjustments in other elements of a dual VET system (for example, training modes, trainer qualifica-

tions, examination types). The question of which incentives are critical for achieving such systemic adjustments, in particular, remains unexplored. A third avenue for future research entails tackling the consequences of ageing societies for economies based on dual apprenticeship training. In ageing societies with fewer labour market entrants, imparting new skills (especially digital skills) to the workforce through dual apprenticeship graduates will no longer suffice. Rather, lifelong learning that systematically adds to older workers’ initial education and training will become crucial. The empirical evidence on the adaptability of dual apprenticeship graduates with broad occupational skill bundles (summarised in this entry) suggests that dual VET systems are, thus far, comparatively well equipped for meeting changing skill demands. However, future research needs to investigate more closely how the economic and technological developments challenge the functionality of dual apprenticeships, and how those apprenticeships can be systematically complemented by continuous education programmes. Uschi Backes-Gellner and Patrick Lehnert

Acknowledgements

The authors would like to thank Marco Henriques Pereira for excellent research assistance and Natalie Reid for language consulting. The work on this entry was partly funded by the Swiss State Secretariat for Education, Research and Innovation through its ‘Leading House VPET-ECON: A Research Center on the Economics of Education, Firm Behavior and Training Policies’.

References

For further references, see the extended working paper version of this entry: U. Backes-Gellner and P. Lehnert (2022), ‘Apprenticeships’, Swiss Leading House ‘Economics of Education’ Working Paper No. 194. Acemoglu, D. and J-S. Pischke (1998), ‘Why do firms train? Theory and evidence’, Quarterly Journal of Economics, 113(1), 79–119. Aghion, P., U. Akcigit and P. Howitt (2014), ‘What do we learn from Schumpeterian growth theory?’, in P. Aghion and S.N. Durlauf (eds), Handbook of Economic Growth, Volume 2B, Elsevier, 515–563.

Uschi Backes-Gellner and Patrick Lehnert

18  Elgar encyclopedia of labour studies Backes-Gellner, U. (1996), Betriebliche Bildungs- und Wettbewerbsstrategien im deutsch-britischen Vergleich (International vergleichende Schriften zur Personalökonomie und Arbeitspolitik, Band 5), Rainer Hampp. Becker, G.S. (1964), Human Capital: A Theoretical and Empirical Aanalysis, with Special Reference to Education, National Bureau of Economic Research. Blatter, M., S. Mühlemann, S. Schenker and S.C. Wolter (2016), ‘Hiring costs for skilled workers and the supply of firm-provided training’, Oxford Economic Papers, 68(1), 238–257. Chuard, P. and V. Grassi (2021), Switzer-Land of opportunity: Intergenerational income mobility in the land of vocational education. Available at SSRN: https://​doi​.org/​10​.2139/​ssrn​.3662560. Dionisius, R., S. Mühlemann, H. Pfeifer, G. Walden, F. Wenzelmann and S.C. Wolter (2009), ‘Costs and benefits of apprenticeship training: a comparison of Germany and Switzerland’, Applied Economics Quarterly, 55(1), 7–37. Dustmann, C. and U. Schönberg (2009), ‘Training and union wages’, Review of Economics and Statistics, 91(2), 363–376. Eggenberger, C., S. Janssen and U. Backes-Gellner (2022), ‘The value of specific skills under shock: high risks and high returns’, Labour Economics, 78, 102187. Eggenberger, C., M. Rinawi and U. Backes-Gellner (2018), ‘Occupational specificity: a new measurement based on training curricula and its effect on labor market outcomes’, Labour Economics, 51, 97–107. Harhoff, D. and T.J. Kane (1997), ‘Is the German apprenticeship system a panacea for the U.S. labor market?’, Journal of Population Economics, 10, 171–196. Hoeschler, P., S. Balestra and U. Backes-Gellner (2018), ‘The development of non-cognitive skills in adolescence’, Economics Letters, 163, 40–45. Kiener, F., A.-S. Gnehm, S. Clematide and U. Backes-Gellner (2019), ‘Different types of IT skills in occupational training curricula and labor market outcomes’, Swiss Leading House ‘Economics of Education’ Working Paper No. 159. Lazear, E.P. (2009), ‘Firm-specific human capital: a skill-weights approach’, Journal of Political Economy, 117(5), 914–940. Lehnert, P., C. Pfister and U. Backes-Gellner (2020), ‘Employment of R&D personnel after an educational supply shock: effects of the introduction of Universities of Applied

Uschi Backes-Gellner and Patrick Lehnert

Sciences in Switzerland’, Labour Economics, 66, 101883. Mohrenweiser, J. and U. Backes-Gellner (2010), ‘Apprenticeship training: For investment of substitution?’, International Journal of Manpower, 31(5), 545–562. Mohrenweiser, J., T. Zwick and U. Backes-Gellner (2019), ‘Poaching and firm-sponsored training’, British Journal of Industrial Relations, 57(1), 143–181. Moretti, L., M. Mayerl, S. Mühlemann, P. Schlögl and S.C. Wolter (2019), ‘So similar and yet so different: a firm’s net costs and post-training benefits from apprenticeship training in Austria and Switzerland’, Evidence-Based HRM, 7(2), 229–246. Mühlemann, S., P. Ryan and S.C. Wolter (2013), ‘Monopsony power, pay structure, and training’, Industrial and Labor Relations Review, 66(5), 1097–1114. Mure, J. (2007), Weiterbildungsfinanzierung und Fluktuation: Theoretische Erklärungsansätze und empirische Befunde auf Basis des Skill-Weights Approach (Beiträge zur Personalund Organisationsökonomik, Band 16), Rainer Hampp. Pfister, C., S.N. Tuor Sartore and U. Backes-Gellner (2017), ‘The relative importance of type of education and subject area: empirical evidence for educational decisions’, Evidence-based HRM, 5(1), 30–58. Rupietta, C. and U. Backes-Gellner (2019), ‘How firms’ participation in apprenticeship training fosters knowledge diffusion and innovation’, Journal of Business Economics, 89(5), 569–597. Sadowski, D. (1980), Berufliche Bildung und betriebliches Bildungsbudget, Poeschel. Schultheiss, T., C. Pfister, A.-S. Gnehm and U. Backes-Gellner (2021), ‘Tertiary education expansion and task demand: Does a rising tide lift all boats?’, Swiss Leading House ‘Economics of Education’ Working Paper No. 154. von Bardeleben, R., U. Beicht and K. Fehér (1995), Betriebliche Kosten und Nutzen der Ausbildung: Repräsentative Ergebnisse aus Industrie, Handel und Handwerk, Bertelsmann. Wolter, S.C. and P. Ryan (2011), ‘Apprenticeship’, in E.A. Hanushek, S. Machin and L. Woessmann (eds), Handbook of the Economics of Education, Vol. 3, Elsevier, 521–576. Wolter, S.C., J. Schweri, J. and S. Mühlemann (2006), ‘Why some firms train apprentices and many others do not’, German Economic Review, 7(3), 249–264.

B

Beveridge curve, matching functions

ployment to the flows into and out of unemployment and to the continuous job creation and destruction that takes place every month. In the benchmark search and matching model, job creation is determined by firms that post vacancies to fill jobs and unemployed workers who search for jobs. The labour force is assumed to be fixed (often normalized to one) and workers may be either employed or unemployed. There is free entry of firms to the market, so vacancies are posted until their value is driven to zero. Filling an open vacancy or finding a job is not instantaneous, due to labour market frictions and mismatches related to, for example, skills or physical distance. The model acknowledges that finding a worker for a job, or for a worker to find an open job vacancy, is a time-consuming and costly activity that is characterized by search frictions. A matching function, that is in many ways comparable to a production function that uses vacancies and unemployed workers as inputs and produces firm‒worker matches as output, summarizes the frictions and heterogeneities of the labour market. The matching function is often assumed to have constant returns to scale, typically Cobb‒ Douglas ​m(​ u, v)​  =  A ​u​ α​ ​v​ 1−α​​, where u and v characterize unemployment and vacancies, and A is a shift parameter that encapsulates matching efficiency. Once matched, the firm‒worker pairs are productive until the match is terminated by an exogenous job destruction shock δ. Job creation and destruction determine the inflow to and outflow from unemployment, characterized by the flow equation for unemployment:

Introduction

The Beveridge curve is widely used to describe the cyclical state of the labour market and the efficiency of the labour market in terms of matching unemployed workers to job vacancies. It traces a negative relationship between unemployment and vacancy rates over the course of a business cycle, tracing the evolution of the economy from expansionary phases (with lower unemployment and higher vacancies) to contractions in activity (with higher unemployment and lower vacancies). Movements along the Beveridge curve have typically been interpreted as reflecting cyclical labour market dynamics, whereas shifts in the Beveridge curve have typically been interpreted as reflecting changes in matching efficiency or structural change. Shifts in the Beveridge curve are of particular interest in times of intense reallocation due to technological shocks or crisis periods since they are suggestive of structural changes in the unemployment‒vacancy relationship. The downward-sloping Beveridge, with occasional medium-term shifts, is an empirically robust finding across countries and over time. The origins of the curve arise from the work of William Beveridge (1944), who emphasized the role of job vacancies in the relationship between the demand of workers and the rate of unemployment, although the curve was presented only later.

Theories of search and labour market frictions

​u̇ ​  ​=  δ​(1 ​−  u)​ − ​   m​(u, v)​

The Beveridge curve is closely related to and easily understood in the context of the search and matching model (Diamond, 1982; Pissarides, 1985; Mortensen and Pissarides, 1994). The model relates equilibrium unem-

The equation states that fluctuations in unemployment are determined by separations of employed workers (unemployment inflow) 19

20  Elgar encyclopedia of labour studies

Figure 1

The Beveridge curve, Finland, 1988–2021

and the matching of unemployed workers to vacancies (unemployment outflow). The job finding rate of a worker is given by​ m​(u, v)​ ​ u  ​​,   where ​θ  =  ​ _uv ​​is labour market p​(θ)​  =  _ tightness. A higher number of vacancies raises the job finding rate for workers, and a higher number of unemployed reduces it as competition for the open vacancies is higher. Setting the flows equal (​u​  ̇​  =  0​) gives the equilibrium level of unemployment: ​u  ​=  _ ​  δ ( ) ​.  ​ δ ​+  p​ θ ​

This is the Beveridge equation that produces a downward-sloping curve in the u,v-space, where the matching function plays a key role. Figure 1 shows the Beveridge curve for Finland for a period of about 30 years.

Cyclical and structural changes in the labour market

For a given rate of job destruction and matching efficiency, an increase in vacancies raises the job finding rate and leads to lower unemployment, and vice versa. Such movements in opposite directions of vacancies and unemployment along the Beveridge curve are typically interpreted as cyclical changes in the labour market: at the peak of a business cycle the unemployment rate is low and the job vacancy rate is high; in downturns firms post fewer vacancies and the unemployment Juuso Vanhala

rate rises. Changes in the efficiency of matching induce shifts of the Beveridge curve. A reduction in matching efficiency shifts the Beveridge curve outwards from the origin. For a given level of vacancies the equilibrium level of unemployment is higher; or, a higher level of vacancies is required to achieve a given level of unemployment. A change in the job separation rate also results in a shift of the Beveridge curve. The distinction between cyclical changes and outward or inward shifts of the Beveridge curve is not always immediate, as the Beveridge curve tends to make counterclockwise loops over the business cycle (Blanchard and Diamond, 1989). In the early stages of a recovery, the posting of vacancies is faster than matching unemployed workers to jobs, so the cyclical fall in unemployment lags the increase in vacancies. As the recovery matures, posting vacancies moderates and the fall in unemployment catches up, and the curve approaches its earlier position. Outward shifts occur when the curve does not close the counterclockwise loop. Changes in the efficiency of matching may arise from various sources. Outward shifts in the Beveridge curve are often attributed to skill mismatch, where technological change reduces the demand for some types of skills while increasing demand for other skills that are less readily available. Another source of mismatch is the geographical dis-

Beveridge curve, matching functions  21

persion of unemployment and vacancies. This type of mismatch may arise due to idiosyncratic shocks within local labour markets. Moreover, a shift in the distribution of job vacancies, from sectors and occupations with a high matching rate to sectors with low matching rates, lowers aggregate matching. These sources of weaker matching may lead to the simultaneous rise in unemployment and vacancies, and thus to an outward shift in the Beveridge curve. Also, changes in labour market institutions played an important role for movements in the Beveridge curve in European countries from the 1960s to the 1990s, particularly those which are important for search and matching efficiency (Nickell et al., 2001). Improved matching efficiency and inward shifts of the Beveridge curve may be due to, for example, improved job search assistance or improvements in job search technology such as online job postings. Matching efficiency may also display cyclical fluctuations (Barnichon and Figura, 2015). When the average characteristics of the unemployed deteriorate substantially in a recession, matching efficiency will decline. For example, a rise in the share of long-term unemployed during a recession will reduce matching if skill obsolescence during unemployment makes workers less employable and lowers their job finding rate. Matching efficiency may also move over the cycle when dispersion in labour market conditions – the degree to which some labour submarkets fare worse than others – increases markedly. Dispersion in labour market conditions drives down matching efficiency because of the concavity of the matching function.

Some extensions

The basic model is a stylized characterization of the labour market and abstracts from many features that are important for job flows, and may thus potentially affect the behaviour of the Beveridge curve. Many of these features, which have been studied both theoretically and empirically, are related to the job creation margin, but the job destruction margin is also relevant. I refer to some extensions that have featured prominently on research agendas and where gaps in our knowledge remain. See Pissarides (2000) for an exposition of a theoretical treatment of these extensions, and Elsby et al. (2015) for an overview of empirical results.

By choosing their intensity of search, workers and firms can aim to increase their matching probability. The intensity of search enters the matching function as efficiency units of search, m ​   =  m​(su, av)​​, where s and a represent the intensity of search by the unemployed workers and firms having open vacancies. Optimization takes place between a higher probability of matching and a higher search cost, which the greater search effort entails. In the basic theory, workers’ search effort is procyclical, but available evidence challenges this prediction (Pissarides, 2000; Elsby et al., 2015). The cyclical properties of firms’ search effort depend on the elasticity of vacancy creation (Davis et al., 2013). A persistent rise in search intensity induces an inward shift in the Beveridge curve, ceteris paribus. The matching process may be characterized in more detail by separating the meeting of the firm and worker that results from search from the decision of accepting the match (that is, that the meeting actually leads to productive match). Job acceptance may be modelled by incorporating an endogenous reservation wage (for the worker) or productivity (for the firm) that is required for an acceptable match. The acceptance decision implicitly affects the efficiency of matching, and it features in the matching function _ ​  )​  ​]​m(​ u, v)​​, where m  =  [​ 1 − F​(x  accordingly, ​ a match specific productivity _ x is drawn from ​​  ​​  is the reservation a distribution ​F(​ x)​​and x  productivity. A realization below this level is rejected and the parties continue search. The reservation level may depend on institutional factors, such as the level of unemployment insurance. All else equal, a higher threshold for acceptance will lower the matching rate and push the Beveridge curve outwards. The pool of searching workers may also be extended beyond the unemployed. Job search may take place while employed or from the ranks of the nonparticipants. The job search of workers enters the matching function as an input ​ m  =  m​(​u​ unemp​ + ​u​ emp​, v)​​, where ​​u​ unemp​​ are unemployed searching workers and u​​ ​ emp​​ are employed searching workers. The labour market participation margin can also be relevant for the Beveridge curve, as workers may enter employment directly from the labour force without passing through intermittent unemployment: for example, from studies or parental leave. As with the long-term Juuso Vanhala

22  Elgar encyclopedia of labour studies

unemployed, it may be plausible to assume that their attachment to the labour market and contribution in the matching function may differ from those of the unemployed. The Beveridge curve may also be affected by the job destruction margin. In the basic model it is to be assumed as fixed, but in more elaborated models shocks to the profitability of the match can determine job destruction endogenously. Empirically the inflow to unemployment is countercyclical to varying degrees across countries and according to the severity of downturns. This can make the adjustment of unemployment more cyclical and alter the slope of the Beveridge curve. Labour force composition or demographics may also influence the Beveridge curve in the medium to long run. For example, the job flows of elderly workers are lower than those of younger workers, and exits from employment may more frequently lead to exits from the labour force than to unemployment. Incorporating more realistic features into the model often has direct effects on the Beveridge curve through the matching function. However, they may also indirectly affect wages through bargaining in the full-fledged matching model of the labour market. In the full model, matching frictions and costly search imply that the match as such has value, and that there is a match surplus that is divided between the firm and the worker typically by Nash bargaining of the wage. The parameters of the model, but also labour market tightness, θ​ ​, affect the surplus to be shared and the wage outcome. A higher match surplus will provide incentives for firms to create more vacancies. Also, higher surplus for the worker may affect search effort or the labour market participation decision. Although the Diamond‒Mortensen‒ Pissarides framework provides an intuitive qualitative description of the behaviour on unemployment and vacancies, the standard model has difficulties in generating the observed business cycle frequency fluctuations in these variables in response to shocks of a plausible magnitude (Shimer, 2005). A large literature has enriched the model to improve its performance in this respect, where much of the focus has been on the flexibility of wages and the assumption of free entry for posting vacancies.

Juuso Vanhala

Data on vacancies

The Beveridge curve traces a negatively sloped locus in the unemployment‒vacancy space, and meaningful analysis of the Beveridge curve – especially shifts – requires long time series. Unemployment data is readily available in long harmonized series and is a rather uncontroversial and well-understood variable. But even here, some issues remain. For example, the unemployment figures from the Labour Force Survey collected by statistical authorities in European countries may differ somewhat in level and fluctuations from those provided by public employment officials (employment agencies), due to different definitions of unemployed and employed workers. Vacancy data vary more across countries and over time, even though there has been harmonization. In earlier studies on the United States Beveridge curve, the Conference Board’s Help Wanted index was used. Subsequently, survey data on vacancies has become available, such as the data from the Job Openings and Labor Turnover Survey (JOLTS) in the United States since 2000, and the job vacancy data from Eurostat since 2006. For individual European countries, longer series collected by statistical authorities are often available, but these data are not fully harmonized across countries (for example, cross-country sectoral and coverage differences). Vacancies reported to the public employment services is another source of vacancy data, used in the Organisation for Economic Co-operation and Development’s (OECD) Job Vacancies Statistics and in numerous cross-country studies sourcing that data (e.g., Hobijn and Sahin, 2013; Nickell et al., 2001). This data may be biased for several reasons. For example, public sector vacancies may be required to be announced in employment offices by law, whereas many private sector firms may not post vacancies in employment offices. With the evolution of internet-based job search, the role of employment office vacancies has diminished. Many recruitments are also made through direct contacts without ever posting an open vacancy. Moreover, measuring labour demand by the stock of open vacancies at the end of the month involves a time aggregation bias, as vacancies that are opened and filled within the month are not included in the measure. Finally, a proxy for vacancies used

Beveridge curve, matching functions  23

in a number of studies (e.g., Bonthuis et al., 2016) is the European Commission’s survey series of employers’ perceptions of labour shortages.

Acknowledgements

The opinions expressed in this entry are those of the author, and do not necessarily reflect the views of the Bank of Finland or the Eurosystem. The author thanks Petteri Juvonen and Matti Virén for very helpful comments. Juuso Vanhala

References

Barnichon, R. and A. Figura (2015), ‘Labour market heterogeneity and the aggregate matching function’, American Economic Journal: Macroeconomics, 7(4), 222–249. Beveridge, W.H. (1944), Full Employment in a Free Society. London: G. Allen. Blanchard, O.J. and P.A. Diamond (1989), ‘The Beveridge curve’, Brookings Papers on Economic Activity, 20(1), 1–76. Bonthuis, B., V. Jarvis and J. Vanhala (2016), ‘Shifts in euro area Beveridge curves and their determinants’, IZA Journal of Labor Policy, 5(1), 1–17.

Davis, S.J., R.J. Faberman and J.C. Haltiwanger (2013), ‘The establishment-level behaviour of vacancies and hiring’, Quarterly Journal of Economics, 128(2), 581–622. Diamond, P.A. (1982), ‘Aggregate demand management in search equilibrium’, Journal of Political Economy, 90(5), 881–894. Elsby, M.W.L., R. Michaels and D. Ratner (2015), ‘The Beveridge curve: a survey,’ Journal of Economic Literature, 53(3), 571–630. Hobijn, B. and A. Sahin (2013), ‘Beveridge curve shifts across countries since the Great Recession’, IMF Economic Review, 61(4), 566–600. Mortensen, D.T. and C.A. Pissarides (1994), ‘Job creation and job destruction in the theory of unemployment’, Review of Economic Studies, 61(3), 397–415. Nickell, S., L. Nunziata, W. Ochel and G. Quintini (2001), ‘The Beveridge curve, unemployment and wages in the OECD from the 1960s to the 1990s ‒ preliminary version, Centre for Economic Performance DP 502, London School of Economics and Political Science. Pissarides, C.A. (1985), ‘Short-run equilibrium dynamics of unemployment, vacancies, and real wages’, American Economic Review, 75(4), 676–690. Pissarides, C.A. (2000), Equilibrium Unemployment Theory, 2nd edition. Cambridge, MA: MIT Press. Shimer, R. (2005), ‘The cyclical behaviour of equilibrium unemployment and vacancies’, American Economic Review, 95(1), 25–49.

Juuso Vanhala

D

Dynamic employment adjustment of firms

immediately. The level of production may change, the price of the good may change, but nothing is dynamic about the adjustments that you make. It is thus possible to observe changes of prices and quantities over time. Nevertheless, as long as those changes are in the error term—that is, as long as they are unpredictable but immediate—even though changes may occur frequently, you live an entrepreneurial life in a flexible, stochastic, but static world. It may in fact, however, not be so easy to find more space, to buy new equipment, to find people who want to help you or people who can. How do you make these decisions, and what does it take to make them instantly? Is that at all possible? To find out how a firm is going about making such decisions, four economists at the Carnegie Institute of Technology, Charles Holt, Franco Modigliani, Herbert Simon, and John Muth, teamed up spending long days inside a firm to observe and find out how precisely the planning of inventories, production, and the workforce takes place (Holt et al., 1960). They found that rather than immediately changing things—which was what most (static equilibrium) economic models at the time implicitly assumed—it takes time to update, because it is costly to adjust.

Introduction

What precisely is dynamic about the level of employment of a firm that changes all the time? Suppose you have an idea to make stuff, to produce things, to create something new. If you like what you are doing, you may decide to spend a lot of time on your creation. This can be worthwhile and rewarding for many reasons. One reason might be that you think you can make a living from selling what you make, because you think, or others have told you, that other people like the thing you make. You decide to start selling the product and become an entrepreneur. After a while, it turns out that what you make is a hit: people want more and more of it. What do you do? In principle you have two variables to play with: the price of the product and the quantity you produce. You might want to raise the price for the thing you make. Although some people may end up not buying your product any longer, you will earn more per item sold. Alternatively, you may wish to increase the amount you produce. That may take more storage space, more equipment, and maybe the help of other people. The outcome of it all depends on the uniqueness of the product, the investment costs of extra equipment, the complexity of the production process, the available expertise of people who might help, and the competition of others like you who make similar things. Economists will argue that you will decide to go so far that the costs of making one more item will be equal to the revenues of that extra item. Then marginal costs of production equal marginal returns. You are in a static equilibrium. As long as you can easily adjust to all the unforeseen fluctuations in the demand for your product, a new equilibrium price and quantity are within reach

Need for speed

So, when all else is there except the extra hands you are looking for, you may decide to employ new workers for your operation to produce more of that thing you make. Apart from the wage that you can afford to pay, you need to search for good people who are willing to supply their time and effort to your operation. In the case of a match, they may also need instructions and training to learn how to produce exactly the way you wish. These overhead costs (Clark, 1923) or hiring costs may be fixed, or they can depend on the number of people you wish to employ. Hiring 24

Dynamic employment adjustment of firms  25

costs result in labor hoarding and determine the heterogeneity of the firm. The irreversibility of fixed disbursements plays a pivotal role in the design of optimal trajectories of change. At some point it may well be that your good is out of fashion. You will have to decide whether the drop in demand for your product is just a temporary blip or a permanent decline. If it is temporary, waiting may be optimal. Employees in possession of non-tradable firm-specific skills will be less likely to quit. When more information about the duration or deepness of the downturn becomes available, it may become inescapable to fire at least some of your employees. However, how many, and how fast, that is hard to say. While hiring costs are largely the reason for dynamic optimization from the firm’s perspective, firing costs are largely induced by legislation and meant to protect workers from being laid off too easily or too quickly; to provide job and earnings security to workers. Firing costs are typically a function of age, tenure, contract hours, and the hourly wage. Workers and firms should be willing to trade wage for security. Where firing costs are high, the flexibility of worker turnover (the extensive margin) is low, but hours worked (the intensive margin) fluctuate more. When the idea of firing costs is to protect workers’ employment, it is important to realize that the way firing costs are imposed upon firms is almost always such that these costs protect the best-paid workers most, and they are typically the ones with the highest levels of education, the best contracts, and the longest tenure. That may never have been the intention of policymakers, but studying and understanding worker heterogeneity in the labor market cannot lead to any other conclusion (Pfann, 2006). Hiring and firing costs are fundamentally different types of expenses. There is absolutely no reason to assume them to be symmetric. Hiring and firing costs act together, and they define what may be the best policy for personnel changes. For this reason alone, employers do not hire new workers at the beginning of every new working day, and fire all of them at the end of that day. Employment decisions of the firm are outcomes of dynamic long-term contingency

plans; long-term plans, which need updating all the time. Adjustment costs associated with these updates of labor demand are the reason that expectations and uncertainties about future development of prices, wages, product demand, technological developments, investment costs, and labor supply are coming into play. Good overviews that explain how to model workforce changes under uncertainty when sizes and structures of adjustment costs may differ between various types of employment are helpful, and are available in the literature; see Nickell (1986) and Hamermesh and Pfann (1996). These well-cited reviews present and explain solutions of dynamic stochastic optimization problems, which are technical road maps leading to optimal paths for new equilibrium levels of employment. You might think that flexible labor markets are preferred over dynamic labor markets. Job security is costly because it limits your control as an employer wishing to adjust promptly to changing market conditions. Labor market flexibility is the characteristic that fluctuations in the number of employees coincide with the volatility of input prices and product demand. Employees wish for job and income security. You may be convinced that the ultimate flexibility comes from paying independent contractors to do the job that must be done. Independent contractors are self-employed. Independent contractors demand high remunerations, and you will be paying only for the right to control or direct the result of their work. You cannot determine what will be done and how it will be done. Consequently, independent contracting can only be a solution to temporary imbalance between the desired and the actual workforce of the firm. In general, it is fair to say that, on the one hand, aggregate employment data show smoothed dynamic adjustment. Firm-level data of workforce adjustments, on the other hand, show lumpy adjustment patterns. Firms tend to change productive labor inputs only occasionally, and simply produce without adjusting their workforce at the extensive margin most of the time. Lumpiness in employment fluctuations at the firm level is an optimal outcome of the existence of fixed and irreversible adjustment costs expenditures.

Gerard A. Pfann

26  Elgar encyclopedia of labour studies

Interrelationships

So far so good, but what do you do if you wish to upgrade or expand your business by means of new investments, new technology, and new workers with matching knowledge and skills for these modern techniques, all at the same time? Do you train workers first before installing new machines? Is it better to implement new technology first and train workers later? Can it be that the optimal adjustment speed of employment depends on the slowness of building up new physical production sites? Vice versa, does one wait to invest in new production technologies if too few people are capable of working effectively with these new machineries? How does it all fit together? Is it possible to tell all this apart? This is the problem of interrelatedness of production factors. Mohammed Ishaq Nadiri and Sherwin Rosen were the first to address it (Nadiri and Rosen, 1969). With a few exceptions, the existing literature on the irreversibility of disbursements and the lumpiness of production factor input adjustments is concerned with the analysis of one production factor only, be it capital investments and disinvestments, or job creation and job destruction. Empirical studies based on micro-data have shown, however, that substantial and important amounts of inaction observations exist for employment changes and capital investments. Most of the time, lumpy adjustments happen concurrently. This revealed preference of firms waiting to adjust may result from complementarities in the production process, or from reduced expenses when adjustments can be made all at the same time. Firms prefer simultaneously updating factor inputs over exercising control sequentially. A more recent study (Asphjell et al., 2014) finds that the relative cost advantages related to simultaneous investments in productive capital and workforce changes are 3 percent for the former and 31 percent for the latter. The prevailing results demonstrate that factor demand models with fully specified interrelated and irreversible adjustment structures are outperforming all other dynamic factor input models with different adjustment costs structures.

Gerard A. Pfann

Finale

The time needed for upward or downward adjustment to new balance, to new equilibrium levels of employment, depends largely on the size and structure of hiring and firing costs. Clearly, the optimal speed of adjustment rests on many factors. Dynamic employment adjustments of firms are large-scale frames of operational complexity, with employment protection legislation triggering increased complicacy of result optimization. These complexities, however, should not be used as reasons to deter thinking about the dynamics of employment adjustment of firms. Observed differences in adjustment costs, which define heterogeneity among workers and among firms, have marked the limits of our current state of unawareness; for example of the fact that in today’s system, workers who need protection the most obtain the least job security. At least it will be worthwhile to research new sizes and new structures in order to understand and design firms’ dynamic employment adjustment in ways that have an eye for the most vulnerable workers: “a bit of ingenuity might get somewhere” (Solow, 1998). Gerard A. Pfann

References

Asphjell, M.K., W.H. Letterie, Ø.A. Nilsen, and G.A. Pfann (2014), “Sequentiality versus simultaneity: Interrelated factor demand,” Review of Economics and Statistics 96 (5), 986‒998. Clark, J.M. (1923), Studies in the Economics of Overhead Costs. Chicago, IL: University of Chicago Press. Hamermesh, D.S. and G.A. Pfann (1996), “Adjustment costs in factor demand,” Journal of Economic Literature 34 (3), 1264–1292. Holt, C.C., F. Modigliani, J.F. Muth, and H.A. Simon (1960), Planning Production, Inventories, and Work Force. Englewood Cliffs, NJ: Prentice-Hall. Nadiri, M.I. and S. Rosen (1969), “Interrelated factor demand functions,” American Economic Review 59 (4), 457‒471. Nickell, S.J. (1986), “Dynamic models of labour demand,” in Handbook of Labour Economics, Volume I, edited by O. Ashenfelter and R. Layard. Amsterdam: Elsevier Science Publishers, 473–522.

Dynamic employment adjustment of firms  27 Pfann, G.A. (2006), “Downsizing and heterogeneous firing costs,” Review of Economics and Statistics 88 (1), 158–170.

Solow, R.M. (1998), “What is labour-market flexibility? What is it good for?” Proceedings of the British Academy 97. Lectures and Memoirs, 189–211.

Gerard A. Pfann

E

Early retirement

prompt questions of the sustainability of public pension systems; and (3) the actuarial fairness of early retirement when life expectancy is increasing.

Regular, or ‘normal’, retirement allows a person to retire with a full (social security) pension without penalties. The Organisation for Economic Co-operation and Development (OECD) defines a country’s normal retirement age (NRA) as the age of eligibility to a full pension, based on a full career after labour market entry at age 22. In many countries, people may retire earlier with a lower pension, adjusted for fewer contributions and an expected longer retirement period. Early retirement age (ERA) is the earliest age at which a person can claim an old age pension. (Many countries provide provisions for early retirement if a person is incapable of working due to illness or accident.) For example, in Australia, people may retire early at the age of 55 from their ‘funded defined contribution’ (FDC) plans, and at age 65 from the targeted pension plan (OECD, 2019, Table 4.4). (FDCs are based on individual pension accounts that accumulate contributions which are converted to a pension at retirement.) In Portugal, early retirement is possible at the age of 62 years, and the regular retirement age is at 65.2 years of age (OECD, 2019, Table 4.4). Early retirement typically carries a penalty, while later retirement is rewarded with a bonus. For example, in Finland, an individual with an uninterrupted career after entering the labor market at age 22 would be able to retire early at age 65 at a penalty of 4.8 per cent for each year that they retired before the regular age of 67.9, and a bonus of 4.8 per cent if they retired after an age of 67.9 years (OECD, 2019, Table 4.5). The current discussion in economics and social policy debates focuses on: (1) the determinants of early retirement, including incentives through arguably generous pension systems; (2) population ageing and expected decreases in the labour force which

Determinants of early retirement

The economic literature considers retirement typically as an informed choice where people choose to retire depending on their expected future wage and pension incomes (e.g., Burkhauser, 1979). Generous retirement benefits and low ERAs create incentives for early retirement and lead to low labour force participation rates at older ages. For example, a lower ERA in Norway led to a substantial increase in early retirement. Population ageing poses significant challenges to public pension systems. Most OECD countries have therefore revised their retirement policies since the 1990s to encourage longer working lives to counteract the decline of the working-age population. Reforms include stricter eligibility requirements and increases in the pension eligibility age, the introduction of actuarial deductions for early retirement, increases in the NRA, increases in the statutory retirement age, or a combination of these policies (Böheim et al., 2021). However, financial incentives alone do not explain retirement patterns, and evidence suggests that retirees also use the statutory retirement age as an anchor for their retirement decisions. As a result, the average retirement age has increased over the past 20 years. The assumption that people can freely choose their optimal retirement age has been challenged. Several studies distinguish between voluntary and involuntary retirement, with involuntary early retirement seen as a severely constrained choice. While voluntary early retirement is considered a worker’s conscious decision, involuntary early retirement is retirement due to employment constraints, such as discrimination against older workers. Dorn and Sousa-Poza (2010) 28

Early retirement  29

find that involuntary early retirement is an empirically important phenomenon in Europe, especially in countries where labour market participation rates of older workers is low. In contrast, very few retirements are involuntary in the United States. More flexible labour markets and non-standard employment contracts increase the labour market participation of older persons, and in countries where part-time employment among older people is more frequent, the labor supply of older persons is higher, at both the extensive and the intensive margin. In general, stricter access to early retirement or greater penalties for retiring early induce people to retire later. However, depending on the strictness or generosity of the rules for early retirement, people might resort to unemployment or disability pensions instead (e.g., Duggan et al., 2007). Negative health shocks, or expectations about a decline in health, increase the probability of early retirement. Consequently, the employment rates of persons who report poor health are consistently below those of persons in good health (Böheim and Leoni, 2018). Rising disability benefit use and falling employment rates (especially among older workers with poor health) have led many countries to reform their disability benefit systems and policies for workers with health problems (Baumberg Geiger et al., 2019). OECD countries have strengthened the activation and the reintegration components of their sickness and disability policies, while tightening benefit conditions for working-age persons to keep older people with health problems in employment longer. Böheim et al. (2021) emphasize that improvements in the labour market integration of people with health limitations are a particularly promising way to increase the labour force participation of older people. However, there are also concerns ‒ perhaps more in the medical than the economic literature ‒ about the consequences of longer working lives on workers’ health (e.g., Hengel et al., 2021).

Technology and retirement

Technological change may have opposite consequences for early retirement. It could erode skills, leading to lower productivity and lower wages, thus increasing the incentive to retire. Alternatively, technological change could complement existing skills and main-

tain or even increase a worker’s productivity, increasing the opportunity costs of retirement and reducing the incentive to retire. The erosion of skills by technological advances is an important factor in early retirement that could be addressed through retraining and continuing education. Several countries have recently introduced an automatic link between future pensions and life expectancy (Ayuso et al., 2021). Countries use different approaches, such as linking life expectancy to initial pensions (for example, Italy) or indexing the retirement age to life expectancy (for example, Denmark). However, linking early retirement penalties to life expectancy, conditional on contribution years and age of retirement, does not ensure actuarial neutrality between contributions and benefits. An increase in life expectancy requires higher penalties for early retirement to ensure actuarial neutrality. (See Knell, 2021 for a discussion of actuarial deductions for early retirement and actuarial supplements for late retirement, particularly the assumptions required and their implications for welfare considerations.)

Partial retirement

Partial retirement programmes allow for a more flexible entry into retirement by gradually reducing working hours before full retirement. Schemes differ, in particular, in the age at which a worker may enter partial retirement and in the amount of working time that can be reduced (Albanese et al., 2020, Table B.1). Theoretically, the effects of partial retirement on labour supply are ambiguous. The overall effect is composed of two opposing effects: an increase in the labour supply of those who choose partial retirement over early retirement (or other forms of out of labour force, OLF) and the decrease in labour supply of workers who choose partial retirement (Haan and Tolan, 2019). The empirical evidence is mixed, not only because of the ambiguous effects, but also because the schemes vary substantially (Albanese et al., 2020). There is also evidence that workers who reduced their hours from full-time to part-time work towards the end of their working lives retired significantly earlier than those who continued to work full-time. Most of the literature does not consider spillover effects within the household, René Böheim

30  Elgar encyclopedia of labour studies

although there are theoretical reasons and empirical evidence for spouses deciding to retire at the same time (Hamermesh, 2002). While there are theoretical models of joint retirement decisions, there is limited and mixed empirical evidence on the effect of pension reforms on couples’ joint retirement decisions. This is clearly an area where more evidence would be helpful to inform plans for pension reforms. René Böheim

References

Albanese, A., B. Cockx and Y. Thuy (2020), ‘Working time reductions at the end of the career: Do they prolong the time spent in employment?’, Empirical Economics 59 (1), 99–141. Ayuso, M., J.M. Bravo, R. Holzmann and E. Palmer (2021), ‘Automatic indexation of the pension age to life expectancy: When policy design matters’, Risks 9 (96), 1‒28. Baumberg Geiger, B., R. Böheim, and T. Leoni (2019), ‘The growing American health penalty: International trends in the employment of older workers with poor health’, Social Science Research 82, 18–32 Böheim, R., T. Horvath, T. Leoni and M. Spielauern (2021), ‘The impact of health and education on labor force participation in aging societies — projections for the United States and Germany from a dynamic microsimulation’, NBER Working Paper 29534. Cambridge, MA: National Bureau of Economic Research.

René Böheim

Böheim, R. and T. Leoni (2018), ‘Sickness and disability policies: Reform paths in OECD countries between 1990 and 2014’, International Journal of Social Welfare 27 (2), 168–185. Burkhauser, R.V. (1979), ‘The pension acceptance decision of older workers’, Journal of Human Resources 14 (1), 63–75. Dorn, D. and A. Sousa-Poza (2010), ‘“Voluntary” and “involuntary” early retirement: An international analysis’, Applied Economics 42 (4), 427–438. Duggan, M., P. Singleton and J. Song (2007), ‘Aching to retire? The rise in the full retirement age and its impact on the social security disability rolls’, Journal of Public Economics 91 (7–8), 1327–1350. Haan, P. and S. Tolan (2019), ‘Labor supply and fiscal effects of partial retirement – The role of entry age and the timing of pension benefits’, Journal of the Economics of Ageing 14, 100–187. Hamermesh, D. (2002), ‘Timing, togetherness and time windfalls’, Journal of Population Economics 15, 601–623. Hengel, K.O., C. Riumallo-Herl, J. Schram, D. Nieboer, A. van der Beek and L. Burdorf (2021), ‘Effects of changes in early retirement policies on labor force participation: The differential effects for vulnerable groups’, Scandinavian Journal of Work, Environment & Health, 47(3), 224–232. Knell, M. (2021), ‘Actuarial deductions for early retirement’, Journal of Demographic Economics 87(2), 141–167. OECD (2019), Pensions at a Glance 2019: OECD and G20 Indicators. Paris: OECD Publishing. https://​doi​.org/​10​.1787/​b6d3dcfc​-en.

Employee ownership

401Ks can be considered another form of employee ownership, as an investment option for employee participants in 401Ks includes the stock of the firm for which they work. Unlike ESOPs, however, employees make contributions to 401Ks, often with matching contributions from the firm they work for. KSOPs are a hybrid of 401Ks and ESOPs, where the firm uses ESOPs to fund its matching contributions in stock to 401Ks. While there are not as many firms with 401Ks or KSOPs as those with ESOPs, in terms of the total value of employee-owned stock and the total number of employee participants, 401Ks and KSOPs are five times larger than ESOPs. Similar to ESOPs, 401Ks and KSOPs have tax benefits (Blasi et al., 2005; NCEO, 2019). Under differed profit sharing plans, the firm links the amount of bonus to profitability and pays the profit sharing bonus in company stock instead of cash. This form of employee ownership has become less prominent among US firms. An employee stock purchase plan (ESPP) is an optional program in which employees participate voluntarily. It is used by the firm to promote employee ownership by offering employees an option to purchase its own stock, often at a discounted price. All shares purchased through ESPPs are individually and directly owned and freely traded like any other shares, except that employee shareholders enjoy a tax benefit by keeping all shares purchased through ESPPs for a year (Blasi et al., 2005; NCEO, 2019). Broad-based stock option plans are standard stock options, with one exception. Unlike traditional stock options which are mostly for executives, broad-based stock option plans include at least 50 percent of the firm’s non-executive employees. As in the case of traditional stock options, all shares owned by employees as a result of exercising the option are owned individually and traded with little restriction. There are also a number of different forms of employee stock ownership in the United Kingdom (UK). Aside from rare majority employee-owned ESOP firms, the most common forms of employee stock ownership in the UK are direct ownership plans in minority employee-owned firms in which employee own stock individually rather than collectively in a trust. There are three option-based plans: Save As you Earn (SAYE), Company Share Option Plans (CSOPs), and the most recently introduced Enterprise Management

Firms around the world have been experimenting with new and innovative management practices in recent decades. They include shared capitalism schemes which link the financial well-being of workers to group performance such as firm performance, as well as employee participation in decision making in the workplace and sometimes even at the top corporate level (Freeman and Kato, 2019). Shared capitalism schemes range from employee ownership (arrangements in which a company’s employees own shares in the company’s stock) to profit sharing (at least part of the compensation for employees is dependent on firm performance, typically profit) and various team-incentive pay (at least part of the compensation for employees is linked to team performance). The objective of this entry is to describe employee ownership (EO), a chief example of shared capitalism, and its effects on firm performance and worker outcomes.

Forms of employee ownership

There are diverse forms of EO around the world. In the United States (US), there are six primary forms of EO: Employee Share Ownership Plans (ESOPs), 401Ks, Japanese share ownership plans, deferred profit sharing plans, employee stock purchase plans (ESPPs), and broad-based stock option plans. ESOPs are the most popular form (the majority of US firms with at least one of those six forms of employee ownership use ESOPs according to Blasi et al., 2005). The firm creates an ESOP trust, and contributes its own stock or cash to purchase it to the firm-controlled trust which holds stock for its employees until they retire. There is no direct contribution to the trust from employees, although when the firm provides its employees with an ESOP in exchange for a wage concession, lost wages can be considered indirect contributions to the trust from its employees. The contributions to the ESOP trust by the firm are tax-deductible, and this tax advantage becomes greater if the firm borrows money to make such contributions to the ESOP trust (leveraged ESOP). In addition, there are a number of additional tax benefits (NCEO, 2019). 31

32  Elgar encyclopedia of labour studies

Incentives (EMIs). In addition, there are the recently introduced Share Incentive Plans (SIPs), which are similar to ESPPs in the US. All four plans have tax benefits, and participation in any of these plans is voluntary. SAYE combines a stock option feature with an employee saving plan, and if deciding to exercise the option, the participating employee uses what they have saved for the plan till the exercise date (typically three to five years) to purchase company stock, typically with a 20 percent discount on the market price on the day of the purchase. CSOPs and EMIs are more traditional option plans with no saving component, and EMIs are aimed at smaller firms (Bryson and Freeman, 2010; Pendleton, 2011).

The Japanese Employee Share Ownership system

In Japan, till 1997 there was only one form of employee ownership, Jyuugyoin-mochikabu-sei, which we refer to as JESO (Japanese Employee Share Ownership) hitherto. Several new forms of employee ownership such as Nihon-ban ESOP (Japanese ESOP) and stock options have been introduced in recent years. However, they are still in their infancy, and JESO remains the dominant form of employee ownership in Japan (Kato et al., 2019). The firm voluntarily establishes a JESO plan for its non-executive employees by forming a JESO union (called mochikabukai). Unlike the US and the UK, there is no tax incentive for the establishment of JESO. Employees voluntarily participate in their firm’s JESO plan by joining its JESO union (executives are not eligible). This voluntary nature of JESO plans makes employee ownership in Japan similar to ESPP in the US, and most plans in the UK. As a way to induce individual employees to partici­pate in JESO plans, Japanese firms offer subsidies, typically the firm matching each employee’s contribution by giving 5‒10 percent of the contribution as well as bearing ad­ministrative costs­. While individual participants’ shares (and divi­dends) in JESO plans are held collectively in JESO unions, each par­ticipant has a right to withdraw the shares in round lots from their JESO plan, and share with­ drawals are pri­ vately owned. While members may freely exit com­ pletely from JESO unions, re‑entry is restrict­ ed. Upon retirement, model rules adopted by most Takao Kato

JESO plans require retiring workers to exit completely from JESO unions, and withdraw all of their shares. Such withdrawn shares are owned privately and thereby can be sold freely at the prevailing market price. As such, JESO plans are not intended and designed for employees to engage in short-term repeated trading of company stock. They can be viewed as supplementary retirement schemes. Finally, each JESO union selects the general director (rijicho), on a one-par­ticipant, one-vote basis. The general director represents stock­holders in its JESO union. At the general meeting of share­holders, the general director votes the stock held by the JESO union, deciding indepen­ dently, rather than by tabulating votes of employee par­ticipants. The general director must be a member of its JESO union and thus is not an execu­tive. JESO’s collective ownership and voting at the general meeting of shareholders, along with restrictions on short-term repeated trading, make Japanese employee ownership differ from ESPP in the US and its UK counterpart SIP, and share some commonality with ESOP in the US and the UK. However, the voluntary participation of employees, and direct contributions to JESO plans by employee participants, make Japanese employee ownership distinct from ESOP, and similar to ESPP and SIP. In sum, JESO is a hybrid between ESOP and ESPP/SIP in the US and the UK. Unlike the US ESOP, JESO is typically not leveraged. However, leveraged JESO appeared in the past decade, although still holding only a tiny share of employee stock ownership. JESO grew re­ markably in Japan during Japan’s rapid growth era and managed to weather Japan’s Great Recession in the 1990s and early 2000s (Kato, 2003). According to the Tokyo Stock Exchange (TSE), the proportion of firms listed on the TSE that use JESO has been stable at around 90 percent over the last decade. Using most up-to-date data on key attributes of JESO plans for a balanced panel of 572 firms provided by the TSE, in terms of the breadth of JESO, the proportion of the labor force in TSE-listed firms with JESO plans who participate in the plans has been on a gradual upward trend, from below 50 percent in early 1990s to over 60 percent in mid-2000, and dropped again to near 50 percent after the financial crisis. Concerning the depth of JESO, in 2009 the average participant owned stock worth close

Employee ownership  33

to 1.5 million yen that constitutes close to 40 percent of the value of total financial asset holdings of the average employee household (according to the 2009 National Survey of Family Income and Expenditure). However, these plans do not own large percentages of company stock. For TSE-listed companies the propor­tion of stock owned by JESO plans has been rising recently, yet it is still around 2 percent (2.09 percent in 2013). In sum, JESO is deep-rooted and widespread among publicly traded firms in Japan, with broad participation (up to 50 percent of employees participating) and considerable stakes (up to 40 percent of their total asset holdings). In theory, the JESO trust can become a vehicle for turning conventional capitalist firms into employee-owned firms. Today the average JESO trust owns only 2 percent of total outstanding shares of the firm. While there are a variety of forms of EO in other parts of the globe, we focus on EO in the US, the UK and Japan in part due to its salience in each country and a rich body of research on EO in those countries.

Employee ownership impacts on the firm

Theoretically, EO can benefit or harm the firm. On the positive side, perhaps most popular is the goal alignment theory: enterprise success is reflected in a higher price of its stock, and thereby wealth gains for employees who own the stock through EO, resulting in the interest of the firm being more aligned with the interest of its employ­ees. Such goal alignment would lead to improved workplace productivity through the following three channels. First, employees whose interest is better aligned with that of the company will exert more discretionary effort in general. Second, more specifically, such workers with well-aligned interest participate more wholeheartedly in various productivity-enhancing activities such as small group activities by employees, and to smoother and less costly collective bargaining (see, e.g., Kato, 2003). Third, with improved goal alignment, a broader range of relational contracts between the company and its labor force will become more feasible. For instance, in order to induce workers to put in sufficient discretionary effort, the amount of compen-

sation for hard work will have to be large, yet the large compensation will raise the firm’s incentive to renege on the promise to pay this large compensation. In other words, there is a tension between the power of incentive and the firm’s incentive to renege. Since EO serves as an alternative incentive mechanism by linking the worker’s hard work to their financial well-being, the amount of compensation for hard work will not need to be as large as otherwise, and therefore the firm’s incentive to renege on the promise will be reduced (Baker et al., 1994). Moreover, with EO, the firm may find it more costly to default on the relational contracts, for it provides its labor force with an additional channel to punish the firm: withdraw shares from the trust and sell them on the stock market, or vote against the management proposals at the shareholders’ meetings. In sum, EO can help the firm to develop effective new relational contracts or reinforce the existing ones which encourage workers to put in sufficient effort and remain in the firm to maintain firm-specific human capital, resulting in higher enterprise productivity. Turning to the negative effects of EO, the traditional literature argues that EO can weaken the residual claimant status and incentive of managers, and make it difficult for managers to take actions to improve firm performance, such as wage cuts, lay-offs, or reorganization (Jensen and Meckling, 1976). In addition, EO may result in more managerial entrenchment, for employee owners are insider owners, and thereby in principle managers and employee owners can form an insider coalition against the shareholder interest. Such insider coalition against the outside shareholders may lead to insider entrenchment and deteriorating firm performance.

Evidence

Whether EO benefits or harms the firm depends on the relative size of the aforementioned positive effect compared to the negative effect, and ultimately it is an empirical question. There is a rich body of evidence on this question (for reviews of the literature, see Bloom and Van Reenen, 2011; Jones, 2018). Early studies use large representative surveys of firms/establishments, and provide cross-sectional estimates on the relationship between the incidence of EO and firm/establishment productivity. Takao Kato

34  Elgar encyclopedia of labour studies

The subsequent studies make use of their newly acquired access to firm/ establishment-level panel data and attain fixed effect estimates to demonstrate that earlier correlational evidence from cross-sectional data does not simply reflect an association between the incidence of EO and unobserved characteristics of organizations (for example, managerial quality), and that EO may have a causal relationship with productivity. Jones and Kato (1995) provide the first rigorous fixed-effect estimates on the productivity effect of EO by using a longitudinal data of publicly traded firms in Japan: the introduction of EO results in a 4‒5 percent increase in enterprise productivity, yet the payoff takes 3‒4 years. More recent studies identify a set of conditions under which the positive productivity effects of EO become more eminent: see Kato and Morishima (2002) for the importance of employee involvement programs used in conjunction with EO; see Kim and Ouimet (2014) for the significance of firm size measured by the number of employees as a negative moderating factor, as interpreted as the relevance of the free-rider problem. Overall, most studies point to the positive effect of EO on productivity. The existing evidence on the effect of EO on profitability is less conclusive, for EO tends to raise labor cost (for example, wages) and other costs of production. At least there is little evidence suggesting that EO harms firm profitability (Kato et al., 2019). Most prior studies on the effect of EO on firm performance focus on productivity and profitability. Promising areas of future research include the effect of EO on other firm outcomes such as innovation. Turning to the effects of EO on worker outcomes, the literature points to positive effects of EO on worker outcomes such as pay and benefits, voice, training, and employment stability; while the existing evidence on the effect on job satisfaction is less sanguine (Kruse et al., 2010; Kurtulus and Kruse, 2018). As in the case of the effect of EO on enterprise productivity, Kim and Ouimet (2014) provide evidence that the positive effect of EO on pay is more pronounced for firms with fewer employees and more limited proportions of outstanding shares owned by EO, suggesting again that EO is less efficacious for large firms with many employees, in which the free-rider problem is likely to be more acute. Finally, in light Takao Kato

of the rising wealth inequality in the US and other advanced market economies (Saez and Zucman, 2020; Alvaredo et al., 2017), there is growing interest in EO as a potentially effective channel through which otherwise wealth-poor employees accumulate wealth (see, e.g., Kato and Sorresso, 2022). Takao Kato

References

Alvaredo, F., L. Chancel, T. Piketty, E. Saez, and G. Zucman (2017), “Global inequality dynamics: New findings from WID.world,” American Economic Review, 107(5), 404–409. Baker, G., R. Gibbons, and K.J. Murphy (1994), “Subjective performance measures in optimal incentive contracts,” Quarterly Journal of Economics, 109(4), 1125–1156. Blasi, J., D. Kruse, J. Sesil, and M. Kroumova (2005), “An assessment of employee ownership in the United States with implications for the EU,” International Journal of Human Resource Management, 14(6), 893–919. Bloom, N. and J. van Reenen (2011), “Human resource management and productivity,” in O. Ashenfelter and D. Card (eds), Handbook of Labor Economics. Amsterdam: Elsevier, 1697‒1767. Bryson, A. and R.B. Freeman (2010), “How does shared capitalism affect economic performance in the United Kingdom?,” in D.L. Kruse, R.B. Freeman and J.R. Blasi (eds), Shared Capitalism at Work: Employee Ownership, Profit and Gain Sharing, and Broad-Based Stock Options, NBER Conference Report series. Chicago, IL, USA and London, UK: University of Chicago Press, 201–224. Freeman, R.B. and T. Kato (2019), “Journal of Participation and Employee Ownership (Jpeo) in the changing world of participative work practices and pay,” Journal of Participation and Employee Ownership, 1(1), 2–3. Jensen, M.C. and W.H. Meckling (1976), “Theory of the firm: Managerial behavior, agency costs and ownership structure,” Journal of Financial Economics, 3(4), 305–360. Jones, D.C. (2018), “The economics of participation and employee ownership (Peo): An assessment,” Journal of Participation and Employee Ownership, 1(1), 4–37. Jones, D.C. and T. Kato (1995), “The productivity effects of employee stock-ownership plans and bonuses: Evidence from Japanese panel data,” American Economic Review, 85(3), 391–414. Kato, T. (2003), “The recent transformation of participatory employment practices,” in S. Ogura, T. Tachibanaki, and D. Wise (eds), Labor Markets and Firm Benefit Policies in Japan and the United States. Chicago, IL: University of Chicago Press, 39–80.

Employee ownership  35 Kato, T., H. Miyajima and H. Owan (2019), “Does employee stock ownership work? Evidence from publicly-traded firms in Japan,” IZA Discussion Paper No. 11671 (revised). Kato, T. and M. Morishima (2002), “The productivity effects of participatory employment practices: Evidence from new Japanese panel data,” Industrial Relations: A Journal of Economy and Society, 41(4), 487–520. Kato, T. and G. Sorresso (2022), “Shared capitalism and employee wealth: New evidence,” Paper presented at 2022 Kelso Workshop, Rutgers, January 15–16. Kim, E.H. and P. Ouimet (2014), “Broad-based employee stock ownership: Motives and outcomes,” Journal of Finance, 69(3), 1273–1319. Kruse, D.L., R.B. Freeman, and J.R. Blasi (2010), “Do workers gain by sharing? Employee outcomes under employee ownership, profit sharing, and broad-based stock options,” in D.L. Kruse, R.B. Freeman, and J.R. Blasi (eds), Shared Capitalism at Work: Employee Ownership, Profit and Gain Sharing, and

Broad-Based Stock Options, NBER Conference Report series. Chicago, IL, USA and London, UK: University of Chicago Press, 257–289. Kurtulus, F.A. and D.L. Kruse (2018), “An empirical analysis of the relationship between employee ownership and employment stability in the US: 1999–2011,” British Journal of Industrial Relations, 56(2), 245–291. NCEO (2019), Annual Report. Oakland, CA: National Center for Employee Ownership. Pendleton, A. (2011), “Employee ownership in Britain: Diverse forms, diverse antecedents,” in E.J. Carberry (ed.), Employee Ownership and Shared Capitalism: New Directions in Research, Labor and Employment Relations Association Series. Ithaca, NY, USA and London, UK: Cornell University Press, ILR Press, 311–339. Saez, E. and G. Zucman (2020), “The rise of income and wealth inequality in America: Evidence from distributional macroeconomic accounts,” Journal of Economic Perspectives, 34(4), 3–26.

Takao Kato

Employer and employee learning

from the previous model: human capital accumulation and job assignment. Specifically, human capital is a linear function of experience, and there are several possible jobs such that productivity at higher level jobs depends more strongly on ability. Each worker has an unknown innate ability, and his productivity depends on effective ability, defined as a function of innate ability and experience, which are assumed to be complementary; thus, the rate at which human capital accumulates with experience is greater for employees of greater ability. This complementarity implies that there will be a positive serial correlation in wage increments. The model also implies that wages may decrease over time and that there can be demotions, but both pay reductions and demotions should be rare compared to job increases and promotions. In this model employees are paid according to their expected effective ability, which in turn depends on the expected innate ability and experience. On the one hand, expected innate ability may increase or decrease depending on the learning, but on the other hand human capital increases with experience. Hence a reduction in expected ability would have to be greater than the increase in productivity due to human capital accumulation in order to produce a wage reduction or a demotion. The model also implies that wage increases upon promotion should be large: first, because productivity at higher-level jobs is more sensitive to ability; and second, because employees who are promoted have greater ability than those who are not. These models explain some empirical findings, specifically the fact that wages raise with experience, the high wage increases that are associated with promotions, the positive serial correlation in wage increments, and the fact that wage dispersion increases over time. However, some of these findings can also be explained by alternative models: thus, the tournament model and the model of promotion as signals also predict high wage increases associated with promotions (Waldman, 2012), and the positive serial correlation of wage increments is also consistent with a model of human capital accumulation with perfect information about ability (Gibbons and Waldman, 1999b).

Some individual characteristics that influence employee productivity cannot be perfectly observed, and require time to be known. As an employee’s career evolves, a number of work outcomes, for example, employee performance, can be observed and initial uncertainty can to some extent be overcome. For this reason, models that take learning into account are useful to understand some labor issues, such as wage dynamics, labor contract design, or job assignment decisions. Research on learning can be classified into three groups: learning and wage dynamics; learning and employees’ incentives; and asymmetric learning.

Learning and wage dynamics

The first group are models whose aim is to explain wage dynamics. Harris and Holmström (1982) proposed a symmetric learning model assuming that individual productivity depends on the employee’s ability, which is unknown to him and to the labor market (that is, other potential employers). The employee and the market have beliefs about ability which are modelled as a random variable. All firms are identical and risk-neutral, and employees are risk-averse. It is also assumed that firms compete to hire employees, and that employees cannot commit to stay with a given employer. The optimal contract exhibits downward wage rigidity: from one period to the next, the firm will keep the salary constant unless the beliefs about the worker’s ability imply that their expected ability is greater than in the previous period. Downward wage rigidity implies that firms offer partial insurance to their employees, and the reason why full insurance – that is, a constant wage — is not optimal is the employees’ lack of commitment. The model also implies that wage dispersion increases with experience, and that wages increase with experience even after controlling for productivity. Gibbons and Waldman (1999a) proposed an alternative model in which employees are risk-neutral, with two features that are absent

36

Employer and employee learning  37

Learning and employees’ incentives

The second group of papers study the effect of learning on employees’ incentives. The “career concerns” model (Holmström, 1999) assumes that the employer and the employee are risk-neutral, and the employee’s productivity is a separable function of ability and effort. The market can observe the employee’s performance but cannot tell which part is due to effort and which part is due to ability. At the beginning of each period employers offer a salary equal to the employee’s expected productivity based on all the performance information observed up to that point. The employee then decides how much effort to supply, and this, together with their ability and other factors that are beyond their control, determines his performance for the period. In this model incentives are implicit and dynamic. At the beginning of each period the market has some beliefs about the employee’s ability and these prior beliefs are updated at the end of the period once the employee’s performance is observed. If the employee works harder they can lead the market to believe that their ability is greater, which will translate into a higher salary in the following period. The rate at which the market updates its beliefs determines the employee’s incentives: when the market updates at a high rate the employee will have a strong incentive to exert effort because an increase in effort will greatly change market beliefs and therefore their wage in the following period. However, when the market updates at a low rate, incentives are weaker. When employee ability is constant, beliefs become stronger over time as more information about performance is received, hence the rate of updating falls over time and so do employees’ incentives. The model therefore predicts that employees’ incentives will change along their career. When firms are allowed to offer explicit incentives such as a performance-based bonus the optimal contract is one in which explicit incentives increase over time to compensate for the reduction in implicit incentives (Gibbons and Murphy, 1992). Learning has also been studied in the context of tournaments: it has been shown that in a multi-stage contest a bias that favors the winner leads to an increase in incentives (Meyer, 1992). Learning is also important to

understand the effect of feedback on worker performance: employees can use the information they receive to update their beliefs about the marginal costs and benefits of their effort, leading to changes in their effort choices and their performance (Awaysheh et al., 2021). Empirical research on learning and incentives is small. There is empirical evidence indicating that the pay‒performance sensitivity of chief executive officers (CEOs) increases as CEOs approach retirement (Gibbons and Murphy, 1992), evidence consistent with biases in multi-stage tournaments (Barmby and Bridges, 2003), and evidence that employees’ reactions to feedback are consistent with learning (Awaysheh et al., 2021).

Asymmetric learning

The third group of studies has been interested in the asymmetric nature of learning, in particular in the fact that an employer has better information about its employees than other potential employers. The employer does not have an incentive to disclose this information to other employers, as this may imply losing its best employees or being forced to increase their salaries to retain them. Employees would want to disclose their ability if this helps them to improve their salaries, but it is difficult to disclose one’s ability in a credible way. A credible way to convince alternative employers of their ability is to show that they have been promoted. An employee’s job title and responsibilities are easier to verify than their ability, and being promoted to higher-level jobs would be an indication of having greater ability than other employees who did not get the promotion. However, if employers know that promotions can be used by employees to find better employment opportunities, they will be aware that promoting an employee discloses information to alternative employers, and therefore forces them to increase salaries to match potential outside offers. Waldman (1984) shows that this will lead to “under-promotion”: that is, some employees who would be promoted in the absence of this signaling effect will not be promoted. This distortion will be lower when the promotion discloses less information to the market. This will be the case, for instance, when employees have observable characteristics that are positively correlated with ability, Jaime Ortega

38  Elgar encyclopedia of labour studies

such as high levels of education (Bernhardt, 1995). For similar reasons, wage increases upon promotion are lower for employees with more education, and the relationship between promotion wage increases and education must be stronger for first promotions than for subsequent promotions (DeVaro and Waldman, 2012). There is strong empirical evidence consistent with these predictions (Waldman, 2012). Jaime Ortega

References

Awaysheh, A., R. Bonet, and J. Ortega (2021), “Performance feedback and productivity: evidence from a field experiment,” Mimeo. Barmby, T. and S. Bridges (2003), “An analysis of the theory of biased contests in an organisational setting,” International Journal of Manpower, 24(7), 836–843. Bernhardt, D. (1995), “Strategic promotion and compensation,” Review of Economic Studies, 62(2), 315–339. DeVaro, J. and M. Waldman (2012), “The signaling role of promotions: Further theory and empirical evidence,” Journal of Labor Economics, 30(1), 91–147.

Jaime Ortega

Gibbons, R. and K.J. Murphy (1992), “Optimal incentive contracts in the presence of career concerns: Theory and evidence,” Journal of Political Economy, 100(3), 468–505. Gibbons, R. and M. Waldman (1999a), “A theory of wage and promotion dynamics inside firms,” Quarterly Journal of Economics,  114(4), 1321–1358. Gibbons, R. and M. Waldman (1999b), “Careers in organizations: Theory and evidence,” in R. Gibbons and J. Roberts (eds), The Handbook of Organizational Economics Vol. 3. Princeton University Press, 2373–2437. Harris, M. and B. Holmström (1982), “A theory of wage dynamics,” Review of Economic Studies, 49(3), 315–333 Holmström, B. (1999), “Managerial incentive problems: A dynamic perspective,” Review of Economic Studies, 66(1), 169–182. Meyer, M.A. (1992), “Biased contests and moral hazard: Implications for career profiles,” Annales d’Economie et de Statistique, 25/26, 165–187. Waldman, M. (1984), “Job assignments, signalling, and efficiency,” RAND Journal of Economics, 15(2), 255–267. Waldman, M. (2012), “Theory and evidence in internal labor markets,” in R. Gibbons and J. Roberts (eds), The Handbook of Organizational Economics Vol. 3. Princeton University Press, 520–572.

Employer search

ginal costs and marginal benefits of search. If the wage offer is higher than the reservation wage, it is accepted; if it is lower, the search continues. Stigler (1962) briefly discusses employer search, arguing that there are two reasons for workers searching for a job rather than employers searching for workers. First, it is cheaper for workers to search than for employers:

Introduction

The labor market is dynamic with continuous worker movements, that is, workers losing their jobs and finding new ones, young workers entering the market and old workers leaving the market. Employers are often searching for new workers to replace quits or to expand their business. There is a substantial literature on job search theory and there are many empirical job search studies, in particular on unemployed workers looking for a job. There is far less research, theoretical or empirical, on employer search, that is, employers searching for employees to fill job vacancies. How these vacancies arise is not a topic that will be discussed here. Whether an employer might focus on retaining incumbent workers rather than hiring new workers will not be discussed either. Employer search is not limited to finding unemployed workers willing to work at the going wage, but may also concern employed workers, that is, workers switching between jobs. There are various elements of employer search: how they search for employees, how they select from applicants, and how long the search lasts; that is, the focus is on search channels, decisions on whether an applicant is suitable for the vacant position, and how much time the employer needs to fill a vacancy.

The main reason for workers undertaking the burden of solicitation is that it is cheaper for them than for employers. When an employer has numerous employees the probability that a given employer needs additional workers is much greater than the probability that a given worker will accept a job offer.

Second, the identification problem is usually more simple for a worker than for an employer. Whereas workers search for a high wage, employer search involves “more than the identification of potential workers: they must be processed to a degree set by the personnel practices.” Rees (1966) makes a distinction between extensive and intensive margin of employer search. The extensive margin is about the number of applicants; the intensive margin is about the information on workers who applied for the vacant job. The more heterogeneous workers and vacancies are, the more important is the intensive margin. With a more homogeneous labor market the extensive margin is more important. In a heterogeneous labor market it does not make sense to generate many applications, but it is wiser to invest in acquiring knowledge about the applicants. As Rees puts it: “Variation in the quality of applicants in many dimensions is one reason why employers invest so much in the selection of new employees.” According to Rees, employers do not want to get in touch with the largest possible number of potential applicants, but with the few applicants promising enough to be worth the investment of thorough investigation. Rees also mentions that employers have a strong preference for using informal information networks, for example because these provide good screening by employees who are satisfied with their present job. Lipmann and McCall (1976) present an early survey on the theory of job search in which they also discuss the theory of

Theory of employer search

Search theory in the labor market starts with Stigler (1962) focusing on job search. He presents a theory in which a job seeker determines in advance the optimal numbers of draws from a wage distribution, and after obtaining this number the highest wage is chosen. The optimal number of draws is where the marginal costs of doing one more application is equal to the marginal benefits in terms of expected wage increase because of that additional application. The non-sequential job search model has faded away in favor of the sequential job search model, which has an empirical component in the analysis of duration of search, that is, unemployment duration. In this theory, job seekers sample from a wage offer distribution and every sample is assessed immediately by comparing the wage offer with a predetermined reservation wage which balances mar39

40  Elgar encyclopedia of labour studies

employer search. Like workers search for jobs, employers look for workers to fill their job vacancies. The elementary employer search model mirrors the sequential job search model. For given wage offers, employers searching for new employees face a distribution of marginal products. While wage rates constitute an unambiguous market signal to the job searcher, the productivities of job searchers are much more difficult to transmit to searching employers. In order to learn about an individual’s productivity, employers have an incentive to obtain information about characteristics positively correlated with job productivity (for example, performance on aptitude tests, education, training). Obtaining this information is costly in time and money, and it is these costs which limit employer search. In the basic employer search model the optimal search strategy for the employer is to accept an applicant for whom the marginal productivity exceeds some minimally acceptable productivity, and to keep on searching if this is not the case. Basic search theory is not very helpful in understanding the intricacies of employers searching and finding new workers. The sequential model of employer search has some popularity because it fits in a macroeconomic approach to the labor market. There are hazard rate models applied to analyzing vacancy durations, but the number of these studies is tiny compared to unemployment duration studies. It is difficult to imagine that employers search non-sequentially and workers search sequentially. During the selection period in which job applicants are interviewed and perhaps tested, workers are waiting to get an offer. If they do not get a job offer, they have been involved in a time-consuming process that is not modeled in the sequential search approach. If they do get an offer, they are very likely to accept that offer. Of course, it does not have to be the case that all employers or all job seekers search either sequentially or non-sequentially. That may depend on whether it is easy for workers to signal their ability to the employer, and for employers to accurately establish that productivity. So, it could be that part of the labor market is characterized by mutual sequential search and another part by mutual non-sequential search. There is also the issue of whether it actually matters to model search as sequential or non-sequential. The natural answer is: it depends. For macro-models of Jan C. van Ours

the labor market it may be sufficient to use sequential search models, which are elegant and easy to interpret. For a policy-related analysis of employer search it may be more helpful to use the non-sequential model.

Search channels

There are various ways for employers to look for new workers, that is, there is a variety of search channels. There are two types of search channel: formal (for example, advertisements, public employment agencies, the Internet) and informal (for example, former and present employees, social networks). Employers often prefer informal recruitment channels because they provide good initial screening and give potential applicants more information about the job, which may improve potential matches. The type of search strategy that employers use may depend on the search channels and the characteristics of the vacant job. When screening of job applicants is important and search is through formal channels such as advertisements, employers likely use a non-sequential strategy. When screening is less important and more information about the applicant is readily available through informal channels such as employee referral, a sequential search strategy may be used. Since employers often use various search channels simultaneously it is not always clear which search channel was decisive when a new worker is hired. In terms of search channel, employers may also use a sequential strategy. They start using the search channel with highest expected benefits, but if no applicant from the pool meets the minimum requirement they start using the second search channel, and so on.

Empirical evidence

According to Oyer and Schaefer (2011), empirical studies on employer search often focus on vacancies, in particular vacancy durations. Employers are thought to sequentially sample from a pool of potential employees with a known ability distribution. A vacancy duration is the amount of time until a vacancy is filled. Empirical studies typically estimate hazard rates of job filling, examining how these rates vary with characteristics of the job or firm. Van Ours and Ridder (1992) argue that in a sequential search model the rate by which a vacancy is filled is equal to the

Employer search  41

product of the arrival rate of applicants and the probability that an applicant is acceptable. Their results from an empirical study of Dutch vacancy durations and numbers of applicants indicate that employer search is non-sequential. Their main contribution is the investigation of the timing of applications. The arrival rate of applicants is very high for the first two weeks, essentially zero for the next two weeks, thereafter rising again. The acceptance probability is zero for the first two weeks after the vacancy’s posting. Employers spend far more time on selection than on search. Thus, a vacancy duration should not be thought of as the duration of search, but rather as a combination of search duration and screening duration. Employer search is likely to be non-sequential, whereby a pool of applicants is formed just after the vacancy has been posted, and employers select a suitable applicant from this pool. So, employers do not play a waiting game but simply choose the best available applicant. A companion study, Van Ours and Ridder (1991), concludes that job requirements may be used as an instrument to control the duration of the vacancy. Employers may lower the job requirements in terms of education and experience if there is no applicant that satisfies them. They find in their sample of Dutch job vacancies that one in four vacancies are filled by workers who do not meet the original requirements stated in the vacancy.

Internet search

The Internet allowed online job posting to grow spectacularly. According to Autor (2001), there are two types of information about workers’ attributes: low bandwidth varieties (education, experience, salaries, credentials), and high bandwidth varieties (quality, motivation, and fit). The Internet makes information about low bandwidth varieties cheap, and has thus lowered search costs and application costs. Lower costs induce many more applications, but do not reduce the problem of gathering high bandwidth information. Freeman (2002) speculates that Internet search and recruitment will produce better matches because posting vacancies for jobs on the Web is much lower in cost for employers and provides a rapid response.

Oyer and Schaefer (2011) argue that it is easier for workers to apply for many vacancies simultaneously. This implies that job seekers will self-select less efficiently because the costs of applying for vacancies that they themselves find less suitable are sufficiently low. For employers this means that they have to invest more in screening. Hoffman et al. (2018) find that pre-employment online testing can improve the match quality and tenure of newly hired workers. They also find that managers overruling algorithmic recommendations end up with worse outcomes. This does not rule out the possibility that algorithmic search and selection leads to unconscious discrimination in hiring practices, as not all applicants can provide sufficient accurate information about their productivity.

Further reading

DeVaro and Gürtler give a fascinating historical account of the way employers and job seekers found each other through advertising. They distinguish four eras in advertising history. In the pre-advertisement era, from the early seventeenth century until the nineteenth century, newspaper ads posted by either employers or job seekers were rare. In the early advertisement era, starting in the nineteenth century, ads began to appear regularly, predominantly posted by job seekers. In the modern era, in the twentieth century, ads were increasingly posted by employers. In the current Internet era, electronic ads can be posted at low cost, resulting in a significant volume of online posting by both sides of the labor market. Oyer and Schaefer (2011) provide a nice overview of relevant elements of employer search. They argue that the fundamental economic problem in hiring is well understood, but the methods that firms use to solve hiring problems still need a lot more research. Jan C. van Ours

References

Autor, D.H. (2001), “Wiring the labor market,” Journal of Economic Perspectives 15 (1), 25–40. DeVaro, J. and O. Gürtler (2018), “Advertising and labor market matching: A tour through the times,” Journal of Labor Economics 36 (1), 253–307.

Jan C. van Ours

42  Elgar encyclopedia of labour studies Freeman, R.B. (2002),”The labor market in the new information economy,” Oxford Review of Economic Policy 18 (3), 288–305. Hoffman, M., L.B. Kahn, and D. Li (2018), “Discretion in hiring,” Quarterly Journal of Economics 133 (2), 765–800. Lipmann, S.A. and J.J. McCall (1976), “The economics of job search: A survey; Part I: Optimal job search policies,” Economic Inquiry 14 (2), 155–189. Oyer, P. and S. Schaefer (2011), “Personnel economics: Hiring and incentives,” in O. Ashenfelter and D. Card (eds), Handbook of Labor Economics, Volume 4B. Elsevier, 1769–1823.

Jan C. van Ours

Rees, A. (1966), “Information networks in labor markets,” American Economic Review 56 (1/2), 559–566. Stigler, G.J. (1962), ”Information in the labor market,” Journal of Political Economy 70 (5), 94–105. Van Ours, J.C. and G. Ridder (1991), “Job requirements and the recruitment of new employees,” Economics Letters 36, 213–218. Van Ours, J.C. and G. Ridder (1992), “Vacancies and the recruitment of new employees,” Journal of Labor Economics 10 (2), 138–155.

Employment protection legislation impacts

employers and employees less cooperative than would be desirable, by ensuring the revelation of relevant, productivity-enhancing information. It follows, then, that a number of interventions in the labor market can indeed be viewed as an efficient response to the presence of some market failure, and in that sense a given institutional-economics program is a test on how successful institutions can be viewed in solving a resource-allocation problem. Empirical evaluation of the employment effects of employment protection is therefore crucial, in particular because economic theory has not been able to generate unambiguous predictions. Given that employment protection is alive and well everywhere, perhaps the implication is that the rule of employment-at-will has not proved to be ex post efficient.

Introduction

Employment protection can be seen as a set of legal requirements that impose a separation cost upon firms wishing to terminate or replace workers. In this sense the implied firing or dismissals costs are restrictions on employers’ ability to utilize labor. Dismissals protection comprises severance payments, advance notice to termination, and a variety of legal procedures, including penalties for wrongful discharges (for example, reinstatement of the employee). More broadly, employment protection legislation (EPL) can cover limitations on the use of fixed-term and temporary agency work, unemployment insurance benefits, worker representation rights and collective bargaining recognition, and minimum wages, as well as labor standards, such as regulations on parental/ maternity leave, teleworking, and health and safety, inter alia. A labor market free of employment protection will be able in principle to signal workers’ true ability through the corresponding market wage and, as a result, constitutes a powerful resource-allocation mechanism. The issue, however, is whether the competitive, employment-at-will solution is capable to protect workers from unjust or arbitrary separation, and whether the distribution of the gains from trade is sufficiently equitable, given that unprotected workers usually do not have sufficient bargaining power. Whether employment-at-will provides adequate insurance for workers against unforeseen labor shocks is also an issue. If there is excessive job turnover (given by the sum of job destruction and job creation), on a purely resource-allocation basis the competitive solution will not constitute an efficiency-improving mechanism, namely where significant relationship-specific investments have been made. The rationale for employment protection may therefore arise from market failures stemming from the need to avoid excessive job (and worker) turnover, as well as the need to assure that incentives to invest in training are adequate. Job security can also be instrumental in reducing asymmetric information problems with the potential to make

Theoretical observations and main predictions

In a world of complete contracts (that is, with no gaps) and low transaction costs, any firing cost arising from, say, a state-mandated severance payment at termination can be undone by a labor contract that sets an appropriate up-front fee to be paid by the worker to the employer. In this setting, employment is unchanged from the pure contract-at-will situation, which means that the introduction of job security will not change the number of jobs in the economy (Lazear, 1990). In turn, an employment contract with severance pay and no initial fee will serve to insure a risk-averse, financially constrained employee who is not willing to pay the fee, fearing that the firm may “take the money and run.” In this case, however, it remains to be seen whether the insurance function is accomplished at a cost of lower employment. In a world of incomplete contracts and pervasive transaction costs (arising from risk and asymmetric information, the latter being truly ubiquitous in all employment relations), the introduction of employment protection is expected to be binding, that is, not to be undone by a contract, in which case employment (and unemployment) effects will not necessarily be neutral (MacLeod, 2011). Employment effects of employment protection can also be discussed on the basis of turnover costs and inter-temporal firm optimization over the business cycle. Given the presence of binding firing (and hiring) 43

44  Elgar encyclopedia of labour studies

costs, it may be optimal to a cost-minimizing firm to reduce labor demand in good times and increase labor demand in bad times. Although it may be difficult in this case to predict the average effect, the absence of strong employment effects of EPL cannot be excluded: it will depend after all on the particular configuration of the labor demand schedule (that is, whether it is flatter or steeper in recessions than in boom times) (Boeri et al., 2015). Even admitting that the employment effects of job protection are approximately zero, there is still the issue of employment composition. If employment protection coverage is incomplete, in the sense that definite-term employment is excluded or less tightly regulated than open-ended employment, the former is likely to be used as the main firm’s mechanism of adjustment to output demand shocks. In this case, there will be an increase in short-term jobs (that is, fixed-term and temporary agency work), and efficiency considerations necessarily arise, as incomplete EPL coverage will entail the risk of generating excess job turnover of temporary employees. This outcome is also undesirable to the extent that too much job competition between the unemployed and the short-term job holders, for a necessarily limited number of job slots, is likely to be generated. Reduced transitions from unemployment to employment, and an increase in long-term unemployment, will therefore be the predicted result (Boeri et al., 2015).

Evidence

Studies using relatively long time-series of cross-country data have provided a good number of estimates of the employment effects of statutory job security. Typically, these studies looked at an extended subset of Organisation for Economic Co-operation and Development (OECD) countries where the employment‒population ratio, the labor force participation rate, and the unemployment rate were easily available, as well as some constructed measure of EPL, most notably the OECD employment protection index (OECD, 2020). After controlling for a number of covariates, such as the population of working age and the growth in per capita gross domestic product, these early studies found a strongly negative association between stringent employment protection and Paulino Maria Freitas Teixeira

employment, and a positive relationship with unemployment. Some estimates suggested that an increase from zero to three months of severance pay would raise the unemployment rate by 5.5 percent in the United States (Lazear, 1990). These results proved very influential, launching many replication exercises that were not always supportive of the original results (Addison et al., 2000). The main critique, however, was related to the fact that the selected measures of dismissals protection were in general too narrowly defined, offering only a partial view of the regulatory apparatus. Another shortcoming was the absence of sufficiently long time-series on variables related to the collective bargaining environment, and indeed subsequent studies found that the adverse effects of employment protection were reduced in countries characterized by either centralized and coordinated or decentralized collective bargaining, compared with nations where sectoral wage bargaining predominates with limited coordination. The analysis based on employment and unemployment stocks showed the immediate need to improve both the data and methods. Changes in flows, rather than in stocks, were as a consequence emphasized in a wave of new studies, while aggregate (country-level) data were increasingly replaced by microeconomic (firm-level) data. The evidence on the impact of job security on labor market outcomes then proved to be more robust, in the sense that they have generated a wide consensus that job and worker flows decrease with employment protection. Accordingly, job creation and job destruction rates were found to be lower in countries with higher job protection, while worker transitions between employment and unemployment were in general less. Unemployment duration was also found to increase with job security, although in this case aspects of the unemployment insurance system (the maximum duration of unemployment insurance benefits and the replacement rate, inter alia) were seemingly relevant. In particular, the combination of low job protection with high unemployment protection was found to be more functional, that is, with the ability to generate more favorable labor market outcomes; while some specific characteristics of the employment protection system were likely to be harmful to labor reallocation, namely costly

Employment protection legislation impacts  45

legal procedures and strict reinstatement rules (Boeri et al., 2015). The empirical analysis of worker and job flows was also crucial for the analysis of issues related to labor market segmentation or dualism. Under stringent job protection, labor market flows tend on average to be smaller; but the impact varies with worker groups. Given that employment protection tends to shelter workers with longer tenure, adjustment of the labor input to negative output demand shocks is made by laying off at the margin, that is, by separating workers with short tenure. The evidence is that countries with a greater index of labor protection do tend to have a higher percentage of temporary workers. The impact of EPL on investment, productivity, and innovation is less clear cut. Although some increase in worker protection—in situations where the employment-at-will doctrine tends to predominate—was found to be innovation-enhancing, the cross-country evidence suggests that EPL is predominantly distortionary, that is, it tends to promote low-risk, low-productivity growth sectors, thus generating a lower aggregate productivity growth (Bartelsman et al., 2016). The crucial aspect here is that risky technologies have on average a higher payoff but at a cost of much higher volatility. If dismissals protection is high and the investment proves unsuccessful, prohibitive dismissal costs will deter firms from investing in high-risk sectors, so that countries with a high employment protection will benefit less from the arrival of new technologies, thus compromising economic growth in the long run. In contrast, firms in countries without strict labor protection are attracted by high-return investments, in a process implying resource reallocation from low-risk to high-risk sectors, thus fueling economic growth given that negative shocks are more easily offset by shedding labor in the presence of lower firing costs. Evidence based on large numbers of countries from the OECD, Latin America, Africa, and Asia, and sectoral (that is, manufacturing) data, also shows that that strict job security hampers the creative destruction process (Caballero et al., 2013). Specifically, in countries with a strong rule of law, moving from the lower to the higher percentiles of the distribution of job security lowers microeconomic flexibility (that is, the speed of adjust-

ment to sector-specific shocks), while annual productivity growth can be cut by as much as 1 percent. The same movement for countries with a weak rule of law, however, seems to have a negligible effect, a result that serves to show that the degree of law enforcement matters. Some further evidence, based on quasi-natural experiments, can be added in this context. For example, it has been found that the introduction of mutual consent in employment termination facilitates labor adjustment and worker reallocation in a regime of strict labor protection and high litigation costs. In contrast, reforms that entail reduction in firing costs only for a specific group of workers (for example, for those whose seniority is lower than a certain threshold) are likely to substantially increase the destruction of low-tenured jobs, for either unskilled or skilled workers, although the impact on the former is expected to be stronger. Recent evidence exploiting long time-series of industry-level data from a group of OECD countries shows that short- and long-run consequences of certain reforms are distinct (Bassanini and Cingano, 2019). Flexibility-enhancing reforms in particular are likely to have immediate effects on job termination, while the contribution to higher flows into employment may be delayed by market frictions and uncertainty. It is estimated in particular that in the year following the typical EPL reform, aggregate employment will be significantly below the pre-reform level. If the reforms entail strong short-run disemployment effects, the implication is that they should not be implemented during a recession. The evidence is that employment costs of deregulation are negligible in good times (Boeri et al., 2015). Finally, the costs and benefits of EPL reforms seem to be dependent on the interaction between cultural, political, and labor market institutions (Algan and Cahuc, 2009). Specifically, there is evidence that countries more attached to civic virtues will tend to rely less on employment security legislation, compared to nations that seem to value them less, with the latter group finding it more costly to protect workers through the alternative—but more prone to cheating— unemployment insurance system. It may also be the case that reforms will be more effecPaulino Maria Freitas Teixeira

46  Elgar encyclopedia of labour studies

tive under more democratic and transparent political institutions.

Some gaps in current knowledge

Perhaps too many studies on the area of employment protection evaluation have shared the same data and methods, thus generating in excess results that are mutually reinforcing. On the other hand, it seems that too much effort has been allocated to the construction of a summary EPL measure that is based on a relatively simplified, if not crude, representation of the law, which does not always use a good understanding of how the law works, and ultimately shows considerable path dependence. After all, court decisions are based on careful review of the evidence, on how the parties have breached a duty. A particular shortcoming has been the inability to measure actual labor costs entailed by changes in regulation (Heckman, 2007). Prices and quantities are informative, and perhaps it would be more profitable to focus on how changes in labor regulations impact labor demand and labor supply schedules. Reverse causation has also been underplayed. It is clear that many protective labor market institutions are caused by an underperforming level of the labor market, not the other way around. Data and methods that fail to recognize this reverse causality will not serve a sound policy recommendation; they only allow prevalence of prior beliefs in the interpretation of the evidence. Employment protection comprises specific rules concerning advance notice, severance pay, and conditions under which a protected worker can be dismissed. There is in particular a large consensus that an increase in employment laws tends to reduce employment for individuals less attached to the labor market. In most cases, however, to put it bluntly, policy recommendations are simply to “reduce them all” (MacLeod, 2011). It would therefore be great progress if more precise, quantitative empirical responses were obtained. Thus, a specific gap to be filled in the future should involve the construction

Paulino Maria Freitas Teixeira

of mechanisms that protect individuals who are badly treated by poor management. This approach will perhaps be more realistic, given that employment law exists to deal with the extreme cases, not the average employment relationship. This seems to be of particular relevance, as the evidence indicates that a job loss is a particularly harmful experience, well beyond the implied pecuniary costs. Paulino Maria Freitas Teixeira

References

Addison, J.T., P. Teixeira, and J.-L. Grosso (2000), “The effect of dismissals protection on employment: More on a vexed theme”, Southern Economic Journal, 67(1), 105–122. Algan, Y. and P. Cahuc (2009), “Civic virtue and labor market institutions,” American Economic Journal: Macroeconomics, 1(1), 111–145 Bartelsman, E.J., P.A. Gautier, and J. de Wind (2016), “Employment protection, technology choice, and worker allocation”, International Economic Review, 57(3), 787–825. Bassanini, A. and F. Cingano (2019), “Before it gets better: The short-term employment costs of regulatory reforms,” Industrial and Labor Relations Review, 72(1), 127–157. Boeri, T., P. Cahuc, and A. Zylberberg (2015), “The costs of flexibility-enhancing structural reforms: A literature review,” OECD Economics Department Working Papers No. 1264. Caballero, R.J., K.N. Cowan, E.M.R.A. Engel, and A. Micco (2013), “Effective labor regulation and microeconomic flexibility,” Journal of Development Economics, 101, 92–104. Heckman, J.J. (2007), “Comments on: Are protective labor market institutions at the root of unemployment?,” Capitalism and Society, 2(1), Art. 5. Lazear, E.P. (1990), “Job security provisions and unemployment,” Quarterly Journal of Economics, 105(3), 699–726. MacLeod, W.B. (2011), “Great expectations: Law, employment contract, and labor market performance,” in: Ashenfelter, O. and D. Card (eds), Handbook of Labor Economics, Vol. 4b, Elsevier, 1591–1696. OECD (2020), “Chapter 3: Recent trends in employment protection legislation,” OECD Employment Outlook, OECD Publishing.

Executive compensation

building. One mechanism to achieve this is via mergers and acquisitions. Third, the CEO may make excessive or unwarranted use of company perks that may not be beneficial for shareholders. This may include use of corporate aircraft, financial services, or club memberships. Fourth, in the extreme case, moral hazard also includes intentional misappropriation of shareholder funds, including fraud and theft. Enron in 2001 and Theranos in 2018 are examples. The standard agency model is solved by finding an optimal sharing rate between the principal and the agent. This provides the underlying logic for CEO “pay-for-performance” contracts. Agency models predict that executive pay contracts contain stock options, accounting earnings, and other individual and firm performance metrics that signal the CEO effort (Holmström’s “informativeness” principle). Additionally, the sharing rate (pay-for-performance) is predicted to be inversely correlated to the degree of agent risk-aversion. What evidence supports the optimal contracting view, or other shareholder value explanations? First, the models predict that firms compete in the managerial labor market for the best talent. Talented CEOs will be assigned to the largest firms, where their marginal productivity has greatest effect on firm value. Econometric studies show that firm size (measured by firm sales, assets, or market value) is positively and significantly related to CEO pay. In a log-log regression model, an estimated CEO pay size elasticity in the range of 0.3 to 0.5 is not uncommon. Moreover, firm scale explains a large proportion of the variation in compensation (Conyon and Murphy, 2000; Murphy, 2013). Second, is pay related to performance? The literature on this topic is vast. Jensen and Murphy’s (1990) seminal paper showed that CEO wealth increased only $3.25 for every $1000 change in shareholder wealth, suggesting weak incentives to promote shareholder value. CEOs were paid like “bureaucrats.” The reason was a lack of equity ownership, such as stock options or restricted stock. This changed in the 1990s as boards granted vast amounts of options to managers. Hall and Liebman (1998) were the first to show a strong positive correlation between CEO compensation and firm performance, arising almost entirely via changes in the value of CEO holdings of stock and stock options. The

Introduction

Executive compensation is a complex and controversial issue. Two dominant views exist. The optimal contracting view asserts that shareholders design chief executive officer (CEO) compensation contracts to attenuate CEO opportunistic behavior (moral hazard). Conversely, the rent extraction view asserts that CEO pay outcomes are decided by CEOs themselves to maximize the amount they can extract without receiving intervention from owners. Corporate scandals, frauds, and financial crises have led to increased scrutiny of CEO pay. The Sarbanes‒Oxley Act (in 2002) and Dodd‒Frank (in 2008) were introduced to correct the perceived flaws in the governance system. This entry provides a brief overview of executive compensation. Extensive and compelling surveys are provided in Edmans et al. (2017) and Murphy (1999, 2013). It first discusses the optimal contracting and rent extraction perspectives, and then provides empirical evidence on the level and composition of United States (US) executive pay practices, before highlighting some open questions.

Theories of executive compensation The optimal contracting view The economics of executive compensation is motivated by the principal‒agent model (Holmström, 1979; Holmström and Milgrom, 1987). It posits a rational risk-neutral shareholder (maximizing profits) and risk- and effort-averse CEO (maximizing utility). Shareholders (the “principal”) delegate decision making authority to the CEO (the “agent”). A linear contract is offered to the agent. The asymmetry of information in the model is moral hazard, which is not perfectly observable at zero cost by shareholders. Moral hazards are pervasive. First, the CEO can enjoy the quiet life by picking easy-to-manage tasks. A risk-averse CEO may avoid undertaking risky projects with positive net present values, especially if the ex post outcomes of such projects turn out to be poor and lead to termination. Second, the CEO may engage in self-interested empire 47

48  Elgar encyclopedia of labour studies

granting of equity compensation has continued beyond the 1990s (see below). There is a debate on how to measure incentives empirically. One approach is the CEO’s effective ownership stake in the firm (Conyon and Murphy, 2000), measured as the total amount of stock and other equity claims expressed as a percentage of common shares. This number is typically low. Edmans et al. (2017) report that the CEO’s effective percentage ownership between 2000 and 2005 in a sample of large companies was 0.11 percent. Another measure is the CEO’s effective dollar ownership in the firm. This is the product of the effective percentage ownership and the firm’s equity market capitalization. Even if a CEO’s percentage ownership stake is small, their dollar value incentives are typically not, especially in large firms. The typical CEO can gain millions of dollars from only small improvements in firm performance. Or lose millions for small performance loss. Edmans et al. (2017) report that the CEO’s effective dollar ownership between 2000 and 2005 was $31.3 million. On this measure, incentives are strong. The rent extraction view The rent extraction view asserts that both the level and the structure of pay are decided by the CEOs and executives themselves (Bebchuk and Fried, 2004). Boards are viewed as compliant or complicit in the rent extraction process. The resulting level and structure of CEO pay is not set in shareholders’ interests, and CEO pay is often excessive. One version of the theory is that CEOs exercise power and influence over their boards and use this to lobby for high pay levels. Directors on boards and their compensation committees have incentives to please the CEO. The CEO might pack the board with their friends and colleagues. They might have shared social interests or sit on each other’s boards. This creates bonds between the CEO and the board members that may be stronger than the ties between the board members and shareholders. The CEO may be able to offer the directors extended tenure on the board. If these cozy arrangements exist, then CEO pay negotiations are not arm’s-length. CEO pay might be too high. The “excess pay” constitutes an economic rent, an amount greater than required for the CEO to provide labor services to the firm. Martin J. Conyon

Corporate boards are viewed as relatively weak compared to the CEO. The issue then turns to: what is a weak board, or one that is prone to capture by the CEO? What constitutes a weak board? Various proxies for non-ideal corporate governance board arrangements have been proposed by academics. First, a board might be weak if it is too large and has too many members. In this situation it is difficult for directors to oppose the CEO. Second, if the CEO has appointed outside directors who are beholden to the CEO for their jobs, then the board might be prone to be captured by the CEO. Third, directors may serve on too many other boards, making them too busy to be effective monitors. Fourth, if the CEO is also chair of the board, then there might be potential conflicts of interest. Does the CEO serve their own best interest (wearing the CEO hat) or do they serve the shareholders’ best interest (wearing the chairperson’s hat)? Fifth, the board may be too friendly with the CEO, coming from the same social or friendship groups, and therefore pay insufficient attention to their fiduciary duties to shareholders. For all these reasons, and others, the board might become weak in relation to the CEO. When boardroom governance is weak, managerial power theories predict that excess pay will result. There are limits on how high CEO pay can be, of course. If CEO pay is far too high, then activist investors might intervene, or the firm might be subject to takeover. Alternatively, the CEO’s reputation is at risk, or they may be caused social embarrassment. Bebchuk and Fried (2004) term this “outrage costs.” What is the evidence for rent extraction? First, CEOs enjoy perks, including corporate jets, club memberships, and personal security. The consumption of these might not be in shareholders’ interests. Second, there is evidence of the practice of stock option backdating prior to 2005. Evidence shows that option grant dates were set ex post to minimize the exercise price of option grants to maximize their value to CEOs. This is not in shareholders’ interests. Third, CEOS are paid for “luck” rather than “skill.” The lack of relative performance evaluation in compensation contracts (for example, options are not indexed) means that value increases just because market prices drift up; and this is not connected to CEO skill. Studies also show that CEOs in oil companies enjoy the

Executive compensation  49

benefits of oil price movements, and CEOs of companies in import-intensive sectors benefit from exchange rate movements. Such variables are not controlled by the CEOs, yet their compensation increases.

Executive compensation: empirical evidence

The US Securities and Exchange Commission (SEC) mandates public companies to reveal considerable information about CEO compensation, from salaries and bonuses to the (expected) value of stock options and stock, to other features in the pay contract such as pension arrangements. The information is contained in the firm’s definitive proxy statement filed with the SEC (the DEF 14A form). The level of CEO compensation Figure 1 shows the level of real CEO pay from 1992 to 2020. The data are for the CEOs of S&P1500 firms. There are two bars per year. The height of the first (pale grey) is the average level of CEO pay, the height of the second (dark grey) is the median level of CEO pay. Total CEO pay is the sum of salary, bonus, stock options, restricted stock and other equity, and other payments

(Execucomp variable TDC1). Options are the grant date value using the Black‒Scholes method. The data show several important facts about CEO pay. First, in each year, average CEO pay exceeds median pay, and often by a significant margin. CEO pay is positively skewed. There are a few highly paid CEOs who pull the average upwards, but the mass of the pay distribution is to the left. Second, real CEO pay both increases and decreases. Average CEO pay increased rapidly from 1992 to 2000: it climbed from about $4 million to approximately $12 million. This was driven by the grants of stock options (see below). During the Internet bubble and subsequent dot.com crash, average CEO pay fell. It then regained momentum during the economic upswing before falling again at the onset of the Great Recession. From 2010 to 2020, average CEO pay has been increasing. The same pattern is observed for median CEO pay, but the peaks and troughs are not as marked. Third, CEO pay is correlated to changes in business cycle activity. This does not imply that every CEO has their pay adequately tied to performance, only that the typical CEO’s compensation changes with economic

Note: CPI inflation-adjusted, 2015 = 100. Source: Data from Execucomp.

Figure 1

The level of CEO compensation

Martin J. Conyon

50  Elgar encyclopedia of labour studies

activity. Nor does it imply a causal relation between performance and pay. The composition of CEO compensation The composition (or mix) of CEO pay refers to how much of CEO annual total pay is delivered in the form of a salary, a bonus, equity compensation (such as restricted stock and stock options) and other pay (such as perks, the imputed cost of aircraft use, and so on). The more executive compensation is delivered in the form of performance-related pay, such as bonuses and equity, the more aligned are shareholders and manager interests. The potential negative externalities or unintended consequences are not explored. Figure 2 shows the composition of CEO pay from 1992 to 2020. The following are noteworthy. First, CEO salary as a percentage of total CEO pay has declined significantly since 1992. In 1992, the typical CEO’s salary accounted for about 40 percent of their total annual pay. By 2020, salary was approximately 20 percent of total pay. In real terms, CEOs typically earn about $1 million per year. Second, bonuses account for about one-fifth of total CEO compensation. The

Figure 2

The composition of total CEO pay

Martin J. Conyon

fraction is stable over time. Typically, bonus payments are triggered by accounting performance measures such as budgeted earnings, rather than market-based measures like shareholder returns. Third, the amount of pay delivered in the form of stock options ballooned during the 1990s, increasing from about 22 percent of total pay in 1992 to about 40 percent in 2000. Options are valued at their expected value using the Black‒Scholes method. This is different from the value realized from the CEO’s sale of stock options. Murphy (2013) attributes the increase to tax and accounting rules that favored stock options over common stock. Generally, because the value of a CEO’s options increases as the underlying stock price increases, this aligns CEO and shareholder interests. Fourth, restricted stock became more important from the early 2000s. Historically, Rule FAS123 in 1995 recommended, but did not mandate, companies to expense stock options using the Black‒Scholes (or similar) method. Few companies did. In 2005, accounting standards changed and FAS123R required companies to expense options using a fair market value method such as

Executive compensation  51

Black‒Scholes. This levelled the playing field between stock options and restricted stock. The evidence shows that restricted stock became a more important compensation vehicle. By 2020, about 50 percent of total compensation is delivered in the form of stock, and less than 10 percent in the form of stock options. Restricted stock has displaced stock options. Fifth, ‘other pay’ (comprising aircraft, country club memberships, and so on) is a small fraction of total annual pay. The figure is about 5 percent of total pay. One implication is that the imputed cost of the CEO’s use of aircraft, country clubs, financial advice, and so on, is a small fraction of total CEO pay. Of course, the dollar amount for any given individual might be high. However, this is partial evidence against a pure rent-extraction story. Sixth, the overall evidence shows a shift towards performance-based compensation and away from fixed pay from 1992 to 2020. CEO salary is the main variable that is not contingent upon an objective performance measure, at least explicitly. As the salary percentage of pay has fallen over time, the importance of performance-based components has increased. This is consistent with the shareholder value view of optimal contracting.

Extensions History History matters. Most studies of CEO pay use data from the last 30 years. Historically, this is a short time period. Frydman and Saks (2010) analyzed longer-run trends in US executive compensation from 1936 to 2005. They find that the median value of real executive compensation was remarkably flat from around the late 1940s until the mid-1970s. This suggests that the correlation between executive pay and firm growth was weak during this time. Second, the authors show that very large increases in CEO pay, sometimes more than 10 percent per annum, have occurred since the mid-1970s, but especially since the 1980s. Prior to this, modest growth in executive pay of 1 percent per annum was the norm. A key takeaway from this analysis, then, is that the growth in executive pay, in the wider historical context, is a compara-

tively recent phenomenon. But during this time in history there was much innovation and economic growth, while at the same time CEO incentives were supposedly weak. What explains this? Culture Culture matters. There is an emerging literature on international executive compensation (Conyon et al., 2011). There are similarities in pay design across countries, but will American pay practice prevail? Conyon et al. (2011) document that the US CEOs command a pay premium relative to European company CEOs. This premium declined from a mean (median) of 200 percent (118 percent) in 1997, to 81 percent (23 percent) in 2003. It disappears if CEO pay is suitably adjusted for the risk of holding more equity-based pay. What about Asia, and especially China? The cultural differences relative to North America and Europe are profound. What are CEO incentives like in Chinese firms, especially state-controlled firms? Conyon and He (2011) show that executive compensation is positively correlated to firm performance, but CEO incentives are lower in state-controlled firms and firms with concentrated ownership structures. They also found that US CEOs enjoy a significant pay premium: US pay is about 17 times higher in the US relative to China. Conyon and He (2011) is an early study, and comparative international studies can help to understand the functioning of high-wage labor markets.

Conclusions

CEO pay is complex and controversial. Two views are dominant in the literature: the optimal contracting view and the rent extraction view. Both have their logic, and both have supporting evidence, as well as unanswered questions. Future research is about designing studies to successfully refute clearly stated hypotheses based on well-grounded theory. Martin J. Conyon

References

Bebchuk, L.A. and J.M. Fried (2004), Pay Without Performance: The Unfulfilled Promise of Executive Remuneration. Cambridge, MA: Harvard University Press. Conyon, M.J., J.E. Core, and W.R. Guay (2011), “Are US CEOs paid more than UK CEOs?

Martin J. Conyon

52  Elgar encyclopedia of labour studies Inferences from risk-adjusted pay,” Review of Financial Studies, 24(2), 402–438. Conyon, M.J. and L. He (2011), “Executive compensation and corporate governance in China,” Journal of Corporate Finance, 17(4), 1158–1175. Conyon, M.J. and K.J. Murphy (2000), “The prince and the pauper? CEO pay in the United States and United Kingdom,” Economic Journal, 110(467), 640–671. Edmans, A., X. Gabaix, and D. Jenter (2017), “Executive compensation: A survey of theory and evidence,” in: B. Hermalin and M. Weisbach (eds), The Handbook of the Economics of Corporate Governance, 383–539. Amsterdam: Elsevier. Frydman, C. and R.E. Saks (2010), “Executive compensation: A new view from a long-term perspective, 1936–2005,” Review of Financial Studies, 23(5), 2099–2138.

Martin J. Conyon

Hall, B.J. and J.B. Liebman (1998), “Are CEOs really paid like bureaucrats?,” Quarterly Journal of Economics, 113(3), 653‒691. Holmström, B. (1979), “Moral hazard and observability,” Bell Journal of Economics, 10(1), 74–91. Holmström, B. and P. Milgrom (1987), “Aggregation and linearity in the provision of intertemporal incentives,” Econometrica, 55(2), 303–328. Jensen, M.C. and K.J. Murphy (1990), “Performance pay and top-management incentives,” Journal of Political Economy, 98(2), 225–264. Murphy, K.J. (1999), “Executive compensation,” in: O. Ashenfelter and D. Card (eds), Handbook of Labour Economics, Vol. 3, 2485–2563. Amsterdam: Elsevier Murphy, K.J. (2013), “Executive compensation: Where we are, and how we got there”, in: G.M. Constantinides, M. Harris, and R.M. Stultz (eds), Handbook of the Economics of Finance (Vol. 2), 211–356. Amsterdam: Elsevier.

F

Firms and wages

tenure or firm size. Second, a time-invariant person effect (or worker/individual effect), θi, which aims to capture individual heterogeneIn the standard (textbook) competitive ity including both observable traits such as labour market model, any wage differentials gender, birth cohort or some components of between individuals can only be the result family background, and unobservable ones of differences in the marginal productiv- such as soft skills, charisma, attitude, perity of the given workers, hence individuals severance, and so on. Hence this parameter are expected to earn the same wage when also acts as a proxy of individual ability or working for any employer. However, recent a measure of worker quality, capturing the evidence suggests a quite substantial role of part of wage determinants that is rewarded firm heterogeneity in the shaping of indi- equally at all firms. Accordingly, if some vidual wages. Although theories deviating workers are systematically discriminated from the baseline model, such as models against at all employers, their wage disadof monopsonistic labour markets and the vantage will be also reflected in the indiconcepts of compensating differentials, or vidual terms. Third, ψj is the time-invariant efficiency wages, already suggested the firm effect aimed to capture firm-specific presence of wage differentials for identical wage premia, that is, the (percentage) wage workers in different firms for some time, differential firm j pays to all of its workers the empirical importance of such features (regardless of their qualities), above or below became evident only over the past two or a benchmark firm of the given market. As three decades. This process was in a large the firm and person effects enter the equapart driven by the increasing availability of tion simultaneously, their estimated values linked employer‒employee panel datasets, can be considered as firm-level averages of which allowed empirical economists not only residual wages being controlled for observed to focus on individuals embedded in firms and unobserved worker composition, and (instead of broad sectors), but also to observe individual-level mean residuals controlled for the universe of firms and the mobility of indi- employment history. Finally, an independent random error term, εijt, is assumed, capturing vidual workers between these firms. the idiosyncratic deviation from the wage level predicted by the above factors. The AKM model Let us assume that we have access to A cornerstone of this shift is without doubt a two-dimensional panel of workers and the seminal paper by Abowd et al. (1999), firms. That is, we can not only observe the often simply referred as AKM, in which wages and other characteristics of employers the authors propose a log-additive statistical or individuals across multiple time periods, model of wage setting of the following form: but we can also observe any mobility of the workers between different employers ​(1)​ (between any two time periods). If indi​In  ​w​ ijtijt​  ​=  ​X​ ijtijt​ β ​+   ​θ​ ii​  ​+   ​ψ​ j​  ​+  ​ ε​ ijtijt​  viduals do move between employers, then In this formulation, the (logarithmic) wage the observed wage change from a given level of individual i at firm j in period t is mod- switch will identify the difference in the firm elled as the sum of the following four additive premia between the sending and the receiving components. First, a covariate index, Xβ, firms. Although there is no need to observe capturing the wage effects of time-varying a direct move between any two employers, person and firm characteristics such as age, firm effects will be comparable only within 53

54  Elgar encyclopedia of labour studies

the same mobility groups, that is, within the same connected sets of workers and firms. In practice, the bipartite network of firms and individuals in most labour markets has a giant component, containing more than 90–95 per cent, or even 99 per cent, of firms, workers or total wage observations. Within this largest connected set, the chain of one-on-one firm comparisons will lead to the comparability of all identified firm effects and ultimately the identification of worker effects (as residuals) as well. However, firm effects across different mobility groups are not directly comparable. Following the above logic, the model is identified and can be estimated even by ordinary least squares, with the observable variables in X and the design matrices – large, sparse matrices defining which individual and firm a given row of observation belongs to – of firm and person effects as right-hand-side variables. Technically, however, this requires the inversion of very large matrices. Although computational power has also increased tremendously in the past decades, fortunately novel econometric techniques have been developed as well, which help in the efficient estimation of multidimensional panel models, even of larger dimensions. In Stata, for instance, a state-of-the-art implementation is provided by Correia (2017) under the command reghdfe. Whether the estimated firm and person effects are meaningfully interpretable (economically) rests on the underlying assumptions of the model specification, namely the independence of the error term, and the additive separability of these effects. Card et al. (2013) present three main threats to the latter assumption, by considering that the error term is the composite of an employer‒employee match component, a person-specific (unit root) drift component, and the idiosyncratic error term. The assumption of exogenous mobility will be violated and the estimated firm and person effects will be contaminated by selection components if any of these three components is correlated with the matrix of employer identifiers. Specifically, we require that individuals do not switch jobs based on either the difference in the match effect components of employers, based on a drift in their person effect – due to human capital accumulation or receiving high outside offers – or based on any transitory shocks to wages at their current and potential employers. Person István Boza

and firm effects, along with time-varying characteristics, may correlate arbitrarily. The additive separability of the model is violated if, for instance, higher-premium firms value high-skill workers more than low-skill ones and hence do not pay a constant premium to all workers. A large set of empirical findings suggests that although these assumptions could be violated in particular data, the biases they cause are often of negligible magnitude, hence the baseline AKM specification proved to be a robust benchmark model in many studies. Still, models that ease some of the assumptions of the baseline model have been developed, as discussed later in this entry. With the above assumptions fulfilled, worker and firm effects can be estimated without bias. However, lower-dimension (quadratic form) summaries of the estimated parameters could still be biased. The main examples include variances and covariances of the estimated person or firm effects, and standard errors of parameters in any linear projections on the vector of these terms. The source of this bias is large estimation error in the linear parameters if the observed mobility (the number of job movers per firm) in the used data is low, for instance, in short panels. Most notably, the correlation of worker and firm effects will be downward-biased, due to an incidental parameter problem: a high (positive) estimation error for an observation in the firm effect comes with a low (negative) error in the person effect. Fortunately, the limited mobility bias problem is now well documented, with multiple bias-correction methods proposed as potential solutions. Notable works include Andrews et al. (2008), Bonhomme et al. (2020) and Kline et al. (2020).

Empirical relevance of firm differentials

Although by now AKM-style person and firm effects have been used in studies from a wide range of fields, they may have had the most influential effect on the literature of wage and earnings inequalities. Along with the natural role of individual diversity in skills, opportunities and ambitions, the heterogeneity of firms’ wage premia – originating either in differences in firm productivity alongside the rent-sharing propensity of firms, or in compensating differentials or efficiency wage schemes – can be an important source of

Firms and wages  55

cent higher sorting shares, with substantially high correlations between individual and firm effects, even in the range of 0.3–0.4. Torres et al. (2018) and Boza (2021) show that direct measures of firm productivity (sales or value added) also correlate with AKM worker effects, allowing for an interpretation of observed wage sorting as the manifestation ​Var(w) ​= Var(θ) ​+  Var(ψ​)​ ​+  Var(Xβ​)​ ​+  Var(ε​)​ ​ of assortative matching in productivity. +  2Cov(θ, ψ​)​ ​+  2Cov(θ,Xθ,Xβ) ​+  2Cov(β) ​ Besides the clear empirical evidence on +  2Cov(ψψ, Xβ)β) ​(2)​ the presence and substantial importance of variation in firm effects, theoretical literAs this decomposition reveals, firms may ature focuses on the mechanisms driving affect the overall wage dispersion by them- the variation. One branch of explanation for selves, Var(ψj ), and through a sorting channel, emerging firm effects relates to compen2Cov(ψj , θi), as well. The complementarity of sating differentials paid by employers for firm-level and individual productivity would disamenities of working at the employer, lead productive firms to hire more productive or the lower wages workers are willing to workers. Assuming that AKM firm effects are accept in the presence of positive amenities. monotonous in firm productivity, this would These non-wage factors could include the lead to the systematic sorting of ‘high-wage’ workplace’s distance from residence, flexiindividuals into ‘high-wage’ firms, and hence bility of working times, natural or personal would be reflected in a positive correlation environment, stress or health risks associated between estimated individual and firm fixed with the given job. Naturally some of these effects of the AKM model; with the contri- differences emerge between industries, but bution to overall inequality being captured in part of them can differentiate firms even the double covariance term. within the same market. A closely related While early empirical studies following factor is the reliance on specific wage-setting the Abowd et al. (1999) approach find no or schemes, such as paying high (efficiency) negative such correlation, the study by Card wages in order to incentivize workers to work et al. (2013) find a strong role of wage sorting hard. Another set of explanations relate to in overall wage dispersion, also showing the observation that more productive firms that both an increase in the dispersion of pay somewhat higher wages to (all) their firm effects and in the correlation between employees. This phenomenon can emerge workers and firm effects contributed signif- either due to some sort of bargaining power icantly to the observed overall increase in of workers in models of job search and of wage inequality in West Germany between collective bargaining, or due to firms facing 1985 and 2009. Many consequent studies upward-sloping supply curves. The latter adapted the same approach and provided could be observed in non-competitive labour evidence on positive wage sorting for a large markets, with at least some monopsonistic set of countries from Western and Northern nature. In such instances, more productive Europe, from the United States and Brazil. monopsonies can only increase their size by At the same time, as bias correction methods hiring workers at increasing wage levels, also became available, the importance of while usually they cannot wage-discriminate the downward bias on the sorting parame- and hence end up paying higher wages to all ter – due to limited mobility in the sample workers. – became evident. Bonhomme et al. (2020) The extent to which productivity differand Boza (2021) both survey findings of such ences translate into wage differences is a key studies, showing that while studies without empirical question in the rent-sharing literaany bias correction find on average around ture. Some novel studies of rent-sharing elas18 per cent direct contribution of firms to ticities actually rely on AKM firm effects to wage dispersion, and around 5 per cent con- solve a major estimation problem regarding tribution through sorting, bias-corrected esti- the pass-through of productivity shocks into mates from the literature include on average wages. Namely, comparing firms of different 6 per cent lower firm shares, and 10 per productivity, standard measures of average wage variation in the economy in itself. Most studies capture this by relying on wage variance decompositions building on the AKM model specification. Due to the additive, linear nature of the model, the variance of wages could be decomposed according to the following simple formula:

István Boza

56  Elgar encyclopedia of labour studies

wages may be confounded by the differing skill composition of different firms’ workforce. Besides focusing only on the wage changes of individual workers (staying at the same employer), another way of eliminating this composition effect is proposed by Card et al. (2016). The authors substitute standard wage outcomes with AKM firm premia as an alternative wage measure, which controls for unobserved worker heterogeneity. Rent-sharing parameters estimated using this method turn out to be substantially lower than standard cross-sectional estimates, signalling a large role of changes in the skill composition of the workforce in response to productivity shifts. On the other hand, a branch of novel research focuses on the role of compensating differentials in the variation of firm premia. When data on working conditions are widely available, hedonic wage regressions could be used to evaluate the monetary values assigned to overtime, flexible working hours or similar non-wage compensation components. Even with the lack of such direct data, the role of compensating differentials may be observed as the residual component of wage differentials, which is not explained by differences in productivity. Sorkin (2018) evaluates firms based on revealed preferences (observed mobility patterns), and interprets residual variation as compensating differentials, finding that only one-third of firm effect variation could be attributed to firm ‘value’ and the rest is due to compensating differentials. Lamadon et al. (2022) also build a structural model with monopsonistic labour markets and identify both firm-specific rent-sharing and amenity parameters, showing that while both differences in shareable rents and amenities contribute strongly to wage variation, their contribution is mutually offset by the strong correlation between the two. That is, high-productivity firms tend to compensate workers in positive amenities rather than wages, decreasing the variation in estimated firm effects. Finally, due to its linearity, the components of the AKM model could be also used to decompose the overall wage difference among groups of workers or firms into the sum of the differences in the AKM worker effects (individual quality), differences in firm premia and the difference in the covariate index of time-variant observables; or in an extended model also the differences István Boza

in occupation wage effects. For instance, one could decompose the gender gap as in Cardoso et al. (2016), or characterize gaps between workers of different education levels, between firms of different industries, ownership or union membership, or even compare formal and informal hiring channels to assess the contribution of systemic sorting of higher-ability workers into higher-wage firms (or occupations) to the specific wage gaps. By relaxing the constant firm premium assumption and allowing firms to pay different premia for male and female workers, Card et al. (2016) also propose a way to decompose the differences in the average firm effects faced by males and females into a bargaining (within-firm) and a sorting component. These flexible models clearly present that women sort out of high-wage firms, but also that to some extent they receive smaller premia within the same firm. Female wage premia also move less strongly together with firm productivity, further suggesting the relevance of bargaining differences.

Model extensions

The past decade has also seen the emergence of alternative specifications of the AKM model, which either augment the model with additional terms, or relax the restrictive assumptions of the classic formulation. One important alternative to the additive separable benchmark model contains a match effect for all observed employer‒employee matches, capturing average residual wages of employment spells after controlling for the (within-match) effect of time-varying factors. If the AKM assumptions hold, these match effects could be further decomposed into firm and person effects identical to the baseline model, but also providing as a residual the orthogonal match components, which could be then used for wage variance or wage gap decompositions (Woodcock, 2015). Card et al. (2016) and Torres et al. (2018) augment the AKM model by introducing a third high-dimensional fixed effect for capturing job title (occupational) heterogeneity. This approach is motivated by the notion that beside firms, given occupations may possess such (time-invariant) general features that dictate different wage levels even for workers of the same skill and at firms of similar productivity. The source of such occupational differentials could be differences in working

Firms and wages  57

conditions, or even social appreciation. The distinction of firm and occupation heterogeneity can be important as the sorting of high-wage workers into specific occupations and the clustering of such occupations in high-wage firms could both increase the level of inequality, while the joint presence of these phenomena could even confound standard measures of wage sorting. The final set of alternative specifications relaxes the assumption that the firm has to pay the same premia for all their workers and in all periods. As a middle ground between the classic AKM and the match models, one may allow firm effects to vary over groups of workers. For instance, Card et al. (2016) allow for differing premia for female and male workers of the same firms, but this could be extended to workers of different education levels (Card et al., 2018), workers of different (discretized) skill levels (Lamadon et al., 2022), workers in different jobs (firm‒occupations cells) or groups based on worker age (Boza, 2021). Either estimated on different subsamples, or with one equation, these specifications may require further normalization assumptions to make the estimated firm‒group premia parameters comparable across disjoint mobility groups. A model with time-varying firm effects is presented in Lachowska et al. (2020) and also in Lamadon et al. (2022), with both finding a rather small role of intertemporal variation in firm premia in explaining wage dispersion. Still, as the identification and estimation of such models relies on both wage differences experienced by job-switchers and the wage changes of stayers within the firm – a variation that longitudinal studies of rent-sharing utilize – this model can prove to be a more versatile tool in some applications. István Boza

References

Abowd, J.M., F. Kramarz and D.N. Margolis (1999), ‘High wage workers and high wage firms’, Econometrica 67 (2), 251–333. Andrews, M.J., L. Gill, T. Schank and R. Upward (2008), ‘High wage workers and low wage firms: negative assortative matching or limited mobility bias?’, Journal of the Royal Statistical Society Series A, 171 (3), 673–697.

Bonhomme, S., K. Holzheu, T. Lamadon, E. Manresa, M. Mogstad and B. Setzler (2020), ‘How much should we trust estimates of firm effects and worker sorting?’, SSRN Electronic Journal. https://​doi​.org/​10​.2139/​ssrn​.3625745. Boza, I. (2021), ‘Wage structure and inequality: The role of observed and unobserved heterogeneity’, KRTK-KTI Working Papers 31. Card, D., A.R. Cardoso, J. Heining and P. Kline (2018), ‘Firms and labor market inequality: Evidence and some theory’, Journal of Labor Economics, 36 (S1), S13‒S70. Card, D., A.R. Cardoso and P. Kline (2016), ‘Bargaining, sorting, and the gender wage gap: Quantifying the impact of firms on the relative pay of women’, Quarterly Journal of Economics, 131 (2), 633–686. Card, D., J. Heining and P. Kline (2013), ‘Workplace heterogeneity and the rise of West German wage inequality’, Quarterly Journal of Economics, 128 (3), 967–1015. Cardoso, A.R., P. Guimarães and P. Portugal (2016), ‘What drives the gender wage gap? A look at the role of firm and job-title heterogeneity’, Oxford Economic Papers, 68 (2), 506–524. Correia, S. (2017), ‘Linear models with high-dimensional fixed effects: An efficient and feasible estimator’, Unpublished Working Paper. scorreia​.com/​research/​hdfe​.pdf. Kline, P., R. Saggio and M. Sølvsten (2020), ‘Leave-out estimation of variance components’, Econometrica, 88 (5), 1859–1898. Lachowska, M., A. Mas, R. Saggio and S. Woodbury (2020), ‘Do firm effects drift? Evidence from Washington administrative data’, Mimeo. Lamadon, T., M. Mogstad and B. Setzler (2022), ‘Imperfect competition, compensating differentials, and rent sharing in the US labor market’, American Economic Review, 112 (1), 169–210. Sorkin, I. (2018), ‘Ranking firms using revealed preference’, Quarterly Journal of Economics, 133 (3), 1331–1393. Torres, S., P. Portugal, J.T. Addison and P. Guimarães (2018), ‘The sources of wage variation and the direction of assortative matching: Evidence from a three-way high-dimensional fixed effects regression model’, Labour Economics, 54, 47–60. Woodcock, S.D. (2015), ‘Match effects’, Research in Economics, 69 (1), 100–121.

István Boza

Footballers’ labour market

broadcast rights, then teams will use the extra revenue to spend more on player talent, since they are assumed to break even. There are two substantial problems with this theoretical picture. First, teams might vary in their budget constraints. Teams with rich owners and/or with preferential access to credit might spend more on players such that financial losses are incurred. The possibility of a ‘soft budget constraint’ is covered below. Second, the basic model assumes uniform talent, and it is clear that players vary considerably in skills and ability. In practice, footballers’ pay is determined by a bilateral bargaining process between team and player (via an agent). In this process, the team considers the value of a player to the team in comparison with the value of the next-best alternative player. The team needs to consider who will perform a player’s role in the team if the negotiation breaks down. The player’s agent will consider the rewards to the player at this team compared to potential offers from alternative teams. Ideally, a player’s agent would like to generate an auction for a player’s services with several bidding clubs, but in practice the number of credible bidders will be small. If a bargain over pay is struck, then the pay level will be determined by a number of factors:

Introduction

This entry outlines the economic principles associated with the footballers’ labour market. The footballers’ labour market is one example of labour markets for team sports, but the themes covered here are generally applicable to other team sports. However, it should be stressed that European football and North American sports leagues differ considerably in degrees of regulation. North American sports labour markets feature player entry by draft from college, and substantial restrictions on player mobility. For example, it takes six years of service before a Major League Baseball player can become a free agent, able to seek contract offers on the open market. In contrast, all football players over 21 years old are effectively free agents. See Simmons and Berri (2018) for further details on the differences between North American and European sports labour markets. This entry proceeds as follows. It first establishes some basic theoretical concepts, then examines patterns of player mobility. It investigates superstar effects for elite footballers, and puts forward some principles of contract determination. Finally, it summarises discussion of effects of financial regulation on the players’ labour market.

● The importance of the player to the team in terms of substitutability of skills and ability. ● The team’s ability to pay (financial resources). ● The degree of team owners’ impatience in terms of team building. ● The player’s willingness to move. ● Rival bids for the player’s services.

Theoretical concerns

Unlike standard markets, firms (teams) in professional football leagues can be viewed as maximising wins rather than profits (Késenne, 2014). They do this subject to a breakeven constraint. With revenues equal to costs, we can assume that teams acquire units of talent, abstractly considered. Then talent is hired so that revenue per unit of talent equals the ‘price’ of talent. If we then assume that total talent of players in a team can be measured by the team’s payroll of the playing squad, then greater spending on payroll relative to the league as a whole should result in more team wins. Increased wins mean improved league standing and this translates into increased team revenues. If there is a sudden increase in team revenues, for example from increased sale of television

Player mobility

Footballers have short careers of around 12 years as first-team professionals; longer for goalkeepers compared to outfield players. Leagues are organised in hierarchical structures with divisions that are linked by promotion and relegation. Also, leagues vary in overall playing quality and total revenues, and it is commonly accepted that the top divisions in England, France, Germany, Italy and Spain have better quality and revenues than the other top divisions in Europe. A set of best-performing clubs can enter the lucrative 58

Footballers’ labour market  59

Union of European Football Associations (UEFA) Champions League, and playing in this competition is an important aspiration for top-level footballers. Under freedom of movement in the European Union, and since the Bosman Ruling of 1995, players from European Union countries can move between European leagues without constraint if they are out of contract. If players are intending to switch clubs while under contract then the incumbent club may insist on compensation for loss of the player’s services. This will be a transfer fee paid by the hiring club to the selling club. The principle of ‘assortative matching’ states that the best players will gravitate to the best teams by means of switching teams. High-ability players from small-market teams will move on to large-market teams. In this process, the teams that develop and train promising young players are compensated by transfer fees to richer buying clubs. Ribeiro and Lima (2019) have wage and career data for players in four divisions of Portuguese football from the 2002/03 to 2009/10 seasons. They show that players who move to new clubs in a higher division earn a wage premium of 60 per cent. Players who move to lower-division teams incur a 34 per cent wage penalty. A player who stays in a relegated club suffers a wage penalty of 17 per cent. There is no gain in wage growth for players who move to new clubs or stay in the same club in the same division. These results are consistent with assortative matching. The Portuguese league is known to be a net exporter of playing talent. Several players arrive there from South American countries such as Brazil, due to similarities in language and culture. The best immigrant and domestic Portuguese players then move on to higher-status leagues such as the English Premier League. Follow-up research based on Ribeiro and Lima (2019) needs to track these ‘stepping stone’ movements and correlate them with wage growth.

Superstar effects

Three footballers (Lionel Messi, Cristiano Ronaldo and Neymar Jr) occupy slots in the top ten of the Forbes 2021 list of highest-paid athletes worldwide. These players can be thought of as ‘superstars’, identified by

exceptional, consistent performance and popularity. Such players can magnify team revenues by their exposure in external, foreign markets such as Asia and North America. To a large extent, revenues from sales of replica shirts abroad can be captured by superstar players in the form of ‘image rights’ deals as part of their employment contracts. But we would expect superstar status to be reflected in basic salary as well. An early paper on superstar effects in Italy in the 1995/96 season, Lucifora and Simmons (2003), showed that a small number of superstar players had salaries considerably greater than the nearest group in the pay distribution. These superstar players had very high sensitivity of pay to playing performance (goals and assists). In a study of 469 players in Italy Serie A over the 2010/11 to 2014/15 seasons, Carrieri et al. (2018) measure superstar effects on pay via numbers of Google searches. Italy’s Serie A is the only European football league to offer a consistent archive of footballer salary data, compiled by journalists at Gazzetta dello Sport since 2008. Carrieri et al. (2018) measure player performance by goals, assists and journalist player ratings. They also include a measure of player bargaining power represented by total market value of all players contracted to a given player’s agency. The greater this value is, the greater the bargaining power will be, and this is predicted to translate into higher pay. The authors’ regression model includes a set of control variables for confounding influences. These include player age, career appearances for the player’s national team as an indicator of esteem, minutes played in previous season, and player’s position in the team. The results of Carrieri et al. (2018) show that performance, popularity and bargaining power are all economically and statistically significant determinants of player basic pay, excluding bonuses. Considering the whole distribution of salaries, the authors find that popularity is the most important determinant of pay for footballers at the top quantiles of the pay distribution. This confirms the presence of superstar effects in Italian football. Further work would usefully replace subjective (and possibly biased) journalist ratings with deep performance statistics, such as those published at www​.fbref​.com. Robert Simmons

60  Elgar encyclopedia of labour studies

Player contracts and shirking

Many professional footballers’ contracts are renewable multi-year agreements with complex contingency clauses. A large number of players have single-year agreements. In this setting, strong performances will tend to be rewarded by a contract extension on the same or improved terms, not least to guard against potential offers from rival teams. Multi-year contracts with guaranteed basic pay give rise to the possibility of opportunistic and shirking behaviour. A player might raise their performance levels in the season prior to expiry of a contract, in anticipation of renewal. Also, a player might reduce their performance levels in the season immediately after signing a new long-term contract. At first glance, these examples of strategic behaviour seem unlikely. A player’s contract can be renewed because of good performance at any time, and the player does not have to wait for contract expiry before getting a new deal. Shirking is typically regarded in labour economics as a moral hazard problem linked to imperfect monitoring of worker effort. In football, performances are visible to fans, experts, coaches and club executives. Nevertheless, there is a case to answer, as poor performance of a particular player could be due to their own lack of effort, but this is hard to disentangle from other influences, even when player performance is observable. In a team production setting, a player’s performance may reflect inadequacies of team-mates, inappropriate team formations and player task assignments, fan behaviour, and even refereeing decisions. Buraimo et al. (2015) analysed just under 3000 player-season observations of contracts in the German Bundesliga over the 2002/03 to 2012/13 seasons. On average, players were 1.6 years into their current contract and had just over two years remaining on their contract. The authors set up a two-stage model to show the effects of contract length variation on player performance measured by Kicker magazine journalist ratings. In the first stage, the number of years remaining on current contract for a particular player is determined by the average length of contract remaining of the player’s team-mates. In the second stage, journalist player ratings (as season averages) are regressed on predicted number of years remaining. A negative correlation is indicative of shirking, since more years remaining Robert Simmons

on the current contract might induce poorer performance. The authors show a positive relationship between the number of years remaining on the current contract and player performance. Essentially, players with longer contracts tend to deliver better performances than players with shorter contracts. This evidence contradicts the notion of shirking behaviour, and suggests a degree of rationality in contract determination in the German Bundesliga. This result does need to be generalised to other leagues, preferably with the use of direct performance metrics that are now readily available in the public domain.

Financial regulation of the players’ labour market

In the 2000s there was a growing concern that a large and increasing number of European football clubs were incurring serial deficits and hence accumulated debt. Financial solvency is a serious threat to football leagues. If clubs become bankrupt, the social welfare of affected fans is reduced, and failure to fulfil fixtures will damage league reputation, resulting in reduced sponsorship and broadcast revenues. Excessive growth of player pay was seen as an important factor behind parlous club balance sheets in the 2000s (Franck, 2018). Clubs with wealthy owners were seen as driving salary inflation by bidding against each other for available players who might help teams to raise their league positions or even win championships. Franck (2018) points to the combined effects of a soft budget constraint and an ‘arms race’ of bidding for player talent in European football. With the explicit objective of securing clubs’ financial sustainability, UEFA launched its Financial Fair Play initiative, with first implementation in the 2013/14 season. This policy amounted to a ceiling on ‘allowable losses’ of teams over a three-year accounting period, initially set at €45 million and then intended to be reduced over time. Allowable losses meant excess of costs as ‘relevant expenses’, primarily amortised transfer fees and payments to players, over ‘football-related revenues’. The major sanctions for breaking the allowable losses rule would take the form of exclusion of teams from UEFA competitions, since UEFA is an umbrella organisation that does not administer domestic European leagues. Softer pen-

Footballers’ labour market  61

alties, such as restrictions on the eligible number of players on a team roster, and fines, could also be applied. The intention was to reduce the growth rates of both transfer fees and player compensation. Later, both the English Premier League and Spain’s La Liga adopted cost control policies similar to Financial Fair Play. In the former case, this was by informal agreement. Spanish clubs had to demonstrate financial viability before being able to register players with their football association. According to Garcia-del-Barrio and Rossi (2020), UEFA’s Financial Fair Play policy was successful in reducing the ratios of club payroll to revenue over time, and particularly around 2013, the year of first implementation of the policy. Reduced payroll to revenue ratios suggests improved financial sustainability and lower risk of club insolvency. The industry benchmark for alarm is a wage bill to revenue ratio of 70 per cent. The English Premier League had rising profitability and fewer cases of club losses between 2007/08 and 2017/18. By 2017/18, The Guardian reported that profits per team in the Premier League were £22 million, and only seven teams made losses. The payroll to revenue ratio was 63 per cent on average, below the critical 70 per cent benchmark. However, the improvement in payroll to revenue ratio could be due to factors other than Financial Fair Play. Revenues from sales of broadcast rights and sponsorships increased considerably from 2010 onwards. Whereas previously these injections of revenue would be translated directly into player pay and transfer fees, this did not happen in the 2010s decade. An alternative explanation to the introduction of Financial Fair Play is that team owners became dissatisfied with sequential losses and hired tougher negotiators in player contract negotiations. Bargaining power in contract negotiations shifted in favour of clubs. The sustained increases in broadcast income and prize money from competitions, especially the UEFA Champions League,

meant that player salaries and bonuses for starting players at highly ranked clubs were still increasing. Moreover, salaries of players for lesser-ranked teams unlikely to qualify for the Champions League also increased. This was all possible without a greater threat to club insolvency. However, Szymanski (2017) documents that financial sustainability remains a problem for lower-division teams that receive smaller revenue streams. Robert Simmons

References

Buraimo, B., B. Frick, M. Hickfang and R. Simmons (2015), ‘The economics of long-term contracts in the footballers’ labour market’, Scottish Journal of Political Economy, 62 (1), 8–24. Carrieri, V., F. Principe and M. Raitano (2018), ‘What makes you “super-rich”? New evidence from an analysis of football players’ wages’, Oxford Economic Papers, 70 (4), 950–973. Franck, E. (2018), ‘European club football after “five treatments” with Financial Fair Play ‒ Time for an assessment’, International Journal of Financial Studies, 6 (97), 1–19. Garcia-del-Barrio, P. and G. Rossi (2020), ‘How the UEFA Financial Fair Play regulations affect football clubs’ priorities and leagues’ competitive balance’, European Journal of Government and Economics, 9 (2), 119–142. Késenne, S. (2014), The Economic Theory of Professional Team Sports: An Analytical Treatment, 2nd edn. Cheltenham, UK and Northampton, MA, USA: Edward Elgar Publishing. Lucifora, C. and R. Simmons (2003), ‘Superstar effects in sports: Evidence from Italian soccer’, Journal of Sports Economics, 3 (1), 35–55. Ribeiro, A. and F. Lima (2019), ‘Football players’ career and wage profiles’, Applied Economics, 51 (1), 76–87. Simmons, R. and D. Berri (2018), ‘Labor markets in professional team sports’, in: García, J. (ed.) Sports (and) Economics. Madrid: Funcas. Szymanski, S. (2017), ‘Entry into exit: Insolvency in English professional football’, Scottish Journal of Political Economy, 64 (4), 419–444.

Robert Simmons

G

Global value chains and employment relations

ers to improve compliance with the codes, by linking future sourcing decisions to their compliance records (penalizing or dropping noncompliant suppliers, and rewarding more compliant ones). We have seen explosive growth in this form of private regulation. From its beginnings in apparel and footwear in the 1990s (for example, Nike, GAP, and Adidas), it has diffused to many other industries over the last two decades: horticulture, home furnishings, furniture, fish, lumber, Fairtrade coffee, and others. This diffusion has created a large and growing ecosystem of actors and institutions to support private regulation. Multi-stakeholder institutions (MSIs) such as the Fair Labor Association (FLA), the Ethical Trading Initiative (ETI), and other business and human rights organizations, provide collective fora for private regulation. A growing number of companies, such as Intertek, ELEVATE and TÜV Rheinland, for example, engage in social auditing for global companies and MSIs. Others facilitate the exchange of audit information among global companies (for example, Sedex) and provide a myriad of consulting services to global brands and supplier factories. A number of critical NGOs use investigative reporting to pressure brands into improving their labor standards performance (for example, Oxfam, Labour Behind the Label). More recently, organizations such as the Sustainable Apparel Coalition and the Better Buying initiative have emerged that seek to improve the efficacy of private regulation by layering new policies onto the basic model. Socially responsible investment companies engage in evaluating private regulation programmes. International institutions such as the United Nations Global Compact promote the growth of private regulation. University centers have emerged that are devoted to the study of private regulation and workers’ rights, and the number of scholarly books, special journal issues, and articles devoted to the subject continues to grow.

Introduction

As is well established, globalization has made it possible to source production and services from anywhere. The globalization of supply chains and the absence of regulation at the global level generated demand for new forms of global governance from anti-globalizers calling for global governance, and from consumer and activist movements calling for global corporations to be more socially and environmentally responsible. Naming and shaming campaigns by non-governmental organizations (NGOs) tying Nike, Reebok and other brands to sweatshops and the use of child labor in the early 1990s provided the impetus for the development of private regulatory efforts with regard to labor and human rights in the supply chains by global athletic shoe and apparel retailers. Although there are different methods of voluntary private regulation, such as certification schemes (where organizations such as the NGO GoodWeave certify that a product such as a carpet has been made without child labor under sweatshop-free conditions) or reporting methods (where organizations such as the Global Reporting Initiative set particular standards of public disclosure of what member companies are doing to meet labor standards in their supply chains), the most common model of private regulation, however, are corporate codes of conduct. There are three elements here. First, standards are set in a code of conduct for suppliers generally based on the conventions of the International Labour Organization (ILO). Second, suppliers are audited by global brands or by social auditing companies that monitor whether they comply with the Code of Conduct of the global company. Finally, the companies provide incentives for suppli62

Global value chains and employment relations  63

The key question, and a key debate that remains unresolved, concerns whether private regulation has brought about meaningful improvements in working conditions in the global supply chain over the last 25 years. Since private regulation is “private” (most companies hesitate to share data with researchers), it has been difficult, though not impossible, for scholars to analyze the effectiveness of the model. In this entry I review what we know about its effectiveness, provide explanations for why private regulation does not appear to be working well, and highlight challenges for future research.

Does private regulation improve labor standards in global supply chains?

As noted, it has been difficult to find a comprehensive answer to this question. What we know is obtained from case studies, investigative reports by journalists, frequent disasters such as fires in garment factories or the collapse of the Rana Plaza factory in Bangladesh, and from recent analyses of auditing data. The evidence from the burgeoning body of case study research on the impact of private regulation with respect to labor standards suggests limited effects. Reviewing prior research, Bartley and Egels-Zandén (2016: 161) note that private regulation has: had some meaningful but narrow effects on working conditions and the management of human resources, but the rights of workers have been less affected, and even on the issues where codes tend to be most meaningful, standards in many parts of the (apparel) industry remain criminally low in an absolute sense.

This body of research shows unstable and limited improvements in working conditions (for example, health and safety), no improvement in freedom of association or collective bargaining rights, as well as a lack of steady improvement in compliance over time. Locke (2013) suggests that supplier factories “cycle in and out of compliance,” even in the case of Nike, a leading exponent of the private regulation model. More recent comprehensive cross-sectional evidence using data from a global auditor (Kuruvilla, 2021) appears to confirm the conclusions drawn above through case studies.

On the positive side, using evidence on violations of labor standards provisions in company codes from 12 countries and industries over a seven-year period from 2011 to 2018, there is clear evidence that the incidences of child labor are close to zero. Violations of child labor provisions still exist (some of these relate to record-keeping), but they are still negligible. With regard to forced labor as well, the incidences of violations are insignificant; but, as we know, it is not always easy to detect forced labor violations in a brief two-day audit. The difficulty of detecting violations with regard to discrimination in a brief audit is also well known; women workers experiencing discrimination and harassment are unlikely to confide in an unknown auditor in a brief meeting inside the factory premises. Hence, while the discrimination violation data show a positive trend (less than 0.1 violations per audit), they must be taken with a grain of salt. The same dataset also indicates a clear and discernible decline in total working hours and continuous working without rest, two provisions that are included in most codes of conduct. Weekly working hours have declined on average from 60 hours per week to 55 hours per week, but there is substantial variation across the 12 countries and industries examined. While the overall trend is generally towards reduced hours, Kuruvilla reports increased hours in the toy and footwear industries, particularly after 2017. With regard to the number of days of continuous work without a rest day, there is an average decline across 12 countries and all industries from 10.5 days to 8.5 days over the period from 2010 to 2018, which is significant progress. This progress could be attributed to private regulation’s effectiveness, but just as plausibly it could be the result of better enforcement of national laws by the respective governments. Either or both explanations are relevant here. The data also show that the percentage of workers in a factory who are paid correctly (according to the relevant laws) has increased, in every country and industry, from an average of 90 per cent to about 94 per cent. Thus, there are some positive effects. Unfortunately, negative results, in terms of violations per audit, and a number of other findings discussed below, outweigh the positive ones. The overall number of violations Sarosh C. Kuruvilla

64  Elgar encyclopedia of labour studies

on all labor issues combined increased rather than decreased during the 2010–2018 period. Similarly, the number of violations per audit with regard to health and safety shows a clear and discernible increase over the period in all countries and industries. Perhaps this reflects the heightened focus on health and safety issues in most corporate private regulation programmes after Rana Plaza. We should note that these results indicate an overall picture of progress, since we know that factories often cycle in and out of compliance (Locke, 2013), but also because the factories in the audit database keep changing as they enter or exit being suppliers. In order to examine the progress in the same factories over time (a better measure than the gross progress measure used above), Kuruvilla (2021) reports results from a set of factories that were audited multiple times during a four-year period from 2014 to 2018. The results of this analysis are even more damning of private regulation’s effectiveness: there was no sustained improvement in wages, the percentage of workers paid correctly, weekly working hours, or various other measures, even though some of these factories had been audited more than ten times during the four-year period. More importantly, although most codes of conduct require that suppliers pay wages high enough to include discretionary spending, the data shows that in general, in most countries, workers in the garment industry are paid just slightly above the local minimum. For example, the wages of Bangladeshi workers would have to increase by 223 per cent to meet a low estimate of living wages (provided by the Wage Indicator Foundation), and by 428 per cent to meet the highest estimate of living wages (provided by Asia Floor Wage). Finally, Kuruvilla’s (2021) data show that freedom of association and collective bargaining are not at all well developed in global supply chains.

Why is private regulation ineffective?

Many explanations have been advanced for the apparent failure of private regulation to improve labor standards in global supply chains. Locke (2013) suggests that the model is, in effect, badly designed, as it is based on several unwarranted assumptions. The model, he argues, assumes that asymmetric Sarosh C. Kuruvilla

power relations exist between buyers and suppliers, where powerful global buyers have the leverage to force suppliers to comply with their codes of conduct. This assumption does not hold, for two reasons. First, factories may produce for several brands, which allows them to spread their risk and insure themselves against an unforeseen change in consumer demand for any single brand’s product. This reduces the brand’s leverage over specific factories. Second, some large suppliers and vendors such as Foxconn actually have leverage over the companies they supply to. Apple, for example, is 80 per cent dependent on Foxconn for its most popular products. Locke (2013: 36) also points to the faulty assumption that through a system of incentives it is possible to achieve an alignment of interests of the different actors in the supply chain that are responsible for increased compliance. Buyers and suppliers often have different interests that “provoke mixed and often contradictory behaviors.” Buyers want suppliers to invest in improving labor standards, but they also consistently “squeeze” suppliers with lower prices for their products, thus prompting suppliers to hide or falsify compliance data. A third assumption which Locke (2013) suggests to be unwarranted is that the auditing mechanism is capable of producing high-quality and reliable information to drive sourcing decisions. Problems with auditing have received the most attention. To summarize briefly, audits are too short to uncover violations, so auditors “satisfice” through cursory document checks, quick walk-throughs, box-ticking exercises, without focusing on the root causes of problems found. Findings suggest most auditors are not well trained and that auditing has become commoditized and outsourced to large third-party generalist firms which charge too little to do a good job. Worker interviews are also problematic: most of those conducted at the factory site are staged and workers are coached to provide desirable answers; workers are not interviewed at their homes, where they might provide franker assessments of whether the factory engages in discrimination or union suppression. There is also considerable fraud and bribery, with auditors “willing to turn a blind eye when necessary to please factory managers” and increasing evidence that a large number of factory records are falsified to pass;

Global value chains and employment relations  65

easy, because most audits are announced in advance. Suppliers also evade audits by subcontracting production to factories not subjected to brand auditing. A further problem with private regulation concerns the third element, that is, whether companies are incentivizing their suppliers to improve on compliance. However, we know little about whether global buyers have done so, as empirical investigations of this third element of private regulation are rare; brands have not shared their sourcing data with researchers. In what is the first empirical study on the question of whether purchasing and compliance are aligned in a global corporation, Amengual et al. (2019) show, by analyzing purchasing data and compliance data, that there was no relationship between the two: factories with mediocre compliance records were getting more orders, while factories with improving records were getting fewer orders. And there is new research which shows that poor compliance is a function of poor purchasing practices as well. If orders are late, then factories have to work overtime to complete the process. More recently, scholars have begun to rely on concepts of organizational decoupling drawn from institutional theory to explain the lack of progress in private regulation (Kuruvilla, 2021; Bartley and Egels-Zandén, 2015). Organizational decoupling could account for the so-called policy‒practice gap, that is, why organizations adopt policies but do not implement them. Institutionalists argue that policy‒practice gaps typically occur when organizations respond to rationalizing pressures from the environment; for example, by adopting an externally induced rule to gain legitimacy or by avoiding legal sanction through symbolic adoption. The arguments made in prior research (e.g., Appelbaum and Lichtenstein, 2016) and the accumulated case study evidence (Bartley and Egels-Zandén, 2015) suggest that the symbolic adoption thesis is largely supported. A new strand of institutional theory hypothesizes the existence of a second decoupling form: practice‒outcomes decoupling, which occurs when formal structures are adopted, work activities are changed, and policies are implemented; but there is no evidence of the outcomes at the supplier level, that is, there are practice‒outcomes gaps. Institutional theorists argue that practice‒

outcomes decoupling is more likely to be prevalent in opaque institutional fields. Wijen (2014: 302) defines an opaque institutional field as one where “observers have difficulty identifying the characteristics of prevailing practices, establishing causal relationships between policies and outcomes and precisely measuring the results of policy implementation,” as opposed to more transparent fields where these issues are clear. In general, private regulation conforms to this definition. Companies do not share the results of the impact of their private regulation programs on workers. There is little or no transparency. Even MSIs do not share their analyses of each of their corporate members’ private regulation programmes publicly, or even internally with their members. The difficulty of identifying best practices is the key problem, as institutional theorists have long maintained that companies through processes of imitation converge on best practices.

Avenues for further research

With the growing demands for more transparency, it is possible that global corporations may be willing to share their private regulation data with researchers. There is a clear need for additional evidence using multiple sources of data. In particular, there is also a need for more research into the purchasing practices of global brands, and the extent to which poor purchasing practices are responsible for poor compliance with labor standards. Even more importantly, there is need for research inside the global corporation to see whether their sourcing practices are linked to their codes of conduct in ways that incentivize suppliers to comply with their codes. The ability of research to answer these questions depends heavily on whether corporations will share their sensitive sourcing and compliance data. One positive development that might influence research and practice is the new raft of mandatory due diligence laws emanating from Europe. Beginning with France in 2017, and Germany in 2021, and currently being discussed by the European Union, these laws will allow workers in supplier factories in far-flung global supply chains to sue, in Europe, the global brands if there are violations of their codes of conduct. This is likely to be a game-changer that will stimulate more research and better compliance such that Sarosh C. Kuruvilla

66  Elgar encyclopedia of labour studies

workers’ lives in global supply chains could be improved. Sarosh C. Kuruvilla

References

Amengual, M., G. Distelhorst, and D. Tobin (2019), “Global purchasing as labor regulation: The missing middle,” Industrial and Labor Relations Review 73 (4), 817–840. Appelbaum, R.P. and N. Lichtenstein (eds) (2016), Achieving Workers’ Rights in the Global Economy. Ithaca, NY: Cornell University Press. Bartley, T. and N. Egels-Zandén (2015), “Responsibility and neglect in global production networks: The uneven significance of codes of conduct in Indonesian factories,” Global Networks 15 (S1), S21–S44.

Sarosh C. Kuruvilla

Bartley, T. and N. Egels-Zandén (2016), “Beyond decoupling: Unions and the leveraging of corporate social responsibility in Indonesia,” Socio-Economic Review 14 (2), 231–255. Kuruvilla, S. (2021), Private Regulation of Labor Standards in Global Supply Chains: Problems, Progress and Prospects. Ithaca, NY: Cornell University Press. Locke, R.M. (2013), The Promise and Limits of Private Power: Promoting Labor Standards in a Global Economy. Cambridge: Cambridge University Press. Wijen, F. (2014), “Compliance to achievement in sustainability standardization,” Academy of Management Review 39 (3), 302‒323.

H

HRM practices and productivity

true that any one empirical case study done by the researcher should feel generalizable, which means that the conclusions of the study would apply to many similar firms. Most of this entry is aimed at answering the question “What do we know?” The entry concludes with a short section on “What is to come?” in research. If optimal management practices were not changing over time, we could stop doing research, having learned what we need to know. We have certainly learned a lot, and have a body of research knowledge. However, optimal practices are changing over time. This creates some natural experiments that enable us to learn more, and also implies that there is a need for new work.

Introduction

Economists have been studying the importance of management practices for elevating workers’ productivity for only about 40 years. The research began as largely theoretical, and then moved over time to be more empirical. Empirical work that estimates the effects of management practices on performance pay usually relies on the cooperation of firms. Early on, firms cooperated by giving their data to economists, and more recently, by letting economists run field experiments within the firms. In this entry two broad questions are addressed: (1) What do we know? and (2) What is to come? The discussion draws on research done by personnel and organizational economists. “Management practices” refer to topics such as performance pay, promotions, hiring practices, teamwork or problem-solving teams, or the allocation of decision rights within firms. These practices could apply to fairly low-skilled or to high-skilled workers, but because productivity is most measurable for the least-skilled, most of the empirical work is on the least-skilled. A short subsection below discusses research surrounding managers. The study of management practices effects differs from studies of labour market policies, such as minimum wages, because the effects of management practices are, and always should be, endogenous to the strategy of the firm. For example, we would expect the optimal management practices for a low-cost airline, such as Southwest Airlines in the United States (US), to be different from the optimal management practices of a higher-end airline, such as business class in United or Emirates Air. There is no “one size fits all.” However, it should still be

Four methods of studying the impact of practices on productivity

When thinking about past work and where the future agenda is moving, there are four empirical approaches to identifying the impact of management practices on performance. All of these can be seen as types of “insider econometrics,” when one becomes an “insider” to firms by understanding what the firm does, and by modelling it; see Shaw (2009). New management technologies are being discovered In the early 1990s, no academics, even in business schools, were thinking about how to model the inside functioning of a firm or plant, aiming to model its performance. At this time, after visiting an initial set of steel mills, Ichniowski et al. (1997) obtained five years of productivity data from virtually all US steel mills by visiting them in person, and obtaining detailed information on the management practices that they had used over time, such as teamwork, incentive pay, and so on. These “innovative” practices were being imported from Japan. 67

68  Elgar encyclopedia of labour studies

More recently, the innovative practices of lean manufacturing were carried to developing countries and shown to raise productivity. A major field experiment was conducted in mid-sized textile plants in India, showing that lean manufacturing substantially raised performance (Bloom et al., 2013). These papers, and some recent work by Florian Englmaier of Ludwig-Maximilians-University (LMU), in Munich, followed the adoption of major sets of management practices. But these studies are very expensive, and alternative methods, such as internet scraping, are being used. The firm decides to adopt management practices that are new to the firm, perhaps because new digital technologies make new management practices optimal The empirical methodology here could be called a natural experiment. The natural experiment is when the firm introduces a new management practice and then the researcher obtains the firm’s data to model it. These papers are fairly rare, because researchers lack information about when firms made changes, or do not have the cooperation of the firms. One of the most famous of these papers is Lazear’s (2000) paper on Safelite glass, in which the firm switched from hourly pay to performance pay. As a natural experiment, some exogenous shock had to induce change. In the Safelite study, the firm introduced performance pay to replace hourly pay because it installed computers that kept track of employee output. Thus, a new management practice was introduced due to use of the new digital technology of computerization. In addition to digitization, Covid-19 shocks are studied today for the changes in management that followed. The firm adopts a management practice: field experiments An increasing number of recent research papers use field experiments in which the researchers have obtained the cooperation of firms to conduct the experiments. In this work, it is the firm that is lacking information, not the researcher. This lack of information is important, because if firms were fully informed they would have enacted a change in optimal management practices themselves, Kathryn L. Shaw

and would reject field experiments that researchers recommend. The advantage of the field experiment lies in the econometrics. In the field experiment, the researcher convinces a firm to put in a new practice, such as a bonus or a training program, and follows it over time, with a treated group and a control group (if possible); see Sandvik et al. (2021). The study of management practices by using observational data from firms, or from the internet The big disadvantage of a field experiment is that it might tackle only a small issue in the firm. Most firms cannot, or will not, change everything about how they run their firm. One reason for using observational data is that the researcher can follow firms over time when the firm is making many management practice changes (Ichniowski et al., 1997). Observational data can now be scraped off the internet. An example is a study of the introduction of artificial intelligence (AI) into law firms (Jedras et al., 2022). A large number of research papers on management practices, such as hiring, are using data from the company Burning Glass Technologies, that is, an aggregator of job postings on the internet (see, e.g., Frederiksen et al., 2020). Overall, today’s research papers on management practices may often combine several methodologies in one paper. For example, a research paper that introduces a field experiment is also likely to survey the employees both before and after the field experiment.

The lens for interpreting empirical papers: strategy and “treatment of the treated”

A key point in personnel economics is that the research case studies should feel generalizable; if that is not true, why publish a paper on one firm? But results should not be generalizable to all firms. First, firms must have individual strategies to create a sustainable competitive advantage, so it must be the case that the management practices match the strategy of the firm. Consider Southwest Airlines versus United Airlines. Traditionally, teamwork was valued at Southwest because its strategy was to be a low-cost airline, whereas attention to indi-

HRM practices and productivity  69

vidual customers was valued at United, when it served business customers. In this case, the value of a management practice, such as teamwork, not only depends on the productivity that it might produce, but it must also align with the strategy of the firm to be of value. This introduces an empirical quandary: the researcher will only observe the use of teamwork when it is of value to the firm’s strategy. Second, because a management practice is only adopted when it is aligned with the firm’s strategy, the econometric exercise of estimating the practice’s impact on productivity is one in which the empirical model is identifying the “treatment of the treated.” Firms should only permit experiments that might have a positive impact, and reflect the strategy of the firm.

What have we learned? Some key results

As space is limited, only a short list of papers is discussed here. Topics covered are aspects of performance pay, hiring, teamwork, promotions, and performance reviews. Performance pay A classic paper is Lazear (2000), studying the impact of performance pay on individual output in the Safelite glass company. This is a company in which each employee drives a Safelite truck to a homeowner’s house, and installs a new windshield in their car, replacing a broken one. Safelite switched from hourly pay to paying for performance, measuring the number of windshields installed by each person. There followed a 40 percent increase in the number of windshields installed. What would be expected is that each employee is working harder: that was true, and raised productivity by 20 percent. The rest of the increase in productivity arose from sorting: people who were skilled at installing windshields started taking jobs with Safelite, increasing output by another 20 percent. This “sorting effect” insight makes this paper one of the most widely cited in personnel economics. But why did Safelite adopt performance pay in the 1990s? It was because it had installed a computer system that began to keep track of individual performance.Thus, the “treatment” of the performance pay was

driven by the introduction of computer technology, which identifies why the performance pay was right for this firm at this time. Bartel et al. (2007) followed the introduction of computerization in the valve industry, and the management practice changes that followed. Hiring The process of hiring employees was for some time a management practice that was very difficult to study, because firms often do not keep data on the employees who they did not hire, or do not keep data on the interview process that is a big part of hiring. In one natural experiment, Hoffman et al. (2018) show that job testing when hiring can result in better new hires, because it eliminates the potential biases that managers may have. A number of researchers have had observational or experimental data which shows that the management practice of asking employees to refer their friends results in better hiring. Another approach is to run experiments in firms, as Cowgill (2020) did regarding the use of AI in hiring. The third approach is to scrape data off the internet, as is being done using the Burning Glass data. Practices regarding managers For the most part, empirical work on management practices and performance has been conducted on workers for whom one can measure productivity, and thus on those in the less-skilled job frame. For most employees, paid hourly or salaried, employee performance can only be judged on a relative scale, where a manager is able to rank one employee relative to another. This point is emphasized in the tournaments literature, beginning with Lazear and Rosen (1981). This is an example in which, once you read the article, you know it to be true: that many firms are run as tournaments for managers. As in the Safelite paper, the remarkable theory behind the tournaments paper is what makes it so valuable. Recent work asks the question, “How do managers matter?” The focus on managerial quality arose in the 2010s when many companies began putting in place “engagement surveys” aimed at better management and lower turnover. Using observational data on several different divisions in a very large Canadian bank, Bartel et al. (2022) show Kathryn L. Shaw

70  Elgar encyclopedia of labour studies

first that the value of managers, and the pay schemes for managers, depends on the strategy of the division in which the manager works. Moreover, the most talented employees are found to work with the most talented managers. Thus, there is assortative matching for managers. In a different environment of lower-skilled technology jobs, Lazear et al. (2015) do not find assortative matching, as high turnover implies that employees are randomly assigned to managers when hired. In a natural experiment, Hoffman and Tadelis (2021) find that the best managers in a high-skilled company use practices that produce lower turnover for their workers.

Why adopt? The pervasive roles of management complementarities and technology

The work of Milgrom and Roberts (1990) is a foundational theory of complementarity: management practices interact. Their theoretical model is one of “super-modularity” in the production function, such that management practices have an impact when introduced as sets of practices in firms. Ichniowski et al. (1997) took this model to the data, showing that in steel mills, sets of practices that make use of all the talents of the workers running the mills, produce the highest output for the mills. That is, when you have teamwork, performance pay, information sharing, and job security, the mill performs at its very best. If a steel mill moved from a traditional union setting by only introducing team meetings, there would be very little increase in mill output. Researchers are increasingly running multi-armed experiments, so that one practice and its interaction with another practice is in the regression.

What is to come? Studying artificial intelligence

To suggest where studies of management practices might go in the future, consider the introduction of AI in firms. AI either replaces some workers, or augments the work of others, including managers, so that employees work faster and produce a higher-quality product. Thus, there are three ways in which AI is entering the firm. While AI may be used to replace workers, it can also be used to augment the work of Kathryn L. Shaw

employees. AI, for example, is being used by lawyers to do simple research, so new young lawyers are being trained to have more authority and work with clients (Jedras et al., 2022), and similar things are occurring for financial analysts (Grennan and Michaely, 2020). AI can also be used to improve the hiring process for recruiters. Though much has been written about hiring bias in AI, in Cowgill’s (2020) experiments the AI better identifies relevant skills that the humans were overlooking. A large number of firms will likely introduce AI in the near term and the future. This changes the optimal management practices, improving productivity and product quality. It moves some workers to use their non-routine cognitive skills (as in Autor’s and Deming’s many papers), but it also results in “human-in-the-loop,” in which workers are doing the same jobs as before AI, but are needed to increase product quality (Jedras et al., 2022).

Conclusion and agenda

What do we know? Beautiful theoretical models have led, and should lead, to the empirical testing of these models. Some theoretical models, such as Lazear and Rosen’s tournament theory, led to numerous empirical tests, yet the model still stands, because it is so obviously correct in its simple form. Other theories do not have a clear empirical prediction, and need to be taken to the data. There are largely four empirical frameworks, including the recent use of firm-level field experiments, to test for the impact of management practices. However, one needs to be cautious. The research papers rarely point out that a management practice is only of value to a firm if the practice is aligned with the strategy of the firm, and thus the effects of management practices on productivity are, and should be, thought of as “treatment of the treated” effects. The management practice is also of more value if it is complementary to other practices, and this fact is omitted from most field experiments estimating management practice effects. What is to come? New management practices are introduced for two reasons. One is that there was a “technology shock” such as computerization or AI, or even better management ideas. The other is that firms are

HRM practices and productivity  71

not well run, and thus in field experiments researchers can try out practices new to the firm. The future of management research revolves in part around following the effects of AI usage by firms in the future. New theoretical models are arising as we face the need to understand AI, not just for its impact on skill demand, but also for its impact on management practices. Kathryn L. Shaw

References

Bartel, A.P., C. Ichniowski, and K. Shaw (2007), “How does information technology affect productivity? Plant-level comparisons of product innovation, process improvement and worker skills,” Quarterly Journal of Economics 122(4), 1721–1758. Bartel, A., J. Jedras, and K. Shaw (2022), “How managers matter: follow your stars, monitor your guardians, persuade your foot soldiers,” Working paper. Bloom, N., B. Eifert, A. Mahajan, D. McKenzie, and J. Roberts (2013), “Does management matter? Evidence from India,” Quarterly Journal of Economics 128(1), 1–51. Cowgill, B. (2020), “Bias and productivity in humans and algorithms: theory and evidence from resume screening,” Working paper. Frederiksen, A., L. Kahn, and F. Lange (2020), “Supervisors and performance management systems,” Journal of Political Economy 128(6), 2123–2197. Grennan, J. and R. Michaely (2020), “Artificial intelligence and high-skilled work: evidence from analysts,” Swiss Finance Institute Research Paper No. 20–84.

Hoffman, M., L. Kahn, and D. Li (2018), “Discretion in hiring,” Quarterly Journal of Economics 133(2), 765–800. Hoffman, M. and S. Tadelis (2021), “People management skills, employee attrition, and manager rewards: an empirical analysis,” Journal of Political Economy 129(1), 243–285. Ichniowski, C., K. Shaw, and G. Prennushi (1997), “The effects of human resource management practices on productivity,” American Economic Review 86(3), 291–313. Jedras, J., E. Lazear, and K. Shaw (2022), “Adjusting to AI: the impact on the legal profession,” Working paper. Jedras, J. and K. Shaw (2022), “Productivity and product quality gains from AI: mitigating the rising impact of wildfires in the insurance industry,” GSB Stanford working paper. Lazear, E.P. (2000), “Performance pay and productivity,” American Economic Review 90(5), 1346–1361. Lazear, E.P. and S. Rosen (1981), “Rank-order tournaments as optimum labor contracts,” Journal of Political Economy 89, 841–864. Lazear, E.P., K. Shaw, and C.T. Stanton (2015), “The value of bosses,” Journal of Labor Economics 33(4), 823–861. Milgrom, P. and J. Roberts (1990), “The economics of modern manufacturing: technology, strategy and organization”, American Economic Review 80(3), 511–528. Sandvik, J., R. Saouma, N. Seegert, and C. Stanton (2021), "Employee responses to compensation changes: evidence from a sales firm", Management Science 67(12), 7687–7707. Shaw, K. (2009), “Insider econometrics: a roadmap with stops along the way,” Labour Economics 16(6), 607–617.

Kathryn L. Shaw

I

Intergenerational income mobility

average, richer parents have better-endowed children, but there is no cost to parents of this transmission. Because endowments are rewarded in the labour market, inherited endowments account for some of the association between parents’ and children’s incomes. Investments can also depend on family income if children (or their parents) cannot borrow to support education. If there are such credit constraints, the children of poorer parents will receive too little education, and intergenerational persistence will be inefficiently high. Researchers have therefore searched for evidence of credit constraints. The evidence for their existence has been mixed, although recent contributions have highlighted that investment may be too low in early life for high-ability children from poor backgrounds, meaning that they are unable to develop their potential. The Becker–Tomes model (Becker and Tomes, 1986) illustrates the difficulties of interpreting estimates of intergenerational persistence because inherited characteristics mean that intergenerational elasticities would be expected to take a non-zero value. In that case, how high is too high? There are many reasons why parents’ and children’s outcomes are related, and deciding which of these are ‘unfair’ requires value judgements. Policy intervention in this area requires balancing the possible benefits of more mobility against the disadvantages of blunting incentives for parents to help their children, and for children themselves to work hard. Researchers have made some progress in understanding the drivers of mobility by comparing estimates of intergenerational mobility between countries and regions and over time. The idea is that any intergenerational persistence which stems from endowments (for example, hereditary aspects) should be broadly constant across countries and localities. Thus, comparing international mobility in this way has led researchers to speculate on the factors

There is a large body of evidence demonstrating that those from richer backgrounds have better educational outcomes, earn more and enjoy higher living standards than those who grow up in poorer families (see, e.g., Blanden, 2018). There are several reasons why we might be concerned about this. First, a strong association between incomes across generations – with children from poor families likely to be poor as adults, and elite positions in society closed to most people – is frequently considered an indicator of insufficient equality of opportunity. Second, it may indicate that talented young people cannot access high-productivity jobs; this will have an impact on economic growth and the welfare of society. Finally, rigid intergenerational inequality may threaten the social fabric of society and lead to political instability. The strength of the relationship between parents’ income and that of their children as adults is known as intergenerational income persistence. Its corollary is intergenerational income mobility, the topic of this entry. The measurement and interpretation of these concepts have been of great interest to social scientists for five decades.

The Becker‒Tomes model

The Becker–Tomes model provides a framework for analysing the mechanisms that drive intergenerational income persistence. The model starts with the idea that parents care about both their own consumption and that of their children as adults. Children’s adult incomes are determined by endowments, and by investments made in their education and skills. Investments are costly for parents. Endowments are to some extent inherited: on 72

Intergenerational income mobility  73

Figure 1

British cohort study: 1970 birth cohort

that lead to relatively strong and weak intergenerational persistence.

Measuring intergenerational income persistence

One simple approach to measuring intergenerational persistence is to assign parents and children to categories and then to calculate the share of children who end up in each category depending on the category they started in. An example of this is shown in Figure 1, which compares parental income categories when sons are age 16 with sons’ earnings categories (in 1986) when they are aged 42 (2012) using data from the United Kingdom. There is clear evidence of some intergenerational persistence, as almost 35 per cent of sons who grow up in the bottom 20 per cent are themselves in the bottom 20 per cent, and more than 40 per cent of the richest sons find themselves in the highest-earning 20 per cent. If there were no intergenerational persistence, we would expect all the bars to be the same height. If incomes are measured in terms of log values, the interpretation of a slope of 0.4 is that if there are two friends, Paul and John, and Paul’s parental income is twice as large as John’s, then Paul could expect to earn 40 per cent more than John. Beta is known as the ‘intergenerational elasticity’ and provides

a single summary measure of the extent of mobility in the population studied. In recent years it has become more standard to measure outcomes in terms of ranks rather than absolute values. The statistical technique of regression allows the estimation of a single summary measure of income persistence. This can be visualized graphically as plotting the combination of parent and child outcomes on a graph and then considering the slope which best fits the average relationship between the two. This slope of regression line in Figure 2 is estimated using regression, and is calculated by minimizing the sum of the squared gaps between each observation and its prediction on the line (errors in Figure 2). It has also become common to measure mobility between the parents’ and children’s ranking in their own income distribution. The methods used to estimate intergenerational persistence are therefore straightforward, and provide a description of the association in outcomes across generations. However, obtaining data good enough to make these calculations is challenging. It is important to observe parents and children at around the same age, which means that families must be followed for around 30 years. In addition, it can be difficult to collect accurate data on income. Income measured at a point Jo Blanden

74  Elgar encyclopedia of labour studies

Figure 2

Parent–offspring regression model

in time is not necessarily representative of longer-term experience. This leads to what economists call ‘measurement error’ and means that the intergenerational association is underestimated. This was a problem faced in many early studies of mobility. The importance of measurement error for estimates of intergenerational mobility was first exposed in two papers published together in the American Economic Review in 1992. Although Solon (1992) and Zimmerman (1992) use different datasets, they both find that when measurement error is properly accounted for, the intergenerational elasticity in the United States (US) for men was 0.4, not 0.25 as had been found in previous studies. Björklund and Jäntti (1997) continued to paint a less than rosy picture of mobility in the US by comparing intergenerational income mobility in the US and Sweden. They found that mobility is higher in Sweden, and therefore provides an early hint that intergenerational persistence and cross-sectional inequality might be positively related (see below). In order to overcome measurement problems, estimates of intergenerational mobility now increasingly rely on matched administrative data from tax records. Tax records have the dual advantage of being accurate records of income (assuming that tax evasion is not widespread), and being available over a longer period, so that lifetime income can Jo Blanden

be better approximated. Mazumder (2005) shows that the estimates from Solon and Zimmerman are substantially smaller than those he derives from income tax records. A disadvantage is that they lack information on other key variables, such as race and education level. So far, these data sources have been primarily exploited for the Nordic countries, Canada and the US. It is important to stress that intergenerational elasticity does not measure the causal relationship between incomes across generations; the intention is not to find out how the child’s income as an adult would respond to a change in the parents’ income, everything else held constant. Rather, measures of intergenerational persistence provide a summary measure of intergenerational inequalities, assessing how children’s outcomes vary with parental income and all the other factors that can vary alongside it, such as parents’ education, family culture, motivation, and child’s ability. Evidence suggests that among rich countries mobility is high in the Nordic nations and Canada, middling in Western Europe, and low in the US. Evidence for middle-income and less-developed countries tends to find lower mobility than in most rich nations. The strict data requirements mean that there is often considerable uncertainty about both country rankings and changes over time.

Intergenerational income mobility  75

Differences in intergenerational income mobility

Nonetheless, comparisons of intergenerational persistence across countries have been used to draw powerful policy conclusions. For example, the ‘Great Gatsby’ curve indicates that in societies with high levels of income inequality, the extent of intergenerational persistence is higher. US government advisor and academic economist Alan Krueger called this relationship the ‘Great Gatsby curve’ after the novel by F. Scott Fitzgerald depicting social stratification in American society. Of course, as is frequently noted, correlation does not imply causality. An alternative approach is to look at local areas within countries, and Chetty et al. (2014) look at this issue across very small areas in the US, while Connolly et al. (2019) add data from Canada to the analysis. These studies relate the extent of intergenerational persistence to inequality at the local level and other local characteristics, while the addition of Canadian data attempts to distinguish cultural and political effects by considering the influence of the border. All studies find a correlation between cross-sectional and intergenerational inequality, as well as evidence that human capital investment by both the family and the state has an important role in promoting mobility. This perspective that areas matter is strengthened by evidence finding important causal effects of the time children spend growing up in an area on the extent of mobility they achieve. However, recent commentary has questioned the precision of the underlying estimates (Mogstad and Torsvik, 2021).

Family background

So far, this entry has discussed measures of mobility based on linking the economic status of parents and children. However, these can only tell a partial story of the extent to which family background affects individuals. One alternative approach is to consider the correlation between the outcomes of siblings. This includes the influence of family factors that are broader than the outcome focused on (usually income or education), plus the influence of other shared factors such as the neighbourhood they lived in growing up. Evidence from siblings suggests (Björklund and Salvanes, 2011) that family factors are

more important than is indicated by simple parent‒child measures, with neighbourhoods contributing only a little to the story. In more recent research, economists are also considering relationships across three generations, looking from grandparents to grandchildren. Once again, estimates show that the standard parent‒child story is missing something, and that broader factors such as ‘family culture’ are likely to have an effect that persists across several generations (Braun and Stuhler, 2018). Practices in measuring mobility are likely to continue to improve, giving us access to more accurate estimates of mobility for different times, places and family relationships. However, these are summary measures driven by myriad factors, and they are therefore rather difficult to interpret. Studies are increasingly trying to isolate some of these mechanisms, but we are still far from fully understanding the drivers of mobility, how it could be improved and, crucially, what the ideal level of intergenerational mobility should be. Although a hugely important area of research, policy recommendations based on estimates of intergenerational mobility should be approached with caution. Jo Blanden

References

Becker, G.S. and N. Tomes (1986), ‘Human capital and the rise and fall of families’, Journal of Labor Economics Part 2, 4(3), S1–39. Björklund, A. and M. Jäntti (1997), ‘Intergenerational income mobility in Sweden compared to the United States’, American Economic Review, 87(5), 1009–1018. Björklund, A. and K.G. Salvanes (2011), ‘Education and family background: Mechanisms and policies’, in: E. Hanushek, S. Machin and L. Woessmann (eds), Handbook of the Economics of Education, Volume 3, 201–247. Elsevier. Blanden, J. (2018), ‘Intergenerational income persistence’, IZA World of Labor. Braun, S.T. and J. Stuhler (2018), ‘The transmission of inequality across multiple generations: testing recent theories with evidence from Germany’, Economic Journal, 128(609), 576–611. Chetty, R., N. Hedren, P. Kline, and E. Saez (2014), ‘Where is the land of opportunity? The geography of intergenerational mobility in the United States’, Quarterly Journal of Economics, 129(4), 1553–1623.

Jo Blanden

76  Elgar encyclopedia of labour studies Connolly, M., M. Corak and C. Haeck (2019), ‘Intergenerational mobility between and within Canada and the United States’, Journal of Labor Economics, 37(S2), S595‒S641. Mazumder, B. (2005), ‘Fortunate sons: new estimates of intergenerational mobility in the United States using Social Security earnings data’, Review of Economics and Statistics, 87(2), 235–255.

Jo Blanden

Mogstad, M. and G. Torsvik (2021), ‘Family background, neighbourhoods and intergenerational mobility’, NBER Discussion Paper No. 28874. Solon, G. (1992), ‘Intergenerational income mobility in the United States’, American Economic Review, 82(3), 393–408. Zimmerman, D.J. (1992), ‘Regression toward mediocrity in economic stature’, American Economic Review, 82(3), 409–429.

International migration

simplest form, bilateral migration is related to the relative size of the origin and destination countries (as measured by population or gross domestic product) and the distance between them. The gravity model has become one of the most commonly used theories for predicting gross bilateral migration flows between regions and countries (Grogger and Hanson, 2011, among many others). One of the most crucial assumptions of early gravity models is that the rate of bilateral migration flows between two countries depends solely on the characteristics of the origin and destination countries. More recent studies relax this assumption and extend the basic model to control for the influence of potential alternative destinations (Beine et al., 2016). The migration process is highly selective. A significant share of studies that investigate the self-selection mechanism of immigrants are rooted in the theory developed by Roy (1951) and mathematically expressed the theory developed by Borjas (1987). According to the theory, immigrants are expected to be positively selected if income inequality (or returns to education/skills) in the receiving country is higher than in the sending country. Recently, Lazear (2021) developed a model where immigration policies and rationing of immigrant slots across different origin countries have a more important role in the process of self-selection of immigrants than is assumed in previous studies. As a consequence of policies leading to a rationing of migration slots, educational attainment of immigrants will be inversely related to the number of immigrants from a given origin country, and positively related to the total population residing in the origin country.

Introduction

Migration has always been an important contributor to population dynamics throughout the history of mankind. During recent decades, migration has been on the rise; the number of international migrants in the world has grown, according to the recent global estimates, from about 84 million in 1970 to 281 million in 2020, or from 2.3 per cent to 3.6 per cent of the world’s population (IOM, 2022). The potential migration based on people’s migration intentions is even higher. According to Gallup World Poll surveys, around 15 per cent of the world’s population, or more than 750 million people, would like to move to another country if they had the opportunity. Furthermore, not only is the volume of migration on the rise, but also the mix of typical migrants’ receiving and sending countries is changing. There are many factors explaining the migration trends and the changing composition of immigration, in particular different migration pull and push factors, barriers to migration, as well as differences in propensity to stay in the destination, and return migration behaviors. This entry provides an overview of major determinants of international migration.

Basic theory

The large majority of studies investigating what motivates both internal and international migration are rooted in the human capital investment theory of migration, wherein immigration is seen as a form of investment in human capital (Sjaastad, 1962; Todaro, 1969). Perfectly rational and fully informed potential migrants assess the expected discounted returns to and costs of migration, and move to destinations where the discounted net return to migration is greatest. The basic, also called the neo-classical, model was later extended for the probability of being employed in destination areas, to account for the uncertainty of finding a job after migration (Todaro, 1969; Harris and Todaro, 1970). From the macro-economic perspective, the gravity theory of migration was developed to explain aggregate migration flows between regions and countries. The model is based on Newton’s 1687 law of gravity, and in the

Empirical specifications

To motivate empirical model specification, the empirical literature usually employs some type of gravity model in earlier literature, or some type of human capital investment model in more recent literature. Both, however, arrive at a similar empirical specification. The existing empirical literature usually utilizes aggregated migration data on a regional or country level. Earlier literature started with studying inter-regional migration, followed by analyses of international migration. The latter began with studies analysing immigration into one country, and later, as international migration data became more readily 77

78  Elgar encyclopedia of labour studies

available, came analyses of migration to and from multiple countries. Empirical studies on determinants of migration typically estimate a gravity-type model that includes a number of push/pull factors in origins and destinations and distance measures. They typically work with a version of the following equation: ​In  ​mijt​  ​  ​=  F​(​ ​GDP​i ​, ​GDP​j ​, ​Unemp​i ​,​

​Unemp​j ​, ​Dij​ ​, ​Stock​ij ​, ​Xijt​  ​, ​δi​ ​, ​δj​ ​, ​θt​ ​)​

where the dependent variable ​ ​mijt​  ​​ denotes the migration rate defined as gross or net flows of migrants from country i to country j divided by the population of the country of origin i at time t. Independent variables include measures of economic pull and push factors on destination and origin country level, respectively, such as income per capita ) and unemployment rates (​G ​ DP​j ​​; ​​GDP​i ​​ (Unemp), and measures of distance as a proxy for costs of migration, such as physical distance between countries in km and common border, or cultural and linguistic distances (​​ D​ij ​). In addition, researchers often control for the existence of migrants’ networks, which is often proxied by the stock of migrants (​​ Stock​ij ​) from origin i who live at destination j. Furthermore, models often include a set of controls Xijt capturing political, environmental and institutional push/pull factors. Finally, models using panel data typically include origin (​​ ​ ​δi​ ​)​​, destination country ​​(​ ​δj​ )​ ​​

low income levels in origin countries serve as a push factor increasing emigration from the home country. At the same time, low income levels in origin countries can also make people less mobile, as some potential migrants from poorer environments face the so-called poverty constraints, that is, potential migrants cannot afford to pay migration costs and are forced to stay at home (Clark et al., 2007; Pedersen et al., 2008; Kennan and Walker, 2011; McKenzie et al., 2014; Batista and McKenzie, 2021). Furthermore, existing studies confirm employment opportunities as often proxied by the unemployment rates in destination and origin countries acting as significant push and pull factors, with greater weight on the destination’s job opportunities as a pull factor (see, e.g., Hatton, 1995; Pedersen et al., 2008). Another pull factor for migration that has received considerable attention both among policy makers and in academia is the generosity of welfare systems in destination countries, the so-called ‘welfare magnet’. Countries with generous welfare benefits may attract certain types of immigrants who are more likely to rely on welfare payments to support themselves. Furthermore, welfare benefits may act as an insurance against potential unemployment or health shocks that immigrants are likely to face after arriving. Welfare benefits may thus reduce the costs and risks associated with migration, and provide larger incentives to migrate. However, the empirical evidence on the causal effects of welfare benefits on migration flows is not conclusive (Borjas, 1999; Giulietti et al., 2013).

or pair-wise ​​(​ ​δij​ ​)​​fixed effects, time effects​​ (​ ​θt​ ​)​to account for unobserved factors. In the following section, we discuss the empirical evidence on the importance of dif- Migrants’ networks ferent migration determinants coming from The ‘networks’ of migrants (family members, the existing empirical evidence. friends and people from the same source country) may reduce both their direct and psychological migration costs as well as the Empirical findings need to learn the local language. Through networks potential migrants receive information Economic factors about the immigration country, on the likeliPerhaps one of the most discussed determi- hood of getting a job, on economic and social nants of migration in the literature are the systems, immigration policy, people and net income differentials and income inequal- culture. The community may also provide ity between origin and destination countries. public services, such as language training and There is a consensus in the empirical literature education of children that better fit needs of that higher income levels in potential destina- the newcomer and their family (Bertrand et tion countries serve as a pull factor, attracting al., 2000; Munshi, 2003). migrants from low-income countries, while Mariola Pytlikova and Davit Adunts

International migration  79

There is a consensus in the migration literature that migration is greater to destinations with a larger stock of individuals from the same origin (Pedersen et al., 2008). Empirical evidence suggests that diasporas explain a large part of the variability and selection in migration flows (McKenzie and Rapoport, 2010), as well as the diminishing importance of other migration determinants with larger networks (Palmer and Pytliková, 2015; Gorinas and Pytliková, 2017). Network effects also explain the persistence of migration flows even after the original push and pull forces that initiated migration have disappeared or diminished. Cultural and linguistic barriers Differences in language and culture between countries imply costs that potential migrants likely consider in deciding whether to migrate and where to go (Ortega and Peri, 2013; Adsera and Pytlikova, 2015). Fluency in the language of the destination country – and ease of learning it – can facilitate the transfer of migrants’ skills to the new labour market, contributing to the global interchange of skills and stimulating economic growth. In fact, Adserà and Pytliková (2015) show that linguistic proximity matters more for migration flows from source countries with better-educated populations, which is in line with results from Belot and Hatton (2012), who find positive effect of language proximity on skill selectivity. A greater need for skill transferability for highly skilled migrants may account for the findings. Beine and Salomone (2013) find that a common official language tends to raise the proportion of skilled migrants at the expense of less-skilled ones, and this holds regardless of the gender of migrants. The role of ethnic conflicts and wars Ethnic conflicts and wars induce flows of refugees out of the affected areas. This is unfortunately a common headline in the current news. How important the role of political instability, ethnic conflicts and wars is in explaining migration flows, however, is not well understood, due to the rather limited empirical literature (e.g. Abel et al., 2019). Some previous studies on migration determinants that include variables on civil and political rights of citizens in their country

of origin, such as Freedom House indexes, show that these are important in explaining migration (Pedersen et al., 2008; Adsera and Pytlikova, 2015). What is not entirely understood, however, is the role of different types of conflicts and wars on migration, the dynamics of politically induced migration flows, and the interplay of political factors with other pull and push factors and policy measures. Hopefully, future research will shed light on this. Environmental factors Recently, the question of whether and how environmental factors influence migration has gained both academic and public interest. This is reflected in a large and growing number of recent contributions focusing on the role of climate change and variability, natural disasters and environmental pollution on migration. The evidence is not conclusive, and differs: (1) across different countries and between internal versus international migration; (2) according to which type of data and measures of environmental changes and migration are used; and (3) according to the methodologies applied. Some studies focus on rural‒urban within-country migration (Thiede and Gray, 2017; Mikula and Pytlikova, 2021), and only a limited number of studies try to tackle environmentally induced international migration (Cai et al., 2016, among others). Thus, there is little consensus among the studies regarding both the direction and the magnitude of such effects. Relatively little is also known about the role of different mechanisms. The role of the state The state influences a number of areas, which may affect migration. First, immigration policy regimes and bilateral visa restrictions limit the inflow of international migrants, and influence their length of stay and return. The extent to which a country protects its borders and imposes its immigration policy restrictions is important; countries which are known to be relatively easy to enter may be more attractive to potential migrants. In contrast, when countries enforce their immigration policy (such as after a major recession), the costs of migrating may become sufficiently high to discourage migration. In most theoretical models of migration, immigration Mariola Pytlikova and Davit Adunts

80  Elgar encyclopedia of labour studies

policy serves as a proxy for migration costs. Countries with a relaxed immigration policy will be less costly to move to, while countries with a tight policy will be costlier. Selective immigration regimes, such as a points-based system, affect both the number as well as the composition of migrants. There are few empirical contributions on this topic, mainly due to data limitations. For instance, Mayda (2010) investigates how changes in the strictness of entry requirements affected the size and direction of migration flows to 14 developed countries during 1980–1995. Stricter policies that require applicants to have higher-level skills and some knowledge of the national language (such as in Australia or Canada) may result in a more positively selected pool of migrants, and decrease the relevance of traditional pull and push factors (such as differences in income per capita) in explaining migration flows. The study finds that stricter immigration rules reduce the relevance of push factors more than pull factors (such as linguistic closeness), and as a result, cultural and linguistic differences may become more relevant after such a policy change (Mayda, 2010). Another study finds that shifting to a points-based immigration system raises the share of high-skilled workers in the pool of migrants by about six percentage points. As a result, introducing a skill-based immigration policy is likely to increase the degree of proficiency in the national language among new arrivals (compared with the absence of such a policy), and to shift the composition of new immigrants toward origin countries whose languages are the same or relatively close to that of the destination country (Belot and Hatton, 2012). Furthermore, the state also influences the extent of rights ‒ economic, legal, social ‒ that migrants receive at their destinations. Palmer and Pytliková (2015) show that migration between origin‒destination pairs is positively associated with the loosening of destination labour market restrictions on new immigrants. In addition, they find that the interplay of change in policy of different countries distorts migration: specifically, they find that migration to a given country is negatively associated with the loosening of competing destinations’ labour market restrictions. The influence of destination labour market access seems to be weaker for destinations where migrants have larger networks, and for Mariola Pytlikova and Davit Adunts

migrants from countries who speak languages more similar to the destination language. Natives’ attitudes towards immigrants Recent literature reveals natives’ hostility towards migrants to be an important deterrent factor for potential migrants, in particular for labour migrants (Gorinas and Pytliková, 2017). Thus, potential migrants might view natives’ hostility at a potential destination as a barrier for their integration and success in the labour market and society. Compared to other traditional migration determinants, natives’ attitudes towards migrants seem to have a lesser effect than that of migrants’ networks and the destination’s gross domestic product pull factor, but greater than that of the destination’s unemployment rate (Gorinas and Pytliková, 2017). Risk-aversion, income diversification and relative deprivation Empirical studies also emphasize the importance of risk-aversion, income diversification and relative deprivation in migration decision making processes. The decision to migrate is sometimes made at the family or household level rather than at the individual level (Stark and Bloom, 1985). In most developing countries market failures – labor market, credit market or insurance market ‒ are quite common. Households may consider sending one of the household’s members to work abroad to insure against possible negative shocks in local economic conditions. Indeed, Chen et al. (2003) find that the market correlation and risk diversification between origin and destination countries are important determinants of the migration decision. Dustmann et al. (2020) find that risk attitudes of individual migrants as well as the risk attitudes of other household members affect rural-to-urban migration decisions in China. Furthermore, migration may occur to increase income relative to other households in the origin community, rather than to improve household income in absolute terms (Stark and Taylor, 1989). Incomplete information Most studies on migration determinants assume that potential migrants have full information about the costs, risks and returns to migration before making a migration

International migration  81

decision. However, in practice, migrants are likely to have incomplete or inaccurate information about their expected earnings, as well as about the risks associated with migrating to a given destination (see Bertoli et al., 2020 for a detailed discussion). Furthermore, there might be significant costs of acquiring necessary information about all destination countries. Therefore, an individual’s decision on where or whether to migrate might be distorted by the availability and costs of information about different characteristics of destination countries. Empirical evidence about the importance of information and information costs in migration decision making is relatively new and scarce, but seems to confirm that misinformation about the expected income and risks of migration lowers actual migration (McKenzie et al., 2013; Shrestha, 2020; Batista and McKenzie, 2021).

Summary

The increase in the volume of migration and the changes in the mix of typical migrants’ receiving and sending countries explain the interest of policy makers and academia in understanding the determinants of migration. This entry summarizes the main theories and empirical evidence on the pull and push factors of migration. There is a consensus that high incomes, employment opportunities, large ethnic networks, and cultural and language proximity in the destination countries, explain a large fraction of migration flows. However, the impacts of other factors, such as welfare generosity in destination countries and environmental factors, on migration flows are either unclear or much smaller in magnitude (such as the effects of natives’ attitudes) compared to factors such as high income levels and ethnic networks. Furthermore, migrants can be ‘pushed’ out of their origin countries due to low incomes, lack of jobs, ethnic conflicts, and wars. While there is extensive evidence on both economic and non-economic factors of migration, empirical evidence on the effect of immigration policies is relatively meagre. Few studies focus on explicit policies, while the effects of implicit policies on the selection and sorting of migration are rarely examined; Lazear’s (2021) study is an important exception. Future studies may address this issue, with the availability of better datasets on the

measures of migration policies and international migration flows. Furthermore, most studies of the determinants of migration focus on immigration into high-income countries. However, immigration to middle-income countries is growing, and countries such as China and Turkey are quickly catching up. Future studies may provide more empirical evidence on drivers of migration in those countries. Finally, future research on migration determinants is likely to shed more light on the role of environmental changes and political factors, such as conflicts and wars, and the interplay of various mechanisms, as the evidence on these is still rather thin and inconclusive. Mariola Pytlikova and Davit Adunts

Acknowledgments

The work on this entry was funded in part by the Czech Science Foundation grant on ‘International Migration: Drivers and Impacts’ (GA20–31615S) and by the NPO ‘Systemic Risk Institute’, grant no. LX22NPO5101.

References

Abel, G.J., M. Brottrager, J. Crespo Cuaresma and R. Muttarak (2019), ‘Climate, conflict and forced migration’, Global Environmental Change, 54, 239–249. Adsera, A. and M. Pytlikova (2015), ‘The role of language in shaping international migration’, Economic Journal, 125(586), F49‒F81. Batista, C. and D.J. McKenzie (2021), ‘Testing classic theories of migration in the lab’, IZA DP No. 14717. Beine, M., S. Bertoli and J. Fernández-Huertas Moraga (2016), ‘A practitioners’ guide to gravity models of international migration’, World Economy, 39(4), 496‒512. Beine, M. and S. Salomone (2013), ‘Network effects in international migration: education versus gender’, Scandinavian Journal of Economics, 115(2), 354–380. Belot, M.V.K. and T.J. Hatton (2012), ‘Immigrant selection in the OECD’, Scandinavian Journal of Economics, 114(4), 1105–1128. Bertoli, S., J.F.H Moraga and L. Guichard (2020), ‘Rational inattention and migration decisions’, Journal of International Economics, 126, 103364. Bertrand, M., E.F.P. Luttmer and S. Mullainathan (2000), ‘Network effects and welfare cultures’, Quarterly Journal of Economics, 115(3), 1019–1055.

Mariola Pytlikova and Davit Adunts

82  Elgar encyclopedia of labour studies Borjas, G.J. (1987), ‘Self-selection and the earnings of immigrants’, American Economic Review 80(1), 305–308 Borjas, G.J. (1999), ‘Immigration and welfare magnets’, Journal of Labor Economics, 17(4), 607–637. Cai, R., S. Feng, M. Oppenheimer and M. Pytlikova (2016), ‘Climate variability and international migration: the importance of the agricultural linkage’, Journal of Environmental Economics and Management, 79(C), 135–151. Chen, K.P., S.H. Chiang and S.F. Leung (2003), ‘Migration, family, and risk diversification’,  Journal of Labor Economics,  21(2), 353–380. Clark, X, T.J. Hatton and J.G. Williamson (2007), ‘Explaining US immigration, 1971‒99’, Review of Economics and Statistics, 89(2), 359‒373. Dustmann, C., F. Fasani, X. Meng and L. Minale (2020), ‘Risk attitudes and household migration decisions’, Journal of Human Resources, 1019-10513R1. Gallup World Poll (various years), https://​www​ .gallup​.com/​analytics/​318875/​global​-research​ .aspx. Giulietti, C., M. Guzi, M. Kahanec and K.F. Zimmermann (2013), ‘Unemployment benefits and immigration: evidence from the EU’,  International Journal of Manpower, 34(1), 24–38. Gorinas, C., and M. Pytliková (2017), ‘The influence of attitudes toward immigrants on international migration’, International Migration Review, 51(2), 416–451. Grogger, J. and G.H. Hanson (2011), ‘Income maximization and the selection and sorting of international migrants’, Journal of Development Economics, 95(1), 42–57. Harris, J.R. and M.P. Todaro (1970), ‘Migration, unemployment and development: a two-sector analysis’,  American Economic Review, 60(1), 126–142. Hatton, T. (1995), ‘A model of U.K. emigration, 1870–1913’, Review of Economics and Statistics, 77(3), 407–415. Kennan, J. and J.R. Walker (2011), ‘The effect of expected income on individual migration decisions’, Econometrica, 79(1), 211–251. Lazear, E.P. (2021), ‘Why are some immigrant groups more successful than others?’, Journal of Labor Economics, 39(1), 115–133, Mayda, A.M. (2010), ‘International migration: a panel data analysis of the determinants of bilateral flows’, Journal of Population Economics, 23(4), 1249–1274. McKenzie, D., J. Gibson and S. Stillman (2013), ‘A land of milk and honey with streets paved with gold: do emigrants have over-optimistic expectations about incomes abroad?’, Journal of Development Economics, 102(C), 116–127.

Mariola Pytlikova and Davit Adunts

McKenzie, D. and H. Rapoport (2010), ‘Self-selection patterns in Mexico‒U.S. migration: the role of migration networks’, Review of Economics and Statistics, 92(4), 811–821. McKenzie, D., C. Theoharides and D. Yang (2014), ‘Distortions in the international migrant labor market: evidence from Filipino migration and wage responses to destination country economic shocks’, American Economic Journal: Applied Economics, 6(2), 49–75. Mikula, S. and M. Pytlikova (2021), ‘Air pollution and migration: exploiting a natural experiment from the Czech Republic’, IZA Discussion Paper no. 14863. Munshi, K. (2003), ‘Networks in the modern economy: Mexican migrants in the U. S. labor market’, Quarterly Journal of Economics, 118(2), 549–599. Ortega, F., and G. Peri (2013), ‘The effect of income and immigration policies on international migration’, Migration Studies, 1(1), 47–74. Palmer, J.R.B and M. Pytliková (2015), ’Labor market laws and intra-European migration: the role of the state in shaping destination choices’, European Journal of Population, 31, 127‒153. Pedersen, P.J., M. Pytlikova and N. Smith (2008), ‘Selection and network effects – migration flows into OECD countries 1990–2000’, European Economic Review, 52(7), 1160–1186. Roy, A.D. (1951), ‘Some thoughts on the distribution of earnings’, Oxford Economic Papers, 3(2), 135–146. Shrestha, M. (2020), ‘Get rich or die trying: perceived earnings, perceived mortality rates, and migration decisions of potential work migrants from Nepal’, World Bank Economic Review, 34(1), 1–27. Sjaastad, L.A. (1962), ‘The costs and returns of human migration’, Journal of Political Economy, 70(5, Part 2), 80–93. Stark, O. and D.E. Bloom (1985), ‘The new economics of labor migration’, American Economic Review, 75(2), 173–178. Stark, O. and J.E. Taylor (1989), ‘Relative deprivation and international migration’, Demography, 26, 1–14. Thiede, B.C. and C.L. Gray (2017), ‘Erratum to: Heterogeneous climate effects on human migration in Indonesia’, Population and Environment, 39, 173–195. Todaro, M.P. (1969), ‘A model of labor migration and urban unemployment in less developed countries’, American Economic Review, 59(1), 138–148.

J

Japanese labor market

of the Japanese labor market. In particular, how do these compositional changes result in one of the most important recent phenomena in the Japanese labor market: the increase in nonstandard workers?

The Japanese labor market has several distinct macroeconomic features, for example, low and stable unemployment rates, flexible adjustment in hours worked, and so on. According to Organisation for Economic Co-operation and Development (OECD) statistics between 1956 and 2020, the annual unemployment rate hit 5.4 percent at its highest in 2002, and there were only five years during 60 years that it exceeded 5 percent. On the other hand, the Japanese labor market uses flexible adjustment of labor input by hours worked and total labor cost. For example, when the financial crisis struck in 2009, the total hours worked and the total labor cost fell by 10 percent and 5 percent, respectively. These adjustment patterns during the recession are known as labor hoarding, and have been practiced in the Japanese labor market for more than half a century (see Hijzen et al., 2015). At the same time, the Japanese labor market has also been changing, especially since the 1990s. The composition of workers has changed, accompanied by the demographic change of the society, such as the decline in total fertility rates from 1.53 in 1991 to 1.32 in 2006. More concretely, female labor participation has risen, along with a disappearance of the M-shaped participation-by-age curve. The rate of female labor force participation for the 25–34 age group grew from 56.6 percent in 1990 to 80.2 percent in 2018. Elders remain in the labor market longer than before. Thus, the labor force participation rate for the 60–64 age group increased from 55.5 percent in 1990 to 70.6 percent in 2018. Enrollment rates for tertiary education moved from around 40 percent during the 1970s to around 60 percent during the 2010s. How these compositional changes have affected the labor markets is the most classical topic in the empirical research

Long-term employment relationships

The changing and unchanged facets of the Japanese labor market mentioned above are strongly related to the institutional background. First, there is the expectation of a long-term relationship between the employer and the employee. The turnover of workers is still low, especially once the workers have passed the initial trial period. The probability of continuing to work for the same employer for more than ten years (ten-year retention rate) is around 80 percent for college-educated men at age 30–34, if they were already employed for more than five years. These figures are notably higher and more stable than, for instance, those in the United States (Kambayashi and Kato, 2016a). The annual separation rates due to dismissals are at most around 2 percent of employment, even during the financial crisis in 2009. It should be noted that low dismissal rates are based on implicit agreements between employers and employees, as the governmental regulation plays at most an indirect role. The OECD’s Employment Protection Legislation Index shows that the strictness of dismissal regulation in Japan is weaker than the OECD average. Second, the social norm that mutually respects the long-term relationship enables the steep tenure‒wage profile. Theoretically, a deferred payment schedule can provide workers with the incentive to work and to accumulate in-house skill formation. In Japan, the deferred payment schedule is reinforced by the bonus system and severance payments. According to the Basic Survey on Wage Structure in 2019, about 20 percent 83

84  Elgar encyclopedia of labour studies

of annual earnings are paid as a bonus for an average worker. The General Survey on Working Conditions in 2018 reports that the average amount of severance payment for college-educated workers of mandatory retirement age was equivalent to 38.6 months salary. Because the amount of bonus and severance payments is decided at the discretion of employers, these are not committed to in labor contracts in advance; the deferred payment schedule becomes steep and is a strong incentive for workers in the Japanese labor market (Hashimoto and Raisian, 1985). Workers’ incentives are important for improving productivity of workplaces in Japan. Especially in manufacturing, the accumulation of multitasking and contextual skills enable workers not only to flexibly help neighbors in their workplaces, but also to respond to problems in situations not described in the instruction manual (Koike, 2002). Third, the long-term relationship is also complemented with three-tier communication channels between the employer and the employee. At the top of the tier, there is formal and collective communication between the labor union and the employer that is regulated by the Labor Union Act. The employer has a legal obligation to communicate with the labor union about specific topics such as wages, hours, and so on. In the middle, many companies construct a labor management council (LMC). This is a collective but informal channel of communication, in the sense that the LMC does not have any legal basis. Usually, information about business conditions and training strategy is shared between employers and employees through this channel. At the bottom, informal but more individual-based meetings are organized at the workshop level. Often called a workshop committee, information about everyday activities, such as the idea of improving productivity, is shared among the managers and their subordinates through this channel. These three channels are bundled as a communication tool to create common knowledge and mutual monitoring of employers and employees. These institutional arrangements are often summarized as the ‘Japanese employment system’. It is also classified as a high-performance work system, and is regarded as a source of improvement in the productivity of Japanese firms (Kato and Morishima, 2002). Ryo Kambayashi

Standard versus nonstandard workers

In contrast, because the Japanese employment system covers only a portion of workers, there are severe backlashes in the Japanese labor market. First, there is a sharp duality between standard and nonstandard workers in the workplace. Not only wages, but also actual hours worked, promotion opportunities, and training programs are differently distributed among these two categories. Moreover, since the Japanese employment system works well, with mutual communication between employer and employees, its coverage is also determined by informal private agreements, not by public third-party clarification. In other words, the divide between standard and nonstandard workers is determined by a de facto definition, such as titles in the workplace, rather than by a de jure definition, such as contract length. Based on de facto definition, the hourly wages of nonstandard workers are about 15‒25 percent lower than those of standard workers, whereas the wage differences due to de jure definition is less than 5 percent (see Kambayashi and Kato, 2016b). Second, the divide between standard and nonstandard workers is in accordance with the large gender gap in the Japanese labor market, since the majority of standard workers are male, whereas the majority of nonstandard workers are female. One interpretation is that the disparity of two employment categories is consistent with the comparative advantage model in a household. Given the large wage disparity in the labor market, the division of labor between a breadwinner and a housewife has been a rational choice of a household. As a result, the labor supply of nonstandard workers is highly elastic, and employers do not expect a long-term relationship with them. Therefore, the disadvantage in the promotion and training opportunities for nonstandard workers is directly reflected in the disadvantage of women in Japanese society. Third, while standard workers receive better working conditions in terms of long-term relationship, higher wages, and greater promotion opportunities, they must accept long working hours, which gives rise to work‒life balance problems. According to the Labor Force Survey, in 2019, the percentage of employees who work more than 49 hours a week is 30.7 percent for

Japanese labor market  85

male standard workers, while 11.4 percent work more than 60 hours. The long working hours clearly reduce the time available for housework and to care for family members. In 2016, the average Japanese man used only 45 minutes per day for housework, whereas the average Japanese woman spent 218 minutes on it (Japanese Time Use Survey). Long working hours also increase the risk of death from overwork (karoshi). The Japanese government has officially admitted that about 100‒150 deaths per year are caused by overwork that should be compensated by Industrial Accident Compensation Insurance. Furthermore, the disparity in the Japanese labor market is related to industrial structure, especially firm size. While standard workers in larger firms receive better working conditions based on the long-term relationship, workers in smaller firms accept worse working conditions regardless of their titles. According to the Employment Status Survey, the share of workers who work in large firms with over 1000 employees was 17.9 percent in 2017, while that of workers who work in small firms with under 20 employees was 27.0 percent, including the self-employed. Although part of the disparity between firm sizes is due to differences in productivity, the economic reasons are not fully explored.

Weak government regulation in motion

To provide an overview of the Japanese labor market from the point of view of the institution, the weakness of governmental intervention should be noted. Although Japan has strict legal regulations in many ways, collective agreements between employers and employees can counteract many regulations. A typical example is the overtime regulation: while the Labor Standard Act regulates the maximum working hours as eight hours per day and 40 hours per week, if there is a collective agreement employers can order overtime work up to an additional 45 hours per month. Courts are likely to give priority to mutual agreements between employers and employees, rather than to evaluate conflicts from the point of view of the social norm outside the firm. However, the labor market mechanism based on mutual agreements between employers and employees has recently been

shaken up. First, there has been an increase in the large gap where mutual agreements do not work. The informal sector, including the self-employed, has continuously shrunk to about 10 percent of the working population in the 2010s. On the other hand, the Japanese employment system could not expand to cover the increase in employees, since the share of standard workers within the working population has remained constant over several decades. This implied an increase in nonstandard workers created by the shrinking self-employment. Second, governmental intervention is increasing. For example, the level of the minimum wage relative to the median wage of full-time workers has increased rapidly. It is 45 percent in 2020, while it had been around 30 percent from the 1980s to the 2000s. The Law for Employment Promotion of Persons with Disabilities was revised in 2013 to not allow any exemption from employers’ obligation to employ disabled workers. The relation between Japanese institutions and the labor market is a research area that has still not been fully investigated. Ryo Kambayashi

References

Hashimoto, M. and J. Raisian (1985), “Employment tenure and earnings profiles in Japan and the United States,” American Economic Review, 75(4), 721–735. Hijzen, A., R. Kambayashi, H. Teruyama, and Y. Genda (2015), “The Japanese labour market during the global financial crisis and the role of nonstandard work: A micro perspective,” Journal of the Japanese and International Economies, 38(C), 260–281. Kambayashi, R. and T. Kato (2016a), “Long-term employment and job security over the past 25 years: A comparative study of Japan and the United States,” Industrial Labor Relations Review, 70(2), 359–394. Kambayashi, R. and T. Kato (2016b), “Good jobs, bad jobs in Japan: 1982–2007,” Center for Japanese Economy and Business (Columbia University), Working Paper Series, No. 348. Kato, T. and M. Morishima (2002), “The productivity effects of participatory employment practices: Evidence from new Japanese panel data,” Industrial Relations, 41(4), 487–520. Koike, K. (2002), “Intellectual skills and competitive strength: Is a radical change necessary?,” Journal of Education and Work, 15(4), 391–408.

Ryo Kambayashi

Job design

costs to specialization, and benefits to multitasking. One cost may be poor motivation: Smith noted that a specialized worker may become bored performing a few repetitive tasks. I return to that point below. A second cost of specialization is coordination problems. Many parts of a business process need to be integrated with each other. When two employees perform related tasks, there is a risk of poor coordination. Suppose two employees manufacture engine parts that work closely together (for example, a piston and a cylinder). By making only one part, neither employee is likely to fully understand how their part fits and functions with the other part. That may lead to quality problems. If, instead, one employee manufacturers both parts, quality may be higher. An important type of coordination problem is imperfect communication. By performing some task (for example, speaking with a customer in a jewelry store) an employee may gain information that is relevant to a colleague who performs other tasks (the jewelry repair specialist). Communication takes time, and information may be garbled when it occurs. If, instead, the tasks are combined, no communication is necessary. Multitasking often has benefits because some tasks are complementary to each other. There are several considerations. First, tasks may be complements in skill requirements. If several tasks require the same or closely related skills, it may make sense to put them in the same job in order to economize on investments in human capital. For example, accountants may also provide financial advice to their clients, because the concepts, tools, and methods for both are closely related. Second, tasks may be complements in their effects: output from one may be a key input into the other. For example, a professor’s research may improve teaching, and vice versa, if the assigned courses match the research area. A closely related idea is that multitasking may generate ”intertask learning” (Lindbeck and Snower, 2000), in which performing related tasks generates learning that improves productivity across these tasks. This can arise because multitasking provides a broader perspective on how various tasks fit together in the overall business process. Moreover, it allows the employee to abstract general principles about how to do the work, which might be applied to other tasks.

Introduction

How should a firm design jobs? What are the key job characteristics? What are the key goals of job design? This chapter provides a brief overview of the economics of job design. The core elements of a job include the number and types of tasks assigned, the extent to which autonomy is granted, and the degree of collaboration with others. The goals of job design include productivity, coordination, quality, learning (of various types), adaptation, and motivation. These goals generate important tradeoffs in job design, and how it may vary with the circumstances of the work environment.

Specialization versus multitasking

Begin with the most basic question: which task(s) should be assigned to this job? Perhaps the most famous analysis of job design was the first chapter of Adam Smith’s (1776) Wealth of Nations, which provided an argument for specialization: assigning only one or a very few tasks to each employee. He described a manufacturer of “pins” (nails) that divided the 18 steps in the process among ten employees; all performed only one or two tasks. According to Smith, when compared to the traditional approach in which artisans performed all tasks in the nail manufacturing process, productivity soared with specialization. Specialization is a powerful force in job design, for several reasons. First, the firm can better match comparative advantages in employee skills to job requirements if jobs are narrowly defined. Second, the firm can economize on training costs since each employee needs to learn fewer things. Moreover, skills will increase more quickly, and possibly to a greater extent. Third, focus may lead to greater improvements in the process as the employee becomes an expert in the few assigned tasks. Fourth, the employee saves time that may be lost in switching between tasks. Similarly, switching between multiple tasks can easily be cognitively challenging. For these reasons, most jobs involve significant degrees of specialization. This logic suggests separating all tasks into distinct jobs. However, there are also 86

Job design  87

Task complementarity helps explains why most jobs are not highly specialized. It also provides a framework for thinking about which tasks to bundle together in the same job. Finally, it suggests that an important step in optimal organizational design is to seek modularity in the overall business process. All else equal, a firm should bundle closely related tasks together, and assign them to the same person or team of close colleagues. The firm will still face the problem of coordinating the activities of each module. However, by effectively modularizing the business process, it can reduce these issues as much as possible.

Centralization versus decentralization

A special type of task deserves its own analysis: decision making. Job design specifies not just which tasks an employee performs, but also the extent to which the firm decentralizes, granting autonomy to the employee, or centralizes. This depends on the importance, location, and evolution of knowledge (broadly defined) within the firm, as well as communication costs. On one end of the spectrum, a firm may adopt a highly centralized approach in which the employee is expected to perform their tasks with close adherence to standard rules and procedures. Such an approach was pioneered by Frederick Taylor (1923). Under Taylorism, a firm would use experts (“industrial engineers”) to figure out best practices for performing each task. To the extent that this can be done, it follows that employees should be expected to follow those best practices. Job design would then involve very little autonomy. At the other end of the spectrum, the job may grant the employee significant autonomy in how to perform the work, and even the goals of the work. For example, an investment banker is expected to seek new clients and develop new products and services. An engineer in a firm’s research and development (R&D) group may conduct basic research with the hope that it might someday lead to process improvements or new products. An employee at a luxury hotel may be expected to continually identify and solve unique problems as they arise for guests.

Think of these two extreme approaches as ex ante or ongoing optimization. A primary goal of organization design—and therefore of job design—is to create, locate, combine, and use knowledge effectively. If a firm can develop best practices and codify knowledge, then it should certainly do so, and that will tend to lead to jobs with less autonomy, except for the specialists who develop those practices. Investing in understanding best practices is costly, and will be optimal only if there is adequate return on that investment. UPS made extensive use of Tayloristic methods over the twentieth century. That made sense because the business of package delivery is very simple, and has changed little over a long period of time. If a business is more complex, the cost of ex ante optimization will be higher. If the business is more dynamic, unpredictable, and evolving, the return on investment in ex ante optimization will be lower. An additional consideration is that knowledge and opportunities for learning are often widely dispersed among employees throughout the organization. For example, employees may have intangible understanding of each customer’s mood and preferences. That information is valuable to the business, since it can be used to improve customer service. In principle it might be communicated to the center of the organization (which might improve coordination), but communication takes time, and may be costly and imperfect (Jensen and Meckling, 1992). Thus, when a business is more complex, uncertain, and evolving, the firm is less likely to succeed at ex ante optimization, so employees are more likely to generate valuable knowledge as a result of performing their jobs. In such a context, especially when knowledge is costly to communicate, job design is likely to grant greater autonomy in order to develop and use that dispersed and dynamic knowledge more effectively. This emphasizes continuous improvement rather than ex ante optimization. These observations also explain a widely observed phenomenon: there has been a growing trend towards adoption of jobs which involve more multitasking, participation in decision making, and emphasis on continuous improvement. That is almost certainly driven by the increasingly rapid pace of Michael J. Gibbs

88  Elgar encyclopedia of labour studies

innovation with the information technology revolution, as well as greater complexity and uncertainty due to deregulation and globalization of business. Of course, firms combine ex ante optimization and continuous improvement. Consider the modern practices of Six Sigma and Total Quality Management (TQM). The philosophy behind Six Sigma is Tayloristic: identify best practices, and then implement them with as little variation as possible. However, in contrast to Taylor’s approach, development of best practices is decentralized via TQM, in which employees are trained in problem solving techniques, and granted autonomy to diagnose problems, develop and test solutions, and implement them.

Teams

Another special type of task is collaboration. Virtually all employees collaborate with colleagues to some extent, in order to coordinate their work and share relevant knowledge and skills. To the extent that this coordination is regular and extensive, the job design may include assignment to work closely with colleagues in a team. A useful way to think about a team is as a method by which a firm can “extend the person.” Our analysis above suggests that optimal organization design involves modularity of the business process. It will often be the case that an optimal module is too complex for a single employee to handle effectively. In such a case, the firm might assign that module to a team which works closely together. Team members will develop relationships, culture, and methods to reduce communication costs, improve coordination, and foster intertask learning. Job rotation within the team may reinforce these benefits, as over time each employee accumulates an understanding of the full picture of the team’s work. An additional aspect of “expanding the person” via teams is that it expands the skills, knowledge, and perspectives that will be combined. Diversity of perspectives can in principle improve innovation and learning (subject to the caveat that it may also increase communication and collaboration costs within the team).

Michael J. Gibbs

Motivation

Job design affects both extrinsic motivation (behavior driven by rewards such as incentive compensation) and intrinsic motivation (behavior driven by the nature of the work itself). Job design affects extrinsic motivation in two ways. First, some jobs provide significant promotion opportunities through skill development, exposure to senior management, development of useful social networks, and so on. Second, incentive compensation requires effective performance evaluation. Ideally, a firm can develop relatively meaningful and accurate numeric performance measures on which to base rewards. When that is not the case, the firm may have to emphasize subjective evaluation, which tends to lead to weaker incentives due to its less objective nature. Some tasks lend themselves to quantified evaluation, while others do not (for example, total sales revenue versus customer satisfaction for a sales representative). Therefore, one consideration for which tasks to bundle together in the same job is whether or not the tasks have similar evaluation properties (Holmström and Milgrom, 1991). Recall Smith’s concern that specialized jobs may generate low intrinsic motivation. An important literature in social psychology and organizational behavior analyzes how job design can instead generate high intrinsic motivation. Neuropsychologists have found that from birth humans have a strong interest in exploratory activities, and a desire to resolve cognitive pressures from novelty, complexity, and surprise. Thus, learning is arguably the most important driver of intrinsic motivation. To implement this idea, social psychologists suggest “enriched” job design involving a variety of tasks and skills, greater autonomy, and feedback (Hackman and Oldham, 1976). Though these ideas come from outside economics, they are quite consistent with the economic approach to job design, while adding consideration of motivation (Gibbs, 2021). Task variety means multitasking, and economists have emphasized the value of intertask learning. Autonomy means decentralization, which allows an employee to test ideas and implement solutions as they learn. Feedback supports this learning by providing the employee with evidence about perfor-

Job design  89

mance, in order to assess the effects of their actions and decisions. Designing a job to foster learning can be powerful, as it may yield two benefits. First, it increases motivation, which is generally beneficial. Second, that motivation is of a particular kind, in which the employee is more cognitively engaged with learning about their work, which may significantly enhance innovation and continuous improvement.

Complementarity

A theme of organizational economics research is complementarity: the idea that a firm’s policies can and should be mutually reinforcing, so we may see firms adopt sets of policies in a few common patterns (Milgrom and Roberts, 1995). This idea also applies to job design. To see why, recall the two broad approaches to optimizing a business process, which I described above: ex ante optimization and continuous improvement. To the extent that a firm can effectively develop best practices, it should implement them. In order to develop best practices, firms often break business practices down into discrete steps, and then attempt to fully understand and perfect each step. Doing so naturally tends to lead to more specialization in job design, as one or a few discrete steps are assigned to one employee. Moreover, employees will tend to be granted less autonomy if best practices are known. In addition, the skills required to perform the job will be narrower. Such jobs are likely to have significantly less intrinsic motivation; but employee learning is not important if ex ante optimization works well. By contrast, when ex ante optimization is not effective, the organization has greater opportunities for ongoing learning. That suggests the value of designing jobs so that employees are intrinsically motivated to learn. In such situations, firms are more likely to deploy multitasking, multiskilling, and decentralization. They may also make more extensive use of teamwork, as well as policies designed to encourage collaboration between employees and management (Ichniowski and Shaw, 2003). We can therefore see two broad patterns for job (and organizational) design (Lindbeck and Snower, 2000). In the first, employees have more of a technical role, performing tasks in ways in which they have been

trained. In the second, they are a fundamental part of organizational learning.

Directions for future research

The economy theory of job design is relatively well understood, as it is a culmination of active research in a variety of areas. Ironically, however, little economic theory focuses explicitly on the question of how to design a job. Further work could certainly be done on exploring characteristics of tasks which increase complementarity, how firms might bundle tasks into modules, and how teams might be organized. In addition, much more work could be done on trying to understand how firms can design jobs to increase learning (driven by specialization or by intertask learning). There are extensive opportunities for empirical research on job design, which has not been extensive to date. Most studies draw from detailed data on jobs in one or a small number of firms (as is the case in much empirical personnel economics research), since it is very difficult to construct cross-firm datasets with detailed and consistently measured within-firm data. It is hoped that government agencies will start to collect data of this type so that we can get a better sense of broad patterns in job design (e.g., Gibbs et al., 2010). Michael J. Gibbs

References

Gibbs, M. (2021), Job Design, Learning & Intrinsic Motivation. Working paper, University of Chicago. Gibbs, M., A. Levenson, and C. Zoghi (2010), “Why are jobs designed the way they are?,” Research in Labor Economics 30, 107–154. Hackman, R. and G. Oldham (1976), “Motivation through the design of work: Test of a theory,” Organizational Behavior and Human Performance 16, 250–279. Holmström, B. and P. Milgrom (1991), “Multitask principal‒agent analyses: Incentive contracts, asset ownership and job design,” Journal of Law, Economics and Organizations 7, 24–52. Ichniowski, C. and K.L. Shaw (2003), “Beyond incentive pay: Insiders’ estimates of the value of complementary human resource management practices,” Journal of Economic Perspectives 71(1), 155–180. Jensen, M. and W. Meckling (1992), “Specific and general knowledge and organizational structure,” in Contract Economics, ed. L. Werin and H. Wijkander. Oxford: Blackwell Publishers.

Michael J. Gibbs

90  Elgar encyclopedia of labour studies Lindbeck, A. and D. Snower (2000), “Multi-task learning and the reorganization of work: From Tayloristic to holistic organizations,” Journal of Labor Economics 18(3), 353–376. Milgrom, P. and J. Roberts (1995), “Complementarities and fit: Strategy, structure, and organizational change in manufacturing,” Journal of Accounting and Economics 19(2–3), 179–208.

Michael J. Gibbs

Smith, A. (1776), An Inquiry into the Nature and Causes of the Wealth of Nations. London: W. Strahan & T. Cadell. Taylor, F. (1923), The Principles of Scientific Management. New York: Harper.

Job insecurity

Measurement

Research suggests that people can validly report their expectations of future events, including those surrounding their job (Green, 2006: Chapter 7). Good indicators of job insecurity elicit workers’ ex ante perceptions of their probability of job loss in a given future time span. They can answer on a ranked scale of likelihood, or give probabilities on a numerical scale from 0 to 100. People can also state their perceptions of their chances of regaining employment in a similarly good job, and they can state their perceptions of other forms of insecurity. These expectations are termed subjective, because it is the individual worker who is reporting them. An alternative approach is to measure the end result of insecurity among groups of workers. Common measures that have been used are the fraction of workers (in a group) who, ex post, are made redundant, or the fraction of workers on temporary employment contracts. These measures are termed objective, because they do not depend on individual workers’ reports; yet they are really only indirect proxies for, rather than direct measures of, job insecurity. Direct subjective measures, where available, are therefore better. Fortunately, longitudinal studies show that subjective expectations of job loss match reasonably well the ex post frequency of job loss.

Introduction: types of job insecurity

Job insecurity is the downside component of a job’s ‘prospects’, one of several domains of job quality. The most commonly discussed form of job insecurity concerns the possible loss of a job, with potential pecuniary and non-pecuniary consequences. Suppose there is a certain probability that in, say, the coming year there is a certain chance that the job-holder will be made redundant, and subsequently experience a period of unemployment followed perhaps by re-employment at lower pay. This possibility of job loss lowers the expected present value of the job; it constitutes one aspect of pecuniary job insecurity. This aspect depends on the multiplicative combination of two components: the probability of job loss, and the consequent income loss should redundancy occur (termed the ‘cost of job loss’). Economics studies these components both separately and in combination. Some employees also face the pecuniary risk that their take-home pay will be lowered through variation in the number of hours they are required to work; the greatest risk being faced by those on zero-hours contracts. Another aspect of pecuniary job insecurity stems from the uncertainty itself. The majority of people are risk-averse, so would prefer to know with greater confidence what their future income will be. Those who lose their jobs may lose not only income, but also other valued job features, including something that gives them an identity and which provides satisfaction and meaning. These potential losses constitute non-pecuniary job insecurity. Non-pecuniary aspects are central to less commonly discussed, hidden forms of job insecurity, which stem from the risk that even if the job is kept, valued characteristics of the job may be altered, such as the autonomy afforded, or the extent to which it utilises the worker’s skills and provides interesting work. Employees may also feel insecure about future discrimination or unfair treatment by management.

Determinants and trends

A job is an exchange between an employer and employee in which the latter agrees to work under the former’s direction for a period of time. In the majority of cases, it is presumed that after each period of time that relationship will be renewed. Economics identifies a spectrum between implicit very long-term relationships (sometimes referred to as ‘jobs for life’) offering very high job security, and at the other extreme, short-term ‘spot’ labour markets where each job is highly insecure and may last for only a short time. Most jobs fall in-between these extremes. Employees prefer more job security, because they aim to maximise the extent to which the job is meeting their material and other needs from work. But to employers, a short-term spot market for labour has both benefits and costs. One benefit is that it enables employers to meet variations in 91

92  Elgar encyclopedia of labour studies

market demand. High levels of insecurity also mean that employers have greater bargaining power in negotiating wages. Employees for whom the only alternative is a spell of unemployment may be obliged to accept lower wages. The cost is that employees with no security and only the prospect of a short stay with the company will have little incentive to work hard and acquire job-specific skills. For the latter reason, employers often offer jobs with implicit long-term contracts, in which wages may increase over time, enabling the retention and motivation of valued highly productive workers. When a firm encounters a period of reduced demand, as might occur during a macroeconomic economic downturn, it will prefer to keep employing its high-skilled workers so as not to lose their skills and have to support further training costs later when demand picks up again; it is more likely to make its low-skilled workers redundant, and then rehire as necessary later when demand picks up. Thus, high-skilled jobs tend to be more secure. Given these differing interests, the resulting level of security is partly an outcome of the bargain between employer and employee, where bargaining power is likely to differ according to the skill level, to the strength of unions, and to the state of the external labour market (the local unemployment rate and the availability of jobs). In some countries the labour market partitions into distinct segments of high and low job security, with low mobility between them. Complementing this market-based theory of how jobs vary in their security, government regulation is enormously significant. Governments in the European Union, for example, must comply with the European Fixed-Term Work Directive, which mandated that member states should enact laws which limit employers from endlessly renewing contracts on a fixed-term basis. Governments also regulate the terms of agency work and the use of temporary employment contracts, with some countries imposing substantive limitations on employers’ freedom to terminate employment contracts. The rationale for such social regulation is that job insecurity constitutes a risk to health which has negative external effects; moreover, by hoping to limit unemployment it contributes to meeting the need for social insurance, complementing the provision of unemployment benefits. However, governments may face a trade-off. Francis Green

Too much protection from redundancy may reduce new job openings which provide opportunities for the unemployed. Similarly, strong regulation of open-ended (‘permanent’) contracts, making it very costly for employers to make redundancies, may incentivise employers to design more of their jobs as temporary, thereby increasing the insecurity of those unable to gain permanent jobs. In developing countries, the most significant insecurity divide lies between the formal and in formal sectors of the economy. The consequence of these market-driven and regulation-driven divisions is inequality in the experience of insecurity among the workforce. The evidence from survey studies finds considerable variation between different socio-economic groups. For the large majority of countries, being better educated significantly lowers the chances that a worker is in an insecure job. There is substantial variation among countries, related to their employment protection laws. On the whole, job insecurity is somewhat lower for men than for women, but this is not the case everywhere. Age is another factor, but its effect varies between countries: in East Asian and transitional economies, younger workers are less concerned about insecurity; while in countries bordering the Mediterranean it is the young who are much more likely to be in insecure jobs. An argument of some sociological accounts of job insecurity is that for several decades there has been a prevailing trend towards increasing insecurity in many countries, diminishing the proportion of jobs where job security is high. The prime explanations are thought to be increased globalisation, reduced union strength, and sometimes an erosion of regulatory protections. These tendencies may have reduced the bargaining strength of employees, leading to more of them agreeing to accept insecure jobs. Nevertheless, job security regulation persists in many parts of the world, notably Europe. By contrast, accounts of job insecurity in economics emphasise the external job market. In this perspective, insecurity is expected to follow the trend in the unemployment rate, which in turn depends on macroeconomic growth, the business cycle, and the risk of major financial crisis. If the unemployment rate over the long term does not rise, job insecurity is expected to remain steady. Evidence shows that perceptions of job insecurity do

Job insecurity  93

indeed rise during economic downturns, but there is no evidence of a long-term upward trend in indicators of job insecurity, at least for several major countries. For example, Brochu and Zhou (2009) found that over three decades in Canada there was no evidence of structural change, and that the high levels of perceived insecurity during the early 1990s were only cyclical. For many years a dominant policy theme referred to as ‘flexicurity’ in many European countries has been the intention to reduce regulations that limit the probability of job loss, while increasing the probability of re-employment and thereby lowering the cost of job loss. Despite this pressure for changing policy, labour market regulations to restrict insecurity remain more pervasive across many European countries than in other zones of the global economy. In the United States, by contrast, employers are less restricted and are usually free to ‘hire and fire’ employees with little notice or due process.

Effects on health and well-being

Job insecurity has negative effects on employees’ health and well-being, both for economic and for psychological reasons. The economic reason is straightforward: the present value of a job, taking into account future expected income streams, is lessened when a job becomes more insecure. This loss reduces the extent to which the job meets material needs, just as with a pay cut. Psychology theory, however, extends beyond this material threat. It theorises that job insecurity is a direct ‘stressor’, which produces work strain. Job insecurity can also be interpreted as contributing to a repudiation of the implicit ‘psychological contract’ between worker and employer. This negative effect of job insecurity on health and well-being is distinct from the loss of well-being among those people who actually lose their jobs. But since unemployment generates insecurity across the wider population of those in work, any rise in unemployment potentially has multiplied effects on health and well-being. Unemployment directly affects those who have lost their jobs, but in addition it affects the many more people who do not lose their jobs, by making them feel less secure. A further risk is that job

insecurity can be felt not just by workers, but also by their families. Evidence to support these theories is provided both by psychologists, in many small-scale studies, and by economists using large-scale data. One can accept with greater confidence studies of exogenous variation in insecurity, which can be taken as establishing a causal effect. One good example of the latter is a study that uses data from across Europe. Caroli and Godard (2016) found that people’s perceptions of their job insecurity vary between countries which have different employment protection laws, especially for people working in industries that typically have high dismissal rates. The job insecurity that was induced by low employment protection caused headaches, eyestrain and skin problems, but it was not the cause of insomnia, muscular pain, depression or cardiovascular disease. Another large-scale study uses longitudinal data on job insecurity to look at the impact on mental health and life satisfaction in Australia (Green, 2011). Compared with being in a fully secure job and being fully employable (that is, being sure of getting an equally good job), the mental health of someone who had a 50 per cent probability of job loss within a year was significantly reduced. Moreover, their mental health was lowered twice as much if they had no employability. Indeed, the consequences of being insecure could be comparable to the effects of actually being unemployed. Similar findings are reported in other countries where there is suitable large-scale longitudinal survey data (Otterbach and Sousa-Poza, 2016; Cottini and Ghinetti, 2018). The most striking conclusion from this body of evidence, considered as a whole, is that the effect of job insecurity on health and well-being is larger in magnitude than would be expected just from the associated pecuniary losses. This conclusion is consistent with a broader theory of work, in which jobs potentially mean much more to many workers than just a way of gaining an income. What is required now is more, good research using quasi-experimental methods that allow scientists to have more confidence in their findings, which will build up our knowledge of the health effects of job insecurity in its various forms. Francis Green

94  Elgar encyclopedia of labour studies

Job insecurity in the future

There is also a call for better understanding of future trends in job insecurity, especially in the context of ongoing automation and potential threats to jobs from artificial intelligence (AI) and wider technical change, which may lower the demand for human labour and thereby cause greater job insecurity. There are widely varying estimates of the proportion of jobs that are expected to be threatened by replacement. While these estimates may become more precise, what is needed is better understanding of how the associated feelings of insecurity are likely to develop, their consequent effects on a range of health conditions, and how they can be mitigated by economic and social policies. Francis Green

Francis Green

References

Brochu, P., and L. Zhou (2009), ‘Is job insecurity on the rise? Evidence from Canadian perception data.’ Canadian Journal of Economics-Revue Canadienne D’Economique 42(4), 1305–1325. Caroli, E., and M. Godard (2016), ‘Does job insecurity deteriorate health?’ Health Economics 25(2), 131–147. Cottini, E., and P. Ghinetti (2018), ‘Employment insecurity and employees’ health in Denmark.’ Health Economics 27(2), 426–439. Green, F. (2006), Demanding Work: The Paradox of Job Quality in the Affluent Economy. Woodstock, UK: Princeton University Press. Green, F. (2011), ‘Unpacking the misery multiplier: how employability modifies the impacts of unemployment and job insecurity on life satisfaction and mental health.’ Journal of Health Economics 30(2), 265–276. Otterbach, S., and A. Sousa-Poza (2016), ‘Job insecurity, employability, and health: an analysis for Germany across generations.’ Applied Economics 48(14), 1303–1316.

Job satisfaction

A basic set of job satisfaction determinants often stand as controls in representative examinations of the workforce. These include the inverse U-shape in age with greater job satisfaction for the young and old (Clark et al., 1996) controls for the level of education (sometimes found not to play a large independent role, but with results varying by gender and quality of job match). Other typical controls include hours of work, usually associated with decreased job satisfaction; and wages, usually associated with increased job satisfaction. The typical Likert scale measure of job satisfaction is most commonly the dependent variable in ordered probit or logit estimates including these controls and other variables of interest. In addition to representative studies, the literature explores the job satisfaction of the highly educated, of workers in specific occupations (university professors, athletic trainers, salespeople, and many more), in individual locations and individual firms, of young workers, and government workers. I focus on paradoxes that emerge, on the study of compensating differences, and on how earnings, comparison earnings, and management influence job satisfaction.

Introduction

Workers, the labor input, uniquely consume even as they produce. Thus, the full value of any job consists not only of its earnings, but of the sum of all the negative and positive on-the-job characteristics consumed. It takes but a moment to generate a long list of such characteristics: the quality of co-worker and boss relationships, the chances for advancement, health and safety on the job, the degree of autonomy, the work schedule, leave policies, the nature of worker representation, and many more. Measures of job satisfaction hope to capture the sum of these characteristics and thus come closer to the “full compensation” of a job. It is this full compensation that influences labor supply, worker effort, and firm productivity. It is this full compensation that allows workers to compare a current job to other opportunities. Despite a long history in psychology and sociology, job satisfaction came into labor economics and labor studies in force during the 1970s. The seminal work on the exit-voice view of unions attempted to explain why union members had higher earnings, better benefits, and arguably better methods for dealing with grievances, but expressed lower job satisfaction. The argument was that unions create worker discontent as a tool to improve the quality of the workplace (see Chapter 9 in Freeman and Medoff, 1984). The use of job satisfaction measures has grown well beyond this, as both a dependent variable used to examine labor market phenomena, and a critical control when focusing elsewhere, as will be made clear. A typical criticism of self-reported Likert scale measures of job satisfaction was that they were subjective and that different workers would have different perceptions of the same objective conditions. Despite this, such measures have proven to be very good indicators of which workers are likely to leave a current job. Moreover, the literature emphasizes the general reliability of such measures in accurately characterizing worker well-being. Indeed, workers with higher job satisfaction have been shown to be more productive. Enough objective indicators closely associate with job satisfaction measures that their use has proliferated.

Paradoxes and comparisons

Studies of the discontented union member paradox, alluded to earlier, and the contented female worker paradox, have aided the proliferation of studies examining job satisfaction. These followed the famous Easterlin paradox (Easterlin, 1973): that at a point in time both among and within nations, happiness varies directly with income; but over time, happiness does not increase when a country’s income increases. Union members have been argued to have lower job satisfaction because of the voice function of unions. The strength of the evidence in favor of this depends on sample, time-period, and estimation technique. It has not generally been confirmed for continental Europe, where worker representation is often associated with greater job satisfaction. Even among Anglophone countries in which it was originally identified, it may be reversing (Blanchflower et al., 2021). Similarly, women have traditionally expressed greater job satisfaction than men, despite potential segregation and earnings disadvantages. This has given rise to a large literature exploring the characteristics that 95

96  Elgar encyclopedia of labour studies

differ by gender in determining job satisfaction. Women’s job satisfaction is determined much more by schedule flexibility and family-friendly policies. It is determined much less by earnings. Indeed, evidence (although far from monolithic) suggests that having a child in the family diminishes the job satisfaction of women, but not of men. Despite these potential differences, there is recent evidence that the extent of the female satisfaction premium has been shrinking over time and may now be negligible. This reflects the role of the determinants becoming more nearly similar over time (Green et al., 2018). The workplace characteristics that influence job satisfaction may do so in different ways, and this has allowed the quality of jobs to be examined in many different contexts. Thus, the harm to workers from being overqualified for a job may include lower earnings, slower advancement, less intrinsic enjoyment from the work, and an increased likelihood to quit. Researchers frequently summarize this by estimating the influence of overqualification in diminishing job satisfaction. This can be done either by comparing to workers who are not overqualified, or by examining the job satisfaction of workers as they move into and out of positions for which they are overqualified. As a second example, flexible employment (flexible to the employer) such as fixed-term contracts, agency work, and temporary on-call work, are often less secure for workers. The theory of compensating differentials argues that workers would take these jobs only if other aspects of the work are rewarding. This leads researchers to examine the influence of flexible employment on job satisfaction, and on various domains of job satisfaction. The overall influence of flexible employment on overall job satisfaction may be negative but modest in size when controlling for typical covariates. The real story is that it is associated with dramatically lower satisfaction with job security, even as other domains seem to be modestly higher in flexible employment (satisfaction with the work itself, satisfaction with pay, and so on). Another way to examine these tradeoffs has been to control for the domains of job satisfaction when estimating overall job satisfaction. If all the domains are included, this allows for identifying the relative importance of satisfaction across the domains. This often reveals differences in the relative importance John S. Heywood

of domains by gender, that match the lessons from including actual job characteristics. Thus, satisfaction with pay is less important for women than it is for men. In a variation, sometimes a critical job satisfaction domain can be included to examine a variable of interest on the remaining variation in overall job satisfaction. Thus, when satisfaction with job security is controlled for in estimates, flexible employment often correlates with higher overall job satisfaction, potentially supporting the theory of compensating differentials. A further illustration of examining workplace tradeoffs is provided by the extensive literature on the job satisfaction of the self-employed. The self-employed take on greater risk and routinely report being less satisfied with their job security compared to paid employees. Yet, they often report greater overall job satisfaction, as they are more satisfied with the work itself and with their autonomy and control. At the other extreme, government workers may report greater satisfaction in their jobs. Their pay may be lower even as their intrinsic satisfaction and job security are greater. Again, job satisfaction provides a measure that allows for focusing on these workplace tradeoffs. Compensating differentials are often given precise estimates, using job satisfaction estimates. A job disamenity, such as risk of injury or death on the job, may be associated with lower job satisfaction by a specific amount in the typical estimate. This amount would come from the coefficient in an ordinary least squares (OLS) estimate, or from the average marginal effect in the ordered probit. This is then matched by the amount by which earnings would need to increase so that the overall job satisfaction measure remained unchanged. This increase would again come from the relevant coefficient or average marginal effect of earnings on job satisfaction. The increment in the job characteristics is given a dollar earnings value providing the marginal tradeoff. Thus, analysis of job satisfaction provides an alternative to the more traditional hedonic earnings estimates for quantifying compensating differentials or for showing their absence. The ability to examine compensating differentials results from the strong association between job satisfaction and own earnings. This association would be anticipated in a self-regarding utility function, but an inter-

Job satisfaction  97

dependent utility function might include the wages of others. A large literature examines this possibility by using job satisfaction measures. In early papers the actual wage and the predicted wage based on an individual’s characteristics were both entered in job satisfaction equations. The “comparison earnings” would often have a negative influence on job satisfaction, holding actual earning constant. The presumption was that workers felt worse as their own wage fell further behind comparison earnings. Alternative specifications included simply entering the difference between the actual and predicted wages as a single variable, and including self-reported measures of how workers felt about their pay relative to what they felt they deserved. This strain of examination has been criticized for not accounting for individual wage trajectories, or for individual worker fixed effects. Studies using individual reported data can be contrasted with those observing multiple workers within a given establishment. This second set of studies enters the true wages of co-workers into job satisfaction estimates. Clark et al. (2009) show higher individual job satisfaction when other workers in the same establishment are better paid. This runs counter to the evidence above, but they argue that it reflects the reference group. The earnings of co-workers may reflect jealousy, but they also provide a signal about the worker’s own future earnings. This second positive signal influence dominates especially for younger workers. A natural experiment examined the consequences of informing a randomly selected group of University of California workers about a website listing co-worker earnings (Card et al., 2012). This treatment group responded asymmetrically. Workers with below-median pay for their unit and occupation showed decreased job satisfaction relative to the control (the uninformed), while those with above-median pay showed no change in job satisfaction. Similarly, those with below-median pay were more likely to report an intention move, with no change again for those with above-median pay (Artz et al., 2019). Despite the asymmetry, job satisfaction depended on relative pay comparisons.

Another broad inquiry examines the structure of earnings, the nature of human resource management, and job satisfaction. More-frequent, smaller raises may generate greater satisfaction than larger, less-frequent raises. Tying pay to performance can increase net job satisfaction by enhancing a sense of earned success, but it may also bring greater job stress. Employers who provide clear direction, reward merit, and provide opportunities for growth tend to have workers with greater satisfaction. Female workers have been shown to have higher job satisfaction with a male supervisor. Workers with a boss who is more technically qualified (in terms of ability to do the worker’s job, accomplish objectives within the firm, and so on) report higher job satisfaction, all else equal. Despite this emphasis on earnings and management, job satisfaction also positively associates with sharing values with one’s employer, and with a sense of service in the job.

Use of job satisfaction as a control variable

While the applications above use job satisfaction as an independent variable in estimations, a strain of literature uses it as an important right-hand-side control (Johnson and Neumark, 1997). Researchers use self-reported claims in survey data to explore the determinants of employment discrimination. Researchers rely on such survey data as court records are not standardized, and most employment discrimination cases are dropped or settled, leaving no record. The self-reported claims of discrimination are seen as a necessary antecedent to seeking remedy, and the pattern is viewed as illustrative of the sources of costly litigation. At issue for these researchers is whether larger gender wage differentials are associated with increased discrimination claims, whether family friendly practices can reduce gender discrimination claims, and whether unionsor employer-provided involvement reduce racial and gender discrimination claims. Yet, dissatisfied workers may both report discrimination and earn less than peers generating a spurious association. To control for this possibility, these authors frequently include job satisfaction as a control to see whether the determinants of discrimination remain robust. John S. Heywood

98  Elgar encyclopedia of labour studies

Conclusion

The uses of measured job satisfaction are many. As an indicator of total compensation, it is particularly well suited for examining tradeoffs among on-the-job characteristics, and examining compensating differentials. Several persistent paradoxes have emerged generating ongoing study about their continued prevalence and meaning. The role of earnings, comparison earnings, and human resource practices on satisfaction have generated robust literatures. To the extent that management or worker representation increases job satisfaction, it has increased total compensation, and so the competitiveness of openings in the job market. Finally, workers may accept greater dissatisfaction with on-the-job consumption to accomplish other goals outside of work. They seek a particular location that allows specific activities, or is near family and friends. While these activities might be considered outside of the realm of job satisfaction and to belong in another satisfaction domain, it is far from clear that domains are so easily divided, or that workers might not report greater overall job satisfaction because such off-the-job consumption is permitted because of a job characteristic. This explains the growing use of both job satisfaction and overall life satisfaction, even when examining job characteristics. John S. Heywood

John S. Heywood

References

Artz, B., A. Goodall, and A. Oswald (2019), “Boss competence and worker well-being,” Industrial and Labor Relations Review 70(2), 419–450. Blanchflower, D.G., A. Bryson, and C. Green (2021), “Trade unions and the well-being of workers,” British Journal of Industrial Relations, Early View, https://​doi​.org/​10​.1111/​ bjir​.12627. Card, D., A. Mas, E. Moretti, and E. Saez (2012), “Inequality at work: The effect of peer salaries on job satisfaction,” American Economic Review 102(6), 2981–3003. Clark, A.E., N. Kristensen, and N. Westergård-Nielsen (2009), “Job satisfaction and co-worker wages: Status or signal?,” Economic Journal 119(536), 430–447. Clark, A.E., A. Oswald, and P. Warr (1996), “Is job satisfaction U-shaped in age?,” Journal of Occupational and Organizational Psychology 69, 57–81. Easterlin, R.A. (1973), “Does money buy happiness?,” Public Interest 30(3), 3–10. Freeman, R.B. and J.L. Medoff (1984), What Do Unions Do? Basic Books, New York. Green, C., J.S. Heywood, P. Kler, and G. Leaves (2018), “Paradox lost: The disappearing female job satisfaction premium,” British Journal of Industrial Relations 56(3), 484–502. Johnson, R.W. and D. Neumark (1997), “Age discrimination, job separations, and employment status of older workers: Evidence from self-reports,” Journal of Human Resources 28(4), 779–811.

Jobs, tasks, authority

by the simultaneous growth of the share of employment in high-skill, high-wage occupations (such as managers and professionals) and low-skill, low-wage occupations (such as personal care, cleaning, and protective services). In turn, the share of employment in middle-skill, middle-wage occupations has declined. These include production workers, office staff, and administrative staff. Because occupations are bundles of tasks, and occupations have different task intensities, the change in the occupational structure of employment reflects changes in the tasks performed in the labor market; specifically, it reflects the shift towards non-routine cognitive and manual tasks, and away from routine cognitive and manual tasks (more details on the task categories will be given below). Earnings inequality has evolved less uniformly across industrialized countries and over time. Most industrialized countries have experienced episodes of increases in earnings inequality since the mid-1970s. More recently, those episodes have turned into periods in which the lower end of the earnings distribution has been gaining on the middle in many industrialized countries, that is, episodes of simultaneous growth of high and low wages relative to the middle. This means that wage growth has not increased monotonically in workers’ skill levels. Thus, it has polarized, analogously to what has been observed for the evolution of employment. The similarity across industrialized countries suggests that global factors, such as technological changes, play an important role in driving labor market developments. Overall, it is striking how uniform the broadly observed trends of the past three to four decades are, in particular when it comes to changes in the occupational structure of employment and the tasks that workers perform. In contrast, the evolution of wages across industrialized countries is much more diverse than what we observe for changes in employment, in terms of both pattern and timing. This reflects the fact that there are also other forces at play, which might be specific to single countries or a set of countries, and which often play out more on prices than on quantity. The prime candidate is differences in wage regulation, but countries can also differ in the price elasticity of product demand or the elasticity of labor supply (Autor, 2015).

Introduction

For many decades, trends in labor markets, in particular the evolution of the returns to skills and earnings inequality, were explained using models of skill supply and demand, with the latter linked to technological changes (referred to as the “canonical model” by, for example, Acemoglu and Autor, 2011). Crucial components of the canonical model are the assumption of a one-to-one mapping of workers’ skills to the tasks that they perform on the job, and the implication that technological changes bring about an increase in labor demand that is monotonically increasing in workers’ skill levels. Both components turned out to be difficult to maintain in the context of recent labor market trends. Against this background, task-based frameworks have proven to be fruitful innovations (Autor et al., 2003; Acemoglu and Autor, 2011). Task-based models have many advantages over the canonical model. They allow more nuanced mappings of differently skilled workers to workplace tasks. They provide a direct link to occupations (henceforth, the terms “occupations” and “jobs” are used interchangeably), which are bundles of tasks, and help to explain the pronounced changes in the occupational structure of employment, as well as the changes in earnings inequality (albeit less directly). In addition, they open the black box of technological change and its impact on the labor market by offering specific mechanisms of how humans and technology interact. Task-based approaches are also powerful tools in explaining some aspects of the impact of trade on the labor market. Recently, they have been adjusted to also help explain the effects of artificial intelligence (AI) on the labor market.

Stylized facts

Among the most striking labor market trends in industrialized countries since the end of the 1980s and the beginning of the 1990s is the change in employment shares that is non-monotonically increasing in workers’ skill levels (Acemoglu and Autor, 2011; Goos et al., 2014, among others). That is, industrialized countries have experienced employment polarization, characterized 99

100  Elgar encyclopedia of labour studies

Important features of the task-based framework

in the occupational structure of employment (Spitz-Oener, 2006; Atalay et al., 2020).

Mapping of tasks and skills One key innovation of the task-based models is the distinction between tasks that workers perform on the job, and the skills that workers possess. A task is a unit of work activity that produces goods and services. By contrast, a skill is a worker’s endowment of capabilities for performing various tasks. This endowment is the result of a worker’s investment in human capital, among other things. Workers use their skills to perform tasks in the workplace. Task-based frameworks break open the one-to-one mapping of job tasks to workers’ skills that is assumed in the canonical model, which in turn allows a more nuanced description of the allocation of equally skilled (that is, not only differently skilled) workers across workplaces. Specifically, workers of a given skill level can perform different tasks. Moreover, workers can change the tasks that they perform on the job in response to technological changes, among other things.

Links to technology

Mapping of jobs and tasks The characterization of modern labor markets along occupation boundaries has proven to be particularly fruitful, and not only when it comes to the changes in the occupational structure of employment. For example, the explanatory power of occupations when it comes to examining wage differences across workers has increased over time as well. Task-based models have a natural link to occupations, since occupations comprise bundles of tasks. Because of this link, changes in the occupational structure of employment reflect changes in tasks performed by workers in the labor market. In empirical work, researchers typically rely on (aggregated) administrative occupational classifications in combination with task information on the occupational level. In that respect, occupations are proxies for job tasks, and (due to data limitations) the task content of occupations often remains constant over time. Empirical work using data that also facilitate the measuring of task changes within occupations over time suggests that the changes in tasks performed by workers have been considerably greater than is indicated by studies that only rely on the changes Alexandra Spitz-Oener

While task-based models are now also used to model trade’s impact on the labor market, among other things, their original and still most common application is in the context of how technology affects the labor market. The seminal study by Autor et al. (2003) opened the black box that, at the time, characterized studies which investigated the way in which humans and machines interact. The innovation was that one classifies tasks from a machine’s perspective into routine and nonroutine tasks. Routine tasks are expressible in rules, such that they are easily programmable and can be performed by software installed on machinery. Hence, machinery is a substitute for workers in carrying out routine tasks. Nonroutine tasks are not programmable, and cannot be accomplished by machinery. However, the software installed on machinery may increase the productivity of workers performing nonroutine analytical and interactive tasks. The seminal categorization when thinking about human–technology interaction includes the following task categories: nonroutine analytical tasks, such as research, planning, or evaluation activities; nonroutine interactive tasks, such as selling or coordinating and delegating work; routine cognitive tasks, such as double-entry bookkeeping and calculating; routine manual tasks, such as machine feeding or running a machine; and nonroutine manual tasks, such as gardening, housekeeping, or restoring houses (Autor et al., 2003; Spitz-Oener, 2006; Atalay et al., 2020). One important advantage of the task-based perspective is its amenability. On one hand, the task perspective asks what types of tasks is technology a substitute for workers, and the answer to this question may change over time. In fact, recent advances in AI based on new machine learning techniques, and the availability of massive data sets, have greatly altered the task sets that technology can now perform (e.g., Acemoglu et al., 2022). On the other hand, there are also new tasks that arise. While the original task framework assumed the set of tasks that are performed in the labor market to be static, recent literature stresses the fact that new tasks can be added over time, and some might disappear. This later

Jobs, tasks, authority  101

feature was already important in the past, but it might become particularly useful with the continuing advances in AI (Acemoglu and Restrepo, 2018; Autor, 2019).

Link to recent labor market trends

As discussed in the section on “Stylized facts,” observed employment and earnings changes in industrialized countries have been increasing non-monotonically in workers’ skills, even though the exact pattern and timing might vary across different countries. What are the mechanisms driving those labor market trends through the lens of task-based frameworks? The distinction between workers’ skills and workplace tasks is the important feature of the task-based perspective that allows a characterization of new machinery which substitutes workers of various skill levels in performing job tasks. In addition, because the technology of the 1980s to 2000s was substituted for workers in performing routine tasks that were predominantly performed by workers in middle-skill, middle-wage occupations, it rationalizes the observed changes in the occupational composition of employment in industrialized countries that have been observed in recent decades.

Authority

The question of who has the decision-making authority in a firm is of central importance, and is the subject of a large body of literature in both economics and management science (Aghion and Tirole, 1997). Evidence suggests that workers’ autonomy in decision-making is an important determinant of a worker’s job satisfaction and productivity, as are related factors such as working in smaller firms with flatter hierarchies and higher levels of independence on the job or the absence of monitoring. In a similar vein, experimental evidence indicates that reducing an individual’s decision-making autonomy negatively affects their work effort. The fact that the occupational employment structure has moved away from routine jobs, such as production workers, office workers, and administrative staff, and towards nonroutine cognitive jobs, such as managers and professionals, indicates that the importance of decision-making tasks has increased in recent decades (e.g., Deming, 2021). At the same

time, however, delegating control rights to the worker (agent) also always bears the risk for the employer (principal) that workers reduce their effort choice. The advent of AI has brought more impetus to the discussion on who should exert decision authority from the point of view of the principal, the worker, or the machine (e.g., Athey et al., 2020). As a prediction tool, AI reduces the cost of making predictions that may form the basis for decision-making. But delegating decision-making to the machine might have negative consequences for the effort level asserted by the worker (“fall asleep at the wheel”; Athey et al., 2020: 80). As a result, AI—even if it provides better predictions for decision-making—might not turn out to be more profitable for the firm. Alexandra Spitz-Oener

References

Acemoglu, D. and D.H. Autor (2011), “Skills, tasks and technologies: Implications for employment and earnings,” in O. Ashenfelter and D. Card (eds), Handbook of Labor Economics, Vol. 4b. North-Holland: Elsevier, 1043–1171. Acemoglu, D., D.H. Autor, J. Hazell, and P. Restrepo (2022), “AI and jobs: Evidence from online vacancies,” Journal of Labor Economics, 40(51), S293‒S340. Acemoglu, D. and P. Restrepo (2018), “The race between man and machine: Implications of technology for growth, factor shares, and employment,” American Economic Review, 108(6), 1488–1542. Aghion, P. and J. Tirole (1997), “Formal and real authority in organizations,” Journal of Political Economy, 105(1), 1–29. Atalay, E., P. Phongthiengtham, S. Sotelo, and D. Tannenbaum (2020), “The evolution of work in the United States,” American Economic Journal: Applied Economics, 12(2), 1–34. Athey, S.C., K.A. Bryan, and J.S. Gans (2020), “The allocation of decision authority to human and artificial intelligence,” AEA Papers and Proceedings, 110, 80–84. Autor, D.H. (2015), “Why are there still so many jobs? The history and future of workplace automation,” Journal of Economic Perspectives, 29(3), 3–30. Autor, D.H. (2019), “Work of the past, work of the future,” AEA Papers and Proceedings, 109, 1–32. Autor, D.H., F. Levy, and R. Murnane (2003), “The skill content of recent technological change: An empirical exploration,” Quarterly Journal of Economics, 118(4), 1279–1333.

Alexandra Spitz-Oener

102  Elgar encyclopedia of labour studies Deming, D. (2021), “The growing importance of decision-making on the job,” NBER Working Paper 28733. Goos, M., A. Manning, and A. Salomons (2014), “Explaining job polarization: Routine-biased technological change and offshoring,” American Economic Review, 104(8), 2509–2526.

Alexandra Spitz-Oener

Spitz-Oener, A. (2006), “Technical change, job tasks and rising educational demands: Looking outside the wage structure,” Journal of Labor Economics, 24(2), 235–270.

L

Labor income share The functional distribution of income has long been a topic of interest for economists. David Ricardo’s statement published in 1817 serves as testimony to this fact: “To determine the laws which regulate [this] distribution is the principal problem in political economy.” In recent years, a large body of research has documented a global decline in the labor income share, challenging the stability of the labor income understood as “the stylized fact of long-term economic growth” at least since Kaldor (1957). Resurgence of interest in factor income shares has resurfaced debates over fair distribution of personal incomes, and led research inquiries into a deeper understanding of the problems associated with measurement of the labor income share.

Definition and measurement issues

The labor income share simply measures the share of national income which accrues to labor. An equivalent concept of the labor income share at the firm level would be compensation of employees over firm’s income in a specific accounting period. Compensation of employees typically includes wages and salaries paid in cash, and contributory benefits related to social insurance payable by employers. In this sense, any measure of the labor income share views compensation of employees as a cost to the employer, and does not include the cost of unpaid work undertaken voluntarily. A bias in the cross-country comparisons of the labor income share could arise from differences in the payroll tax rate, paid by the employers on the wage bill. Changes in the labor income share underpin the relationship between national income and personal income. In a standard framework with capital and labor as the only input factors for production, the total income

is composed of labor income and capital income (for example, corporate profits). If capital income is more unequally distributed than labor income, as is typically the case in advanced countries, then a transfer from labor income to capital income could increase inequality in personal incomes. Thus, large differences in the distribution of labor and capital incomes can drive inequality in personal incomes independently of the labor income share. Since a sizable portion of the national income comes from the self-employed, and self-employed incomes are not systematically reported, the accuracy of the statistical procedures to impute wages of the self-employed play havoc with the measurement of the labor income share. Various adjustments have been proposed to minimize the downward bias in the labor income share due to self-employment (Gollin, 2002). Moreover, an increase in housing value-added and a rise in the capital income of business owners could force the labor income share to fall. Excluding self-employment and housing assets from the measurement of the labor income share could address this problem; however, cross-country comparisons of the labor income share solely based on corporate incomes could also produce biased results due to differences in the delineation of corporate sectors across countries; see Gutiérrez and Piton (2020).

Theory

A production function with a constant elasticity of substitution (EoS) technology and constant returns to scale shows a stable relationship between the labor income share, the elasticity of factor substitution, and the capital–labor ratio. Any value of the EoS between capital and labor different from 1 could thus play a decisive role in the movement of the labor income share. An increase in the capital–labor ratio lowers

103

104  Elgar encyclopedia of labour studies

the labor income share if capital and labor are gross substitutes (that is, when the EoS between capital and labor is greater than 1). An increase in the capital–labor ratio can also produce an increase in the labor income share if capital and labor are gross complements (that is, when the EoS between capital and labor is lower than 1). The role of capital accumulation becomes crucial as a drop in the relative price of investment (capital) leads to an increase in the capital–labor ratio, which forces the labor income share to decline when capital and labor in the production technology are substitutes. However, the role of capital accumulation as a driver of the labor income share requires capital and labor to be substitutes, which appears paradoxical in a world predominantly characterized by capital and labor as complements. The composition of skills in the labor force could resolve this puzzle. With capital–skill complementarity, a decline in the relative price of capital can cause a decline in the aggregate labor income share when a decrease in the labor income share for the unskilled labor force outweighs an increase in the labor income share for skilled labor, given that the aggregate EoS between labor (across different skill groups) and capital is less than 1 (Paul, 2020).

Empirical evidence

Labor’s share of gross domestic product has declined in many countries since the early 2000s; see Karabarbounis and Neiman (2014) and Dao et al. (2017). While the empirical evidence is drawn predominantly from the developed countries, the downward trend in the labor income share is also recognized in many emerging markets. On average, the labor income share in developing countries is slightly smaller than that in developed countries. At the sectoral level, labor income shares have declined in seven out of ten major industries worldwide, with the largest decline in the tradable sectors such as manufacturing, transportation, and communication. The government service sector accounts for the largest share of labor income, at 46 percent; whereas public utilities and mining have the smallest shares of labor income, at 16 percent and 20 percent, respectively (Paul, 2020). The average labor income share trends at the sector level are noticeably different Saumik Paul

across regions. Labor receives the smallest income share in the agricultural sector in all regions except the Middle East, North Africa, and sub-Saharan Africa. In East Asia and the Pacific, and North America, the labor income share is largest in the services sector; whereas in Europe and Central Asia, the manufacturing sector shows the greatest labor share. The labor income share in manufacturing has fallen in the majority of countries and has been accompanied by a rise in the shares for both agriculture and services; see Paul (2020). The existing literature offers several explanations for the decline in the labor income share. These forces can be classified under three broad categories: (1) technological change and capital intensity; (2) institutional factors; and (3) globalization. The main forces related to technological change include availability of cheaper capital relative to labor (the accumulation channel), falling information and communication technology cost, worker displacement due to automation, routinization of tasks, and finally, the rise of the superstar firms. The role of labor market institutions and social norms have also been documented in the literature. Globalization can affect the stability of the labor income share in various ways. Trade should lower the labor income share in capital-abundant advanced economies, but it produces an opposite outcome in labor-abundant developing economies, since a country has comparative advantages in industries where a relatively abundant production factor (labor or capital) is intensively utilized (the Heckscher‒Ohlin model). Trade liberalization could be associated with a lower labor income share if trade barriers are associated with the relative bargaining power of capital owners, as they can relocate their resources to foreign destinations with higher returns. Participation in global value chains together with exposure to routinization of tasks that makes labor more substitutable by capital can lower the labor income share through the accumulation channel (Dao et al., 2017). Finally, if the offshoring tasks become more capital-intensive in the receiving countries, it would require less labor to perform these tasks, resulting in a fall in the labor income share.

Labor income share  105

Way forward

Despite a large body of research documenting a global decline in the labor income share, difficulties in measuring the labor income share cast doubt on the validity of these outcomes. While a consensus is yet to be reached on the actual decline of the labor income share, the emerging field of macro development, which aims to understand how heterogeneous firm characteristics are linked to fluctuations in several macroeconomic trends, provides deeper insights into the labor income share dynamics. A negative relationship between markups and the labor income share (LIS) follows directly from the firm’s optimization decision. The rise of market power through reallocation of economic activities from lowto high-markup firms causes firms to grow in size (that is, the growth of superstar firms) and the labor income share to decline; see Autor et al. (2020). The welfare outcomes at disaggregated levels could worsen even if the labor income share at the aggregate level remains stable. Not only among firms: the analysis of the labor income shares by skill and sectors also remains a promising area of future research. Labor is heterogeneous in skills, and the elasticity of substitution between capital and labor differs across different skill levels. Since the early 1980s, the labor income share for skilled workers in the United States has shown an upward trend, which suggests that the decline in the aggregate labor income share is entirely driven by the falling labor income share for unskilled workers (Paul, 2020). Evidence based on a global input–output database also suggests that the labor income share has been increasing for high-skilled workers and declining for middle- and low-skilled workers. A rising

skill premium mainly drives the results for high-skilled workers, while participation in global value chains remains the dominant factor behind the decreasing labor income share for middle- and low-skilled workers (Dao et al., 2017). To conclude, the resurgence of interest in factor income shares has generated research inquiries that go beyond the concerns over the stability of the labor income share. Movements of resources across firms and industries provide novel insights into the micro foundation of factor income trends at the aggregate level. Saumik Paul

References

Autor, D., D. Dorn, L.F Katz, C. Patterson, and J. van Reenen (2020), “The fall of the labor share and the rise of superstar firms,” Quarterly Journal of Economics 135 (2): 645–709. Dao, M.C., M.M. Das, Z. Koczan, and W. Lian (2017),  Why is Labor Receiving a Smaller Share of Global Income? Theory and Empirical Evidence, International Monetary Fund, Working Paper 17/169. Gollin, D. (2002), “Getting income shares right,”  Journal of Political Economy, 110(2), 458–474. Gutiérrez, G. and S. Piton (2020), “Revisiting the global decline of the (non-housing) labor share,” American Economic Review: Insights, 2 (3): 321–338. Kaldor, N. (1957), “A model of economic growth”, Economic Journal, 67 (268): 591‒624. Karabarbounis, L. and B. Neiman (2014), “The global decline of the labor share,” Quarterly Journal of Economics, 129 (1): 61–103. Paul, S. (2020), “Understanding the global decline in the labor income share,” IZA World of Labor, 472. Ricardo, D. (1817), On the Principle of Political Economy and Taxation. John Murray, London.

Saumik Paul

Labor market discrimination: method and measurement

(1986). For a recent review of economic theories of labor market discrimination, see Lang and Lehmann (2012). In this entry, we briefly review the most common methods that have been used to detect discrimination.

Introduction

Detecting discrimination

The oldest economic theory of discrimination is Becker’s (1957) theory of taste-based discrimination in a perfect labor market. In its most basic form, employers act on their ethnic or racial preferences when recruiting workers. Employers favor workers that belong to their own group, or have animosity against workers belonging to other groups than their own. Taste-based theories of discrimination do not assume that employers perceive productivity differences between workers. Discrimination against a group of workers is simply caused by employers’ tastes. Such tastes may, in the short run, result in wage and unemployment differences between different groups of workers, and preferential hiring of own-group workers. Subsequent work on taste-based discrimination has extended the original theory to include search frictions in the labor market using search-and-matching models. Such models have shown that taste-based discrimination can be persistent and occur even with mild employer tastes. Another equally important strand of theoretical literature on discrimination stems from the influential work of Phelps (1972) and Arrow (1973) on statistical discrimination. In this line of theories, employers do not have tastes for a particular group of workers, but lack important information about the true ability of potential workers. Uncertainties about workers’ skills make employers use group-level information in order to draw inferences about individual workers’ skills. Hence, an employer may find it less risky to employ workers from their own group than workers from other groups if the own group on average is known or believed to perform better than other groups. As a result, employers will pay less or be less likely to hire workers from other groups than their own. Theoretical extensions of the statistical discrimination model have shown that group-level negative stereotypes can become long-lasting and self-confirming. For a comprehensive review of the economic analysis of labor market discrimination, see Cain

Studies on observational data Detecting discrimination is not an easy task. The first approach that naturally comes to mind is to compare wages and unemployment rates of workers from minority groups and from the majority group, using observational data. A large body of research has used advanced statistical methods on observational data just to do that, and has provided evidence of large and widespread differentials between different minority and majority groups in labor market outcomes across the globe. Observational data can be historical accounts, or data gathered through surveys. A well-known problem with studies on observational data is that the researcher does not have full control over the information that is being considered. This problem raises the question of how reliable conclusions drawn from studies on observational data are. One central part of the problem has to do with the so-called omitted variable bias. This bias means that there might be other factors besides group membership that may explain labor market differences between minority and majority groups. Such factors might be observable to employers and relevant for the study outcome, but are not available and observable for the researcher. Another part of the problem has to do with the fact that the set of information that is available and used by the researcher to explain differences in labor market outcomes can itself have been affected by discrimination. One such example is discrimination in the education system. Hence, although studies on observational data are useful, and frequently used to illustrate potential ethnic and racial biases in the labor market, their findings do not normally meet the usual requirements that we place on empirical evidence of discrimination. Field experiments Given the problems with studies using observational data, researchers have turned to experimental methods in order to provide

106

Labor market discrimination: method and measurement  107

more unambiguous evidence of discrimination. Field experiments have become a powerful and widely used tool to collect evidence of discrimination where biased employers are basically caught red-handed. In a nutshell, fictitious job applications are sent to real employers with vacant jobs. Except for ethnicity or race, which is randomly assigned, work-related skills and all other individual characteristics are kept fixed across fictitious applicants. Hence, in contrast to studies on observational data, the researcher has complete control over the information that is available to the employers in their decision making. The outcome of interest in field experiments is the number of positive responses that the minority versus majority applicants receive from employers. Field experiments can be done using both a between-subject design (only one application is sent to each employer) as well as a within-subject design (multiple applications are sent to each employer). Furthermore, some field experiments use audit studies where trained auditors make the job applications, while other field experiments rely on correspondence tests where only written job applications are sent. Correspondence tests have become the standard approach for conducting field experiments of discrimination because of their appealing qualities related to the researcher’s control over the study. A field experimental study that has become a landmark in the literature of labor market discrimination was conducted by Bertrand and Mullainathan (2004). First, they generated templates for fictitious resumés that they could send to employers in Boston and Chicago. In order to design realistic resumés, they used as inspiration real resumés posted by people on online job markets, which they altered to create their own set of resumés. They prepared low-quality and high-quality resumés with various features for four occupational categories: sales, administrative support, clerical services, and customer service. Second, they created identities for the fictitious job applicants, including contact information but also, and most importantly, their names. They used name frequency data to create a list of distinctive African American and white names that could be used in their experiment. Third, they carried out their experiment in Boston and Chicago by responding to job ads published in the

Boston Globe and the Chicago Tribune. For each ad two high-quality and two low-quality resumés were sampled to fit the vacant job. One of each quality type of resumé were randomly assigned African American names, while the other two resumés received white names. All in all, they sent almost 5000 job applications to more than 1300 employers during a period of less than one year. Finally, they documented employer responses to the fictitious applications. Their results showed that white applicants were 50 percent more likely than African American applicants to receive a positive employer response, both in Boston and in Chicago; that employers were more responsive to high-quality resumés when the applicant was white; and that the results were consistent across occupational categories and employer size. Discrimination against minorities is remarkably persistent both across the globe and over time. Subsequent studies have provided solid evidence of labor market discrimination against minorities across many countries, using similar or refined field experiments (Bertrand and Duflo, 2017). A meta-analysis (Quillian et al., 2017) on 28 studies in the United States (US) labor market consisting of over 55 000 applications to over 26 000 employers showed that there are only small changes in the magnitude of discrimination over time. For the past 30 years, whites have been 36 percent more likely to receive a positive response from employers than African Americans, and 24 percent more likely to receive a positive response from employers than Latinos. Although the field experimental approach is an appealing and powerful way of collecting evidence on labor market discrimination, some important limitations should be noted. The most obvious one has to do with the fact that it only measures labor market discrimination in the very first stage of the recruitment process and is less efficient in studying discrimination in the later stages or in wage setting. The most well-known critique of audit and correspondence studies of discrimination was put forward by Heckman and Siegelman (1993), who showed that group differences in the perceived variance of ability can bias estimates of discrimination in field experiments. Other issues that researchers must bear in mind are related to the ethical sensitivity of these field experiments (Granberg, Ali Ahmed and Mats Hammarstedt

108  Elgar encyclopedia of labour studies

2022). Employers are observed without their consent, since otherwise the field experiments would lose their purpose. Furthermore, employers are exposed to fictitious applications which may inflict additional work on employers. Last, introducing fictitious applications in the labor market may affect the competition among and prospects of real job candidates. Hence, a researcher needs to always consider the ethical concerns in relation to the potential benefits and usefulness of a study when planning a field experiment of labor market discrimination. Natural experiments Sometimes researchers can observe changes in society that occur naturally and that can facilitate prerequisites for an experimental-like study. Such circumstances can result in skillfully executed studies, but are rare. One example is the work of Arai and Skogman Thoursie (2009), who studied the effect on yearly earnings and on the probability of having nonzero earnings in Sweden, of a name change for immigrants with originally Asian, African, and Slavic names to a typical Swedish name. They identified 219 women and 422 men during the 1990s who changed their surnames from foreign-sounding names to Swedish names. The variation that existed in the data from people changing their names at different points of time was exploited by the researchers to examine annual earnings and the probability of having nonzero earnings before and after a name change, while controlling for other relevant information. Their results showed that people who changed their foreign-sounding surname to a typical Swedish surname considerably increased their annual earnings, and were significantly more likely to have nonzero earnings. Thus, they concluded that discrimination exists. Another example of a natural experiment is when researchers have used terrorist attacks to examine whether attitudes toward certain groups of people change as a result of an attack. One such incident is the 9/11 attacks in the US, which has been used to study discrimination against Arab Muslim people in several countries. Lab experiments Another important strand of empirical literature on discrimination has used laboratory Ali Ahmed and Mats Hammarstedt

experiments. Laboratory settings might be less like real-life situations, and are often conducted among undergraduate students, but they are far more superior than other types of studies when it comes to testing theories and attaining as much control as possible over a study. They are especially useful for disentangling predictions of competing theories. Most empirical studies on discrimination are unable to distinguish between taste-based and statistical discrimination. However, with clever designs in the laboratory, researchers are able to examine the roots of discriminatory behavior. The pioneering study of Fershtman and Gneezy (2001) is one such example. They illustrated the usefulness of laboratory experiments to examine various facets of discrimination. They did not study labor market discrimination per se, but tried to distinguish between taste-based and statistical discrimination against people of Eastern origin by people of Ashkenazic origin in the Israeli Jewish society. Nearly 1000 Eastern and Ashkenazic undergraduate students were recruited for this experiment. Participants played a number of simple two-person games with each other, where players’ names and thereby their ethnic origin were made salient. Using the trust game (a situation where both players have a strategic role and affect the outcome of the game), the researchers first recognized that there was a systematic mistrust directed towards participants of Eastern origin. This mistrust could have been a result of both taste-based and statistical discrimination. However, by using the dictator game (a situation where only one of the players determines the outcome of the game) they could rule out taste-based discrimination against participants of Eastern origin. Consequently, they could conclude that the observed discrimination against participants of Eastern origin in the trust game was due to statistical discrimination.

Conclusions

The two main theories of labor market discrimination in economics have been taste-based and statistical theories of discrimination, and the most compelling evidence of discrimination has been generated through labor market field experiments. Both theoretical perspectives on discrimination assume that unequal treatment of people is attribut-

Labor market discrimination: method and measurement  109

able to individuals, and that this behavior is intentional. Research in social psychology has shown that implicit ethnic or racial biases may inadvertently result in discriminatory behavior. Sociological research has provided the notion of structural or institutional discrimination to emphasize the possibility that discrimination by ethnicity and race may occur on a societal level. For a discussion of alternative perspectives on ethnic and racial discrimination, see Small and Pager (2020). Future economics research needs to incorporate these alternative, reasonable perspectives. From an empirical point of view, little is known about labor market discrimination beyond the initial stage of the recruitment process. Ali Ahmed and Mats Hammarstedt

References

Arai, M. and P. Skogman Thoursie (2009), “Renouncing personal names: An empirical examination of surname change and earnings,” Journal of Labor Economics, 27(1), 127–147. Arrow, K.J. (1973), “The theory of discrimination,” in O.C. Ashenfelter and A. Rees (eds), Discrimination in Labor Markets (3–33). Princeton University Press, Princeton, NJ. Becker, G.S. (1957). The Economics of Discrimination. University of Chicago Press, Chicago, IL. Bertrand, M. and E. Duflo (2017), “Field experiments on discrimination,” in A.V. Banerjee and E. Duflo (eds), Handbook of Economic Field Experiments, Vol. 1 (309–393). North-Holland, Amsterdam.

Bertrand, M. and S. Mullainathan (2004), “Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination,” American Economic Review, 94(4), 991–1013. Cain, G.G. (1986), “The economic analysis of labor market discrimination: A survey,” in O.C. Ashenfelter and R. Layard (eds), Handbook of Labor Economics – Volume 1 (693–785). Elsevier/North-Holland, Amsterdam. Fershtman, C. and U. Gneezy (2001), “Discrimination in a segmented society: An experimental approach,” Quarterly Journal of Economics, 116(1), 351–377. Granberg, M. (2022), Discrimination in Hiring: Some Experiments, Perspectives, and Implications. Linköping University Press, Linköping. Heckman J.J. and P. Siegelman (1993), “The Urban Institute audit studies: Their methods and findings,” in Fix M. and R. Struyk (eds), Clear and Convincing Evidence: Measurement of Discrimination in America (187–258). Urban Institute Press, Washington, DC. Lang, K. and J.Y.K. Lehmann (2012), “Racial discrimination in the labor market: Theory and empirics,” Journal of Economic Literature, 50(4), 959–1006. Phelps, E.S. (1972), “The statistical theory of racism and sexism,” American Economic Review, 62(4), 659–661. Quillian, L., D. Pager, O. Hexel, and A.H. Midtbøen (2017), “Meta-analysis of field experiments shows no change in racial discrimination in hiring over time,” Proceedings of the National Academy of Sciences, 114(41), 10870–10875. Small, M.L. and D. Pager (2020), “Sociological perspectives on racial discrimination,” Journal of Economic Perspectives, 34(2), 49–67.

Ali Ahmed and Mats Hammarstedt

Labor market discrimination: sexual orientation Introduction

Current review studies have indicated that being a sexual minority (for example, gay men, lesbian women, bisexual men and women) is illegal in approximately 70 countries (Drydakis, 2022a, 2019). The reviews found the persistence of sexual orientation discrimination in everyday life, such as at school, work, looking for housing, and accessing healthcare or social services. In the European Union (EU), the United Kingdom (UK), the United States (US), Canada, and Australia, sexual minorities are found to experience more constraints in getting a job, higher unemployment, wage discrimination (especially gay men and bisexual men and women), lower job satisfaction, and more societal bias, including bullying and harassment, than their heterosexual counterparts (Drydakis, 2019). Internationally, studies utilizing experimental data found that applicants who identified as gay men or lesbian women during the initial stage of the hiring process are discriminated against in favor of comparable heterosexual applicants (Drydakis, 2019). Occupational access barriers varied for gay men between 3 percent and 40 percent, while for lesbian women the figure ranged from 6 percent to 27 percent. Hiring discrimination potentially leads to increased rates of unemployment and poverty, which can adversely affect physical and mental health. Indeed, studies indicated that sexual minorities experience poor physical and psychological conditions relative to their heterosexual peers (Drydakis, 2022b). Observational studies indicate that gay men are negatively affected by school-age bullying in terms of labour force participation, employment probability, and wages (Drydakis, 2014). Decompositions indicate that labour force participation gaps, employment gaps, and wage gaps between gay and heterosexual men can be explained by school bullying incidents. Bullying is a chronic problem for sexual minorities, which could continue from school to the workplace (Drydakis, 2019). School-age

bullying experienced by sexual minorities bore a positive association with workplace bullying, and a negative association with job satisfaction. Bullying can adversely affect the physical and mental health of sexual minorities. Moreover, it is found that during the EU economic recession there was an increase in homophobic attitudes and hate crimes, and annual aggregate unemployment enhanced discriminatory workplace experiences for sexual minorities (Drydakis, 2022a). Based on the available evidence, it seems that despite the passage of labor legislation against sexual orientation discrimination in the labor market, those who have a minority sexual orientation experience adverse workplace experiences compared to their heterosexual counterparts. These findings imply that legislative protection constitutes only a small step toward improving the employment circumstances and general well-being of sexual minorities, and highlights the need for effective policy interventions. The European Union regards the reduction of sexual orientation discrimination, and the entailed workplace and health inequalities, as a fundamental public policy goal and one of the foremost public challenges facing its member states (Drydakis, 2019).

Theoretical considerations

The labor market biases for sexual minorities bear a direct connection to the strength of firms’ antipathy against them (Becker, 1957), and/or uncertainties regarding their vocational behavior (Arrow, 1973). The taste hypothesis describes discrimination as a preference for which the discriminator is willing to pay (Becker, 1957). The taste for discrimination by employers is based on the idea that they want to maintain a physical distance from certain minority groups, or that they fear that their customers or co-workers dislike transacting with minorities (Becker, 1957). The theory indicates that the discrimination coefficient incorporates the influence of characteristics unrelated to productivity (Becker, 1957). Employers may offer sexual minorities a lower wage compared to heterosexual people in order to equalize the unit cost of labor once psychic costs are factored in. However, if the antipathy for the sexual minorities is high enough, employers will prefer not to employ them in their firms. So long as employers’ prejudices persist, the size

110

Labor market discrimination: sexual orientation  111

of the discrimination penalty will be directly related to the strength of the employers’ prejudice, and discrimination will be practiced consistently against the minorities by prejudiced agents. The statistical discrimination theory indicates that discriminatory actions are a result of a profit-maximizing response by employers to uncertainty about the actual productivity of minority groups (Arrow, 1973). In a world of imperfect information, employers face risks regarding employees’ profitability, and certain demographic characteristics can become inexpensive screening criteria (Arrow, 1973). If employers believe that there is a systematic difference between the sexual majority and sexual minorities in their reliability, productivity, and work commitment, then differences in employment outcomes may arise. If employers’ uncertainty regarding sexual minorities’ reliability, productivity, and work commitment is strong enough, they will not employ them. These acts are not based specifically on antipathy against sexual minorities, but rather are motivated by negative stereotypes and are grounded in what the decision maker believes to be valid inferences about productivity and risk (Arrow, 1973). The case for sexual minorities could become even more problematic if the information on an average group characteristic is unreliable or wrong.

Empirical evidence Earnings gaps between heterosexual and gay/bisexual men Since the seminal work of Badgett (1995), an increasing number of observational studies have examined the earnings differences between sexual majorities and minorities. In a review study, Drydakis (2019) found that between 1989 and 2014, gay men tended to experience 9 percent earnings penalties compared with heterosexual men. The earnings penalties varied from 16 percent in the US to 4 percent in the Netherlands. Drydakis (2022a) offered a meta-analysis by utilizing observational studies published between 2012 and 2020 in the US, UK, EU, Canada, and Australia, examining earning differences based on sexual orientation. The data sets covered 1991 to 2018. The outcomes indicated that gay men experienced

6.8 percent lower earnings than comparable heterosexual men, and bisexual men experienced 10.3 percent lower earnings than comparable heterosexual men. The study found that after 2010, gay men and bisexual men continue to experience 4 percent earnings penalties. In the US, Jepsen and Jepsen (2022), using the 2000–2019 American Community Survey on cohabiting individuals, found that the gap in wages by sexual orientation narrowed between 2001 and 2008. After that, the gap remained relatively flat for men in same-sex couples, at around 11 percent for annual wages, earnings, and income. The authors indicated that the persistence of a wage penalty for men in same-sex couples is concerning in the face of anti-discrimination policies and rising overall tolerance by Americans with respect to sexual orientation. Hypotheses state that if gay and bisexual people do not conform to traditional gender roles, such a situation might result in antipathy, unfavorable evaluations, and earnings penalties (Drydakis, 2015). For bisexual people, there is a belief that they are homosexual people who falsely declare a desire for the opposite sex to “improve” their position in the society (Drydakis, 2012). Bisexual people might face two penalties: one penalty for being attracted to same-sex partners, and another penalty for being seen as lying about their attraction to the opposite sex. Earnings gaps between heterosexual and lesbian/bisexual women Drydakis (2019) found that during the period 1989–2014, lesbian women earned on average 12 percent more than comparable heterosexual women. The highest earning premium was estimated to be 20 percent in the US, while the lowest was 3 percent in the Netherlands. Only in Australia and Greece did lesbian women receive lower earnings than comparable heterosexual women, with figures of 28 percent and 8 percent, respectively. Similarly, Drydakis’s (2022a) meta-analysis found that lesbian women’s earnings were 7.1 percent higher compared to the earnings of heterosexual women. On the other hand, bisexual women experienced 5.1 percent lower earnings than comparable heterosexual women. The study indicated that after 2010, lesbian women continue to experience 5.5 percent earnings premiums, Nick Drydakis

112  Elgar encyclopedia of labour studies

while bisexual women continue to experience 4 percent earnings penalties. Jepsen and Jepsen (2022) found that in the US, the premium for women in same-sex couples has declined, with a gap of around 8 percent in 2018. Arguments focusing on lesbian women’s earnings premiums in relation to masculine characteristics, which stereotypically characterize lesbian women and demonstrate leadership, have been utilized to evaluate their experiences (Drydakis, 2019, 2022b). In addition, if lesbian women invest more heavily in market-oriented human capital by staying in school and choosing a major that leads to higher earnings and longer working hours, such choices can influence their workplace outcomes. Moreover, a peripheral explanation for the lesbian earnings premium may revolve around women with children earning less than women without children. Lesbian women might prove less likely to have children than married women, so it is possible to earn more because of their commitment to the labor market. If this is the case, lesbian women might invest more in a workplace career. A puzzling issue is the earnings premiums found for lesbian women (Drydakis, 2022a, 2019). On the one hand, most qualitative-oriented studies suggested that lesbian women could experience adverse working experiences (for example, occupational access constraints, workplace bullying, job dissatisfaction) in the labor market. On the other hand, most country studies find that lesbian women earned more than heterosexual women of comparable skills and experience. Whether biased treatment of lesbian women can lead to salary premiums later in their careers is an open question. Job satisfaction gaps between sexual majorities and minorities Observational studies for the period 2007–2016 on sexual orientation and job satisfaction indicated that, on average, in the EU, the US, and Canada, gay men and lesbian women experienced lower levels of job satisfaction than heterosexual men and women (Drydakis, 2019). Gay men experience 14.8 percent lower job satisfaction compared to heterosexual men. Lesbian women face 12.2 percent lower job satisfaction compared to heterosexual women. Nick Drydakis

The satisfaction that gay and lesbian employees derive from their job may reflect how they respond to characteristics of their job and workplace. Because gay men and lesbian women experience high levels of workplace bullying and harassment, and face biases in earnings (especially gay men), promotions, and increased job responsibilities, these conditions may affect their job satisfaction levels (Drydakis, 2019).

Policy implications

Based on the taste theory, policymakers should adopt anti-discrimination legislation (Becker, 1957). Such policies raise the cost of discrimination by setting fines against homophobic firms. At the margin, this approach could discourage such firms from practicing unequal treatment, and might raise the sexual minorities‒majorities wage ratio (Becker, 1957). Policymakers can help through campaigns promoting respect and equality of treatment in the workplace, and by publishing annual data on progress toward equality objectives. Firms should evaluate recruitment and promotion policies to ensure equality of opportunity, and should address incidents of harassment. The statistical hypothesis indicates that unequal treatment can lessen if employers receive reliable information about employees (Arrow, 1973). Hence, firms should dedicate more resources to ascertaining employees’ productivity. Drydakis (2019, 2022b) reviewed how in the US anti-discrimination laws can reduce hourly earnings penalties for gay men relative to heterosexual men. In the EU, good employer–employee relations are shown to increase job satisfaction for gay and lesbian employees. In the US, the reviews indicated that anti-discriminatory laws prohibiting discrimination in the workplace based on sexual orientation can spur innovation, resulting in improved firm performance. Moreover, in the same region, the reviews indicated that workplace equality policies could positively influence customer satisfaction levels, firm performance, stock returns, market valuations, and marketing capability. In addition, antidiscrimination policies help to reduce homophobic incidents, and positively impact sexual minorities’ progression, self-esteem, income, and well-being. Badgett (2020) indicated that eliminating earnings penalties based on sexual orientation

Labor market discrimination: sexual orientation  113

requires not only a fundamental change in attitudes and behaviors, but also effective workplace policies and procedures. This situation could imply a critical role for social dialog between policymakers, trade unions, employees, employers, and minority population groups in improving the workplace experiences of sexual minorities (Drydakis, 2019). The social dialog should form an integral part of changing attitudes and establishing more inclusive workplaces, as it creates the opportunity to raise issues and determine solutions. Moreover, an accepting and welcoming family and school environment for sexual minorities could prevent internalization of the adverse effect of homophobic incidents, enabling them to meet developmental demands critical in dealing with homophobic demonstrations (Drydakis, 2019, 2022b). For sexual minorities, family acceptance could enable them to prevent, avoid, and/or deal with victimization, and avoid the adoption of maladaptive coping strategies and/or risky health behaviors (Drydakis, 2022b). In addition, public health services should ensure that policies are inclusive of the physical and mental health needs of sexual minority groups. Finally, addressing financial hardships for minority population groups should form part of the policymakers’ agenda. Finally, because the detrimental effects of school-age bullying may have lasting effects on human capital and employment, policymakers and educational authorities should address victimization through initiatives such as funding intervention programs and setting up legal requirements against school-age bullying (Drydakis, 2014). These initiatives should be viewed as one aspect of the increasing concern for the rights of people not to experience harassment, which is now being extended to children in school.

Gaps in current knowledge

Due to limited data sets on sexual minorities, there exists a dearth of studies on the topic. There exists a need for representative longitudinal data on sexual orientation in order to examine the level of earnings differences, poverty, unemployment, and well-being indicators as per sexual orientation and gender identity groups. Whether earnings discrimination, job dissatisfaction, bullying, and harassment against gay men and lesbians lessen over time in response to policy inter-

ventions is hard to evaluate on the basis of current quantitative research and in the absence of longitudinal data. Representative longitudinal data might allow policymakers to evaluate “what works” in reducing bias in the labor market. Prompt evaluations should determine how supportive families, schools, law, anti-bullying policies, and social and workplace strategies might boost sexual minorities’ progression. In addition, quantitative research on labor market outcomes is especially scarce for other sexual orientation classifications, including bisexual, transsexual, queer, intersexual, and sexually questioning. Without data, firm generalizations based on previous studies cannot be made for countries that have not yet been examined. Because more and more individuals have started to self-identify as queer, asexual, and sexually questioning, it is important to capture this information and to examine how labor market outcomes are affected by these additional sexual orientation classifications. It is vital to evaluate the channels through which inclusive workplace environments might reduce homophobia and biases, and increase vocational behaviors. Moreover, in family and school settings, how supportive parents and educators, inclusive curricula and facilities, and anti-bullying policies and strategies might increase sexual minorities’ self-esteem and labor market performance should be adequately examined for prompt evaluations. Nick Drydakis

References

Arrow, K.J. (1973), “The theory of discrimination,” in: O. Ashenfelter and A. Rees (eds), Discrimination in Labor Markets (3–33). Princeton, NJ: Princeton University Press. Badgett, M.V.L. (1995), “The wage effects of sexual orientation discrimination,” Industrial and Labor Relations Review, 48(4), 726–739. Badgett, M.V.L. (2020), The Economic Case for LGBT Equality: Why Fair and Equal Treatment Benefits Us All. Boston, MA: Beacon Press. Becker, G.S. (1957), The Economics of Discrimination. Chicago, IL: University of Chicago Press. Drydakis, N. (2012), “Sexual orientation and labour relations: New evidence from Athens, Greece,” Applied Economics, 44(20), 2653–2665. Drydakis, N. (2014), “Bullying at school and labour market outcomes,” International Journal of Manpower, 35(8), 1185–1211.

Nick Drydakis

114  Elgar encyclopedia of labour studies Drydakis, N. (2015), “Measuring sexual orientation discrimination in the UK’s labour market: A field experiment,” Human Relations, 68(11), 1769–1796. Drydakis, N. (2019), “Sexual orientation and labor market outcomes,” IZA World of Labor, 111(v2), 1–10 Drydakis, N. (2022a), “Sexual orientation and earnings: A meta-analysis 2012–2020,” Journal of Population Economics, 35(2), 409‒440.

Nick Drydakis

Drydakis, N. (2022b), “Social rejection, family acceptance, economic recession and physical and mental health of sexual minorities,” Sexuality Research and Social Policy, 19, 1318‒1340. Jepsen, C. and L. Jepsen (2022), “Convergence over time or not? U.S. wages by sexual orientation, 2001–2018,” Labour Economics, 74, 102086.

Labour market discrimination: ethnicity and race Introduction

new knowledge regarding ethnic and racial discrimination in the labour market.

Ethnic discrimination in hiring: where do we stand?

The extent to which individuals with different ethnic backgrounds are integrated in the labour market in a country is of importance for various reasons. From the individuals’ point of view, equal opportunities are decisive for employment and earnings prospects. For society, a well-functioning labour market integration will lead to ethnic groups born abroad, as well as their children, contributing to a country’s economy in various ways. However, in most Organisation for Economic Co-operation and Development (OECD) countries, differences in labour market outcomes can be related to an individual’s ethnicity or race. It is well known that certain ethnic and racial groups have lower employment propensities and lower employment earnings than the native population in the United States (US), in Canada, in Australia, as well as in different European countries. Much research has been devoted to enhancing the understanding of why ethnic and racial differences in labour market outcomes exist. A large literature has documented that factors such as differences in educational attainment, language skills and access to social networks all contribute to the existence of ethnic and racial labour market differentials in different countries. Furthermore, several studies from various countries also show that the existence of discrimination is an important explanation for ethnic and racial labour market inequalities. As shown in our entry on ‘Labour market discrimination: method and measurement’, detecting discrimination is not an easy task, and different methods such as field experiments, lab experiments and natural experiments have been used in the literature. This entry aims to present the main insights from a number of relatively new contributions in the area. We focus on field experiments that have been conducted in order to detect ethnic and racial discrimination in hiring in the US as well as in European countries. We thereafter discuss some areas in which more research is needed in order to gain

The pioneering work regarding racial discrimination in hiring by Bertrand and Mullainathan (2004) inspired a series of studies across the globe. Since then, a large number of field experiments have been conducted in several OECD countries, and the results point in the same direction: ethnic and racial minorities are discriminated against in the hiring decision (for an overview, see Baert, 2018). Turning to research from recent years, focus has often been on how hiring discrimination varies between different ethnic and racial minority groups; and how different types of labour market experiences, such as previous spells of unemployment, experience from different types of jobs, and language proficiency, are rewarded when certain ethnic and racial minority groups apply for job openings. In the US, research from recent years has documented hiring discrimination against job applicants with African, Hispanic or Arab-sounding names (Lambert and Akilade, 2019; Mobasseri, 2019; Leasure and Andersen, 2020; Yemane, 2020). Similar results are found in Australia (Chowdhury et al., 2020), as well as in European countries such as Germany (Koopmans et al., 2019), France (Manant et al., 2019), the Netherlands (van den Berg et al., 2020), Finland (Ahmad, 2020) and Sweden (Aldén et al., 2021). Thus, one conclusion from research using field experiments to detect ethnic and racial discrimination in hiring is that unequal treatment of minority groups is prevalent, and that this result is consistent over time and across countries.

Some recent takes on ethnic and racial discrimination

Besides documenting ethnic and racial discrimination, some recent studies have also examined important aspects of discrimination and the ways in which discrimination can manifest itself. One study in the US, for example, examined how the magnitude of discrimination against black people changed after the “ban the box” policy was imple-

115

116  Elgar encyclopedia of labour studies

mented in many US states (Agan and Starr, 2018). In a nutshell, the policy means that employers no longer are allowed to ask potential employees on the job application form whether they have been convicted for a crime in the past. This policy was implemented to ease the way for ex-offenders to enter the labour market. Black men, however, are overrepresented in this group. This before-and-after field experiment showed that discrimination against black men increased after the implementation of the “ban the box” policy. Hence, when employers do not know whether a jobseeker has a clean criminal record, they might form expectations about criminality on the basis of the jobseeker’s race. This study showed how race may interact with other characteristics, and how policies to some extent may backfire. Another example is the massive field experiment in the US where 83 000 fictitious applications were sent out to the largest employers in the country (Kline et al., 2022). The purpose of this study was to examine variation of discrimination against black people between companies. The study found substantial differences in the racial gap in response rates between companies. This means that a set of large firms are responsible for the severest form of discrimination, and that these discriminatory firms can be reliably identified.

What more do we need to know?

Future studies should investigate broader aspects of discrimination. For example, what other characteristics may play a role in relation to ethnic and racial discrimination, other than ethnicity and race itself: criminality, religion, and so on? Furthermore, the large body of evidence on ethnic and racial discrimination is mainly generated through field experiments. For recent reviews of empirical research on labour market discrimination, with a special focus on field experimental work, see Bertrand and Duflo (2017), Neumark (2018) and Baert (2018). The most significant and obvious drawback of labour market field experiments is that they only test for discrimination in the very first stage of the recruiting process: that is, whether there are any differences between two groups of people in the probability of receiving a positive response from employMats Hammarstedt and Ali Ahmed

ers. There is rather limited knowledge about the extent of ethnic and racial discrimination at the interview stage, during wage bargaining, on the job, and during promotions and job advancements. We also have very little knowledge about ethnic and racial discrimination that vulnerable groups of people may face prior to their job search. Is there discrimination in the education system, in access to trainee programmes, and in various labour market policy measures? This knowledge is valuable to identify the policies and efforts that are required, and to implement them in the best way in the right places. Mats Hammarstedt and Ali Ahmed

References

Agan, A. and S. Starr (2018), ‘Ban the box, criminal records, and racial discrimination: A field experiment’, Quarterly Journal of Economics, 133(1), 191–235. Ahmad, A. (2020), ‘When the name matters: An experimental investigation of ethnic discrimination in the Finnish labor market’, Sociological Inquiry, 90(3), 468–496. Aldén, L., S. Bastani and M. Hammarstedt (2021), ‘Ethnic background and the value of self-employment experience: Evidence from a randomized field experiment’, Oxford Bulletin of Economics and Statistics, 83(6), 1287–1310. Baert, S. (2018), ‘Hiring discrimination: An overview of (almost) all correspondence experiments since 2005’, in: S.M. Gaddis (ed.), Audit Studies: Behind the Scenes with Theory, Method, and Nuance, Cham, Springer, 63–77. van den Berg, C.J.W., L. Blommart, C.C.J.H. Biejleveld and S. Ruijter (2020), ‘Employment opportunities for ex-offenders: A field experiment on how type of crime and applicants’ ethnic background affect employment opportunities for low-educated men in the Netherlands’, Research in Social Stratification and Mobility, February, 100476. Bertrand, M. and E. Duflo (2017), ‘Field experiments on discrimination’, in: A.V. Banerjee and E. Duflo (eds), Handbook of Economic Field Experiments, Vol. 1, Amsterdam, Elsevier, 309–393. Bertrand, M. and S. Mullainathan (2004), ‘Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination’, American Economic Review, 94(4), 991–1013. Chowdhury, S., E. Ooi and S. Slonim (2020), ‘Racial discrimination and white first name adoption: Evidence from a correspondence study in the Australian labour market’, IZA Discussion Paper, No. 13208.

Labour market discrimination: ethnicity and race  117 Kline, P.M., E.K. Rose and C.R. Walters (2022), ‘Systemic discrimination among large U.S. employers’, Quarterly Journal of Economics, 137(4), 1963‒2036. Koopmans, R., S. Veit and R. Yemane (2019), ‘Taste or statistics? A correspondence study of ethnic, racial and religious labour market discrimination in Germany’, Ethnic and Racial Studies, 42(16), 233–252. Lambert, J.R. and E.Y. Akilade (2019), ‘Immigrant stereotypes and differential screening’, Personnel Review, 49(4), 921–938. Leasure, P. and T. Stevens Andersen (2020), ‘Race, criminal records, and certificates of relief: An experimental pilot study’, Deviant Behavior, 41(9), 1065–1083.

Manant, M., S. Pajak and N. Soulié (2019), ‘Can social media lead to labor market discrimination? Evidence from a field experiment’, Journal of Economics and Management Strategy, 28(2), 225–246. Mobasseri, S. (2019), ‘Race, place, and crime: How violent crime affects employment discrimination’, American Journal of Sociology, 125(1), 63–104. Neumark, D. (2018), ‘Experimental research on labor market discrimination’, Journal of Economic Literature, 56(3), 799–866. Yemane, R. (2020), ‘Cumulative disadvantage? The role of race compared to ethnicity, religion, and non-white phenotype in explaining hiring discrimination in the US labor market’, Research in Social Stratification and Mobility, 69, 100552.

Mats Hammarstedt and Ali Ahmed

Labour market discrimination: gender The theme of labour market discrimination against women occupies a growing place in contemporary economic analysis, driven by a combination of three trends. First, public opinion is increasingly sensitive to all forms of discrimination and to inequality. Second, while women’s educational levels are equal to, or even higher than, those of men, and their labour market participation is high, earnings inequalities are not disappearing, and women remain underrepresented among senior executives and board members. Third, from a methodological viewpoint, the persistence of inequalities unexplained by standard models has aroused considerable interest both in microeconomic research, with increased outreach towards other disciplines (psychology, sociology, history) and in statistical analysis (decompositions of the gender wage gap, distribution analysis, construction of counterfactuals, experimentation). Labour market discrimination against women is doubtless difficult to eradicate, but research on effective policy measures is advancing rapidly and opening encouraging new perspectives.

How to explain labour market discrimination against women?

Discrimination in the labour market against women has not always been illegal. In the United States, for example, as in Australia or New Zealand, it was allowed until the mid-20th century to pay a lower hourly wage to women than to men doing the same job. It was argued that men needed to earn a living for their families, whereas women only had themselves to look after. (This was also the viewpoint of the 18th century school of classical economists who calculated subsistence wages by sex.) Another legal, but less well-known, form of discrimination was the marriage bar in the United States, studied by C. Goldin (2021), which outlawed the employment of married women or had them fired after marriage. The practice concerned women in skilled jobs (teachers, clerical workers) and disappeared with the shortage of qualified personnel after the Second World War.

Discriminatory practices did not disappear after being condemned by laws, and represent a challenge for economic theory: employers deprive themselves of competent workers by excluding a large share of the available labour supply, and hence overpay employees selected on the basis of criteria other than their skills. How can this be explained? For Gary Becker, in his 1957 thesis defended in Chicago (see Becker, 2010), discrimination is a question of ‘taste’ and ‘preferences’ that may be manifested by employers, consumers or workers who refuse to mix with certain groups. The important point in Becker’s theory is that these preferences are costly, both for the individuals who pay the cost of frequenting a limited circle of people, and for society as a whole, which loses out by not making full use of each person’s skills. Becker (optimistically) argues that, as a consequence, increased competition eradicates discrimination by eliminating companies whose discriminatory behaviour reduces their profitability. This theoretical line was challenged by Phelps (1972) and then Arrow (1973), who sought to rationalise discriminatory behaviours as a problem of imperfect information that results in so-called ‘statistical’ discrimination. More precisely, not having the necessary resources to identify individual characteristics, employers judge people in terms of the mean characteristics of the group to which they belong. Being risk-averse, they prefer to hire someone whose capabilities are less uncertain (less variance around the mean). In the case of women, this uncertainty is centred on the risk of childbearing and a partial or total withdrawal from the labour market. The reasoning is simple: as women often leave their job after having a child, it is better to give them jobs with few promotion prospects rather than invest in their careers. With little chance of climbing the career ladder, there is logically less reason for them to stay in the company, thus confirming the initial assumption. It is a mechanism of self-fulfilling prophecy. In a similar model (Lazear and Rosen, 1990), the distribution of labour productivity is the same for men and women, but women perform slightly better on domestic tasks. This advantage works against them, as specialisation within the couple excludes them from promotion by their employers. All in all, whatever the theoretical model adopted,

118

Labour market discrimination: gender  119

labour market discrimination against women is an economic loss for society.

How to measure gender discrimination in the labour market?

An abundance of ever more econometrically sophisticated empirical approaches have been developed to estimate the extent of discrimination against women. The methods most widely used are based on the work of Blinder (1973) and Oaxaca (1973). The idea is to identify what can be explained by structural effects (for example, difference in education) in the estimated average pay gap. The part not explained by these factors indicates the existence of discrimination. As women’s educational levels rise and their working careers become more continuous, the variables of human capital (education, experience) are no longer sufficient to explain the gender wage gap in statistical terms. Other dimensions are brought into the analysis: horizontal occupational segregation, whereby women and men not do not work in the same sectors; and, above all, vertical segregation, whereby women have less lucrative careers than men and are underrepresented in the most senior positions. This is what has been popularized as the ‘glass ceiling’ effect, measured empirically by a growing unexplained gender difference in wage distributions (Albrecht et al., 2003). But using decomposition methods to measure discrimination faces a central difficulty: how much of the observed inequality is attributable to composition effects, and how much to pure discrimination? This is a thorny problem, as structural inequalities may themselves be the result of discriminatory practices. For example, a study by Ransom and Oaxaca (2005) examined the case of a unionized company in the United States that was prosecuted for discrimination under a class action lawsuit. As hourly wage rates were negotiated with the trade unions for each work position, there was no gender difference in wages and hence no pure wage discrimination. Yet almost no management positions were occupied by women. The company claimed that there were no female candidates for these positions; an argument rebutted by the court. In this case, discrimination was clearly proven. However, it is

generally difficult to differentiate between labour supply (an individual’s choice not to ask for promotion) and demand (employers’ reluctance to promote women). So, economists conducting empirical research on gender discrimination are drawing insights from other disciplines to look beyond ‘pure’ wage discrimination and identify the mechanisms that produce occupational segregation. Akerlof and Kranton (2000) use sociology to explain the role of social norms in gendered choices of education or occupation. In short, they see the social norm as an argument of the individual’s utility function. Infringing this norm – for example, for a woman, choosing a masculine career – has a double cost: for colleagues who respond negatively to this infringement, and for the person concerned, who must both contravene the norm and cope with the negative externalities that ensue. Occupational segregation persists as a consequence. Similar reasoning applies to choices of educational specialisation, and even to gendered behaviour from early childhood. Collaboration between economics and psychology has also opened a new and fruitful avenue of research. Laboratory and field experiments regularly find that women are more risk-averse, less competitive and more cooperative (for an overview of this research, see Bertrand, 2010). These psychological traits shaped by social norms may explain why women are less successful than their male counterparts in negotiating starting salaries (Babcock and Laschever, 2003), pay increases or promotions.

The family pay gap

In recent years, economists have focused their attention on the impact of childbirth on labour market inequalities. Actually, the family pay gap (or motherhood penalty) is a longstanding topic of interest (for an overview, see Ponthieux and Meurs, 2015), and has been studied by comparing panel data on the trajectories of women with and without children, and by measuring differentials at various stages of the life course, frequently using standard Oaxaca decompositions. This question has recently been revisited using the event studies technique to compare what happens after a birth, whatever the year in which it occurred (so that comparisons can be made across women who have a child in Dominique Meurs

120  Elgar encyclopedia of labour studies

different years). Kleven et al. (2019) follow cohorts of women and men and measure the gender earnings differential over time. Focusing on Denmark, they estimate the child penalty, that is, the total loss of earnings for mothers due to childbearing compared with women who did not experience the same events at the same time. This penalty encompasses labour market withdrawal, transition to part-time working, and earnings losses due to slower career advancement or missed promotions. A similar analysis was conducted for fathers, with the finding that parenthood has practically no effect on their careers. This paper shows that almost all the gender earnings gap not linked to differences in human capital is now attributable to motherhood and its effect on work commitment. Clearly, it is not the children who are responsible for this state of affairs, but the way in which time is shared within the family. In her most recent book, Goldin (2021) shows how this trade-off between work and motherhood has evolved for the successive cohorts born since the start of the 20th century. Time is a multidimensional constraint: how should the 24 hours in a day be divided between work time and childcare time? How should unexpected events and irregular working hours be accepted? How should the life cycle be planned? When should one begin a career, or start a family? While in the early 20th century women had to choose between a career or a family, in the early 21st century effective contraception has made it easier to control the timing of births and to achieve a work‒family balance. The risk of childlessness increases, however, if childbearing is postponed for too long.

Current debates

How are economists addressing the question of gender discrimination today, and where is the cutting edge of research? The debate initiated by Goldin (2014) on whether equality will be achieved through changes in the workplace or in society is far from settled. Goldin’s argument is as follows: some employees in non-standardised jobs are more productive, so their overtime pay is disproportionately high. Within households, therefore, it is advantageous for each member to specialise in order to maximise income. But if work content is standardised and workers are interchangeable, this advantage disappears, Dominique Meurs

providing opportunities to rebalance the time devoted to children and domestic tasks. It is for employers, therefore, to address the question of occupational equality, with a change in job content and in work organisation. For other researchers, labour supply is the heart of the question. Observed inequalities cannot be eradicated without behavioural changes and a shift in social norms towards greater paternal involvement in childcare. As for public policy, the debate no longer centres on prohibiting discrimination. Discriminatory practices have long been outlawed, and can be detected using a range of testing techniques. Thus, audit and correspondence studies are widely used to document discriminatory attitudes of employers and their reluctance to hire women with comparable qualifications as men (see Bertrand and Duflo, 2017). Today, the aim is to develop policies to modify habits and social norms. It is often possible to estimate the causal effect of these measures using well-established econometric methods (differences-in-differences, regression discontinuity design, and so on), with public decision-makers now becoming less reluctant to collect and share the necessary data. These policies mainly concern three major areas: education, corporate career management and organisation of family life. In education, the challenge is to reduce gender bias in the choice of specialty subjects, and to encourage girls to enter scientific careers. Possible measures include early encounters with role models. For example, high school visits by female scientists have proved highly effective in increasing the share of girls opting for scientific tracks. In the workplace, the chosen method is often to impose quotas for female board members or senior management positions in order to break through the glass ceiling. The effects of these policies are unclear, and there is little evidence of improved female career advancement in companies where more women hold senior positions. Research is also focusing on promotion processes, based on the assumption that women are more hesitant to apply for promotion than men. Hospido et al. (2022) show that when the promotion process is changed (all promotable individuals are automatically considered for promotion rather than only those who decide to apply), a higher proportion of women are promoted. With the emergence of movements such as #metoo,

Labour market discrimination: gender  121

current research is also starting to encompass toxic workplace behaviours, their impact on working lives, and the methods that can be used to change workplace attitudes. Last, in the domestic sphere, parental leave and childcare provision do not appear to reduce inequalities (Kleven et al., 2020). Extending paternal leave could potentially produce a rebalancing of roles within the family, but it is still too early to draw conclusions in this respect; the first tests (the ‘daddy-month’ in Sweden) did not find evidence of a more balanced gender division over the long term (Ekberg et al., 2013). This experiment was limited, however, and behaviours may have evolved since then. In conclusion, when examining the numerous policies set in place, the main task of economists today is to look beyond identifying causal effects, and to rank the relative efficacy of the various approaches developed to eradicate discrimination against women. Dominique Meurs

References

Akerlof, G.A. and R.E. Kranton (2000), ‘Economics and identity’, Quarterly Journal of Economics, 115(3), 715–753. Albrecht, J., A. Björklund and S. Vroman (2003), ‘Is there a glass ceiling in Sweden?’, Journal of Labour Economics, 21(1), 145–177. Arrow, K.J. (1973), ‘The theory of discrimination’, in O. Ashenfelter and A. Rees (eds), Discrimination in Labor Markets. Princeton University Press, 3–33. Babcock, L. and S. Laschever (2003), Women Don’t Ask: Negotiation and the Gender Divide. Princeton University Press. Becker, G.S. (2010), The Economics of Discrimination, 2nd edition. University of Chicago Press. Bertrand, M. (2010), ‘New perspectives on gender’, in O. Ashenfelter and D. Card (eds), Handbook of Labor Economics, Volume 4b. Elsevier, 1543–1590.

Bertrand, M. and E. Duflo (2017), ‘Field experiments on discrimination’, in A.V. Banerjee and E. Duflo (eds), Handbook of Economic Field Experiments. Elsevier, 309–393. Blinder, A. (1973), ’Wage discrimination: reduced form and structural estimates’, Journal of Human Resources, 8(4), 436‒455. Ekberg, J., R. Eriksson and G. Friebel (2013), ‘Parental leave ‒ A policy evaluation of the Swedish “Daddy-Month” reform’, Journal of Public Economics, 97, 131–143. Goldin, C. (2014), ‘A grand gender convergence: Its last chapter’, American Economic Review, 104(4), 1–30. Goldin, C. (2021), Career and Family: Women’s Century-Long Journey toward Equity. Princeton University Press. Hospido, L., L. Laeven and A. Lamo (2022), ‘The gender promotion gap: Evidence from central banking’, Review of Economics and Statistics, 104(5), 981–996. Kleven, H., C. Landais, J. Posch, A. Steinhauer and J. Zweimüller (2020), ‘Do family policies reduce gender inequality? Evidence from 60 years of policy experimentation’, National Bureau of Economic Research Working Paper 28082. Kleven, H., C. Landais and J.E. Søgaard (2019), ‘Children and gender inequality: Evidence from Denmark’, American Economic Journal: Applied Economics, 11(4), 181–209. Lazear, E.P. and S. Rosen (1990), ‘Male‒female wage differentials in job ladders’, Journal of Labor Economics, 8(1, Part 2), S106‒S123. Oaxaca, R. (1973), ‘Male‒female wage differentials in urban labor markets’, International Economic Review, 14(3), 693‒709. Phelps, E.S. (1972), ‘The statistical theory of racism and sexism’, American Economic Review, 62(4), 659–661. Ponthieux, S. and D. Meurs (2015), ‘Gender inequality’, in A.B. Atkinson and F. Bourguignion (eds), Handbook of Income Distribution Vol. 2. Elsevier, 981–1146. Ransom, M. and R.L. Oaxaca (2005), ‘Intrafirm mobility and sex differences in pay’, Industrial and Labor Relations Review, 28(2), 219–237.

Dominique Meurs

Labour market integration of immigrants Introduction

In the wake of globalisation, the number of immigrants around the globe has increased and the labour market integration of different types of immigrant has been of major public and political concern as well as of key interest for research. Labour market integration can be defined in different ways, but is commonly understood to be when immigrants reach parity or close parity with natives concerning key labour market indicators such as labour force participation, employment or unemployment level, wages and earnings, and are also an education‒occupation match. The questions that are asked when endeavouring to understand labour market integration are: Do immigrants integrate into the receiving society labour markets? Do immigrants catch up with natives, and how quickly do they reach parity? Are there differences between immigrant groups? Which factors, on both the supply and the demand side, enhance or hinder this integration process? Can immigration or integration policy be part of the answer?

Basic theory

The large majority of studies analysing immigrant labour market integration are rooted in human capital theory, where immigration is seen as an opportunity and it is expected that immigrants are positively selected according to their abilities, skills and health. However, opportunities and restrictions created by immigration law in the receiving country, together with the individuals’ ability, schooling, skills, age at arrival, health, occupational status and reasons for migration, can influence the actual immigration flow and subsequently the key indicators of labour market integration such as employment and earnings prospects, in both the short and the long term. Moreover, a growing diaspora diminishes selection, and non-economic migrants, family reunion and/or refugee migrants are all selected on important factors other than economic ones. Typically, upon arrival in the country, immigrants can be expected to be at an employment and earnings disadvantage rel-

ative to natives, because they lack certain country-specific skills and information that natives have; in other words: to what extent is a transferability of skills possible? Over time, as immigrants learn about the host-country institutions, learn how to best apply the skills they have in the new country, improve their host-language proficiency and gain other country-specific human capital, their employment levels and earnings may catch up with, and possibly overtake, those of comparable natives. This theoretical model is called the immigrant assimilation model, and is still widely used to study immigrant labour market adjustment. An extension of this model, the so-called immigrant human capital investment model, is the proposition that skilled immigrants with low-transferability skills have lower opportunity costs in adjusting to the host society skills, and will therefore invest more in them than high-skilled immigrants who have higher-transferability skills. Immigrants with low-transferability skills will therefore also have lower initial earnings, but higher earnings growth. Since the decision to migrate is often a family one, a family perspective is important for a more complete analysis of labour market integration; this perspective includes decisions by families, and the activities of wives or partners in the household. For example, the family investment hypothesis (Duleep and Sanders, 1993) proposes that women, as secondary earners in a household, engage in labour market activities to finance the investment in their husbands’ (primary earners) host-country-specific human capital, which increases a family’s future income. In so doing, women become primary earners, at least temporarily.

Empirical findings

From a supply-side perspective, Chiswick (1978), using the 1970 United States (US) census, estimates – with control for standard individual characteristics – that white male immigrants overtake natives of similar characteristics within 15–20 years. However, Borjas (1985) counters these findings with the argument that, in a single cross-section, a positive effect of years since migration could be picking up not only assimilation but also changes in cohort quality. The argument is that earlier cohorts may simply perform better at every stage than recent

122

Labour market integration of immigrants  123

ones, so that a difference in their earnings reflects a difference not only in the time they have had to assimilate, but also in ‘quality’. Moreover, Borjas finds much lower catch-up rates than those found by Chiswick. He attributes most of the cross-sectional difference between cohorts to significant declines in the ‘quality’ of successive cohorts of immigrants going to the United States, and to the change in immigration policy which prefers family reunion migration. Implementing the immigrant human capital investment model, Duleep (2015), using a synthetic cohort and following them over subsequent censuses, found strong evidence to show that when immigrants’ entry earnings decreased, earnings growth increased. Empirical support for the family investment hypothesis was shown in early studies for the US and Canada, which found that immigrant women initially worked longer hours but decreased them in line with years since migration, and that immigrant men with potentially high earnings growth had higher labour force participation by their wives. Longitudinal studies using the family investment hypothesis refute the results from earlier cross-sectional studies.

Extensions

While important as a framework, current research has pointed to several shortcomings. Many studies on the economic inte-

gration of immigrants in host counties show unexplained differences between groups with similar human capital and other observable characteristics. Moreover, similar immigrant groups have different labour market outcomes in different contexts or countries. This research has extended the scope of possible explanations for immigrants not catching up with the employment and earnings levels of the native populations, as well as explaining the variation in labour market integration by group and host country. The social capital and network levels of immigrants on both supplyand demand-side factors such as the economic structure of the labour market, ethnic labour market segmentation, and different kinds of discrimination towards immigrants and particular ethnic or religious groups, are additional factors studied in a number of countries for both entrance into the labour market as well as earnings differentials. Besides, the large number of refugees in the world means that they have become a larger part of the immigrant intake, and the number of studies tracing the employment patterns and earnings trajectories of refugees compared to other immigrants has increased substantially (Brell et al., 2020). Some stylised facts for Sweden can be useful to demonstrate the integration process. As can be seen in Figure 1, both male and female labour migrants start at a higher employment level upon arrival and, after about ten years

Source: Author’s own calculations based on Swedish register data provided by Statistics Sweden.

Figure 1

Employment level for different immigrant categories by year since migration, men and women, Sweden, 2016

Pieter Bevelander

124  Elgar encyclopedia of labour studies

in the country have an employment level of 75 per cent. Family reunion migrants, not selected primarily for labour market reasons but having family or friends acquainted with the host society, have a low employment level at entry but subsequently increase their employment level, though this never reaches the level of labour migrants. Refugee migrants as well as their families have the lowest levels of employment at entry. This group have years in the host country at higher employment levels, but never reach the level of labour migrants. In short, many studies conducted in Europe, North America and Australia which analyse the differences between refugees, family reunion migrants and labour migrants have concluded that refugees are in a disadvantaged position when it comes to labour market entry and earnings trajectories. The fact that refugees arrive in different and often difficult circumstances, and have not primarily migrated for labour market reasons but are admitted according to other (non-economic) criteria and endure settlement policies, appears to affect their labour market integration. Moreover, as both the migration and the admittance processes can be lengthy and cumbersome, health issues and the loss of human capital can hinder individuals’ adaption to the labour market of a new country. Whether refugees and family reunion migrants obtain permanent or temporary residence can also affect their investment in the host language and receiving-country-specific human capital, and their labour market integration process (Bevelander and Irastorza, 2021).

Future research

Future research should thus focus on the accumulation of statistical data for each immigrant entry category, and on the analysis of specific pre- and post-migration aspects of a successful labour market integration of refugees. Longitudinal statistical information that makes it possible to follow individuals over time is of crucial importance to assess immigrants’ labour market entrance, as well as the occupational and income mobility of

Pieter Bevelander

refugees versus other immigrant categories and the native population. Comparative country research is also necessary in order to assess whether immigration and refugee integration policies are efficient, and whether they induce the desired effects. Studies going beyond supply-side models and including various demand-side factors such as the state, or the change in the sending and receiving economy, or other macro-economic factors such as welfare provisions, and how receiving societies accept immigrants, are still needed to further understand the labour market adjustment of immigrants. Finally, the legal status of the individual immigrant, the permanence versus temporariness or even their undocumented status and the possibility for them to acquire citizenship of the host country, could affect country-specific human capital investment and duration of stay, and in turn, affect the human capital accumulation of the children of immigrants and their labour market integration. Pieter Bevelander

References

Bevelander, P. and N. Irastorza (2021), ‘The labour market integration of humanitarian migrants in OECD countries: an overview’, in: K. Kourtit, B. Newbold, P. Nijkamp and M. Partridge (eds), The Economic Geography of Cross-Border Migration (157–184). Cham: Springer. Borjas, G.J. (1985), ‘Self-selection and the earnings of immigrants’, American Economic Review, 77(4), 531–553. Brell, C., C. Dustmann and I. Preston (2020), ‘The labor market integration of refugee migrants in high-income countries’, Journal of Economic Perspectives, 34(1), 94–121. Chiswick, B.R. (1978), ‘The effect of Americanization on the earnings of foreign-born men’, Journal of Political Economy, 86(5), 897–921. Duleep, H.O. (2015), ‘The adjustment of immigrants in the labor market’, in: B.R. Chiswick and P.W. Miller (eds), Handbook of the Economics of International Migration (105–182). Amsterdam: Elsevier. Duleep, H. and S. Sanders (1993), ‘The decision to work by married immigrant women’, Industrial Labor Relations Review, 46(4), 677–690.

Labour supply and taxes Introduction

The literature on labour supply and taxes is vast and continues to expand. It places a strong emphasis on the taxes and benefits that directly affect pay and unearned income, including income taxes and social security contributions, tax credits, negative income taxes and benefits for low-income and non-working individuals and families. This entry briefly overviews just part of that literature. The focus is on empirical work using microdata to quantify the impacts of taxes and benefits on labour supply, its theoretical underpinnings and empirical strategies. Perspectives from the equally active macro and optimal taxation literatures are not discussed here. More complete reviews can be found in Blundell and MaCurdy (1999), Meghir and Phillips (2009), Keane (2011), Saez et al. (2012), among others.

Theoretical background

Labour supply models provide a framework to measure and understand how taxation shapes incentives to work and affects behaviour. The discussion below briefly overviews these models and their use for tax analysis; much more thorough discussions can be found in Blundell and MaCurdy (1999) and Keane (2011). Basic model The basic model of labour supply has been widely used as a workhorse for studying the design and effects of income taxes and welfare benefits. It formalises the problem of individuals choosing how to allocate their time between market work and non-market activities (hereafter: leisure). Key to this problem is the trade-off between consumption and leisure: while individuals want more of both, they can only exchange between the two through the supply of labour. Two essential elements determine this trade-off. The first is preferences for consumption and leisure, represented in a utility function. The second describes income available for consumption, which includes unearned and earned income. The latter increases with labour supply at the net-of-tax hourly wage rate.

The simplest version of the model posits a linear income tax (constant tax rate) and considers responses to changes in wages at the intensive (hours) margin only. The resulting labour supply function depends on unearned income and the net-of-tax hourly wage rate. It may also vary with exogenous individual characteristics that shape their valuation of consumption and leisure. Even in this stylised framework, the direction of labour supply responses to changes in the marginal return to work is undetermined. Depending on their circumstances, income or working hours, the same individual may increase or reduce their labour when hourly wages change (Manski, 2014). This feature reflects the action of two well-known opposite forces: substitution and income effects. On the one hand, an exogenous drop in hourly wages reduces the cost of leisure and should therefore increase demand for it, while reducing labour supply. In turn, a lower wage reduces take-home pay. The poorer individual may want to compensate for the loss in income by working more. Despite the theoretical possibilities, the consensual view among empirical labour and public economists is that substitution dominates in practice, and so labour supply tends to increase with net wages. Progressive taxes and welfare benefits It is conceptually simple to allow for a more realistic description of the tax system. An essential extension considers progressive income taxes that form piecewise linear (convex) budget sets. Early work demonstrates that the resulting labour supply function is similar to that obtained for proportional taxation, but where the marginal return to work is that realised at the individual-specific tax bracket, and unearned ‘virtual’ income is obtained by extending the budget line in that bracket to zero hours (Hausman, 1985). In this setting, individual labour supply depends on the entire tax schedule, because the virtual income is a function of both the relevant marginal tax rate and all those that apply at lower earnings. This implies that labour supply responses to a specific change in the tax schedule depend on the overall tax system. Meghir and Phillips (2009) discuss these dependencies. Means-tested benefits can also be considered. These are transfers to low-income

125

126  Elgar encyclopedia of labour studies

individuals that are gradually withdrawn as income increases (for example, income support, tax credits). Typically, the phasing out of benefits happens at a rate that adds to the income tax, thus effectively increasing the marginal tax rate that applies in the withdrawal region of earnings. In most cases the resulting budget set is no longer convex, since marginal tax rates can be higher where benefits are phased out further up the earnings distribution. This is not inconsequential: while marginal changes in net-of-tax wages carry only small labour supply responses when the budget set is convex, the same does not necessarily apply to cases where budget sets are not convex. Larger responses involving switching tax brackets or stopping working altogether are expected, invalidating the typical marginal analysis of taxation. Dimensions of labour supply responses Taxation may impact upon dimensions of labour supply other than working hours. At least as empirically relevant, particularly for women and older workers, are extensive margin responses (employment). Accounting for extensive and intensive margin of labour supply requires that selection into work is explicitly modelled by taking into account fixed costs of work, including those created by the benefit system. Other possible margins of response are not detectable in either participation or working hours. These include effort, income shifting, tax avoidance or business organisation, and so on. Responses along these dimensions can be most significant among high earners. One way of summarising total responses along the many possible margins is to measure the sensitivity of taxable income to net wages (see Saez et al., 2012, for an overview of this literature). Extensions The basic model of labour supply has been extended to represent more realistic setups and provide richer, more nuanced descriptions of the responses to work incentives. The discussion below identifies some of these developments. First is intertemporal labour supply. Modern life-cycle models of labour supply explicitly allow for market work to accumulate productive skills, and consider employment and hours choices in the presence Monica Costa Dias

of uncertainty and credit constraints (see Keane, 2011, for a detailed discussion of these models and their use for tax analysis). In these models, human capital accumulated in work partly determines future wages and provides insurance against income shocks. By the same token, policies incentivising work at one stage of life – for instance, tax credits promoting work among mothers – can have long-lasting effects if they permanently impact upon wage rates. These models allow for the long-term effects of tax reforms to differ from those realised immediately, because of how wages propagate and magnify initial effects. It also suggests that individual responsiveness to changing incentives to work will vary over the course of life, as the horizon to reap the returns from current work gradually declines. Second are frictional labour markets. Frictions limit the ability of workers to locate available jobs or adjust hours of work, hence partly suppressing labour supply responses to tax reforms by putting a cost on these adjustments (Chetty et al., 2011). In the presence of frictions, larger reforms will have a disproportionate impact by increasing the payoff to adjust labour supply. Like in models of skill formation, frictions can generate differences between short- and long-term responses to changes in taxes, as the passage of time allows for information to spread or new job offers to arrive. Third is benefit take-up. If agents face barriers to claim their benefit entitlement due to stigma or other costs, then the budget constraint they face varies with claiming status. Measuring the responsiveness of labour supply to benefit reform therefore needs to consider how it affects claimants and claiming behaviour. To explain variation in take-up across benefits and individuals, models have added detail of the policy parameters, including on administrative and application processes, and considered the role of information and stigma. Ko and Moffitt (2022) discuss this literature.

Empirical strategies

Much of the empirical literature has focused on estimating labour supply elasticities with respect to wages or marginal tax rates, with most studies aiming to quantify a single (at least at the broad group level) parameter. As many have noted, these elasticities are not

Labour supply and taxes  127

unique or structural parameters (Attanasio et al., 2018). Instead, they are heterogeneous, vary with wages and other individual characteristics and circumstances, and depend on the macroeconomic environment and the tax system as a whole. Hence, labour supply elasticities that apply in one context do not necessarily apply in another. The view, however, is that single ‘average’ elasticities summarise aggregate responses to changes in wages within the economic environment in which they are measured, and are therefore key statistics to quantify the potential impact of small reforms and to inform tax design. The literature sets out various concepts of labour supply elasticity, depending on the margin of response (for example, employment, hours, taxable income), what is varied (before- or after-tax wages, marginal tax rates) and what is kept constant (total income, unearned income, utility). In a life-cycle context, the timing of the change, and whether it is anticipated or permanent, also matter. The relevant elasticity depends on the question at hand. Two key concepts for tax analysis are the Marshallian and Hicksian elasticities. In a dynamic framework, the Marshallian elasticity measures the sensitivity of labour supply with respect to an unexpected permanent change in wages, when agents are not compensated for the resulting changes in their lifetime income or utility. Marshallian elasticity is useful to study reforms that result in some redistribution of resources; it is also smaller than other elasticities because it includes income effects. The Hicks elasticity is similar, but keeps lifetime resources constant. It is relevant to evaluate the impacts of policies that keep the distribution of resources across individuals unchanged. The two elasticities can also be defined in a static framework, which would be relevant for tax analysis in cases where that formulation captures the relevant determinants of labour supply (for instance, in cases where workers are liquidity-constrained). Another widely used concept is that of Frisch elasticity, also called the intertemporal elasticity of labour supply. It measures responses to anticipated marginal changes in wages, keeping the marginal utility of wealth constant. This elasticity is less relevant for tax analysis and will not be further discussed here.

There are two broad approaches to estimate labour supply elasticities using individual level data. The microeconometric approach directly specifies a regression model of labour supply on log net-of-tax wages, a measure of unearned income, and a set of other (exogenous) characteristics of the individual and environment. The coefficient of interest is that on the log wage and refers to the intensive margin elasticity. Especially for high earners it is also common to estimate the elasticity of taxable income. Estimation needs to tackle various sources of biases, including (but not exclusively) unobserved heterogeneity in preferences that can be correlated with wages and unearned income; measurement error in wages per hour, perhaps due to division bias; non-linear income taxes causing net-of-tax wages to vary mechanically with labour supply; and the wages being observed only for those in work, who are unlikely to be randomly selected. All these issues affect the interpretation of regular ordinary least squares (OLS) estimates. Various solutions have been advanced, mostly exploiting tax reforms and panel data to single out unanticipated variation in wages and unearned income (Gruber and Saez, 2002). With progressive taxation, one expects a mass of workers to choose labour supply precisely at the kink where the marginal tax rate changes. These are workers who would optimally choose to work less than they actually do at the higher tax rate, and more at the lower one, and end up bunching at the kink. For them, the marginal optimality conditions implicit in the linear labour supply function do not hold. Bunching methods explicitly exploit this feature to estimate compensated labour supply elasticities from the excess mass at the kink (Saez, 2010). Alternatively, the structural dynamic approach relies explicitly on a fully specified life-cycle model of economic behaviour to estimate labour supply elasticities. Originally, the models were very stylised and more often used in the macro literature. Here the focus is on models that build in fine individual-level heterogeneity and are estimated on microdata. Modern versions incorporate intensive and extensive labour supply choices, related choices including those on family formation, fixed costs of participation in work, human capital formation, uncertainty, credit Monica Costa Dias

128  Elgar encyclopedia of labour studies

constraints, and detailed descriptions of the relevant tax and benefit systems to describe the incentives faced by workers. By explicitly considering the (contemporaneous and intertemporal) trade-offs involved in labour supply choices, these models provide a structural interpretation to estimated elasticities. When using structural models, labour supply elasticities are not directly estimated; instead, the parameters determining preferences, wages and other structural relationships are. Individual-level elasticities can then be calculated from the parameterised model, and grouped to construct aggregate parameters. The additional insight provided by the structural approach comes at the cost of strong assumptions about preferences, information and the environment. Some of these may be selected more for their convenience than their content, and are therefore often disputed. Empirically, the identification of model parameters is also not always clear. Recent work is addressing these critics by combining panel data with tax variation for estimation, and by systematically testing structural assumptions in validation exercises, against experimental or quasi-experimental evidence.

Empirical evidence

Despite the mountain of work estimating labour supply elasticities, there is still a lively debate as to what their values are. It is perhaps unsurprising that studies disagree on the size of elasticities, given that these are heterogeneous parameters which also depend on the economic environment and model specification. Yet, the conflicting views can be categorised in two main camps: those in the microeconometric literature, claiming that elasticities are small, often close to zero; and those in the macro structural literature, claiming that they need to be one order of magnitude larger, between 2 and 5, to capture business cycle and between countries variation. Most of the controversy relates to the intertemporal Frisch elasticities and will not be pursued here (Chetty et al., 2012, and Keane and Rogerson, 2015, review this work). Various studies summarise estimates of labour supply elasticities (e.g., Meghir and Phillips, 2009; Keane, 2011). For women, much work has focused on mothers, parMonica Costa Dias

ticularly in low-income families, and on the impacts of tax credits and other welfare programmes. Microeconometric estimates of intensive margin uncompensated elasticities have mostly been in the range of 0.2 to 0.4. Extensive margin elasticities are generally higher, often around or above 0.5. Recent structural estimates reach similar conclusions, but also document considerable heterogeneity in both hours and participation responses. For instance, Attanasio et al. (2018) estimated a median uncompensated intensive margin elasticity of 0.42 for women, varying between 0.22 and 0.80 when looking across quartiles of the distribution of elasticities and pay. Extensive margin uncompensated elasticities were also found to vary strongly with earnings, between 0.17 and 0.62 for higher- and lower-paid women, respectively. Other studies found that the participation of lone mothers is especially elastic, with estimates for the extensive margin uncompensated elasticities as high as 1.3 (Blundell et al., 2016). For men, most studies favour hours rather than employment responses, in line with the view that their traditional breadwinning role in the family implies negligible responses on the latter. Estimates point to muted uncompensated wage elasticities for working hours, at or below 0.1, and small income effects. However, estimates of the elasticity of taxable income, which collapse the many dimensions of labour supply adjustments and other responses to changes in taxes, paint a different picture: they are typically above 0.3, and can be close to 1 for high earners. The magnitudes of these responses highlight that the design of taxes and benefits should take into account features other than the tax rates and tax brackets. A few papers questioned the magnitude of microeconometric estimates exploiting tax variation, and suggested that they may be downward biased. One view is that workers face barriers to adjust labour supply, for example, due to frictions, inattention, hours constraints, and so on (Chetty et al., 2011). This hypothesis is consistent with various empirical facts, including that workers whose labour supply is easier to adjust, such as the self-employed, respond by more. It implies that, at least in the short term, estimated responses provide a lower bound for the labour supply elasticities.

Labour supply and taxes  129

Possible avenues for future work

Without intending to be exhaustive in any way, the discussion below highlights some promising areas for further research. Much of the existing empirical work focuses on estimating a single parameter that summarises the aggregate impact of marginal tax rates on labour supply or taxable income. In general, the magnitude of these effects depends on features of the tax system other than tax rates. These other features define, for instance, the tax base, opportunities for tax avoidance, benefit eligibility, or the salience of a benefit or tax. Reforms often involve changing various parameters simultaneously, and will have impacts on margins other than labour supply. Widening the focus of analysis to look systematically beyond tax rates, labour supply and taxable income responses will provide further insight into how the different policy instruments interact. Empirical structural models provide a framework for explicitly considering many margins of response and the detail of the tax and benefit system. They are also well suited to study heterogeneity and the distributional effects of taxation. However, existing models still simplify important dimensions for tax analysis. For instance, they seldom consider how decisions are taken within families, to supply labour and distribute resources; most models assume away frictions that limit workers’ ability to respond to incentives; behavioural biases such as salience or misperceptions have largely been ignored too. All these have been shown to be relevant determinants of choices, and need further attention. Other relevant topics have received little attention in the empirical literature. For instance, not much work exists on the incidence of income taxes and in-work subsidies. Interestingly, this topic has received more attention in relation to payroll taxes, which are explicitly split between employers and employees (see Uusitalo’s entry on ‘Payroll taxes: incidence and employment effects’ in this Encyclopedia for a review). Similarly, not much work exists on how the design of taxes and benefits may affect the types of jobs or the hours of work that workers can choose from. Considering these and other market-level effects would improve understanding of the impacts of taxation, and help

to bridge the gaps between the micro and macro literatures. Monica Costa Dias

References

Attanasio, O., P. Levell, H. Low and V. Sánchez-Marcos (2018). ‘Aggregating elasticities: intensive and extensive margins of women’s labor supply’, Econometrica, 86(6), 2049–2082. Blundell, R. M. Costa Dias, C. Meghir and J. Shaw (2016). ‘Female labor supply, human capital, and welfare reform’, Econometrica, 84(5), 1705–1753. Blundell, R. and T. MaCurdy (1999). ‘Labor supply: a review of alternative appproaches’, in O. Ashenfelter and D. Card (eds), Handbook of Labor Economics, Vol. 3, Elsevier Science. Chetty, R., J. Friedman, T. Olsen and L. Pistaferri (2011). ‘Adjustment costs, firm responses, and micro vs macro labor supply elasticities: evidence from Danish records’, Quarterly Journal of Economics, 126(2), 749–804. Chetty, R., A. Guren, D. Manoli and A. Weber (2012). ‘Does indivisible labor explain the difference between micro and macro elasticities? A meta-analysis of extensive margin elasticities’, NBER Macroeconomics Annual, 27, 1–56. Gruber, J. and E. Saez (2002). ‘The elasticity of taxable income: evidence and implications,’ Journal of Public Economics, 84, 1–32. Hausman, J. (1985). ‘Taxes and labor supply’, in A. Auerbach and M. Feldstein (eds), Handbook of Public Economics, Vol. I, North-Holland. Keane, M. (2011). ‘Labor supply and taxes: a survey’, Journal of Economic Literature, 49(4), 961–1075. Keane, M. and R. Rogerson (2015). ‘Reconciling micro and macro labor supply elasticities: a structural perspective’, Annual Review of Economics, 7, 89–117. Ko, W. and R. Moffitt (2022). ‘Take-up of social benefits’, in K. Zimmerman (ed.), Handbook of Labor, Human Resources and Population Economics, Springer. Manski, C. (2014). ‘Identification of income– leisure preferences and evaluation of income tax policy’, Quantitative Economics, 5(1), 145–174. Meghir, C. and D. Phillips (2009). ‘Labour supply and taxes’, in T. Besley, R. Blundell, M. Gammie and J. Poterba (eds), Dimensions of Tax Design: The Mirrlees Review, Oxford University Press. Saez, E. (2010). ‘Do taxpayers bunch at kink points?’, American Economic Journal: Economic Policy, 2(3), 180–212. Saez, E., J. Slemrod and S. Giertz (2012). ‘The elasticity of taxable income with respect to marginal tax rates: a critical review’, Journal of Economic Literature, 50(1), 3–50.

Monica Costa Dias

M

Monopsonistic labour markets Introduction

The last two decades saw a surge in interest in the idea that employers possess substantial market or monopsony power in many labour markets. This observation is hardly surprising against the background of falling labour shares, increasing wage inequality, and the erosion of organised labour, all of which suggest a widening power imbalance between employers and workers. Literally, monopsony refers to a labour market with a single employer who, as the single buyer of labour, is able to set the wage; like a monopolist in a goods market who, as the single seller of the good, is able to set the price. So it may come as little surprise that a quarter of a century ago most economists regarded monopsony as a mere curiosity, present in some highly concentrated local labour markets, notably company towns, or under employer collusion in very specialised jobs or heavily regulated occupations, such as academia, professional sports or healthcare. Yet, lately a vibrant literature has been evolving that finds monopsony to be a widespread phenomenon (see the authoritative, recent surveys by Manning, 2011, 2021).

Reasons for monopsony power

The core observation of this literature, which has its roots in Robinson’s (1933) classic account that was the first to analyse monopsonistic labour markets, is that employers possess some monopsony power if workers’ labour supply to individual employers is imperfectly elastic to wages. This holds because an employer offering a lower wage than its competitors will still be able to attract some workers if workers’ sensitivity to wages is limited.

In this case, unlike in a competitive labour market, employers have some discretion in wage setting and are expected to exploit their monopsony power to set below-competitive wages. Setting below-competitive wages, in turn, implies not only that welfare is redistributed away from workers to employers, but also a deadweight loss, because employers achieve below-competitive wages by restricting labour demand, which results in inefficiently low employment. Workers’ responsiveness to wages is likely to be limited for numerous reasons beyond employer collusion and concentration, though these latter reasons actually seem much less rare than typically thought (see Manning, 2021, and the many references given there). In consequence, employers are likely to have substantial monopsony power even in labour markets with many employers competing for workers. Reasons include search costs, which result in limited knowledge of alternative job offers; heterogeneous preferences over non-wage job attributes, such as the workplace environment or working hour arrangements, which induce workers to tolerate some wage differences across employers; and mobility and moving costs, which prevent workers from commuting or moving to employers further afield and thus restrict the range of possible employers. All these factors effectively shield employers from competition and thus provide them with some monopsony power.

Measuring monopsony power

Two key questions are: how is employers’ monopsony power to be to measured, and what evidence is there that employers really make use of it? We turn to the issue of measurement first, and then to the evidence on monopsony’s importance in wage formation. The natural measure for employers’ monopsony power is the wage elasticity of labour supply to the individual employer,

130

Monopsonistic labour markets  131

which gives the percentage change in the labour supplied to the employer from an increase in the wage by 1 per cent. Numerous studies have estimated this parameter, as documented by a recent meta-analysis of 1320 elasticity estimates from 53 studies by Sokolova and Sorensen (2021), and have found that the labour supply elasticity is far from infinity, as it would be under perfect competition.

The separations approach

The first approach to estimate the labour supply elasticity, which has been pioneered by Manning (2003: 96–104), and which we refer to as the separations approach, is by far the most common approach in the literature. It rests on a simple dynamic model in which paying a higher wage makes it easier for employers to recruit and retain workers. Let N ​​ (​ ​w)​ ​​denote the supply of labour to an employer paying wage ​w,​ let ​R​(​w)​ ​ denote its number of recruits, which depends positively on ​w,​ and let ​s(​ ​w)​ ​ denote the separation rate of incumbent workers, which depends negatively on w ​ ​. Now consider a steady state with unchanging employment, so that the employer’s inflow of workers ​R(​ w)​equals the outflow of workers s​ ​(w)​N(​ ​w)​ ​. Then, the labour supply to this employer gets: R​(w)​ ​N(​ w)​  ​=  _ ​ s​(w) ​​   

(​ 1)​

From equation (1), the wage elasticity of the labour supply to this employer ​εNw ​  ​ is the difference of the wage elasticities of recruitment​ ε​Rw   ​and the separation rate ​εsw ​  ​​: ​  ​  ​=  ​εRw ​  ​ ​−  ​εsw ​  ​  ​εNw

(​ 2)​

Equation (2) simplifies by observing that in many models of imperfect competition in the labour market, the recruitment and separation rate elasticities are the same in absolute value. Intuitively, this holds because one employer’s wage-related hire is another employer’s wage-related quit (Manning, 2003: 96–100). This is a neat result because it permits estimating the labour supply elasticity from the wage elasticity of the separation rate of incumbent workers only as: ε​ Nw ​  ​  ​=  ​−  ​2ε​sw   ​ 

(​ 3)​

Using equation (3) thus circumvents the problem of how to estimate the wage elasticity of recruitment, which would require one to observe employers’ recruitment pools or workers’ choices among all their potential employers, information that is typically absent in data. The vast majority of studies covered by Sokolova and Sorensen (2021) use some variant of the separations approach (868 out of 1320 estimates). That said, this approach hinges on the equality of the recruitment and separation rate elasticities, and so it would be important to have independent estimates of the recruitment elasticity to justify this approach.

Production function approach

Dobbelaere and Mairesse (2013) proposed an alternative approach to measuring monopsony power that is based on estimates of employers’ production function. Specifically, they show that labour and product market imperfections drive a wedge between the output elasticities of labour and intermediate inputs and their revenue shares that, in turn, is informative on both types of market imperfections. They refer to this wedge as the joint market imperfections parameter​ ψ​: ​ε​  ​

​ε​  ​

M

N

QM QN ​ψ  ​=  _ ​ ​α​  ​ ​  ​−  ​ _ ​α​  ​ ​  

(​ 4)​

where ​εQM ​  ​​ (​​εQN ​  ​) denotes the output elasticity ​ N​  ​​) of intermediate inputs (labour) and ​​αM​  ​​ (​α their revenue share, which can be obtained from production function estimates and information of employers’ input choices. Taking the market for intermediate inputs as the competitive benchmark ‒ that is, assuming price-taking behaviour in this market ‒ Dobbelaere and Mairesse demonstrate that a negative ψ ​ ​indicates monopsonistic outcomes. The intuition is that under monopsony the economic profits originating from the employer’s labour input, which result in a gap between the output elasticity of labour and its revenue share, dominate those from its intermediate inputs, and thus a negative ​ψ​signifies less competition in the labour market than in the competitive market for intermediate inputs. What is more, Dobbelaere and Mairesse show how to recover the labour supply elasticity ​​εNw ​  ​​from the estimated ​ψ​and thereby Boris Hirsch and Elke J. Jahn

132  Elgar encyclopedia of labour studies

provide an alternative way of measuring employers’ monopsony power. In line with evidence from the separations approach, studies using the production function approach find marked monopsony power for a large fraction of employers.

Concentration approach

A third method of inferring employers’ monopsony power which has been recently advocated in the literature is to measure the link between wages and employer concentration, captured by their employment shares within occupations or sectors in a local labour market. Several studies of this strand of the literature (surveyed in Manning, 2021) document a significant negative relation between employer concentration and wages and interpret this as evidence of monopsony. Though intuitively appealing, this concentration approach suffers from a fundamental issue. As is well known from the structure–conduct–performance paradigm of the industrial organisation literature, it is questionable to lend a causal interpretation to any reduced-form relationship between market shares and prices. The argument is that market shares and prices are simultaneously determined, so an effect running from employer concentration to wages is unlikely to inform on the underlying structural monopsony power parameter (unless one is willing to impose restrictive assumptions). That said, a clear advantage of the concentration approach, on top of being intuitively appealing, is that is easy to implement and allows for straightforward comparisons across countries, regions, sectors, and so on.

Evidence on the importance of monopsony in wage formation

One obvious caveat to these approaches of measuring monopsony power is that the labour supply elasticity only measures employers’ potential monopsony power, and it needs to be established that it translates into the wages employers pay to their workers. Yet, up to now there exists no direct causal evidence on this pass-through of potential monopsony power to wages. That said, there is ample indirect evidence suggesting that employers actually exercise their monopsony power. As a case in point, several studies (surveyed in Manning, Boris Hirsch and Elke J. Jahn

2021) have documented that wages are lower in more concentrated labour markets. Moreover, numerous studies have provided indirect evidence for so-called monopsonistic discrimination, or third-degree factor price discrimination, where groups of workers who are less responsive to wages receive lower wages. This latter indirect evidence is indeed quite suggestive, because in many cases wage differences have been found to be of the order of magnitude predicted by theory. Investigated groups of workers include men versus women (see Hirsch, 2016, for a detailed discussion and a survey) and immigrants versus natives (Hirsch and Jahn, 2015). Along similar lines, some studies have documented that employers’ monopsony power moves counter-cyclically, thereby predicting part of the pro-cyclical behaviour of wages (see Sorensen, 2017, for a survey). Furthermore, Hirsch et al. (2022) find larger labour supply elasticities in denser local labour markets, and show that greater competition in denser markets explains a substantial part of the higher wages in urban labour markets.

Future research and conclusions

While the existing evidence is indeed highly suggestive of some pass-through of employers’ monopsony power to wages, we still lack decisive evidence that differences in labour supply elasticities cause differences in outcomes. Besides, there are good reasons to believe that employers operating in low-wage labour markets are particularly powerful, and low-wage workers are thus particularly vulnerable to monopsony. It would thus be particularly insightful to check whether the pass-through is more pronounced in the lower part of the wage distribution. What is more, extant studies remain largely silent on specific mechanisms behind monopsony. As stressed before, three different approaches of measuring monopsony power have been extensively used in the literature, but no attempt has been made yet to compare the findings of the separations, production function and concentration approaches in any detail. Providing such a comparison may be particularly fruitful. It would not only allow to check whether the approaches arrive at similar or divergent conclusions, which provides a neat opportunity of cross-validation, but also to learn about their strengths and

Monopsonistic labour markets  133

weaknesses and to gain insights into the mechanisms behind monopsony. Learning about mechanisms is also key to design adequate policies to fight employers’ monopsony power. As a case in point, if employer concentration is important, standard antitrust measures that are pursued to limit producers’ monopoly power on goods markets may also be effective to fight employers’ monopsony power on labour markets. Among the constraints faced by employers when exercising their monopsony power, organised labour in terms of collective bargaining or worker co-determination comes quickly to mind. Clearly, the suspicion is that the presence of organised labour mutes the pass-through of monopsony power to wages, thereby limiting employers’ wage-setting power. Whether this is really happening is a highly relevant question, because strengthening organised labour, which has been eroding in many countries, may be one element in fighting monopsony. Closing these research gaps will not only be crucial to the question of whether there exists a power imbalance between employers and workers, but will also offer a most welcome opportunity of assessing monopsony’s importance in actual wage formation. We expect the findings to be of significant interest to scientists and policymakers alike, in that they not only provide new insights into the wage formation process, but also inform important recent public policy debates surrounding rising wage inequality and eroding organised labour. Boris Hirsch and Elke J. Jahn

Acknowledgements

We would like to thank Michael Oberfichtner for helpful comments and suggestions.

References

Dobbelaere, S. and J. Mairesse (2013), ‘Panel data estimates of the production function and product and labor market imperfections’, Journal of Applied Econometrics, 28(1), 1–46. Hirsch, B. (2016), ‘Gender wage discrimination: Does the extent of competition in labor markets explain why female workers are paid less than men?’, IZA World of Labor, 310. Hirsch, B. and E.J. Jahn (2015), ‘Is there monopsonistic discrimination against immigrants?’, Industrial and Labor Relations Review, 68(3), 501–528. Hirsch, B., E.J. Jahn, A. Manning and M. Oberfichtner (2022), ‘The urban wage premium in imperfect labor markets’, Journal of Human Resources, 57(S), S111‒S136. Manning, A. (2003), Monopsony in Motion: Imperfect Competition in Labor Markets, Princeton, NJ: Princeton University Press. Manning, A. (2011), ‘Imperfect competition in the labor market’, in O.C. Ashenfelter and D. Card (eds), Handbook of Labor Economics, Vol. 4B, pp. 973–1041, Amsterdam: Elsevier. Manning, A. (2021), ‘Monopsony in labor markets: A review’, Industrial and Labor Relations Review, 74(1), 3–26. Robinson, J. (1933), The Economics of Imperfect Competition, London: Macmillan. Sokolova, A. and T. Sorensen (2021), ‘Monopsony in labor markets: A meta-analysis’, Industrial and Labor Relations Review, 74(1), 27–55. Sorensen, T. (2017), ‘Do firms’ wage-setting powers increase during recessions? Monopsony models question the classic view of wage-setting and reveal a new reason why wages may decrease during recessions’, IZA World of Labor, 355.

Boris Hirsch and Elke J. Jahn

N

Non-financial motivation in the workplace

and practical applications in the workplace, between the worker and co-workers, and between workers and their employers.

Introduction

Distributional models Models of social preferences generally fall into one of several categories. Consequential models presume that people care solely about the distribution of payoffs. The first approach is simple altruism, whereby one puts an unvarying weight on the payoffs of another. This weight may lead to different actions, depending on the price-effectiveness of sacrifice. Other models involve reducing differences in material payoffs, helping the poorer individuals, or increasing the total payoff for the group. The primary consequential social preference models are Bolton (1991), Fehr and Schmidt (1999), and Bolton and Ockenfels (2000). In Bolton (1991), people care about their own money, but don’t like to have less than others. In the latter two models, receiving more than others may also bother an individual. People trade off money to reduce differences in material payoffs. The Fehr and Schmidt (1999) model has functional form:

This entry discusses non-financial motivations that have been found to affect worker responses in experimental labor environments. We first present various theories of social preference motivations, including distributional, competitive, and reciprocal models. We then discuss laboratory experimental evidence that credibly isolates these social motivators and explain when each motivator may play a role in the workplace. Finally, we consider the role of other important motivators in the workplace: respect, symbolic rewards, and identity.

Models of social preferences

Many economic environments feature a high degree of small-scale interpersonal interaction; this is where social preferences are most powerful. Individual workers function in a rich social environment in the workplace, with concomitant norms, organizational procedures, and interactions with co-workers. Issues of fairness, reciprocity, and identity may play major roles. Since the workplace is not a static environment, one’s actions may reflect instrumental considerations. But laboratory experiments can isolate any non-instrumental influences. Social preferences have received considerable attention in economics and the broader social science literature in recent years. A still-growing literature has demonstrated their prevalence, their social and economic implications, and the conditions under which they manifest. People deliberately sacrifice to help or to hurt other people, to establish equity or equality, or to increase the economic surplus available for the group. This has important consequences for both theory

1 ​ n − 1   ​  ​∑ j≠i​  max​[​xj​ ​ − ​xi​ ​, 0]​​ ​Ui​ ​(x)​  =  ​xi​ ​ − ​αi​ ​ _ _ 1 ​− β ​ i​ ​ ​ n − 1  ​  ∑ ​  j≠i​ max​[​ ​xi​ ​ − ​xj​ ​, 0​]​​.

where β ​ i​ ​  ≤  ​αi​ ​​and ​ ​ ​0  ≤  β​i ​