5. -~~ -( m E ; 9 ; L~ R;; FF)N 9 nB wio,n-,-- s.0 aAl 9~~~~9 BX a tg i4nngJ A !EconomzdEfftc%sj;ngj Global B4vtronm OZ A F I FR I ZT N A T °S ,4. - ' 24'730 /e.fis : 2002 ,,~~~~~~~~~~~~~~2 i. - .~~~~~~~~~~~~~~~~~~~~~~~l DIRECTIONS IN DEVELOPMENT Unions and Collective Bargaining Economic Effects in a Global Environment Toke Aidt Zafiris Tzannatos THE WORLD BANK Washington, D.C. i 2002 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW Washington, DC 20433 All rights reserved. 123405040302 The findings, interpretations, and conclusions expressed here are those of the author(s) and do not necessarily reflect the views of the Board of Executive Directors of the World Bank or the governments they represent. The World Bank cannot guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply on the part of the World Bank any judgment of the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is copyrighted. No part of this work may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photo- copying, recording, or inclusion in any information storage and retrieval system, without the prior written permission of the World Bank. The World Bank encourages dissemination of its work and will normally grant perrmission promptly. For permission to photocopy or reprint, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, www.copyright.com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. ISBN 0-8213-5080-3 Library of Congress Cataloging-in-Publication Data has been applied for. Contents List of Tables, Figures, and Boxes ................................. .................v Foreword ................................................. vii Mamphela Ramphele Foreword ................................................. ix Robert Holzmann Preface and Acknowledgments ................................................. xi About the Authors ................................................. xiii List of Acronyms ................................................. xv 1. Introduction and Summary ..................................................1 The Findings of the Book ..................................................4 2. Economic Effects of Labor Standards ............................................. 17 Labor Standards and Economic Performance ............................ 17 Labor Standards and International Trade ........................ ........... 20 3. Collective Bargaining and Economic Performance-A Short Review of the Theory ................. ................................ 23 What Unions Do .............. ................................... 23 What Employers' Organizations Do ............................................ 27 Organization of Collective Bargaining ........................................ 28 Dispute Resolution ................ ................................. 36 Unions: Net Costs or Net Benefits? .............................................. 37 iii IV UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT 4. Empirical Evidence from Microeconomic Studies ....................... 39 The Wage Markup in Different Countries ................................... 39 Variations in the Wage Markup .................................................. 49 Other Union Effects ................ .................................. 60 Conclusion .................................................. 75 5. Empirical Evidence from Macroeconomic Studies ............ .......... 79 Conceptual Issues .............. .................................... 80 Measuring Labor Market Institutions and Economic Performance .................................................. 80 Estimation Methodology .......................... ........................ 92 Union Density and Bargaining Coverage .................... ............... 94 Bargaining Coordination and Comparative Economic Performance: The Big Picture .................................................. 98 Dissecting Bargaining Coordination .......................................... 107 Bargaining Coordination and the Flexibility of the Labor Market .................................................. 110 The Interaction between Bargaining Coordination, Union Density, and Coverage ..................................... ............. 111 Strikes, Social Partnership, and International Trade ......... ...... 115 Conclusion .................................................. 120 Appendix 1: Definition and Description of the Indicators of Bargaining Coordination .................................................. 121 Appendix 2: Substudies .................................................. 133 Overview .................................................. 133 Selected Glossary .................................................. 143 References .................................................. 147 Index .................................................. 163 Tables, Figures, and Boxes Tables 1-1 Labor Standards in Selected Countries, 1970-94 ......................... 3 2-1 Economic Performance before and after an Improvement in Labor Standards ........................................... 18 2-2 Relationship between Core Labor Standards and International Trade: Evidence from OECD and Non-OECD Countries ........................................... 21 3-1 The Economic Costs and Benefits of the Centralization of Collective Bargaining ........................................... 30 3-2 Six Important Externalities Associated with Decentralized Wage-Setting .......................................... 32 4-1 Union/Nonunion Wage Markup for Selected Countries ......... 42 4-2 Union/Nonunion Wage Markup and Characteristics of Collective Bargaining Systems for 15 Countries ............... 45 4-3 Summary of Union Effects on Nonwage Dimensions of Economic Performance ......................................... 61 5-1 Definitions of Union Density and Bargaining Coverage .......... 81 5-2 Union Density and Bargaining Coverage in Selected OECD Countries ......................................... 82 5-3 Union Density in Selected Developing Countries and Newly Industrialized Countries in the 1980s ........................ 83 5-4 Aspects of Bargaining Coordination ......................................... 85 5-5 Characterization of 28 Indicators of Bargaining Coordination ......................................... 86 5-6 Country Rankings Based on Alternative Valuations of Bargaining Coordination ......................................... 89 5-7 Definitions of Macroeconomic Performance Indicators ........... 91 5-8 Performance Indexes ......................................... 91 v VI UNIONS AND COLLECTIVE BARGAINING: EcONONIc EFFECTs IN A GLOBAL ENVIRONMENT 5-9 Labor Market Flexibility Indicators ............................................. 92 5-10 Union Density and Economic Performance in the OECD Countries: A Summary of Relevant Studies .............. 95 5-11 Bargaining Coverage and Economic Performance: A Summary of Relevant Studies .............................................. 99 5-12 Bargaining Coordination and Economic Outcomes: A Summary and Evaluation of Results ................................. 102 5-13 Percentage of Substudies That Find a Relationship between Bargaining Coordination and Economic Outcomes, Disaggregated according to the Estimation Approach and Data Material Used ................... 104 5-14 Percentage of Substudies Testing the Hump Hypothesis That Find a Relationship between Bargaining Coordination and Economic Outcomes, Disaggregated according to Test Procedure and Estimation Approach Used ............................................... 105 5-15 Percentage of All Substudies That Find a Relationship between Bargaining Coordination and Economic Outcomes, Disaggregated according to Formal and Informal Bargaining Coordination ................................ 108 5-16 Percentage of Substudies Testing the Hump Hypothesis That Find a Relationship between Bargaining Coordination and Economic Outcomes Disaggregated according to Formal and Informal Bargaining Coordination ............................................... 109 5-17 The Relationship between Indicators of Labor Market Flexibility and Bargaining Coordination: A Summary of the Relevant Studies ............................................... 112 5-18 Spearman Rank Correlation between Selected Indicators of Bargaining Coordination, Union Density, and Coverage of Collective Bargaining ........................................ 113 5-19 The Garrett and Lange Hypothesis of Coherence A-1 The Relationship between Bargaining Coordination and Economic Outcomes: Aggregate Evidence from 125 Substudies ....................................... 134 A-2 The Relationship between Bargaining Coordination and Economic Outcomes: Results from 125 Substudies ............ 135 Figure 5-1 Economic Performance and Bargaining .................................... 100 Box 5-1 Regression to Explain Log Unemployment Rate Percentage ............................................... 114 Foreword The form industrial relations take in different countries can be critical in balancing competitiveness with concern for fairer wages and better working conditions. Accepting that workers should have a fair share of the benefits associated with economic growth without being penalized for crises for which they are not responsible is not just another view of looking at globalization; it is the very essence of humanity. This is not a new idea it can be traced back to the Aristotelian notion that a good government is one whose policies and actions are determined by justice and common interest. While history shows that there can be periods of balanced economic and social outcomes, a variety of short-run effects of unionism may exist simultaneously. Approaching these issues with rhetoric dichotomies (such as, "Are unions good or bad?") is not constructive. This book reviews the extensive literature on trade unions not just from the conventional quan- titative lens of unionization rates (for example, how many workers are unionized) but from the broader perspective of how workers and em- ployers interact in a dynamic economy, and what role the government has in such a setting. Broadening the analysis in this way shows that the relationship between economic performance and different economic set- tings, industrial relations, and governmental policies is often blurred and certainly nonlinear. While full employment, wage adequacy, and quality of work can certainly coexist (as in the case of many OECD economies), history also suggests that these three desirable aspects sometimes emerge at the expense of others. For example, fast employment generation is often associated with wage levels and employment conditions that give rise to social concerns. At the same time, fast internationalization of pro- duction can imply declining power of governments to pursue autono- mous national policies. Does the answer to these problems rest with the vii vln UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECmS IN A GLOBAL ENVIRONMENT type and extent of coordination among the social partners? If so, to what extent can coordination be confined within national boundaries and stat- utes, and what is the role of transnational organizations and international agreements? There are no simple answers to these questions. And the current study does not attempt to provide unambiguous conclusions where they do not exist. However, it provides the much needed basis for understand- ing economic performance not just in the presence of trade unions but in the broader context of industrial relations. My own conclusion from the study is that coordination among social partners can promote better investment climates while also fostering a fairer distribution of output. Mamphela Ramphele Managing Director for Human Development World Bank Foreword Providing equitable and inclusive labor markets is one of the most im- portant components of social protection. This is one of the central con- clusions of the World Bank's recent (2001) Social Protection Sector Strategy: From Safety Net to Springboard. Since labor is often a poor person's main or only asset, equitable access to safe and well-paid employment is a key method of reducing the risk of unemployment and poverty for indi- viduals. Formalization of the labor relationship is reflected in labor stan- dards, including the freedom of association and the right to collective bargaining. Sound industrial relations between employers and employ- ees can lead to a stable economy and prevent settlements that are detri- mental to the functioning of the economy. To achieve this potential win-win outcome, developing labor standards needs to go hand in hand with building institutional capacity and trust between workers, employ- ers, and the government. The need for workers, employers, and govern- ments to find solutions that reduce poverty through both growth and distributional efficiency is becoming increasingly important in an era of rapid globalization. The World Bank is committed to approaching issues surrounding labor market standards on a pragmatic, country-by-country basis. The Bank supports the promotion of all core labor standards in many ways. It offers training to its staff on these issues, engages in regular dialogue with trade unions, fosters partnerships with other international organi- zations working on these issues, and conducts research. The Bank is working to build up its knowledge base in this area and learn from the experience of partner institutions and the international community. This book is part of this broader effort. It examines the economic effects of trade unions and collective bargaining by looking at the inter- national evidence on the micro- and macroeconomic impacts of unions ix X UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT and different bargaining institutions in a comparative cross-country con- text. The study is based on a large and detailed literature survey cover- ing more than 1,000 primary and secondary studies. It is aimed at taking stock of existing knowledge on these issues in order to pave the way for a better understanding of this subject in the future. We hope that the information presented here stimulates research and dialogue in this very important area. Robert Holzmann Director, Social Protection World Bank Preface and Acknowledgments Although the massive literature on the effects of unionism ought to iden- tify some central tendency, the results of this research are often quoted selectively. This book therefore aims to provide a systematic and syn- thetic review of the global research on the economic effects of unions and collective bargaining at the micro- and macroeconomic levels. Ear- lier drafts of the manuscript were shared extensively among peers and at meetings, including a major international conference in Geneva in November 2000. This book incorporates recent developments in the lit- erature, as well as feedback and valuable additions from many interna- tional organizations and workers'/employers' representatives as well as World Bank staff. Nevertheless, the literature on labor standards, col- lective bargaining, and trade unions is developing rapidly in many di- rections and we apologize for any omissions. We would like to thank, without implicating for any errors, Gordon Betcherman, Amit Dar, Pe- ter Fallon, Morley Gunderson, Robert Holzmann, Steen Jorgensen, Wil- liam Martin, Martin Rama, Sabine Schlemmer-Schulte, and Vania Sena for comments on early versions of the manuscript. We would also like to thank the World Bank's Office of the Publisher for its advice on im- proving this publication. The study was supported by the Social Protec- tion Department of the World Bank's Human Development Network and by Danish Trust Funds. Zafiris Tzannatos Washington, D.C. Toke Aidt Cambridge, U.K. December, 2001 xi About the Authors Toke Aidt is a member of the Faculty of Economics and Politics, Univer- sity of Cambridge, U.K. His research interests include labor market in- stitutions and political economics, and previous publications include articles in Journal of Public Economics, Public Choice, and Economics and Politics, among others. Zafiris Tzannatos is currently adviser to the World Bank's managing director for human development. He has worked widely on social pro- tection issues in the Middle East and North Africa and led the Bank's Global Child Labor Program. Prior to joining the World Bank, he held a series of senior academic posts in Europe and served as an advisor to governments and international organizations. He has published widely in the areas of applied economics, gender, education, and children. xiii Acronyms CGE computable general equilibrium EPZ export processing zone GDP gross domestic product GLS generalized least squares GNP gross national product ILO International Labour Organization NMR negative monotonic relationship NR no relationship OECD Organisation for Economic Co-operation and Development OLS ordinary least squares R&D research and development PMR positive monotonic relationship WLS weighted least squares xv 1 Introduction and Summary This book examines the economic effects of unions and collective bar- gaining. Although the study is based on a large and detailed literature survey (covering more than 1,000 primary and secondary studies), it cannot claim to be conclusive. The main contribution is to take stock of existing knowledge, thereby paving the way for more innovative future research on the link between labor standards, collective bargaining, and economic performance. The specific findings of our study are summa- rized in this chapter, which also offers some additional remarks on what can be learned from our reading of the literature. Governments around the world and international organizations such as the Organisation for Economic Co-operation and Development (OECD) and the World Bank (World Bank, 1995) have in recent years become aware that labor standards are a potentially important determi- nant of economic performance. By labor standards we mean the rules that govern working conditions and industrial relations. The precise link between labor standards and economic performance is as yet not clear and many controversies remain, but the fact is that labor standards now appear on the international agenda and are likely to stay there for a foreseeable future. One of the driving forces behind the current interest in labor stan- dards around the world is the expansion of international trade and the liberalization of financial markets-sometimes known as globalization- that has occurred during the past decades. As globalization proceeds, differences in labor standards between countries and regions arguably become more important than they used to be. This is not only because such differences might give a cost advantage in internationally traded goods to countries with low standards, but also because new technology enables labor services to be directly subcontracted to workers in 1 2 UNIONS AND COLLECTIVE BARGAINING: ECONoMIc EFFcTS IN A GLOBAL ENvIRoNMENT low-standard countries. For example, a number of data entry procedures are performed in the Caribbean for U.S.-based companies and are trans- mitted to them electronically Another example is the work carried out by skilled Indian engineers who receive initial drawings from American companies by satellite and send the final products back to the United States in the same way. Thus, labor standards can no longer be the con- cern only of individual governments but must also become a concern of the international community. The need for international engagement is also highlighted by the fact that individual countries often have very different views on what constitute proper labor standards and what the consequences of adopting them might be. One view holds that labor regu- lation reduces economic efficiency and growth, and as this is more im- portant for countries with a high incidence of poverty, this view is often held by developing countries (Herzenberg 1990). Another view often found among industrial countries is that differences in labor regulation tend to discriminate against those countries that have higher standards and greater respect for workers' rights. The United States, for example, regards violations of basic workers' rights and minimum labor standards as unfair trade practices. It has adopted legislation to this effect (such as the Omnibus Trade and Competitiveness Act of 1988) that restricts trade and investment guarantees in countries that either do not enforce or vio- late labor rights and standards (Perez-Lopez 1988,1990). The International Labour Organisation (ILO) defines five core labor standards: (a) the prohibition of slavery and compulsory labor, (b) the elimination of discrimination, (c) the prohibition of exploitative child labor, (d) freedom of association (the right of workers to form unions of their own choice and of employers to form employers' organizations), and (e) the right to collective bargaining (the right of unions and em- ployers' organizations to negotiate work conditions on behalf of work- ers and employers, respectively). Among these standards, the right to collective bargaining and the right to freedom of association probably give rise to the most contro- versy, and they are the focus of this book. A recent OECD study (OECD 1996) has revealed significant differences in the extent to which these two standards are "guaranteed by law and practice" across a large sample of developing and industrial countries. The study divides countries into four groups as shown in table 1-1. The countries in group. 1 are those that permit freedom of association and collective bargaining, and in- clude almost all the OECD countries. The countries in groups 2 and 3 place some restrictions on workers' rights, while the countries in group 4 seriously suppress these rights. The classification in table 1-1 points to enormous differences around the world, not only in workers' rights but also in the organization and INTRODUCTON AND SUMMARY 3 Table 1-1. Labor Standards in Selected Countries, 1970-94 Group number Definition Countries Group 1 Freedom of association, on All OECD countries, except the whole, is guaranteed by the Republic of Korea, law and practice. Mexico, and Turkey. In addition, Bahamas, Barbados, Israel, Malta, and Suriname. Group 2 Some restrictions exist, but Argentina, Brazil, Chile, it is possible to establish Ecuador, Ethiopia, Fiji, Hong independent workers' Kong, India, Jamaica, organizations and union Mexico, Niger, Papua New confederations. Guinea, Peru, South Africa, Republica Bolivariana de Venezuela, and Zambia. Group 3 Restrictions on freedom of Algeria, Bangladesh, Bolivia, association are significant, Taiwan (China), Colombia, that is, stringent registration Ghana, Guatemala, requirements exist, and Honduras, Kenya, Mali, political interference or acts Malaysia, Morocco, Nigeria, of antiunion discrimination Pakistan, Philippines, Sri make it very difficult to Lanka, Thailand, Tunisia, form independent workers' Turkey, and Zimbabwe. organizations or union confederations. Group 4 Freedom of association is Cameroon, China, Egypt, practically nonexistent. Indonesia, Iran, Kuwait, Syria, and Tanzania. Source: OECD (1996). conduct of industrial relations more generally. A casual look at the table moreover suggests huge differences in the standard of living particu- larly between the countries in group 1 and 4, and not surprisingly OECD (1996) does find a positive correlation between GDP per capita and com- pliance with the two labor standards. Although it is clear that this can- not be ascribed to differences in workers' rights alone, it does raise an interesting and very important question: what is the link between labor standards and economic performance? The purpose of this book is to investigate this question. To this end, we ask what can be learned from existing economic literature about the economic consequences of adopting or enforcing the two labor standards. It turns out that very little systematic evidence exists on this question. 4 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFECTs IN A GLOBAL ENVIRONMENT This is partly due to the fact that it is very hard to isolate the contribution of these labor standards from other determinants of economic performance in cross-country studies and partly due the fact that it is hard to measure differences in labor standards across time and space. Some progress has, however, been made, and we examine the evidence in chapter 2. The re- mainder of the book focuses on two more specific questions: what is the impact on the economic well-being of individual workers and the perfor- mance of firms of basing industrial relations on collective bargaining be- tween unions and employers rather than relying on individual contracting (question A) and what is the impact on the macroeconomy of adopting different institutional approaches to collective bargaining (question B). It is our hope that answering questions A and B will improve our under- standing of the merits of the two underlying labor standards, and pro- vide a starting point for future research aimed directly at clarifying the link between labor standards and economic performance. Questions A and B have been thoroughly researched by economists, industrial relations scholars, and political scientists, and there exists a vast literature on the subject. In chapter 3, we briefly discuss the rel- evant theoretical literature related to unions, employers' organizations, and collective bargaining. This provides the theoretical background for what follows in chapters 4 and 5. In chapter 4, we examine what microeconometric studies of union behavior and collective bargaining at the firm and industry level can tell us about question A. In chapter 5, we change the focus and examine how different systems of collective bargaining affect macroeconomic performance (question B). The Findings of the Book The rest of this chapter is devoted to a summary of our findings. We begin with three general observations: 1. Comparative studies reveal little systematic difference in eco- nomic performance between countries that enforce the two rel- evant labor standards and countries that do not. This is partly a reflection of the difficulties of isolating the effects of labor stan- dards from other determinants of economic performance, and suggests that the impact of labor standards perhaps best can be analyzed on a case-by-case basis. 2. The microeconomic consequences of collective bargaining are context-specific, and although unions in both industrial and de- veloping countries are successful in securing a wage markup for their members and other workers covered by collective agree- ments, no general conclusions about the net costs (or benefits) of INTRODUCTION AND SUMMARY 5 unions can be reached. Depending on the economic, institutional, and political environment in which unions and employers inter- act, collective bargaining as opposed to individual contracting can contribute negatively or positively to the economic perfor- mance of firms and to the well-being of workers. 3. The macroeconomic impact of collective bargaining is hard to disentangle from other determinants of economic performance. While the available evidence from comparative studies of the OECD countries is fragile, two general features should be em- phasized. First, the impact of collective bargaining on various aspects of macroeconomic performance depends on the economic, legal, and political environment in which collective bargaining takes place and can vary over time. Second, important complementarities exist between key aspects of the bargaining system. Therefore, the impact of individual aspects such as union density or centralization of bargaining cannot be assessed in iso- lation. It is the package of institutions that matters. We elaborate on these themes in the following two sections where we attempt to summarize in more detail the specific findings related to questions A and B. Findings Regarding Question A (Microeconomic Effects) The human rights argument in support of workers' rights is compel- ling. But from an economic point of view the key questions are: What are the costs and benefits of unions? Is collective bargaining a useful institution that contributes to the achievement of desirable economic outcomes at the firm and/or sector level, or is it another labor market distortion that prevents the market from doing its job? The existence of unions arises from the asymmetry in contracting be- tween individual workers and employers, the concern for basic workers' rights, and the different perceptions about the merits of employment rela- tions governed by individual contracts or collective agreements. Textbook reasoning suggests that the alternative to a unionized labor market is one characterized by the atomistic, perfectly competitive structure that en- sures that individual workers choose whether or not to work by compar- ing the given perfectly competitive wage with the marginal utility of nonmarket activity. However, the reality facing policymakers is far less clear-cut than this suggests. First, the "removal" of unions may not reveal an underlying perfectly competitive situation in the labor market; instead, it may expose market imperfections on the labor demand side in the form of monopsony, that is, a situation in which there is only one buyer of the 6 UNIONS AND COLLEcIvE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT relevant labor services. Alternatively, firms in some industries may vol- untarily pay workers more than the going market rate to motivate ex- isting workers or to attract new ones. Hence, policy decisions whose central objective is the "return" to a perfectly competitive labor market (with all its well-known potential benefits) can succeed only if they are accompa- nied by policies designed to free up the demand side of the market. In- deed, the presence of unions in such circumstances may offer a second-best alternative to free competition. In this case, the countervailing influence of unions may result in a set of outcomes closer to the competitive equi- librium than those that would result from competition on the supply side of the labor market and monopsony on the demand side. The removal of unions may also reveal imperfections on the supply side of the labor mar- ket unrelated to unionism. For example, workers with specific skills or those protected by high turnover costs can gain "insider power," which can be used to raise wages above the competitive level. Moreover, the potential benefits (referred to as participatory benefits) associated with the presence of unions in the form of "voice" (empowerment) as opposed to "exit" (separation) effects should be counted against the costs (in the form of welfare losses due to misallocation effects). We see that theoreti- cal reasoning does not allow us to reach unambiguous conclusions about the net benefits (or cost) of unions. Whether collective bargaining contrib- utes to the achievement of desirable economic outcomes or it prevents the market from doing its job is, at the end of the day, an empirical question. To evaluate the costs and benefits of unions empirically, we need, in principle, to know how the labor market would work in their absence. The counterfactual is, of course, never observable in reality, nor can it, as argued above, be deduced from theory. Therefore, evaluations of the costs and benefits of unions must necessarily be based on a comparison of eco- nomic outcomes in those sectors of the economy that are unionized with those that are not, rather than comparing outcomes in currently union- ized sectors with the likely outcomes if those sectors had not been union- ized. In practice, this is done by estimating a union/nonunion differential or markup from individual worker or establishment data. The union/ nonunion differential is the difference between the target variable (wages, employment growth, productivity, and so on) in an average unionized firm (for a unionized worker) and an average nonunionized firm (for a nonunionized worker). Much of the empirical evidence on union/non- union markups comes from the United States and the United Kingdom, but studies from other industrial countries as well as from some develop- ing countries do exist, and are induded in our survey. We summarize the findings in 23 separate points. To aid exposition, we have grouped the findings into several related subject areas, preceded by an explanatory statement and a judgment about the robustness of the INTRODUCTION AND SUMMARY 7 particular findings. The first group of findings relates to the wage markup; these results are very robust. 1. Union members and other workers covered by collective agree- ments in industrial as well as in developing countries do, on av- erage, get a wage markup over their nonunionized (or uncovered) counterparts. 2. The markup is somewhat larger in the United States (15 percent) than in most other industrial countries (5 to 10 percent). In devel- oping and middle-income countries, the markup can be higher or lower. For example, it appears high in Ghana, Malaysia, Mexico, and South Africa but relatively low in the Republic of Korea (in 1988, before the expansion of unionism). 3. Unions compress the wage distribution. In particular, the wage differentials between skilled and unskilled workers and the pri- vate return to education are reduced when unions are present. 4. One, albeit incomplete, way to assess the adverse effects of unions is to evaluate the welfare loss that the wage markup creates through the misallocation of resources in the whole economy. In general, these effects have been found to be small and of comparable mag- nitude to the deadweight loss arising from monopolies in product markets-no more than 0.2 to 0.5 percent of gross domestic prod- uct (GDP). However, even these low estimates may overstate the allocative loss of unions because they do not take full account of the unions' potential effects on the productivity of their members (see points 21-23). On the other hand, they do not include all the potential costs of unions, such as the adverse impact that unions may have on firms' investment behavior, so the estimates may understate the allocative loss of unionism (see point 20). The size of the wage markup depends on a variety of worker and workplace characteristics. These include the following: 5. There is no significant difference between the wage markup for female workers and that for male workers in the United States and Australia. In some other countries such as Germany, Japan, Mexico, South Africa, and, perhaps, the United Kingdom, how- ever, unionized women workers have a greater pay advantage over their nonunionized counterparts than unionized men. 6. There is some evidence from Canada and Malaysia to suggest that unions contribute to a reduction in the overall gender pay gap. British studies on the subject are inconclusive. 7. In the United States and the United Kingdom, unionized non- white workers tend to get a higher wage markup than white 8 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT workers, although the U.S. evidence is mixed. In South Africa, "black" unions are associated with a smaller markup than "white" unions. In Mexico and Canada, unions have been found to re- duce the discrimination against indigenous people. 8. The wage markup tends to be higher in the private sector than in the public sector in industrial countries. The markup also depends on the economic environment in which unions and firms operate. 9. The impact of competitive conditions at the product market on the wage markup is not clear-cut and depends on how the com- petitiveness of the product market is measured. When firm- specific indicators of the competitive environment are used, unions are more successful in establishing a high wage markup if the relevant firms operate under less competitive conditions in the product market. This is not the case if industry concen- tration ratios are used as an indicator of product market com- petition. Arguably firm-specific indicators of competition are preferable to industry-wide indicators and so, on balance, prod- uct market competition seems to prevent unions from establish- ing a high wage markup. Finally, the size of the wage markup also depends on the specific aspects of how collective bargaining is organized, and from the evidence it is possible to identify particular aspects of industrial relations that add to the markup. 10. Industries with high overall union density tend to have a higher wage markup. 11. Although in some countries, such as Malaysia and the United States, industry-level collective bargaining is associated with a higher markup than firm-level bargaining, this is not so in other countries. For example, recent studies from the United Kingdom fail to find a difference. 12. Multiunionismi at the firm level (when different unions compete to recruit or organize the same workers) does not lead to a higher markup per se. However, evidence from the United Kingdom shows that the markup is high in multiunion firms that negotiate separately with each union. 13. Pre-entry closed shops (union membership is a prerequisite to obtain employment) but not post-entry closed shops (union mem- bership is required after hiring) are associated with an additional wage markup. Again, this evidence comes from the United King- dom only. INTRODUCTION AND SUMMARY 9 The union impact on aspects of economic performance other than wages is less well understood. Although a number of conclusions can be drawn with some certainty, most should be treated as tentative. The most robust results relate to hours worked, job mobility, and profitability. 14. Voluntary job turnover is lower and job tenure longer in union- ized firms. The evidence on this finding from Australia, Japan, Malaysia, the United Kingdom, and the United States seems quite robust. On the other hand, layoffs, particularly temporary lay- offs, are more frequent in unionized firms than in nonunionized ones. 15. Net company profits (price-cost ratios, Tobin's q, subjective prof- itability assessments, and the like) tend to be lower in unionized firms than in similar nonunionized firms (in Japan, the United Kingdom, and the United States). There seems to be a relatively large negative impact on profitability in firms that have product market power. 16. Hours worked is lower among unionized than nonunionized workers. This is true for both total and normal hours. In addi- tion, unionized workers are more likely to get paid for the over- time work that they do. The evidence concerning employment-related benefits, spending on research and development (R&D) and physical investment, and employ- ment growth is less robust, but the following could be noted. 17. Fringe benefits are more commonly found among unionized workers than among nonunionized ones (in Australia, Japan, Ma- laysia, the United Kingdom, and the United States). Benefits can include severance pay, paid holidays, paid sick leave, pension plans, and so on. At the same time, there is evidence that part of the wage markup is compensation for an inflexible and struc- tured work environment. 18. Employment growth can be slower in unionized than in nonunionized firms (as suggested by evidence from Canada, Ja- maica, Malaysia, the United Kingdom, and the United States), but the evidence is not particularly strong, and the observed dif- ferences most likely represent situations of disequilibrium. 19. Although spending on R&D tends to be lower in unionized than in nonunionized firms, unionized firms seem to adopt new tech- nology as fast as nonunionized ones (in Canada, Malaysia, the United Kingdom, and the United States). 20. The investment rate (physical capital) tends to be lower in union- ized than in nonunionized firms with otherwise similar 10 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT characteristics (in the United Kingdom and the United States). The adverse impact seems to be relatively larger when firms op- erate in competitive product markets, although. only one study (from the United Kingdom) has addressed this issue directly. The least robust results relate to productivity, training, and pay systems. 21. The impact of unions on productivity levels (in terms of both la- bor productivity and total factor productivity) is empirically in- determinate. Some studies suggest a positive impact, but- others imply a negative impact or no impact at all. For example, unions appear to have a negative impact on productivity levels in the United Kingdom but a positive impact in Malaysia. In the United States, there is no discernible impact, on average, but there is con- siderable variation across industries. Industries operating in com- petitive product markets and firms with "high quality" industrial relations (as measured by grievances among workers, strikes, and the like) have, on average, high productivity. 22. The relationship between unions and productivity growth is not clear either. In the United States, the union/nonunion differen- tial is- found to be negative or insignificant. In the United King- dom, some studies suggest that the weakening of British unions is one factor explaining the high productivity growth in the United Kingdom in the 1980s. 23. Unionized workers tend to receive more training than their nonunionized counterparts, especially company-related training. Overall, these findings show that the extent to which particular costs prevail or particular benefits materialize depends on the economic en- vironment in which unions and employers operate, as well as the way in which collective bargaining is organized. It is of primary interest to note that specific aspects of the economic and institutional environment, such as product market competition, absence of pre-entry closed shops and so on, can help to minimize the net costs or maximize the net ben- efits of unions. In devising union regulations, policymakers must recog- nize this fact and seek to remove the costs of unions while at the same time retaining their benefits. Findings Regarding Question B (Macroeconomic Effects) The impact of collective bargaining on macroeconomic performance can best be assessed through comparative studies where the perfor- mance of countries with (very) different bargaining systems is system- atically compared. Most studies look at the economic performance of INTRODUCrlON AND SUMMARY 1 1 the OECD countries during the period from 1960 to 1998, and ask how the framework of collective bargaining affects a large number of mac- roeconomic performance indicators (such as unemployment and in- flation) and labor market flexibility indicators (such as real wage flexibility) in an environment in which workers' rights can be taken as granted. The importance of collective bargaining as opposed to other ways of organizing contracting in the labor market can be measured by union density (the proportion of workers who are union members) and bargaining coverage (the proportion of the work force that is cov- ered by a collective agreement). With respect to these indicators of col- lective bargaining, we find: 1. Union density per se has a very weak association, or perhaps no association, with economic performance indicators such as the unemployment rate, inflation, the employment rate, real compen- sation growth, labor supply, adjustment speed to wage shocks, real wage flexibility, and labor and total factor productivity. There is, however, one significant exception: union density correlates negatively with labor earnings inequality and wage dispersion. 2. Bargaining coverage tends to be associated with higher real wage growth (with no impact on productivity growth), lower employ- ment rates, higher unemployment rates, and higher inflation. As with union density, bargaining coverage correlates negatively with labor earnings inequality and wage dispersion. Collective bargaining is potentially a powerful means to facilitate bargaining coordination; that is, the extent of coordination between unions and employers' organizations in wage setting and other aspects of in- dustrial relations (for example, working conditions, holidays and leave provisions and so on). Six different aspects of bargaining coordination can be identified: union centralization, union concentration, employer centralization, level of collective bargaining, informal coordination, and corporatism. Bargaining coordination is increasingly seen as an influen- tial determinant of labor market and macroeconomic performance. For example, the Japanese system of wage setting is decentralized (firm- based) but coordinated in the sense that it follows company rules based on seniority (hence, they are transparent) rather than individual con- tracting. In this system, workers are not paid wages equal to their indi- vidual reservation wage (that is, the wage level below which the worker will not supply his or her labor), as would have been the case under individual contracting, but this difference does not adversely affect effi- ciency. The Netherlands and Germany also have coordinated systems through strong employer organizations, coordination among giant com- panies or across industries, and coordination among unions. In France 12 UNIONS AND COLLECnVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT the government provides coordination in the form of public services, utilities, and large nationalized industries. In Italy, there is informal employer coordination (via the big firms and regional employers' asso- ciations) and between some union confederations. Finally, Sweden has a centralized employers' organization as well as centralized union con- federations. The comparative literature focuses on two hypothesizes about the relationship between bargaining coordination and economic performance: Hypothesis 1. Coordinated collective bargaining leads to better economic outcomes compared to semicoordinated collective bar- gaining, which, in turn, performs better than uncoordinated col- lective bargaining. Hypothesis 2. (The hump hypothesis) Semicoordinated collec- tive bargaining leads to worse economic outcomes than both co- ordinated and uncoordinated collective bargaining. The evidence suggests that bargaining coordination did have a ben- eficial impact on macroeconomic performance in the 1970s and 1980s, but the evidence is fragile and in the 1990s the impact seemed to disap- pear for most indicators. More specifically, we find: 3. Countries with highly coordinated collective bargaining tend to be associated with lower and less persistent unemployment, less earnings inequality and wage dispersion, and fewer and shorter strikes compared to countries with semicoordinated (for example, industry-level bargaining) or uncoordinated (for example, firm- level bargaining or individual contracting) collective bargaining. In terms of productivity growth and real wage flexibility, coun- tries with highly coordinated collective bargaining tend to per- form slightly better than countries with semicoordinated collective bargaining but may not perform differently than countries with uncoordinated collective bargaining. This lends some support to hypothesis 1, but only for the 1970s and 1980s. For most economic indicators, the differences disappear in the 1990s. Two exceptions are earnings inequality and wage dispersion. These indicators are comparatively low in countries with highly coordinated col- lective bargaining throughout the whole period. 4. Although countries with either uncoordinated or coordinated collective bargaining tend to be associated with lower and less persistent unemployment and higher productivity growth than semicoordinated collective bargaining during the period 1960 to 1990, the evidence in favor of the hump hypothesis is, in general, very weak, particularly for the 1990s. INTRODUCTION AND SUMMARY 13 5. In terms of inflation and the employment rate, there seems to be little difference between coordinated, semicoordinated, and un- coordinated collective bargaining. These conclusions refer to one dimension of industrial relations and take other dimensions as given (either by controlling for them or by inappropriately ignoring them). This ignores the possibility of complementarities between union density/bargaining coverage and bargaining coordination. Such complementarities are important for the impact of collective bargaining on economic performance, and it can therefore be misleading to focus on one particular aspect in isolation. In particular, the following fact should be emphasized: 6. High union density and bargaining coverage do not contribute to poor unemployment performance so long as they are comple- mented by high bargaining coordination (particularly among employers). Bargaining coordination is related to a number of different aspects of industrial relations, such as the centralization of collective bargain- ing, corporatism, informal coordination between employers or unions, and so on. As far as different types of coordination are concerned, the following points can be emphasized: 7. Informal coordination of wage bargaining (informal consultations between firms and/or unions or pattern bargaining) tends to miti- gate the potential disadvantage (in terms of relative high unem- ployment) associated with semicoordinated (such as industry-level) wage bargaining, and can arise in countries with relatively low union density and bargaining coverage. 8. Coordination among employers tends to be more important in producing low unemployment than coordination among employ- ees. This suggests that employers' organizations are more effec- tive in controlling wage drift than union confederations. 9. Countries that have competing unions (multiunionism) and many different union confederations tend to perform worse (in terms of unemployment and inflation) than other countries. 10. The effects of coordination can be compromised or accentuated depending on the political orientation of the government. "Good" economic outcomes (in terms of economic growth) can arise ei- ther when strong, centralized unions are paired with a strong left- wing government or when weak, decentralized unions are paired with a right-wing government. A mismatch (weak unions paired with a strong left-wing government or strong unions paired with a right-wing government) can lead to poor economic outcomes. 14 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT This concludes the long list of specific macroeconomic findings. Although some patterns emerge, we feel that the evidence is too weak and fragile to warrant grand generalizations about the performance of specific labor market institutions. Instead, we want to stress that the relationship between collective bargaining and economic performance cannot be fully understood unless the general economic and political environment in which bargaining takes place is taken into account. One should therefore be careful not to infer that institutional forms that work well in one environment would also work well in other- often very different-environments. With this caveat in mind, the syn- thesis of the literature embodied in our list of findings can provide a useful starting point for more specific studies of labor market reform. Labor Standards in a Global Environment Assessing the economic effects of unions and collective bargaining is as important as it is difficult. A compelling argument is that workers should have a fair share of the benefits associated with economic growth, and when output falls, they should not be penalized for crises for which they are not responsible. The best way for governments and the interna- tional community to protect workers' interests and their families' wel- fare may be to promote economic efficiency and mechanisms that ensure a fair distribution of efficiency gains. The involvement of social partners may be a prerequisite for designing and implementing policies that re- flect the preferences of society at large. However, systems of coordination are neither easily replicable nor necessarily a panacea. The degree and kind of coordination at the la- bor market achieved in each case are country specific in terms of eco- nomic conditions and institutional and cultural characteristics. In most countries where coordination exists, it evolved gradually through piece- meal legislation over decades rather than as a massive policy inter- vention at a specific point in time. Although some policies may have created insiders and outsiders in the labor market, policies usually blend social concerns with the economic realities of the time. Of course, labor regulation introduced at a time when particular circumstances prevailed should be reconsidered when economic conditions change. Most of the countries with coordinated systems, especially in Europe, are in the process of changing, partly because of increasing exposure to external competition and partly because of the decline in manufac- turing, where collective bargaining is more common than in white- collar sectors. By extension, assessing the economic impact of core labor standards that relate to unions and collective bargaining is important for the INTRODUCTION AND SUMMARY 15 international community, which is concerned with the effects of labor regu- lations on international trade. However, it is also important to know the economic effects of labor standards on individual countries. If freedom of association and the right to collective bargaining can be shown to have positive economic effects for the countries concerned, this will dissipate some of the heat in the "North-South" debate around the notion that when labor rights and labor standards differ between countries, such differences can give an "unfair" cost advantage in internationally traded goods to those countries that have lower standards. Although some of the studies discussed in chapters 4 and 5 came from developing countries, most are from industrial countries. This raises the question of whether our conclusions are relevant to developing coun- tries. One of the key findings of our survey is that the impact of unions and collective bargaining at both the microeconomic and the macroeco- nomic levels is context specific. The economic, legal, and political envi- ronment differs in many respects between the average industrial country and a typical developing or middle-income country. Most industrial countries have stable, liberal democracies and respect the two relevant labor standards in law and practice. This is not the case in many devel- oping and middle-income countries. Nelson (1991) has pointed out that the type of political regime-ranging from democracy to dictatorship- significantly affects the way in which industrial relations develop. The same is true for the economic environment. The economic impact of unions in an environment of ill-designed labor and product markets in which rent seeking is profitable is very different from the economic im- pact of unions in a well-designed environment. To illustrate, in many developing countries, unions with close ties to the government have played an important role in sustaining import substitution policies. Krueger (1993: 86-87), for instance, writes: Because domestic private sector industry was protected by import prohibitions and licensing, most firms had considerable monopoly power. Labor unions, whose bargaining power had been strengthened by be- nevolent social guardian governments, were able to negotiate with pri- vate sector firms whose incentive to resist wage increases, given their monopoly position, was relatively weak. Although employment in the private sector industry grew very slowly, ..., those fortunate enough to have employment in the private sector industries became yet another group supporting economic policies [i.e. import substitution policies]. Moreover, in many developing countries, unions are concentrated in the formal sector, and in the public sector. The concentration of unions in the public sector makes them a powerful pressure group that can be a significant obstade to structural reforms (see, for example, Freeman 1993a). 16 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT The relationship between social partners, however, need not always be a simple one. Some political scientists have made attempts to quan- tify the effects and have looked at the interaction between the strength of labor, party control, and economic outcomes. The argument here is that labor market parties, particularly unions, expect the government to deliver certain welfare goods and policies in exchange for wage mod- eration and peace in the labor market. There can be many scenarios of such arrangements. For example, if unions are powerful and the government is left-wing, economic perfor- mance can be predicted to be "good." This is because the pursuit of welfare policies by left-wing parties is likely to lead to voluntary wage moderation. Moreover, if unions organize the majority of workers, they are less likely to engage in wasteful rent-seeking, since unionized work- ers would themselves bear most of the costs associated with these ac- tivities. Alternatively, if unions are politically weak and the government is right-wing, "good" economic performance can also be expected. This is because unions are restricted in their wage demands by competitive pressure from product markets that are left unregulated by the right- wing government. In contrast, "bad" economic performance can result when there is a mismatch between the power of the labor movement and the political orientation of the government. If, for instance, a right-wing government coexists with powerful unions, the unions are unlikely to restrict their wage demands voluntarily because the government cannot be expected to deliver any welfare goods in return. Likewise, a left-wing govern- ment coexisting with weak unions cannot count on any voluntary wage moderation because individual unions are likely to pursue their own interests (wage pressure) without taking into account the economy-wide consequences of their actions. These scenarios of the political orienta- tion of the government and the organizational power of unions ("the Garrett and Lange hypothesis of coherence") find some empirical sup- port in a sample of OECD countries. These considerations imply that one should be careful to draw policy conclusions for developing and middle-income countries directly from the OECD evidence. In particular, the discussion of bargaining coordi- nation may be largely irrelevant at the current state for many develop- ing and middle-income countries in which union density is low, unions are concentrated in the public sector, and the legal framework of indus- trial relations is only partially designed. Nevertheless, bargaining coor- dination can become increasingly an issue as industrial relations develop and unionization is extended to more sectors. 2 Economic Effects of Labor Standards The substantial cross-country variation in compliance with the two la- bor standards identified in table 1-1 begs the question: do these differ- ences have any detectable impact on economic performance? The purpose of this chapter is to examine this question from a cross-country comparative perspective. Labor Standards and Economic Performance It is a difficult task to measure cross-country variation in labor stan- dards and to disentangle their economic effects from other determinants of economic performance. One approach, which side steps the difficul- ties of obtaining comparable measures of labor standards across coun- tries, is to identify countries that have undertaken major labor market reforms in the areas of freedom of association and the right to collective bargaining, and then to compare the performance of the economy before and after the reform. Using this approach, OECD (1996) has identified 17 countries that have undertaken significant labor market reforms over the past 20 years and has compared the average growth rate of GDP, manufacturing output, and exports in the five-year period before and the five-year period after the reforms. Table 2-1 summarizes the results of this exercise. The evidence shows that on average, GDP grew at 3.8 percent per year before the improvement in labor standards and at 4.3 percent after- wards. Growth in manufacturing output increased by a smaller amount (from 3.4 to 3.6 percent). In contrast, export growth declined by 2.3 per- centage points on average (from 6.6 to 4.3 percent). These averages mask considerable variation and can therefore be misleading. First, the aver- ages hide huge differences in individual country experiences. For example, growth rates in Panama increased by as much as 8 to 10 percentage points 17 18 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFncIs IN A GLOBAL ENVmRoNMENT Table 2-1. Economic Performance before and after an Improvement in Labor Standards (average growth rates, percent) Manufacturing Reform GDP growth output growth Export growth Country year Before After Before After Before After Argentina 1983 -0.2 1.0 -0.5 0 0.6 2.8 Brazil 1988 5.3 0.9 4.5 -2.2 9.5 4.8 Dominican Republic 1990 4.4 4.5 1.7 4.2 9.1 5.6 Ecuador 1979 7.1 1.3 11.6 2.1 0.4 2.3 Fiji 1987 9.8 5.8 4.2 -0.6 14.3 6.7 Guatemala 1992 4.1 4.1 - - 5.6 8.6 Honduras 1990 3.0 3.3 4.0 3.8 1.9 1.8 Korea, Republic of 1987 10.7 8.6 15.7 8.3 15.6 6.9 Panama 1989 -0.5 10.5 -2.5 8.9 0.2 8.9 Peru 1990 -0.9 1.8 - - -3.8 -23.2 Philippines 1987 -1.3 4.0 -2.4 3.1 2.4 7.2 Suriname 1991 1.7 0.6 -3.2 -2.4 - - Taiwan (China) 1987 9.6 6.9 - - - - Thailand 1992 10.7 8.2 14.7 11.5 17.3 13.2 Turkey 1986. 6.1 2.7 7.9 5.7 16.1 8.1 Uruguay 1985 -7.6 4.4 -5.4 3.7 2.7 6.8 Venezuela, Rep. Bolivariana de 1990 2.7 5.2 -3.3 4.5 6.8 3.6 Average 3.81 4.34 3.36 3.61 6.58 4.27 - Not available. Source: OECD (1996, table 7). after the reform, whereas export growth in Peru collapsed. Second, in most countries, economic performance actually deteriorated after the reform. These results are crude, as growth performance depends on many other factors such as initial level of GDP, investment levels, political institutions and, in the shorter term, the growth rates in other countries. Hence, to isolate the (potential) impact of a reform, these factors have to be taken into account. Palley (1999) analyzes the relationship between GDP growth and improvements in labor standards in a multiple regression model. He estimates a pooled time-series/cross-country model using data for 15 of the countries included in table 2-1 five years before and five years after an improvement in labor standards took place.' He includes the following 1. Taiwan (China) and Turkey are excluded. ECONOMIC EFFECTS OF LABOR STANDARDS 19 control variables: lagged GDP growth, the average growth rate in the rel- evant region, the average growth in industrial countries, a linear time trend, a set of country dummy variables, and a dummy variable to cap- ture the effect of the reform. He finds that improvements in labor stan- dards have a positive and statistically significant effect on economic growth, with significance levels ranging from 0.01 to 0.12 depending on the precise specification of the estimated equation. It is unclear, however, what can be learned about the impact of labor standards on economic growth from table 2-1 and these regressions. First, improvements in labor standards are typically an integral part of more comprehensive reformns that change the entire institutional structure of an economy. To the extent that this is the case, one cannot attribute the change (positive or negative) in economic performance to improvements in labor standards alone: the "reform dummy variable" used in the re- gressions picks up all changes in the political and economic environment. Second, while it is possible that changes in institutions and the legal struc- tures can have an impact on economic performance in the short term, it normally takes a long time for such changes to spread throughout the economic system. Therefore, the impact is typically not observable until long after the reform has been initiated: the period under consideration is too short to allow us to make robust generalizations. Freeman (1993a) takes a different approach. He provides a detailed investigation of the relationship between economic growth (and other economic performance indicators) and observance of core labor standards in six East Asian "success" countries.2 Freeman (1993a: 45) concludes that the observance of core labor standards is neither an obstacle nor a neces- sity for economic growth and that "what is clear is that these countries as a group are not examples of developing countries that succeeded by avoid- ing labor interventions." This conclusion is supported by World Bank (1993), which argues that the suppression of unions in the Republic of Korea (prior to 1988) and wage repression in Singapore cannot be con- sidered significant factors in the successful growth of these economies during the 1980s when compared to China, Hong Kong (China), and Tai- wan (China). In addition, Fields (1994) argues that Hong Kong (China), Singapore, the Republic of Korea, and Taiwan (China) have performed equally well during the considerable transformation of their economies over time despite significant differences in the industrial composition of their out- put and employment and despite huge differences in their labor mar- kets. Thus, neither growth nor adjustment seems to relate singularly to 2. China, Hong Kong (China), Malaysia, Singapore, the Republic of Korea, and Taiwan (China). 20 UNIONS AND COLLECTIVE BARGAINING: ECoNoMIC EFFECTS IN A GLOBAL ENVIRONMENT specific labor market characteristics or policies (Tzannatos 1996). In fact, the relationship between specific policy variables and economic growth is generally weak (Levine and Renelt 1992), and a comparative study of 31 industrial and developing countries estimates that labor market dis- tortions cannot account for more than 10 percent of the cross-country variation in economic growth (Agarwala 1983). Labor Standards and International Trade The discussion in the previous chapter suggests that labor standards and international trade have figured prominently in the policy debate. On the one hand, low labor standards might be interpreted as social dumping and unfair trade practices. On the other hand, trade sanctions imposed with the aim of forcing low-standard countries to impose higher standards might be interpreted as protectionism. According to either of these arguments, compliance as well as non-compliance with core labor standards distorts international trade-compliance by favoring coun- tries with high standards and noncompliance by favoring countries with low standards. A crucial question then is if there is any evidence of this actually happening. The OECD (1996) examines the interaction between freedom of association and the right to bargain collectively and various indicators of international trade using the country classification from table 1-1. Table 2-2 summarizes some of their conclusions. Based on the evidence presented in table 2-2, there does not seem to be a systematic relationship between freedom of association and the right to collective bargaining and international trade, as measured by export market shares, revealed comparative advantage, foreign direct invest- ments, and trade prices. However, the two labor standards appear to be associated positively with the success of trade reforms, although, as ar- gued earlier, it is difficult to distinguish labor reforms from more gen- eral reforms in the political and economic spheres. Also, it is unclear if successful reform creates greater respect for workers' rights or if respect for workers' rights makes trade reforms successful. Finally, while ex- amples of countries that have systematically suppressed labor rights in export processing zones (EPZs) do exist, the OECD concludes that se- lective suppression of workers' rights is not a general tendency (see also the review of EPZs in Kusago and Tzannatos 1998). Elliott and Freeman (2001) argue that EPZs can become a testing ground for demonstrating that trade and labor standards can reinforce one another in raising stan- dards of living in poor countries. Overall, it is clear that this important issue requires more research before it can be settled in a satisfactory way. Nevertheless, the contro- versy surrounding it seems unmatched by evidence that labor standards actually distort trade patterns in any significant way. ECONOMTc EFFECrs OF LABOR STANDARDS 21 Table 2-2. Relationship between Core Labor Standards and International Trade: Evidence from OECD and Non-OECD Countries Trade indicator Definition Resuilt Trade Growth in the share of a county's No correlation. performance exports m total world trade (measured as total export, raw material export, or manufacturing export), 1980-90. Revealed Index for comparative advantage No effect on the pattern of comparative calculated for 71 sectors on the revealed comparative advantage. advantage basis of foreign trade performance Comparative advantage is by for these sectors. and large determined by the abundance of factors of production and technology. Trade prices U.S. import prices of textiles from No effect on border prices in the a group of other OECD and United States for similar non-OECD countries. imported textiles from different countries. Likewise, U.S. import of textiles from "high-standard" OECD countries has not been "crowded out" by imports from "low-standard" non-OECD countries. Trade The change in tariffs and Positive correlation with the liberalization quantitative trade restrictions, trade liberalization. 1980-90. Foreign direct OECD investment outflow to Low labor standards not an investment non-OECD countries, 1975-93. important factor for investment (FDI) decisions in OECD firms. Export- Firms in an area that offers Existence of evidence In only 6 of processing privileges with regard to the 73 countries that have zones (EPZs) government policies. established EPZs of deliberate government attempts to restrict freedom to associate and the right to bargain collectively. The countries are Bangladesh, Jamaica, Pakistan, Panama, Sri Lanka, and Turkey. Note: Labor standards refer to the freedom of association and the right to bargain collectively. Source: OECD (1996 part 11). 3 Collective Bargaining and Economic Performance-A Short Review of the Theory Unions arise from the asymmetry in contracting between individual workers and employers, the concern for basic workers' rights, and dif- ferent perceptions about the merits of employment relations governed by individual contracts and collective agreements. The desirability of unions depends on many factors, including (a) what unions do, (b) how collective bargaining is organized, and (c) the effectiveness of dispute resolution mechanisms. Freedom of association refers not only to the workers' right to form unions of their liking, but also to the right of employers to form employers' organizations. As with unions, the desir- ability of employers' organizations depends on what they do and the context in which they do it. What Unions Do Unions are engaged in many different activities. We make a distinction among three aspects of union behavior: monopoly, participatory, and rent-seeking behavior. The Monopoly Cost of Unions Traditionally, economists have focused on the social costs of unions, which arise when they secure favorable pay and work conditions for their members by sharing supernormal profits with firms (Booth 1995).' 1. In a competitive market, the union is able to share in the quasi-rent to fixed capital in the short run, but in the long run, the unionized firm is forced to close down, and capital moves to a nonunionized sector where the return is higher. 23 24 UNIONS AND COLLECTIVE BARGAINING: ECONoMIc EFFECTS IN A GLOBAL ENVIRoNMENT Supernormal profits are typically associated with product market dis- tortions and/or government regulation; thus, labor market and product market distortions are often viewed as complements. Unions can force firms to relinquish some of their profits only if they can monopolize labor supply.2 This is because unions wield the strike threat: firms are willing to give up some of their profits to avoid industrial conflict. Com- petition from a large nonunionized labor market reduces the union's monopoly command over labor supply; if nonunion workers can readily replace union workers, the union's bargaining position is substantially weakened (Ulph and Ulph 1990).3 According to this view, when unions succeed, they impose a number of costs on society, which we call the monopoly costs of unions. These costs are as follows: * Firms will try to pass on the wage demands to consumers as higher prices. This increases the consumer price index and re- duces the real (consumption) wage of all workers. It also increases the real price of intermediate inputs harming other producers. These effects are comparatively small if firms operate in a highly competitive (product) market environment. * The wage markup increases the relative price of labor in the union sector. This induces a reallocation of labor to the nonunion sector as firms decide to lay off unionized workers (Rees 1963). This tends to reduce the nonunion wage and the welfare of nonunion members and leads to an output loss because workers are now being employed where their marginal productivity is lower than before (see Sapsford and Tzannatos 1993: 325-28). These effects are mitigated when unions and firms bargain over wages and employment (McDonald and Solow 1981)! as employment in- creases rather than decreases in the unionized sector, reducing the negative spillover on nonunionized sectors. 2. Some rents are capitalized in the value of the firm and so are not available for sharing. This effect can be illustrated as follows. Assume a monopoly situation is estab- lished as a result of an innovation. If the prospect for high profits is real, the inventor is likely to sell the right and make a large capital gain instantaneously. Thereafter, sales grow and the firm reverts to a public company. The monopoly power of the company is now reflected in the value of its shares, not in the rate of operating profit. It is the rate of return to the shares (in the form of dividends and capital gains) that is relevant for collective bargaining and this is determined competitively in the stock market. Hence, the firm's ability to provide high wages to its labor force has disappeared (Sapsford and Tzannatos 1993). 3. The extent to which nonunion workers are a cost-effective substitute for union workers depends, among other things, on job security rules (the cost of hiring and firing), the availability of nonunion workers with the necessary skills, and institutional rules that allow the union to run a closed shop or otherwise link employment to union membership. COLLECTlVE BARGAINING AND ECONOMIC PERFORMANCE-A SHOrT REVIEW OF THE THEORY 25 * Unionized firms share their profits with the union. This creates a hold-up problem that reduces investiients in physical capital and R&D in unionized firms below the socially optimal level (Grout 1984). * The more senior members, who typically have a disproportion- ate influence on union decisions; may institutionalize a seniority principle in relation to layoffs and other aspects of deployment, such as promotion, recall, and training. This can create insider/ outsider dynamics that can lead to persistently high levels of unemployment. The discussion of the monopoly costs of unions is often based on the (implicit) assumption that the labor market in the absence of collective bargaining would be competitive. This assumption is overly optimistic, as the "removal" of unions may reveal market imperfections on the la- bor demand side in the form of monopsony.4 Under these circumstances, the presence of unions may offer a second-best alternative to free com- petition and the countervailing influence of unions can result in out- comes closer to the competitive equilibrium than are offered by competition on the supply side of the labor market and monopsony on the demand side.5 In addition, it is by no means clear that a net welfare gain will fol- low if a unionized system of wage and employment determination is replaced by one in which contracts are negotiated between individual workers and their employers. This is because large transaction costs may be associated with numerous individual contract negotiations compared with those arising from setting a small number of "pattern bargains" by negotiation between unions and an employers' associa- tion. The outcome of these negotiations can subsequently be adopted throughout the economy much as in the centralized ("Scandinavian") labor market structure. The beneficial effect of centralized negotiations may be stronger in industrial economies where, given the inherently complex nature of the labor market, individual negotiation (at least in some sectors of the labor market) may be the only feasible alternative to a unionized system. 4. Employers derive monopsony power from the fact that it is costly for a worker to leave the firm (because of the firm-specific human capital he or she has accumulated) and move to another city to get a new similar job. 5. Many more imperfections are likely to coexist with unionism, some arising from motivational problems (efficiency wages) and others from insider power (Lindbeck and Snower 1989). 26 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT The Participatory Benefits of Unions The "organizational view" of unions (also known as the second face of unions), associated with Freeman (1980a) and Freeman and Medoff (1979, 1984) among others, focuses on the economic benefits of unions. Unions facilitate worker-participation and worker-manager cooperation in the workplace. This can have efficiency-enhancing effects that jointly ben- efit workers and management. More specifically, the economic benefits, which we refer to as participatory benefits, arise from a number of sources, such as the following: * Unions are institutions with a collective voice operating within internal labor markets. The union's role within this framework is to communicate worker preferences directly to the manage- ment, as well as to paiticipate in the establishment of work rules and seniority provisions in the internal labor market. This -changes the exit-voice tradeoff of workers by providing a chan- nel through which they can express their grievances without having 'to leave the firm. This reduces turnover (voting with the feet), increases the incentive of employers to provide firm- specific training, and facilitates long-term working relationships that benefit all parties. In addition, unions can help to establish seniority provisions, which lessen the rivalry between experi- enced and inexperienced workers, among other things. This can increase the amount of informal, on-the-job training that the former is willing to provide to the latter (Freeman and Medoff 1979, 1984). * Unions facilitate procedural arrangements and other agency ser- vices that help to reduce the likelihood of costly disputes about wage and employment conditions (Faith and Reid 1987). * Unions help to enforce contracts between workers and manage- ment (Malcornson 1983). For example, if the market demand for a product is uncertain, workers may be reluctant to acquire firm- specific skills unless the firm can promise not to fire them if de- mand turns out to be low. Without a credible enforcement mechanism, the firm cannot make such a promise and workers acquire too few firm-specific skills. However, a union can help to enforce the promise if the firm prefers to honor the implicit agree- ment rather than to become embroiled in a strike. * Unions can increase productivity by providing a channel through which labor can draw management's attention to changes in working methods or production techniques that may be benefi- cial to both parties. This channel also offers a mechanism by which COLLECnVE BARGAINING AND ECONOMIC PERFORMANCE-A SHORT REVIEW oF THE THEORY 27 the union can "shock" management into better practices (reduce X-inefficiency) .6 Unions as Rent-seekers Unions represent the special interests of their members in collective bar- gaining and in the political process. As pointed out by Pencavel (1995), unions generally promote policies that reduce competition in labor and product markets. This includes support for minimum wage legislation, trade protection, and so on. Unions support such policies if they increase the surplus available for sharing with the firm (the effect of less com- petitive product markets) or increase the union's bargaining power (less competition from nonunion labor markets).7 The political activities of unions (and other rent-seekers) generally involve three types of (static) costs. First, to the extent that the union is successful in getting government regulation, an economic distortion is created, and the resulting deadweight loss is a loss to society. Second, real resources are withdrawn from production to be used in rent-seeking. To the extent that the shadow price of these resources is positive, this constitutes a loss to society. Third, since the union's distributive success typically comes at the expense of nonunion workers and consumers, a union's political activities may be associated with large distributional costs. In addition to these static costs, rent-seeking can lead to dynamic costs. As Murphy, Schleifer, and Vishy (1993) argue, rent-seeking is par- ticularly harmful to innovations and thus hampers economic growth. One argument is that innovators lack the political base needed to obtain the necessary licenses and permits and that innovative projects are typi- cally long term and risky, which exposes them to expropriation. In short, where unions are part of a rent-seeking society, they impose a substan- tial political cost on society. What Employers' Organizations Do The members of an employers' organization are individual firms, typi- cally within a particular industry. Each employers' organization may in 6. X-inefficiency refers to a situation in which a firm's total costs are not mini- mized because the actual output from given inputs is less than the maximum feasible level. 7. With regard to the regulation of the product market, the union and the firm have a common interest, and they may form a very effective distributional coalition (Rama 1997a; Rama and Tabellini 1998). On the other hand, they disagree with respect to labor market regulations such as job security legislation and minimum wages. 28 UNIONS AND COLLECTrVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT turn be a member of a national employers' organization. A firm may decide to join an employers' organization to improve its bargaining posi- tion with workers (possibly organized in a union). Firms derive their bar- gaining power from their ability to lock out workers. The cost of an industrial conflict from the point of view of an individual firm is larger than the cost to the industry as a whole. This is because an individual firm involved in a strike is likely to lose its market share to other firms in the industry that produce close substitutes. Accordingly, whereas each firm has an incentive to give in to wage demands (to avoid a local conflict), the industry as a whole has less incentive to do so, and by joining forces, it is easier for firms to resist wage demands from unions (see Dowrick 1993). Employers' organizations can also help firms to avoid leap frogging. Leap frogging occurs when individual firms increase their wage rate to extract more effort from existing workers or to attract skilled workers from other firms (Layard, Nickell, and Jackman 1991). When all firms engage in this kind of behavior, the net result may well be that relative wages are unchanged but the level of all wages has increased substantially. A strong employers' organization that coordinates the behavior of individual firms can be helpful in internalizing this "efficiency wage externality" and pre- venting wage drift (as discussed above). In addition, employers' organi- zations play an important role in providing training (Soskice 1990). Since general training is a public good, firms are unlikely to provide much of it unless they are subject to external pressure. A strong employers' organi- zation can provide training facilities for firms and can impose sanctions if a firm does not pay its share of the cost. Organization of Collective Bargaining The economic effects of unions depend on the way in which collective bargaining is organized. Of particular interest here is the degree of bargaining coordination. This section outlines four aspects of bargaining coordination from a theoretical perspective. The aspects are centraliza- tion, union concentration, informal coordination, and corporatism.' Centralization of Collective Bargaining Collective bargaining is centralized when the national union confedera- tion and the national employers' organization can influence and control 8. The relevant theoretical literature has been surveyed in great detail by Calmfors (1993), Henley and Tsakalotos (1993), Layard, Nickell, and Jackman (1991: chapter 2), and Moene and Wallerstein (1993a). COLLECTIVE BARGAINING AND ECONOMIC PERFORMANCE-A SHORT REVIEW OF THiE THEORY 29 wage levels and patterns across the economy. The capacity to do so de- pends on many factors, including the level at which bargaining prima- rily takes place (the plant, the industry, or the national level) and whether or not the national organization(s) can control the behavior of their con- stituent organizations and avoid wage drift. Table 3-1 summarizes eight important aspects of bargaining centralization and evaluates the associ- ated (static) costs and benefits. The idea that centralization of collective bargaining can facilitate in- ternalization of externalities has received particular attention in the lit- erature and warrants a more detailed discussion than the one given in table 3-1. To fix ideas, imagine a society in which all workers are orga- nized in unions. Then suppose that each firm negotiates with a com- pany union. In this case, wage-setters bear only a small fraction of the total economic cost associated with a given increase in their real wage as they impose external costs on others. Table 3-2 defines six such exter- nalities in more detail. Because of these externalities, the negotiated wage is "too" high and the result is, other things being equal, "too" little total employment. By centralizing the bargaining process to the industry or national level, wage-setters are forced to bear a larger share of the cost of their actions, as more (and ultimately all) workers are included in the bargaining coalition. This creates incentives in favor of wage restraint, which, again other things being equal, leads to more total employment. As pointed out by Cahinfors and Driffill (1988), this argument ignores the fact that the competitive pressure from product markets and the mod- erating effect it has on wage demands change systematically with the level of centralization. For example, consider what happens when a union de- mands (and gets) a high nominal wage. To avoid an increase in the prod- uct real wage, firms pass on the cost to consumers as higher prices. From the union's point of view, not only does this reduce the consumption real wage but it has the unpleasant side effect of reducing the demand for the goods produced by the host firm, endangering the jobs of union mem- bers. Anticipating this outcome, the union moderates its wage demand. At the firm level, the competitive pressure from other firms in the same industry (producing close substitutes) provides a strong incentive to mod- erate demands. At the national level, the federation of unions bears the full cost of its actions, social partnerships become possible, and unions and employers' organizations are sufficiently encompassing to make rent- seeking unprofitable (Olson 1982; Heitger 1987). At the industry level, neither of these effects produces much wage moderation. On the con- trary, firms in an industry can pass on a substantial portion of the wage demands to consumers at a relatively low employment cost. In addition, industry-based unions often form effective lobbying groups that seek dis- tributive favors from the government at the expense of society at large. Table 3-1. The Economic Costs and Benefits of the Centralization of Collective Bargaining Issue Benefit Cost 1. Internalization of externalities: Unions and firms acting Centralization increases the size of the bargaining independently of the rest of the market coalition, thereby intemalizing negative extemalities. (decentralization) can have unintended negative The effect is larger, the more workers are unionized. effects (extemalities) on the rest of the economy (for example, higher wages can be passed on to consumers in the form of higher prices, higher inflation, and an increase in unemployment). 2. Competitive pressure: Competition in product markets As bargaining becomes more centralized, competitive disciplines unions and firms, and this effect is pressure is reduced because firms acting in unison are strongest at a decentralized level (more competition less likely to lose their market share (product demand reduces the ability to pass wage increases on to is more irnelastic at the industry level than at the firm consumers as higher prices). level). This increases wage pressure and leads to higher unemployment. This effect is less important in an open economy. 3. Wage comnpression: Under centralized collective Although wage compression can force less effident A reduCtion in wage dispersion leads to an economic bargaining, solidarity/egalitarnan wage goals are firms out of the market (to the extent that low wages misallocation of resources and lower output. easier to achieve, and firm-specific conditions are less move upward), it can encourage the entry of new and likely to enter the wage contracts. This tends to more efficient firms: the net effect can, under certain reduce wage dispersion. conditions, increase output. Again, under certain assumptions, wage compression can act as a form of social insurance. 4 Areas of bargaining: Some issues can only be subject to For example, general training of workers is more likely Efficient bargaining (over employment and wages) is collective bargaining at certain levels of centralization to be part of centralized collective bargaining because it feasible only under decentralized bargaining. or above (training, health and safety, and so on). has the characteristics of a pubhc good. Subsequently, Workplace cooperation and other participatory training can lead to higher economy-wide labor activities between unions and firms decrease under productivity and overall economic growth. centralized bargaining. (table continues onfollowing page) Table 3-1 continued Issue Bmefit Cost 5. Hold-up problems: Firms undertake investment The hold-up problem is reduced under centralized decisions today that affect future profits. If workers, bargaining because an individual firm cannot affect the via collective bargaining, can get a share of these outcome of collective bargaining by its pre-bargaming profits without contributing to the costs, firms would investment decisions. This encourages finms to invest underinvest. more. 6. Insider-induced hysteresis: Only the group of insiders Under centralized bargauiing, more workers can be (for example, union members and employed workers) perceived to be insiders (including the unemployed) to counts in wage bargaining. When the number of the extent that unions are concerned about aggregate insiders is reduced (for example, after layoffs in a unemploymenL recession), they can push for higher wages in the next bargaining round and cause unemployment to remain persistently high (insider-induced hysteresis). 7. Strikes: Imperfect information can lead to more Centralization increases the level of information about Centralization increases the risk of a general strike. strikes. demand condhtions, reducing the likelihood of strikes, especially wild-cat strikes. 8. Bargaining power: The relative bargaining power of Centralization can reduce wage pressure by increasing Centralization can increase wage pressure if unions workers and employers depends on the "fall-back" employers' bargaining power because if all firms lock derive their bargaining power from the monopoly option of the two parties (what they will get if an out workers dunng an industrial conflict, workers will command over labor supply It is easier for a single agreement is not reached). have fewer altemative job opportunities. firm than for an entire industry or nation to replace workers in the event of a strike. Source: Besides the surveys by Calmfors (1993), Henley and Tsakalotos (1993), Layard, Nickell, and Jackman (1991: chapter 2), and Moene and Wallerstein (1993a), the following references are relevant: (I) and (2) Calmfors and Driffill (1988); (3) Agell and Lommerud (1992), Harcourt (1997), Moene and Wallerstein (1993b, 1993c), and Agell (1998); (4) Soskice (1990), (5) Grout (1984); (6) Blanchard and Summers (1986); and (7) Kennan (1986). 32 UNIONS AND COLLECTIVE BARGAINING: ECONoMIc EFFEcrs IN A GLOBAL ENVIRONMENT Table 3-2. Five Important Externalities Associated with Decentralized Wage-Setting Externality Definition Input price Decentralized wage gains are passed on as higher product prices, thus increasing the real cost of inputs for other firms. Fiscal Decentralized wage gains lead to unemployment. The cost in terms of unemployment benefits is borne by all taxpayers, not just those involved in wage-setting. Unemployment Decentralized wage gains increase overall unemployment, making it more difficult for all unemployed workers to find a new job. Envy Decentralized wage gains create envy among other workers. Consumer price Decentralized wage gains are passed on as higher product prices, thus lowering the real wage of all workers. Efficiency wage At the decentralized level, firms have an incentive to try to raise the relative wage of their workers to increase their motivation. Source: Calmfors (1993: 5-6). It follows from this discussion that the relationship between economic performance and centralization of collective bargaining can be nonlinear (U- or hump-shaped): relatively good performance for decentralized and centralized systems, but relatively poor performance for systems based on industry-level bargaining (Calmfors and Driffill 1988). It should be noted, however, that this prediction is sensitive to many of the underly- ing assumptions. For example, Rama (1994) shows that the nonlinear relationship tends to disappear in an open economy as competitive pres- sure becomes more intense at all levels of bargaining. It is also clear that centralization will not help to internalize external costs unless most work- ers are union members or have their pay and work conditions deter- mined by collective agreements. More critical, perhaps, is the fact that the analysis takes a static view of the economy. Arguably, one of the key advantages of a centralized bargaining system is that it enables a coordi- nated and fast response to changing economic conditions. A further complication arises in multi-bargaining systems where wage drift can be a problem whenever the national organizations cannot per- fectly control the behavior of their constituencies. The term "wage drift" refers to wage increases negotiated at the firm level in addition to those specified in the agreement between the national organizations. The suc- cess of firm-level unions in obtaining wage drift depends on their COLLECTIVE BARGAINING AND ECONOMIC PERFORMANCE-A SHORT REVIEW O0 THE THEORY 33 bargaining power relative to that of (the local) management. Although management in general is interested in low wages, they may neverthe- less be willing to accept local wage increases for efficiency wage reasons. Accordingly, not only does the union confederation have to control the behavior of its constituency, but the national employers' organization also has to make sure that its members stay in line. This is typically done by requiring that all local agreements be approved by the national em- ployers' organization. Moreover, to reduce militancy at the firm level, strikes and lockouts may be banned after an agreement has been signed by the two national organizations (a peace clause). As noted by Calmfors (1993) and others, this is not likely to reduce wage drift significantly be- cause unions can use informal sanctions (such as work-to-the-rule and go-slow actions) to improve their bargaining position. Therefore, in prac- tice, it is difficult to prevent wage drift. To avoid its adverse impact, how- ever, national organizations can try to internalize the expected wage drift in the agreements that they reach at the national level. Union Concentration The term "union concentration" is used to describe the horizontal orga- nization of unions at a given level of centralization. Two characteristics of the horizontal organization of unions are normally stressed. These are multiunionism versus single-unionism and open versus closed unions. Multiunionism refers to a situation in which many competing unions offer to represent the same worker. An example of a country with multiunionism is Belgium where workers can choose from among a so- cialist, liberal, or Christian union or the United Kingdom where unions commonly compete at the plant level. In contrast, Denmark is character- ized by single-unionism in the sense that the social-democratic labor movement almost has a monopoly on organizing workers. A closed union restricts membership to workers of a particular craft, profession, or job. The United Kingdom and the Scandinavian countries are examples of countries where closed unions are prevalent. In contrast, an open union allows workers with different levels of education and from different trades to be members. Examples of open unions include company and industry unions (which are open to all employees of the relevant com- pany or industry), as in Japan and Germany. We say that an industrial relations system based on multiunionism and closed unions is less con- centrated than one based on single-unionism and open unions. To understand why the distinction between open and closed unions is important, imagine a firm whose work force of blue- and white-col- lar workers is organized in two closed unions. Each of these unions bargains separately with management over wages and work conditions 34 UNIONS AND COLLECTVE BARGAINING: EcoNoMIc EPPcrs IN A GLOBAL ENVIRONMENT for its particular constituency The agreement between management and the firm's blue-collar workers affects the employment opportunities of the white-collar workers and vice versa. This is because the firm ad- justs its employment mix in response to the change in the relative wage rate between the two types of workers. In other words, each of the two closed unions imposes an externality on the workers of the other union. This externality is internalized if all workers in the firm join one (open) union. The impact on wage demands depends on whether blue- and white-collar workers are substitutes or complements in production. Suppose that they are substitutes. Under a closed union structure where the two groups bargain separately, blue-collar workers are pleased if the white-collar union is successful in getting a pay raise and vice versa. This is because the pay raise for blue-collar workers increases the de- mand for white-collar workers as the firm substitutes toward the less expensive alternative. Under an open union structure, the fact that the two groups of workers bargain together ensures that the (positive) wage externality is internalized. Consequently, an open union pushes harder for wage increases than the two closed unions together. As a result, overall employment is likely to be lower. If, on the, other hand, white- and blue-collar workers are complements in production, the external- ity associated with a closed union structure is negative. Accordingly, a single open union reduces wage pressure, and overall employment is likely to be higher. To see why the distinction between single- and multiunionism is im- portant, consider the costs and benefits of multiunionism. First, when unions compete to organize the same pool of workers, they are likely to end up organizing workers who are substitutes in production. As dis- cussed above, the (positive) wage externality associated with separate bargaining tends to produce real wage moderation and high employment. On the other hand, if the competing unions need to prove their ability to raise wages to attract members, the outcome of competition between unions may be less rather than more employment. Second, workers are likely to have different preferences regarding the services provided by the union, such as the tradeoff between wage gains and employment and the nature of the political activities that the union engages in. Multiunionism allows workers to self-select into the union that best fits their preferences. Third, the preferences of the union leaders and those of the ordinary members may be different. For instance, while the members may want wage gains and job security, the leadership.wants to maximize its power position. Because of asymmetric information, a moral hazard problem may arise. Under multiunionism, the ordinary members can vote with their feet. This may help to control this moral hazard problem and COLLECTrVE BARGAINING AND ECONOMIC PERFORMANCE-A SHORT REvIEw oF THE THEORY 35 improve welfare. Fourth, the fact that each firm has to deal with more than one union may lead to excessive bargaining costs. Informal Coordination Bargaining coordination need not be embodied in the formal institu- tional framework of collective bargaining. It can be informal. Informal coordination typically takes two forms. The first form is internal coordi- nation among employers and/or employees. On the employers' side, this involves coordination between industry-based employers' organi- zations or individual firms. This plays an important role in Austria, Ja- pan, and Switzerland (Soskice 1990; OECD 1994). On the employees' side, internal coordination typically involves coordination between com- pany- and industry-based unions. The second form of informal coordi- nation is pattern bargaining. Here, a dominant industry or company enters a collective agreement that is followed by other firms and industries. This has been important in Germany, where the metal industry has tra- ditionally acted as the leader. It is clear that informal coordination mecha- nisms are more fragile than those embodied in the formal institutions and are more likely to break down in times of rapid economic and social change or instability, when they are most needed. Corporatism The term corporatism is used to describe situations in which the eco- nomic and political activities of unions and employers' organizations take place within a well-defined framework of social partnership be- tween workers, capitalists, and government (see, for example, Aidt 1997; Bruno and Sacks 1985; Cameron 1984; Henley and Tsakalotos 1993; Lehmbruch 1984; Tarantelli 1986). Within this framework, labor market parties, in particular unions, expect the government to deliver certain welfare goods and policies in exchange for wage restraint (Lange and Garrett 1985). In addition, social partnership can create social consensus and reduce the level of conflict in the labor market. It reduces the cost of implementing economic reforms when they are needed and helps miti- gate coordination failures arising when, in the face of changing econonic conditions, the economy needs to move from one equilibrium to an- other. It also facilitates income policy, economy-wide agreements on wages and weekly hours, health and safety standards, and so on. All these aspects help to bring about "good" economic outcomes. It is, how- ever, important to notice that social partnerships have a tendency to break apart. The point is simple: unions, employers' organizations, and 36 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT individual firms have an incentive to break away from their respective confederations to act on their own. Dispute Resolution The breakdown of negotiations between individual workers and their employers can take various forms, ranging from poor relations in the workplace (with potential costs including decreased levels of labor pro- ductivity through poor morale) to labor turnover (the "exit" option, with the potential loss to the employer of previously made investments in the workers' human capital). At the level of collective contracting, the stakes are arguably much higher for both workers (and their unions) and employers, with the ultimate cost of a negotiation breakdown be- ing lost incomes for the workers and lost profits for employers. Given the potentially high level of these costs to both contracting parties, it is likely that workers and employers have a strong incentive to achieve a solution in preference to conflict. Like all good threats, the employer's threat of a lockout and the union's threat of a strike are best if they en- sure that an agreement is reached before they are implemented. In reality, collective bargaining does sometimes break down, and pro- duction, labor earnings, and profits are lost. It is certainly not safe to as- sume that the total of such costs is greater under the collective bargaining system than under the individual contracting system. We simply do not know whether these costs to society are greater or less than those that would arise from a breakdown in individual employer-employee pay negotiations. Indeed, given economies of scale in the production and dis- semination of information, there are grounds for believing that the collec- tive system, through its ability to resolve disputes, may be a less costly option from a social point of view than individual contracting. There is a strong presumption that when disputes do occur under collective bargaining, it is because of asymmetries in the information possessed by the involved parties (Hicks 1932). A common example is when the trade union "misjudges" the maximum wage that the employer is willing or able to pay. Under such circumstances, the existence of regu- lation can prove decisive in resolving disputes through its information- gathering and -disseminating roles. To understand the process, it is important to recognize the distinc- tion between the union proper (sometimes called the official union) and its rank and file membership. Under this tripartite framework, the offi- cial union (often as a well-informed professional body) acts as an inter- mediary between the union membership and the employer. As such, its role is to reconcile the aspirations of the former against what it judges (on the basis of its more complete knowledge of the overall situation COLLECTIVE BARGAINING AND ECONOMIC PERFORMANCE-A SHORT REVIEW OF THE THEORY 37 than that possessed by the union membership) that the employer would agree to pay. This reconciliation between worker aspirations and labor market realities may be achieved without either party having to resort to its no-trade sanction. However, should negotiations break down and a dispute occur, the role of the official union as a purveyor of informa- tion continues, with information being disseminated in both directions regarding concessions acceptable to each party and any new informa- tion that may materialize as the dispute progresses. This transmission of information continues until demands fall into balance with offers, at which time a settlement is achieved. Viewing a union as an information-gathering and -disseminating body suggests that governments might want to adopt policies that in- crease the efficacy with which unions fill this role. The introduction of so-called cooling-off periods, during which all parties take time to as- sess the situation fully before implementing no-trade strategies, is one example. Other such policies might require that the employer (gener- ally seen as the party in possession of more complete information) di- vulge to the union and its members certain types of information, perhaps in a standard form, to minimize the possibility that disputes will arise because workers incorrectly estimate the employer's ability to pay. Some conflict is inevitable when wages and other employment con- ditions are set by negotiation (either collective or individual), rather than by the invisible hand of the market. Recognizing this fact, there are grounds for believing that a centralized, union-based system of wage bargaining may be less costly to society than an individually based ne- gotiating system in terms of both total transaction costs and dispute costs. We have also seen that unions have a role in resolving disputes if they should occur. Unions: Net Costs or Net Benefits? We can summarize the theoretical discussion of the costs and benefits of unions using the following simple equation: Net benefit of unions = participatory and dispute resolution benefits - monopoly costs - rent-seeking costs. Alternatively, one can talk of cost and rewrite this equation as Net cost of unions = monopoly costs + rent-seeking costs - participatory and dispute resolution benefits. From a theoretical perspective, the net benefit/cost of unions is am- biguous and dependent on the relative size of the three components. These in turn depend significantly on the economic, political, and organizational 38 UNIONS AND COLLECTIVE BARGAINING: ECONONMC EFFEcTs IN A GLOBAL ENvmoNMET environment in which collective bargaining takes place. The economic environment affects both the monopoly costs and the participatory ben- efits. The political environment determines the rent-seeking activities of unions. The organizational environment (bargaining coordination, social partnership, and dispute resolution) affects all three components. Thus, judging the contribution of unions and collective bargaining more gener- ally to the achievement of economic and social outcomes is, at the end of the day, an empirical question. 4 Empirical Evidence from Microeconomic Studies In this chapter, we examine a large body of empirical evidence about the economic effects of unions derived from microeconomic data on individual workers and establishments. Most of the evidence comes from the United States and the United Kingdom but some evidence is available from other industrial countries and from some developing countries. The chapter is organized as follows. First, we present evidence on the union/nonunion wage markup. This is by far the most well-researched aspect of union behavior, with more than 200 studies having been conducted in the United States alone. We examine cross-country differences in the average wage markup, as well as variations in the wage markup across skills, gender, occupation, and ethnicity along with the underlying economic and insti- tutional environment. We then examine the effect of unions on other eco- nomic variables such as employment growth, hours worked, productivity, job mobility, the implementation of new technology, physical investments, spending on R&D, training of workers, profitability, fringe benefits, mode of pay, and pension schemes. Finally, we discuss and evaluate the costs and benefits of unions and relate the net cost of unions to the underlying economic and institutional environment. The Wage Markup in Different Countries The union/nonunion wage markup (the "wage markup") is defined as the difference between the average (nominal) wage of unionized and nonunionized workers with similar individual and workplace character- istics divided by the average wage of a nonunionized worker' The markup 1. In principle, we are interested in comparing the union wage with the wage that would have prevailed in the absence of unions (this is called the "wage gain"). However, the wage gain is unobservable, and the literature focuses on the wage markup. 39 40 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT can.be estimated in different ways. First, it can be estimated as a member- ship markup. The membership markup is based on information about an individual's union status and calculates the difference in wages between individual unionized and nonunionized workers. Second, it can be esti- mated as a recognition markup. Here workers are being categorized ac- cording to whether or not their pay is determined by a collective agreement between a recognized union and a firm. In the latter case, individual union membership is not crucial. What matters is whether the workers' pay is determined by a collective agreement. The distinction between the mem- bership and the recognition markup is important when not all the work- ers whose wages are determined by a collective agreement are union members. For example, when many workers are covered by collective agreements although they are not members of a union (such as in the United Kingdom and Germany), estimates based on the membership markup underestimate the "true" markup, and it is preferable to use the recognition markup to measure the impact of unions on wages. The estimation of the wage markup can be based on data on the wages of individual workers or average sectorwide or occupational wage rates (for example, average wages in different industries or broad sectors, such as manufacturing versus services or manual versus nonmanual work- ers). As it is generally agreed that the markup calculated using sectorwide data is biased upwards, we focus on studies that have used individual cross-sectional data (Lewis 1986; Booth 1995). One estimation approach is to analyze separately how wages are determined for unionized and nonunionized workers. The markup can then be estimated as the differ- ence that arises from differences in the coefficients of the two regres- sions weighted with the average characteristics of workers in the sample. Another estimation approach is to pool all observations for individual workers and run a regression of wages on key characteristics plus an additional (dummy) variable indicating union membership. Historically, both approaches have been used, but more recent estimates are predomi- nantly based on the former approach. This is because the latter approach implicitly assumes that there is a uniform wage determination process for unionized and nonunionized workers-an assumption that is not necessarily justified in practice. A major econometric problem involved in estimating the wage markup is to control for all other factors besides unionization that affect wages. These factors have typically been proxied by variables such as education, work experience, gender, family status, hours worked, firm size, indus- try, occupation, and so on. However, some of these variables may be highly correlated with union status. For example, the fact that a given unionized worker works longer hours than a given nonunionized worker may be the result of union negotiated overtime, or, in another example, working in a particular industry may be a critical factor for belonging to a union EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 41 (such as in the mining sector). In addition, Duncan and Stafford (1980) point out that part of the observed difference between the wages of union- ized and nonunionized workers is compensation for different work con- ditions. They hypothesize that unions develop most often in situations where collective decisions must be made on work conditions that affect all workers and where work conditions cannot be tailored to individual needs. Since these are precisely the kinds of situations in which (union- ized) workers need to be compensated for being in a work enVironment that is worse than the environment in which nonunionized workers oper- ate, the higher earnings of union members may reflect, in part, compen- sating wage differentials rather than rents. Another problem is union-status selectivity. This problem arises because the union-status selection process is not random. For example, workers with high productivity may decide not to join a union or decide to work in the nonunion sector because they hope to get a higher wage than the wage that was collectively agreed on. This phenomenon gives rise to selectivity bias and reduces the reliability of the estimated markup.2 Bearing these remarks in mind, we present be- low a summary of country-specific studies on the wage markup as well as cross-country comparisons.3 Country-Specific Studies In all countries where the wage markup has been estimated, it has been found to be nonnegative. There are, however, significant cross- country variations as well as variations of estimates within countries. There is also some evidence, albeit weak, that the wage markup is, on average, lower in high-income countries than in low- and middle-in- come ones. More specifically, table 4-1 presents summary estimates of the wage markup for six high-income economies (Australia, Canada, Japan, the United Kingdom, the United States, and West Germany), four middle-income economies (Malaysia, Mexico, South Africa, and the Republic of Korea), and one low-income economy (Ghana). 2. One way to deal with the union-status selectivity problem is to estimate the wage equation and a union-status selection equation simultaneously. Another is to use a panel data set that contains a time-series of cross-sections for the same set of workers. This makes it possible to control for unobserved heterogeneity. For example, if unobserved characteristics of workers stay constant over time, they can be taken into account by comr- paring the wage of each worker at one date with his or her wage at a later date. One problem with this solution is that identification is based on workers who change their status, and that is a relatively infrequent event. Lewis (1986) is unable to determine the likely size and direction of the selectivity bias. 3. Extensive discussion on methodological issues on the estimation of the wage markup can be found in Booth (1995), Lewis (1986), Pencavel (1991), and Sapsford and Tzannatos (1993). 42 UNIONS AND COLLECTiVE BARGAINING: ECONOMIC EFFECI5 IN A GLOBAL ENVIRONMENT Table 4-1. Union/Nonunion Wage Markup for Selected Countries Country Period/Year Wage markup (percent) High-income economies Australia 1984-87 7-17 Canada 1969-94 10-25 Japan 5 United Kingdom 1969-95 10 United States 1963-95 15 West Gerrnany 1985-87 0-6 Middle-income economies Korea, Republic of 1988 2-4 Malaysia 1988 15-20 Mexico 1989 10 South Africa 1993-95 10-24 Low-income economies Ghana 1992-94 21-28 Note: Figures in the table indicate the range of the estimated average markup for each country. When a single figure is reported, it refers to a single study except in the case of the United Kingdom and the United States where the reported figures are "summary-best- guesses" (Lewis 1986; Booth 1995). For studies based on cross sectional U.S. data, the range of the markup is 14 to 17 percent. For studies based on panel U.S. data, the range is 7 to 10 percent (Lewis 1986). The corresponding range for the United Kingdom is 3 to 19 percent. Source: Australia: Christie (1992) (16 to 17 percent in 1984), Kornfeld (1993) (7 to 10 percent in 1984-87), Mulvey (1986) (7 percent for women and 10 percent for men in 1982); Canada: Grant, Sivindinshyn, and Vanderkamp (1987) (12 to 14 percent in 1969 and 13 to 14 percent in 1970), Green (1991) (15 percent in 1986); Gunderson (1982) (10 to 20 percent); Gunderson, Ponak, and Taras (2000) (10 to 25 percent), MacDonald (1983), MacDonald and Evans (1981), Robinson and Tomes (1984), Simpson (1985) (11 percent in 1974); Japan: Nakumura, Sato, and Kamiya (1988); Korea, Republic of: Kim (1993), Park (1991); United Kingdom: Booth (1995), Pencavel (1991), Sapsford and Tzannatos (1993); United States: Lewis (1986, 1990); West Germany: Schmidt (1995), Schmidt and Zimmermann (1991), Wagner (1991); Malaysia: Standing (1992); Mexico: Panagides and Patrinos (1994); South Africa: Dabalen (1998), Moll (1993); Ghana: Teal (1996). The most reliable picture comes from the United States and the United Kingdom by virtue of the many studies that have been carried out and the broad consistency of the results. The U.S. wage markup has been estimated in more than 200 studies. While the estimates range from 12 to 22 percent, there is consensus that the average markup is approximately 15 percent (Blanchflower 1996a, 1997; Booth 1995; Lewis, 1986). Filer, Hamermesh, and Rees (1996: 499), however, argue that the wage markup significantly overestimates the wage gain and that the true impact of unions on wages in the United States is only between 8 and 12 percent. In the United King- dom, more than 20 studies have estimated the markup to be in the range of EMPIRICAL EVIDENCE FROM MICROECONOMIC STUnIES 43 3 to 19 percent. Booth (1995: chapter 6) argues that the markup is approxi- mately 10 percent on average (see also Blanchflower 1997). The evidence is sparser for other industrialized economies, but some generalizations are possible. The wage markup for Australia has been estimated to vary between 7 and 17 percent. Similarly, the Canadian wage markup has been estimated to be in the range of 10 to 25 percent, although most estimates seem to fall in the 10 to 15 percent range. Gunderson, Ponak, and Taras (2000) find that the Canadian wage markup was increasing during the 1970s but has been decreasing ever since, reach- ing a low of 10 percent in the beginning of the 1990s. In (West) Germany, where most unions are industry unions and work and pay conditions contained in collective agreements are largely extended to nonunionized workers, the wage markup is found to be small, especially for male workers. Similarly, a study on Japan has found a small average wage markup of about 5 percent. As in (West) Germany, the average estimate hides large gender differences (see below). The evidence is even more limited for low- and middle-income econo- mies. In the Republic of Korea, Park (1991) estimates the markup for male workers in manufacturing industries to be below 4 percent in 1988. The union membership wage markup in South Africa has been estimated in three studies. Moll (1993) estimates the markup to be around 24 per- cent for black blue-collar workers. Dabalen (1998) finds that the mean wage markup for workers is about 19 percent. However, a recent study by Butcher and Rouse (2001) argues that these figures overestimate the wage markup. In any case, these estimates hide interesting variations between workers with different skills and with different racial charac- teristics. We return to this fact below when we discuss the differences in the markup by ethnic group and the markup to workers with different skills. Standing (1992) estimates the wage markup for Malaysia to be in the range of 15 to 20 percent depending on the type of union involved. These effects are somewhat larger than in most industrial countries. Standing attributes this high markup to the fact that Malaysian nonunionized workers can, in the absence of minimum wage legisla- tion, be vulnerable to very low wages. He concludes that the markup can reasonably be in the estimated range even though the political and economic environment in Malaysia is difficult for unions. For Mexico, Panagides and Patrinos (1994) find a 10 percent mem- bership wage markup after having controlled for a large number of income-generating characteristics in a cross-sectional sample of union- ized and nonunionized nonagricultural workers in 1989. They attribute the relative low wage markup to the fact that the 1980s were a particu- larly difficult time for unions in Mexico because of a recession coupled with government austerity measures. It is unclear, however, if this ex- plains the result. First, recession and austerity measures should affect 44 UNIONS AND COLLECTIVE BARGAINING: ECONoNIc EFFECTS IN A GLOBAL ENVIRONMENT unionized and nonunionized sectors alike. Second, one could argue that unions would be in a better position than nonunionized workers to resist downward pressure on wages, thereby increasing the wage markup during a recession. For Ghana, Teal (1996), using three surveys of manufacturing firms in 1992, 1993, and 1994, finds that the wage markup varies between 21 and 28 percent.4 Again, these estimates are significantly higher than those in industrial countries. Teal notes that his cross-section estimates can be biased because of omitted variables (such as unobserved productivity differences) that are correlated with earnings and the union status of the firms. His panel estimates, however, suggest that the selectivity bias may not be a serious problem. Comparative Studies Some studies have estimated the membership wage markup in a cross- country context (Blanchflower 1996b; Blanchflower and Freeman 1992; Blanchflower and Oswald 1994). These studies use comparable indi- vidual worker data for the period 1985 to 1993 to estimate hourly earn- ings equations using similar control variables (such as age, gender, years of schooling, hours worked, and so on). This makes it possible to compare the estimates for different countries directly. Unfortunately, there is no information on industry and establishment size for most countries (with the notable exceptions of the United States and the United Kingdom). These variables are likely to be positively correlated with union status and wage rates. For example, large establishments tend to pay higher wages and have more unionized workers than small establishments. As a consequence, the estimates presented in table 4-2 are likely to overstate the wage markup. On the other hand, the esti- mates refer to a membership wage markup. As discussed above, in countries such as Austria and Germany where almost all workers, unionized or not, are covered by collective agreements, such estimates tend to understate the true effect of unions on wages. As table 4-2 indicates, the markup is positive in all countries, but not significantly so in Canada, Israel, the Netherlands, Spain, and Switzer- land.5 On the other hand, the estimates for Ireland and Japan are extremely 4. Contrary to most of the other studies reviewed, Teal (1996) uses a cross-section of firm-level average wages rather than a cross-section (or panel) of worker wages to esti- mate the markup. Therefore, the markup refers to the union status of the firm and not to an individual's membership in a union. 5. The estimates for Israel, Spain, and Switzerland are, however, based on only a few hundred observations. EMPIRICAL EVIDENCE FROM MICROECONOMNC STUDIES 45 Table 4-2. Union/Nonunion Wage Markup and Characteristics of Collective Bargaining Systems for 15 Countries Union Bargaining Change in Wage Centralizationl density, coverage, union markup, coordination, 1990 1990 density 1985-93 1990 (percent) (percent) 1970-93 (percent) Country (1) (2) (3) (4) (5) Australia 1 41 80 B 9.2 Austria 1 46 98 A 14.6 Canada 3 36 38 C 4.8 Germany 1 33 83 C 3.4 Ireland - 52 - B 30.5 Israel - - - 7.0 Italy 3 39 23 C 7.2 Japan 2 25 71 A 47.8 Netherlands 2 26 67 C 3.7' New Zealand 3 45 75 B 8.4 Norway 1 56 79 C 7.7 Spain 2 13 76 D 0.3 Switzerland 2 27 53 B 0.8 United Kingdom 3 39 47 B 14.7 United States 3 16 18 A 23.3 - Not available. Note: Column 1: The numbers indicate the following: 1. highly centralized/coordi- nated wage bargaining system; 2. semicentralized/coordinated wage bargaining system; 3. decentralized/uncoordinated wage bargaining system (OECD 1997: table 3.3). Column 2: Union density is the proportion of all wage- and salary-earners that is unionized (OECD 1997: table 3.3). Column 3: Bargaining coverage is the number of workers covered by collective agree- ments as a percentage of all wage- and salary-earners (OECD 1997: table 3.3). Column 4: The letters indicate the following: A. declining union density over the pe- riod 1970-93; B. increasing union density in the 1970s but declining thereafter; C. declin- ing density in the 1980s, but stabilizing in the 1990s; D. a sharp increase in union density from 1970 to 1993 (Blanchflower 1996b: table 2). Column 5: An asterisk (*) indicates that the estimate is not significantly different from zero (Blanchflower 1996b). Source: Blanchflower (1996b); OECD (1997). high (31 percent and 48 percent, respectively), perhaps reflecting the lack of control for industry and establishment size. Compare, for example, the estimate for Japan in table 4-1 (5 percent) with the estimate in table 4-2 (48 percent). In addition, the estimates for the United States and the United Kingdom are about 50 percent higher than the widely accepted average estimates of 15 percent and 10 percent, respectively, reported in table 4-1. 46 UNIONS AND COLLECTIvE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT This gives some idea of the extent to which the estimates from the other countries may be exaggerating the true impact of unions. We notice that the wage markup in the United States is higher than the markup in all the other OECD countries except Ireland and Japan. The markup in Austria and the United Kingdom is also high by international standards, whereas the markup in other European countries and Australia is more modest. Can this pattern be attributed to institutional differences among the countries? In table 4-2, we capture institutional differences by the following variables: the degree of centralization/coordination of col- lective bargaining, union density, and bargaining coverage (see chap- ter 5 for a detailed discussion of these indicators). Following OECD (1997), the sample of countries can be classified according to the level at which collective bargaining takes place (the firm, the industry, or the national level) and according to the degree of informal coordina- tion between workers and employers. The countries are ranked from "1" (centralized/coordinated) through "2" (semicentralized/coordi- nated) to "3" (decentralized/uncoordinated) in colunn 1 of tab;le 4-2. Excluding Japan and Ireland for the aforementioned reasons, coun- tries with centralized/coordinated or decentralized/uncoordinated bargaining systems tend to have a higher markup than those with a semicentralized/coordinated bargaining system. The average markup in the last group is 1.6 percent versus 8.7 percent and 11.6 percent, respectively, in the other two groups.6 The markup can also vary depending on the percentage of wage- and salary-earners that are unionized (union density, column 2 in table 4-2 and the percentage of wage- and salary-earners covered by union agreements irrespective of whether or not they are members of unions (bargaining coverage, see column 3 in table 4-2). While union density per se appears to be largely unrelated to the wage markup (correlation coefficient of 0.02),7 bargaining coverage is negatively correlated with the markup (correlation coefficient of -0.58). In other words, the more workers who are covered by a collective agreement, the smaller, other things being equal, the wage markup appears to be." This result is largely driven by the fact that the United States has the largest wage markup and the lowest bargaining coverage (18 percent), and countries like the 6. If a small markup is taken to indicate less distortionary conditions, this result runs prima facie counter to the proposition that a semicentralized bargaining system per- forms worse than both decentralized and centralized bargaining systems (Calmfors and Driffill 1988). 7. Again, excluding Ireland and Japan. 8. Of course, this result could also be caused by the fact that in countries where bargaining coverage is high, the union/nonunion membership markup underestimates the true impact of unionization. EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 47 Netherlands and Spain have a low markup and high bargaining cover- age. Nevertheless, all this points to the possibility that the more work- ers become unionized or are covered by collective agreements, the lower is the markup that they can secure. This may be the case because the labor supply in the noncovered sector decreases when more workers become covered, pushing the nonunion wage up. Taking this argument one step further, we may say that unions are able to secure a high markup only where the marginal cost to society (in terms of impact on the macroeconomy) is small. In effect, unions are ultimately constrained by the wage share in the total economy: they can have wide coverage and a small markup or a high markup at the cost of coverage. The estimates in tables 4-1 and 4-2 refer to a given point in time. An interesting additional question is whether the markup is stable over time or it fluctuates with economic conditions. This question has been inves- tigated in only two countries-the United States and the United King- dom. Blanchflower (1997) concludes that the wage markup in the United States has moved procyclically, but that it does not appear to have a trend over the period 1983-95.9 Stewart (1995) finds that the markup in the United Kingdom has declined a bit during the 1980s.10 These results suggest that union power has not been curtailed significantly, despite the reduction in union density observed over the same period in the two countries."' Blanchflower and Freeman (1992) and Blanchflower (1996b) argue that the high and stable wage markup is one reason why union density has declined in the private sector in the United States.12 The high wage markup, so the argument goes, has created substantial opposition from employers, which, together with a highly adversarial electoral process 9. Historically, the wage markup has moved countercyclically in the United States. For example, it increased significantly duririg the Depression of the 1930s (Filer, Hamermesh, and Rees 1996: table 13.4). This can be related to the fact that union wage contracts are typically long-term contracts. 10. The main reason for this decline is that unions have had a hard time establish- ing a wage markup in new firms (those started after 1984), whereas the markup in "old" firms (those that existed in 1984) is approximately the same in 1984 as in 1990. Moreover, the drop in the additional markup associated with pre-entry closed shops also contrib- uted significantly to the overall drop in the average markup. 11. Union density in the United States has declined from 27 percent in 1970 to 16 percent in 1994. The corresponding numbers for the United Kingdom are 45 percent in 1970 and 34 percent in 1994. 12. This view has been challenged by, for example, Farber and Krueger (1992). They analyze the decline by focusing on the demand and supply of unionized jobs. The de- mand for unionized jobs arises from workers who would prefer a union job without being willing to invest in organizing a union to provide the job. The supply of union jobs arises from workers who are willing to invest in organizing unions, and it is affected by the legal framework and the general resistance of employers to unions. Contrary to the claim by 48 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT to determine union recognition,"3 makes it difficult to unionize new and expanding industries to make up for the contraction of old and union- ized sectors. Stewart (1991) investigates in detail whether the decline in unionization from 1980 to 1984 in the United Kingdom is related to the size of the wage markup. He finds little evidence of this. It is therefore likely that the decline in union density in the United Kingdom is caused by changes in labor market legislation and industrial structure during the period rather than by the size of the wage markup per se. To gain more insight into this issue, we include in table 4-2, column 4, information about the change in union density during the period 1970 to 1993 for the broader sample of countries. An "A" indicates a country where union density declined between 1970 and 1993; a "B" indicates increasing union density in the 1970s but declining thereafter; a "C" in- dicates declining density in the 1980s but stabilizing in the 1990s; and a "D" indicates increasing density throughout the whole period. The wage markup tends to be higher in those countries (groups A and B) that have had declining density rates in the 1980s and 1990s than in those coun- tries (groups C and D) that have had increasing or constant density rates in the 1990s. The average markup for the two groups is 11.8 percent and 4.5 percent, respectively.'4 Although these calculations are very crude and too static to indicate anything about causality, they do demonstrate that changes in union density may be related to the markup. The Efficiency Cost of the Wage Markup When unions are successful in getting a wage markup, workers tend to be displaced from the unionized sectors to nonunion sectors. As pointed out by Rees (1963), this creates a deadweight loss. A number of studies have estimated this deadweight loss and found it to be quite small. The original study by Rees (1963) estimates that the welfare loss in the United States in 1957 was only 0.14 percent of GDP. Johnson and Mieszkowski (1970) find a similar result. Freeman and Medoff (1984: chapter 3) calculate the cost of the average wage markup of 15 percent in the United States in 1980 to be between 0.2 and 0.4 percent of GDP. Studies that use large-scale Computable General Equilibrium (CGE) Blanchflower and Freeman (1992) and others, Farber and Krueger find that demand forces are much more important than supply forces in explaining the decline and that changes in industry structure can account for only one-quarter of the decline. 13. For example, there has been a significant increase in the number of states that enforce "right to work" laws, and at the national level, there have been changes in the interpretation of various labor laws, all of which have made it more difficult for new unions to organize. 14. Excluding Ireland and Japan. EMPRICAL EVIDENCE FROM MICROECONOMIC STUDIES 49 models find comparable small losses. DeFina (1983) uses a 12-sector CGE model to simulate the welfare loss associated with a 25 percent union/nonunion wage markup. He finds modest effects: the welfare loss is no larger than 0.2 percent of GDP. Interestingly, the U.S. results are similar to those for Australia, where the average markup is 7 to 17 percent and where 80 percent of the work force is covered by collec- tive agreements. Christie (1992) estimates the welfare loss associated with the union/nonunion wage markup in Australia to be similar, at about 0.5 percent of GDP. Variations in the Wage Markup The average wage markup disguises the variations that exist across dif- ferent types of workers or different types of collective bargaining. In this section, we review some of these issues. In particular, we consider how the wage markup varies across gender, ethnicity, occupation, skills, education, economic environment, and various characteristics of the collective bargaining system. The Difference in the Markupfor Women and Men Unions are just one of many determinants of the gender wage gap. The gender wage gap is the percentage difference between the wage of a female worker and a male worker who otherwise have the same per- sonal and workplace characteristics. The effect of unions arises in three ways: first, from different unionization rates among men and women; second, from the ability of unions to influence wages in some sectors but not in others; and third, from differences in the wage markup for men and women. The net effect of unions upon female wages relative to male wages is uncertain, and we shall not attempt to resolve the issue here. Instead, we review the evidence related to gender differences in the wage markup and a few studies that focus directly on the effect of unions on the gender wage gap. Although a higher wage markup for women than for men can reduce the gender wage gap, it can also de- crease the wages of nonunionized women to such an extent that the gender wage gap actually increases. In his survey of the U.S. literature, Lewis (1986) concludes that there is very little, if any, difference between the markup for female and male workers. Main and Reilly (1992) and Blanchflower and Freeman (1996) have recently confirmed this conclusion for the 1990s. The same result emerges from Australian studies (Christie 1992; Mulvey 1986). Most stud- ies in Britain show that the impact of unions on women's wages is greater than that on men's wages (Blanchflower 1996b, 1997; Blanchflower and Freeman 1996; Main 1991; Main and Reilly 1992). A typical estimate is 50 UNIONS AND COLLECrIVE BARGAINING: ECONOMIC EFFECTs IN A GLOBAL ENVIRONMENT that the markup for women is 4 to 6 percentage points larger than that for men. However, a few studies (Green 1988; Yaron 1990) find the op- posite result. In any case, taking into account the fact that women work- ers are less likely to be unionized than men workers, the net effect on the average gender wage gap is likely to be small (Doiron and Riddell 1994). Evidence from other OECD and middle-income countries unam- biguously supports the view that the wage markup is greater for women. Nakumura, Sato, and Kamiya (1988), in their study of Japan, find a wage markup of 10 percent for women but fails to find any for men. Likewise, Schmidt (1995) shows that the small average wage markup in West Ger- many is mainly due to a wage markup among unionized female work- ers. The findings for Mexico by Panagides and Patrinos (1994) suggest that the markup for women is 9.8 percentage points higher than that for men with similar characteristics. Finally, Moll (1993) finds that the wage markup among black blue-collar workers in South Africa in 1985 is about 11 percentage points higher for women than for men (31 percent com- pared to 19 percent). Some studies focus directly on the gender gap rather than on gender differences in the markup. Simpson (1985), for example, estimates the gender wage gaps in the unionized and nonunionized sectors in Canada. He finds that the gap is 22.9 percent in the unionized sector and 20.3 per- cent in the nonunionized sector. This indicates that unions have little im- pact on the gender wage gap. Doiron and Riddell (1994) incorporate the effect from the increase in the female unionization rate and the decrease in the male unionization rate in their analysis of the gender wage gap in Canada. They show that, had it not been for union effects, the gender wage gap would'have increased by 7 percent in the 1980s. Also, the gen- der wage gap in the nonunion sector makes a larger contribution to the gender wage gap than does the gap in the union sector. Overall, this sug- gests that unions in Canada have helped to reduce gender discrimina- tion, albeit this may not be directly related to differences in the wage markup for the two groups of workers. Standing (1992) compares the wage ratio of male and female workers in nonunionized and unionized firms in Malaysia. His result suggests that the presence of a union reduces the ratio of male to female wages, and he concludes "the data gives primafacie support to the view that in terms of wages, at least, women gain more than proportionately from unionization" (Standing 1992: 341). Differences in the Markup by Ethnic Group Discrimination among workers with different ethnic backgrounds but otherwise similar productivity characteristics can lead to a wage differ- ential. Here we are interested in the impact of unions on the markup for workers with different ethnic characteristics. EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 51 In the United States, it is not clear whether there is a substantial dif- ference between the wage markup for white and nonwhite workers. Some studies fail to find any difference, whereas others find that the markup is 5 to 10 percentage points higher for blacks than for whites (Lewis 1986: chapter 7). In the United Kingdom, the sparse evidence available shows that a nonwhite unionized worker gets a higher markup than a similar white worker (Blanchflower 1997). In South Africa, Dabalen (1998) finds that white workers get a markup of about 30 percent, whereas the markup for black workers is in the range of 16 to 20 percent. Butcher and Rouse (2001) find that after controlling for worker heterogeneity the markup for both groups is somewhat smaller, in particular for white workers (10 percent). As we discuss below, these numbers hide interest- ing differences between workers with different skills. Patrinos and Sakellariou (1992) decompose the difference between the average wage of employed Indians and non-Indians in Canada into (a) the part that is explained by differences in income-generating characteris- tics such as years of schooling, experience, unionization, and other vari- ables and (b) the part that cannot be explained by these variables."5 The unexplained part is taken as an indicator of discrimination against Indi- ans in the labor market. They find, on the one hand, that unions margin- ally reduce discrimination (but other variables such as education and experience are much more important). On the other hand, since Indians are less likely to be unionized than non-Indians, this tends to increase the total earnings differential between the two groups. Panagides and Patrinos (1994) investigate the impact of unions on the wages of indigenous people in Mexico. They include a variable in their wage regressions for union- ized and nonunionized workers that measures the percentage of the popu- lation in a particular county who are indigenous. They show that a worker who lives in a county with a large indigenous population gets a bigger wage markup than a similar worker living in a less "indigenous" county. The Private versus the Public Sector It is unclear if public sector workers are in a weaker position than private sector workers to exert wage pressure. Historically, they have been re- stricted from forming unions in many industrial countries. Even when public sector unions are legal, they are often legally barred from striking. This suggests that the average wage markup in the public sector may be smaller than the corresponding markup in the private sector. On the other hand, public goods and publicly provided private goods are produced in an environment with no or little competition. Moreover, producers, in the 15. Indians refer to Canada's Aboriginal or Native people. 52 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFEcrs IN A GLOBAL ENVIRONMENT public sector are not motivated by a profit concern; rather, they have a politically imposed budget constraint. The lack of competitive pressure and soft budget constraints makes it easier to pass on the costs of high wages and overstaffing to taxpayers. Finally, as argued by Freeman (1986), public sector unions may be able to influence employers' behavior through the political process. In many developing countries, unionization is con- centrated in the public sector, and there are few legal constraints on the kind of behavior that these unions may engage in. In such an environ- ment, unions are able to exercise substantial political pressure. This may contribute significantly to the rent-seeking cost of unions in addition to the impact it may have on the wage markup. In this section, we examine studies that compare the wage markup between the public and private sectors in Canada, the United Kingdom, and the United States. Most studies divide individual workers into two groups. One group contains all employees (police officers, firefighters, truck drivers, teachers, white-collar public administration workers, and so on) or a subset of employees (such as craftspeople or white-collar work- ers) in the public sector. The other group contains all workers in the economy. Each of these groups is then divided into unionized and nonunionized workers, and the wage markups for workers in the public and private sectors are estimated. Lewis (1990) identifies a number of spe- cific problems associated with estimating the public sector wage markup. First, in the private sector, wage and working conditions of unionized workers are determined by a union-negotiated contract, whereas non- union workers are typically excluded from the benefits of the union con- tract. In the public sector, the "wage-comparability" criterion is often used, and it is not uncommon for both union and nonunion workers to get the same wage and working conditions. This makes it more difficult to clas- sify workers according to how their wage is being determined and makes it preferable to use the recognition markup rather than a membership markup. Second, the work force mix differs between the private and pub- lic sectors. For example, the work force in the public sector generally con- sists of a disproportionately large share of white-collar workers. Since the markup for white-collar workers tends to be smaller than that for blue- collar workers, failing to take the work force mix into account can under- estimate the markup in the public sector. Third, workers in the two sectors receive different amounts of fringe benefits. Lewis (1990) reviews 75 U.S. studies that have estimated the wage markup for the public sector at large or for some specific groups of work- ers within the public sector (such as teachers). After correcting for the problems discussed above, he concludes "[t]he mean wage gap [wage markup] (after adjustments for fringe benefits and workforce mix) in the public sector in 1973-84 moved approximately parallel to that in all EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 53 sectors but at a level lower by about 0.03 to 0.07. I estimate that the public- sector gap in this period averaged about 0.08-0.12... " (Lewis 1990: 321). More recently, Blanchflower (1997), who uses data from 1993-94, ob- tained a similar result. Moreover, Lewis (1990) finds that, within the public sector, the wage markup is lowest for federal employees and high- est for employees of local governments in the United States. In fact, the average markup for workers employed by local governments in 1973- 84 was slightly higher than the economy-wide average. Green (1988) considers the difference in the wage markup for workers in the public and private sectors in the United Kingdom. He finds that the wage markup is smaller for both manual and nonmanual workers in the pub- lic sector than in the private sector. Blanchflower (1996b, 1997) and Blanchflower and Freeman (1992), however, find that the wage markup in the private and public sectors is very similar (the difference being about 2 percentage points in favor of the private sector) in the United Kingdom. Robinson and Tomes (1984) and Simpson (1985) find a simi- lar result for Canada. The Markupfor Workers with Different Skills Collective bargaining can insert a wedge between worker productivity and wages. Although this can be desirable as a way to avoid wage in- equality from a societal point of view, it can also distort the relative wages of skilled and unskilled workers or the relative rewards for different types of jobs. As a consequence, resources can be misallocated, and the resulting efficiency loss has to be traded off against the distributive gain. In the United States (Lewis 1986: chapter 7) and the United King- dom (Booth 1995: table 6.1), manual workers get a larger markup than nonmanual workers. Likewise, semiskilled workers get a larger markup than skilled workers. Similar results have been obtained in Canada (Simpson 1985). In South Africa, the wage markup for workers with dif- ferent skills varies between different ethnic groups. For example, the average wage markup for unskilled nonwhite workers is 19 percent, while the wage markup for semiskilled nonwhite workers is much smaller and the markup for skilled nonwhite workers is practically zero. On the other hand, semiskilled and skilled white workers got a wage markup of 13 percent in 1985 (Moll 1993). Moll (1993: 256) concludes that "black unions tended to compress wages by skill level."16 Although Dabalen 16. The comparison between black and white workers with respect to the markup for different skill groups should not be carried too far. This is because the sample of white workers is small and also because white and black workers are probably distributed dif- ferently across industries. 54 UNIONS AND CoLLEcnvE BARGAINING: EcoNoMIc EFFECTs IN A GLOBAL ENViRoNMENT (1998) finds somewhat smaller estimates, his study confirms the general pattern observed by Moll (1993). In Malaysia, unions reduce the intrafirm wage differential between skilled and unskilled workers. In particular, industry unions tend to reduce the differential between workers with different skills more than company unions do (Standing 1992). Unions, Wage Dispersion, and the Return to Schooling The facts that unionized workers get a wage markup and that unioniza- tion is concentrated among low-paid workers suggest that unions re- duce the wage dispersion across an economy. Moreover, the impact of unions on the wage dispersion within the unionized sector can also con- tribute to low overall wage dispersion. For example, the differences be- tween the wage markup for different skill groups (see above) indicate that unions contribute to the compression of wages within the union- ized sectors of the economy. There are many reasons why unions may be keen to promote a compressed wage scale across different groups of workers employed within the unionized sectors of the economy. One reason may be that they have egalitarian wage goals. Egalitarian wage goals can arise if productivity differs among union members and if the median member has low (compared to average) productivity. Under these circumstances, a democratic union tends to enter wage contracts that compress the wage structure (Freeman 1980b). Evidence from the United Kingdom and the United States indicates that unions reduce wage dispersion significantly between industries, between (similar) firms within an industry, and among workers within a firm (Freeman 1980b; Gosling and Machin 1994).17 Evidence from Mexico points in the same direction. Panagides and Patrinos (1994) com- pare the Gini coefficient associated with the wage distribution for union- ized and nonunionized workers.'8 For nonunionized workers, the estimated Gini coefficient is 42.1. The corresponding number for union- ized workers is 33.5. The fact that unions tend to reduce the wage dispersion can affect the decisions people make about their children's and their own education. These decisions depend on a number of factors. One important factor is the return to education in terms of higher (future) wages. If unions re- duce the return to schooling, say, by compressing the wage differential between workers with different skills, they can have an adverse impact 17. In the United Kingdom, the bulk of the overall rise in earnings inequality is, however, due to a large increase in earnings dispersion across nonunion establishments. 18. The Gini coefficient is a measure of the degree of wage inequality. The larger the value of the coefficient, the more unequal is the distribution of wages. EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 55 on the formation of human capital. On the other hand, when the relative wage of unskilled workers increases, firms substitute away from unskilled workers. To avoid being unemployed, (unskilled) workers have to ac- quire more skills, so the compression of the wage distribution may in- duce more, rather than less, human capital formation."9 The empirical evidence suggests that the wage markup is usually higher for less edu- cated workers in countries such as Canada, the United Kingdom, and the United States.20 The same pattern appears in Mexico, where Panagides and Patrinos (1994:18) show that the wage markup decreases as the edu- cation level of workers increases. The Economic Environment The economic realities facing firms can make it difficult for unions to get a high wage markup. As discussed in chapter 3, competitive pres- sure from both the product market and the nonunion labor market can be particularly effective in serving this role. A number of studies from the United States and the United King- dom have investigated the effect of competitive pressure on the wage markup. Most of these studies use industry concentration as a proxy for a firm's market power. Concentration as a measure of competitiveness has been proxied by either the volume of sales or employment accounted for by the three or five largest firms in the industry or some more so- phisticated measure such as an index of concentration. In the United States, the majority of studies (Lewis 1986: 154) find a negative correla- tion between industry concentration and the wage markup in manufac- turing industries. This is also the case in the United Kingdom (Stewart 1983). In a study of Canada, Martinello and Meng (1992) find indirect evidence that industry characteristics such as concentration, import pen- etration, and labor substitutability have little impact on the wage markup. These results do not support the theoretical predictions. However, the relationship between monopoly power and the wage markup can be masked in these estimates if wages are high in concentrated indus- tries even in the absence of unions. For example, this would be the case if firms in these industries wish to forestall unionization. Another rea- son is that firms in concentrated sectors would like to escape possible enforcement of competitive laws or they want to avoid the bad press associated with high profits and low wages. As argued by Sapsford and 19. See Ravn and S0rensen (1997) for a recent model that makes this type of argu- ment in a model with a minimum wage that compresses the wage distribution from below. 20. See Blanchflower (1997), Booth (1995), Christie (1992), Lewis (1986), Panagides and Patrinos (1994), and Simpson (1985). 56 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECiS IN A GLOBAL ENVIRONMENT Tzannatos (1993: 203-4), these reasons may be more apparent than real. For example, the fear of provoking the response of the competitive au- thorities is hypothetical in many countries where there is little faith in competition laws and little effort is spent on enforcing-them. Also, firms may receive better press coverage by spending part of their excess prof- its on health and safety improvements in the workplace or by donating to charities rather than by paying higher wages. Another interpretation of the evidence is that firms in concentrated industries use their mo- nopoly rent to withstand the wage demands of unions. This may induce workers to be content with greater job security and other nonmonetary benefits as a substitute for high wages. Finally, the industry concentra- tion ratio may not be a good proxy for a firm's monopoly power in the product market. This suspicion seems to be confirmed by a few studies that have used indicators other than industry concentration to measure the monopoly power of firms. These studies find that the wage markup is larger in firms with monopoly power than in those without it. First, Mishel (1986) uses a mixture of industry concentration ratios and a subjective measure of en- try barriers to the industry as a proxy for monopoly power in a sample of unionized U.S. manufacturing firms (in 1968 to 1972). He finds that the wage markup is significantly higher in noncompetitive industries than in competitive ones. Second, Stewart (1990) measures the degree of product market competition among U.K. firms by simply asking the management of each firm about the number of competitors that they are facing in the product market. This provides a firm-specific measure of monopoly power. When it is defined in this way, he finds that competition in the product market significantly reduces the average wage markup. In particular, in firms that operate in a competitive product market, the wage markup is, on average, zero. On the other hand, firms that have little or no competi- tion in the product market grant a wage markup in the range of 8 to 10 percent. Moreover, unions are unable to create a wage markup in firms that primarily operate, in international markets. In addition, unions are able to create a markup only in industries that are sheltered from foreign competition when the whole industry is unionized. The Design of Collective Barga-ining We have seen that workers, on average, get a wage markup if they are members of a union or otherwise have their pay conditions determined by collective agreements. However, the size of the markup may depend on how collective bargaining is organized. In this section, we focus on the impact that different bargaining structures and institutions have on the wage markup in addition to the membership or coverage effect. We consider four aspects of the institutional framework: EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 57 * The extent of unionism (average union density in the industry or the percentage of firms in the industry that recognizes a union). * The level at which bargaining takes place (the firm, industry, or the national level). * Multiunionism (more than one union can potentially represent the same worker). * Closed shops (a worker can obtain or retain a particular job only if he or she is a member of a particular union). The extent of unionism. There seems to be a strong relationship between the extent of unionism in an industry (or occupation) and the wage markup. In industries where unionization is low in terms of ei- ther density or the percentage of firms that recognize a union, unions generally have little impact on wages. This is because attempts to raise the wages paid by a few unionized employers (above what their com- petitors pay) put union employers at a severe disadvantage in the prod- uct market. This increases employers' resistance to union wage pressure and encourages the union to moderate its wage demands. On the other hand, in industries where almost all firms are unionized, unions will have more bargaining power and will therefore be able to secure a higher wage markup. This is known as the "extent of unionism" effect (Green 1988: 186). Many studies have estimated this effect in the United States. They use union density in specific industries as the relevant indicator (the percentage of all workers in the industry who are members of a union). The main finding is that union density increases the membership markup, although there is substantial disagreement about the magni- tude of this effect (Lewis 1986: chapter 7). In Canada in the late 1970s, the wages of otherwise identical unionized Canadian workers in indus- tries with high union density were 13 percent higher than in industries with almost no unionized workers (Robinson and Tomes 1984). Green (1988) investigates the relationship between the membership wage markup and the union density of the relevant industry in the United Kingdom at the beginning of the 1980s. He finds that the wage markup is always larger in industries with more than 70 percent union density. For example, in these industries, the (hourly) wage markup for manual workers is 34 percent, compared to 7 percent in industries where union density is less than 30 percent. The corresponding markups for nonmanual workers are 13 percent and practically zero. These findings are consistent with earlier results (Stewart 1983). In the United States, there is no significant difference between union density and coverage of collective agreements. In the United Kingdom, on the other hand, a large number of workers have their pay conditions determined by collective agreements without actually being members 58 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVLRONMENT of a union. Therefore, focusing on the membership wage markup may bias not only the estimate of the wage markup itself but also the esti- mate of the extent of the unionism effect. More recent British studies take this into account and analyze the relationship between the recogni- tion wage markup and union density at the establishment level. Metcalf and Stewart (1992) find that the recognition wage markup (for semi- skilled manual workers in 1984) is significant only for firms where more than 95 percent of the work force is unionized. The markup is in the range of 7 to 10 percent. In firms where a smaller fraction of semiskilled workers are union members, the wage markup is insignificant. This sug- gests that workers in a workplace where management does recognize a union benefit in terms of higher wages only if almost all the workers in the workplace are actually organized in unions. The level at which collective bargaining takes place. The level at which collective bargaining takes place affects how workers and employ- ers interact, and this has implications for the size of the wage markup. From the discussion in chapter 3 of centralization of collective bargain- ing, we would expect the wage markup to be higher when collective bar- gaining is at the industry level rather than at the firm level. Using individual worker data from U.K. establishments in 1980 and 1984, Stewart (1987) fails to find any evidence that the level of bargaining affects the markup for semiskilled and skilled manual workers. This is in contrast to earlier studies (Mulvey 1976). Mishel (1986) found, for example, that the wage markup was higher in U.S. industries that used centralized bar- gaining in 1968-72 than in industries that used firm-level bargaining. Standing (1992) analyzes the impact of industrial and company unions on the wage markup in Malaysian manufacturing firms in 1988. In Standing's terminology, an industrial union is a union that organizes workers from a given industry irrespective of their trade. A company union, on the other hand, is a union that only organizes workers who are employed in the relevant firm-again irrespective of their trade. He finds that the average wage markup paid by a firm that deals with an industrial union is 19.7 percent, compared to 14.9 percent in firms that deal with a company union. Bhattacherjee (1987) examines similar issues for India. He uses a data set of 119 plant-level agreements between manufacturing firms and blue-collar unions in Greater Bombay and Pune in 1978-84. He distinguishes two types of unions: (a) external unions (unions that are explicitly affiliated with a trade union federation, which in turn is affili- ated with a political party) and (b) independent plant-based unions (unions run and managed by workers employed in the plant). His main finding is that members of independent plant-based unions get significantly higher wages and bonuses than other workers. While these findings may reflect the specific circumstances that prevailed in Bombay at the time of the EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 59 study, they do suggest that plant unions (the Japanese model of industrial relations) do not always produce wage constraint. Multiunionism. The prime source of information on the effect of multiunionism is the United Kingdom where multiunionism has tradi- tionally played an important role. In.the 1980s, about 30 percent of all unionized plants in the private sector recognized more than one union for collective bargaining purposes. Under multiunionism, the unions may bargain together (multiple bargaining) or separately (separate bar- gaining) with management. Stewart (1987) finds that multiunionism is associated with a higher wage markup in the United Kingdom. Subse- quent studies have refined this result. Machin, Stewart, and van Reenen (1993) have shown that it is not multiunionism per se that is associated with the additional wage markup; it is the combination of multiunionism and separate collective bargaining that produces the additional wage markup. If all the unions that represent workers at a given workplace bargain together, then the wage markup is no larger than in firms where workers (of the same type) are represented by a single union. Closed shops. A closed shop exists when an employee can obtain or retain a particular job only if he or she is a member of a particular union. The closed shop can be either pre- or post-entry. A pre-entry closed shop requires that the employee is accepted as a member of the relevant union ("holds a union card") before he or she can be employed in the particular trade. Historically, craft unions have managed to run a pre-entry closed shop.2' One example is the International Typographical Union in the United States, which at its peak required that all individuals hired for the composing room must already have union cards. A post-entry closed shop requires that the employee joins the union upon getting a specific job. In the United Kingdom, post-entry closed shops have been impor- tant in industries such as metal engineering, transport, and communi- cations (Stevens, Millward, and Smart 1989). From a theoretical point of view, a closed shop increases a union's control over labor supply and as a result its bargaining power.2 The question, therefore, is whether the presence of a closed shop increases the wage markup over and above the basic recognition or membership effect. 21. Some professions such as doctors and journalists regulate themselves and are effectively running pre-entry closed shops. 22. A British study by Stevens, Millward, and Smart (1989) finds that employees in pre-entry closed shops face a formidable range of sanctions if they lose or give up their union membership. For example, about two-fifths of workers expected to be dismissed or made to resign by their employers, and a quarter argued that the union did limit entry by restricting the number or types of people that it takes into membership. 60 UNIONS AND COLLECI-1VE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT In the United Kingdom, closed shops used to be common before various changes in the industrial relations framework took place dur- ing the 1980s.23 In 1989, about 12 percent of all employees worked un- der some kind of closed shop arrangement. Half of these were pre-entry and half were post-entry closed shops (Stevens, Millward, and Smart 1989). Stewart (1987, 1991) uses data from the 1980 and 1984 work- place survey to analyze the effect of closed shops on the wage markup. Stewart (1991) finds that the pre-entry premium on semiskilled pay is about 14 percent and the post-entry premium is about 3 percent rela- tive to the pay in an establishment with union recognition but no form of closed shop. Metcalf and Stewart (1992) investigate if the wage pre- mium found in earlier studies was due to the presence of a closed shop per se or if it arose because union membership by definition is high in establishments that run a closed shop. To this end, they use a British workplace survey from 1984. Their main finding is that the post-entry closed shop does not increase the wage markup above what it would have been had the majority of the firm's workers been unionized. On the other hand, pre-entry closed shops can increase the wage markup by as much as 100 percent. They conclude, "[T]he pre-entry variety is-on the pay dimension-a separate institutional form with an effect additional to density. This is, however, not so for the post-entry closed shop, which brings no extra reward in terms of pay in addition to that resulting from high density" (Metcalf and Stewart 1992: 507). How- ever, subsequent research using a British workplace survey from 1990 (Stewart 1995) finds that the premium associated with the pre-entry closed shop has been reduced and is roughly the same as that found in firms where management recommends union membership. Other Union Effects In this section, we examine the impact of unions on other aspects of economic performance than wages. These include employment growth, hours worked, productivity (level and growth), job mobility, implemen- tation of new technology, spending on R&D, training, profitability, fringe benefits, mode of pay, and provision of pensions. Table 4-3 summarizes our conclusions and provides a classification of union effects on nonwage dimensions of economic performance accord- ing to the degree of confidence that we have in them. In column one, we list the relevant dimension of economic performance. In column two, we 23. In the late 1980s, the British government banned pre-entry and post-entry Closed shops. Table 4-3. Summary of Union Effects on Nonwage Dimensions of Economic Performance Number Variation Degree of Indicator Union effect of studies in results Evidencefrom robustness Hours worked Unions reduce total and normal hours 25 Agreement Australia, Canada, Germany, Ireland, High but increase paid overtime. Italy, Japan, Netherlands, New Zealand, Spain, Switzerland, United Kingdom, and United States Job mobility Unions reduce voluntary turnover and 16 Agreement Australia, Japan, Malaysia, United High increase job tenure and temporary Kingdom, and United States layoffs. Profitability Unions reduce profitability. The effect is 18 Agreement Japan, United Kingdom, and High larger when firns have product-market United States power. Productvity level Unions have a negative impact on 28 Disagreement Germany, Japan, Malaysia, United Low productivity levels unless (a) unionzed Kingdom, and United States firms operate in competitive product markets and/or (b) industrial relations in the workplace are of high quality. Productivity Unions have an ambiguous effect. In 18 Disagreement United Kingdom and United States Low growth the United States, the effect is nonpositive. In the Unuted Kingdom, the impact depends on the time penod. New technology There is little or no difference between 6 Some Canada, Malaysia, United Kingdom, Low unionized and nonunionized firms with disagreement and United States respect to the implementation of new technology. Training Unions tend to increase company- 4 Some Malaysia, United Kingdom, and related training. disagreement United States Low (table continues on following page) Table 4-3 continued Number Variation Degree of Indicator Union effect of studies in results Evidencefrom robustness Pay systems Unions reduce the use of individual 3 Agreement Malaysia, United Kingdom, and Low performance pay and increase the use United States of seniority pay. Employment Unions reduce employment growth. 8 Some Canada, Jamaica, Malaysia, Medium growth disagreement United Kingdom, and United States Physical Unions reduce the investment rate of 6 Agreement United Kingdom and United States Medium investments physical capital. Spending on R&D Unions reduce spending on R&D. 6 Agreement United Kingdom and United States Medium Employment- related benefits Unions increase employment-related 8 Agreement Australia, Japan, Malaysia, Medium benefits such as severance pay, paid United Kingdom, and United States holidays, paid sick leave, and pension plans. Source: See text. The judgement of robustness is that of the authors. EMPIRICAL EVIDENCE FROM MICROECONOMiC STUDIES 63 summarize our evaluation of the effect of unions on the relevant dimen- sion of economic performance. In columns three to six, we provide infor- mation about the robustness of the results. First, we indicate how many studies we identified and surveyed (column three).34 Second, we indicate if there is agreement among the studies about the direction of the union effect (column four). Third, we list the set of countries from which evi- dence is available (column five). In the last column, we provide our evalu- ation of the robustness of the results. The degree of robustness can be high, medium, or low. It is said to be high if (a) there is agreement about the direction of the union effect, (b) if more than 10 studies can be identi- fied, and (c) if the evidence from the United States and the United King- dom is confirmed by evidence from other countries. The degree of robustness is said to be low if there is disagreement about the direction of the union effect or fewer than five studies have been identified and sur- veyed. The degree of robustness is said to be medium if (a) there is agree- ment among the studies about the sign of the union effect, and (b) between 5 and 10 studies, including a study from at least one country other. than the United States and the United Kingdom, can be identified. Unions and Employment The wage markup reduces total employment as long as the demand curve of labor in the unionized sector of the economy is sloping down- ward and the management of unionized firms retains the right to man- age (that is, management independently decides on employment after wages have been agreed with the union). However, as pointed out in the section in chapter 3, the adverse employment effect can be reduced and even reversed if (a) unions and firms bargain over wages and em- ployment and enter an efficient contract or if (b) firms have monopsony power in the absence of collective bargaining. Oswald and Turnbull (1985) and Oswald (1993) have investigated if unions and firms typi- cally bargain over employment. They find that this is rarely the case in either the United Kingdom or the United States. In the United States, many contracts explicitly state that the right to determine employment remains with the management. While this is not true in the United King- dom, U.K. unions do not generally bargain over employment.35 How- ever, although the employment level per se is not subject to formal 24. With respect to the evidence from the United States and the United Kingdom, we have drawn heavily on the surveys provided by Bellman (1992), Booth (1995), Lewis (1986), and Pencavel (1991). 25. There are exceptions such as in the printing and mining sectors. 64 UNIONS AND COLLEcTIVE BARGAINING: ECONOMIC EFFEcrs IN A GLOBAL ENVIRONMENT bargaining, recruitment, staffing norms, redundancy pay, and deploy- ment are included in formal bargaining, and this can have indirect ef- fects on employment (Booth 1995: table 4.1). It is possible to test econometrically the right to manage model (unions push up wages and reduce employment) against the efficient bargaining model (unions push up wages and employment).26 The results consistently reject both mod- els. Ulph and Ulph (1990: 102) conclude "on the Whole neither theory seems to be able to account satisfactorily for the data on negotiated wages and their associated employment levels." Although it is tempting to ar- gue that the truth should lie somewhere in the middle, data limitations and the procedure used to test the two models may be seriously flawed (Booth 1995: 137-40).27 Another way to assess the impact of unions on employment is to look at employment growth. Here the available evidence from Canada, Jamaica, Malaysia, the United Kingdom, and the United States suggests that employment grows more slowly in unionized firms than in nonunionized ones. Studies from Canada, the United Kingdom, and the United States2l typically find a growth differential in the range of 3 to 5 percentage points per year in favor of nonunionized firms.29 Evidence 26. See Alogoskoufis and Manning (1991), Bean and Tumbull (1988), Brown and Ashenfelter (1986), Card (1986), MacCurdy and Pencavel (1986), and others. 27. In short, the idea of the test is this. If the right to manage model is true, then conditional on the wage negotiated in the contract, the alternative wage should have no independent impact on employment. On the other hand, if the efficient contract model is true, then the alternative wage should have an impact on employment. Unfortunately, the different impact of the alternative wage on employment depends on the specification of the union's objective function (Pencavel 1991: 210). Moreover, as pointed out by Oswald (1993), the difference between the efficient contract model and the right'to manage model disappears when membership dynamics are taken into account. 28. Boal and Pencavel (1994), Bronars, Deere, and Tracy (1994), Dunne and MacPherson (1994), Freeman and Kleiner (1990), Lalonde, Marschke, and Troski (1996), and Leonard (1992) provide evidence from the United States. Long (1993) estimates the differential for a sample of Canadian firms. Blanchflower et al (1991) provide evidence from the United Kingdom. The study by Blanchflower et al (1991) has been subject to criticism. Machin and Wadhwani (1991) argue that the difference in the growth rate of employment in unionized and nonunionized firms is due to the fact that unionized firms experienced a reduction in restrictive practices during the sample period 1980-84. This implies that unionized firms were more likely than nonunionized firms to lay off workers during this period. This strongly suggests that the estimated effect is associated with an adjustment to a long-run equilibrium rather than with an equilibrium position per se. 29. In their study of coal-mining data from West Virginia during the period 1897- 1938, Boal and Pencavel (1994), however, find that the employment growth differential is approximately zero if the variation in working days is explicitly taken into account. Machin and Wadhwani (1991) find, using U.K. data, that employment grew faster in unionized establishments in the 1970s. Blanchflower and Burgess (1996) find that unions have a nega- tive effect on employment in the United Kingdom but not in Australia. EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIEs 65 on the employment growth differential is mixed in Malaysia. The em- ployment growth differential is about 5 percentage points per year in firms that bargain with industrial unions but is insignificant in firms that deal with a company union (Standing 1992). Rama (1998) estimates the employment differential between sectors with high and low union membership rates for Jamaica over the period 1986-93 to be 2 to 5 per- centage points per year. He argues that the most plausible explanation for the growth differential is slow productivity growth in the unionized sectors of the Jamaican economy. This growth differential is quite substantial and may well represent disequilibrium phenomena. Rama (1998) provides a number of poten- tial explanations for the observed employment growth differential: * It takes time and effort to organize a union. Consequently, at a given point in time, old firms are more likely to be covered by unions than newer firms are. If newer firms expand faster than old firms, we would expect to observe higher employment growth in the newer, nonunionized firms. * Unions are more likely to be concentrated in sectors that enjoy large rents. If these sectors are less dynamic because of monopoly inefficiencies and their activities are limited by the size of the domestic market, employment would tend to grow more slowly in these sectors. * Unions may encourage labor hoarding by increasing hiring and firing costs. This would make unionized firms more reluctant to hire new workers during a boom, thus reducing employment growth over the cycle. * Labor costs grow faster in unionized firms than in nonunionized ones. * Productivity grows slower in unionized firms than in nonunionized ones. Voluntary Turnover, Layoffs, and Job Tenure The evidence from Australia, Japan, Malaysia, the United Kingdom, and the United States unanimously shows that voluntary turnover (mea- sured by the "quit" rate) is lower and job tenure is longer in unionized firms than in nonunionized ones.3x Freeman and Medoff (1984: 109-10) estimate the welfare gain associated with a reduction in labor turnover to be equivalent to a 0.2 to 0.3 percent increase in GDP in the United 30. See Elias and Blanchflower (1989), Freeman (1980a), Kupferschmidt and Swidensky (1989), Miller and Mulvey (1991,1993), Muramatsu (1984), Osawa (1989), and Standing (1992). 66 UNIONS AND COLLECTIVE BARGAING: ECONOMIC EPFEcTS IN A GLOBAL ENVIRONMoNT States in the 1980s. For unionized firms, they estimate the gain to be equivalent to a 1 to 2 percent reduction in costs. To calculate the gain to workers, Freeman and Medoff (1984) first calculate the increase in wages necessary to reduce the nonunion quit rate to the union quit rate, and, second, they weight this by the difference between the quit rate in the unionized and nonunionized sectors. While these calculations are crude, it is interesting to notice that the welfare gain associated with participa- tory benefits of this kind is of the same order of magnitude as the esti- mated monopoly cost of unions. As pointed out by Freeman and Medoff (1984), it should be kept in mind, however, that the participatory benefit accrues to organized workers (and firms) only, whereas the monopoly cost of unions is borne by society at large. While voluntary turnover tends to be lower in unionized firms than elsewhere, unions increase the use of layoffs, particularly temporary layoffs. In a study of layoff patterns in U.S. manufacturing firms in the 1960s and 1970s, Freeman and Medoff (1984: chapter 7) find that unions significantly alter the firm's choice between layoffs, wages, and hours worked in response to business cycle fluctuations. Unionized firms ad- just by making temporary layoffs rather than by reducing weekly hours (Work sharing) or wages.31 In particular, unionized (blue-collar) work- ers are 50 to 60 percent more likely to be laid off temporarily than nonunionized workers. One explanation for this may be the fact that junior workers can be laid off more easily. Also, senior workers typi- cally have more influence on the union's policy than junior workers. Faced with the choice between a reduction in their earnings or a tem- porary layoff of junior Workers, unions are likely to prefer layoffs. An- other explanation is that the cost of temporary layoffs can be shifted onto the unemployment benefit system. As long as there is less than a 100 percent experience rating (in other words, as long as the amount that firms contribute to unemployment benefits is less than the costs of the unemployment that they generate), those firms with above-aver- age layoffs are subsidized at the expense of firms with below-average layoffs. Allen (1988) points to an interesting difference between the pri- vate and public sectors. While it is true that the existence of unions increases layoffs in the private sector, the opposite is true in the public sector. Public sector unions do not increase the wages of their members as much as private sector unions do. Instead, they reduce layoffs and protect employment. 31. Temporary layoffs refer to a situation in which a worker is laid off for a shorter period of time (less than a month) and is recalled or rehired by the same firm. EMPIRICAL EVIDENCE FROM MICROFCONOMIC STUDIES 67 Unions and Hours Worked The effect of unions on the total number of hours worked by their mem- bers (compared to nonunionized workers) is not a priori clear. On the one hand, unions typically demand lower normal hours, more holidays, and so on. Conversely, they may be able to secure overtime work at higher rates of pay. The union/nonunion hours differential has been extensively stud- ied in the United States and the United Kingdom.32 Overall, the finding is that unions reduce the total number of hours worked. In particular, the evidence suggests that workers in unionized firms work fewer nor- mal hours. Moreover, unions reduce the number of unpaid overtime hours and, in some cases, increase the amount of paid overtime work (Oswald and Walker 1993; Trejo 1993). Furthermore, Green (1995) pro- vides evidence that unions increase the likelihood that workers receive paid holidays and finds that unionized workers get, on average, almost an additional week of holidays compared to nonunionized workers in the United Kingdom. Finally, Blanchflower (1996b) estimates the union/ nonunion total hours differential for 14 OECD countries in 1985-93.33 He confirms the general result that unionized workers work less than nonunionized ones in all but two countries (Spain and Switzerland). The estimates range from a one- to two-hour differential per week in the United States and the United Kingdom to a four- to six-hour differential per week in Austria and Ireland. Unions and Profitability It is a commonly held view that unions reduce the profitability of firms because they appropriate part of the rent that would otherwise have been available to shareholders. The fact that unions are able to get a wage markup supports this view. As pointed out by Clark (1984), how- ever, it is unwise to deduce the effect of unions on profitability by look- ing at the wage markup alone. This is because the union's ability to extract rent from a firm depends on the bargaining power of the union and on the size of the rent. The bargaining power and the size of the rent in turn 32. See Lewis (1986: table 6.5) for a sunmmary of 16 studies for the United States. See also DiNardo (1991), Earle and Pencavel (1990), Oswald and Walker (1993), Perloff and Sickles (1987), and Trejo (1993). 33. They are Australia, Austria, Canada, Germany, Ireland, Italy, Japan, Nether- lands, New Zealand, Norway, Spain, Switzerland, the United Kingdom, and the United States. 68 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECrS IN A GLOBAL ENVIRONMENT depend on a mixture of factors, including the structure of collective bar- gaining, the structure of the product market, the production technology used, and so on. In addition, by improving morale and job satisfaction among workers and by facilitating worker-employer cooperation, unions can contribute positively to profitability, as discussed in chapter 3. There- fore, instead of trying to capture a given rent, unions may help to create profits from which they can achieve wage gains (Filer, Hamermesh, and Rees 1996: 506). A large number of studies have estimated the impact of unions on profitability. These studies use a number of different measures of profit- ability such as price/cost margins, net (of wages) return to capital, Tobin's q,34 and subjective profitability judgments by management and estimate the impact using industry, firm, or stock market data. Bellman (1992) surveys 14 studies from the United States. All find that unions have a negative impact on profitability as measured by one or more of the indi- cators mentioned above. The impact tends to be larger in industries or firms that have some monopoly power in the product market. Some of the evidence suggests that the unions' share of monopoly profits may be as large as between 47 and 77 percent (Karier 1988). While these fig- ures are hardly representative, they do show that under specific circum- stances unions are able to appropriate a substantial share of monopoly profits. Booth (1995: table 7.6) surveys seven studies from the United Kingdom. While a few of them find that unions haves no impact on profitability, the general impression is that unions have a significant negative impact on profitability in British manufacturing firms. This adverse impact is larger when firms have some product market power. A Japanese study (Brunello 1992) finds that unions reduce the rate of return on equity by 20 to 25 percent. The ratio of profits to sales is re- duced by about 40 percent. The effect of unions on profitability seems to be clear: unions reduce profitability. The evidence reviewed above shows that the most pro- nounced union effects are found in industries where firms have mo- nopoly power. This suggests that unions typically share in supernormal profits rather than cutting into normal profits (Reynolds 1986). Productivity Differentials As discussed in chapter 3, unions can contribute positively to labor pro- ductivity by improving work morale, facilitating cooperation with man- 34. Tobin's q is the market value of the firm relative to the replacement cost of the firm's assets. EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 69 agement, reducing grievances (through their "collective voice" function), and so on. These participatory benefits can, however, be countered if unions impede management's ability to adjust to changing economic circumstances or if they impose restrictive practices (such as overstaff- ing or guaranteed overtime). In most empirical studies, productivity is defined as either labor productivity or total factor productivity.35 The union/nonunion pro- ductivity differential is typically estimated from a production function model. Productivity is explained by the input mix (employment, capi- tal, and hours worked), a vector of observed firm and industry charac- teristics (for example, industry concentration), a union dummy variable, and other control variables (such as business cycle indicators or the level of union coverage in the industry). The production function ap- proach has a number of problems. First, measured productivity in unionized firms can be higher than in nonunionized firms without implying that unionized firms are more efficient. This is because the wage markup, other things being equal, reduces employment in union- ized firms. As a consequence, the marginal product of labor would be higher in unionized firms than in nonunionized ones.36 Second, union- ized firms are likely to change their input mix in response to the wage markup. Hence, the input mix cannot be considered an exogenous determinant of productivity, and a simultaneity bias can develop. A third problem arises because management's role is largely ignored. Since the interaction between management and unions affects produc- tivity levels, ignoring management can give a biased view of the im- pact of unions (Denny 1997). This problem is more generally related to unobserved heterogeneity and can best be dealt with by estimating productivity growth models instead of productivity level models. With these methodological issues in mind, we now review studies that esti- mate the productivity level and growth differential. We start with the evidence related to the productivity level differential. 35. Typically, value added per employee or per working hour is used to proxy pro- ductivity. Only a few studies use physical output. The use of value added rather than physical output tends to overestimate the productivity differentials because of price ef- fects. For example, if the wage markup leads to higher product prices, then productivity will appear to increase without unions having any beneficial impact on output. Total fac- tor productivity takes into account that more than one factor of production is used to produce output. More specifically, total factor productivity growth is defined as the growth of output less the weighted sum of the growth of the relevant inputs, where the weights are given by the shares of each input. 36. This argument assumes that the firm is operating at its labor demand curve rather than entering efficient contracts. 70 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFEcTS IN A GLOBAL ENvIRoNMfENT The productivity level differential. Bellman (1992) has surveyed the empirical literature from the United States.37 His conclusion, based on the evidence from 17 studies, is a qualified one, as there is considerable variation across industries. In those industries in which firms are sub- ject to substantial product market competition, unionized firms have higher productivity levels than nonunionized ones. The quality of in- dustrial relations is also important. The "quality" of industrial relations is proxied by the number of grievances filed, the number of unresolved grievances, the number of strikes and quits, and the use of long-term collective agreements. Firms with high-quality industrial relations are associated with higher productivity levels and higher product quality than firms with low-quality industrial relations.38 On the other hand, the significantly higher absenteeism among union workers than among nonunion workers can have a negative impact on productivity; some studies find that absenteeism is 30 percent higher among unionized workers than among nonunionized ones (Allen 1984; Katz, Kochan, and Gobeille 1983). Booth (1995) and Metcalf (1993) have surveyed the evidence from the United Kingdom. Booth tentatively concludes that U.K. unions appear to have a negative impact on the level of productivity, but this conclusion is far from robust. For example, Denny (1997) shows how the productivity effect of unions varies over time. His evidence suggests that British unions had no impact on productivity levels before 1979, but in the Thatcher era in the early 1980s, unions appeared to have a negative impact on produc- tivity. It is noteworthy that the results are affected by how unionism is measured. For example, those studies that use union density as an indica- tor of unionism find a negative productivity effect. However, studies that use strikes as an indicator tend to find positive or insignificant effects.39 This suggests that the adverse impact of unions on the productivity level is not due to industrial conflict. In Japan, unions are enterprise based and concentrated in larger firms, and the attitude of Japanese unions is often viewed as cooperative with management. Hence, Japanese unions seem like an obvious place to look for the "collective voice" effect of unions. Nevertheless, empirical studies 37. See also Addison and Hirsch (1989), Booth (1995: chapter 7), and Filer, Hamermesh, and Rees (1996: chapter 13). 38. It is not obvious that the number of grievances filed is an indicator of high- quality industrial relations. While a large number of grievances filed can indicate that workers are able to express their dissatisfaction within the firm, it can also indicate that something is wrong in the workplace. It is therefore not surprising that Katz, Kochan, and Gobeille (1983) have found that the number of grievances filed in a union workplace nega- tively affects productivity. 39. See Knight (1989) and Moreton (1993). EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 71 from Japan find that unions have mixed effects on productivity. Muramatsu (1984) observed that unions had a positive impact on productivity levels for 1978 when technology and labor-quality variables are held constant, 4 while Brunello (1992) found that productivity in unionized firms was 15 percent lower than in similar nonunionized firms. In Germany, unions appear to have no impact on productivity. This may be related to the fact that all German workplaces have work councils that provide the collec- tive voice function of unions, even in nonunionized firms (Schnabel 1991). Malaysia is the only middle-income economy for which evidence on productivity (level) differentials is available. Standing (1992) uses the value of total sales relative to the total work force to proxy productivity. He finds that unionized firms have higher productivity levels than nonunionized firms and that the positive productivity differential is primarily associated with industrial rather than company unions. While Standing argues that this is prima facie evidence that unions have been associated with dynamic efficiency effects in Malaysia, it is somewhat puzzling why the strongest productivity effects are associated with in- dustrial unions. Although industrial unions typically have shop-level facilities, we would expect that company-based unions would be just as good and perhaps in even a better position than industry-based unions to provide a "voice" and other efficiency-enhancing services. The productivity growth differential. The U.S. evidence on the union/nonunion productivity growth differential has been reviewed by Booth (1995: chapter 7) and Bellman (1992). Bellman finds that in five out of nine cross-sectional studies and in all time-series studies that he reviewed unions decrease productivity growth. In the remain- ing four cross-sectional studies, no significant difference was found between unionized and nonunionized firms. While the available evi- dence indicates that British unions may have a negative impact on pro- ductivity levels, the evidence regarding productivity growth is mixed. Some studies, for example, suggest that unionized firms have higher productivity growth than nonunionized firms during the period 1979- 84 (Nickell, Wadhawani, and Wall 1989). In the 1980s, many firms derecognized unions and repudiated closed shop arrangements. Gregg, Mashin, and Szymanski (1993) consider how these changes in union arrangements affect the productivity growth differential. While they find no difference between productivity growth in unionized and nonunionized firms between 1984 and 1987, firms that experienced a 40. The difference is about 20 percent. However, since net value added per em- ployee is used as a measure of productivity, the difference may reflect the effects of unions on prices through cost-push. 72 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT change in union arrangements between 1987 and 1989 had higher pro- ductivity growth than both unionized firms with constant union ar- rangements and nonunionized firms. These results indicate that the weakening of British unions is one factor that explains the high pro- ductivity growth in the 1980s in the United Kingdom (Booth 1995: 208). Finally, Bean and Symons (1989) estimate a reduced form productivity growth equation for 19 OECD countries for the period 1950-80. Their analysis mainly concludes that union density does not have a statisti- cally significant impact on productivity growth. Unions and Implementation of New Technology Unions' attitude toward new technology (for example, computers and new machinery) is unclear. On the one hand, unions may resist techno- logical changes because they fear immediate short-run employment losses. On the other hand, they may take a long-run view and welcome new technology that increases productivity and the prospect for future increases in wages. The available empirical evidence suggests that new technology is adopted as fast in unionized firms as in nonunionized ones and that unions have little impact on technological innovation in firms. Keefe (1992) surveys research on the relationship between unions and techno- logical change in the United States, the United Kingdom, and Canada in the 1980s. He concludes that "unions have no effect on firms' use of advanced manufacturing and microelectronic technologies" and that "in most cases unions welcome technological modernization; sometimes encouraging it, most often accepting it, infrequently opposing it but usu- ally seeking to protect their members" (Keefe 1992: 110-11). Betcherman (1991) reaches a similar conclusion in his study of the impact of unions on technological change in Canada in 1980-85 but observes that unions do have an impact on the way in which technological change is imple- mented. In particular, he finds that unionized firms were more likely to introduce technological changes than nonunionized firms for cost- cutting or production control reasons. Likewise, Daniels (1987), Latreille (1992), and Machin and Wadhwani (1991) find that unions had a small positive impact on the introduction of new microelectronic equipment in U.K. firms in the mid-1980s. Finally, Standing (1992), in his study of industrial relations in Malaysia, concludes that unions actually stimu- late capital, product, and labor process innovations. Unions, Physical Investments, and R&D The reviewed evidence on the wage markup and the effect of unions on profitability shows that unions share rents with firms. Besides the static EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 73 impact on the functional distribution of income, this can have signifi- cant dynamic efficiency effects. These arise when firms realize that work- ers are going to appropriate part of the profits associated with investments in physical capital and R&D. Consequently, a unionized firm can be expected to invest less than a similar firm operating in a competitive labor market because of the resulting "hold-up problem" (Grout 1984; Ulph and Ulph 1990). A handful of studies have looked into the issue of under-investment by unionized firms, using firm- or industry-level data from the United States and the United Kingdom. For the United States, Bronars and Deere (1986) and Hirsch (1990) and for the United Kingdom, Denny and Nickell (1991) find that unionization has a negative impact on investment in physical capital. For example, Denny and Nickell (1991) find that, hold- ing wages and productivity constant, the rate of investment in firms that recognize a union and have an average manual union density is, on average, 23 percent lower than in other firms. Furthermore, distinguish- ing between competitive and noncompetitive sectors and taking second round wage effects into account, they find that the net reduction in the investment rate is 13 percent for a competitive, unionized firm but only about 4 percent for a noncompetitive, unionized one.41 - This result is surprising in suggesting that product market competition has an ad- verse impact on the behavior of unions. Likewise, the available evidence suggests that unionization can re- duce spending on R&D (Acs and Audretsch 1987; Connolly, Hirsch, and Hirschey 1986; Ulph and Ulph 1989). Van Reenen (1993) estimates that firm-level innovations in unionized British manufacturing firms are asso- ciated with higher wages for up to seven years. This suggests that unions do share in the surplus from innovation and may explain why the spend- ing on R&D is lower in unionized firms than in nonunionized ones. Unions, Fringe Benefits, and Health and Safety Regulations Unions do significantly increase wages. While this can be interpreted as evidence that unionized workers earn substantial rents, Duncan and Strafford (1980) argue that as much as two-fifths of the wage markup is compensation for an inflexible and employer-controlled work environ- ment. In addition to their monthly paycheck, however, unionized work- ers may be concerned with other issues, such as bonuses, severance pay, health and safety regulations, and paid sick leave. 41. The distinction between competitive and noncompetitive industries is based on a workplace survey in which firms are asked to indicate if the market they are operating in (a) is dominated by the main supplier, (b) has only a few compc-itors, or (c) has many competitors. 74 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFEcTS IN A GLOBAL ENVIRONMENT The evidence suggests that workers in unionized firms are more likely to receive these benefits than are workers in nonunionized firms. For instance, in Malaysia, Standing (1992) finds that firms with unions are more likely to provide paid sick leave, retirement benefits, cheap loans, and transportation. Green, Hadjimatheou, and Smail (1985) find that unions increase the likelihood of improved health and safety mea- sures in the United Kingdom. Nakumura, Sato, and Kamiya (1988) find that unions in Japan increase the use of severance pay and the size of the yearly bonus. Some of these benefits obviously contribute to increasing labor costs in general and turnover costs in particular. On the other hand, cheap loans, free transportation, paid sick leave, and safety regulations may improve worker motivation and pay off in terms of higher productivity. Moreover, to the extent that inadequate safety and health provisions generate a sub- optimal allocation of labor, union-sponsored (as well as government- sponsored) safety and health regulations increase not only individual worker's welfare but also aggregate welfare. Maskus, Rutherford, and Selby (1995) consider this issue of disclosure of information in the context of the risk attributes of different jobs. Commodities produced in different sectors use production technologies that expose workers to different levels of physical risks, such as exposure to toxins and industrial accidents. Safety is desirable from the point of view of workers. Therefore, safety has an opportunity cost, and safer jobs pay a lower wage. In an unregulated labor market, workers may be unable to appreciate the dangers inherent in dif- ferent jobs. As a consequence, firms would not be required to compensate workers fully for the hidden risks involved in their jobs. This would lead to an inefficient allocation of labor across sectors, with too many workers doing jobs that are too dangerous. A labor market reform that induces full disclosure of safety levels would remove this distortion. Maskus, Ruther- ford, and Selby (1995: table 2) estimate the welfare effects of a labor market reform of this type using a CGE model of the Mexican economy. They find that the well-being of workers would increase by 0.5 percent of baseline GDP per year. Moreover, the real income of the owners of the firms would increase as well because the reform increases the demand for capital for risk-abatement purposes. The total gain is estimated to be 0.6 percent of baseline GDP per year. This is a substantial gain and is of the same order of magnitude as the estimated monopoly cost of unions. Individual Performance Pay and Seniority Freeman and Medoff (1984) find that individual performance pay is much more prevalent among nonunionized firms in the United States than among unionized firms by as much as 16 to 23 percentage points. EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 75 Similar results are found in Britain. Blanchflower and Oswald (1988) use a questionnaire to identity the factors that influence the level of pay in wage settlements in the private sector in Britain. They find that the most significant difference between unionized and nonunionized firms is that individual performance is important in wage determina- tion only in nonunionized firms. This can have an adverse effect on productivity if individual perfor- mance pay is used as an incentive to increase workers' efforts. How- ever, there are counterarguments. For example, the fact that unions are able to reduce the use of individual performance pay could be seen as evidence that the presence of unions reduces the need for this control instrument. Seniority-based wages can be interpreted as an efficiency wage that is designed to motivate workers to stay with the same firm for a longer period of time. Therefore, unions can increase productivity by extending seniority-based systems to smaller firms.42 Unions and Pensions Evidence from Canada, Malaysia, and the United States shows that unions increase the likelihood that workers are enrolled in pension schemes.43 As pointed out by Freeman (1985), if union-sponsored pen- sion plans do not replace private saving, national saving can increase. At the macroeconomic level, the implied reduction in the real interest rate will increase investment demand and may even have a (temporary) impact on economic growth. Conclusion In chapter 3, we argued that the net cosf of unions has three compo- nents: the monopoly cost, the participatory benefits, and the rent- seeking cost. To conclude the discussion of the microeconomic effects of unions, we integrate the evidence surveyed in the previous sections to make an overall evaluation of the monopoly costs and the participatory benefits of unions. A detailed summary of the results can be found in the chapter 1. Our examination of the empirical evidence has little to add to our theoretical discussion of unions as rent-seekers (see chapter 3) and so this (potentially important) aspect of union behavior is not subject to separate consideration below. 42. Large firms tend to have seniority-based wage systems largely irrespective of unionization (Brown 1990: 178). 43. See Freeman (1985), Kupferschmidt and Swidensky (1989), and Standing (1992). 76 UNIONS AND COLLECnVE BARGAWING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT The Monopoly Cost of Unions The evidence on the wage markup shows that unions everywhere are successful in raising wages for their members and others covered by col- lective agreements. As pointed out by Rees (1963), this creates a misallo- cation of resources. However, attempts to quantify the size of this deadweight loss have consistently shown that it is relatively small, no larger than 0.5 percent of GDP per year. It should be kept in mind, how- ever, that these estimates do not include all the potential costs of unions. While they do take into account that the level of employment is likely to be reduced, they do not include the costs associated with a reduction in employment growth, nor do they indude rent-seeking costs. The estimates also disregard a number of potential dynamic costs of unionism. First, the evidence reviewed above shows that unions compress the reward to skill accumulation. This, other things being equal, reduces workers' incentives to engage in training and education. Second, the evidence reviewed above also suggests that unions can have an adverse impact on firms' incentive to invest in physical capital and R&D, although there is no evidence that unions reduce the speed at which new technologies are adopted. Theoretical considerations suggest that firms' exposure to compe- tition from product markets and nonunion labor markets can poten- tially help to reduce the monopoly cost of unions. Here we have found some evidence to support this view, but it is far from conclusive. First, the measured impact of product market competition on the wage markup is sensitive to the method used to measure the degree of prod- uct market competition. When industry concentration indicators are used, the markup is typically found to be relatively small for unionized workers employed in industries with high concentration. Arguably, however, industry concentration is not a good indicator of the competitive situation facing individual firms, and other studies, which use firm-specific indicators of competitive pressure, find that competitive pressure from product markets does reduce the wage markup. In addition, the evidence shows that the adverse impact of unions on profitability is relatively small when firms have little rmo- nopoly power in the product market. There are good theoretical reasons to believe that the monopoly cost is systematically related to the institutional framework in which collec- tive bargaining takes place. The empirical evidence is, however, mixed. Some studies show that the wage markup is small when bargaining is conducted at the firm level rather than at the industry level but the dif- ferences are not large, and some studies from the United Kingdom are unable to detect any differences. Nevertheless, it is possible to identify particular aspects of collective bargaining, such as the combination of EMPIRICAL EVIDENCE FROM MICROECONOMIC STUDIES 77 multiunionism and separate collective bargaining and pre-entry closed shops, that definitely add to the wage markup as well as other aspects, such as post-entry closed shops, that do not. Participatory Benefits of Unions We would expect the participatory benefits of unions to show up as pro- ductivity differences between unionized and nonunionized firms. The evidence shows that unions have little impact on productivity levels and growth on average. Under special circumstances, however, unions can have a positive impact. In firms where industrial relations are of a "high" quality (in terms of a low number of unsolved grievances, low strike activity, and so on), the presence of unions tend to increase pro- ductivity levels. The same is true in firms that are operating in a com- petitive product market environment. An important source of participatory benefits is a reduction in labor turnover. Here the evidence shows that turnover is significantly lower in unionized firms than else- where. The gain associated with this reduction is estimated to be 0.2 to 0.3 percent of GDP. This gain is on the same order of magnitude as the estimated deadweight loss associated with the wage markup. In addi- tion, by making it more profitable for workers to engage in firm-specific training and for firms to fund general training programs, the observed reduction in turnover can explain why union members are more likely to be trained than other workers. In condusion, the costs and benefits of unions depend on the economic (and political) environment in which unions and employers interact. It is crucial to keep this in mind when contemplating policy reforms. 5 Empirical Evidence from Macroeconomic Studies The labor market has important influences on the functioning of the macroeconomy. This suggests that different institutional approaches to collective bargaining can lead to very different macroeconomic outcomes, and that some institutional arrangements might be more appropriate to achieve desirable outcomes than others. A central idea is that collective bargaining facilitates coordination. Coordination can take make forms. For example, the Japanese system of wage setting is decentralized (firm based) but coordinated in the sense that it follows company rules based on seniority rather than individual contracting. The Netherlands and Germany also have coordinated systems through strong employer orga- nizations, coordination between giant companies or across industries, and between unions. Coordination in France is through the government in the form of public services, utilities, and large nationalized industries. In Italy, there is informal employer coordination (via the big firms and re- gional employers' associations) and between some union confederations. Finally, centralized employers' organizations as well as centralized union confederations have dominated Sweden and more generally Scandina- vian labor markets. It is clear from these examples that the specific institu- tions and the extent to which pay and work conditions are determined by collective agreements as oppose to individual contracts differ quite a lot across the OECD. These differences combined with the observed differ- ences in macroeconomic performance (primarily in terms of unemploy- ment and inflation) between the OECD countries over the last 30 years has spurred a large literature that tries to explain cross-country variation in economic performance by cross-country differences in labor market institutions. The purpose of this chapter is to examine this literature in detail and ask what can be learned about the impact of collective bargain- ing on macroeconomic performance from it. 79 80 UNIONS AND COLLECTIVE BARGAINING: ECONOMIC EFFECTS IN A GLOBAL ENVIRONMENT Conceptual Issues From a normative perspective, we would like to design labor market in- stitutions that in some sense maximize social welfare subject to economic and non-economic constraints (where noneconomic constraints may in- clude geographical, historical, cultural, political, social, and religious ones): Max (welfare) Subject to (constraints) Since welfare is not observed, economists typically think of maxi- mizing some aggregate economic outcome that captures aspects of so- cial welfare, such as the output level or employment. The aggregate outcome (Y) relates to the n suboutcomes (y) in some general form: (1) Y =f(y,, Y -,- Yn) Although conceptually clear, the measurement of the aggregate eco- nomic outcome requires that we know how to aggregate as well as mea- sure the suboutcomes. One way to solve the aggregation problem is to impose some assumption such as (2) Y.Y,+y2+...+y. There is, however, no reason why suboutcomes should be given equal importance (or measured in the same units). They could be given differ- ent weights. Unfortunately, the weights usually have to be chosen arbi- trarily, and in any case it is also possible that a multiplicative rather than an additive relationship is more appropriate. Mainly due to all these diffi- culties, the prevalent approach in the literature is to assess the impact of the institutional framework of collective bargaining on economic perfor- mance by looking at one sub-outcome as measured by an economic indi- cator at the time, and then try to aggregate (often implicitly) the suboutcomes in a meaningful way. Measuring Labor Market Institutions and Economic Performance To evaluate the impact of collective bargaining on macroeconomic per- formance in a comparative context, aspects of the relevant labor market institutions have to be measured empirically and examined against in- dicators of macroeconomic performance. Measuring differences in la- bor market institutions over time and space in a consistent and meaningful way is perhaps the greatest obstacle for comparative study, and we shall devote considerable attention to this issue in what follows. EMPIRICAL EVIDENCE FROM MACROECONOMIC STUDIES 81 Institutional Indicators The comparative literature attempts to measure cross-country differences in labor market institutions by looking at a few crude indicators. The simplest way to capture the importance of collective bargaining in an economy is to measure the proportion of the economy in which pay and employment conditions are determined by collective bargaining between employers and employees rather than by individual contracts. In practice, union density and bargaining coverage are used to capture this aspect of collective bargaining (see table 5-1). These two indicators go some way in measuring the "importance" of collective agreements as opposed to individual contracts, but they can hardly be seen as indica- tors of union power. In particular, it should be kept in mind that there can be substantial spillover effects from unionized/covered sectors to nonunionized/uncovered sectors that are not captured by the two indi- cators. For instance, firms in noncovered sectors may set wages at the collectively agreed level to avoid being subject to other effects of union- ization or to motivate their workers who may be concerned about rela- tive wages (Mazumdar 1993; Pencavel 1991). Table 5-2 shows data on union density and bargaining coverage for 19 OECD countries for the period 1970-94. Average union density increased from 43 percent to 47 percent during the 1970s but declined to 40 percent during the 1980s and 1990s. However, the average hides a lot of variation. Some countries, such as Japan, the Netherlands, the United Kingdom, and the United States, have experienced a significant reduction in union density. Other countries, such as Finland and Sweden, have encountered Table 5-1. Definitions of Union Density and Bargaining Coverage Term Definition Union density The number of workers who are members of a union, as a percentage of all workers, unionized and nonunionized. Bargaining The number of workers, unionized or not, whose pay coverage and employment conditions are determined by a collective agreement, as a percentage of all workers, unionized and nonunionized. Note: Depending on the study, the term "all workers" refers to all wage and salary workers (employees) or total labor force (employees plus self-employed, family workers, and so on). Source: OECD (1997). 82 UNIONS AND COLLECTIVE BARGA