THE ECONOMICS OF IMMIGRATION AND SOCIAL DIVERSITY
i
RESEARCH IN LABOR ECONOMICS Series Editor: Solomon W. Polachek
ii
RESEARCH IN LABOR ECONOMICS VOLUME 24
THE ECONOMICS OF IMMIGRATION AND SOCIAL DIVERSITY EDITED BY
SOLOMON W. POLACHEK Department of Economics, State University of New York at Binghamton
CARMEL CHISWICK Department of Economics, University of Illinois at Chicago
HILLEL RAPOPORT Department of Economics, Bar-Ilan University
Amsterdam – Boston – Heidelberg – London – New York – Oxford Paris – San Diego – San Francisco – Singapore – Sydney – Tokyo iii
ELSEVIER B.V. Radarweg 29 P.O. Box 211 1000 AE Amsterdam, The Netherlands
ELSEVIER Inc. 525 B Street, Suite 1900 San Diego CA 92101-4495 USA
ELSEVIER Ltd The Boulevard, Langford Lane, Kidlington Oxford OX5 1GB UK
ELSEVIER Ltd 84 Theobalds Road London WC1X 8RR UK
r 2006 Elsevier Ltd. All rights reserved. This work is protected under copyright by Elsevier Ltd, and the following terms and conditions apply to its use: Photocopying Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying, copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are available for educational institutions that wish to make photocopies for non-profit educational classroom use. Permissions may be sought directly from Elsevier’s Rights Department in Oxford, UK: phone (+44) 1865 843830, fax (+44) 1865 853333, e-mail:
[email protected]. Requests may also be completed on-line via the Elsevier homepage (http://www.elsevier.com/locate/permissions). In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UK through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London W1P 0LP, UK; phone: (+44) 20 7631 5555; fax: (+44) 20 7631 5500. Other countries may have a local reprographic rights agency for payments. Derivative Works Tables of contents may be reproduced for internal circulation, but permission of the Publisher is required for external resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including compilations and translations. Electronic Storage or Usage Permission of the Publisher is required to store or use electronically any material contained in this work, including any chapter or part of a chapter. Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission of the Publisher. Address permissions requests to: Elsevier’s Rights Department, at the fax and e-mail addresses noted above. Notice No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made. First edition 2006 British Library Cataloguing in Publication Data A catalogue record is available from the British Library. ISBN-10: 0-7623-1275-0 ISBN-13: 978-0-7623-1275-7 ISSN: 0147-9121 (Series)
∞ The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper). Printed in The Netherlands.
Working together to grow libraries in developing countries www.elsevier.com | www.bookaid.org | www.sabre.org
iv
CONTENTS LIST OF CONTRIBUTORS
ix
PREFACE Carmel Chiswick, Solomon W. Polachek and Hillel Rapoport
xiii
PART I: IMMIGRATION AND THE ECONOMY IS IMMIGRATION GOOD OR BAD FOR THE ECONOMY? ANALYSIS OF ATTITUDINAL RESPONSES Christian Dustmann and Ian Preston
3
THE EFFECTS OF INCOMPLETE EMPLOYEE WAGE INFORMATION: A CROSS-COUNTRY ANALYSIS Solomon W. Polachek and Jun (Jeff) Xiang
35
THE IMPACT OF IMMIGRANT DYNASTIES ON WAGE INEQUALITY Michael Ben-Gad
77
DOES IMMIGRATION AFFECT LABOR DEMAND? MODEL AND TEST O¨rn B. Bodvarsson and Hendrik Van den Berg
135
LENIENT POLICY PROPOSAL FOR THE STRUGGLE AGAINST ILLEGAL IMMIGRATION Nava Kahana and Tikva Lecker
167
v
vi
CONTENTS
PART II: GROUP DIFFERENCES AND ECONOMIC ACHIEVEMENT THE LINGUISTIC AND ECONOMIC ADJUSTMENT OF SOVIET JEWISH IMMIGRANTS IN THE UNITED STATES, 1980–2000 Barry R. Chiswick and Michael Wenz
179
MULTI-GENERATION MODEL OF IMMIGRANT EARNINGS: THEORY AND APPLICATION Joseph Deutsch, Gil S. Epstein and Tikva Lecker
217
ETHNIC ORIGIN AND MULTIDIMENSIONAL RELATIVE POVERTY IN ISRAEL: A STUDY BASED ON THE 1995 ISRAELI CENSUS Joseph Deutsch and Jacques Silber
235
IMMIGRANTS IN THE ISRAELI HI-TECH INDUSTRY: COMPARISON TO NATIVES AND THE EFFECT OF TRAINING Sarit Cohen-Goldner
265
WHAT DO WAGE DIFFERENTIALS TELL ABOUT LABOR MARKET DISCRIMINATION? June E. O’Neill and Dave M. O’Neill
293
PART III: SOCIAL DIVERSITY AND INSTITUTIONS CULTURAL DIVERSITY, STATUS CONCERNS AND THE ORGANIZATION OF WORK Chaim Fershtman, Hans K. Hvide and Yoram Weiss
361
Contents
vii
ETHNIC DIVERSITY, MARKET STRUCTURE AND RISK SHARING IN DEVELOPING COUNTRIES Mohamed Jellal and Yves Zenou
397
ON THE LAW OF RETURN IN RURAL–URBAN INTERACTIONS: AN ECONOMIC APPROACH TO SOLIDARITY WITH RETURN MIGRANTS Carine Drapier, Hubert Jayet and Hillel Rapoport
427
AN ECONOMIC PERSPECTIVE ON RELIGIOUS EDUCATION: COMPLEMENTS AND SUBSTITUTES IN A HUMAN CAPITAL PORTFOLIO Carmel U. Chiswick
449
This page intentionally left blank
viii
LIST OF CONTRIBUTORS Michael Ben-Gad
Department of Economics, University of Haifa, Israel
O¨rn B. Bodvarsson
Department of Economics, St. Cloud State University, MN, USA
Barry R. Chiswick
Department of Economics, University of Illinois at Chicago, IL, USA IZA, Bonn, Germany
Carmel U. Chiswick
Department of Economics, University of Illinois at Chicago, IL, USA IZA, Bonn, Germany
Sarit Cohen-Goldner
Department of Economics, Bar-Ilan University, Israel
Joseph Deutsch
Department of Economics, Bar-Ilan University, Israel
Carine Drapier
CADRE, Universite´ de Lille, France
Christian Dustmann
Department of Economics and CReAM, University College London, UK IZA, Bonn, Germany
Gil S. Epstein
Department of Economics, Bar-Ilan University, Israel IZA, Bonn, Germany
Chaim Fershtman
The Eitan Berglas School of Economics, Tel-Aviv University, Israel
Hans K. Hvide
Norwegian School of Economics and Business, Bergen, Norway ix
x
LIST OF CONTRIBUTORS
Hubert Jayet
MEDEE, Universite´ de Lille, France
Mohamed Jellal
CES, Universite´ Mohamed V, Morocco, ConseilsEco, Toulouse, France
Nava Kahana
Department of Economics, Bar-Ilan University, Israel
Tikva Lecker
Department of Economics, Bar-Ilan University, Israel
Dave M. O’Neill
Center for the Study of Business and Government, Zicklin School of Business, Baruch College, NY, USA
June E. O’Neill
Department of Economics and Finance, Baruch College, City University of New York, New York, USA
Solomon W. Polachek
Departments of Economics and Political Science, State University of New York at Binghamton, NY, USA IZA, Bonn, Germany
Ian Preston
Department of Economics and CReAM, University College London, UK Institute for Fiscal Studies, UK
Hillel Rapoport
Department of Economics, Bar-Ilan University, Israel CADRE, Universite´ de Lille, France
Jacques Silber
Department of Economics, Bar-Ilan University, Israel
Hendrik Van den Berg
Department of Economics, University of Nebraska, NE, USA
Yoram Weiss
The Eitan Berglas School of Economics, Tel-Aviv University, Israel
Michael Wenz
Department of Economics, University of Illinois at Chicago, IL, USA
xi
List of contributors
Jun (Jeff) Xiang
Department of Political Science, University of Rochester, NY, USA
Yves Zenou
IUI, Stockolm, Sweden GAINS, Universite´ du Maine, France
This page intentionally left blank
xii
PREFACE This volume is a collection of papers first presented at a conference held in June 2004 dedicated to the memory of the late Tikva Lecker, hosted by BarIlan University in Ramat Gan, Israel, and co-sponsored by the University of Illinois at Chicago. A warm and lively member of the Department of Economics at Bar Ilan University, Professor Lecker’s many interests included topics in labor economics, women and the economy, the economics of Judaism, the economics of migration, and every aspect of the economic experience of immigrants and their descendants. Each chapter in this volume honors the memory of Professor Lecker by presenting research on a topic in which she was especially interested. The papers in Part I: Immigration and the Economy look at immigration from the perspective of the receiving country. They include examples of both positive and normative economics, concerned with estimating the impacts of immigrants and also with the characteristics of an optimal immigration policy. Part II: Group Differences and Economic Achievement contains papers that look at various aspects of income distribution. Some of these compare the earnings and other socioeconomic characteristics of immigrants to the United States and Israel with that of natives in the same country, and others look at differences by race, ethnicity, national origin, and gender. The papers in Part III: Social Diversity and Institutions consider some economic aspects of ethnic pluralism and of institutions that help support various aspects of cultural diversity. The initial five chapters deal with the various channels through which immigration impacts the economy. In ‘‘Is Immigration Good or Bad for the Economy? Analysis of Attitudinal Responses,’’ Christian Dustmann and Ian Preston look at public attitudes toward immigrants. Although often overlooked in economics, attitudes are important because they often reveal how citizens vote, which in turn affects legislation. By examining the European Social Survey (ESS) in conjunction with a simple equilibrium model, they are able to assess public opinion in three areas: the labor market, the public burden, and overall economic efficiency. In general, they find self-interest to be important. Individuals who strongly benefit from migration have positive attitudes xiii
xiv
PREFACE
toward migration. Thus former immigrants, their children, and the moreeducated view immigration as beneficial; whereas the unemployed and the less educated perceive migration as detrimental. Minorities fear economic losses, but perceive immigration to help fulfill the tax burden and fill jobs in areas where there are shortages. Polachek and Xiang take an econometric approach. In ‘‘The Effects of Incomplete Employee Wage Information: A Cross-Country Analysis,’’ they define a tractable procedure to measure worker incomplete information about available wages. Their procedure, which makes use of earnings distribution skewness, is based on frontier estimation techniques, and is consistent with search theory. They apply the technique to 11 countries over various years, and find that incomplete information leads workers to receive on an average about 30–35% less pay than they otherwise would have earned, had they information on what each firm paid. Generally, married men and women suffer less from incomplete information than the widowed or divorced; and singles suffer the most. Women suffer more from incomplete information than men. Schooling and labor market experience reduce these losses, but institutions within a country can reduce them, as well. For example, they find that workers in countries that strongly support unemployment insurance (UI) receive wages closer to their potential, so that doubling UI decreases incomplete information and results in 5% higher wages. Relating the technique to immigration, they find that foreign worker inflows increase incomplete information, and at the same time reduce average wage levels, at least in the short-run. In ‘‘The Impact of Immigrant Dynasties on Wage Inequality,’’ Michael Ben-Gad concentrates on migration of unskilled workers to the US. He notes there is a disproportionate share of unskilled workers among certain immigrant groups. The key question he asks is: Did these immigrants likely cause wage inequality to increase? His answer is no. To get there, Ben-Gad develops an overlapping dynasties optimal growth model. Changes in immigration affect wages directly but also alter rates of return, which as a consequence causes individuals to modify their own and future generations’ investments in human capital. Indeed, Ben-Gad finds that the native-born children and grandchildren of immigrants are often more likely to exceed educational levels of their parents. By using US Current Population Survey data, he demonstrates that the arrival of a disproportionate share of unskilled workers among certain immigrant groups did not cause wage inequality. Many current studies claim that immigration has small if any effects on the local economy, even when immigration flows are large. Paradoxically,
Preface
xv
instead of generating unemployment and driving down wages, these studies find the enhanced labor supply to have almost no detrimental effects on the local labor market. In ‘‘Does Immigration Affect Labor Demand? Model and Test,’’ O¨rn B. Bodvarsson and Hendrik Van den Berg cogently argue that these surprising results may occur because all current studies neglect an important macroeconomic consideration: migrants stimulate local labor markets via their own consumption. To test this hypothesis, Bodvarsson and Van den Berg examine a unique case study: a large export-driven meatpacking plant that attracted many Hispanic immigrants to Dawson County, Nebraska. These immigrants spent their earnings in the local economy, but language barriers precluded them from obtaining sales and retail jobs in the local labor market. As a result, wages in the retail sector rose and housing prices increased to the tune of 4–5%, confirming that immigration is capable of exerting significant effects on local labor demand. Policies that stimulate immigration can thus be effective in invigorating a local economy. Whereas there are benefits to legal migration, many believe the costs of illegal migration to be inordinate. As such, governments seek to deter illegal immigration. But how should a country go about preventing such illegal migration? The conventional wisdom is to raise the probability of apprehension by beefing up border patrols and increasing the punishment, thus reducing expected wages of undocumented workers. In the next chapter, Nava Kahana and Tikva Lecker suggest an alternative strategy. They argue in favor of supplementing border and domestic control with a more lenient approach to illegal immigrants who identify themselves to the immigration authority. Whereas such a policy would reduce the cost of apprehension to the immigrant thereby increasing the incentive for illegal immigration, it would on the other hand boost the number of self-reporting immigrants leaving voluntarily. Depending on the parameters of the expected utility model, the number of remaining illegal immigrants could actually decrease under this plan. Moreover, they argue that the self-reporting immigrants leaving the host countries would be predominantly from the relatively lower socioeconomic groups. The next five chapters deal with differences in economic success, viewed either as achievement in the labor force or as income and consumption, between different groups of the population. In ‘‘Immigrants in the Israeli Hi-Tech Industry: Comparison to Natives and the Effect of Training,’’ Sarit Cohen-Goldner observes that the arrival and assimilation in Israel of an exceptionally large wave of immigrants from the former Soviet Union (FSU) coincided with the growth of a new technologically based family of industries. She looks at whether there was a
xvi
PREFACE
causal link between the two, i.e. whether rapid growth in the high-tech sector was a consequence of skills formed in the former Soviet Union. Her findings suggest not, although immigrants from the FSU clearly benefited from rapid economic growth, since their rate of participation in high-tech employment was similar to that of Israeli natives and depended on postmigration human capital investments provided by training programs in Israel. In ‘‘The Linguistic and Economic Adjustment of Soviet Jewish Immigrants in the United States, 1980 to 2000’’ Barry R. Chiswick and Michael Wenz investigate differences in language proficiency as well as the economic integration of immigrant men from different countries and regions. This paper focuses attention on the (mostly Jewish) immigrants from the former Soviet Union. Their findings based on the 2000 US Census of population are consistent with those found in Chiswick’s studies of earlier waves of Soviet immigrants. FSU immigrants arrive with lower English proficiency that improves more quickly over time and rapidly obtains parity with other immigrant groups, a pattern that is typical of refugee populations in general. FSU immigrants to the US also have lower initial earnings than other groups but more rapid improvement over time, reaching earnings parity with other European immigrants in a few years. As is the case for Jews born in the United States, the effect of schooling on earnings is somewhat larger for the Soviet Jews than it is for other immigrant groups, ceteris paribus. Thus despite their disadvantage as a refugee population, Soviet Jews have had a highly successful labor market adjustment in the United States. In ‘‘Multi-Generation Model of Immigrant Earnings: Theory and Application,’’ Joseph Deutsch, Gil Epstein, and Tickva Lecker use data from the 1995 Israeli Census to analyze differences in the schooling and earnings of first, second, and third-generation immigrant men relative to each other and to native-born Israelis. As expected, they find that the sons of immigrants have a higher level of schooling and earnings than their immigrant fathers. In contrast to previous studies that did not distinguish between the grandchildren of immigrants and other natives, this paper finds that the third generation does less well than their fathers and their counterparts with grandparents born in Israel. The paper develops a model that attributes this phenomenon to sons motivated by the difficulties faced by their immigrant fathers to achieve higher levels of schooling and economic success. The converse applies to the sons of the high-achieving second-generation, who are less motivated to work hard than either their fathers or grandfathers. Another approach to measuring the intergenerational absorption of immigrants is taken by Joseph Deutsch and Jacques Silber in ‘‘Ethnic Origin
Preface
xvii
and Multidimensional Relative Poverty in Israel: A Study Based on the 1995 Israeli Census.’’ This paper uses living standards rather than work as the yardstick for economic integration, and measures success in terms of consumer durables rather than income. Several different indices of poverty are considered, and the results are used to compare the standard of living for households headed by immigrants from Asia and Africa with that of households headed by immigrants from Europe and North America, using households with Israeli-born heads as a benchmark. Because of the overwhelming importance of the foreign born in Israel’s population, the definition of immigrant generation is modified by treating people who arrived in Israel as young children as similar to native-born Israelis. ‘‘What do Wage Differentials Tell Us about Labor Market Discrimination’’ by June E. O’Neill and Dave M. O’Neill looks at earnings differences in the United States by gender, race, and ethnicity. They use the 2000 US Census of Population to look at differences for ethnic groups, estimating equations separately for men and for women. They also use data for persons aged 35–43 from the 2000 National Longitudinal Survey of Youth (NLSY), which permits them to measure skill levels with Armed Forces Qualifying Test (AFQT) scores and lifetime work histories instead of the usual variables of schooling and age-related experience. The NLSY analysis compares earnings between African-Americans, Hispanics, and Non-Hispanic Whites as well as between men and women. As in studies from earlier data sets, controlling for the human capital determinants of earnings diminishes the size of the between-group earnings differentials. The differentials persist in the estimates from the Census data, which relies on the conventional variables, but virtually disappear in the NLSY analysis with its more sensitive controls for the human capital workers bring to the labor market. The last four papers of this volume focus on social diversity and institutions and are also strongly connected to migration and minorities issues. In their paper on ‘‘Cultural Diversity, Status Concern and the Organization of Work,’’ Chaim Fershtman, Hans K. Hvide, and Yoram Weiss investigate the impact of diversity in preferences for social status on the labor market equilibrium and the organization of work within the firm. In a model where workers’ effort is endogenous and there are only two types of workers (those who care about social status and those who don’t), they first find that mixing workers with different preferences and offering different contracts to different types of workers is optimal from the firm’s perspective. They then explore how diversity in status concern (measured by the degree of concern of those who care) affects a number of outcomes of interest such as wages, total output, and the distribution of effort, with more diversity
xviii
PREFACE
generally bringing in more efficient outcomes as diversity acts as an effort extraction device. Finally, they apply their model to immigration, with particular reference to the recent experience of Israel. The discussion reinterprets the fall and rise of Russian immigrants’ socioeconomic status in the light of the theory. Among other things, the authors emphasize that their model contributes to explaining why as skilled immigrants start accessing better jobs, natives react by putting more effort into training and production activities, which in the end preserves the wage gap between natives and immigrants in spite of the latter assimilation and integration. In suggesting a new channel through which immigration can have a positive effect on wages, this also goes part of the way toward explaining why most empirical studies on the local labor market impact of immigration concluded a neutral or at the worst a weakly negative effect. Social interactions are also at the heart of Mohamed Jellal and Yves Zenou’s paper on ‘‘Ethnic Diversity, Market Structure and Risk Sharing in Developing Countries.’’ The paper – an extension of Salop’s model of spatial competition – explores how ethnic diversity impacts on rural labor markets in developing countries, an important yet neglected issue in the rural labor contracts literature. The model derives interesting predictions about the circumstances under which a fixed-wage or a piece rate contract is optimal (and thus likely to prevail) or, instead, under which circumstances the two types of contracts should (are likely to) coexist. It shows theoretically that in equilibrium ethnic diversity increases the piece-rate component of the contract while ethnic interaction costs increase the fixed-wage component, providing that there is coexistence between the two methods of pay. These and other results relating the type of contract to employers and employees degrees of risk-aversion are then shown to be supported by empirical evidence from selected empirical studies. In the same context of traditional agrarian societies, Carine Drapier, Huber Jayet, and Hillel Rapoport investigate the economic foundations of community solidarity with return migrants, an institution very common in the rural areas of developing countries. Their paper, entitled ‘‘On the Law of Return in Rural–Urban Interactions: An Economic Approach to Solidarity with Return Migrants,’’ first documents the phenomenon and then shows that in a context of decreasing marginal productivity of labor, such solidarity brings about efficiency gains beyond those of mutual insurance emphasized in previous literature. The authors then show that an equal sharing rule is Pareto-optimal but raises time-consistency problems that can be overcome in an intergenerational setting with parental altruism. Finally, they show that the degree of altruism required for cooperation to prevail is
Preface
xix
proportional to the optimal migration rate; in other words, changes that make migration socially more desirable also put at risk the institution of solidarity with return migrants. The last paper of this volume is Carmel Chiswick’s paper on ‘‘An Economic Perspective on Religious Education: Complements and Substitutes in a Human Capital Portfolio.’’ It builds on a simple model of investment in two forms of human capital (one specifically religious and the other general), with the degree of complementarity between the two determining whether individual choices will lead to polarization or integration within the religious community and between the religious community and the larger society. The model is applied to perceived hostility to religion in American public education and the coping strategies of fundamentalist Protestants and Jews, two religious groups with potentially large negative complementarities between the sacred and secular education. Some implications of educational choices of individuals for the religious community are also discussed, including its distinctiveness, its internal cohesion, and its success at intergenerational transmission of religious beliefs and behaviors. For insightful editorial advice in preparing this volume, we thank Gary Anderson, Heather Antecol, Emmanuelle Auriol, Eli Berman, Sarit CohenGoldner, Deborah Cobb-Clark, Augustin De Coulon, Carlo Devillanova, Don DeVoretz, Philippe Devreyer, Slobodan Djajic, Frederic Docquier, Christian Dustmann, Gil Epstein, Arthur Fishman, Ira Gang, Astrid Kunze, Sylvie Lambert, Anna Mayda, Prachi Mishra, Tobias Mu¨ller, Peter Norman, June O’Neill, Fabien Postel-Vinay, William Sander, Helena SkytNielsen, Ian Smith, Anne Winkler, and Madeline Zavodny. Beginning with the next volume the Institute for the Study of Labor (IZA) will cosponsor Research in Labor Economics (RLE). At that time, Olivier Bargain will become co-editor, and the series will soon benefit from an electronic submission process available on the IZA website. In the meantime readers who have prepared manuscripts that meet the stringent standards of RLE are encouraged to submit them to Solomon Polachek or Olivier Bargain for possible inclusion in a future volume. Carmel Chiswick Solomon W. Polachek Hillel Rapoport
This page intentionally left blank
xx
PART I: IMMIGRATION AND THE ECONOMY
1
This page intentionally left blank
2
IS IMMIGRATION GOOD OR BAD FOR THE ECONOMY? ANALYSIS OF ATTITUDINAL RESPONSES Christian Dustmann and Ian Preston ABSTRACT In this paper, we study attitudinal responses of host country residents towards further immigration that are triggered by economic considerations. We develop an economic model motivating the empirical work that takes a broader view on these issues than previous papers. We provide empirical analysis that is based on data more specific and better suited to pick up the many channels of economic interest through which benefits and costs of immigration may be felt. Results support previous literature in establishing strong associations between individual characteristics and a wide range of responses to questions relating to perceived impact of immigrants on economic outcomes. Our analysis points towards harmful effects of immigration on the economy being felt through immigration being a fiscal burden rather than having adverse effects on the labour market.
1. INTRODUCTION A large empirical literature in economics is concerned with identifying the effects of immigration on the economy. A particularly strong focus has been Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 3–34 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24001-3
3
4
CHRISTIAN DUSTMANN AND IAN PRESTON
on the effect on wages and employment (see Borjas, 1994, 1999b; Friedberg & Hunt, 1995; Dustmann & Glitz, 2005 for overviews). Many of the papers in this literature fail to find large effects, although there is controversy about this (see recent papers by Borjas, 2003; Card, 2005). In any case, effects of labour market competition are frequently perceived to be one of the main driving forces determining public attitudes towards immigration. Those for whose skills immigrant labour is likely to be a substitute may oppose immigration whereas those for whose skills it is complementary may view immigration more sympathetically. Of course, immigration also has other economic aspects. Individuals may fear that immigrants burden public finances – either through using public services intensively or by contributing to costly social problems such as unemployment – or they may by contrast welcome the contributions made by immigrants to the public exchequer.1 Furthermore, immigrant inflows are frequently suggested as a solution to specific sorts of skill shortages. In political debate this is often used as an argument in support of more liberal migration policies (see, for instance, the debate in European member states on allowing for free movement of labour after EU enlargement as of May 1, 2004). Individuals taking a wider view may in addition appreciate the efficiency gains to be expected from free international movement of labour. Economic enquiry can contribute to arguments on all of these issues by attempting to extend knowledge of the nature and extent of any such economic gains and losses. Recently, a literature has evolved that addresses the formation of opinion and attitudes towards immigrants and immigration in a more direct way than papers that attempt to quantify the economic impact of immigration itself. These papers are based on empirical analysis of attitudinal responses towards immigration and immigrants (see, for example, recent papers by Scheve & Slaughter, 2001; Gang, Rivera-Batiz, & Yun, 2002; Mayda, 2004; Fertig & Schmidt, 2002; O’Rourke & Sinnott, 2003; Dustmann & Preston, 2004; Bauer, Lofstrom, & Zimmermann, 2000 among others).2 Typically, such papers relate responses about individual attitudes to further immigration to individual specific characteristics. Interpretation of the coefficient estimates in the economic part of this literature often relies on well-established economic theory, most prominently the Hecksher-Ohlin model (see, for example, Scheve & Slaughter, 2001), assigning particular interpretation to variables such as education and skills. Other researchers emphasise the importance of non-economic determinants of attitudes to immigration (see, for example, Espenshade & Hempstead, 1996). Dustmann and Preston (2004) developed a model, which allows for three factors in
Is Immigration Good or Bad for the Economy?
5
determining anti-immigrant feelings: labour market considerations, welfare considerations, and racial attitudes. They find all three determinants to be important in affecting attitudes, but identify a dominant role for the race factor, in particular, for the lower educated. The plausibility of the conclusions reached in these papers about the nature and role of economic considerations is not helped by the need to rely on secondary analysis of responses to questionnaires, which are rarely explicit about economic issues. Most of the papers mentioned above rely on attitudes to further immigration as a measure for individual perception of harmful or beneficial effects of immigration in the host economy. The association of this response with skill or education of the respondent is then interpreted within a Hecksher—Ohlin-type framework, where differential responses across skill groups are compatible with differently perceived labour market competition from new immigrants. In this paper, we add to this literature in several ways. Firstly, we broaden the economic argument, by allowing for consideration not only of factors relating to labour market competition, but also to public burden and efficiency considerations. Secondly, we discuss the way such consideration may affect welfare of residents of different skill background in a simple general equilibrium framework. Thirdly, our empirical investigation is based on more specific survey responses than have been used previously. In particular, we not only study the association between economic opinion and demographic characteristics, but also, having conditioned on such effects, seek to structure the interrelation between overall opinions on whether immigration is good or bad for the economy and opinions on more specific economic effects. In our analysis, we allow this overall response to be related to three more specific concerns: labour market competition, public burden, and efficiency considerations. We identify these three response sets from specific survey questions that are directly related to each of these factors. We draw on new and informative data from the European Social Survey (ESS). This survey includes attitudinal information for some 22 European and associated countries, and has a specific module on migration and minorityrelated issues. This module not only provides information on the overall attitudinal response of individuals to further immigration, but also direct responses to a battery of questions concerning the effect on the economy. Our interpretation of the data follows structure imposed by economic theory. We first present a simplified theoretical model which describes the manner and the circumstances under which immigration may benefit or harm different groups in the population. This model is in its nature similar to standard equilibrium models in the literature, but, besides allowing analysis of
6
CHRISTIAN DUSTMANN AND IAN PRESTON
labour market effects of immigration, allows, in addition, analysis of welfare effects of immigration through taxation and welfare payments. According to Borjas (1999a) these are the two main concerns people have when forming attitudes about migration. We also discuss generalisations to capture other possible dimensions of economic concern. It is not so important in the current context whether this model accurately reflects the working of the economy as whether it captures the way individuals might sensibly think about the gains or losses they will incur as a consequence of immigration. We then discuss the empirical implications of this model, and the way these may be reflected in the data we have available. Our empirical analysis has a descriptive part, where we relate responses to the overall evaluation of immigration as well as to demographic information of respondents. We then proceed to a more structural analysis, where we impose a factor structure on the responses concerning particular issues, and where we seek to distinguish between labour market concerns, public burden, and efficiency considerations. Each of these factors is related to a set of responses regarding particular implications of immigration, and we determine the way these factors, in turn, relate to the overall evaluation of whether immigration is good or bad for the economy. The structure of the paper is as follows. In the next section, we present the background theoretical model (Section 2). We then discuss briefly implications for empirical analysis (Section 3). We describe the data set which we use for our analysis in Section 4, and provide descriptive information in Section 5. We then explain our estimation method (Section 6) for the factor model and discuss results in Section 7. Finally, Section 8 concludes.
2. IMMIGRATION: WAGES, TAXES, AND GENERAL WELFARE EFFECTS Our prime interest in this paper is to understand the considerations lying behind individuals’ general opinions on whether immigration is harmful or beneficial for the economy. As we explain in the introduction, these drivers of opinion may relate to labour market competition, to the welfare and tax system as well as to distributional aspects of migration and general welfare considerations reaching beyond national borders. Economic theory is well suited to investigate all these aspects in simple equilibrium models. Even though it would be silly to believe individuals form judgements by working explicitly through models of this sort, individuals observe outcomes in the economies which they inhabit, and it is not
7
Is Immigration Good or Bad for the Economy?
unreasonable to look to the basic mechanisms of such models for guidance as to the sort of beliefs it would be sensible to imagine people might form on the workings of the relevant economic processes. As a motivation for our empirical analysis below, to impose an economically motivated structure on these processes, and to help derive wellfounded empirical implications, we therefore commence by presenting a simple equilibrium model, which helps to structure the way individuals might think about the effect of immigration on the labour market and the welfare system. Although the model allows relatively straightforward derivation of the basic welfare and labour market effects of immigration, the complexity of extending it to general welfare considerations forbids formal development in this paper. While we use formal argumentation to cover wage, employment, and tax effects, we use intuitive argumentation to develop ideas about extensions to other aspects.
2.1. A Simple Equilibrium Model Our model distinguishes between two types of workers (skilled and unskilled) to emphasise the differences in opinion that may result from the different ways immigration can affect current residents in various skill categories. Suppose there are two labour types, skilled (S) and unskilled (U), earning wages wS and wU : The numbers of workers of the two types are given by xi ¼ fi N þ ci M;
i 2 I fS; Ug
where N is the total current population, M the total immigrant population, and fi and ci the skill group shares in the two groups. We consider below, the marginal effects of immigration at the current position where the ratio of immigrants to the current population is p ¼ M=N ¼ 0: Changes in skill group shares follow from d ln xi =dp ’ bi
i2I
where bi ¼ ðci =fi Þ is the relative skill share of immigrants, assumed constant. Capital is assumed elastically supplied at a return to capital, r, which is fixed on world markets. We consider two cases differing in the number of goods produced by the economy. Either the economy produces one good in quantity y0 or two goods in quantities y0 and y1 : We denote the set of goods by J which therefore equals f0g or f0; 1g: These goods are assumed traded and the economy is small, so that their prices p0 and p1 are set on world markets.
8
CHRISTIAN DUSTMANN AND IAN PRESTON
This distinction allows us to differentiate between the case where the economy adjusts to immigration through the output mix, and where the only channel of adjustment is through factor prices. Assuming constant returns to scale and excluding the possibility of joint production, we write the unit cost function for the jth output as cj ðwS ; wU ; rÞ; j 2 J: Letting cji ðwS ; wU ; rÞ denote the derivative @cj =@wi ; demand for the ith type of labour is therefore Sj2J yj cji by Shephard’s lemma. Wages and outputs are determined by two equilibrium conditions. Firstly, labour market equilibrium requires equality of demand and supply of labour, i.e. X yj cji ðwS ; wU ; rÞ xi ¼ 0 i 2 I (1) j2J
and second, firms earn zero profits and therefore ln cj ðwS ; wU ; rÞ ln pj ¼ 0
j2J
(2)
2.2. The Government Budget Constraint Both immigrants and those currently resident consume government services. However, since immigrants differ from current residents in their age, health, number of children, and so on, we should expect them to consume different amounts of government services than residents. We assume therefore that each current resident consumes G and each immigrant G of public services.3 Government spending is financed by a proportional tax on labour income4 at a rate t which is determined to secure government budget balance: X xi wi ¼ GN þ GM (3) t i2I
The welfare of currently resident workers depends on their after-tax wages W i ¼ ð1 tÞwi ; i 2 I so that d ln W i d ln wi t d ln t ¼ 1 t dp dp dp
i2I
(4)
In other words, the effect of migration on after-tax wages works through two channels by affecting before-tax wages and the tax rate. 2.2.1. Two Output Goods Consider first the case with two types of output. Reasoning only from (2) and noting that output prices are fixed by the small country assumption, we
Is Immigration Good or Bad for the Economy?
9
have y0S d ln wS þ y0U d ln wU ¼ 0 y1S d ln wS þ y1U d ln wU ¼ 0 where y0i ¼ @ ln c0 =@ ln wi denotes a factor share. From this it follows immediately that d ln wU =dp ¼ d ln wS =dp ¼ 0: This result, essentially similar to the factor price equalisation theorem (Samuelson, 1948), is what Leamer and Levinsohn (1995) call factor price insensitivity. Wages are determined solely by prices through the zero profit condition (2). Insensitivity extends to non-marginal changes Dp; provided that immigration is not so large as to take factor endowments out of the economy’s cone of diversification, in which case the economy would stop producing one of the two goods. Rather than impacting on wages, long-run effects of immigration are felt in the output mix. These responses can also be deduced and follow from (1) given unchanged factor prices: r0S d ln y0 þ ð1 r0S Þ d ln y1 ¼ bS dp r0U d ln y0 þ ð1 r0U Þ d ln y1 ¼ bU dp where rji ¼ yj cji =Sk2J yk cki denotes a sectoral share in a factor market. Therefore, d lnðy0 =y1 Þ bS bU ¼ 0 dp rS r0U and unskilled immigration leads to a relative expansion of the sector using unskilled labour relatively intensively, in line with the Rybczynski (1955) theorem. For fixed levels of output, equilibrium between cost-minimising factor demands and changed labour endowments would imply wage changes. However, these would lead to positive profits being earned in sectors using intensively labour types which become cheaper. Output in such sectors would be expected to expand, driving back up wages, and long-run equilibrium would not be restored until wages were driven back to their initial levels. Given the absence of wage effects, effects of immigration on the welfare of current workers of both types is dependent solely on the change in the tax rate required by government budget balance. From (3) follows: d ln t þ zS d ln xS þ zU d ln xU ¼ g dp where zi ¼ wi xi =Sk2I wk xk is a share in labour income and g ¼ G=G denotes relative publicly funded consumption of immigrants. Thus, noting
10
CHRISTIAN DUSTMANN AND IAN PRESTON
that zi ¼ twi fi =G; d ln t t ¼ g ðwS cS þ wU cU Þ dp G which simply says that immigration is beneficial to current workers if and only if immigrants contribute more in labour taxes than they take out in consumption of government services. Note that if g ¼ bS ¼ bU ¼ 1 so that immigrants are identical to the current population in both skill composition and public service consumption, then d ln t=dp ¼ 0 by simple substitution from the government budget constraint. However, if immigrants differ from the current population in either respect, then there may still be positive or negative welfare effects. For a given public service consumption G; the perceived gain or loss depends on the allocation of immigrants to skill groups. If immigration is mainly skilled, contributions to the welfare system will be larger and immigration more favourable. 2.2.2. One Output Good We now consider the case where the economy is not able to react to immigration that differs in skill composition from the current labour force by adjusting the output mix. The simplest model to reflect insufficient flexibility is the one output model. This model is often used in the labour literature to motivate the way immigration may affect employment and wages (see, for example, Altonji & Card, 1991; Borjas, 1994). In this model, there will be employment- and wage effects whenever the immigrant population differs in skill mix from the current population. In the one output good case, we obtain the following system of equations determining output, unskilled and skilled wage changes in response to immigration: d ln y0 þ 0SS d ln wS þ 0SU d ln wU ¼ d ln xS ¼ bS dp d ln y0 þ 0US d ln wS þ 0UU d ln wU ¼ d ln xU ¼ bU dp y0S d ln wS þ y0U d ln wU ¼ 0 where 0ij ¼ @ ln c0i =@ ln wj denotes a labour demand elasticity. Hence, by substitution, we obtain d ln wU b U bS ¼ y0 y0 dp 0 0 UU ðSU þ yU0 0US Þ þ 0SS yU0 S
S
(5)
Is Immigration Good or Bad for the Economy?
d ln wS y0 d ln wU ¼ U0 dp dp yS
11
(6)
It is immediately obvious from these two equations that, in the case where immigrants resemble in their skill composition the resident labour force, we should again expect no wage effects, as bU ¼ bS : Tax and welfare effects are therefore as in the two good case. If, however, the skill mix differs then we should expect changes in wages (and output mix). Negativity of the denominator in (5) follows from concavity of the cost function5 and therefore immigration should be expected to depress the wages of workers competing with the type relatively more abundant in immigrant labour and to raise the wages of the other labour type. For example, unskilled immigration therefore depresses unskilled wages and raises skilled wages. There are wage effects but they are not uniform and therefore also raise distributional issues. At the margin, immigrating labour is paid the value of its marginal product and therefore the total remuneration of the existing workforce is unaffected – gains to one labour type exactly offset the losses of the other. However, if we consider non-marginal immigration Dp; we need to appreciate that all immigrating labour is paid the marginal product of the last immigrant and the surplus thus generated on the labour of inframarginal immigrants accrues to owners of other factors (see Berry & Soligo, 1969; Johnson & Stafford, 1999). Total remuneration of workers already resident rises, albeit that some benefit and some lose – this is the so-called ‘‘immigration surplus’’. Returning to the government budget constraint, we can infer d ln t t X d ln wi ¼g wi ci þ fi dp G i2I dp so that effects on the tax rate now require that we recognise the consequences of changes in the wage structure of current workers for tax receipts. Effects on welfare of the workers of the two types then follow from (4) and require weighing up labour market and government budget effects.
2.3. Generalisations 2.3.1. Traded and Non-Traded Goods, Multiple Factors and Immigrants’ Skill Composition The nature of the solution in general depends upon a comparison between the numbers of goods produced and of labour types. The observations
12
CHRISTIAN DUSTMANN AND IAN PRESTON
above can be generalised beyond the case of only two labour types, and can also be extended to allow for non-traded goods.6 What is at issue is the ability of the economy to respond to immigration through flexibility in its output mix. With sufficient number of traded goods, there is no need for immigration to induce factor price changes – whatever the skill mix of immigrating labour – and welfare effects on current workers follow simply from a comparison of immigrants’ tax contributions and consumption of public services. The same is true, even with fewer output goods, if the skill mix in immigrant labour exactly matches that of the current workforce. However, if the skill mix of immigrant labour does differ, then a smaller number of traded goods means that there are insufficient degrees of freedom to accommodate changes in the skill mix through changes in the output mix, and wage changes are therefore nonzero even in the long run. In this case, workers of different types are likely to feel differently about the economic effects of immigration, and a full evaluation of welfare effects on current workers of any labour type require that we weigh up tax and wage effects. 2.3.2. Elastic Labour Supply In the analysis above, we have assumed that labour is completely inelastic. Hence, workers will supply labour at any wage, and immigration will not have employment effects. This, of course, can be extended by assuming that labour supply is elastic, so that some current workers will not be willing to work after immigration, as wages fall below their reservation wages. By this mechanism, reasoning about wage effects of migration can be supplemented with an analysis of voluntary employment responses in the already resident workforce. If unemployed workers can claim from the state then this might open up another mechanism through which tax rates could be affected by immigration. 2.3.3. Disequilibrium The analysis offered so far has been based on assumptions of labour market equilibrium. If we allow for mechanisms preventing factor prices from reaching equilibrium then the effects of immigration will clearly differ. In such cases, immigration may act as an alternative means of equilibration by increasing the relative supply of factors in excess demand (though it could also aggravate disequilibrium if tending to bring in factors in excess supply). Arguments about the benefits of migration as a means of, say, alleviating skill shortages could be conceptualised in this way.
Is Immigration Good or Bad for the Economy?
13
2.4. Effects on Migrants and Sending Countries Effects of migration are plainly not confined to the residents of the host country. There are welfare consequences for both the migrants themselves and residents of the country that they leave. To the extent that migration is voluntary and well informed, it has to be assumed that migrants are better off as a consequence, either economically or in some other way. In economic terms in a model such as that above, this could be in wage gains or in access to the public services of the host country. Effects on residents of the sending country can be analysed in ways that mirror the discussion of the host country above. If the sending country is small, produces both goods and trades at world prices then we should also expect no effect on wages from emigration. Analysis of general welfare effects requires that we make assumptions about sources of international wage differences. With no international technology differences and both goods produced, for example, trade in goods assures equalisation of factor prices without factor mobility whereas if productivity of different types of workers differs depending upon location then factor prices can differ internationally and labour movement will be correspondingly encouraged.7 However, if only one good is produced, wages abroad and at home can be affected by labour movement, and immigration creates an aggregate deficit in the sending country just as it creates a surplus in the receiving country. The impact on world welfare requires that we balance the loss in one country with the gain in the other and the gain to the migrants. If we assume that migrants move to the country where their labour type is better paid, then it can be shown that the overall gain is positive – the world as a whole benefits from movement of labour to locations where it is best remunerated (see Berry & Soligo, 1969; Ruffin, 1984).
3. EMPIRICAL IMPLICATIONS The analysis above provides some suggestions about the range of effects compatible with simple economic models. Individuals are likely to be aware of the overall patterns according to which economic processes determine effects of migration.8 Below, we focus attention on responses to a range of questions from the ESS that are concerned specifically with economic impact. To the extent that answers to these questions are rationalisable in terms consistent with such economic theory, the discussion above points to
14
CHRISTIAN DUSTMANN AND IAN PRESTON
the sort of considerations that might underlie responses. These considerations can be thought of as falling into three types.
3.1. Labour Market Competition Firstly, immigration has the potential to alter labour market outcomes if the skill mix among immigrants is expected to differ from that of the current workforce. In our model these effects could arise if the economy reacts to changes in the skill mix through wage adjustments. If there are effects on wages then these will differ across skill groups and we should therefore expect awareness of such effects to differ across different skill groups. In particular, low-skilled migration should be felt as threatening by low-skilled residents. As we point out, with elastic labour supply, adjustment through wages may also induce unemployment – again being harmful for those who are in competition with immigrant labour. Furthermore, as we point out above, although the unskilled may be harmed, skilled labour will generally benefit, thus creating distributional effects. We should therefore also expect to find such considerations manifested in concerns about effects on the income distribution.
3.2. Public Finance Burden Secondly, immigration can increase or alleviate tax burdens if immigrants are expected to differ from the current population either in skill mix or in propensity to consume public services. In particular, if immigrants are expected to consume more out of public services than residents, then this may be perceived as a possible disadvantage of the overall effects of migration. To the extent that the implied additional tax burdens fall more heavily on the rich, then this may again be a source of difference in opinion across skill groups and income classes. Furthermore, any expected impact on unemployment may also be expected to feed through into concerns about public tax and welfare burdens: If immigration creates unemployment, it will increase the tax burden, thus harming the economically active in the resident workforce.
3.3. Efficiency Thirdly, immigration may be felt to enhance efficiency domestically by alleviating disequilibrium in factor markets, as our discussion above suggests.
Is Immigration Good or Bad for the Economy?
15
Also, for those with a broader perspective, the beneficial impact on world allocation of factor resources may be felt as an international efficiency gain. Notice that efficiency gains of this sort should be of potential benefit to all residents collectively and can be less clearly related to individual circumstances.
4. THE EUROPEAN SOCIAL SURVEY The data we use for our empirical analysis come from the European Social Survey (ESS). The first wave of the ESS was conducted in fall 2002. The ESS interviews between 1,200 and 3,000 people in each of the 22 countries, including all the countries belonging to the EU, seven former Eastern European countries and Israel. Included in the first ESS is a topical module on attitudes towards immigrants and minorities. The module includes over 50 questions.9 A subset of these are specifically economic and address themselves to precisely the considerations alluded to in the theory above. We concentrate on seven questions and display means and standard deviations of these questions in Table 1.10 The first question is an overall question on whether immigration is considered to be good or bad for the economy (Overall). In our analysis below, we seek to explain this question by more specific underlying economic concerns as addressed by the latter six questions. These relate to the underlying issues identified above. The questions Wage and Jobs directly relate to concerns individuals may have about the labour market impact of migration – the first set of our empirical implications discussed above. The question Poor addresses directly distributional aspects and is also most plainly thought of in terms of labour market competition. As we have pointed out above immigration may harm some, but will benefit others, even though typically generating an overall surplus for the receiving economy. The question Tax directly addresses concerns individuals may have about the tax and public spending implications of immigration. Answers regarding Jobs and Poor, because relevant also to distributional impacts which may feed through into taxation and government spending, can also be thought relevant to public finance implications. Finally, the last two questions (Fill and Gain) are statements primarily relating to efficiency aspects of migration – also discussed above. Because the filling of labour market vacancies may also affect wages of other workers, the question Fill may in addition relate to the labour market aspects of immigration.
16
CHRISTIAN DUSTMANN AND IAN PRESTON
Table 1. Variable Name
Overall
Wage
Poor
Jobs
Tax
Fill
Gain
Immigration questions in the European Social Survey.
Survey Question
Immigration is good or bad for country’s economy
Range of Responses
0: bad for the economy 10: good for the economy Average wages are 1: agree strongly generally brought down 5: disagree strongly by immigrants Immigrants harm 1: agree strongly 5: disagree strongly economic prospects of the poor more than the rich 0: take jobs away – Immigrants take jobs 10: create new jobs away in country or create new jobs Immigrants take out more 0: generally take out more in services more than they put in through 10: generally put in more taxes 1: disagree strongly Immigrants help to fill jobs where there are 5: agree strongly shortages of workers All countries benefit if 1: disagree strongly people can move where 5: agree strongly their skills are needed
Mean
Standard Deviation Total
Between country
4.98
2.37
0.65
2.95
1.13
0.41
2.69
1.09
0.34
4.54
2.24
0.81
4.24
2.23
0.50
3.50
1.01
0.23
3.73
0.90
0.14
The objective of the analysis below is to estimate a model that allows us to identify the relevance of these concerns to overall assessment of the impact of immigration on the economy as captured in responses to the first question. In order to do that, we impose a particular factor structure described below. We first describe the responses in more detail in the next section.
5. DESCRIPTIVE OVERVIEW In the following figures, we display the mean responses to the seven questions we use for our analysis. The graphs plot the means of responses by age- and education group. We distinguish between the four age groups (14–30, 31–45, 46–60, and above 60) and the three education groups (low, intermediate, and
17
Is Immigration Good or Bad for the Economy?
0: Bad for Economy 10: Good for Economy
high education). Fig. 1 displays group means for the general questions whether immigrants are good or bad for the economy. Responses differ clearly between education groups, with the low educated being less positive than the highly educated. There are only slight differences in responses across age groups. Fig. 2 displays responses to questions on wage- and employment effects. Again, the differences across educational categories are quite dramatic, with the low educated being much more inclined to assume negative wage- and employment effects than the highly educated. This is compatible with our model if immigration is perceived to be mainly unskilled. Fig. 3 displays responses to questions regarding the effects of immigration on taxes and services (left panel) and distributional effects. Regarding taxes and welfare receipts, it is again the lower educated who are more concerned about immigrants being net recipients of the public finance system. Quite interesting are also the relatively large differences between the young and the old, with older respondents tending to be more sceptical about immigrants making a positive contribution. Regarding distributional concerns, the differences in both age- and educational groups are clearly visible. The less educated seem to be more concerned about immigration imposing a larger burden on the poor – a view
Immigrants bad/good for economy 6
5.5
5
4.5
4 Low
Intermediate Education Age 20−30 Age 46−60
Fig. 1.
High
Age 31−45 Age > 60
Immigrants and the Economy. Source: European Social Survey, 2003.
18
CHRISTIAN DUSTMANN AND IAN PRESTON
that is compatible with our model above in the case of unskilled immigration. While the three younger age groups seem to be quite similar in responses, it is the groups of the above 60-year- olds that is most concerned about more harmful effects on the poor. Fig. 4 displays responses to questions relating to efficiency aspects of migration. In the left panel, responses are displayed that relate to whether respondents believe that immigrants fill jobs where there are shortages – an argument that is often used as justification for more liberal migration policies. Responses differ again between educational categories, but there are also substantial differences across age groups, with older workers being more
Immigrants bring wages down
Immigrants take away jobs/create new jobs 5.5 0: Take away jobs 10: Create New Jobs
1: agree; 5: disagree
3.4 3.2 3 2.8
4.5
4
2.6 Low
Intermediate Education Age 20−30 Age 46−60
Fig. 2.
High
Low
Age 31− 45 Age > 60
Intermediate Education Age 20−30 Age 46−60
High Age 31−45 Age > 60
Wages and Jobs. Source: European Social Survey, 2003.
Immigrants and taxes/services
Immigrants harm poor more than rich 3.2
5 1: agree; 5: disagree
0: take out more 10: take out less
5
4.5
4
3 2.8 2.6 2.4
3.5 Low
Intermediate Education Age 20 − 30 Age 46−60
Fig. 3.
High Age 31−45 Age > 60
Low
Intermediate Education Age 20−30 Age 46−60
High Age 31− 45 Age > 60
Taxes and Distribution. Source: European Social Survey, 2003.
Is Immigration Good or Bad for the Economy?
Immigrants help to fill jobs
Benefit of free movement 3.9
3.7
1: disagree; 5: agree
1: disagree; 5: agree
3.8 3.6
3.5
3.7
3.6
3.4 3.5
Low
Intermediate Education Age 20-30 Age 46-60
Fig. 4.
High Age 31-45 Age > 60
Low
Intermediate Education Age 20-30 Age 46-60
High Age 31-45 Age > 60
Efficiency. Source: European Social Survey, 2003.
19
20
CHRISTIAN DUSTMANN AND IAN PRESTON
in agreement with immigrants filling jobs where there are shortages.11 There is a clear ordering across age groups, with the youngest being most sceptical. The right panel in Fig. 4 relates to overall assessment of the benefits of allowing workers to move where their skills are needed. In terms of our model, this should be welfare enhancing, and the majority of respondents are in agreement with that. Furthermore, there are no strong education patterns, as we observed in the previous responses, and small age differentials, with the older ones being more in agreement with beneficial effects from free movement than the younger ones. Notice that the nature of our data (we only observe a cross section) does not allow us to distinguish between age- and cohort effects, so that the differences in responses across different age groups may be either an age effect, or a cohort effect, or both. In our analysis below, we will associate each of these responses to a set of additional individual characteristics, over and above age and education. The variables we use are displayed in Table 2.
6. THE ECONOMETRIC MODEL Our empirical analysis attempts to relate the overall judgement of whether immigration is harmful or beneficial to three distinct underlying factors: labour market concerns, welfare concerns, and efficiency considerations. We assume that each of these underlying factors is related to a specific subset of survey questions. We explain in this section how our estimation method works. Notice first that we consider all the outcomes we observe (denoted by yi ; zi and which are displayed in Table 1) as discrete responses which relate to underlying continuous and latent variables yni ; zni in the following way:
n
yn ¼ f L þ XA þ u
(7)
zn ¼ fM þ XB þ v
(8)
where y is an n 1 vector of latent attitudinal responses to the question on overall economic effect of immigration for n individuals (variable Overall), zn an n q matrix of latent attitudinal responses to the questions on specific economic effects, and X an n k matrix of observed characteristics. In our analysis, these refer to those characteristics that we display in Table 2. The matrix f is an n p matrix of factor scores capturing the p underlying dimensions to economic concerns, L a p 1 vector of loadings reflecting the
21
Is Immigration Good or Bad for the Economy?
Table 2. Variable Name
Variable Names and Descriptive Statistics. Proportion
Standard Deviation Total
Between countries
Labour market status Unemployed Inactive Retired House work Student
0.036 0.020 0.208 0.235 0.106
0.187 0.140 0.406 0.424 0.308
0.019 0.011 0.041 0.109 0.047
Education Secondary Higher
0.635 0.208
0.481 0.400
0.165 0.069
Immigrant status Immigrant Father immigrant Mother immigrant
0.092 0.136 0.132
0.289 0.343 0.339
0.080 0.145 0.139
Other Male Minority City Town
0.498 0.039 0.335 0.297
0.500 0.193 0.472 0.457
0.031 0.028 0.106 0.089
Variable Name
Age (in years)
Mean
45.60
Standard Deviation Total
Between countries
17.54
2.26
Note: The participating countries in the ESS that are included in our sample are: Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Luxembourg, Netherlands, Norway, Poland, Portugal, Slovenia, Spain, Sweden, Switzerland, and United Kingdom.
importance of these concerns to overall assessment of economic impact, and M a p q matrix which maps the factor scores into the opinions on specific effects. Following our discussion above, we take p ¼ 3 with factors corresponding to concerns about labour market competition, public finance burden, and efficiency. The n 1 vector u and n q matrix v contain disturbance terms assumed normally distributed and uncorrelated with either X or f or each other so that
22
CHRISTIAN DUSTMANN AND IAN PRESTON
0 u N ; 0 v
Su
0
0
Sv
!!
(9)
with Sv diagonal. The factors are themselves allowed to be influenced by the regressors X : (10)
f ¼ XC þ w
where c is a k p matrix of coefficients in the underlying lower dimensional model. We assume that w Nð0; Sw Þ with w uncorrelated with u and v: Notice that we do not need to assume diagonality of Sw : This structure implies an estimable reduced form, which can easily be obtained by substitution. Let Y n denote the stacked vector of latent responses, n y n Y ¼ zn We then obtain Yn ¼ XG þ
(11)
where G¼C
L M
þ
A B
(12)
is the ð1 þ qÞ k matrix of reduced form coefficients and L u ¼w þ M v
(13)
Then Nð0; S Þ; where S ¼
Su þ L0 Sw L
L0 Sw M
M 0 Sw L
Sv þ M 0 Sw M
!
S11 S012
S12 S22
!
(14)
is the ðq þ 1Þ ðq þ 1Þ variance–covariance matrix of the reduced form residuals. In implementing this it is the restrictions placed on M which are crucial to identification. We could assume each factor to load solely on strict subsets of zn : In this we are guided by the core idea that the variables Wage and Poor are the most relevant for the labour market competition factor, it Jobs and Tax for the public burden factor, and Fill and Gain for the efficiency
Is Immigration Good or Bad for the Economy?
23
factor. We have estimated a specification allowing only for these loadings – which is to say with block diagonal M – but also other specifications allowing for a certain overlap. Results are qualitatively robust across these specifications. The results we present below are based on the following relatively permissive specification as suggested by the discussion in Section 4: the first factor, reflecting labour market competition, loads solely on the responses Wages, Poor, Jobs, and Fill; the second factor, reflecting public finance burden, loads solely on the responses Poor, Jobs, and Tax; the third factor, reflecting efficiency, loads solely on the responses Fill and Gain.
6.1. Estimation We proceed to estimate the model in several stages. First we estimate the reduced form parameters of the model in (11), consisting of the matrix G and the covariance matrix S ; without imposing the restrictions in (12) and (14). We estimate G by independent ordered probit estimation for the seven responses. We then estimate S by a series of pairwise bivariate ordered probits fixing the value of G at their estimated values. Results are reported in Tables 3 and A1 (the latter in the appendix), and we discuss them in Section 7. These unrestricted latent covariance estimates allow us to form unrestricted estimates of the implied regression coefficients linking the latent responses @Eðyn jzn ; X Þ=@zn ¼ S1 22 S12 : In other words, we derive the parameter estimates we would obtain, if we regressed the latent underlying variable of the overall assessment of whether migration is good or bad on the latent six other more specific responses, conditional on all the regressors included in the matrix X : These are reported in Table 4 and we discuss them in Section 7.2. We can now impose the additional restrictions in (14), implied by our factor structure on the model. We impose these restrictions by equally weighted minimum distance estimation on the reduced form parameters to identify the parameters in L; M and Sw : Tests of overidentifying restrictions are calculated using formulae from Newey (1985). The present results are given in Table 5 and discussed in Section 7.3.12
Variable
Overall
24
Table 3.
Perceived Economic Impact.
Wages
Poor
Jobs
Tax
t-ratio
Coeff
t-ratio
Coeff
t-ratio
Coeff
0.140 0.071 0.010 0.004 0.260
4.76 1.81 0.48 0.29 11.08
0.199 0.092 0.056 0.020 0.169
6.61 2.30 2.50 1.26 6.91
0.205 0.101 0.058 0.010 0.131
6.70 2.44 2.60 0.61 5.44
0.023 0.0661 0.005 0.002 0.193
0.224 0.349
15.96 21.39
0.194 0.341
13.25 20.25
0.176 0.288
11.90 17.38
0.159 0.264
11.42 15.73
0.124 0.232
8.79 14.16
2.40 8.27
0.030 0.019
1.95 1.14
0.281 0.114 0.181
11.52 4.50 6.91
0.077 0.074 0.104
2.83 2.70 3.74
0.142 0.061 0.147
5.17 2.23 5.20
0.320 0.060 0.184
13.30 2.36 7.07
0.260 0.139 0.134
10.34 0.114 4.09 5.41 0.020 0.71 5.07 0.132 4.68
0.074 0.096 0.009
2.57 3.45 0.31
0.074 3.68 0.067 3.22 0.006 2.98 0.007 3.21 0.137 11.35 0.054 4.29 0.170 2.56 0.105 3.59 0.125 8.68 0.023 1.57 0.059 3.99 0.033 2.20
0.024 0.001 0.005 0.049 0.004 0.020
1.15 0.24 0.40 1.68 0.27 1.31
0.000 0.000 0.019 0.001 0.055 0.014
t-ratio
Coeff
Gain
Coeff
t-ratio
Coeff
Fill t-ratio
Coeff
t-ratio
Labour market status 7.71 0.130 4.42 0.087 2.93 0.007 0.22 1.60 0.082 2.05 0.034 0.83 0.048 1.14 0.24 0.007 0.31 0.005 0.23 0.025 1.10 0.77 0.007 0.49 0.024 1.47 0.014 0.90 8.30 0.187 7.89 0.123 4.97 0.060 2.49 0.035 0.145
Immigrant status Immigrant Father immigrant Mother immigrant Other Age/10 Age2/100 Male Minority City Town Sample size
31,822
Note: All ordered probit estimates. Country dummies included in all cases.
0.01 0.008 0.40 0.09 0.001 0.37 1.60 0.027 2.20 0.03 0.051 1.81 3.79 0.069 4.80 0.95 0.055 3.71
0.026 0.001 0.018 0.072 0.056 0.014
1.23 0.040 1.88 0.36 0.002 1.02 1.44 0.015 1.20 2.46 0.069 2.29 3.68 0.034 2.22 0.90 0.039 2.54
CHRISTIAN DUSTMANN AND IAN PRESTON
Unemployed Inactive Retired House work Student Education Secondary Higher
25
Is Immigration Good or Bad for the Economy?
Table 4.
Implied Latent Regression Coefficients.
Variable
Coeff
t-ratio
Wage Poor Jobs Tax Fill Gain
0.0813 0.0810 0.2901 0.3336 0.1027 0.0605
12.70 12.12 44.03 54.36 21.81 13.04
Sample size
31,822
7. EMPIRICAL RESULTS 7.1. Unrestricted Ordered Probit Estimates In Table 3 we display results from the independent ordered probit models. We distinguish between four sets of regressors: The individual’s labour market status, education, immigrant status, and other variables, including age, gender, minority affiliation, and urbanisation. In all our regressions we condition on country dummies. The reference category is a native-born rural majority female in paid work without secondary education and with native parents. All responses are normalised in such a way that a positive number indicates a more optimistic view about a particular outcome. The first pair of columns shows results for the overall assessment of whether immigration is bad or good for the economy and respective t-ratios. The labour market status variables suggest that individuals in paid work or students have an overall more optimistic view about the impact of immigration than the unemployed. This is similar for questions relating to labour market competition and public burden, but the relationship seems unclear for some of the efficiency questions. Not surprisingly, immigrants have not only a more positive view about the overall effect of immigration than natives (immigrants are in the reference group), but evaluate also the impact of migration on other concerns more positively. Interestingly, an optimistic view seems to be stronger even among second-generation immigrants. The partial regression coefficients on the educational dummies are much in line with our descriptive graphs above. Higher education is associated with a more optimistic view on migration, and the coefficient estimates are highly significant. This interpretation extends to all the questions in the
26
CHRISTIAN DUSTMANN AND IAN PRESTON
Table 5.
Three-Factor Model. M matrix
Variable
Wage Poor Jobs Tax Fill Gain
Labour Market Competition
Public Burden
Efficiency
Sv
Coeff
t-ratio
Coeff
t-ratio
Coeff
t-ratio
0.784 0.651 0.098 — 0.052 —
28.61 17.93 9.69 — 3.57 —
— 0.132 0.638 0.669 — —
— 4.43 75.62 120.23 — —
— — — — 0.579 0.334
— — — — 37.31 37.27
0.386 0.478 0.525 0.553 0.677 0.889
Lmatrix Variable
Overall
Variable
Labour market competition Public burden Efficiency
Labour Market Competition Coeff
t-ratio
0.021
1.89
Public Burden
Coeff
Efficiency
Su
t-ratio
Coeff
t-ratio
0.728 55.78 Sw matrix
0.080
6.70
Labour Market Competition
Public Burden
0.388
Efficiency
1.000
—
0.468
23.68
0.238
10.52
0.468 0.238
23.68 10.52
1.000 0.514
— 37.34
0.514 1.000
37.34 —
Note: Overidentifying restrictions: w25 ¼ 125:596:
table, with the weakest association between education and gains from free movement (last column) – as in the figures above. Age effects differ across responses. This again is in line with our descriptive charts in Section 5. Individuals in more densely populated areas have a more optimistic opinion about the overall effects of immigration throughout.
7.2. Latent Response Regressions In Table 4, we display estimates of the implied regression coefficients as calculated from the estimated residual covariance matrix. These are
Is Immigration Good or Bad for the Economy?
27
unrestricted estimates of the derivative of the conditional expectation Eðyn jzn ; X Þ with respect to the latent variables in zn : Notice that normalisation of the variances in the first step estimation implies that estimates are interpretable as the impact of a one standard error change in the respective latent regressor on the latent overall assessment of whether immigration is good or bad, itself expressed in standard errors.13 All coefficient estimates are significant. The results suggest that tax and job concerns have the largest impact on the overall assessment of immigration, while the wage and distributional aspects are considerably weaker as are perceptions about general welfare gains of immigration.
7.3. Three-Factor Model We now turn to analysis where we structure the pattern of responses according to the three-factor model outlined above, each factor being allowed to be associated with the variable Overall. In Table 5, we report estimated loadings of the factors on the indicator questions (M matrix) and the vector of factor loadings on the overall assessment question (L vector) which we obtain by imposing the restrictions in (14), on our estimated reduced form coefficients. Estimates in this table are obtained without imposing any restrictions on the Sw matrix which determines the correlation between the factors. Results on the M matrix (displayed in the upper panel of the table) show well-determined coefficients in all parts of the matrix. Three of the variables are allowed to enter into more than one factor – namely the Poor, Jobs, and Fill variables. While there is evidence in each case of loading on more than one factor, the association is in each case much stronger with a particular one of them. The estimates on L are displayed in the second panel of the table. They appear to suggest the strongest role for fears about public finance burden, and a lower, but still significant role for efficiency considerations. There seems to be no evidence of importance for labour market competition. This may seem strange, given that the estimates in Table 4 in Section 7.2 do suggest a strong association. Notice, however, that we allow in our specification the different factors to be correlated, and our estimates suggest strong well-determined positive correlations across factors (as seen in the Sw matrix). The results therefore suggest that the variation in the public burden factor absorbs most of the variation in the labour market factor in explaining the overall assessment of whether migration is good or bad.
28
CHRISTIAN DUSTMANN AND IAN PRESTON
This is an interesting result. In the strict interpretation of the model we have sketched above and to the extent that our identification of factors is plausible, this tends to suggest that the perception of possible harmful effects of immigration may be less associated with labour market competition than with worries about immigrants being a fiscal burden on the overall economic system. This is to some extent compatible with apparently contradictory results of previous empirical studies who, although establishing a strong link between education and general opposition to further immigration, did not find evidence for tense local labour markets (represented by local unemployment) being significantly associated with such responses. We should note that the overidentifying restrictions imposed by the model are rejected by the data, as clearly evidenced by the w25 value of 125.596. However, the model may not be badly misspecified when we consider the very large sample size and the possibly questionable auxiliary assumptions of linearity and normality. The uniqueness statistic of 0.388 in the Su column shows that the factor structure is capable of accounting for over 60 per cent of the variation in the overall opinion. If we estimate a model which does not permit correlation between the three factors, restricting the off diagonal elements of Sw to be zero, we do find a much stronger association between the overall evaluation and the labour market competition factor. Results of that model are displayed in Table A2 in the appendix. However, the massive increase in the w2 statistic suggests strong reason to reject the implied orthogonalisation of the factor structure. One conclusion to be drawn from this analysis is that strict interpretation of attitudinal responses within a labour competition context may be misleading and that there are other economic concerns that may be more prominently determining an overall opinion about the costs and benefits of immigration.
8. SUMMARY AND CONCLUSIONS In this paper, we provide analysis of individual perceptions about the effects of immigration on the host country’s economy. Our paper contributes to a large and growing literature that tries to understand the particular concerns that drive residents’ attitudes towards immigration. We contribute to this literature by broadening the economic argument, allowing for consideration not only of factors relating to labour market competition, but also of factors relating to public burden and efficiency considerations. We provide a
Is Immigration Good or Bad for the Economy?
29
theoretical discussion that takes a broader view on the channels of welfare effects than much of the previous literature. We present empirical investigation based on more specific survey responses than have been used previously, studying the response to whether immigration is considered as good or bad for the economy as being determined by three more specific concerns: labour market competition, public burden, and efficiency considerations. Identification of these responses is based on specific survey questions that are directly related to each of these factors. Our analysis yields a set of interesting results. First, our theoretical model suggests that economic self-interest points to an assessment of the benefits and costs from immigration that encompasses not only labour market competition, but also taxes and public burden, as well as general welfare effects determined by efficiency considerations. Interpretations that focus solely on the competition aspect seem therefore quite narrow. Our empirical analysis supports findings in much of the previous literature of a strong relationship between education and more positive attitudes towards various issues relating to migration. We also find that the particular questions that focus on very particular concerns are all strongly related to the overall assessment of migration. When we impose a particular factor structure on the data, we find that concerns regarding economic competition are largely represented by overall concerns regarding public burden. This is an interesting finding which does not dispute the importance of economic concern in the determination and formation of attitudes and opinion about benefits and costs of migration, but the narrow interpretation within a labour market competition framework. By no means do we wish to imply that we have exploited the entire range of factors that affect assessment of costs and benefits of migration. We have concentrated here on some factors that are rationalisable within economic models, broadening the perspective of such analysis. However, we strongly believe that opinions on cultural effects and racial tolerance may be equally germane to responses on the overall desirability of immigration. Evidence provided by Dustmann and Preston (2004) supports this conjecture which we are exploring in greater depth using ESS data. We believe that research into this area is important and that the recent effort undertaken by economists in understanding various attitudes related to immigration and immigrants is most welcome. However, we also believe that the complexity of the processes that contribute to attitude formation requires approaches over and above sole economic argument. We have only just started to understand data regularities and evidence in this important area of research.
30
CHRISTIAN DUSTMANN AND IAN PRESTON
NOTES 1. Research for the British Home Office (Gott & Johnston, 2002) recently suggested, for example, that immigrants to the UK make a positive net contribution to British public finances. 2. Related issues come up also in the literature on preferences on trade policies – see, for instance, papers by Mayda and Rodrik (2001), Scheve and Slaughter (2001), and O’Rourke and Sinnott (2001). 3. We could allow these amounts to depend also on the levels of wages and the skill compositions of the two groups. However, this would complicate the model without adding any appreciable insight. 4. We exclude, for convenience, the possibility that capital income is also taxed domestically. Since, we have assumed capital to be elastically supplied at a fixed (posttax) return, labour inflows should also lead to inflows of capital. To the extent that the returns to that capital attract domestic tax liabilities, this should soften the impact of immigration on public finances. However, since the owners of capital may reside abroad, it is not obvious how their income would be taxed. 5. Note that ! " 0 2 # 0 y0U 0 wU 0 c0U 0 c 0 0 0 yU UU SU þ 0 US þ SS 0 ¼ 0 cUU 2 0 cSU þ U0 c0SS c cS c yS yS U S which is a positive multiple of a quadratic form in the second derivatives of the cost function and therefore negative. 6. The relevant algebra can be drawn from trade theory models – see Ethier (1984), and Woodland (1982). 7. Trefler (1998) discusses welfare effects in such a model, and other cases, drawing attention to terms of trade effects that will occur where the economies are large. 8. Of course, evaluation of the benefits and costs of immigration may also be motivated by considerations which are non-economic (see Dustmann & Preston, 2004 for analysis of racial aspects to attitudes towards immigration). 9. See Card, Dustmann, and Preston (2006) for a more detailed description of the survey’s immigration module. 10. The variable names are chosen by us to reflect the earlier economic discussion rather than taken from the coding. We have reordered responses, so that higher values indicate a more positive response to immigration for each question. 11. As pointed out by a referee, one reason for this may be that immigrants are often employed in jobs related to care for the elderly. 12. For full details of how we calculate standard errors and test statistics, we refer the reader to Dustmann and Preston (2004). We draw on the work of Muthe´n (1984). Note though that identification of parameters in this particular model is subtler than in that paper – in particular the parameters of M are not identified in the current context solely from restrictions on S22 although all parameters of L; M; and Sw are identified from imposition of restrictions jointly on S12 and S22 : 13. The residual correlation matrix on which these results are based is presented in the appendix (Table A1).
Is Immigration Good or Bad for the Economy?
31
ACKNOWLEDGMENTS We are grateful to Maria Demousis for research assistance, to David Card for discussion, and to the Nuffield Foundation for financial support. Parts of this paper were written by Dustmann who visited the Research School of Social Sciences at the Australian National University. He is grateful for the hospitality.
REFERENCES Altonji, J. G., & Card, D. (1991). The effects of immigration on the labor market outcomes of less-skilled natives. In: J. M. Abowd & R. B. Freeman (Eds), Immigration, trade and labor. Chicago: University of Chicago Press. Bauer, T., Lofstrom, M., & Zimmermann, K. F. (2000). Immigration policy, assimilation of immigrants and natives’ sentiments towards immigrants: Evidence from 12 OECDCountries. Swedish Economic Policy Review, 7, 11–53. Berry, R. A., & Soligo, R. (1969). Some welfare effects of international migration. Journal of Political Economy, 77, 778–794. Borjas, G. J. (1994). The economics of immigration. Journal of Economic Literature, 32(4), 1667–1717. Borjas, G. J. (1999a). Heaven’s door. Princeton: Princeton University Press. Borjas, G. J. (1999b). The economic analysis of immigration. In: O. Ashenfelter & D. Card (Eds), Handbook of labor economics, (Vol. 3A, pp. 1697–1760). Amsterdam: Elsevier. Borjas, G. J. (2003). The labor demand curve is downward sloping: Reexamining the impact of immigration on the labor market. Quarterly Journal of Economics, 118, 1335–1374. Card, D. (2005). Is the new immigration really so bad? CReAM Discussion Paper 02/04; forthcoming Economic Journal Features. Card, D., Dustmann, C., & Preston, I. (2006). Understanding attitudes to immigration and national exclusiveness: The migration and minority module of the first European Social Survey. In: R. Jowell, C. Roberts, R. Fitzgerald & G. Eva (Eds), Measuring attitudes cross-nationally – lessons from the European Social Survey. Sage Publications. Dustmann, C., & Preston, I. (2004). Racial and economic factors in attitudes towards immigration. CReAM Discussion Paper no. 1. Dustmann, C., & Glitz, A. (2005). Immigration, jobs and wages: Theory, evidence and opinion. Centre for Economic Policy Research (CEPR), London, May. Espenshade, T. J., & Hempstead, K. (1996). Contemporary American attitudes towards U.S. immigration. International Migration Review, 30, 535–570. Ethier, W.J. (1984). Higher dimensional issues in trade theory. In: R.W. Jones, & P.B. Kenen (Eds.), Handbook of international economics, (Vol I), Amsterdam: Elsevier Science. Fertig, M., & Schmidt, C.M. (2002). The perception of foreigners and jews in Germany – a structural analysis of a large opinion survey. IZA Discussion Paper No. 431. Friedberg, R. M., & Hunt, J. (1995). The impact of immigration on host country wages, employment and growth. Journal of Economic Perspectives, 9, 23–44.
32
CHRISTIAN DUSTMANN AND IAN PRESTON
Gang, I., Rivera-Batiz, F. & Yun, M.-S. (2002). Economic strain, ethnic concentration and attitudes towards foreigners in the European Union. Working Paper no. 2002–2014. Rutgers University Department of Economics. Gott, C., & Johnston, K. (2002). The migrant population in the UK: Fiscal effects. Home Office, RDS Occasional Paper no. 77. Johnson, G., & Stafford, F. (1999). The labour market implications of international trade. In: O. Ashenfelter & D. Card (Eds), Handbook of labor economics, (Vol. 3B, pp. 1697–1760). Amsterdam: Elsevier. Leamer, E. E., & Levinsohn, J. (1995). International trade theory: The evidence. In: G. Grossman & K. Rogoff (Eds), Handbook of international economics, Vol III (pp. 1339–1394). Amsterdam: Elsevier Science. Mayda, A.M. (2004). Who is against migration? A cross-country investigation of attitudes towards immigrants. IZA Discussion Paper No. 1115. Mayda, A.M., & Rodrik, D. (2001). Why are some people (and countries) more protectionist than others? NBER Working Paper no. 8461. Muthe´n, B. O. (1984). A general structural equation model with dichotomous, ordered categorical and continuous latent variable indicators. Psychometrica, 49, 115–132. Newey, W. K. (1985). Generalized method of moments specification testing. Journal of Econometrics, 29, 229–256. O’Rourke, K. H., & Sinnott, R. (2001). What determines attitudes towards protection? Some cross-country evidence. In: S. M. Collins & D. Rodrik (Eds), Brooking trade forum 2001 (pp. 157–206). Washington, DC: Brookings Institute Press. O’Rourke, K.H. & Sinnott, R. (2003). Migration flows: Political economy of migration and the empirical challenges. Trinity College Dublin Economic Papers 20036, Trinity College Dublin Economics Department. Ruffin, R. J. (1984). International factor movements. In: R. W. Jones & P. B. Kenen (Eds), Handbook of international economics, Vol. I. Amsterdam: Elsevier Science (Chapter 5). Rybczynski, T. M. (1955). Factor endowments and relative commodity prices. Economica, 22, 336–341. Samuelson, P. (1948). Internationalization of trade and the equalization of factor prices. Economic Journal, 48, 163–184. Scheve, K. F., & Slaughter, M. J. (2001). Labor market competition and individual preferences over immigration policy. Review of Economics and Statistics, 83, 133–145. Trefler, D. (1998). Immigrants and natives in general equilibrium trade models. In: J. P. Smith (Ed.), The immigration debate: Studies on the economic, demographic, and fiscal effects of immigration (pp. 206–238). Washington, DC: National Academy Press. Woodland, A. (1982). International trade and resource allocation. North Holland: Amsterdam.
33
Is Immigration Good or Bad for the Economy?
APPENDIX The residual correlation matrix and the three-factor model estimation for the case of diagonal Sw are given in Tables A1 and A2. Table A1. Residual Correlation Matrix. Variable
Overall
Wage
Poor
Jobs
Tax
Fill
Gain
Overall Wage Poor Jobs Tax Fill Gain
1.000 0.306 0.342 0.524 0.532 0.242 0.160
0.306 1.000 0.559 0.310 0.237 0.065 0.069
0.342 0.559 1.000 0.352 0.301 0.091 0.065
0.524 0.3061.0 0.352 1.000 0.455 0.207 0.125
0.532 0.237 0.301 0.455 1.000 0.165 0.099
0.242 0.065 0.091 0.207 0.165 1.000 0.189
0.160 0.069 0.065 0.125 0.099 0.189 1.000
Sample size
31,822
Note: Pairwise bivariate ordered probit estimates. Country dummies included in all cases.
34
CHRISTIAN DUSTMANN AND IAN PRESTON
Table A2.
Three-Factor Model. M matrix
Variable
Wage Poor Jobs Tax Fill Gain
Labour Market Competition
Public Burden
Coeff
t-ratio
Coeff
0.939 0.573 0.350 — 0.139 —
61.79 65.16 48.34 — 22.67 —
— 0.307 0.577 0.818 — —
t-ratio
45.56 93.87 94.86
Efficiency
Sv
Coeff
t-ratio
— — — — 0.484 0.391
— — — — 42.20 38.20
0.119 0.578 0.544 0.332 0.747 0.847
L matrix Variable
Overall
Labour Market Competition
Public Burden
Efficiency
Coeff
t-ratio
Coeff
t-ratio
Coeff
t-ratio
0.313
43.18
0.661
98.71
0.410
40.96
Su
0.297
Sw matrix Variable
Labour Market Competition
Labour market competition Public Burden Efficiency
1.000
—
— —
Note: Overidentifying restrictions: w28 ¼ 3998:698
Public Burden — 1.000 —
Efficiency —
—
— 1.000
—
THE EFFECTS OF INCOMPLETE EMPLOYEE WAGE INFORMATION: A CROSS-COUNTRY ANALYSIS Solomon W. Polachek and Jun (Jeff) Xiang ABSTRACT In this paper, we define a tractable procedure to measure worker incomplete information in the labor market. The procedure, which makes use of earnings distribution skewness, is based on econometric frontier estimation techniques, and is consistent with search theory. We apply the technique to 11 countries over various years, and find that incomplete information leads workers to receive on average about 30–35% less pay than they otherwise would have earned, had they information on what each firm paid. Generally, married men and women suffer less from incomplete information than the widowed or divorced; and singles suffer the most. Women suffer more from incomplete information than men. Schooling and labor market experience reduce these losses, but institutions within a country can reduce them, as well. For example, we find that workers in countries that strongly support unemployment insurance (UI) receive wages closer to their potential, so doubling UI decreases incomplete information and results in 5% higher wages. A more dense population reduces search costs leading to less incomplete information. A more industrial economy disseminates wage information better, so workers exhibit less incomplete information and higher wages. Finally, we find that Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 35–75 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24002-5
35
36
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
foreign worker inflows increase incomplete information, and at the same time reduce average wage levels, at least in the short run.
INTRODUCTION Tikva Lecker’s main research centers on two issues. First, she concentrated on earnings – particularly earnings of migrants including the Arabs in Israel. Second, she worked on policies affecting legal and illegal immigrant flows. As will be explained, this paper marries these two topics: immigration and earnings. It does so in a unique way. The paper goes back to George Stigler’s (1961, 1962) path breaking seminal work on incomplete information and applies earnings function concepts along with frontier estimation techniques to devise a measure of incomplete information, which we parameterize as the degree employees end up receiving wages less than their potential, given their skills. Then, once we estimate incomplete information, we explore how institutional factors, such as unemployment insurance and foreign worker inflows, affect incomplete information across eleven countries. Incomplete information leads workers to accept wages below what they could have attained had they full information about each potential employer’s pay.1 Workers could improve their wage by prolonging job search, but information about available jobs is costly. To find a job, workers search the market, but normally terminate their search before discovering the very highest paying job available. As explained by Nelson (1970); Mortensen (1970); McCall (1970) and others, individuals set a reservation wage and search until offered a job at least equal to this reservation wage. On average, the higher the reservation wage the longer the search, but invariably the accepted wage is almost always less than the best possible market wage available for a person of their skill level. Receiving a wage less than the maximum possible wage (given one’s skill) is an important phenomenon because it illustrates an effect of incomplete information that arises from costly search. Collectively, over the whole economy, it reflects foregone gross national product, since so many within the economy are similarly receiving less than they could potentially earn. Therefore in the aggregate, this wage gap reflects incomplete information’s cost to the economy. Measuring the effects of incomplete information is significant for at least two reasons. First, as just noted, one can gauge the overall economic losses associated with costly information. Second, by
The Effects of Incomplete Employee Wage Information
37
having a measure of these losses, one can assess appropriate policies needed to reduce search costs, thereby increasing efficiency within the economy. One result regarding policy seems pretty much universal in past literature: Unemployment insurance (measured by the replacement rate) subsidizes employee search, which lengthens unemployment duration (e.g., Moffitt & Nicholson, 1982; Meyer, 1990). The resulting extra search enables workers to obtain more information and higher wages (e.g., Ehrenberg & Oaxaca, 1976). A gigantic body of literature corroborates these findings regarding UI, both for the US and other countries (e.g., Jurajda & Tannery, 2003; Fougere, Pradel, & Roger 1998; Van den Berg & van der Klaauw, 2001; Micklwright & Nagy, 1995; Card & Levine, 1998; Beach & Kaliski, 1983; Ham & Rea, 1987; Arellano, Bentolila, & Bover, 1998; Belzil, 1995; Gonzalo, 2002). This paper differs from past empirical research in three ways. First, rather than concentrating on unemployment duration, it examines worker wages. It focuses on the extent to which workers receive a wage less than what they could be paid on the basis of their skill level. This focus enables the paper to get a metric defining the monetary effect (and indeed a measure of) incomplete information. Second, the paper examines incomplete information from an international perspective. Rather than examining incomplete information for one single country, it obtains measures for 11 countries over several time-periods. Third, by looking across countries, the paper is able to explain how inter-country institutional differences affect incomplete information. In testing our model, we first corroborate past findings on unemployment insurance. We show that an employee’s incomplete information is smaller where UI is a larger proportion of GDP. Then second, we test whether institutional factors lead to differences in incomplete employee information. In this context, we show that geographic considerations as well as industrial structure likely affect search costs, and hence incomplete information. Third, given Tikva Lecker’s interest in immigration, the paper culminates by examining how foreign migrant workers affect overall employee incomplete information. In this context, we determine that an influx of foreign workers into an economy decreases the effectiveness of search, thereby increasing the degree of incomplete information within the labor market.2
BACKGROUND A country’s distribution of wages defines the benefits of job search. The more dispersed the wages, the greater the gains from the search. Higher
38
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
search gains lead to relatively higher reservation wages, which in turn leads to more search. But as a consequence, earnings dispersion narrows, and the degree of incomplete information diminishes. At the same time, the amount of search is related to search costs. Individual characteristics, such as a worker’s location vis-a`-vis jobs or a worker’s opportunity costs (which would be higher for those already at work) affect search costs. Higher search costs diminish search, leading to wider earnings dispersion. The whole search process provides job seekers with wage (and amenity) information, but clearly information remains imperfect because search is costly.3 Creating an index of the degree of incomplete information is important because, as already mentioned, incomplete information leads to lost opportunities and diminished GNP. But in addition, the level of incomplete information is an indicator of market competition. Whereas prices collapse to a unique single equilibrium in purely competitive full-information markets, this is not the case in imperfect competition where there are multiple prices. But even markets for homogeneous easily transferable commodities contain price variability when there is imperfect information. The amount of market imperfection can also be related to institutional factors. These institutions might include information networks, such as nationally based bargaining units (e.g., Germany) and the availability of unemployment insurance (most developed countries); they also could include other institutions, such as inflows of foreign workers (who might have little information) into an economy. There is much research on how migrants affect the labor market. These analyses include the effect of immigrants on wages. Some of this research looks at how quickly (in terms of number of generations) migrants achieve success equal to natives (Chiswick, 1978). Other of this research examines how this rate of assimilation depends on migrant quality, particularly concentrating on the skills migrants possess upon entering a country (Borjas, 1985). However, almost no research assesses workers’ overall knowledge of wage offer distributions. Similarly, almost no research assesses the effects of foreign workers on incomplete information within a particular labor market. This paper examines both. It develops a metric defining the effects of incomplete information. It assesses the impact of institutional factors including unemployment insurance and population density. Then, in memory of Tikva Lecker, it assesses how in-migration of foreign workers affects the amount of information workers have regarding the labor market. All this is done with international labor market data on 11 countries obtained from the Luxembourg Income Study (LIS).
The Effects of Incomplete Employee Wage Information
39
INCOMPLETE INFORMATION There is a vast literature on the theory of equilibrium prices. Most is theoretical and concerns defining the conditions under which there is an equilibrium price distribution (e.g., Reinganum, 1979; Burdett & Judd, 1983; Bester, 1986; Arnold, 2000; Kamiya & Sato, 2004). There is also a small but growing body of literature that is empirical. That literature relies on overall price variation to measure the degree of incomplete information (e.g., Stigler, 1961; Stigler & Kindahl, 1970; van Hoomissen, 1988; Lach & Tsiddon, 1992; Sorensen, 2000). As recognized by Stigler and others, there is a major drawback of merely using price (or wage) dispersion as a measure of incomplete information. Price dispersion can vary for many reasons other than incomplete information. These include differences in worker quality and differences in worker-firm bargaining power such as through unions; they also can result from noisy data. Dispersion measures do not get at these considerations because they do not net out these effects. Regression models suffer the same biases because the dispersion measures they use as the dependent variable do not net out random price variations, nor do they distinguish between a worker’s (seller’s) and a firm’s (buyer’s) incomplete information. Thus, these past more traditional measures do not reflect accurate estimates of incomplete information. One technique to get at incomplete information that can net out worker quality as well as pure measurement error, yet get a measure of the effect of incomplete employee information, is given in Hofler and Polachek (1985) and Polachek and Yoon (1987). The technique is relatively simple to implement. Basically, it measures worker incomplete information as a parameter obtained from estimating an earnings function using frontier estimation techniques. Essentially, this parameter depicts the degree workers receive wages that are less than they could obtain, had they known the specific firms paying the highest wages. Hofler and Murphy (1992) employ this technique to compute the average worker shortfall using the 1983 US Current Population Survey (CPS). Gaynor and Polachek (1994) apply the technique to compute incomplete information about physician prices. Daneshvary, Herzog, Hofler, and Schlottmann (1992) use the technique to get at assimilation of foreign workers in the US, and Lang (2004) uses it to get at assimilation of German immigrants. Groot and Oosterbeek (1994) corroborate the validity of the approach using a 1985 Dutch national sample of employees. Finally, Polachek and Robst (1998) confirm the technique’s power by showing how the technique’s incomplete information
40
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
measures match those independently obtained from the World of Work test administered to a group of workers in the 1966 US National Longitudinal Survey of Young Men (NLSYM). But to date, no one has systematically studied how the degree of incomplete information differs across countries. Part of the problem is the paucity of data for enough countries to be able to draw valid conclusions. Comparative cross-country micro-data were simply not available. However, new data are now obtainable from the LIS for a sufficient number of countries to get meaningful results. This paper applies the techniques mentioned above to estimate the amount of employee incomplete information. It then looks at different institutional factors across countries to determine how these institutional factors affect incomplete information.
APPROACH We use LIS data to examine incomplete information by country.4 The LIS is a collection of household data compiled from ongoing statistical surveys in approximately 29 countries widely spread across Europe, America, Asia, and Oceania. The data are standardized in order to facilitate comparative research. The LIS began in 1983 and is now jointly sponsored by the Luxembourg government and the Centre for Population, Poverty and Policy Studies (CEPS), the Centre Universitaire (CU) de Luxembourg. Data include country-specific labor force surveys over various labor market structures. These data provide demographic, income and expenditure information on three different levels: household, person and child. We concentrate on extracting earnings, education and age data from the LIS person files for 10 OECD countries. In addition, we include Israel, given that this is the location of the conference honoring Tikva Lecker. For each country, we use an econometric frontier estimation technique developed by Aigner, Lovell, and Schmidt (1977) and Meeusen and van den Broeck (1977), which is described more fully in Kumbhakar and Lovell (2000). This technique was originally employed by Hofler and Polachek (1985) to estimate a numeric index of incomplete worker information. Next, we look at institutional factors within countries to establish which of these institutions affect incomplete information. First standard institutions such as UI are considered. Then, we concentrate on population density, the proportion of rural employment, the proportion of industrial employment, and finally, the influx of foreign workers into an economy.
The Effects of Incomplete Employee Wage Information
41
MODELING INCOMPLETE INFORMATION Assume a labor market with a wage offer distribution such that wages for a worker of given skill vary from a low of wmin to a maximum of wmax.5 A worker entering this job arena is motivated to seek a job at the firm paying the highest possible wage. The only problem is that the worker does not know which firms pay high wages. Indeed, given that firm heterogeneity is not uniform across jobs (i.e., some firms may pay high wages for one particular skill, but low wages for another), wages vary across firms in uneasy to detect ways, a priori. Given this incomplete information, the worker must search. To do this, the worker sets a reservation wage wR i , which is the minimum wage the worker will accept when searching. This reservation wage is based on perceived costs and benefits of search. On average, a worker searches more the higher the reservation wage, and less the lower the reservation wage. The greater the search, the more likely the worker achieves a wage closer to the maximum wmax. Suppose that worker i finds a job paying a wage wi0, such that wmax 0 0 wi wR i ; The gap between wmax and wi will vary across individuals, depending upon relative success in the worker’s job search. The greater one’s knowledge of the labor market, the more likely the chance w0i is closer to, wmax barring the impact of luck (which we will talk about shortly). Let ui be the (logarithmic) gap between wmax and the wage one actually receives (w0i ). This gap represents the proportion by which one could enhance one’s current wage through continued search for the highest possible wage. The variable ui can be expressed as ui ¼ ln wmax
ln w0i þ oi
(1)
and reflects the cost to a worker of incomplete information. In (1), oi reflects luck, which for the sake of assumption can be normally distributed with zero mean. The average u ¼ Si ui =N; over all workers (N) in the market, depicts the mean effect of worker incomplete information. Note that since E ðoi Þ ¼ 0; luck cancels out when computing average incomplete information for all workers. Using similar logic, a firm pays more than necessary to the ith worker, since it pays w0i despite the worker’s willingness to work at wR i . The gap (ln w0i – ln wR ) 0 reflects the effect of the firm’s incomplete information i regarding worker i (neglecting luck). But, in reality, worker i need not be the lowest reservation waged worker, so the true effect of the firm’s incomplete information is Zi ¼ ln w0i – ln wmin, where wmin is the minimum reservation
42
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
wage for all comparable workers who satisfy the job’s requirements. The average Z ¼ Si Zi =N; across all workers (N), depicts the mean effect of firm incomplete information. Polachek and Yoon (1987) tested search theory’s implications by computing average u and Z for various groups of workers. They used the Panel Study of Income Dynamics (PSID) for the US in 1981 and found smaller u values for workers who received UI benefits before their current job. This finding illustrated the efficiency UI brings about through enhanced worker search. Similarly, using the January 1983 Current Populations Survey (CPS), Hofler and Murphy (1992, p. 516) found that workers in areas paying relatively higher unemployment benefits exhibit a smaller u. Groot and Oosterbeek (1994) found that ‘‘males have more labor market information than females [which is] probably caused by the greater market attachment of males, which makes the opportunity costs of ignorance [i.e., incomplete information] greater for males than females (p. 388).’’ They also found that ‘‘employees in the public sector possess more market information than workers in the private sector y probably due to the fact that wage policies in the private sector are in general less public knowledge and more individually based (p. 388).’’ Polachek and Robst (1998) found that workers scoring high on World of Work tests had more information than those with lower World of Work scores. Finally Gaynor and Polachek (1994) applied the technique to the medical profession, and there too found that ‘‘measured incomplete information is higher when y search costs are higher (p. 830).’’ But to date no one has done a comprehensive study across more than one country. Nor have they examined whether in-migration affects incomplete information.
ESTIMATION PROCEDURE The LIS data require all computations be done on the Luxembourg computer. At this time limitations on canned STATA and SAS software preclude estimating both employer and employee information simultaneously. However, Groot and Oosterbeek (1994) find employer incomplete information to be relatively constant across economic sectors, as does Polachek and Yoon (1987). This result means that incomplete employee information varies far more than incomplete employer information. So, because incomplete employer information varies little with institutional characteristics, we concentrate solely on incomplete worker information (ui).
43
The Effects of Incomplete Employee Wage Information
To put meat on the discussion, we now derive the econometric specification. However, first a word regarding the underpinnings of an individual’s maximum possible wage, wmax(x) defined earlier in note 5. Various models can describe the underlying factors of an individual’s wage potential. For example, the human capital model argues that a worker’s wage potential is determined by human capital acquisition. Productivity enhancing contract models suggest contracts whereby a worker’s ability to supply effort affects wage. Finally, collective bargaining implies that union power is important when negotiating a viable contract. Each of these underpinnings can be represented by a vector of an employee’s own individual and his or her firm’s characteristics, which we denote as vector xi.6 As such, a worker’s maximum potential wage (wpi ) is related to xi, plus measurement error.7 In logarithmic form, ln wmax jxi ¼ ln wpi ¼ gðxi Þ þ vi
(2)
Taking a linear model for g(xi), ln wpi ¼ xi b þ vi ,
(3)
where the dependent variable represents the individual’s maximum potential wage offer, xi denotes a vector of individual characteristics defining worker i ’s skill level, and vi is a disturbance distributed as N 0; s2v : With perfect information regarding what each firm pays, a worker would be able to locate the firm paying his or her maximum potential wage wpi : However, in a world of incomplete information, where search is costly, the typical worker defines a reservation wage wR i below his or her potential wage. As indicated above, the offer wage one accepts is w0i ZwR i : Combining (1) and (2), a worker’s observed wage can be represented as ln w0i ¼ ln wpi
(4)
ui
where ui Z 0. This one-sided residual ui represents the extent incomplete information causes one to accept a wage lower than the market’s potential wage. It reflects a monetary measure of incomplete information. Substituting (3) into (4) yields ln w0i ¼ xi b þ i
(5)
0; s2v 8
: We also assume that wherei ¼ vi ui : As already indicated, viN uiN mmi ; s2mi ; as is typical in frontier estimation. On the basis of the above assumptions, the composite error ( ¼ v u) must be rightward skewed for the approach to be valid. Therefore before proceeding, we test for skewness in : To do this, we follow Schmidt and Lin
44
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
(1984) to obtain residuals from OLS regressions of (5) for each country and year. The specific test9 we adopt was developed by Royston (1991) and is a test for normality, which combines two tests, one based on skewness and another based on kurtosis, into an overall test statistic. In no country or year do the test results support the normality assumption (with p-value less than 0.1%). Therefore, our hypothesis about the skewness of the residuals is strongly upheld justifying that we can proceed with estimation of (5) using frontier techniques to disentangle ui and vi. To do this, we adopt the maximum likelihood approach to estimate (5) that incorporates a composed error term first proposed by Aigner, Lovell, and Schmidt (1977) and Meeusen and van den Broeck (1977). It assumes a non-negative one-sided error term, ui, in addition to the traditional normally distributed error term, vi. To understand the approach rewrite (5) as w0i
w0i ¼ gðxi Þ expf ui þ vi g,
(6)
where is the actual observed earnings, as already defined and g(xi) the potential maximum earnings achievable barring random errors vi. Actual earnings (w0i ) are less than potential maximum earnings (g(xi)) by the proportion (1-exp{ ui}). In other words, (1-exp{ ui}) is the effect of incomplete information. As is standard in the literature, we specify the xi vector to adhere to typical earnings functions.10 On the basis of the current literature regarding functional form of earnings equations (Mincer, 1974; Heckman & Polachek, 1974; Murphy & Welch, 1990), we express an individual’s potential wage as gðxi Þ ¼ exp a0 þ a1 Si þ a2 ti þ a3 t2i þ a4 F i (7)
where Si is the individual worker i’s completed highest schooling level, ti the individual worker i’s potential working experience, t2i the quadratic term of i’s experience level, and Fi the individual worker i’s gender dummy, 1 for female and 0 otherwise. The above equation (7) gives a worker’s potential wage in the absence of the incomplete information. In reality, because of the limits of finite search, a worker receives less than his or her potential maximum wage, and gets an observed wage w0i : To incorporate incomplete information, substitute (7) into (6) to obtain w0i ¼ gðxi Þ expf ui þ vi g ¼ exp a0 þ a1 Si þ a2 ti þ a3 t2i þ a4 F i expð ui þ vi Þ (8)
45
The Effects of Incomplete Employee Wage Information
After taking the logarithm, it becomes ln w0i ¼ a0 þ a1 S i þ a2 ti þ a3 t2i þ a4 F i
ui þ v i
(9)
The common approach is to estimate (9) using the (ln) likelihood function: pffiffiffi N X 0 2 2 ln½1 F ði ls 1 Þ ln L wi x; b; l; s ¼ N ln pffiffiffiffi þ N ln s 1 þ P i¼1 N 1 X 2 2s2 i¼1 i
ð10Þ
proposed by Aigner, Lovell, and Schmidt (1977) and Meeusen and van den Broeck (1977), where s2 ¼ s2u þ s2v l¼
su sv
and f and F are, respectively, the standard normal density and distribution functions.11 The expected value of the composite error is pffiffiffi 2 E ðÞ ¼ E ðuÞ ¼ pffiffiffiffi su (11) P
the mean of the one-sided error term. A problem with the above approach is that it assumes ui to be unrelated to any of the independent variables. So, following Wang’s (2002) strategy, we take account of how the independent variables can influence incomplete information by parameterizing ui to be a function of gender, marital status, race, schooling and experience, because each of these variables affects a worker’s opportunity costs of search. Thus, we assume ui N þ mui ; s2ui (12) mui ¼ d0 þ d1 F i þ d2 M 1i þ d3 M 2i þ d4 Ri þ d5 S i þ d6 ti
(13)
s2ui ¼ expfg0 þ g1 F i þ g2 M 1i þ g3 M 2i þ g4 Ri þ g5 Si þ g6 ti g
(14)
vi N 0; s2v
(15)
where ui is truncated normally distributed with mean mui and variance s2ui ; M 1i is the individual worker i’s marital status, 1 for married and 0 otherwise, M 2i the individual worker i’s marital status, 1 for divorced, widowed,
46
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
separated and 0 otherwise, and Ri the individual worker i’s race, 1 for white and 0 otherwise. (For countries other than US, 1 for the majority ethnic group and 0 otherwise.) The empirical estimate of ui is obtained through its conditional expectation on the total error term i ; which is defined as i ¼
ui þ v i .
(16)
As noted by a referee, our choice of independent variables in (9), (13) and (14) is an important issue. Omitting key earnings determinants from (9) that are included in (13) and (14) can bias our estimates of incomplete information (m). Let us take an example. Suppose that education affects earnings positively. Then, including education in equation (9) implies lower potential earnings for the less educated and higher potential earnings for the more educated. On the other hand, omitting education from equation (9) implies a single potential earnings measure for all. But given the frontier estimation adopted in likelihood function (10), potential earnings in this latter case are those of the most educated group. As a result, the gap between the potential wage and the actual wage is smaller for the more educated and larger for the less educated. Thus, incorporating education in (13) but not in (9) would imply a smaller m for the most educated workers (hence less incomplete information) simply because the gap between the potential wage and the actual wage is smaller for the more educated when in reality frontier wages for the less educated should be lower than frontier wages for the more educated. In this paper, we adopt the Mincer earnings specification for (9). Here education, experience, and experience-squared identify potential earnings. In addition, we incorporate gender because gender is known to affect wage level. On the other hand, we include marital status and race in (13) and (14) but not in (9) because marital status does not affect wage uniformly. For women marital status appears to lower earnings, whereas for men it appears to raise earnings. Similarly, we incorporate race in (13) and (14) but not in (9). The LIS data (which we will describe later) includes majority ethnic group rather than race. For the US, this ethnicity variable denotes race, but not for the other countries. Blacks in the US are earning less than white people. However, there is no strong evidence outside the US that minority ethnic groups earn less than the majority ethnicity groups. Thus, we hesitate to include this ethnicity variable in the wage equation. As for the incomplete information equation, marriage and race are included because married, widowed/divorced, and majority ethnicity groups are assumed to have different search behaviors. We expect these groups to search more not only because their marginal gains for an additional search are higher, but also
The Effects of Incomplete Employee Wage Information
47
because their search costs are lower. The determinants of search do not always perfectly match those of earning patterns and therefore, the variables in the two equations are not required to be the same. We admit some biases can creep in because of difficulty in specifying the variables that affect m but not w, and vice versa. Mostly for this reason, later in the paper, we adopt an additional approach to identify country institutional factors that affect m. In this latter approach (comparable to differences-in-differences), the above type errors cancel out as long as they are uniform across countries. Maximum likelihood is used to estimate equation (17), given assumptions (12) to (16) (see Stevenson, 1980). The log-likelihood function for a sample of N workers is X X X m m i li ln L w0i jb; l; s ¼ C ln F ui ln si ln F ui þ sui si li si i i i 2 1 X i þ mui ð17Þ 2 i si where
1=2 si ¼ s2ui þ s2v li ¼ sui =sv
(18) (19)
As discussed above, we use 1–E ðexpf ui gji Þ as a measure of a worker’s incomplete information. The advantage of this measure is that it is bounded by (0,1), which is easily interpreted as the proportion of the potential maximum wage a worker gives up due to the incomplete information. So, the bigger the 1 E ðexpf ui gji Þ; the more incomplete one’s knowledge of available wages. The formula to obtain each worker’s incomplete information is
1 F si ðm~ i =si Þ 1 2 (20) 1 E ðexpf ui gji Þ ¼ 1 exp ð mi Þ þ si 2 1 F m~ i =si
where
s2ni ¼ s2ui s2v =s2i
mi ¼
s2ui i þ mui s2v =s2i
(21) (22)
The estimate of incomplete information is the average of all workers’ incomplete information. So each country year’s incomplete information is
48
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG N
mj ¼
j 1 X ½1 N j i¼1
E ðexpf ui gji Þ
(23)
where Nj is the country’s total number of workers in year j.
THE DATA Our first task is to compute each worker’s incomplete information defined in equation (20), using the maximum likelihood estimation formulated in equation (17). In the LIS data, there are three data files for each country and year. We use the personal file, which has information on work status, personal income, education as well as other basic individual characteristics. For each of the 29 countries, the number of available years differs, ranging from one year for Estonia to nine years for Canada and the United States. For two reasons we use only part of the available countries for the analysis. First, our final goal is to understand how institutional differences explain variations in incomplete information across countries. However, crucial variables to test hypotheses related to institutional perspectives are not available for all LIS countries and years. Generally, LIS has sufficient data for many, but not all OECD countries. So for this reason we first restrict our analysis to the OECD countries and years, which contain the full complement of demographic information that we need for analysis. (As mentioned before, we also include Israel – our only non-OECD country – because LIS contains the relevant information, and because Israel is the site of the Tikva Lecker conference.) Second, we only include countries and years for which the nonlinear maximum likelihood estimation of equation (17) converged. Of the 23 potential OECD countries (and Israel), this left 11 countries for which we have sufficient data. They are the United States, Canada, the United Kingdom, Germany, Sweden, Finland, Ireland, Norway, Netherlands, Czech Republic and finally Israel. They span North America and all parts of Europe.
WITHIN COUNTRY REGRESSION RESULTS The summary statistics are provided in Table 1. For most countries and years, the average schooling is around 11–12 years, and the average potential experience is between 20 and 25 years. So, across these OECD countries,
49
The Effects of Incomplete Employee Wage Information
Table 1.
Variable Means from the Individual Country-Year Regressions.
Potential Majority Divorced Country Year School Experience Female Ethnicity Married Widowed CAN CAN CAN CAN CAN CAN CAN CZECH CZECH FIN FIN FIN FIN GER GER GER GER IRE IRE IRE IS IS NL NL NL NL NL NW NW NW SW SW SW UK UK UK UK US US US US US
1981 1987 1991 1994 1997 1998 2000 1992 1996 1987 1991 1995 2000 1984 1989 1994 2000 1994 1995 1996 1992 1997 1983 1987 1991 1994 1999 1991 1995 2000 1992 1995 2000 1986 1991 1994 1995 1969 1974 1979 1986 1991
11.117 10.822 11.646 11.984 12.066 12.492 13.288 11.804 10.413 9.351 9.706 11.145 11.693 7.638 8.483 8.780 12.607 9.481 9.471 9.492 12.531 12.019 9.881 9.417 11.615 10.868 12.693 11.485 11.804 12.734 11.144 10.740 12.207 10.482 10.744 10.283 11.094 11.491 11.327 12.804 12.227 12.429
21.813 24.353 25.152 25.164 26.507 25.488 18.450 20.900 23.784 22.222 22.374 21.195 22.208 29.046 28.992 29.558 27.399 29.716 30.007 29.905 19.813 22.691 22.597 20.573 17.760 26.795 19.441 19.930 21.152 20.755 25.180 24.248 22.140 21.153 21.295 31.504 22.043 20.311 23.539 20.336 23.362 24.408
0.172 0.519 0.517 0.521 0.513 0.513 0.487 0.471 0.466 0.474 0.486 0.493 0.491 0.511 0.504 0.519 0.516 0.500 0.501 0.495 0.452 0.514 0.102 0.361 0.396 0.526 0.463 0.470 0.479 0.486 0.492 0.495 0.491 0.429 0.478 0.528 0.504 0.421 0.535 0.472 0.535 0.532
0.124 0.126 0.135 0.130 0.229 0.201 0.103 0.108 0.073
0.939 0.965 0.973 0.961
0.618 0.636 0.619 0.646 0.487 0.529 0.733 0.694 0.614 0.629 0.586 0.584 0.638 0.634 0.649 0.629 0.628 0.621 0.611 0.755 0.602 0.793 0.677 0.64 0.648 0.612 0.554 0.571 0.551 0.802 0.765 0.501
0.885 0.815
0.659 0.615 0.649 0.658 0.610
0.140 0.225 0.180 0.099 0.151
0.785 0.745
0.554 0.540
0.175 0.195
0.992 0.933 0.939 0.948 0.942 0.821 0.804 0.813 0.937 0.997 0.993 0.995 0.473 0.431
0.106 0.082 0.125 0.132 0.135 0.196 0.105 0.108 0.116 0.058 0.108 0.053 0.046 0.048 0.113 0.071 0.061 0.074 0.082
0.084
Wage
Sample Size
10.452 12.131 15.110 16.020 14.716 15.256 16.080 518.575 1140.721 63344.97 85465.26 88175.4 112289.7 19.146 25.493 34.482 24.574 5.035 5.217 5.239 38082.47 64932.53 21.685 41329.14 40200.91 26.229 29.216 115304.8 150085.6 209902.6 145644.9 157548.8 189576 8121.4 6.277 7.295 7.474 5566.403 4.650 6.703 9.737 10.467
9063 11976 20514 39635 33052 36566 32548 18910 31435 18793 17444 12822 14137 4844 3914 5849 5537 2762 2396 2327 5429 5817 2162 3054 3828 4707 5110 12786 13538 19214 15623 16828 17235 6251 6398 21659 5472 15715 12087 15470 13048 19636
50
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
Table 1. (Continued ) Potential Majority Divorced Country Year School Experience Female Ethnicity Married Widowed US US US
1994 12.598 1997 12.675 2000 12.818
24.130 25.060 25.219
0.531 0.532 0.521
0.733 0.719 0.701
0.517 0.549 0.581
0.203 0.194 0.158
Wage
Sample Size
11.962 79323 13.795 64369 15.320 63890
Note: Variable definitions: School is average years of schooling; Potential experience is average years of working experience, calculated by age-school-6; Female is a gender dummy, 1 for female and 0 otherwise; Majority ethnicity is ethnicity dummy, 1 for majority ethnic group and 0 otherwise; Married is marital status dummy, 1 for married and 0 otherwise; Divorced/ widowed is a marital status dummy, 1 for divorced/widowed and 0 otherwise; Wage is average wage, either specified as hourly wage or annual earnings. Source: Luxembourg Income Study. Denotes countries with hourly wage; otherwise with annual wage data are given. The mean is based on those in the population with positive wages (earnings). Missing variables are denoted as blanks.
most workers have relatively comparable educational backgrounds, and are similar in age. Also, female workers constitute about half of the population interviewed. (Though, not shown in the table, a greater proportion of women did not work for pay.) The ethnicity variable depicts the proportion of the population constituting a country’s majority racial or ethnic group. In the US, this is the proportion of whites. Married workers are around 60% of the population, and divorced or widowed workers are in the 5–20% range. Mean wages are contained in the last column and are defined either as hourly wages or annual earnings, depending on the available data. We expect more variation in annual earnings because hourly wages fluctuate less as work hours vary. In the statistical analysis to follow, we add a dummy categorical variable to signify country-years with hourly wage data. Table 2 contains the regression results of equation (9). Columns (1) and (2) give the country and year; columns (3)–(6) give the coefficient values for schooling, potential experience, potential experience-squared, and gender; and finally column (7) gives estimates of the extent of incomplete information obtained from equation (23). We begin by explaining columns (3)–(6), since they reflect earnings function parameters typically obtained when estimating Mincer earnings functions. The schooling variable coefficient depicts the average rate of return for an additional year of school. For more than half of the countries, this rate of return is increasing over time. (Only a few counties have decreasing rates of return.) This pattern is consistent with other data sets as well as with
51
The Effects of Incomplete Employee Wage Information
Table 2. Within Country Regression Results. Country
Year
School
CAN CAN CAN CAN CAN CAN CAN CZECH CZECH FIN FIN FIN FIN GER GER GER GER IRE IRE IRE IS IS NL NL NL NL NL NW NW NW SW SW SW UK UK UK UK US US US US US US
1981 1987 1991 1994 1997 1998 2000 1992 1996 1987 1991 1995 2000 1984 1989 1994 2000 1994 1995 1996 1992 1997 1983 1987 1991 1994 1999 1991 1995 2000 1992 1995 2000 1986 1991 1994 1995 1969 1974 1979 1986 1991 1994
0.050 0.066 0.067 0.070 0.080 0.103 0.098 0.077 0.093 0.082 0.083 0.067 0.065 0.089 0.073 0.067 0.105 0.098 0.087 0.091 0.105 0.128 0.083 0.084 0.064 0.060 0.065 0.074 0.069 0.073 0.068 0.076 0.092 0.093 0.141 0.156 0.161 0.087 0.076 0.074 0.092 0.098 0.113
Potential Experience 0.024 0.031 0.021 0.024 0.024 0.036 0.006 0.023 0.020 0.066 0.056 0.043 0.043 0.030 0.021 0.013 0.020 0.041 0.038 0.038 0.048 0.042 0.048 0.061 0.038 0.025 0.026 0.042 0.046 0.046 0.037 0.041 0.036 0.046 0.028 0.028 0.018 0.028 0.028 0.018 0.035 0.030 0.035
Experience Square
Female
0.0003 0.0004 0.0002 0.0002 0.0002 0.0004 0.0002 0.0004 0.0003 0.0009 0.0007 0.0005 0.0005 0.0004 0.0002 0.0001 0.0002 0.0004 0.0004 0.0004 0.0006 0.0005 0.0006 0.0008 0.0005 0.0002 0.0003 0.0007 0.0007 0.0008 0.0005 0.0006 0.0005 0.0007 0.0004 0.0003 0.0001 0.0004 0.0004 0.0002 0.0004 0.0004 0.0004
0.276 0.205 0.211 0.202 0.204 0.322 0.314 0.324 0.321 0.413 0.383 0.381 0.421 0.243 0.169 0.151 0.187 0.116 0.128 0.110 0.486 0.514 0.081 0.172 0.239 0.046 0.058 0.416 0.436 0.460 0.437 0.443 0.392 0.396 0.405 0.390 0.368 0.533 0.390 0.476 0.292 0.275 0.257
Incomplete Information 0.281 0.377 0.347 0.348 0.347 0.251 0.340 0.259 0.274 0.535 0.536 0.563 0.563 0.252 0.267 0.281 0.357 0.324 0.312 0.353 0.320 0.341 0.156 0.256 0.398 0.275 0.297 0.581 0.527 0.484 0.445 0.469 0.471 0.304 0.187 0.207 0.204 0.495 0.345 0.288 0.365 0.405 0.396
52
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
Table 2. (Continued ) Country
Year
School
US US
1997 2000
0.118 0.116
Potential Experience 0.034 0.027
Experience Square 0.0004 0.0003
Female 0.301 0.307
Incomplete Information 0.384 0.384
Note: Variable definitions: School denotes the rate of return to an additional year of schooling in equation (9); Potential experience is experience coefficient in equation (9); Experience square is experience square coefficient in equation (9); Female is the female dummy variable coefficient in equation (9), 1 for female and 0 otherwise; As computed from equation (23), incomplete information is expressed as a percentage of potential maximum wages. Source: Luxembourg Income Study. The dependent variable is hourly wage; otherwise the dependent variable is annual earnings. Coefficients are not statistically significant at 5% p-value level; otherwise are statistically significant at least 5% p-value level.
technological change. It implies that the more educated absorb advanced technology easily, and that their rate of pay per year of schooling is increasing secularly.12 The experience coefficients yield concave earning profiles, shown by the consistently negative squared-term. The female dummy variable is uniformly negative, suggesting that in all the countries (that we consider), women earn less than men, given adjustments for schooling and experience. Overall, these earnings function parameters are typical of those found in the literature. Estimates of incomplete information, based on equation (23), are given in Column 7. They range from 0.16 to 0.58, but average about 0.3. This means that incomplete information causes the average worker to get about 30% less than his or her potential. An examination of the values indicates strong consistency within countries, since these values differ more country-tocountry than within countries. Within countries, incomplete information measures do not reveal any apparent time pattern. Incomplete information seems to be increasing in Sweden, Czechoslovakia, Finland, Germany, Netherlands and Israel, while in Norway it is decreasing. On the other hand, time trends are relatively flat for the United States, the United Kingdom, Canada and Ireland. Usually technological improvements mirror time trends. The lack of a time trend might suggest that technological improvements are not necessarily associated with decreasing incomplete information, as common sense might have implied.13 What factors affect the level and distribution of incomplete information within each country? To answer this question, we adopt Wang’s (2002)
The Effects of Incomplete Employee Wage Information
53
method to compute marginal effects of mui ’s covariates listed in equation (13) (Table 3) and s2ui ’s covariates given in equation (14) (Table 4).14 Beginning with Table 3, we see a number of trends. For example, the mostly negative marginal effects of school indicate that additional schooling reduces a worker’s incomplete information. This finding is consistent with Stephenson’s (1976) argument that workers with more education gather more wage information by searching more efficiently. Similarly, about threequarters of the cases show married workers, and two-thirds of the cases show that widowed and divorced workers, have more information than singles. This result is consistent with higher married and widowed labor force participation rates (Taubman, 1976), which leads to larger marginal gains from search. Further, as hypothesized (McCall, 1973), blacks exhibit lower labor force participation and possibly higher search costs, so they might acquire less information than whites. In our international data, we show that on average, minority ethnic groups have less complete information than majority groups, which is an extension of McCall’s argument. Also, from Table 3, potential experience reduces workers’ incomplete information in over 80% of the observations. Finally, incomplete information is larger among female workers than among male workers, again consistent with less lifetime female labor force participation. Another observation regarding these marginal effects in Table 3 is that the patterns are more consistent within a specific country, than across countries. As such, any covariate’s marginal effect is likely to have a uniform sign within a particular country, but not necessarily across countries. This uniformity within countries underscores the importance of using crosscountry institutional differences to explain how incomplete information differs from country-to-country. Recall from equation (5) that s2ui depicts the dispersion of incomplete information. How individual characteristics affect this dispersion is parameterized in equation (15). The impacts of these characteristics on s2ui are given in Table 4. They indicate the degree an incomplete information fluctuates across socioeconomic groups. As an example, take race. From Table 3, incomplete information is smaller for whites than for blacks. But, the negative ethnicity coefficient in Table 4 implies that whites exhibit smaller dispersion in incomplete information than blacks. This implies a relatively wider range in incomplete information for blacks than whites. Thus the variance of incomplete information is greater for blacks than whites. Similarly, though on average females garner less information, in total, they exhibit greater variance in the amount of information they gather,
54
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
Table 3.
Marginal Effect of Covariates on E(U) by Country and Year.
Country
Year
Female
CAN CAN CAN CAN CAN CAN CAN CZECH CZECH FIN FIN FIN FIN GER GER GER GER IRE IRE IRE IS IS NL NL NL NL NL NW NW NW SW SW SW UK UK UK UK US US US US US
1981 1987 1991 1994 1997 1998 2000 1992 1996 1987 1991 1995 2000 1984 1989 1994 2000 1994 1995 1996 1992 1997 1983 1987 1991 1994 1999 1991 1995 2000 1992 1995 2000 1986 1991 1994 1995 1969 1974 1979 1986 1991
0.139 0.100 0.135 0.093 0.120 0.049 0.059 0.034 0.073 0.262 0.236 0.142 0.173 0.095 0.165 0.083 0.089 0.219 0.170 0.256 0.132 0.040 0.069 0.198 0.486 0.227 0.141 0.230 0.170 0.119 0.003 0.019 0.086 0.247 0.107 0.023 0.058 0.452 0.154 0.015 0.136 0.175
Married
Divorced/ Widowed
0.107 0.143 0.021 0.098 0.060 0.022 0.004 0.006 0.245
Majority Ethnicity
0.162 0.077 0.088 0.076 0.057 0.077 0.004 0.026 0.038 0.081 0.097 0.134 0.132 0.045 0.053 0.032 0.267 0.270 0.276 0.042 0.156 0.236 0.162 0.958 0.713 0.577 0.031 0.065 0.011 0.086 0.112 0.031
0.119
0.152 0.184 0.389 0.336
0.042 0.071 0.020 0.200 0.147
0.049 0.079 0.005 0.183 0.080
0.228 0.072
0.201 0.128
0.183 0.006
0.069 0.088
0.032 0.067 0.344 0.142 0.032 0.067 0.198 0.041 0.068 0.005 0.088 0.389 0.107 0.804 0.930 0.196 0.429 0.350 0.067
0.119 0.016 0.018 0.032 0.040 0.059 0.059 0.138 0.035 0.298 0.258 0.063 0.162 0.215
School 0.005 0.106 0.016 0.012 0.007 0.027 0.012 0.002 0.004 0.081 0.076 0.059 0.061 0.010 0.014 0.008 0.006 0.008 0.008 0.007 0.034 0.035 0.012 0.015 0.044 0.014 0.021 0.055 0.054 0.049 0.025 0.034 0.003 0.014 0.034 0.050 0.048 0.046 0.006 0.005 0.013 0.017
Potential Experience 0.002 0.006 0.008 0.010 0.011 0.009 0.012 0.015 0.010 0.016 0.034 0.044 0.036 0.042 0.039 0.027 0.019 0.003 0.001 0.007 0.001 0.018 0.006 0.009 0.055 0.013 0.028 0.043 0.050 0.031 0.006 0.034 0.058 0.006 0.001 0.004 0.001 0.056 0.028 0.002 0.021 0.020
55
The Effects of Incomplete Employee Wage Information
Table 3. (Continued ) Country
Year
Female
Married
US US US
1994 1997 2000
0.134 0.093 0.073
0.134 0.128 0.157
Divorced/ Widowed 0.005 0.015 0.035
Majority Ethnicity 0.067 0.067 0.069
School
Potential Experience
0.012 0.008 0.012
0.020 0.020 0.020
Note: Definitions of variables: Female is the gender dummy variable in equation (13), 1 for female and 0 otherwise; Married is the marital status dummy variable in equation (13), 1 for married and 0 otherwise; Divorced/widowed is the marital status dummy variable in equation (13), 1 for divorced/widowed and 0 otherwise; Majority ethnicity is the race/ethnicity dummy in equation (13 ), 1 for majority ethnic group and 0 otherwise; School is the years of schooling variable in equation (13); Potential experience is the years of working experience in equation (13), calculated as age–school–6. Source: Luxembourg Income Study. Denotes computations based on hourly wage; otherwise computation based on with annual earnings. Marginal effect for each country year is the average marginal effect of individual workers within that country year.
as well. For the most part being older (having more potential experience) reduces this dispersion, as does being married or widowed, and as does having more education (except for the United Kingdom, Germany, The Netherlands and Israel).
WHY INCOMPLETE INFORMATION DIFFERS ACROSS COUNTRIES? AN EXAMINATION OF INSTITUTIONAL FACTORS So far we have estimated incomplete information and examined how it varies within countries. We have seen that an employee’s characteristics, especially those characteristics affecting one’s incentive to search, influence the amount of wage related information one acquires. However, a country’s institutions may also be important, but identifying the impact of these institutions is difficult to discern with a limited number of cross-sections for a given country. For this reason, we now do a comparative analysis by contrasting institutional differences across each of the 11 countries to explain inter-country differences in incomplete information.
56
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
Table 4.
Marginal Effect of Covariates on V(U) by Country and Year.
Country
Year
Female
CAN CAN CAN CAN CAN CAN CAN CZECH CZECH FIN FIN FIN FIN GER GER GER GER IRE IRE IRE IS IS NL NL NL NL NL NW NW NW SW SW SW UK UK UK UK US US US US US
1981 1987 1991 1994 1997 1998 2000 1992 1996 1987 1991 1995 2000 1984 1989 1994 2000 1994 1995 1996 1992 1997 1983 1987 1991 1994 1999 1991 1995 2000 1992 1995 2000 1986 1991 1994 1995 1969 1974 1979 1986 1991
0.119 0.074 0.190 0.115 0.159 0.004 0.001 0.023 0.062 0.741 0.653 0.380 0.475 0.128 0.174 0.099 0.116 0.364 0.324 0.545 0.148 0.163 0.031 0.539 0.978 0.332 0.130 0.527 0.352 0.199 0.003 0.041 0.188 0.511 0.001 0.015 0.002 0.986 0.210 0.012 0.177 0.429
Married
Divorced/ Widowed
0.208 0.209 0.030 0.155 0.067 0.001 0.001 0.004 0.485
Majority Ethnicity
0.338 0.117 0.145 0.117 0.076 0.009 0.012 0.005 0.230 0.319 0.197 0.420 0.011 0.060 0.093 0.171 0.365 0.308 0.450 0.043 0.122 0.081 0.110 0.343 0.360 0.448 0.567 0.071 0.074 0.167 0.219 0.090
0.042
0.281 0.393 0.761 0.600
0.008 0.079 0.012 0.008 0.206
0.008 0.080 0.002 0.046 0.112
0.402 0.099
0.493 0.213
0.424 0.018
0.133 0.146
0.112 0.015 0.048 0.097 0.097 0.145 0.263 0.052 0.097 0.006 0.216 0.142 0.125 0.211 0.480 0.144 0.946 0.278 0.030
0.107 0.005 0.000 0.010 0.236 0.051 0.051 0.130 0.047 0.444 0.400 0.148 0.179 0.381
School 0.004 0.029 0.025 0.022 0.016 0.013 0.007 0.002 0.002 0.138 0.127 0.002 0.003 0.023 0.014 0.016 0.003 0.012 0.010 0.012 0.038 0.063 0.005 0.020 0.032 0.007 0.017 0.056 0.053 0.058 0.057 0.048 0.013 0.020 0.014 0.038 0.022 0.013 0.009 0.004 0.039 0.006
Potential Experience 0.002 0.014 0.015 0.022 0.026 0.005 0.003 0.020 0.017 0.031 0.006 0.051 0.050 0.006 0.006 0.001 0.001 0.007 0.010 0.019 0.001 0.001 0.003 0.014 0.022 0.002 0.022 0.001 0.024 0.021 0.025 0.041 0.052 0.013 0.002 0.009 0.006 0.012 0.039 0.001 0.057 0.001
57
The Effects of Incomplete Employee Wage Information
Table 4. (Continued ) Country
Year
Female
Married
US US US
1994 1997 2000
0.287 0.188 0.154
0.224 0.224 0.244
Divorced/ Widowed 0.039 0.044 0.085
Majority Ethnicity 0.109 0.120 0.109
School 0.001 0.005 0.004
Potential Experience 0.012 0.024 0.011
Note: Definitions of variables: Female is the gender dummy variable in equation (14): 1 for female and 0 otherwise; Married is the marital status dummy variable in equation (14): 1 for married and 0 otherwise; Divorced/widowed is marital status dummy variable in equation (14): 1 for divorced/widowed and 0 otherwise; Majority ethnicity is race/ethnicity dummy variable in equation (14): 1 for majority ethnic group and 0 otherwise; School is the years of schooling in equation (14): Potential experience is the years of working experience in equation (14), calculated by age-school-6. Source: Luxembourg Income Study. Denotes computations based on hourly wage; otherwise computation based on with annual earnings. Marginal effect for each country year is the average marginal effect of individual workers within that country year.
UI and Incomplete Information Perhaps UI is the most studied institution regarding information and search. Ehrenberg and Oaxaca (1976), Jurajda and Tannery (2003), and numerous other studies find that UI increases unemployment duration, job search and post-unemployment wage. Hofler and Polachek (1985), Polachek and Yoon (1987) and Hofler & Murphy (1992) corroborate this result using one of the frontier estimation techniques described above. They find that having received UI leads to less incomplete information. As already mentioned, this finding is consistent with search theory explanations that UI subsidizes search costs leading to longer search, better wages, and more information (less incomplete information). Whereas these studies test this proposition with US data, to the best of our knowledge, none examine other countries; nor do any do comparative analysis across countries. In what follows, we perform a comparative analysis using the eleven countries mentioned above. We test whether variations in UI are related to worker incomplete information. To achieve this goal, we need UI information for these countries. One measure, found in the ‘‘OECD Employment Outlook,’’ is a country’s public expenditure as a percentage of GDP. The advantage of this measurement is that we have each country’s actual expenditures on UI, instead of some
58
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
nominal benefit measure that might not be implemented exactly for each worker. The disadvantage is that this measure might reflect a country’s business cycle rather than how it subsides an individual’s search. Because UI (as a proportion of GDP) is narrowly distributed (most of the UI expenditures are less than 2% of countries’ GDP, with a few over 3%), we convert the UI variable into a logarithm. This assumption is consistent with UI’s impact being nonlinear. A logarithmic specification implies a larger impact when UI is initially small, which is what we expect.
Other Institutional Factors Besides UI, we examine several other institutional factors that might contribute to the explanation of workers’ incomplete information. These variables are population density, proportion of employment in industry, rural population and the inflow of foreign workers.15 Information on each of these is obtained from World Bank data. The first three variables get at how information is concentrated among the population. As such they reflect the costs of search, since we presume that search costs rise when information is dispersed more widely and harder to find. So, for example, a more dense population implies quicker access to information networks. Also, a more dense population probably means that jobs are closer in proximity. A large rural population implies the opposite, namely sparse harder to find information, with jobs spread over wider distances. Sandell (1980) suggests that geographically concentrated opportunities lowers search costs, which prolongs search, and reduces incomplete information. In contrast to urban areas, rural regions are less concentrated with job opportunities, and therefore likely result in more incomplete information. Unionization rates rise as a country’s industrial employment increases (Polachek, 2004). Also unions provide information to employers and employees regarding wages and jobs (Polachek & Yoon, 1987) and unions compress wage distributions (Freeman, 1980). Thus, because industrialized countries are more unionized, we expect workers in more industrial countries to be more informed and have less incomplete information. Finally, in honor of Tikva Lecker, we examine the inflow of foreign workers. A considerable body of research examines the relative success of immigrants. For example, past studies of US immigration note that wages of newly arrived immigrants lag behind native wages (Chiswick, 1978), but that the assimilation process can cause earnings of the foreign born to eventually
The Effects of Incomplete Employee Wage Information
59
overtake US natives (Chiswick, 1986), though there remains some debate on the issue (Borjas, 1985). In any case, the whole assimilation process involves increased acquisition of information on domestic labor markets, and the use of this information in the search process (Daneshvary, Herzog, Hofler, & Schlottmann, 1992). Further, new lower waged foreign workers, who come to a country with less initial knowledge of wage structures, can affect the overall distribution of wages. Since generally, having less information implies that immigrant workers receive lower wages than natives, wage dispersion increases, as does skewness. Thus the effect of foreign workers is likely to be a more left-skewed earnings distribution, resulting from incomplete information (holding skills constant). But this result is not always found. A study of German immigrants, using data from 2000, finds natives and immigrants at about the same distance from the frontier (Lang, 2004). For this reason, the effect of foreign in-migration on incomplete information is still an open question. Table 5 gives a detailed summary of these variables. Column (1) contains the variable name and definition. Column (2) contains the number of country-year cells for which data are available. Columns (3)–(6) contain summary statistics. Rows (1)–(7) contain statistics for each of the institutional variables just mentioned. Generally, data are available for most of the time periods and countries. The exceptions are for UI (which is missing 8 of 45 observations) and foreign worker inflows (which is missing 24 observations). This latter restriction necessitates running separate analysis of foreign worker inflow effects. Row (8) gives estimates of the incomplete information variable (obtained from Table 2). Because the effects of incomplete information are affected by using annual instead of hourly wages, we normalize those 18 observations that were based on annual instead of hourly wages. This normalization unifies the incomplete information measurement to make it consistent across counties so that inter-country comparisons can be made. These values are given in row (10).16
Cross-Country Analysis Cross-country regression results are given in Table 6. Country (and year) specific incomplete information measures derived from (23) are the dependent variables, and the country-specific institutional variables just described serve as independent variables.17 We present five models that explain incomplete information as a function of these institutional factors. Model (1) concentrates on UI. Model (2) concentrates on the population density and
60
Table 5.
Statistic Summary of the Variables from the Cross Country Regressions.
Unemployment insurance (expenditure as a percentage of GDP) Log of unemployment insurance Population density (people per sq km) Industrial employment (% of total employment) Rural population (% of total population) Inflow of foreign worker (in 1000s) Inflow of foreign worker (% of total population) Incomplete information Hourly wage dummy (1 for hourly Wage) Adjusted incomplete information
Number of Observations
Mean
Standard deviation
Minimum
Maximum
37 37 45 42 45 21 21 45 45 45
1.779 0.352 28.542 27.764 22.107 77.700 0.161 0.359 0.600 0.309
1.181 0.722 147.464 5.644 9.822 184.707 0.108 0.108 0.495 0.088
0.26 1.347 2.7000 21.5 8.841 3.8 0.033 0.156 0 0.130
5.6 1.723 466.499 45.1 42.283 742.3 0.412 0.581 1 0.455
Note: Industry includes mining and quarrying (including oil production), manufacturing, electricity, gas and water, and construction, corresponding to major divisions 2-5 (ISIC revision 2) or tabulation categories C-F (ISIC revision 3). Source: Incomplete information is from Table 2; UI is from OECD Employment Outlook; the remaining variables are from World Bank data. Median
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
Variable
Regression Results of Country Attributes on Country Workers’ Average Incomplete Information (Robust Standard Errors Used; t-Values in Parentheses). Model 1
Constant Wage definition dummy Log of UI
Population density
Rural population
Industrial employment
Inflow of foreign workers(in thousands)
Inflow of foreign workers (proportion of total population)
0.487 (20.62) 0.161 ( 5.82) 0.050 ( 3.82) [ .036] { .048}
Model 2 0.570 (8.94) 0.128 ( 6.86)
Model 3
Model 4
Model 5
0.287 (17.01)
0.279 (10.23)
0.414 (12.00) 0.138 ( 7.12) 0.022 ( 1.99) [ .016] { .021} 0.0002 ( 1.70) [ .028] { .015} 0.003 (3.05) [.031] {.180}
0.0002 ( 2.63) [ .030] { .016} 0.003 (2.65) [.030] {.184} 0.006 ( 2.64) [ .034] { .464}
The Effects of Incomplete Employee Wage Information
Table 6.
0.0002 (3.30) [.037] {.049}
61
0.226 (2.03) [.023] {.116}
N R-squared Probability4F
62
Table 6. (Continued ) Model 1
Model 2
Model 3
Model 4
Model 5
37 0.5185 0.0000
42 0.7301 0.0000
21 0.2922 0.0038
21 0.1391 0.0561
37 0.7787 0.0000
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
Note: [ ] contains the marginal effect of a standard deviation change in the variable. { }contains the elasticity defined as: Elasticity ¼ @y=¯y ; where y¯ and x¯ are the means (or medians indicated by ) of the dependent variable and @x=x¯ the independent variable respectively.
The Effects of Incomplete Employee Wage Information
63
industrial structure variables. Models (3) and (4) concentrate on the inflow of foreign workers. Finally, Model (5) integrates most of the institutional factors into one model.18 We present four parameters for each independent variable. The first is the simple OLS coefficient. The second is the t-statistic based on the robust standard error. The third is the marginal effect caused by a one standard deviation change in the independent variable. Finally, the fourth is an elasticity measure describing the percent change in incomplete information caused by a one percent change in the independent institutional characteristic. To distinguish between incomplete information variables derived from annual versus incomplete information measures derived from hourly wage data, we adjust for hours of work.19 The first model focuses on how UI impacts incomplete information. A one-unit increase in the logarithm of a country’s UI expenditures (relative to GDP) reduces incomplete information by 0.05. This means that a one extra logarithm unit of UI spending (relative to GDP) induces workers to get 5% closer to their maximum attainable wage. The –0.036 coefficient implies that a one standard deviation increase in a country’s logarithm of UI (relative to GDP) causes workers to get 3.6% closer to their potential wage. Finally, the –0.048 elasticity measure indicates that a doubling of the logarithm of UI spending (from its mean value) leads to 4.8% decrease in worker incomplete information (from its mean value). Most important, these coefficients substantiate search theory. Increasing UI reduces search costs. Lower search costs increase search leading to more worker information on the wage structure. The second model focuses on population density, rural population, and industrial employment. If, as we postulate, each of these influences search costs, then we should expect these variables to have an effect on worker incomplete information. Indeed, this is precisely what we find. Both a more dense population and a high level of industrial employment provide information networks, and thus decrease incomplete information. A large rural population increases distances, thereby increasing information costs. This leads to decreases in employee information, and hence a greater level of employee incomplete information. The marginal effect of a one standard deviation change in each of these three variables is very similar in magnitude (0.03). This means that a one standard deviation increase in either population density or industrial employment decreases a country’s incomplete information by about 10%.20 Analogously, one could look at this in another way: Incomplete information of about 0.30 (the mean reported in Table 5) implies that employees on average earn about 70% of their maximum possible wage. A one standard
64
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
deviation increase in population density raises employee information (lowers employee incomplete information) by about 10%. Thus, workers in (a one standard deviation) denser area get wages 3% closer to their maximum. Though this analysis is performed across countries, it is well known that urban areas pay higher wages, and often for unexplainable reasons. Perhaps one reason is that better information networks lead to less employee incomplete information.21 Models (3) and (4) incorporate the inflow of foreign workers. Both these models show that a growing foreign workforce decreases worker information. A standard deviation increase in inflow of foreign workers yields a 3.7% wage loss in model (3) and a 2.4% loss in model (4). The elasticity measurements for these two models are 0.049 and 0.116, respectively. Thus doubling the inflow of foreign workers induces a 4.9% increase in incomplete information in model (3), and a 11.6% increase in model (4).22 Thus, despite major contributions, foreign workers probably contribute to a host country’s economic development, they also affect the wage distribution, which becomes more skewed to the left. As such, the average worker receives a wage proportionally lower than the maximum available wage. These lower wages are consistent with an increase in incomplete information. Combining all variables (Model (5)) does not alter the results.23 Again, UI is associated with significantly increased worker wages, relative to potential wages. This is consistent with increased worker information. Also, a more dispersed population is associated with wages being lower, relative to potential wages. In contrast with UI, this is consistent with less employee information. From a policy perspective, these results show that a country can reduce its incomplete information through governmental efforts, such as spending more on UI. Also, from a policy perspective, the results are consistent with information dissemination being linked to economic development. As a country develops, its geographical structure often changes, so that rural (relative to urban) population is reduced. From the above coefficients, relative rural population reductions are associated with more worker information. In making these claims, it is important to note that information rises because the overall wage distribution becomes more left-skewed, not because wages simply rise equally across the whole distribution. Thus, we are not making a trivial statement regarding development and wage levels. While it is very well known that economic development raises a society’s wages, a rising overall wage does not constitute increasing information. By our original
65
The Effects of Incomplete Employee Wage Information
definition, incomplete information is based on changes in wage distribution skewness, and not on the distribution’s overall mean wage level. Thus our proclaimed result goes beyond traditional development economics.
A Graphical Depiction To get another view of the above results, we graph incomplete information against each of these explanatory variables (Figs. 1–6). Every observation is labeled by country name. Also, the graphs contain a fitted line mirroring each of the above model’s predictions. As easily seen, the cross-country predictions blatantly stand out. Nations with greater UI benefits (Fig. 1) have less incomplete information (more employee information). Workers in more densely populated nations have less incomplete information (Fig. 2). Nations with greater rural populations have more incomplete worker information (Fig. 3). Industrialized nations have less incomplete information (Fig. 4). Finally, nations with a greater influx of foreign workers have more incomplete information (Figs. 5 and 6). Fitted values
adjusted_incomplete_information NW
0.419992 US
US
FI FI US
NW
GE CN CN
CN
CN
FI CN FI
US US
IE IE
NW IE SW NL
SW SW
GE
NL
GE GE
CN
UK UK UK
IS IS UK
NL
0.094879 -1.34707
1.72277 lnui
Fig. 1.
Adjusted Incomplete Information Versus Logarithm of UI.
66
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG adjusted_incomplete_information
0.457992
Fitted values
NW FI FI NWUS US US US US NWUS CN SW SW CN US
CN
GE
IE IE IE
SW
NL US
CN
GE GE
NLNL
GE
CN
IS UK UK IS UK UK NL
CZ CZ
0.132879
NL
2.70037
466.499 population_density
Adjusted Incomplete Information Versus Population Density.
Fig. 2.
Fitted values
adjusted_incomplete_information NW
0.462992
FI
NW
US US CN USUS
FI FI
FI
US
NW US
GE
SW SW
CNCNCN CN
IE
US
SW
IE IE
NL NL NL
CN
GE
US
GE GE
CN
IS UK IS UK UK UK NL NL
0.137879
CZ CZ
8.841
42.2834 rural_population
Fig. 3.
Adjusted Incomplete Information Versus Rural Population.
67
The Effects of Incomplete Employee Wage Information adjusted_incomplete_information
Fitted values
NW
0.446992
FIFI US
FI
NWUS US US CN
FI
US IE
NW CN CN CN CN
SW
GE
SW IE IE
SW NL
CN
NL
US
GE GE
NL CN
IS UK UK IS
UK UK
NL CZ CZ
NL
0.121879 21.5
45.1 industrial_employment
Adjusted Incomplete Information Versus Industrial Employment.
Fig. 4.
adjusted_incomplete_information
Fitted values
0.437136 US US US
US
US GE
IE
CN CN CN CN
IE IE
GE GE GE
CN
UK UK
0.178782
UK UK
742.3
3.8 inflow_foreign_workers_number
Fig. 5. Adjusted Incomplete Information Versus Inflow of Foreign Workers.
68
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG Fitted values
adjusted_incomplete_information US
0.404516
US US
US
US GE
IE
CN CN
CN CN
IE IE
GE GE GE
CN
UK UK UK
0.178782
UK
0.411507
0.032892 inflow_foreign_workers_prop
Fig. 6.
Adjusted Incomplete Information Versus Proportional Inflow of Foreign Workers.
One interesting observation is noted from these graphs. Within-country results are not always as strong as the between country results. Take incomplete information versus the logarithm of UI (Fig. 1). Whereas greater UI decreases incomplete information in the US, Canada, the UK, Israel, the Netherlands, Ireland and Sweden, it does not in Norway, Finland and Germany.24 Rather, these latter within-country patterns seem to indicate rising incomplete information as UI increases. These few counter examples imply that UI alone does not explain all changes of the within country incomplete information. Perhaps, UI must be examined in conjunction with welfare policies, given the preponderance of Scandinavian countries among this latter group. Probably because a country’s population density and rural population change very slowly over short time periods, we observe similarly anomalous results for within-country comparisons in Figs. 2 and 3. As a result, we observe almost no within-country relationship in these figures, even though the between country relation is significant. On the other hand, the within country observations follow the crosscountry trends in Figs. 4, 5 and 6. Incomplete information decreases within
The Effects of Incomplete Employee Wage Information
69
a country when its industrial employment increases (Fig. 4). Incomplete information tends to rise with inflows of foreign workers (Figs. 5 and 6).
CONCLUSION In this paper, we define a tractable procedure to measure incomplete information in the labor market. The procedure is based on econometric frontier estimation techniques, and is consistent with search theory. It is an improvement over past measures because it holds individual characteristics constant and nets out random effects. We apply the technique to 11 countries over various years. We find that incomplete information leads workers to receive on average about 30–35% less pay than they otherwise would have earned, had they information on what each firm paid. Generally married men and women suffer less from incomplete information than the widowed or divorced, and singles suffer the most. Women suffer more than men do. Schooling and labor market experience reduce these losses. But institutions within a country matter, as well. For example, we find that workers in countries that strongly support unemployment insurance (UI) receive wages closer to their potential. Doubling UI decreases incomplete information and results in 5% higher wages. A more dense population reduces search costs leading to less incomplete information and higher wages. A more industrial economy disseminates wage information better, so workers exhibit less incomplete information. Finally, in memory of Tikva Lecker, we examined the effect of foreign worker inflows on incomplete information. Data within countries, as well as data from a cross-country comparison, yield comparable results. Foreign worker inflows skew wages to the left. In the short-run they increase incomplete information, and at the same time they reduce average wage levels. In reaching these conclusions, it is important to note that we achieve our estimates of incomplete information based on the skewness of the wage distribution and not on wage levels. Thus, whereas we find that individual characteristics such as race, gender, and work experience lead to higher wage levels in their own right (i.e., these variables have a direct effect), they also do so through their effects on incomplete information. Similarly whereas we find that a country’s institutional characteristics such as commitment to UI, sanctions against foreign workers, or a more developed industrial structure directly raise wages, these country characteristics also do so through their effects on incomplete information. These latter effects
70
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
of incomplete information on wage level are obtained from wage distribution skewness, and not from symmetric movements in the entire wage distribution.
NOTES 1. Katz and Ziderman (1986) argue that incomplete information also affects the complete wage package including non-wage benefits. Given unobserved worker characteristics, employers provide specific non-wage amenities to attract workers with desired but unobserved characteristics, thereby affecting the equilibrium wage. Unfortunately, we do not have sufficient data to consider non-wage fringes. 2. How institutional characteristics affect employer (as opposed to employee) incomplete information is beyond the scope of this paper. To answer this latter question, one would need to adopt a two-tier estimation technique (Polachek & Yoon, 1987). However, at this point, the two-tier algorithm is not available to use with LIS data, since all statistical analysis must be done on the LIS computer with standard statistical software (SAS, SPSS, and STATA). 3. At this point we are abstracting from life cycle considerations, particularly training and other opportunities available on the job. As will be illustrated later, the estimation procedure accounts for these factors by including life cycle variables. Differences in information between ‘‘inside’’ and ‘‘outside’’ employees may also be a consideration, but getting at these is more difficult. Not all ‘‘inside’’ information is acted upon, and hence it is more difficult to measure. For example, the perspicacious peregrinator (Polachek & Horvath, 1977) searches on-the-job for more beneficial opportunities. The information he/she gathers is unobservable until acted upon, which does not occur until the perspicacious peregrinator actually moves. 4. We do not use ISSP data partially because the earnings are reported in categories in many country-years. This earnings smoothing is likely to distort information regarding incomplete information. However, future work can include those country-years that do not report earnings in categories to extend our current study sample. 5. Wage level w is dependent on the worker’s skill level x and a variable y that will be defined shortly, so that w ¼ w(x,y). This wage can vary somewhat because companies differ in how they use comparably skilled workers. Some, perhaps with good management and/or higher physical capital (both indexed by y) pay more, while other less efficient companies pay lower amount for comparable workers. This results in a distribution of potential wages, for example W Nðw; ¯ yÞ; given the firms’ heterogeneity of utilization of workers of given quality x. Accordingly, there are high wage firms, the maximum wage being wmax ðxÞ; and low wage firms, the minimum being wmin ðxÞ: For now, since we are talking about a worker of a particular skill level, we suppress the vector of worker characteristics (x), defining these available wages as wmin and wmax : 6. The LIS data do not contain employer characteristics. Thus, such information is omitted from xi : To the extent these omitted variables influence ui ; our worker
The Effects of Incomplete Employee Wage Information
71
incomplete information estimates may be biased. However, differences in these measures across countries are unaffected by this bias. These differences are akin to serially correlated errors in fixed-effects panel estimates being better than crosssectional estimates (see Bound & Krueger, 1991). Economy-wide institutional variables are included later in the analysis when comparing incomplete information across countries. 7. This is the measurement error omitted from past studies of incomplete information that simply use wage (or price) dispersion measures to get at incomplete information. 8. It is not necessary to assume ui N þ ðmmi ; s2mi Þ: Other common distributions used in this context include the exponential, the Gamma, and lognormal. However, past studies found little qualitative difference between results using these different distributions. See Olson, Schmidt, and Waldman (1980) and Gong and Sickles (1992). 9. The skewness and kurtosis test for normality (sktest) is described in the STATA Corporation (2001) reference manual, pp. 226–228. 10. As already noted, data on employers were not available, and hence omitted. However, in the latter cross-country analysis, the effect of these omitted variables are smaller, the more uniformly distributed these variables are across countries. Also, without data on tenure it is impossible to examine incomplete information differences between insider and outsider workers. However, if tenure data were available, one could compare information measures of recently hired workers to workers with longer tenure to ascertain the role on-the-job search of ‘‘insiders’’ within the firm. Preliminary evidence on this solely for the U.S. (Polachek & Yoon, 1987, Table 2) indicates that tenured employees have less incomplete information. Similarly, if enough panel data were available, one could employ the techniques utilized by Polachek and Yoon (1996) to net out person-specific heterogeneity. 11. Note that according to Weinstein the likelihood function is based on (1964), the ei distribution being f ðÞ ¼ 2=s f =s 1 F ðls 1 Þ: 12. As schooling levels rise, the wage distribution becomes wider. Although a wider earnings distribution increases search gains, it need not imply more information since search costs also rise. Perhaps this is the reason we see no time trends regarding incomplete information. 13. Whereas technological improvement provides the opportunity for more search, the whole search process can become more overwhelming, making the job choice decision more complicated. For instance, by intensively searching through the Internet, individuals decrease search costs. These decreased search costs might imply more search but less real information if, for example, wages are not posted. Thus, technology could leave workers with less real information. 14. For the details, see Eqs (9) and (10) in Wang’s paper. 15. Refer to Table 5 for definitions of these variables. 16. We adjust the 18 incomplete information estimates by the hourly wage dummy coefficient in a regression of incomplete information on the hourly wage dummy variable. 17. It is argued that our estimates could be biased by unadjusted time-invariant country effects when these effects are correlated with the independent variables. However, the small sample sizes of our regressions – between 21 and 42 observations – prevent us from using country dummies: we will lose about 10 degrees of freedom
72
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
after country dummies are implemented, which would make our regression results much weaker. 18. Foreign worker inflows cannot be integrated because the limited number of observations on this variable precludes sufficient degrees of freedom. In addition, we eliminated the industrial employment variable because it becomes insignificant. 19. This adjustment is omitted from Models (3) and (4), because each of the 21 observations contains incomplete information measures derived solely from hourly wage data. 20. According to Table 5, average adjusted employee incomplete information is about 0.30. A 0.03 change is about 10%. 21. The large industrial employment and rural population elasticities (Table 6) at first surprised us. However, an examination of the means (or median for population density) and standard deviations of these variables reveals values, respectively, five times the size and twice the size of their standard deviations. So these two elasticity measures are really very consistent with the marginal effects. 22. More specifically, a doubling of foreign workers (i.e., a new inflow of 77,700 foreign workers for a country already with 77,700 (median value from Table 5) foreign workers and incomplete information of 0.359) will increase incomplete information by 0.018 units, which is 4.9% of 0.359. Similarly, increasing the proportion of foreign workers from 0.161 to 0.322% yields an increase in incomplete information by 0.142, which is 11.6% of 0.359. 23. As noted in note 13, an insufficient number of observations precludes including the foreign worker inflow. The industrial employment variable has the correct sign, is statistically insignificant. Thus we also eliminate this variable in model (5). 24. Missing UI data preclude plotting Czechoslovakia in the Firgure.
ACKNOWLEDGMENTS We are especially indebted to Hung-Jen Wang for providing a STATA version of the frontier estimation program, and to Subal Kumbhakar, to participants of the Bar-Ilan Conference in Honor of Tikva Lecker, as well as to two anonymous referees for valuable comments and suggestions. All remaining errors are our own.
REFERENCES Aigner, D., Lovell, C. K., & Schmidt, P. (1977). Formulation and estimation of stochastic frontier production function models. Journal of Econometrics, 6, 21–38. Arellano, M., Bentolila, S., & Bover, O. (1998). Unemployment duration, benefit duration and the business cycle. C.E.P.R. Discussion Paper 1840. Arnold, M. A. (2000). Costly search, capacity constraints, and Bertrand equilibrium price dispersion. International Economic Review, 41(1), 117–131.
The Effects of Incomplete Employee Wage Information
73
Beach, C. M., & Kaliski, S. (1983). The impact of the 1979 unemployment insurance amendments. Canadian Public Policy, 9, 164–173. Belzil, C. (1995). Unemployment duration stigma and re-employment earnings. Canadian Journal of Economics, 28(3), 568–585. Bester, H. (1986). Bargaining, search costs and equilibrium price distribution. Germany: Discussion Paper Series A, University of Bonn. Borjas, G. (1985). Assimilation, changes in cohort quality, and the earnings of immigrants. Journal of Labor Economics, 3, 463–489. Bound, J., & Krueger, A. B. (1991). The extent of measurement error in longitudinal earnings data: Do two wrongs make a right? Journal of Labor Economics, 9(1), 1–24. Burdett, K., & Judd, K. L. (1983). Equilibrium price dispersion. Econometrica, 51(4), 955–969. Card, D., & Levine, P. B. (1998). Extended benefits and the duration of UI spells: Evidence from the New Jersey extended benefit program, NBER Working Papers: 6714, National Bureau of Economic Research, Inc. Chiswick, B. (1978). The effect of Americanization on the earnings of foreign born men. Journal of Political Economy, 86, 897–921. Chiswick, B. (1986). Is the new immigrant less skilled than the old? Journal of Labor Economics, 4, 897–921. Daneshvary, N., Herzog, H., Hofler, R., & Schlottmann, A. (1992). Job search and immigrant assimilation: An earnings frontier approach. Review of Economics and Statistics, 74(3), 482–492. Ehrenberg, R., & Oaxaca, R. (1976). Unemployment insurance, duration of unemployment and subsequent wage gain. American Economic Review, 66(5), 754–766. Fougere, D., Pradel, J., & Roger, M. (1998). The influence of the state employment service on the search effort and on the probability of leaving unemployment. Institut National de la Statistique et des Etudes Economiques Working Paper. Freeman, R. (1980). Unionism and the dispersion of wages. Industrial and Labor Relations Review, 34, 3–23. Gaynor, M., & Polachek, S. (1994). Measuring information in the market: An application to physician services. Southern Economic Journal, 60(4), 815–831. Gong, B.-H., & Sickles, R. (1992). Finite sample evidence on the performance of stochastic frontiers and data envelopment analysis using panel data. Journal of Econometrics, 51 (1–2), 259–284. Gonzalo, M. T. (2002). A new look at the UI effect on transitions from unemployment into wage employment in Spain: The limited duration of the UI benefits entitlement. Applied Economics, 34(17), 2177–2187. Groot, W., & Oosterbeek, H. (1994). Stochastic reservation and offer wages. Labour Economics, 1(3/4), 383–390. Ham, J., & Rea, S. (1987). Unemployment insurance and male unemployment duration in Canada. Journal of Labor Economics, 5(3), 325–353. Heckman, J., & Polachek, S. (1974). Empirical evidence on the functional form of the earnings-schooling relationship. Journal of the American Statistical Association, 69, 350–354. Hofler, R., & Murphy, K. J. (1992). Underpaid and overworked: Measuring the effect of imperfect information on wages. Economic Inquiry, 30(3), 511–529. Hofler, R., & Polachek, S. (1985). A new approach for measuring wage ignorance in the labor market. Journal of Economics and Business, 37(3), 267–276.
74
SOLOMON W. POLACHEK AND JUN (JEFF) XIANG
Jurajda, Sˇ., & Tannery, F. J. (2003). Unemployment durations and extended unemployment benefits in local labor markets. Industrial and Labor Relations Review, 56(2), 324–348. Kamiya, K., & Sato, T. (2004). Equilibrium price dispersion in a matching model with divisible money. International Economic Review, 45(2), 413–430. Katz, E., & Ziderman, A. (1986). Incomplete information, non-wage benefits and desirableworker self selection. Australian Economic Papers, 25(47), 252–256. Kumbhakar, S., & Lovell, C. A. K. (2000). Stochastic frontier analysis. Cambridge: Cambridge University Press. Lach, S., & Tsiddon, D. (1992). The behavior of prices and information: An empirical analysis of dissaggregated price data. Journal of Political Economy, 100(2), 349–389. Lang, G. (2004) How different are wages from wage potentials? Analyzing the earnings disadvantage of immigrants in Germany. Institute for Economics Discussion Paper #256, Universitaet Augsburg. Luxembourg Income Study. http://www.lisproject.org/. McCall, J. (1970). Economics of Information and Job Search. Quarterly Journal of Economics, 84, 113–126. McCall, J. (1973). Income mobility, racial discrimination and economic growth. Lexington: D.C. Health. Meeusen, W., & van den Broeck, J. (1977). Efficiency Estimation form Cobb–Douglas production functions and composed errors. International Economic Review, 18(2), 434–444. Meyer, B. (1990). Unemployment insurance and unemployment spells. Econometrica, 58(4), 757–782. Micklwright, J., & Nagy, G. (1995). Unemployment insurance and incentives in Hungary. C.E.P.R. Discussion Papers, CEPR Discussion Papers: 1118. Mincer, J. (1974). Schooling, experience and earnings. New York: Columbia University Press for the NBER. Moffitt, R., & Nicholson, W. (1982). The effect of unemployment insurance on unemployment: The case of federal supplemental benefits. Review of Economics and Statistics, 64(1), 1–11. Mortenson, D. (1970). Job search, the duration of unemployment, and the Phillips curve. American Economic Review, 60(5), 847–862. Murphy, K., & Welch, F. (1990). Empirical age-earnings profiles. Journal of Labor Economics, 8(2), 202–229. Nelson, P. (1970). Information and consumer behavior. Journal of Political Economy, 78, 311–329. Olson, J., Schmidt, P., & Waldman, D. (1980). A Monte Carlo study of estimators of stochastic frontier production functions. Journal of Econometrics, 13(1), 67–82. Polachek, S. (2004). What can we learn about the decline in union membership from international data? In: P. Wunnava (Ed.), Changing roles of unions: New forms of representation (pp. 362–377). Armonk, NY: M.E. Sharpe. Polachek, S., & Horvath, F. (1977). A life cycle approach to migration: Analysis of the perspicacious peregrinator. Research in Labor Economics, 1, 103–149. Polachek, S., & Robst, J. (1998). Employee labor market information: Comparing direct world of work measures of workers’ knowledge to stochastic frontier estimates. Labour Economics, 5(2), 231–242. Polachek, S., & Yoon, B. J. (1987). A two-tiered earnings frontier estimation of employer and employee information in the labor market. Review of Economics and Statistics, 69(2), 296–302.
The Effects of Incomplete Employee Wage Information
75
Polachek, S., & Yoon, B. J. (1996). Panel estimates of a two-tiered earnings frontier. Journal of Applied Econometrics, 11(2), 169–178. Reinganum, J. F. (1979). A simple model of equilibrium price dispersion. Journal of Political Economy, 87(4), 851–858. Royston, P. (1991). Comment on sg.3 and an improved D’Agostino test. Stata Technical Bulletin, 3, 23–24. Sandell, S. (1980). Job search by unemployed women: Determinants of the asking wage. Industrial and Labor Relations Review, 33, 368–378. Schmidt, P., & Lin, T. (1984). Simple tests of alternative specifications in stochastic frontier models. Journal of Econometrics, 24(3), 349–361. Sorensen, A. T. (2000). Equilibrium price dispersion in retail markets for prescription drugs. Journal of Political Economy, 108(4), 833–850. Stata Corporation (2001) Stata Statistical Software: Release 7.0. Reference Annual (College Station: Texas), pp 226–228. Stephenson, S. P. (1976). The economics of youth job search behavior. Review of Economics and Statistics, 58(1), 104–111. Stevenson, R. E. (1980). Likelihood functions for generalized stochastic frontier estimation. Journal of Econometrics, 13(1), 57–66. Stigler, G. (1961). The economics of information. Journal of Political Economy, 69(3), 213–225. Stigler, G. (1962). Information in the labor market. Journal of Political Economy, 70(5, Part 2), S94–S105. Stigler, G., & Kindahl, J. (1970). The behavior of industrial prices. New York: Columbia University Press. Taubman, P. (1976). The determinants of earnings: Genetics, family, and other environments: A study of white male twins. American Economic Review, 66(5), 858–870. van Hoomissen, T. (1988). Price dispersion and inflation: Evidence from Israel. Journal of Political Economy, 96(6), 1303–1314. van den Berg, Gerard, J., & van der Klaauw, B. (2001). Counselling and monitoring of unemployed workers: Theory and evidence from a controlled social experiment. C.E.P.R. Discussion Papers, CEPR Discussion Paper # 2986. Wang, H.-J. (2002). Heteroscedasticity and non-monotonic efficiency effects of a stochastic frontier model. Journal of Productivity Analysis, 18(3), 241–253. Weinstein, M. A. (1964). The Sum of Values of A Normal and A Truncated Normal Distribution. Technometrics, 6, 104–105.
This page intentionally left blank
76
THE IMPACT OF IMMIGRANT DYNASTIES ON WAGE INEQUALITY Michael Ben-Gad ABSTRACT I construct a set of dynamic macroeconomic models to analyze the effect of unskilled immigration on wage inequality. The immigrants or their descendants do not remain unskilled – over time they may approach or exceed the general level of educational attainment. In the baseline model, the economy’s capital supply is determined endogenously by the savings behavior of infinite-lived dynasties, and I also consider models in which the supply of capital is perfectly elastic, or exogenously determined. I derive a simple formula that determines the time discounted value of the skill premium enjoyed by college-educated workers following a change in the rate of immigration for unskilled workers, or a change in the degree or rate at which unskilled immigrants become skilled. I compare the calculations of the skill premiums to data from the US Current Population Survey to determine the long-run effect of different immigrant groups on wage inequality in the United States.
Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 77–134 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24003-7
77
78
MICHAEL BEN-GAD
1. INTRODUCTION In this paper I construct a set of macroeconomic models to analyze how increases in the number of unskilled immigrants may affect wage inequality over time. In these models, a change in the number of such immigrants does not necessarily alter the composition of the workforce permanently. Rather than remaining unskilled forever, a portion of the additional immigrants, or their descendants, join the ranks of skilled workers. Indeed, as is the case for some immigrant groups in the United States, the native-born children or grandchildren of immigrants with low levels of education may not merely assimilate by matching the general level of educational attainment, but exceed it. For the baseline model, I adopt the Weil (1989) overlapping dynasties optimal growth framework. Changes in immigration policy not only affect wages directly by altering the size and composition of the labor force, but also alter the rate of return to capital, inducing changes in savings behavior that gradually affect the size of the capital stock. These changes to the size of the capital stock indirectly affect wages as well. I derive a simple reduced form that encapsulates all these different effects on one measure of wage inequality – the ratio between the discounted values of skilled and unskilled wages. The effect of a change in immigration policy on wages is neither constant nor immediate. A change in immigration policy generates changes in the size and composition of the population that accumulate over time. The effect of these changes on factor returns may or may not be permanent, depending on whether the labor supplied by the immigrants or their descendents perfectly substitutes for the pre-existing labor supply. Therefore by examining the ratio between the discounted value of the two different wages, I can determine to what degree, in the long run, high educational attainment by the descendants of unskilled immigrants either ameliorates or reverses their short-run impact on wage inequality. In my baseline model, capital is endogenously determined but adjusts slowly. Borjas (1999) analyzes the impact of immigration in static models under two alternative assumptions – capital supply is either completely elastic, or fixed. To better understand the sensitivity of my measure of wage inequality to different assumptions about the supply of capital, I compare the behavior of my model to one in which the stock of capital adjusts immediately to policy changes, but where the rate of return is exogenously determined. In addition, I also consider the case where the size of the capital stock grows at a fixed exogenous rate.
The Impact of Immigrant Dynasties on Wage Inequality
79
In Section 2, I briefly review recent U.S. immigration policy. I present data from the U.S. Census Current Population Survey that demonstrates the vast differences in educational attainment among different immigrant groups – differences that span at least two generations. These data also highlight the higher degree of intergenerational mobility between the immigrant generation and the second generation, when compared to the analogous native populations. I present in Section 3 the dynamic optimal growth model with overlapping dynasties, developed by Weil (1989). Ben-Gad (2004) used the model to examine the behavior of an economy that is absorbing immigrant dynasties over time. In this paper, I distinguish between two types of households: households with skilled workers (college-educated adults), with unskilled workers (adults without college degrees). Section 4 describes the dynamic system and the general perturbations method I use to simulate its behavior. I also derive the formula used to calculate the discounted skill premium (the percentage gap between the present value of wages for college and non-college educated workers) in the baseline model. I also present the explicit reduced form for the special case where the elasticities of substitution between the factors of production are all identical. Section 5 briefly describes two alternative assumptions about the elasticity of the capital supply. I demonstrate that for the special case in which all the elasticities of substitution between the inputs are identical, the ratio between the two wages at any moment in time is identical, regardless of what mechanism governs the dynamic behavior of the capital stock. Nonetheless, even in this case, the discounted premium to education is sensitive to the model we choose. In Section 6, I present my procedure for modeling the effect of immigration policy on the composition of the labor force over time. There is first, a direct effect as the ratio of skilled to unskilled workers among the extra new arrivals seldom matches the veteran population. There is a second effect because I permit the descendants of these immigrants to switch between the two categories during the periods after they arrive. One serious limitation to this approach is that membership in either category is determined exogenously – I am modeling the impact of observed changes in educational attainment but make no attempt to explain them. Section 7 explains the choice of parameters I use to calibrate the model. In Section 8, I consider the impact of a 20-year surge in the immigration of unskilled workers on wages and wage inequality, within the context of three different assumptions for the elasticity of capital supply, and three different
80
MICHAEL BEN-GAD
specifications of the production function. I consider three different scenarios. First, what if the immigrants and their descendants remain permanently unskilled? No one ever attends college, and their arrival permanently lowers the share of unskilled workers within the economy. This scenario is a crude approximation of perhaps the most pessimistic outcome for immigration: the creation of a permanent unskilled under class. In the second scenario, the immigrants are initially unskilled, but over time, they or their descendents gradually attain the levels of college education prevalent in the general population. This process of immigrant ‘assimilation’ ultimately restores the distribution of college-educated and non-college educated people in the workforce to its initial level. In the third scenario, all the immigrants or their descendents eventually attend college. This last case is perhaps the least likely, yet paradoxically the most informative. Finally, I compare the results to those obtained for an identically sized influx of college-educated immigrants. Finally, in Section 9, I consider more realistic examples, where not all immigrants are skilled or unskilled and neither are all their descendants. I relate the results for the discounted skill premium to the data on college attainment for the different immigrant groups in the Current Population Survey. In this paper I do not presume to explain the decisions made by households to immigrate. Because legal migration from the developing world to the developed world is regulated by the rationing of visas, and illegal migration by the resources invested in interdiction, or the harshness of penalties imposed on those violating immigration laws, I believe it is possible to treat modest changes in rates of immigration for unskilled workers as exogenous policy decisions.1 More importantly, this paper ignores the decisions made by immigrants or their descendants to acquire education. Instead, I focus on the long-run implications of differential educational attainment among immigrants and their descendents for wage inequality.
2. U.S. IMMIGRATION POLICY AND EDUCATIONAL ATTAINMENT AMONG IMMIGRANT GROUPS 2.1. The Rate of Net International Migration The share of foreign-born within the population of the United States declined steadily between 1910 and 1970, from 14.7 to 4.7%. Since then it
The Impact of Immigrant Dynasties on Wage Inequality
81
has climbed swiftly, reaching 11.7% by the end of 2003. What has generated such a dramatic rise in just over three decades? The official rate of immigration presented in Fig. 1 Panel (a), features the data tabulated by the U.S. Bureau of Citizenship and Immigration Services. These numbers show immigration rising steadily from a rate of 1.5 per thousand in 1960 to 2.6 per thousand in 1988, then rising much more steeply, reaching 7.1 per thousand in 1991, and then declining to 2.3 per thousand in 1999. The rate at which people arrived in the United States did indeed rise between 1960 and 1991, but not by nearly so much, nor was the rate nearly so volatile. The official rate of immigration in Panel (a) of Fig. 1 does not show the date at which foreigners arrive in the United States or join its workforce, but merely captures the number who attain the official status of immigrant. Hence, there was no massive influx of immigrants in 1991, but rather a large number of people, many living and working illegally in the United States for a decade or more, who took advantage of the amnesty provisions in the Immigration Reform and Control Act of 1986 (IRCA), to register as legal immigrants. Net international migration (NIM) in Fig. 1 Panel (b), measures the physical movement of people between the United States and the rest of the world. The rise in the NIM was far less dramatic than either the changes in the official rate of immigration or the steep decline in the rate of natural population growth in Panel (b) of Fig. 1.2 Between 1960 and 1999 the birth rate in the United States dropped from 23.8 to 14.4 per thousand, causing the steep decline in the rate of natural population growth. The sharp decline in the birth rate combined with the gradual increase in net migration between 1970 and 2000 to generate the large increase in the share of the foreign born within the U.S. population over the same period.
2.2. U.S. Immigration Policy Since passage of the Immigrant and Nationality Act of 1965, most legal immigrants have arrived in the United States through some form of family sponsorship. Immediate relatives of United States citizens may enter without limit; during the 1990s about a quarter of a million arrived each year. Other relatives of U.S. citizens are admitted as family sponsored preference immigrants – the Immigration Act of 1990 set the limit for all family sponsored immigrants as either 226,000, or 480,000 minus the number of people admitted under the category of immediate relatives during
82
MICHAEL BEN-GAD
14 13 12 11 10 9 8 7 6 5 4 3 2 1 1960 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 (a) 14 13 12 11 10 9 8 7
Natural Pop. Growth
6 5 4 3 2 1
Net International Migration
1960 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 1996 1999 (b)
Fig. 1. Annual Rates Per Thousand, of Legal Immigration and Components of Population Growth in the United States, 1960–1999. Natural Population Growth is Number of Births, Less Deaths. NIM from 1960 to 1984 Includes Migration by U.S. Civillians, but Excludes Military Personnel, NIM from 1985 to 1994 Excludes both U.S. Civillians and Military Personnel, and 1995–1999 NIM Excludes both U.S. Civillians and Military Personnel. Note: The Dashed Gray Lines Correspond to Decade Averages for NIM. Sources: U.S. Census Bureau (2004), Population Division and Housing and Household Economic Statistics Division and U.S. Census Bureau. a) Legal Immigration to the United States 1960–1999; b) Components of Population Change in the United States 1960–1999.
The Impact of Immigrant Dynasties on Wage Inequality
83
the previous year, whichever is larger. The United States also allocates 140,000 employment-based preference visas for workers with special skills or training (as well as investors), and an additional 55,000 visas are allocated by lottery under the diversity program. Finally, the United States admits refugees and asylum seekers (refugees are admitted from abroad on the basis of a yearly quota set annually by the president). After a year, refugees and asylees are eligible for permanent residence – between 1991 and 2000, just over one million gained admission. In addition to immigration visas, in 1992 the United States began granting 65,000 H-1B visas to temporary workers with special skills – nearly all recipients have college or advanced degrees.3 To ameliorate a perceived shortage of qualified workers in the information technology sector, Congress passed the American Competitiveness in the Workforce Act of 1998, temporarily increasing the number of H-1B visas to 115,000 per year in 1999 and 2000, and 107,500 in 2001. The American Competitiveness in the Twenty-First Century Act of 2000 (AC21) added an extra 347,500 visas by raising the cap to 195,000 for each of the years 2001, 2002, and 2003, for a total of 585,000 H-1B visas over three years. The cap for 2004 and beyond is once again 65,000. Finally, the gross inflow of illegal immigrants is about 350,000 per year.4 The net increment to the population from this source is smaller – 80% of those who leave the United States are foreign born, and a substantial fraction of these are illegal aliens returning home. In the year 2000, there were approximately seven million people living in the United States illegally, of whom 1.5 million arrived between 1991 and 2000 – a net inflow of 150,000 per year.5
2.3. College Attainment and Immigrant Groups Clearly, any government considering a serious change in its immigration policy should be concerned not only with the skills immigrants bring to their new country, but also with the levels of education attained by their children – the members of the second generation. In the two graphs in Fig. 2, I pool data from the Current Population Survey of the U.S. Census for the years 2001, 2002, and 2003. The horizontal axes correspond to the share of people aged 45–64 with four-year Bachelor’s degrees by place of birth. I sample only those immigrants who arrived prior to 1975. So as to focus on people who have immigrated to the United States near the beginning of their working lives, and at about the time they are establishing a household.
84
MICHAEL BEN-GAD
By restricting the sample, I exclude older people whose immigration was perhaps sponsored by their adult children – themselves immigrants to the United States. Examining immigrants by country of origin reveals an enormous degree of heterogeneity in the shares of people with college degrees. By this measure, male immigrants from India are the best educated – nearly 83% have college degrees, followed by male immigrants from China with 63%. By contrast, the average share of college educated U.S.-born males within the same age group is only 31%. Among all the males in this sample, the least educated are Mexican immigrants, of whom just under 7% have completed college, followed by Puerto Ricans (I treat respondents who report Puerto Rico as their birthplace as immigrants rather than natives even though they are U.S. citizens by birth). Although formal education does not completely encompass all labor market skills or perfectly predict labor market outcomes, it is not hard to imagine the direct impact of these immigrants from Mexico or Puerto Rico on the wages of unskilled workers in the United States. Indeed, Borjas (2003) estimates the overall wage elasticity within skill groups to be 0.4 (there are four levels of educational attainment in his model, and he also controls for labor force experience). By contrast the arrival of highly educated immigrants from India or China is likely to depress the wages of skilled workers. Abstracting from the overall level of wages, further immigration from India or China is likely to lower the wage premium enjoyed by male college graduates, whereas immigration from Mexico and Puerto Rico is likely to enhance it. Furthermore, the impact of each particular group of immigrants on the composition of the labor force and wage inequality is likely to last long after the immigrants themselves retire. To understand what this means, consider the vertical axes of the graphs in Fig. 2. These measure the rate of college completion among U.S.-born individuals aged 25–44, according to the birthplace of their fathers (for men) or mothers (for women). These are members of a second generation, counterparts to the immigrant generation whose rates of college education are measured on the horizontal axis. Notice that in the upper panel of Fig. 2, the points representing Mexico, Puerto Rico, and India are above, but fairly close to the solid gray 451 line through the origin, and the point representing China is not too far from it either. If we treat the younger people in our sample, represented on the vertical axes as surrogates for the children of the immigrant cohort on the horizontal axes, we can conclude that the impact of each immigrant group
85
The Impact of Immigrant Dynasties on Wage Inequality 100
US born men aged 25-44 by father's place of birth
India
80
China
y=16.07+0.82x R2 = 0.68 Poland
60
UK Germany Colombia
USSR Ireland
Greece Philippines
40
Italy
Portugal
Canada Cuba
Dom. Rep.
US
20 Puerto Rico
Mexico
0
0
20 40 60 80 Men aged 45−64 born in US or arrived before 1975 by place of birth
100
100
US born women aged 25-44 by mother's place of birth
China
India
80
y=31.34+0.59x Poland
R2=0.30
Greece
60 Ireland USSR Colombia Italy
40
Dom. Rep.
Germany Cuba Canada
UK
Portugal
Philippines
US
20 Puerto Rico Mexico
0 0
20 40 60 80 Women aged 45−64 born in US or arrived before 1975 by place of birth
100
Fig. 2. Percentage of the U.S. Population with Four-Year College Degrees by Age, Sex, Birthplace, and Parent’s Birthplace. Data for the U.S.S.R. Includes all Respondents from any of the Former Republics in the Sample, the Data for the U.K. Includes Respondents from Northern Ireland, and Data for Portugal Includes Respondents from the Azores. Note: Pooled Data for 2001, 2002, and 2003 from the U.S. Census, Current Population Survey. Source: King, Ruggles, and Sobek (2003).
86
MICHAEL BEN-GAD
on the share of college educated within the population is fairly constant over the course of at least two generations. This does not mean that their quantitative impact on wages is constant – the overall level of wages, and under some circumstances the ratio between wages for skilled and unskilled workers, is very dependent on the rate of adjustment of the capital stock to any surge in immigration. Nonetheless, qualitatively, we can confidently predict the impact on wage inequality generated by each of these four immigrant groups across at least two generations and probably more.6 This confidence quickly dissipates when we consider some of the other groups represented in Fig. 2, or worse still, when we attempt to compare between them. Among male immigrants from Poland in our sample, only 21% have college degrees – well below the average for the population as a whole. Yet the levels of college graduation among young American-born sons of Polish-born fathers are immediately below the high levels attained by their Indian and Chinese counterparts – 59% in the sample completed college. How do we compare the impact of male immigrants from places like Poland, with their high levels of intergenerational upward mobility, with the impact of better-educated immigrants from Colombia, Cuba, Germany, Ireland, Italy, or the United Kingdom? How much does it matter that the children of these more-educated immigrants seem to experience relatively little upward mobility, and graduate from college at lower rates than men whose fathers are from Poland? Indeed, what about immigrants from Canada, the Philippines, or the former U.S.S.R? There we see a slight drop across the two generations in the share who report completing college. Chiswick (1978) found that controlling for various factors, including age, and schooling, immigrants to the United States earn more than their native counterparts provided they have worked in the U.S. for a long enough period of time. If we interpret this finding as a measure of motivation, it would seem that some of this motivation spills over to the next generation or is expressed in a greater effort by immigrant parents to provide a college education for their children. Among males, the Polish immigrant group presents the most obvious example of this phenomenon. However, nearly all the points in Fig. 2 cluster along a regression line (R2 ¼ 0:68) above the 451 line.7 The intergenerational outcomes for the women among these same immigrant groups in the lower panel of Fig. 2 are far less predictable. Consider female immigrants from Greece. In terms of college completion, they are as a group, the second least educated people in our sample. Less than 4% of this group have college degrees, while among the American-born women between the ages of 45 and 64 in our sample, the rate of college
87
The Impact of Immigrant Dynasties on Wage Inequality
completion is 25%. Indeed, only women born in Portugal have lower rates of college completion (0.5%). Yet the point in the lower panel of Fig. 2 corresponding to Greece is well above the 451 line. This is because just under 60% of second generation American born women aged 25–44 who report having Greek-born mothers completed college, twice the rate of daughters of American-born women in the same age group, and behind only women with mothers from China, India, and Poland. What does the arrival of immigrants like these women from Greece mean for wage inequality in the U.S. over time? Which dominates, the low levels of education in the first, or the high levels of education in the second generation and perhaps beyond?
3. IMMIGRATION IN A MODEL WITH ENDOGENOUS CAPITAL ACCUMULATION Suppose there are only two types of workers in the economy, either skilled or unskilled, and each supplies a distinct labor input. These workers are members of infinite-lived households that grow in size at a constant rate, and the number of these households is constantly augmented by immigration. To model an economy with both natural population growth and immigration, we treat each resident as a member of an infinite-lived immigrant dynasty. In the absence of uncertainty, the behavior of each new immigrant of type i 2 fU; Sg; unskilled or skilled, and all of his or her descendants, can be characterized as the maximization of the dynasties’ infinite horizon discounted utility function beginning at time s: Z 1 max eðr nÞðs tÞ ln ci ðs; tÞ dt; i 2 fU; Sg, (1) ci
s
subject to a time t budget constraint: k_i ðs; tÞ ¼ wi ðtÞl i þ ðrðtÞ
nÞki ðs; tÞ
ci ðs; tÞ;
8s; t;
i 2 fU; Sg
(2)
where ci(s,t) and ki(s,t) represent the time t consumption and holdings of capital of the members of a type i dynasty with arrival date s; wi(t) and r(t) represent their time t wages and the rate of return of capital; r is the subjective discount rate; and n the rate of natural population increase – the rate of growth of the dynasties themselves. The consumption rule for dynasty s at time t is: ci ðs; tÞ ¼ ðr
nÞ½oi ðtÞ þ ki ðs; tÞ;
8s; t;
i 2 fU; Sg
(3)
88
MICHAEL BEN-GAD
R 1 R u rðvÞdv where oi ðtÞ ¼ t e t wi ðuÞl i du is the present discounted value of all future income from labor of type i from time t forward. Immigrant households of type i enter the economy at time t at a rate of mi(t), and we assume that all immigrants arrive in their new homeland. Aggregate consumption and capital evolve according to: C_ i ðtÞ ¼ ðr
nÞ½rðtÞðOi ðtÞ þ K i ðtÞÞ
C i ðtÞ þ Pi ðtÞmi ðtÞoi ðtÞ;
K_ i ðtÞ ¼ wi ðtÞLi ðtÞ þ rðtÞK i ðtÞ
i 2 fU; Sg (4) (5)
C i ðtÞ
where C i ðtÞ; K i ðtÞ; and Oi ðtÞ are, respectively, the time t consumption, physical capital holdings, and the present value of future earnings aggregated overall the households with skill-level i; M i ðsÞ is the number of households with skill-level i, that have accumulated by time s; and Pi ðsÞ ¼ enðt bÞ M i ðsÞ represents the overall size of each portion of the population.8 The total labor input supplied by a household of type i 2 fU; Sg at time t is li, and the total supply of each type is Li ðtÞ ¼ Pi ðtÞl i ; i 2 fU; Sg: The production function F : R3 ! R has constant returns to scale in both types of labor and aggregate capital. Factors receive their marginal products: rðtÞ ¼ F K ðkU ðtÞ þ ZðtÞkS ðtÞ; l U ; ZðtÞl S Þ
d,
wi ðtÞ ¼ F H i ðkU ðtÞ þ ZðtÞkS ðtÞ; l U ; ZðtÞl S Þ,
(6) (7)
where ZðtÞ ¼ PS ðtÞ=PU ðtÞ is the ratio of households with skilled workers to unskilled workers in the economy at time t, and d is the rate of depreciation for physical capital. The behavior of the economy is determined by four laws of motion for per-capita consumption ci ðtÞ ¼ C i ðtÞ=Pi ðtÞ and capital ki ðtÞ ¼ K i ðtÞ=Pi ðtÞ: c_i ðtÞ ¼ ðrðtÞ
rÞci ðtÞ
k_i ðtÞ ¼ wi ðtÞl i þ ðrðtÞ
ðr n
nÞki ðtÞmi ðtÞki ðtÞ
mi ðtÞki ðtÞÞki ðtÞ
i 2 fU; Sg
ci ðtÞ i 2 fU; Sg
(8) (9)
where ki ðtÞ ¼ ðki ðtÞ ki ðt; tÞÞ=ki ðtÞ is the fractional difference between percapita capital holdings and the capital immigrants bring with them. In our simulations, we analyze the model using a family of production functions whose most general expression is the nested constant elasticity of
89
The Impact of Immigrant Dynasties on Wage Inequality
substitution (nested CES) aggregate production function with constant returns to scale developed by Sato (1967): h i1 W W (10) F ðKðtÞ; LU ðtÞ; LS ðtÞÞ ¼ ð1 aÞLU ðtÞW þ aðbKðtÞu þ ð1 bÞLS ðtÞu Þ u where KðtÞ ¼ K U ðtÞ þ K S ðtÞ is the total stock of capital.9
4. THE DYNAMIC SYSTEM, AND THE DISCOUNTED SKILL PREMIUM The sets of Eqs. (8) and (9) for each skill-type are very similar, as the savings and consumption decisions of each type of household in the economy are not very different from each other. Finding a sufficiently precise approximation of the saddle path that corresponds to this dynamic system is very difficult because the condition number of the Jacobian matrix of the linearized system is very high. To overcome this problem, we define the variables aU ðtÞ ¼ ln c~U ðtÞ and wðtÞ ¼ oS ðtÞ=oU ðtÞ; which equals ½cS ðtÞ ðr nÞkS ðtÞ=½cU ðtÞ ðr nÞkU ðtÞ; and replace the two laws of motion for consumption (8) with: a_ U ðtÞ ¼ rðtÞ
r
ðr
w_ ðtÞ ¼ ðr
nÞ
k_U ðtÞ ¼ wU ðtÞ þ ðrðtÞ
nÞe
aU ðtÞ
kU ðtÞmU ðtÞkU ðtÞ
wðtÞwU ðtÞ wS ðtÞ ðr nÞkU ðtÞ
(12)
eaU ðtÞ n
mU ðtÞkU ðtÞÞkU ðtÞ
k_S ðtÞ ¼ wS ðtÞ þ ðrðtÞ r þ ðr nÞkU ðtÞ
(11)
eaU ðtÞ
mS ðtÞkS ðtÞÞkS ðtÞ eaU ðtÞ wðtÞ
(13)
ð14Þ
Redefining the variables of the system has an additional benefit. The variable wðtÞ directly expresses the ratio between the discounted values of all the future skilled and unskilled wages. In steady state, rates of immigration for skilled and unskilled must be equal – we employ perturbation methods (see Judd, 1998) to study the dynamic behavior of the model following temporary changes in the flow of
90
MICHAEL BEN-GAD
skilled or unskilled immigrants.10 Define m as the initial steady state rate of immigration, and replace mi(t) in (11)–(14) with m þ pi ðtÞ; where pi ðtÞ is a bounded dynamic perturbation to the rate of migration by type-i workers, and e a small positive number that regulates its magnitude. Similarly, we define Z as the steady-state ratio of skilled to unskilled workers and replace the terms s(t) with Z þ xðtÞ; where xðtÞ is a bounded dynamic perturbation to the skill ratio. Defining pðtÞ ¼ fpS ðtÞ; pU ðtÞg1 t¼0 ; consumption and capital for each skilltype are all functions of p, x, and e.11 We differentiate (8), (9) with respect to e at the point ¼ 0: 2
@ _ @ aU ðt; ; p; xÞ
3
2
3 @ @ aU ðt; ; p; xÞ
2
ðr
7 6 @ 7 6 6 @ 6 w_ ðt; ; p; xÞ 7 6 @ wðt; ; p; xÞ 7 6 7 6 @ 7 6 6 7 6 7 6 6 @ k_ ðt; ; p; xÞ 7 ¼ J6 @ 7 6 6 k ðt; ; p; xÞ U U 7 6 @ @ 5 4 4 5 4 @ @ _ k ðt; ; p; xÞ @ U @ k S ðt; ; p; xÞ 3 2 OK 6 ðr nÞðwl U OU l S OS Þ 7 7 6 eaU ðr nÞk U 7 6 7xðtÞ, 6 6 kU OK þ l U OU 7 5 4 kS OK þ l S OS
nÞe
aU
kU kU pU ðtÞ
0 kU kU pU ðtÞ kS kS pS ðtÞ
3 7 7 7 7 7 5 ð15Þ
where J is a 4 4 Jacobian matrix; aU ; w; kU ; and kS are the initial steady state values of log consumption, the ratio between the present values of skilled and unskilled wages, and capital; and OK ¼ l S @2 F =@LS @K þ kS @2 F =@K@K; OU ¼ l S @2 F =@LS @LU þ kS @2 F =@K@LU ; OS ¼ l S @2 F =@LS @LS þ kS @2 F =@K@LS : To better understand how immigration affects the behavior of the model, I divide the shocks in (15) between two separate vectors that operate autonomously. The first contains the terms kU and kS, and if positive (negative) reflects the effects of capital dilution (enhancement) generated by the arrival of capital-poor (rich) immigrants, as described by Borjas (1995) in his static model and Ben-Gad (2004) in a dynamic setting. The terms OK ; OU ; and OS in the second shock vector, capture those changes in the returns to capital, unskilled wages, and skilled wages respectively, that result from the change in the composition of the labor force, i.e., x(t).
91
The Impact of Immigrant Dynasties on Wage Inequality
The portion of a policy change that operates through the channel represented by the first vector are completely transitory if the shocks have bounded support. Even permanent changes in the rate of immigration produce few changes in factor returns, or in the welfare of the native population. By contrast, the second vector can generate permanent changes in the economy even if the policy changes it represents are transitory. Small differences in the rate of immigration between the two types of immigrants accumulate over time, and permanently affect the composition of the labor force. Only if the value of all the perturbations operating through the second vector is zero in the limit – as will be the case if immigrant dynasties gradually assimilate until they replicate the overall skill distribution – do all the variables in the economy return to their original steady-state values. We solve (15) using Laplace Transforms: 2
Lu
@
3
@ aU ðt; ; p; xÞ
7 6 6 Lu @@ wðt; ; p; xÞ 7 6 7 7 ¼ ðuI 6 6 Lu @@ kU ðt; ; p; xÞ 7 4 5 Lu @@ kU ðt; ; p; xÞ
02
@ @ aU ð0; ; p; xÞ
B6 @ wð0; ; p; xÞ 1 B6 @ B6
2 6 6 6 6 6 4
JÞ B6 B6 @4
ðr
nÞe
0 0
aU
3 7 7 7 7 7 5
kU kU Lu ½pU 0
kU kU Lu ½pU
kS kS Lu ½pS 1 3 OK C 6 ðr nÞðwl U OU l S OS Þ 7 7 C 6 eaU ðr nÞk U 7 C 6 7Lu ½xC 6 C 6 kU OK þ l U OU 7 5 A 4 kS OK þ l S OS 2
3 7 7 7 7 7 5 ð16Þ
where @=@ aU ð0; ; p; xÞ and @=@ wð0; ; p; xÞ are the initial changes in the values of the two control variables.12 The matrix J has four eigenvalues, two negative and two positive. Define the two positive eigenvalues as m1 and m2. Each element of the left-hand vector must be bounded for any positive value v, including the eigenvalues m1 and m2, and yet the determinants jmh I Jj;
92
MICHAEL BEN-GAD
h 2 f1; 2g; are zero by definition. The only way for (16) to be bounded when v ¼ mh ; h 2 f1; 2g; is for the numerator, the adjoint of mh I J; h 2 f1; 2g; multiplied by the term in parentheses, to be equal to zero:
adj ½mh I
02
B6 B6 B6 JB6 B6 @4 2
@ @ aU ð0; ; p; xÞ @ @ wð0; ; p; xÞ
0 0
3 7 7 7 7 7 5
2 6 6 6 6 6 4
ðr
nÞe
aU
kU kU Lmh ½pU 0
kU kU Lmh ½pU 3
kS kS Lmh ½pS 1
OK 7 6 C 6 r nðwl U OU l S OS Þ 7 C C 7 6 C 6 eaU ðr nÞkU 7 7Lmh ½xC ¼ 0; h 2 f1; 2g 6 7 C 6 7 C 6 kU OK þ l U OU 5 A 4 kS OK þ l S OS
3 7 7 7 7 7 5
ð17Þ
both of which can then be solved for the values of @=@ aU ð0; ; p; xÞ and @=@ wð0; ; p; xÞ: Whereas the evolution of wðtÞ; the ratio between the present discounted value of skilled and unskilled wages, is not of particular interest, (19) – the change in its value at time t ¼ 0 – gives us all the relevant information for determining how a change in immigration affects discounted wage inequality over time – without explicitly calculating the impulse responses for wages or the rate of return to capital following a policy announcement. The discounted skill (or college) premium SP is the percentage difference between the present discounted value of skilled and unskilled wages: @ S ¼ w þ wð0; ; p; xÞ @ P
1 100
(18)
where w is once again the initial steady-state value of w(t). The first element in the second row of matrix J is zero. The third and fourth are zero as well, if the elasticities of substitution between the various inputs in the production function are equal – if u ¼ W in (10), only the diagonal element J22 of the second row of matrix J is non-zero, which is one
93
The Impact of Immigrant Dynasties on Wage Inequality
of the two positive eigenvalues. Hence, the value of @=@ wð0; ; p; xÞ is the second element in the third vector in (17), where the eigenvalue is J22 ¼ r n: @ wð0; ; p; xÞ ¼ @
ðr
nÞðwl U OU l S OS Þ eaU ðr nÞkU
Z
1
e
ðr nÞt
xðtÞdt
(19)
0
Result 1. If the elasticities of substitution between all the inputs are equal, the discounted skill premium is not a function of capital dilution kU and kS . Result 1 tells us that even though the rate of return to capital is affected by changes in immigration policy, for the special case of Cobb–Douglas or CES production, capital dilution itself has no effect on the discounted wage gap. A surge in immigration of type i certainly lowers the wages of all the workers of type i in the economy as long as the values of kU and kS are not negative. The wages of type jai may either rise or fall depending on whether the relative scarcity of this type of labor has a stronger effect than the decline in per-capita capital. Either way, as long as the elasticity of substitution between the inputs is constant, we can separate the analysis of the discounted skill premium from the effects on the economy generated by the dilution of the capital stock. If F is CRS and all the inputs are complementary in production, F ij 40 for all iaj then OU 40 and OS o0: Furthermore ðr nÞ=½eaU ðr nÞkU ¼ 1=oU ðtÞ; the inverse value of the per-capita present value of unskilled wages. Result 2. If the elasticities of substitution between all the inputs are equal, the production function is CRS and all the inputs are complementary in production, the college premium increases [decreases] if the present value of the shock xðtÞ; discounted by r n; is negative [positive]. From Result 2 we know how skilled and unskilled workers fare relative to each other, but we do not know what happens to the overall level of wages in the economy. To calculate the dynamic behavior of the wage levels we need to know the evolution of capital over time. We apply the inverse
94
MICHAEL BEN-GAD
Laplace transforms to (16): 3 3 2@ 2@ @ aU ðt; ; p; xÞ @ aU ð0; ; p; xÞ 7 6 @ 6 @ 7 6 wð0; ; p; xÞ 7 6 @ wðt; ; p; xÞ 7 7 6 7 Jt 6 @ 7 7¼e 6 6@ 7 6 6 @ kU ðt; ; p; xÞ 7 0 5 5 4 4 @ k ðt; ; p; xÞ 0 @ U 3 2 ðr nÞe aU kU kU pU ðqÞ 7 6 Z t 7 6 0 7 Jðt qÞ 6 e 7dq 6 7 6 k k p ð q Þ 0 U U U 5 4 kS kS pS ðqÞ 3 2 OK xðqÞ 7 6 ðr nÞðwl U OU l S OS Þ Z t 6 eaU ðr nÞk xðqÞ 7 U 7 Jðt qÞ 6 e 7dq. 6 7 6 ðk O þ l O ÞxðqÞ 0 U K U U 5 4 ðkS OK þ l S OS ÞxðqÞ
ð20Þ
The time path of capital is approximated by ki ðt; ; p; xÞ ki þ @=@ ki ðt; ; p; xÞ; i 2 fU; Sg; and to determine the time path of each wage in isolation, as well as the rate of return to capital, we insert ki ðt; ; p; xÞ together with the time path of Z(t) into (6)–(7).
5. INELASTIC AND PERFECTLY ELASTIC CAPITAL SUPPLY In the literature on immigration, capital supply is usually treated in one of two ways (see Borjas, 1999). One approach is to assume that changes in the supply of labor do not induce changes in the stock of capital – i.e., the capital stock is fixed. The other approach is to assume that capital flows freely between countries, and the capital stock adjusts immediately to accommodate the arrival of new immigrants. If capital is fixed, or in the case of this model, constrained to grow at the constant baseline rate of population growth m þ n; the wage responses to changes in immigration will generally be very strong, and changes in the rate of return to capital will be large and permanent. By contrast, if capital adjustment is instantaneous, the rate of return to capital is fixed exogenously, and if labor is completely
The Impact of Immigrant Dynasties on Wage Inequality
95
homogenous, immigration does not affect wages either. Of course, in models with heterogenous labor, a surge in immigration that upsets the composition of the labor force will alter wages. The model developed in Sections 3 and 4 stakes out a middle position between these two extremes. Capital is neither fixed, nor does it adjust immediately. Instead it accumulates gradually through the savings decisions of the agents in the economy. In the long run, the rate of return to capital is fixed as in the open economy – a function of time preference. In the short run, the rate of return to capital does respond to changes in the flow of immigration, and these changes induce changes in savings behavior and the accumulation of capital. To better understand the different implications of what we assume about the capital supply, set u ¼ W in (10). The result is a CES production function, where a and b control the relative shares of the three inputs, and the elasticity of substitution between each pair of inputs is sUSsUK ¼ sSK ¼ 1/1 W: From (7), factors receive their marginal products and the wages under CES production reduce to: w0S ðtÞ ¼ Wað1
bÞLS ðtÞW 1 =Y ðtÞ
(21)
w0U ðtÞ ¼ Wð1
aÞLU ðtÞW 1 =Y ðtÞ
(22)
where Y(t) is total output. Dividing (21) by (22), the ratio of the two wages, the instantaneous college premium, is identical regardless of what we assume about the supply of capital. Result 3. If the production function is CES, shocks x(t) and p(t) induce the same changes in the instantaneous college premium if capital is fixed, perfectly elastic, or endogenously supplied. Does this mean that the endogeneity of the capital supply is only relevant for determining the wage level or the discounted college premium SP if the elasticities of substitution between the inputs are not equal? No. First, the total levels of output Y(t) in (21) and (22) will differ depending on what we assume about the nature of the capital supply. Hence, even if wage ratios are identical, wage levels are not. Second, the rate at which we discount the evolution of each wage over time, the rate of return to capital less the natural rate of population growth, is sensitive to what we assume about the supply elasticity of capital. What we assume about capital supply does not affect the ratio between the two wages, but does affect the ratio between their discounted values. Nonetheless, the ratio between the present values of each stream of wages is
96
MICHAEL BEN-GAD
likely to be close to each other when the elasticity of substitution is constant, even if neither the levels of these wages nor the rate at which each is discounted is the same. Indeed, the lower the elasticity of substitution, the smaller the gap between the rate of return to capital under free capital flows and endogenous capital supply, and the smaller the difference in the discounted college premium.
6. MODELING THE IMPACT OF IMMIGRATION AND EDUCATIONAL ATTAINMENT OVER TIME Most of the empirical work that measures the performance of immigrants and their families focuses on labor market outcomes – either the wages they command when they work, or their rates of employment. Work by Chiswick (1978, 1986), Borjas (1985, 1987, 1992), Smith and Edmonston (1997) and Card, DiNardo, and Estes (2000) document the earnings of immigrants to the United States and work by Altonji and Card (1991), Borjas, Freeman, and Katz (1997) and Johannsson, Weiler, and Shulman (2003) focus on employment. My focus here is on educational attainment as a proxy for labor market skills (see Jasso, Rosenzweig, & Smith, 2000 for a discussion). Wages in the model derive directly from that. In addition, I am interested in the educational attainment not only of the immigrants themselves, but of their descendants. Recent empirical work on assimilation, and more particularly on educational attainment in the second generation, includes Gang and Zimmerman (2001), Riphahn (2003) (both study Germany), and van Ours and Veenman (2003) (who study the Netherlands). My focus here is on unskilled immigrants and their children. Immigrants with baccalaureate degrees remain educated to the end of their lives, while only a small number of uneducated people who arrive in North America, Australia, or Western Europe as adults, subsequently complete college. From Fig. 2 we see that most of the points are above the 451 line, suggesting that at least some of the children of unskilled immigrants do attend college. Indeed, comparing the educational outcomes of the different immigrant groups to the most convenient reference point – the older generation of U.S. natives and the children of U.S. natives – there seems to be far more upward mobility among the immigrants. Bauer and Riphahn (2004) observe a similar pattern when comparing educational attainment among the native Swiss population with that of second-generation immigrants to Switzerland, using micro-level data that includes direct measurement of parental education.
The Impact of Immigrant Dynasties on Wage Inequality
97
The immigration flows of each type and the relative size of the two types of the population, are clearly related, and yet we distinguish throughout between the two. In an economy without assimilation, a surge of immigration, even if temporary, can permanently alter the distribution between the two skill-types. By treating changes in the rates of immigration and changes in the shares of the two skill types as distinct and separate shocks in (15), I distinguish between the immediate impact of an immigrant group’s arrival, and its long-run affect on the economy as the group’s members and their families either assimilate, or exceed the general population’s rate of college completion. To simplify the analysis, I will assume that changes in the rates of immigration begin at time zero, last T years and are constant over the entire period. The change in the overall rate of immigration is defined as e, and the increase in the overall population that results from the new policy is eT 1: The share of skilled workers within the immigration surge is PS ðTÞ; and the share of unskilled workers in the immigration surge is PU ðTÞ ¼ 1 PS ðTÞ: Beginning at time Q, some of these workers, or their descendants, shift between the two categories. By time V, the share of skilled workers within this population stabilizes at PS ðV Þ; and the share of unskilled workers is PU ðV Þ ¼ 1 PS ðV Þ: Define a set of dynamic perturbations: 1 ðeT 1ÞPi ðjÞ ln 1 þ pi ðx; jÞ ¼ UðT xÞ; i 2 fU; Sg; j 2 fT; V g T Pi ð0Þ (23) where the unit step indicator function U : R ! f0; 1g returns the value of one for all numbers greater than, or equal to zero, and the value zero for all numbers less than zero. The perturbations directly affect the ratio of skilled to unskilled workers in the economy: R Zð0Þ t ðpS ðx; jÞ pU ðx; jÞÞdx 0 lðt; jÞ ¼ e 1 ; j 2 fT; V g (24) The perturbation to the ratio of skilled to unskilled workers Z(t) is the sum of two components: ðt QÞUðt QÞ ðt V ÞUðt V Þ xðtÞ ¼ lðt; TÞ þ ðl ðV ; V Þ l ðT; TÞÞ Q V (25) We replace the perturbations pi ðtÞ; i 2 fU; Sg in (15)–(20) with pi ðt; TÞ; i 2 fU; Sg: These represent the actual perturbations to the rates
98
MICHAEL BEN-GAD
of immigration, whereas the terms pi ðx; V Þ; i 2 fU; Sg correspond to a counterfactual policy under which the shares of skilled and unskilled households within the population of additional immigrants are PS ðV Þ and PU ðV Þ; in the short run and not merely in the long run. The first component in (25), lðt; TÞ; expresses the direct cumulative effect of the actual changes in immigration flows, and lðV ; V Þ expresses the long-run effect of the additional immigration, after they have completed their shift from the initial rate of college attainment Pi ðTÞ to the final rate Pi ðV Þ; i 2 fU; Sg: The first component of xðtÞ; lðt; TÞ; expresses the initial, direct effect of the immigration surge on the composition of the work force from the moment the new immigration policy is announced till time T, when the immigration surge has concluded. If lðT; TÞ40; then the share of skilled workers among the additional immigrants is higher than in the local population, and the immigrants initially raise the value of Z(t). If lðT; TÞo0; the share of skilled workers is lower, and the immigrants initially lower the value of Z(t). The second component in (25) expresses movements of the immigrants between the two skill categories and their effect on the overall labor force composition between the periods Q and V. A few examples of possible time paths for Z(t) will help illustrate the behavior of the model. The left-hand panels of Fig. 3 illustrate the evolution of Z(t) following a surge in unskilled immigration only, and the right-hand panels of Fig. 3 illustrate the evolution of Z(t) following a surge in skilled immigration. The black curves represent the behavior of Z(t) if there is no assimilation, the dark gray curves represent the behavior of Z(t) if all the immigrants assimilate, and the light gray curves correspond to instances where the shares of skilled and unskilled workers within the immigrant population reverse over time. Suppose the skill composition of the workers in the immigration surge does not match the prevailing composition of the host country, but neither the immigrants nor their descendants switch between the two categories after they arrive. In this case Pi ðV Þ ¼ Pi ðTÞ; i 2 fU; Sg; which implies lðV ; V Þ ¼ lðT; TÞ: The second term in (25) is zero, the behavior of x(t) is determined by lðt; TÞ; and Z(t) either declines or increases until period T, and then remains fixed at its new steady state value. Each set of black curves in Fig. 3 corresponds to the extreme sub-cases in which an immigration surge is uniformly composed of either unskilled (left-hand side of Fig. 3) or skilled (right-hand side of Fig. 3) workers. Note that in each column the black curves are identical for different values of Q and V, and serve as points of reference.
99
The Impact of Immigrant Dynasties on Wage Inequality
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
Fig. 3. The Time Paths of Z(t), the Ratio of Skilled to Unskilled Workers, for Different Degrees of AssimilationFollowing Different Influxes of Skilled and Unskilled Immigrants. a) k(T, T)o0, Q ¼ 0, V4T; b) k(T, T)o0, Q ¼ 0, V4T; c) k(T, T)o0, Q ¼ 0, V44T; d) k(T, T)40, Q ¼ 0, V44T; e) k(T, T)o0, T4Q; f) k(T, T)40, T4Q; g) k(T, T)o0, ToQ; h) k(T, T)40, ToQ.
100
MICHAEL BEN-GAD
Suppose the surge of additional immigrants initially upsets the balance between skilled and unskilled workers in the labor force, but gradually, over the time period between Q and V, these immigrants assimilate until the shares of skilled and unskilled workers exactly mimics that of the general population. Under this scenario Pi ðV Þ ¼ Pi ðTÞ ¼ Pi ð0Þ; i 2 fU; Sg and therefore lðV ; V Þ ¼ 0: Consider a surge in unskilled immigration. If the process of assimilation begins immediately and ends soon after the last of the additional immigrants arrive (Panel (a) of Fig. 3), the dark gray curve barely declines below its initial value. In Panel (c) assimilation begins immediately, but the process lasts longer and the decline in Z(t) is steeper. Immigrants, or more likely their descendants, may not merely assimilate. As we see in Fig. 2, few of the women among Greek immigrants to the United States have college degrees, but a disproportionate fraction of second-generation Greek-American women do. Similarly, the value of Z(t) may first decline because of a surge of unskilled immigration, but rise above its initial value by time V. Similarly it is possible (though a good deal less likely) that a surge in immigration may initially raise, but ultimately lower the value of Z(t). If the value of Q is set above T, then the value of Z(t) first behaves according to lðt; TÞ; before beginning its ascent (the light gray curves in Panel (g) of Fig. 3) or descent (the light gray curves in Panel (h) of Fig. 3). However, if the value of V is no longer zero, but below T, then the reversal in the direction of Z(t) begins at time Q (the light gray curves in Panel (e) or Panel (f) of Fig. 3). If Q ¼ 0; then as in Panels (a–c) the direction of Z(t) may be completely determined by the value of Pi ðV Þ:
7. PARAMETERIZING THE MODEL Between 1990 and 1999 the net rate of migration to the United States was just under 3.2 per thousand. Although a much larger fraction of immigrants have less than nine years of schooling, the percentage of the foreign-born with baccalaureate degrees closely matches that of the general population – 25.8% of foreign-born people in the United States over the age of 25 have college degrees, as compared to 25.6% of the total U.S. population. For the initial stock of skilled and unskilled workers we set PS ð0Þ ¼ 0:256 and PU ð0Þ ¼ 0:744; and set the steady state rates of immigration for both skill types to mS ¼ mU ¼ 0:0032:13 If the rates of legal and illegal immigration to the United States during the decade of the 1990s
The Impact of Immigrant Dynasties on Wage Inequality
101
carry forward, and the rate of out migration continues to hold steady at one per thousand, foreign migration will augment the U.S. population with close to 10 million additional people over the course of this decade. There is no readily available data on the financial assets or physical capital that new immigrants to the United States bring. Given the large gap in income between sending countries and the United States, and the relative youth of most immigrants when they arrive, it is unlikely that capital holdings for the typical immigrant, skilled or unskilled, approaches U.S. per-capita capital holdings for either skill type. I set kS ¼ kU ¼ 1; which implies that after financing their move to the United States and setting up a household, immigrants have exhausted their savings. I also assume that both types of workers supply the same amounts of labor and set l U ¼ l S ¼ 1: The ratios of mean earnings and income for households, as well as individuals, with bachelor’s degrees to those without, range from 2.13 to 2.71, as measured by the U.S. Census. The 1998 Survey of Consumer Finances reports on net wealth as well as income and earnings. The ratio of mean earnings is 2.35, that of income is 2.3 while net wealth is 3.3. The gap between median earnings and wealth is smaller – 2.4 versus 3.06. In steady state, the ratio of capital held by skilled and unskilled agents must be equal to the ratio of their wages. I choose an intermediate number 2.7, and combine this with the 1991–2000 average share of capital in national income, 28.3%, to set the values of the parameters in the production function for different elasticities of substitution. Both the cross-country estimations of the nested CES production function (10) by Fallon and Layard (1975) and Duffy, Papageorgiou, and PerzSebastian (2004) and the time-series estimations using U.S. data by Krusell, Ohanian, Rı´ os-Rull, and Violante (2000) and Swedish data by Lindquist (2003), find that the difference between the values of the parameters W and u in (10) is statistically significant – implying the existence of the capital skill complementarities first postulated by Griliches (1969). I simulate the baseline model setting W ¼ 0:401 and u ¼ 0:495 in (10) to match the estimates by Krusell et al. (2000) (their distinction between skilled and unskilled workers based on college education matches my own). These parameter values correspond to elasticities of substitution between capital and unskilled labor sUK, and between skilled and unskilled labor sUS, that are equal to 1=1 W ¼ 1:67; and an elasticity of substitution between capital and skilled labor sSK equal to 1=1 u ¼ 0:67:14 When W is set equal to u, the production function (10) becomes the standard CES function with three inputs, and when both approach the value of one in the limit we have the Cobb–Douglas function (Table 1).
102
MICHAEL BEN-GAD
Table 1.
Parameterization of Baseline Model.
Preferences, Technology and Factor Shares r ¼ 0:0495 d ¼ 0:061 a ¼ 0:283 b ¼ 0:482
Matches average rate of return of capital of 5% Average rate of depreciation on capital: 1991–2000a Average U.S. capital’s share of national income: 1991–2000a Matches ratio of initial earningsa,b,c
Population n ¼ 0:0067 mS ¼ mU ¼ 0:0034 kS ¼ kU ¼ 1 PS ð0Þ ¼ 0:256 d ¼ 2:7
Average U.S. natural rate of population growth: 1991–2000b Average U.S. rate of net migration: 1991–2000b Immigrants arrive without physical capital Population with college degreesb Ratio of initial earnings and wealth for households with/without college degreesb,c
a
Bureau of Economic Analysis (2004); U.S. Census Bureau c 1998 Survey of Consumer Finances. b
8. A 20-YEAR SURGE IN IMMIGRATION 8.1. Wage Levels Comparing the period 1980–1989, and the period 1990–1999 in Fig. 1, Panel (b) the rate of NIM rose by just over 0.3 per thousand. In the next two sections I consider the implications of an additional rise of a similar magnitude that lasts for two decades. To begin with, let us suppose that all the additional immigrants are people without college degrees. The overall rate of immigration rises from 3.2 to 3.5, per thousand and the rate of immigration by the unskilled rises to just over 3.6 per thousand. How many additional people does such a rise in immigration imply? If present trends continue, the United States will absorb about 20 million legal and illegal immigrants over the course of two decades, and 15 million of them will not have Baccalaureate degrees. The change considered here need not entail an increase in the number of legal immigrants alone. A slight curtailment in enforcement efforts along the border could easily cause the number of unskilled immigrants to rise by the additional 75,000 people per year (one million-and-three-quarters over the course of 20 years) that we are considering here.
103
The Impact of Immigrant Dynasties on Wage Inequality
Consider first the effect of the policy change on each type of wages when assimilation does not occur. Setting PS ðTÞ ¼ PS ðV Þ ¼ 0; and the elasticities of substitution in the production function (10) sSK ¼ 0:67 and sUK ¼ 1:67; the surge of unskilled immigration produces a permanent change in the skill composition of the work force, and generates the changes in the wages of unskilled workers shown in the upper left-hand corner of Fig. 4. The dotted lines represent the impulse response in an economy with an inelastic supply of capital. Here, because of the permanent dilution of the capital stock, the long-run response of unskilled wages, a drop of just above 0.35% for the Nested CES production function, is 20% higher than the long-run decline in wages if capital is either completely elastic or endogenously determined. By contrast, the rise in skilled wages of over three-tenths of a percent if capital
ΠS (T )=0, ΠS (V )=PS(0) 0.008
0.006
0.006 ∆wU(t)/wU
∆wU(t)/wU
ΠS(T )=0, ΠS (V )=0 0.008
0.004 0.002 10
20
30
40
0.004 0.002
50
−0.002
10
20
−0.004
40
50
40
50
−0.004 t
t
ΠS(T )=0, ΠS (V )=1
ΠS(T )=1, ΠS (V )=1
0.008
0.008
0.006
0.006
0.004
0.004
∆wU(t)/wU
∆wU(t)/wU
30
−0.002
0.002 10
20
30
−0.002
40
0.002
50
10
20
30
−0.002
−0.004
−0.004 t
t
Fig. 4. Nested CES Production Function sSK ¼ 0:67; and sUK ¼ sUS ¼ 1:67: Impulse Response for Unskilled Wages Following a 20 Year Surge in the Rate of Immigration from 3.2 to 3.5 Per Thousand, Q ¼ 25 and V ¼ 45: Note: The Solid, Dashed, and Dotted Curves Represent Respectively, the Impulse Responses Generated by the Model with Capital Supply that is Elastic, Completely Elastic, and Inelastic.
104
MICHAEL BEN-GAD ΠS(T ) = 0, ΠS (V ) = 0
ΠS(T ) = 0, ΠS (V ) = PS (0)
0.005
0.005
∆wS(t)/wS
10
20
30
40
50
10
−0.005
−0.005
−0.01
− 0.01
−0.015
−0.015
−0.02
−0.02
−0.025
20
−0.025
t
30
40
50
40
50
t
ΠS(T )=0, ΠS (V )=1
ΠS(T )=1, ΠS (V )=1
0.005
0.005 t
∆wS(t)/wS
10
20
30
40
50
10
− 0.005
− 0.005
−0.01
−0.01
− 0.015
−0.015
− 0.02
−0.02
− 0.025
−0.025 t
20
30
t
Fig. 5. Nested CES Production Function sSK ¼ 0:67; and sUK ¼ sUS ¼ 1:67: Impulse Response for Skilled Wages Following a 20 Year Surge in the Rate of Immigration from 3.2 to 3.5 Per Thousand, Q ¼ 25 and V ¼ 45: Note: The Solid, Dashed, and Dotted Curves Represent Respectively, the Impulse Responses Generated by the Model with Capital Supply that is Elastic, Completely Elastic, and Inelastic.
is elastic, in the upper left-hand corner of Fig. 5, is well over twice the rise in skilled wages if capital is fixed. The impulse responses in Figs. 4 and 5 are all relatively modest because they were calculated for an economy in which the elasticity of substitution between the two types of labor is high. If the elasticity of substitution between all the inputs is lower, the effect of immigration on both skilled and unskilled wages is larger. Setting all the elasticities of substitution to twothirds, the long-run drop in unskilled wages is close to nearly nine-tenths of a percent if capital is fixed, and nearly sixth-tenths of a percent if capital is elastic. The long-run rise in skilled wages is one-third of a percent if capital is fixed, and nearly twice that amount if capital is elastic (see Figs. A1 and A2 in the appendix).
The Impact of Immigrant Dynasties on Wage Inequality
105
Whether or not there is free movement of capital, or if capital is elastically supplied but only from internal savings, the long-run effect on wages is always the same. The difference in the responses of wages is only a shortterm phenomenon, and for the case of unskilled immigration, this difference is very small. In general, the response of wages if capital is endogenously determined falls between two extremes, i.e., between the case where the capital supply grows at an exogenous rate, and that where capital is perfectly elastic – but is much closer to the latter than the former. Of course it must be emphasized that in the overlapping dynasties model there is no representative consumer. Ensuring aggregability requires a logarithmic utility function, and hence a relatively high degree of intertemporal elasticity of substitution in consumption. If the elasticity of substitution were lower, the accumulation of capital would be slower, and the short-term response of wages in the case of endogenously supplied capital would not be quite so close to the responses generated by the model with perfectly elastic capital. The upper right-hand graphs in Figs. 4 and 5 illustrate the response of wages if PS ðTÞ ¼ 0 but PS ðV Þ ¼ PS ð0Þ and PS ð0Þ ¼ 0:256: In the 25th year (Q ¼ 25), five years after the immigration surge has ended, the descendants of these additional unskilled immigrants enter the labor force. Of these new workers 25.6% are skilled, exactly mirroring the proportion of skilled workers within the larger population. By year 45 (V ¼ 25) this additional population has completely assimilated. The arrival of these additional workers dilutes capital and causes both skilled and unskilled wages to drop in the short run, but will not affect long-run wages unless capital is inelastic. If capital supply is inelastic, unskilled wages drop by between a third of a percent by year 20 (as in Fig. 4), or just under nine-tenths of a percent (as in the appendix), depending on the elasticity of substitution. The drop is caused by the combined effect of an increase in the relative share of unskilled workers in the labor force, and the overall rise in the size of the labor force itself. In the long run, as the descendants of immigrants assimilate, only the latter of these two effects remain, and the unskilled wage recovers approximately half its short-term loss. Wages for skilled workers initially climb, as the additional immigrants upset the balance between skilled and unskilled workers. In each case, this change in the composition of the labor force initially dominates the effects of capital dilution. If capital supply is completely elastic or endogenous, the skilled wage also returns to its initial level, once the immigrants or their descendants completely assimilate. If capital is in fixed supply, the skilled wage initially rises but ultimately declines. If sSK ¼ 0:67 and sUK ¼ 1:67; the wage is one-tenth of a percent
106
MICHAEL BEN-GAD
higher by year 20, but then begins to gradually decline, until it is half a percent below where it was before the new policy was initiated. If both elasticities of substitution are two-thirds, the wage initially rises by one third but is ultimately four-tenths of a percent lower, and if both elasticities are one the wage first rises by two-tenths of a percent and then by half a percent, until it is three-tenths of a percent below its initial level (see Figs. A1 and A2 in the appendix). The greater the complementarity between skill and capital, the more capital dilution mitigates the initial rise in skilled wages generated by the change in the composition of the labor force. In every case, the permanent dilution of the capital stock guarantees a lower skilled wage in the long run. The lower left-hand graphs in Figs. 4 and 5 are perhaps the most interesting – they show the behavior of wages when PS ðTÞ ¼ 0 and PS ðV Þ ¼ 1: A surge of immigrants, all unskilled, enter the country over the course of 20 years. Starting in year 25 they or their descendants begin a remarkable transformation. Rather than merely assimilating as in the previous experiment, this group crosses in its entirety from the skilled to the unskilled category during the course of 20 years. Wages for unskilled workers initially decline, and wages for skilled workers rise as these immigrants arrive and join the workforce. Starting in year 20, if capital is endogenous, or in year 25 if capital is either inelastic or completely elastic, unskilled wages begin to rise and skilled wages to decline. If the capital supply is completely elastic, both sets of wages pass through their original levels before continuing to rise or decline, at precisely the same time – year 30. If the supply of capital is endogenous, the remaining effects of capital dilution will cause unskilled wages to begin to decline below their original level during the second half of the 29th year – approximately half a year before skilled wages have completely recovered. If capital supply is exogenous, the gap between the point when skilled wages reach their initial level and the moment when unskilled wages have completely recovered will be much larger – the former occurs between year 26 or 27 (depending on the elasticities of substitution), and the latter at year 35. This gap of a few years, when both wages are below their initial levels, directly results from capital dilution. In the long run, wages in the lower left-hand sides of Figs. 4 and 5 are identical to the long-run wages in the lower right-hand sides. The latter represent the behavior of wages when all the immigrants arrive as skilled workers. The long-run changes in wages are the same for a completely elastic and an endogenously determined capital supply – capital dilution affects the former not at all, and the latter only in the short-run. Indeed, the
The Impact of Immigrant Dynasties on Wage Inequality
107
long-run drops in skilled wages and increases in unskilled wages are nearly symmetric in this particular example, ranging from nine-tenths of a percent when the elasticity of substitution between the two types of labor is at its highest value, to approximately 1.8% when the elasticity of substitution is lowest. If the supply of capital is fixed or grows at a fixed rate, capital dilution is permanent. As long as all the elasticities of substitution are identical (as in Figs. A1 and A2 in the appendix), capital dilution’s effect on skilled and unskilled wages is the same: each drops by one half or three-quarters of a percent, depending on whether the elasticity of substitution is one or 0.67. If the elasticity of substitution between capital and unskilled labor is raised to 1.67, capital dilution generates a much larger drop in both the short and long-run wages of skilled workers, as their labor complements the capital that is now relatively more scarce. Hence, the permanent rise in wages for unskilled workers is only four-tenths of a percent, while the drop in the wages of skilled workers is 2.3%.
8.2. The Discounted College Premium As we have seen, a change in the rate of immigration does not change the wage structure overnight. A shift in policy means a change in the flow of immigrants that gradually alters the size and composition of the workforce. In addition, the economy does not necessarily adjust immediately to these changes. Unless the amount of capital adjusts immediately or is permanently fixed, a surge of immigration will generate an initial shock to the savings rate, followed by the gradual convergence of the size of the capital stock to its new steady-state level. Finally, immigrants themselves, or at least their descendants, may not permanently remain within their initial skill categories, generating further disturbances to wages long after the surge in immigration has subsided. The impact of immigration on wage inequality must reflect the behavior of wages over time. Encapsulating this behavior into a number requires us to discount by the rate of return available to these workers, and this rate of return is just as sensitive to immigration policy as the wages themselves, unless we assume from the outset that the capital supply is perfectly elastic and that its rate of return is determined exogenously. The need to take into account the endogeneity of the discount rate is apparent if we compare the time paths of wages in Figs. 4 and 5 to the behavior of the discounted college premium in Table 2. The time paths of
108
MICHAEL BEN-GAD
Table 2. The Values of the Discounted Skill Premium SP, Following a Rise in the Rate of Immigration from 3.2 to 3.5 Per Thousand During the Course of Two Decades. PS ðTÞ ¼ PS ðV Þ ¼ 0 PS ðTÞ ¼ PS ðV Þ ¼ PS ð0Þ PS ðTÞ ¼ PS ðV Þ ¼ 1 Cobb–Douglas: sUK ¼ sUS ¼ sSK ¼ 1 Fixed capital Completely elastic capital Endogenous capital
171.45 171.45 171.45
170.00 170.00 170.00
165.89 165.86 165.76
CES: sUK ¼ sUS ¼ sSK ¼ 0:67 Fixed capital Completely elastic capital Endogenous capital
172.16 172.17 172.16
170.00 170.00 170.00
163.91 163.85 163.67
Nested CES: sUK ¼ sUS ¼ 1:67; sSK ¼ 0:67 Fixed capital Completely elastic capital Endogenous capital
170.87 171.08 171.05
169.38 170.00 169.89
165.20 166.88 166.52
Note: The baseline value of SP is 170%.
wages when capital supply is endogenously determined, fall between the two extreme cases of completely elastic or inelastic capital supply, and closer to the former than the latter. By contrast, the effect of immigration on the discounted college premium does not necessarily follow this pattern. The first column of Table 2 illustrates the effect of the immigration surge on the discounted college premium, if all the additional immigrants are and remain unskilled workers. The new policy implies that the premium will rise from an initial value of 170% to anywhere from 170.87 to 172.17%, depending on what we assume about the elasticity of capital supply and, more importantly, about the elasticity of substitution between the factors of production. Changes in inequality of this magnitude might appear small, but considering the magnitude of the change in policy – a rise in the immigration rate of a mere 3 per 10,000 during 20 years – even the smallest of these changes are quite impressive. Even more impressive is the effect of the immigration surge when it is comprised wholly of skilled workers, as in the last column of Table 2. The small rise in the number of skilled immigrants over the course of 20 years has the potential to lower the value of SP from 170 to 166.52% if capital is supplied endogenously, and capital and skill are relative complements.
The Impact of Immigrant Dynasties on Wage Inequality
109
The drop in the discounted skill premium is even larger under nearly all the other scenarios in Table 2. It is important to keep in mind that the number of additional skilled immigrants who arrive under the policy considered here is approximately 75,000 per year. This number is similar to the 65,000 visas available under the H1-B visa program through most of the years of its existence, and gives us some idea of what that program’s impact on wage inequality might be. How robust are these results? The impact of a surge of skilled immigration of this magnitude might cause a much smaller drop in the value of SP – perhaps from 170 to 169% – but only if the elasticities of substitution between the different inputs are very high (in this model close to five). In the middle column of Table 2 – PS ðTÞ ¼ PS ðV Þ ¼ PS ð0Þ – the distribution of skills among the immigrants exactly replicates that of the general population. As long as the cross elasticities of substitution are equal across the different inputs, such an influx of immigration has no effect on the college premium. Even if the supply of capital is not perfectly elastic, capital dilution causes both types of wages to drop by the same proportion. Only if the elasticities are not equal and capital supply is not completely elastic can a surge in immigration of this type lead to a change in the discounted skill premium. The third to last row of Table 2, in the middle column, combines the nested CES production function and fixed capital supply for a surge of immigration whose rates of college attainment match those of the general public. This entry isolates the effect of capital dilution on the discounted college premium – at most a drop from 170 to 169.38%. A rise in immigration lowers wages for everyone because everyone’s labor complements capital. However, the higher degree of complementarity between skilled labor and the factor of production that is being diluted ensures that their wages decline more, and the wage premium declines. If capital is endogenously determined, capital dilution is transitory and the effect of the immigration surge is smaller – the discounted wage premium is 169.89%.
8.3. Assessing the Impact of Skill Acquisition The results in Table 2 reveal that a relatively small influx of unskilled immigrants substantially raises the discounted college premium if these immigrants and their descendants remain permanently unskilled. A similar sized influx of skilled immigrants produces an opposite and much larger change. How might these results change if the immigrants, or at least their
110
MICHAEL BEN-GAD
descendants, switch between the two skill categories? More specifically, how much can college attainment by immigrants subsequent to their arrival in the United States as unskilled workers mitigate or reverse their initial impact on wage inequality? How much does future college attainment among the members of the second generation affect the discounted college premium? Suppose once again that the additional immigrants arrive initially as unskilled workers, PS ðTÞ ¼ 0; but over time a fraction of these immigrants or their children join the ranks of the skilled, until ultimately the distribution between skilled and unskilled within this additional population matches that of the general population – PS ðV Þ ¼ PS ð0Þ: As in the second columns of Figs. 4 and 5, we are again modeling a process of assimilation, at least in terms of educational attainment. Table 3 illustrates the implications of this process on the discounted skill premium for the nested CES specification of the production function, sUK ¼ sUS ¼ 1:67; sSK ¼ 0:67: The same calculations, with all the elasticities of substitution set equal to each other, are presented in Tables A1 and A2 in the appendix. As should be expected, the values in Table 3 fall somewhere between the corresponding entries in the first two columns of Table 2 (last three rows). The lower the values of Q and V – the earlier the date at which the process of assimilation begins, and the earlier the date at which it is completed – the closer the values of the discounted college premium are to the values in the second column of Table 2. Higher values of Q and V correspond to delayed and extended periods of assimilation, and are closer to the higher values in the first column. As in the second column of graphs in Figs. 4 and 5, unskilled wages rise and skilled wages decline starting in period Q. Even if Q is large, implying that only members of the second generation attend college, assimilation mitigates in the long run, the initial rise in inequality. Compare the discounted college premium in Table 2 for the values PS ðTÞ ¼ 0 and PS ðV Þ ¼ 0; and the nested CES production function with their corresponding values in Table 3. If Q ¼ 20 and V ¼ 40; the discounted college premium in Table 3 is 170.26% if the supply of capital is exogenously determined, and 170.64% if capital supply is perfectly elastic. The corresponding values in Table 2 are 170.87% and 171.08%. The less elastic the supply of capital, the higher the rate at which changes in wages that take place after time Q are discounted, and yet when capital supply is inelastic it seems that the relative impact on the discounted college premium is larger. Hence, we conclude that the disproportional downward pressure on skill wages induced by capital dilution dominates the higher discount rate.
111
The Impact of Immigrant Dynasties on Wage Inequality
Table 3. V
Nested CES Specification of the Production Function: sUK ¼ sUS ¼ 1:67; sSK ¼ 0:67.
Q ¼ 0
Q ¼ 10
Q ¼ 20
Q ¼ 30
Q ¼ 40
Q ¼ 50
169.71 169.92 170.08 170.20 170.30
— 170.13 170.26 170.37 170.45
— — 170.40 170.49 170.55
— — — 170.57 170.63
— — — — 170.69
170.24 170.39 170.50 170.59 170.67
— 170.54 170.64 170.71 170.77
— — 170.74 170.80 170.85
— — — 170.87 170.91
— — — — 170.95
170.18 170.34 170.45 170.54 170.61
— 170.49 170.59 170.67 170.72
— — 170.69 170.75 170.80
— — — 170.82 170.86
— — — — 170.90
Inelastic capital 20 30 40 50 60
169.38 169.63 169.82 169.97 170.09
Completely elastic capital 20 30 40 50 60
170.00 170.18 170.32 170.43 170.52
Endogenous capital 20 30 40 50 60
169.93 170.11 170.26 170.37 170.46
Note: The values of the discounted skill premium SP, following a rise in the rate of immigration from 3.2 to 3.5 per thousand during the course of two decades, for different values of V and Q. The immigrants are initially unskilled PS ðTÞ ¼ 0; but beginning at time Q, some of these immigrants, or their descendants, become skilled workers. By time V the share of skilled workers within this population stabilizes at PS ðV Þ ¼ PS ð0Þ where PS ð0Þ ¼ 0:256 is the prevailing share of skilled workers within the general population. The baseline value of SP is 170%.
The arrival of a small number of unskilled immigrants generates a significant rise in inequality in Table 2, regardless of how the production function is specified, or the nature of the capital supply. By contrast, an immigration surge that replicates the existing distribution of skills in the general population, and the underlying immigration flow, slightly lowers the discounted college premium if capital and skill are relative complements and the supply of capital is not completely elastic. Hence, a surge of immigration need not raise inequality, even if the immigrants and their descendants merely match the skill distribution decades in the future, and in the meantime increase the relative size of the unskilled workforce. Of course,
112
MICHAEL BEN-GAD
only if as in the top part of Table 3, the supply of capital is inelastic, and the process of assimilation begins soon after these immigrants begin to arrive. Otherwise assimilation mitigates the effects of immigration on the value of SP, but cannot reverse it. The values of the discounted college premium in Table 3 are far more sensitive to the value of Q than to the value of V. The point at which the share of college-educated people has returned to its initial value of PS(0) is not as important as how quickly the first immigrants begin their transformation into skilled workers. This difference becomes even more obvious if we consider a more radical degree of transformation on the part of the new immigrants. What if unskilled immigrants, or their children, do not merely match the general level of educational attainment but far surpass it? Again, consider the most extreme scenario: the immigrants arrive initially as unskilled workers, PS ðTÞ ¼ 0; but over time they or their children all join the ranks of the skilled, PS ðV Þ ¼ 1: In nearly every case the values of the discounted skill premium in Table 4 (calculated for nested CES specification of the production function, sUK ¼ sUS ¼ 1:67; sSK ¼ 0:67) fall well below 170%. The same is true if the elasticities of substitution are identical, as in Tables A3 and A4. Consider the following scenario: a man and woman arrive in the United States at age 25, at the very beginning of the immigration surge, meet and start a family. Neither is college educated and both spend the next 40 years employed as unskilled workers. These parents place a high value on education and all their American-born children complete college. The first child graduates and joins the labor force perhaps 25 years after the parents’ arrival. Such a high level of college attainment among the second generation does not merely mitigate the negative effect on wage inequality of their immigrant parents’ decades of participation in the labor force, but completely reverses it. Hence, admitting these two people into the United States raises the college premium for 30 years, but seen from the perspective of several generations, discounted by the economy’s prevailing rate of return, the arrival of this family has lowered overall wage inequality.
9. IMMIGRANT GROUPS AND WAGE INEQUALITY The simple examples of immigration surges in Section 8 are completely hypothetical and hardly match the heterogeneity of the bi-generational patterns of educational attainment for the different U.S. immigrant groups
113
The Impact of Immigrant Dynasties on Wage Inequality
Table 4. V
Nested CES Specification of the Production Function: sUK ¼ sUS ¼ 1:67; sSK ¼ 0:67.
Q ¼ 0
Q ¼ 10
Q ¼ 20
Q ¼ 30
Q ¼ 40
Q ¼ 50
166.46 167.26 167.87 168.34 168.70
— 168.08 168.59 168.98 169.28
— — 169.12 169.45 169.99
— — — 169.79 169.99
— — — — 170.21
167.79 168.37 168.83 169.18 169.45
— 168.97 169.35 169.65 169.88
— — 169.74 169.99 170.18
— — — 170.24 170.40
— — — — 170.57
167.63 168.24 168.71 169.06 169.34
— 168.86 169.25 169.54 169.77
— — 169.63 169.88 170.07
— — — 170.13 170.29
— — — — 170.46
Inelastic capital 20 30 40 50 60
165.20 166.14 166.87 167.44 167.90
Completely elastic capital 20 30 40 50 60
166.88 167.57 168.11 168.53 168.87
Endogenous capital 20 30 40 50 60
166.64 167.38 167.94 168.38 168.73
Note: The values of the discounted skill premium SP, following a rise in the rate of immigration from 3.2 to 3.5 per thousand during the course of two decades, for different values of V and Q. The immigrants are initially unskilled PS ðTÞ ¼ 0; but beginning at time Q, all these immigrants, or their descendants, become skilled workers. By time V the share of skilled workers within this population stabilizes at PS ðVÞ ¼ 1: The baseline value of SP is 170%.
in Fig. 2. Compare for example, the shares of college-educated male immigrants to the United States from Poland and the Philippines, and their corresponding second generations. A relatively small share of male immigrants from Poland aged 45–64 have college degrees – 21.3 versus 25.6% for the overall U.S. population. However, a very large fraction of American-born sons aged 25–44 with Polish-born fathers have college degrees – 64.1%. Admitting Polish immigrants prior to 1975 raised the gap between the wages of skilled and unskilled workers, but a generation later, the high levels of educational attainment achieved by the second generation have had the opposite effect on the skill premium. Indeed, compare the experience of Polish immigrants over the course of two generations with immigrants from the Philippines. The share of male
114
MICHAEL BEN-GAD
immigrants from the Philippines aged 45–64 with college degrees is much higher – 45.8% – but among second generation Americans aged 25–44 with Philippine-born fathers, the share completing college is significantly lower than for their Polish-American counterparts – only 42.3% (indeed, there is even some reversion to the mean, as members of the second generation are less educated than members of the first). How do we compare the impact of the entire experience of Philippine immigration, including subsequent generations, with immigration from Poland? Once again PS(0) is the initial steady-state share of college-educated people in the economy prior to the change in policy. The initial share of college educated among the additional immigrants is PS ðTÞ; and PS ðV Þ is the share of college educated among these same immigrants or their descendants at time V. Starting at time Q, the share of skilled workers among these additional dynasties begins changing (linearly) until the share of skilled workers among these same additional immigrants or their descendants is PS ðV Þ: I focus on the case of endogenously determined capital supply, and once again set ¼ 0:0003; the value of T equal to 20, the value of Q to 25, and the value of V to 45, in (23). I insert the resulting values of the perturbations into (24), and use these to calculate the value of x in (25). Setting (19) to different values of SP yields an implicit function in the values PS ðTÞ and PS ðV Þ: These implicit functions are represented as iso-curves in Figs. 6 and 7, corresponding to the Nested CES specification, with the initial steady state discount rate set at 0.05 and 0.03 respectively (the Cobb–Douglas and CES production functions appear respectively in Figs. A3 and A4 in the appendix). These curves are superimposed on the points representing the intergenerational rates of college completion for different immigrant groups in Fig. 2. What do we learn from Figs. 6 and 7? First, though the function SP is a complicated non-linear function, it appears nearly linear on the PS ðTÞ and PS ðV Þ plane. Second, the distances between the iso-curves appear constant – changing the values of either PS ðTÞ or PS ðV Þ generates nearly linear changes in the value of the discounted college premium. Each iso-curve that correspond to S P ¼ 170% bisects a plane between two regions – immigrant groups with values of PS ðTÞ and PS ðV Þ that fall to the right of the curve lower the discounted college premium, while those with values to the left of the curve raise it. In Figs. 6 and 7 the curve corresponding to SP ¼ 170% passes to the left of the point (25.6,25.6). Therefore even if both the immigrant and second generation fall below the general levels of educational attainment, the discounted college premium
China
USSR Ireland
Greece
US
(25.6, 25.6)
20
USSR
40
Canada
Portugal
US
(25.6,25.6)
20
Puerto Rico
68%
P =1
S
%
169
5% 68.
P =1
S
P=
S
5% 69.
P =1
S
0
70%
0 100
P =1
5% 70.
68%
P =1
Mexico
S
P =1
S
S
5% 68.
P =1
20 40 60 80 Men aged 45−64 born in US or arrived before 1975 by place of birth
Philippines
UK
S
91%
P=6
S
5% 69.
P =1
S
70%
P =1
.5%
0
S
70
P =1
S
0
Cuba
Dom. Rep.
Puerto Rico
Mexico
Germany
5% 67.
Dom. Rep.
Cuba
Ireland Colombia Italy
P =1
5% 67.
Canada
Poland Greece
60
S
Philippines
Italy
P =1
Portugal
S
40
61%
UK Germany Colombia
P=7
67%
P =1
Poland
60
India
80
S
China
US born women aged 25−44 by mother's place of birth
80
S
US born men aged 25−44 by father's place of birth
India
20 40 60 80 Women aged 45−64 born in US or arrived before 1975 by place of birth
The Impact of Immigrant Dynasties on Wage Inequality
100
100
100
Fig. 6. Nested CES Production Function: sUK ¼ sUS ¼ 1:67; sSK ¼ 0:67: Iso-Curves for the Values of SP Following a 20 Year 25% Rise in the Rate of Immigration. Note: The Baseline Rate of Return of Capital is Set to 0.05, and Q ¼ 25 and V ¼ 45: Points Represent the Percentage of the U.S. Population with Four-Year College Degrees by Age, Sex, Birthplace, and Parent’s Birthplace from Fig. 2. 115
116
100
100
=1 % 68
US born women aged 25−44 by mother's place of birth
P
S
% 68
20
P
P
. 68 =1
100
Women aged 45−64 born in US or arrived before 1975 by place of birth
Fig. 7. Nested CES Production Function: sUK ¼ sUS ¼ 1:67; sSK ¼ 0:67: Iso-Curves for the Values of SP Following a 20 Year 25% Rise in the Rate of Immigration. Note: The Baseline Rate of Return of Capital is Set to 0.03, and Q ¼ 25 and V ¼ 45: Points Represent the Percentage of the U.S. Population with Four-Year College Degrees by Age, Sex, Birthplace, and Parent’s Birthplace from Fig. 2.
MICHAEL BEN-GAD
80
5%
60
% 69 =1
5%
40
S
. 69 =1
20
S
P
% 70 =1
S
P
S
0
5%
0
. 70 =1
% 71 =1
% .5 68
100
P
P
S
=1
%
80
Men aged 45−64 born in US or arrived before 1975 by place of birth
S
P
69 =1
60
Mexico
S
P
% .5 69
40
=1
%
% .5 70
70 =1
=1
20
S
P
S
P
S
P
% 71 =1
0
US
Puerto Rico
S
P
S
0
Philippines
(25.6,25.6)
Puerto Rico
Mexico
Canada
UK
% .5
=1
US born men aged 25−44 by father's place of birth
5%
Dom. Rep. Portugal
67
P
S
20
40
Cuba
=1
US
(25.6,25.6)
USSR Germany P
Cuba
Dom. Rep.
Ireland Colombia Italy
S
5%
Canada
% 67 =1
Italy
Portugal
. 67 =1
Philippines
60
P
P
40
S
Greece
Poland Greece
S
Ireland
. 66
% 67 =1
Colombia
=1
P
S
USSR
P
5%
UK Germany
India
80
S
. 66 =1
Poland
60
% 66 =1
P
S
China
China
% 66 =1
80
P
P
S
S India
The Impact of Immigrant Dynasties on Wage Inequality
117
may still decline because when capital and skill are relative complements, capital dilution harms the wages of skilled workers more than it harms the wages of the unskilled.15 A group of immigrants with a low value of PS ðTÞ but a high value of PS ðV Þ raise the ratio between skilled and unskilled wages in the short run, but cause it to drop in the long run. The short-run effect of PS ðTÞ is temporary, whereas the long-run effect of PS ðV Þ in this model lasts forever. On the other the long run is discounted at a rate of return that varies (in response to immigration policy) around 5%. The discounting dominates and the angles measured between the iso-curves and the horizontal axis in Fig. 6 (and Figs. A3 and A4) – are above 451. Consider the effect of Greek immigrants on the values of SP in Fig 6. Among members of the second generation, the rate of college completion is 59.60% for women against 42.91% for men. Nonetheless, the very low levels of education among the women of the immigrant generation, 3.62%, means that the combined effect of the mother–daughter pair on the value of SP is positive – they raise inequality, if only slightly. The male immigrant from Greece is more likely to have a college degree (18.34%), though less likely than members of the population at large. Combined with a higher-thanaverage rate of college attainment in the second generation, the father–son pairs fall to the right of the SP ¼ 170 iso-curve. Hence males as a group lower inequality, and the combined effect of all Greek immigrants, male and female, on the value of the discounted college premium is close to neutral. Indeed, women (and their daughters) from Colombia, the Dominican Republic, Greece, and Portugal, together with men from the Dominican Republic, all raise the value of SP even though members of the second generation are better educated than the general population. Female immigrants from Cuba and Ireland, and male immigrants from Greece, Italy, Poland, and Portugal have lower rates of college completion than the general population, but the high rates of college attainment for corresponding members of the second generation mean their overall effect on SP is negative. The best-educated group in our sample are Indian men. For the immigrant generation 82.62% have college degrees, and for members of the second generation this rises to 87.26%. An additional 87,000 visas per year to members of this group over the course of 20 years profoundly lowers the value of SP, to 167.24%. The next group, Chinese men, lowers the value of SP to 168.03%. Among the people in the sample born in the United States, the least educated are those whose parents were born in Mexico or Puerto Rico. Only 6.68% of immigrant men from Mexico have college degrees, and for
118
MICHAEL BEN-GAD
American-born sons of Mexican-born fathers, the share with college degrees is only 10.23%. These values generate a value of 170.7% for SP. The corresponding figures for women, 3.98 and 12.21%, produce an SP of 170.8%. Overall, the values of SP for men rise in the following order: India, China, U.S.S.R., U.K., Germany, Philippines, Ireland, Canada, Colombia, Poland, Cuba, U.S., Italy, Greece, Portugal, Dominican Republic, Puerto Rico, and Mexico. For women the order is China, India, U.S.S.R., Philippines, Poland, Canada, Germany, Cuba, Ireland, U.S., Italy, U.K., Colombia, Greece, Dominican Republic, Puerto Rico, Portugal, and Mexico. If the production function is Cobb–Douglas or CES all the iso-curves shift outward, including the curve corresponding to 170%. Women from the U.K., and Italy now fall to the left of the curve; they now lower inequality in and economy without capital-skill complementarities (Figs. A3 and A4 in the appendix). More generally, the specification of the production function employed in Fig. 6 implies a high rate of substitution between the two types of labor. The lower the elasticity of substitution the closer together are the iso-curves and the more sensitive is the value of SP to different immigration surges. Consider, setting all the elasticities of substitution to 0.67 (Fig. A4 in the appendix). Raising the number of male immigrants by 75,000 per year for two decades lowers the value of SP to 165.02%. At the other end of the ordinal ranking, the arrival of the same number of Mexican women to the United States raises the value of SP to 171.50%. Changing the initial discount rate to 0.03 changes the shapes and not merely the placement of the iso-curves in Fig. 7. The angles measured between the iso-curves and the horizontal axis drops below 451, because the rate of educational attainment in the future designated by PS ðV Þ; is no longer as heavily discounted. This creates a new ranking for some of the immigrant groups. The values of SP for men rise starting with India, and moving through China, U.S.S.R., U.K., Germany, Ireland, Philippines, Poland, Colombia, Canada, Italy, Cuba, Greece, Portugal, U.S., Dominican Republic, Puerto Rico, and Mexico. For women the order is: China, India, U.S.S.R., Poland, Philippines, Germany, Ireland, Canada, Greece, Cuba, Italy, Colombia, U.K., U.S., Dominican Republic, Portugal, Puerto Rico, and Mexico. Comparing these two lists to those obtained under the higher discount rate for men, the order of the first five, India, China, U.S.S.R., U.K., and Germany, does not change; nor does the order of the last three, Dominican Republic, Puerto Rico, and Mexico. Ireland and Philippines switch places at positions six and seven – the rates of college attainment for secondgeneration members of the Irish community displace the higher rates of
The Impact of Immigrant Dynasties on Wage Inequality
119
college attainment of Philippine immigrants. Colombia remains in ninth place, but eighth-placed Canada switches with 10th-placed Poland. Of the remaining countries, Italy jumps two places to number 11, and Greece and Portugal each rise by one place, to 13 and 14 respectively. Cuba drops by one place to number 12, and the U.S. drops three places to 15. The shifts between the different countries are more pronounced for the women in the sample – only six, India, China, U.S.S.R., at the beginning, Italy and Dominican Republic in the middle, and Mexico at the end of the list maintain their positions. The lowering of the discount rate has a particularly large impact on the position of Greek women – they rise from 14th to 9th place. Furthermore, they no longer raise the value of SP even slightly; with the lower discount rate these women lower the value of SP. In summary, those immigrant groups with stable intergenerational rates of college attainment (those closest to the 451) have the same effect on the discounted college premium, regardless of which discount rate is employed in the calibration of the model. However, for the immigrant groups that experience large changes between the immigrant and second generations, the discount rate is an important factor in determining their overall impact. There are four important caveats. First, in this model there are only two skill types. Although educational attainment is by necessity a discrete variable in empirical studies, ordinarily there are more than just two categories. Second, in order to keep the model tractable, I do not treat educational attainment as a choice made by either the immigrants or the other individuals in the economy. Third, I abstract from the important distinction Chiswick (1978) found for the effect of additional schooling prior to immigration and after immigration for immigrant men’s earnings in the United States. Here, a college degree from China is equivalent to a college degree attained in the United States. Finally, I assume that once immigrant groups have achieved a level of college education at time V, this remains their permanent level from there on in. This means not only that a disproportionate number of immigrants from India and China arrive with college degrees, and that the rates of college graduation among members of the second generations are even higher, but also that subsequent generations do not ‘assimilate’ by lowering their educational performance.
10. CONCLUSION Countries throughout the developed world find themselves grappling with the question of immigration. The combination of declining birth rates and
120
MICHAEL BEN-GAD
increased life span is already raising the average age of citizens throughout the west, and increasing substantially the share of pensioners within the population. The ability of governments to maintain generous systems of oldage pensions is now in serious doubt. Some see more liberal immigration policies as a possible alternative to raising minimum retirement ages, slashing benefits, or raising the tax burden on the dwindling population of native younger workers (see Auerbach & Oreopoulos, 1999; Storesletten, 2000; Fehr, Jokisch, & Kotlikoff, 2004). This paper abstracts from the motivations for higher rates of immigration, instead focusing on its consequences for wage inequality, both in the short and long term. The traditional immigration-absorbing countries, the United States, Canada, and Australia, presently implement programs that grant residence to carefully selected foreigners with at least a Baccalaureate degree and a marketable skill. Increasingly other countries, including Germany and the United Kingdom, are also considering programs that target highly skilled engineers and scientists. At the same time, as traditional source countries, particularly in Asia, develop and opportunities there grow, perhaps fewer engineers and scientists will choose to permanently leave their homelands. Although there is still a vast pool of well-educated people willing to move to the West, competition between Western countries to attract them may intensify in the future. By contrast, the supply of unskilled immigrants – from the most impoverished and unstable countries in the world, and even from those portions of the world experiencing rapid growth – is enormous and may even grow during the next few decades. First, the cost of long-distance travel is likely to continue to decline. Second, the wages paid to unskilled workers in the poorest countries are not likely to move anywhere near the wages available in the developed world for the foreseeable future. Furthermore, the spread of mass communication in recent years has greatly raised awareness among inhabitants of poor countries about standards of living in the West. Finally, whereas once international migration entailed the near complete severing of ties to family and culture, the telecommunications revolution of the past two decades now allows migrants to retain links to their homeland by telephone, internet, and satellite television. Given these factors, Western countries need only relax their present interdiction efforts among illegal immigrants and they can receive nearly any influx of lowskilled workers they want. The only question is how many do they want? A surge in the number of unskilled immigrants will at least initially exacerbate wage inequality. In countries with a strong commitment to income equality, governments may find themselves spending relatively
121
The Impact of Immigrant Dynasties on Wage Inequality
Men aged 25−44 born in US or arrived since 1975 by place of birth
100
India
80 y=6.49+0.98x R2 =0.69 Canada
USSR China
60
UK
Germany
Ireland
Italy
Greece
Philipines
40
Colombia
Puerto Rico
20
US
Poland Cuba Portugal
Dom. Rep. Mexico
0 0
20 40 60 80 Men aged 45−64 born in US or arrived before 1975 by place of birth
100
Women aged 25−44 born in US or arrived since 1975 by place of birth
100
80
India
y=16.91+0.86x R2=0.30 60
USSR China
Ireland
Philipines
UK
Greece
Canada
Germany
40 Poland
Italy Colombia
US
Cuba
20
Puerto Rico Dom. Rep. Mexico Portugal
0 0
20 40 60 80 Women aged 45−64 born in US or arrived before 1975 by place of birth
100
Fig. 8. Percentage of the U.S. Population with Four-Year College Degrees by Age, Sex, and Birthplace. Data for the U.S.S.R. Includes All Respondents from Any of the Former Republics in the Sample, the Data for the U.K. Includes Respondents from Northern Ireland, and Data for Portugal Includes Respondents from the Azores. Note: Pooled Data for 2001, 2002, and 2003 from the U.S. Census, Current Population Survey. Source: King et al. (2003).
122
MICHAEL BEN-GAD
less on old-age pensions, but more on other types of transfer payments. Can assimilation or enhanced educational attainment by members of the second generation ameliorate these effects? Since few unskilled adult immigrants attend school after arriving in their new home, their absorption typically entails a rise in the share of unskilled workers in the labor force until the end of their working lives. Nonetheless, a rise in unskilled immigration today need not imply the creation of a self-perpetuating community of unskilled workers for generations to come and a permanent rise in the wage gap. As I have demonstrated, the disproportionate share of unskilled workers among certain immigrant groups that arrived in the United States a generation ago did not necessarily cause wage inequality, if we compare the ratio of discounted wages over time. The effect of high levels of educational attainment among the members of the second generation can easily overwhelm the low levels of education that often characterize the immigrant generation. What is the likely impact of immigration to the United States today? Consider the results in Fig. 8. The horizontal axis is the same as the horizontal axis in Fig. 2, but I replace the vertical axis with U.S. born and young immigrants that arrived after 1975. The large changes are mostly toward greater shares of college education. The most noticeable improvements are among Canadian and Italian men, women from Ireland, and both men and women from Greece. Overall the men are reasonably close to the 451. We cannot predict the rates of educational attainment among members of the second generation, 25 years hence, but perhaps the patterns in Fig. 2 offer some clue. Similarly, the methods developed here for calculating the discounted skill premium provide some guidance for determining how these immigrants and their families will affect wage inequality in the future.
NOTES 1. See Galor (1986), Djajic (1989), Borjas (1994), and Zak, Feng, and Kugler (2002) for models with endogenously determined levels of immigration. 2. Data for the rate of net international migration are available for calendar years up until 1999, and for years 2001 and beyond. Data for 2000 are available for only part of the year. Therefore, I compare the decades 1960–1969, 1970–1979, 1980–1989, and 1990–1999. 3. H-1B visas are granted for a maximum of two consecutive three-year stays. However, workers are no longer required to demonstrate an intention to
The Impact of Immigrant Dynasties on Wage Inequality
123
return to their home countries and most recipients are soon eligible to apply for permanent residency. In the past, at least half of those admitted under the program changed status and ultimately became permanent residents (see Lowell, 2001). 4. U.S. Department of Homeland Security, Office of Immigration Statisitics (2002). 5. U.S. Immigration and Naturalization Service, Office of Policy and Planning (1990–2000). 6. Considering recent studies on intergenerational mobility (see the survey by Solon, 1999) or the model of ethnic capital estimated by Borjas (1992), these effects are likely to be felt in the third generation and beyond. 7. However, we cannot reject the null hypothesis that its slope is equal to one. 8. Define t ¼ b as a date in the arbitrarily distant past bo0; when the economy was founded by an initial cohort of size M U ðbÞ þ M S ðbÞ: Then Ci(t), Ki(t), and Oi(t) are the consumption, capital, and the future earnings for the initial type i population at time b, and all the additional cohorts accumulated at rate mi(s) since b, all growing at the rate of n. Hence, Z t M i ðsÞmi ðsÞci ðs; tÞds C i ðtÞ ¼ enðt bÞ ¼ þ enðt
K i ðtÞ ¼ enðt Oi ðtÞ ¼ enðt
b bÞ
Z
M i ðbÞci ðb; tÞ
t
M i ðsÞmi ðsÞki ðs; tÞds þ enðt bÞ M i ðbÞki ðb; tÞ Z t bÞ M i ðsÞmi ðsÞds þ M i ðbÞ oi ðtÞ, bÞ
b
b
and Rs m ðvÞdv M i ðsÞ ¼ e b i 9. Also known as the two stage CES production function. The first stage combines skilled labor and raw capital to develop and maintain production capital: K n ¼ ðlK n þ ð1 lÞðH S Þn Þ1=n : K is used by unskilled labor 1=Win the second stage to manufacture final goods: Y ¼ mðH U ÞW þ ð1 mÞðK n ÞW : See Goldin and Katz (1998). 10. The general theory of perturbations was first developed by Euler, Laplace, and most importantly Lagrange in the late 18th century to study celestial mechanics. The movement of a planet around the sun was ‘perturbed’ from its elliptical orbit by the gravitational pull of other planets which varied over time (Ekeland, 1988). Judd (1982, 1985) introduced perturbations to economics to study fiscal policy where the perturbations are changes in tax rates. Here, changes in immigration policy perturbs the economy from its balanced growth path. 11. To guarantee convergence to an interior balanced growth path we also impose the restriction on pS ðtÞ and pU ðtÞ that they must satisfy Z T lim o1: ðpsðtÞ puðtÞÞdt T!1
0
124
MICHAEL BEN-GAD
R 112. ThevtLaplace transform of a function f(t) and a positive number v is Lv ½f ¼ dt: 0 f ðtÞ e 13. At the high end, graduate education declines slightly with the degree of nativity: 9.7% of the foreign born have graduate degrees, as do 8.9% of natives with foreign-born parents, but only 8.2% of natives with native-born parents. Grade school education rises more steeply with nativity – 22.2% of the foreign-born and 10.1% of the natives with foreign-born parents have less than nine grades of schooling (7.2% of the foreign-born have less than five), against only 4.5% with less than nine grades among the native-born population with native parents (See U.S. Department of Commerce, Bureau of Census, 2001). 14. The Allen Hicks partial elasticity of substitution between capital and unskilled labor, and between skilled and unskilled labor are also equal to 0.67, but the Allen Hicks partial elasticity of substitution between capital and skilled labor is 1=1 W þ 1=fSK 1=1 u 1=1 W ¼ 0:36
where fSK is the combined share of skilled labor and capital. 15. In Figs. A3 and A4 (see appendix), the elasticities of substitution between all the inputs are equal, and the iso-curve passes through the point (25.6,25.6), the value of PS(0). Again, a surge of immigration that merely replicates the existing skill distribution has no effect on the discounted discount premium, unless the elasticities of substitution between the inputs are not equal. For example, if production is Cobb–Douglas or CES the impact of Italian women on the value of SP is slightly negative, rather than slightly positive, as in Fig. 6.
ACKNOWLEDGMENTS I wish to thank Hillel Rapoport and two anonymous referees for their comments and suggestions.
REFERENCES Altonji, J. G., & Card, D. E. (1991). The effects of immigration on the labor market outcomes of less-skilled natives. In: M. A. John & R. B. Freeman (Eds), Immigration, trade, and the labor market. Chicago: University of Chicago Press. Auerbach, A. J., & Oreopoulos, P. (1999). Analyzing the fiscal impact of U.S. immigration. American Economic Review, 89, 176–180. Bauer, P., & Riphahn, R. T. (2004). Heterogeneity in the intergenerational transition of educational attainment: Evidence from Switzerland on natives and second generation immigrants. IZA Discussion paper, 1354. Ben-Gad, M. (2004). The economic effects of immigration – a dynamic analysis. Journal of Economic Dynamics and Control, 28, 1825–1845.
The Impact of Immigrant Dynasties on Wage Inequality
125
Borjas, G. J. (1985). Assimilation, changes in cohort quality, and the earnings of immigrants. Journal of labor Economics, 3, 463–489. Borjas, G. J. (1987). Self-selection and the earnings of immigrants. American Economic Review, 77, 531–553. Borjas, G. J. (1992). Ethnic capital and intergenerational mobility. Quarterly Journal of Economics, 107, 123–150. Borjas, G. J. (1994). The economics of immigration. Journal of Economic Literature, 32, 1667–1717. Borjas, G. J. (1995). The economic benefits of immigration. Journal of Economic Perspectives, 9, 3–22. Borjas, G. J., Freeman, R. B., & Katz, L. F. (1997). How much do immigration and trade affect labor outcomes? Brookings Papers on Economic Activity, 1997(1), 1–67. Borjas, G. J. (1999). The economic analysis of immigration. In: A. Orley, D. Card (Eds), Handbook of labor economics, (Vol. 3). Amsterdam: Elsevier. Borjas, G. J. (2003). The labor demand curve is downward sloping: Reexamining the impact of immigration on the labor market. Quarterly Journal of Economics, 118, 1335–1374. Card, D., DiNardo, J., & Estes, E. (2000). The more things change. In: G. J. Borjas (Ed.), Issues in the economics of immigration. Chicago: University of Chicago Press. Chiswick, B. R. (1978). The effects of Americanization on the earnings of foreign-born men. Journal of Political Economy, 86, 897–922. Chiswick, B. R. (1986). Is the new immigration less skilled than the old? Journal of Labor Economics, 4, 168–192. Djajic, S. (1989). Skills and the pattern of migration: The role of qualitative and quantitative restrictions on international labor mobility. International Economic Review, 30795–809. Duffy, J., Papageorgiou, C., & Perz-Sebastian, F. (2004). Capital-skill complementarity? Evidence from a panel of countries. The Review of Economics and Statistics, 86, 327–344. Ekeland, I. (1988). Mathematics and the unexpected. Chicago: University of Chicago Press. Fallon, P. R., & Layard, P. R. G. (1975). Capital-skill complementarity, income distribution, and output accounting. Journal of Political Economy, 83, 279–301. Federal Reserve, Board of Governors, (1998). Survey of consumer finances. Washington, DC. Fehr, H., Jokisch, S., & Kotlikoff, L. (2004). The role of immigration in dealing with the developed world’s demographic transition. NBER Working Paper no. 10512. Galor, O. (1986). Time Preference and International Labor Migration. Journal of Economic Theory, 38, 1–20. Gang, I., & Zimmermann, K. F. (2001). Is child like parent? Journal of Human Resources, 35, 550–569. Goldin, C., & Katz, L. F. (1998). The origins of technology-skill complementarity. Quarterly Journal of Economics, 113, 693–732. Griliches, Z. (1969). Capital-skill complementarity. The Review of Economics and Statistics, 51, 465–468. Jasso, G., Rosenzweig, M. R., & Smith, J. P. (2000). The changing skills of new immigrants to the United States. In: G. J. Borjas (Ed.), Issues in the economics of immigration. Chicago: University of Chicago Press. Johannsson, H., Weiler, S., & Shulman, S. (2003). Immigration and the labor force participation of low-skill native workers. Research in Labor Economics, 22, 291–308. Judd, K. L. (1982). Short-run analysis of fiscal policy in a simple perfect foresight model. Economic Letters, 10, 55–59.
126
MICHAEL BEN-GAD
Judd, K. L. (1985). An alternative to steady-state comparisons in perfect foresight models. Journal of Political Economy, 93, 298–319. Judd, K. L. (1998). Numerical methods in economics. Cambridge, MA: MIT Press. King, M., Ruggles, S., & Sobek, M. (2003). Integrated public use microdata series, current population survey: Preliminary version 0.1. Minneapolis: Minnesota Population Center, University of Minnesota. Krusell, P., Ohanian, L. E., Rı´ os-Rull, J.-V., & Violante, G. L. (2000). Capital-skill complementarity and inequality: A macroeconomic analysis. Econometrica, 68, 1029– 1053. Lindquist, M. J. (2003). Capital-skill complementarity and inequality in Swedish industry. Working Paper. Lowell, B. L. (2001). Skilled temporary and permanent immigrants in the United States. Population Research and Policy Review, 20, 33–58. Riphahn, R. T. (2003). Cohort effects in the educational attainment of second generation immigrants in Germany: An analysis of census data. Journal of Population Economics, 16, 711–737. Sato, R. (1967). A two-level constant-elasticity-of-substitution production function. The Review of Economic Studies, 34, 201–218. Solon, G. (1999). Intergenerational mobility in the labor market. In: A. Orley, D. Card (Eds), Handbook of labor economics (Vol. 3). Amsterdam: Elsevier. Smith, J. P., & Edmonston, B. (Eds) (1997). The new Americans: Economic demographic and fiscal effects of immigration. Washington: National Academy Press. Storesletten, K. (2000). Sustaining fiscal policy through immigration. Journal of Political Economy, 108, 300–323. U.S. Department of Commerce, Bureau of Census. (2001). Profile of the foreign-born population in the United States: 2000, December. U.S. Department of Commerce, Bureau of Census. (2004). State population estimates and demographic components of population change, December. U.S. Department of Commerce, Bureau of Economic Analysis. (2004). Survey of current business. U.S. Department of Homeland Security, Office of Immigration Statistics. (2002). Yearbook of immigration statistics. U.S. Immigration and Naturalization Service, Office of Policy and Planning. (1990–2000). Estimates of the unauthorized immigration population residing in the United States. Van Ours, J. C., & Veenman, J. (2003). The educational attainment of second-generation immigrants in the Netherlands. Journal of Population Economics, 16, 739–753. Weil, P. (1989). Overlapping generations of infinitely lived agents. Journal of Public Economics, 38, 183–198. Zak, P. J., Feng, Y., & Kugler, J. (2002). Immigration, fertility, and growth. Journal of Economic Dynamics and Control, 26(4), 547–576.
127
The Impact of Immigrant Dynasties on Wage Inequality
APPENDIX: COBB–DOUGLAS AND CES PRODUCTION The Cobb–Douglas and the CES production are explained in Figs. A1–A4 and also given in Tables A1–A4. Table A1. V
Cobb–Douglas Production Function: sUK ¼ sUS ¼ sSK ¼ 1.
Q¼0
Q ¼ 10
Q ¼ 20
Q ¼ 30
Q ¼ 40
Q ¼ 50
— — — 171.16 171.22
— — — — 171.27
— — — 171.16 171.22
— — — — 171.27
— — — 171.14 171.19
— — — — 171.25
Inelastic capital 20 30 40 50 60
170.00 170.24 170.43 170.57 170.69
170.32 170.52 170.68 170.80 170.89
— 170.73 170.86 170.96 171.04
— — 170.99 171.08 171.14
Completely elastic capital 20 30 40 50 60
170.00 170.24 170.43 170.57 170.69
170.31 170.52 170.67 170.80 170.89
— 170.72 170.86 170.96 171.04
— — 170.99 171.07 171.14
Endogenous capital 20 30 40 50 60
170.00 170.24 170.42 170.56 170.68
170.31 170.51 170.66 170.78 170.88
— 170.71 170.84 170.94 171.02
— — 170.97 171.05 171.12
Note: The values of the discounted skill premium SP, following a rise in the rate of immigration from 3.2 to 3.5 per thousand during the course of two decades, for different values of V and Q. The immigrants are initially unskilled PS ðTÞ ¼ 0; but beginning at time Q, some of these immigrants, or their descendants, become skilled workers. By time V the share of skilled workers within this population stabilizes at PS ðV Þ ¼ PS ð0Þ where PS ð0Þ ¼ 0:256 is the prevailing share of skilled workers within the general population. The baseline value of SP is 170%.
128
MICHAEL BEN-GAD
(a)
(b)
Fig. A1. Cobb–Douglas and CES Production Functions. Impulse Response for Unskilled Wages Following a 20 Year Surge in the Rate of Immigration from 3.2 to 3.5 Per Thousand, Q ¼ 25 and V ¼ 45: Note: The Solid, Dashed, and Dotted Curves Represent Respectively, the Impulse Responses Generated by the Model with Capital Supply that is Elastic, Completely Elastic, and Inelastic. a) Cobb Douglas Production Function, sUK ¼ sUS ¼ sSK ¼ 1; b) CES Production Function, sUK ¼ sUS ¼ sSK ¼ 0:67:
The Impact of Immigrant Dynasties on Wage Inequality
129
(a)
(b)
Fig. A2. Cobb–Douglas and CES Production Functions. Impulse Response for Skilled Wages Following a 20 Year Surge in the Rate of Immigration from 3.2 to 3.5 Per Thousand, Q ¼ 25 and V ¼ 45: Note: The Solid, Dashed, and Dotted Curves Represent Respectively, the Impulse Responses Generated by the Model with Capital Supply that is Elastic, Completely Elastic, and Inelastic. a) Cobb Douglas Production Function, sUK ¼ sUS ¼ sSK ¼ 1; b) CES Production Function, sUK ¼ sUS ¼ sSK ¼ 0:67:
130
MICHAEL BEN-GAD 100 66%
P =1
S
80
China
5% 66.
P =1
S
Poland
60
UK Germany
USSR
Colombia
Ireland
Greece
P =1
S
Philippines
40
Italy
Portugal
76%
US born men aged 25−44 by father's place of birth
India
Canada Cuba
Dom. Rep.
US
(25.6,25.6) PuertoR ico
5.%
68%
P =1
S
68
P =1
69%
5% 69.
P =1
S
P =1
70%
0
S
S P =1
5% 70.
1 7 %
P =1
0
S
P =1
S
S
Mexico
5% 67.
P =1
S
20
20 40 60 80 Men aged 45−64 born in US or arrived before 1975 by place of birth
100
100 P =1
S
India
80
5% 66.
P =1
S
Poland Greece
60
Ireland USSR
Cuba Canada
UK
Portugal
67%
Dom. Rep.
P =1
40
Germany
S
Colombia Italy
Philippines
US
(25.6,25.6)
20 Puerto Rico
68%
P =1
S
5% 68.
P =1
S 69%
P =1
S
5% 69.
P =1
S 70%
P =1
S
5% 70.
0
P =1
71%
P =1
S
0
S
Mexico
5% 67.
P =1
S
US born women aged 25−44 by mother's place of birth
66%
China
20 40 60 80 Women aged 45−64 born in US or arrived before 1975 by place of birth
100
Fig. A3. Cobb–Douglas Production Function: sUK ¼ sUS ¼ sSK ¼ 1: Iso-Curves for the Values of SP Following a 20 Year 25% Rise in the Rate of Immigration. Note: The Baseline Rate of Return of Capital is Set to 0.05, and Q ¼ 25 and V ¼ 45: Points Represent the Percentage of the U.S. Population with Four-Year College Degrees by Age, Sex, Birthplace, and Parent’s Birthplace from Fig. 2.
131
The Impact of Immigrant Dynasties on Wage Inequality
64% P =1
S
100
80
P =1
S
China
5 64. %
P =1
S
Poland
65%
60
UK Germany
USSR
Colombia
Ireland
P =1
S
Greece Philippines
Italy
%
Portugal
5 65.
40
Canada Cuba
Dom. Rep.
S
US
P =1
(25.6,25.6)
66%
US born men aged 25−44 by father's place of birth
India
20 Puerto Rico
67%
P =1
5% 67.
20 40 60 80 Men aged 45−64 born in US or arrived before 1975 by place of birth
100
64% P =1
S
100 China
S
5% 64.
P =1
India
80
Poland
65% P =1
S
Greece
60
Ireland USSR Colombia Italy Dom. Rep.
Germany Cuba Canada
UK
Portugal
5% 65.
40
P =1
S
Philippines
US
66% P =1
S
US born women aged 25−44 by father's place of birth
5% 66.
P =1
S S
68%
P =1
5% 68.
P =1
S S
P =1
69%
5% 69.
P =1
S
P =1
70%
5% 70.
P =1
S
P =1
71%
0
S
S
S P =1
S
5% 71.
P =1
S
0
Mexico
(25.6,25.6)
20
Puerto Rico
80
5% 66.
67%
P =1
S
P =1
S
5% 67. P 1 = 68% P =1
S
5% 68.
60
S
69%
P =1
P =1
S
5% 69.
40
S
70%
P =1
S
P =1
S
71%
P =1
20
5% 70.
P =1
S
0
S
5% 71.
P =1
S
0
Mexico
100
Women aged 45−64 born in US or arrived before 1975 by place of birth
Fig. A4. CES Production Function: sUK ¼ sUS ¼ sSK ¼ 0:67: Iso-Curves for the Values of SP Following a 20 Year 25% Rise in the Rate of Immigration. Note: The Baseline Rate of Return of Capital is Set to 0.05, and Q ¼ 25 and V ¼ 45: Points Represent the Percentage of the U.S. Population with Four-Year College Degrees by Age, Sex, Birthplace, and Parent’s Birthplace from Fig. 2.
132
MICHAEL BEN-GAD
Table A2. V
Q¼0
CES Production Function: sUK ¼ sUS ¼ sSK ¼ 0.67. Q ¼ 10
Q ¼ 20
Q ¼ 30
Q ¼ 40
Q ¼ 50
— — — 171.74 171.82
— — — — 171.90
— — — 171.73 171.82
— — — — 171.90
— — — 171.70 171.78
— — — — 171.86
Inelastic capital 20 30 40 50 60
170.00 170.36 170.64 170.86 171.03
170.47 170.78 171.01 171.19 171.33
— 171.09 171.28 171.43 171.55
— — 171.48 171.61 171.70
Completely elastic capital 20 30 40 50 60
170.00 170.36 170.64 170.86 171.03
170.47 170.77 171.01 171.19 171.33
— 171.08 171.28 171.43 171.55
— — 171.48 171.61 171.70
Endogenous capital 20 30 40 50 60
170.00 170.35 170.63 170.84 171.01
170.46 170.76 170.99 171.17 171.31
— 171.06 171.25 171.40 171.52
— — 171.45 171.57 171.67
Note: The values of the discounted skill premium SP, following a rise in the rate of immigration from 3.2 to 3.5 per thousand during the course of two decades, for different values of V and Q. The immigrants are initially unskilled PS ðTÞ ¼ 0; but beginning at time Q, some of these immigrants, or their descendants, become skilled workers. By time V the share of skilled workers within this population stabilizes at PS ðV Þ ¼ PS ð0Þ where PS ð0Þ ¼ 0:256 is the prevailing share of skilled workers within the general population. The baseline value of SP is 170%.
Table A3. V
Cobb–Douglas Production Function: sUK ¼ sUS ¼ sSK ¼ 1.
Q¼0
Q ¼ 10
Q ¼ 20
Q ¼ 30
Q ¼ 40
Q ¼ 50
— —
— —
Inelastic capital 20 30
165.88 166.81
167.11 167.89
— 168.68
— —
133
The Impact of Immigrant Dynasties on Wage Inequality
Table A3.
(Continued )
V
Q¼0
Q ¼ 10
Q ¼ 20
Q ¼ 30
Q ¼ 40
Q ¼ 50
40 50 60
167.52 168.08 168.52
168.49 168.95 169.31
169.19 169.57 169.87
169.70 170.03 170.28
— 170.36 170.57
— — 170.78
— — — 170.33 170.54
— — — — 170.76
— — — 170.23 170.44
— — — — 170.66
Completely elastic capital 20 30 40 50 60
165.86 166.77 167.49 168.05 168.50
167.06 167.85 168.45 168.91 169.28
— 168.64 169.15 169.54 169.84
— — 169.66 169.99 170.25
Endogenous capital 20 30 40 50 60
165.75 166.68 167.40 167.97 168.42
166.97 167.76 168.36 168.83 169.19
— 168.54 169.05 169.45 169.75
— — 169.57 169.90 170.15
Note: The values of the discounted skill premium SP, following a rise in the rate of immigration from 3.2 to 3.5 per thousand during the course of two decades, for different values of V and Q. The immigrants are initially unskilled PS ðTÞ ¼ 0; but beginning at time Q, all these immigrants, or their descendants, become skilled workers. By time V the share of skilled workers within this population stabilizes at PS ðVÞ ¼ 1: The baseline value of SP is 170%.
Table A4. V
Q¼0
CES Production Function: sUK ¼ sUS ¼ sSK ¼ 0.67. Q ¼ 10
Q ¼ 20
Q ¼ 30
Q ¼ 40
Q ¼ 50
— — — 170.56 170.87
— — — — 171.19
Inelastic capital 20 30 40 50 60
163.90 165.27 166.33 167.16 167.82
165.73 166.89 167.77 168.46 168.99
— 168.08 168.82 169.39 169.83
— — 169.59 170.07 170.43
134
MICHAEL BEN-GAD
Table A4. V
Q¼0
Q ¼ 10
(Continued )
Q ¼ 20
Q ¼ 30
Q ¼ 40
Q ¼ 50
— — — 170.49 170.81
— — — — 171.14
— — — 170.34 170.66
— — — — 170.98
Completely elastic capital 20 30 40 50 60
163.84 165.20 166.26 167.09 167.76
165.63 166.79 167.68 168.38 168.92
— 167.97 168.73 168.31 169.76
— — 169.50 169.99 170.37
Endogenous capital 20 30 40 50 60
163.66 165.05 166.13 166.97 167.64
165.48 166.65 167.55 168.25 168.79
— 167.83 168.59 169.17 169.62
— — 169.35 169.85 170.22
Note: The values of the discounted skill premium SP, following a rise in the rate of immigration from 3.2 to 3.5 per thousand during the course of two decades, for different values of V and Q. The immigrants are initially unskilled PS ðTÞ ¼ 0; but beginning at time Q, all these immigrants, or their descendants, become skilled workers. By time V the share of skilled workers within this population stabilizes at PS ðVÞ ¼ 1: The baseline value of SP is 170%.
DOES IMMIGRATION AFFECT LABOR DEMAND? MODEL AND TEST O¨rn B. Bodvarsson and Hendrik Van den Berg ABSTRACT Numerous studies have concluded that immigration has very small effects on wages or unemployment, even when the immigration flow is very large. Three reasons suggested for this are that immigration: (1) is not supplypush, but may instead be driven by demand-pull factors; (2) is likely to cause some out-migration; and (3) may induce flows of other factors across the economy. Surprisingly, few studies consider another obvious explanation: immigrant workers also consume locally, which means immigration stimulates the local demand for labor. Previous researchers have generally ignored the measurement of immigration’s effects on labor demand, perhaps because when immigration, out-migration, and immigrant consumption occur simultaneously in the same labor market, it is very difficult to isolate immigration’s effect on labor demand. This paper measures the labor demand-augmenting effects of immigration using a two-sector model of a very special case in which the receiving economy consists of: (a) an export industry employing both immigrants and natives; and (b) a retail industry employing native labor that is driven by local demand. The model can incorporate both supply-push and demand-pull immigration as well as out-migration. The model’s important Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 135–166 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24004-9
135
136
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
implication is that since immigration is exogenous to the retail sector, an unbiased estimate of the demand effect of immigration can be obtained without having to use instrumental variables estimation or other statistical procedures that may introduce new sources of bias. Fortunately, the economy in our model matches a very convenient test case: Dawson County, Nebraska. Dawson County recently experienced a surge in demand-pull immigration due to the location of a large export-driven meatpacking plant. This exogenous capital shock pulled in many Hispanic immigrant workers, who did not immediately seek work in the retail sector because of social and language barriers. This immigration led to higher retail wages and housing prices, confirming that immigration is capable of exerting significant effects on local labor demand.
Jacques Chirac, Prime Minister of France: If there were fewer immigrants, there would be less unemployment, fewer tensions in certain towns and neighborhoods, and lower social cost. Liberation [A Paris newspaper]: That has never been formally proven. Chirac: It is easy to imagine, nevertheless (From an October 30, 1984 interview).1
1. INTRODUCTION The conventional wisdom about immigration, reflected in Jacques Chirac’s words above, is that an inflow of foreigners constitutes an increase in the supply of labor that adversely affects native workers. However, the weight of the empirical evidence suggests that immigration actually has little effect on native wages and employment. Studies in this area can be divided into two categories: (1) test cases involving extraordinary levels of international exogenous (supply-push) immigration; and (2) studies that examine differences in immigration’s impact across local labor markets due to differences in geographic clustering of immigrants. The evidence from the first category is best summarized in Friedberg and Hunt’s (1995) widely cited survey, which concludes that supply-push immigration actually has very little effect on wages or unemployment, even when immigration flows are very large.2 The typical study falling into the second category defines the labor market as a metropolitan area and examines the relationship between native employment outcomes in the metropolitan area and the immigrant share of the
Does Immigration Affect Labor Demand? Model and Test
137
labor force in that locality.3 These studies generally find that spatial correlations between local native wages and the locality’s share of immigrants are usually negative, but very weak. For example, in a widely cited study, Card (2001) used 1990 census data for nearly 200 U.S. cities and found that exogenous immigrant inflows during the 1980s had very modest adverse effects on wages of low-skilled natives. What could explain immigration’s relatively mild effects on the destination labor market? First, statistical estimates of the effects of exogenous immigration on local wages may be biased if the observed immigration is endogenous. For example, immigrants may endogenously cluster in areas with thriving economies, i.e. immigration could be of the demand-pull variety. Failure to account for such endogeneity causes simultaneity bias in OLS estimates of the effects of exogenous immigration. Second, immigrant inflows are likely to trigger offsetting out-migration, which can mitigate supply-push immigration’s depressing effect on wages or unemployment. Most researchers have recognized these two problems by controlling for or instrumenting out-migration and by using instrumental variables to control for demand-pull immigration (see, for example, Altonji & Card, 1991; Card, 2001; and Pedace, 1998). Third, as Borjas, Freeman, and Katz (1996) and Borjas (2003) point out, immigration could induce flows of other factors of production across the economy. For example, natives may respond to the local wage impact of immigration by moving their factor services to other localities, generating re-equilibration across the national labor market and perhaps explaining why inter-city differences in native employment outcomes from immigration are very small. Borjas (2003) got around this problem by redefining the labor market as national, sorting workers into particular skill groups based on educational attainment and work experience, and exploiting substantial differences in immigrant shares across these skill groups. He did find, in stark contrast to previous studies, that immigration has significant adverse effects on natives; a 10% increase in immigrant labor supply reduces native wages by 3–4%. We offer another explanation for the observed small labor market effect of immigration: immigration increases not only the supply of labor, but also the demand for labor. Immigrant workers are also, after all, consumers. It is quite surprising that immigration’s potential labor demand effects have seldom been examined in the economics of immigration literature. In fact, there is only a smattering of literature since the 1960s that discusses the possibility that immigration may stimulate local demand. Taking a strictly macroeconomic approach, Mishan and Needleman (1966, 1968) theorized on immigration’s effects on excess aggregate demand
138
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
and the balance of payments. In a little-known study on immigration to Australia, Harrison (1983) argued that immigrant inflows affect both the supply of labor and the local demand for productive resources such as labor. Altonji and Card (1991) present a model of the local wage and employment effects of immigration that explicitly includes the demand effect from immigrants’ consumption of at least a fraction of the output they produce. Altonji and Card’s theoretical model yielded the prediction that when immigrants consume a larger proportion of the output they produce, local wages are less sensitive to exogenous immigrant inflows. However, they did not carry the demand effects of immigration from their theoretical model over to their econometric model with which they tested for immigration’s effect on wages. Using a dynamic model to link immigration with the domestic goods market, Hercowitz and Yashiv (2001) examined the effects of mass immigration from the former Soviet Union to Israel. They found that when immigration is allowed to raise the demand for goods and lower the relative price of imports, this delays any negative employment effect on natives by about a year. However, as with Altonji and Card (1991); Yashiv and Hercowitz (2001) did not theoretically or empirically isolate the effects of immigration on Israeli labor demand. Rhode (2003) finds that the rapid growth of real wages in California after World War II can be explained by the growth of the local retail and housing markets caused by the large inflows of migrants. As with the previously cited studies, however, Rhode’s study stopped short of obtaining estimates of the ceteris paribus effects of immigrant inflows on the local demand for labor. Regardless of the paucity of explicit analysis of immigration’s labor demand effect, there is no doubt that such a demand effect is present. The Inter-American Development Bank (2004) estimates that even immigrants from Latin America, the largest immigrant group in the U.S. and which remits the greatest share of U.S. income back to relatives in Latin America, spent 93 % of their gross 2003 income of $450 billion in their local communities.4 Isolating immigration’s labor demand effects on wages or employment is difficult because they are endogenous, e.g. local consumption depends on the levels of in- and out-migration, which in turn depend, in part, on wages. Consequently, an empirical model that tests the pure labor demand effect of immigration must be able to separate out immigration’s demand effects from the reverse effect of wages on immigration, the wage and employment effects on out-migration, and the effect of out-migration on wages. The analysis also depends critically on the availability of data on local immigration and labor market conditions.
Does Immigration Affect Labor Demand? Model and Test
139
This paper follows a novel strategy to accurately measure immigration’s effect on labor demand. First, we develop and present a general equilibrium labor market model in which the labor demand effect of immigration can be conveniently isolated without having to use instrumental variables. In this model, immigration: (1) is driven by both supply-push and demand-pull sources; (2) induces migratory responses by natives and earlier immigrants; and (3) increases local demand. Specifically, we model a two-sector economy in which immigrants work in an export sector, but spend their wages in a local retail sector. By separating the markets in which immigrants work and consume, the model clearly distinguishes the effects of greater immigrant labor supply from the labor demand effect of immigrant consumption on local retail wages. The second part of our strategy is to find a test case that fits the assumptions of the model and also provides sufficient data for estimating the model. While this model describes a very special case, it accurately describes immigration to Dawson County, Nebraska, which received large inflows of Hispanic workers attracted by a large, new meatpacking industry in the 1990s. Dawson County also provides an especially attractive readily available data set that we can use with the model to generate an unbiased estimate of the local labor demand effect of immigration. Our analysis uncovers a large labor demand effect, which suggests that models that ignore immigration’s effects on labor demand may not accurately describe immigration’s full effect on the labor market.
2. A NOT-SO-GENERAL GENERAL EQUILIBRIUM MODEL OF IMMIGRATION It is not easy to distinguish immigration’s effect on the demand for labor from the many other effects that immigration has on the destination economy’s labor market. The measurement of immigration’s demand-augmenting effect is complicated by the simultaneous presence of supply-push and demand-pull immigration and endogenous out-migration. In this section, we develop a model of a competitive local economy that has the distinct feature of complete separation between the labor markets in which immigrants work and in which they spend. It is obviously a short-run model because it assumes that there is no inter-sector mobility of labor. It is clearly designed for its econometric convenience, not its general applicability. The model nevertheless does accurately fit some recent immigration experiences in
140
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
small rural United States cities. The model can therefore be applied to these cases to test for immigration’s labor demand effect. Suppose a competitive local economy produces two different products: (i) a retail good that is consumed locally; and (ii) an export good that is produced solely for export and whose demand is exogenous. Suppose, also, that the labor force in the retail goods sector is exclusively native, but exports are produced by both immigrant and native workers. The local economy thus consists of three markets – the retail good market, the market for retail labor, and the market for labor in the export industry. All export sector workers, as well as those working in the retail sector, spend their wage earnings on the retail good, whose market price is PR. We first discuss each market separately, and then we highlight the relationships between them.
2.1. The Market for Labor in the Export Industry In the export labor market, demand (LDX) depends on a real wage: pX LDX ¼ b WX
(1)
where PX is an exogenous world price for the export good and WX a nominal wage paid to immigrant and native workers. The constant b reflects other exogenous determinants of labor demand, e.g. the capital stock and technology. Immigrants and natives are assumed to be perfect substitutes in the production of export goods. Export industry labor supply, LSX, equals the sum of native workers willing to work in the export industry, NX, and the total supply of immigrant workers, I: LSX ¼ N X þ I
(2)
The decision to immigrate to the local economy is assumed to depend on the differences in economic conditions between the local and source economies. Specifically, thenumber of immigrants depends on the real wage paid to export workers W X =PR and the real reservation wage of immigrants, VI. The immigrant labor supply function is i WX I¼ (3) VI PR
Does Immigration Affect Labor Demand? Model and Test
141
Note that the reservation wage VI could be the wage in the source country (for immigrants who come directly from abroad) or the wage available in another region of the host country (for those who came from abroad earlier and have been working elsewhere in the host country). Note also that we adjust the nominal export wage by the price of the retail product, PR, as it is assumed that immigrant workers spend their earnings in the retail sector. The coefficient i reflects the presumption that immigration depends on more than just income differences. For example, some people may immigrate because they are following family members that immigrated there in earlier periods, because they prefer to reside in areas with higher concentrations of immigrants, or because they are seeking a quality of life higher in the source economy.5 Similarly, the supply of native labor depends on the real wage paid in the export sector and a real reservation wage VN reflecting labor market opportunities elsewhere: n WX (4) NX ¼ VN PR The reservation wage VN may be taken as a wage available in another region of the host country, for example. The constant n reflects other (fixed) determinants of native labor supply. Eq. (3) allows for either supply-push or demand-pull immigration. Supply-push immigration could result from a lower immigrant reservation wage (VI) caused by weakened economic conditions in the source country. If the reservation wage fell, the immigrant labor supply curve would rotate to the right. In contrast, demand-pull immigration involves a movement up the labor supply curve. For example, an increase in export industry labor demand due to a higher product price (PX) or an exogenous increase in capital (a higher value of b) would push up the export wage and raise the quantity of immigrant labor supplied. Note, however, that supply-push immigration generates a secondary effect in the form of a movement along the immigrant labor supply curve because the fall in export wages causes outmigration of native workers. The equilibrium export sector wage, WX, is found by substituting Eqs. (3) and (4) into Eq. (2), setting Eq. (1) equal to Eq. (2), and solving for the wage: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u bP P u X R W X ¼ t (5) i n þ VI VN
142
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
According to Eq (5), the equilibrium export wage depends on the export price, the local cost of living, and the real reservation wages of natives and immigrants. The equilibrium level of immigrant employment, I, is obtained by substituting Eq. (5) into Eq. (3): ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi v u bP i u X n (6) I ¼ pffiffiffiffiffiffiffi t i PR V I þ n VI
VN
The equilibrium level of native employment in the export sector, NX, is obtained by substituting Eq. (5) into Eq. (4): ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi v u bP n u X N X ¼ pffiffiffiffiffiffiffi (7) t i PR V N þ n VI
VN
The equilibrium stocks of both immigrant and native labor depend upon the same variables, namely export price, the local cost of living, and real reservation wages. Furthermore, demand-pull immigration and supply-push immigration influence the equilibrium employment of both native and immigrant labor in the export sector. raises Demand-pull immigration always the employment of immigrants @I n =@PX 40 and natives @N nX @PX 40 : On the other hand, supply-push immigration always reduces native em ployment @N nX @V I 40 ; but raises immigrant employment @I n =@V I 40 : 2.2. The Retail Labor Market In the market for labor to produce the locally consumed retail product, only native workers are employed, and they are assumed to face the same reservation wage as their native counterparts in the export sector. This assumption clearly makes this a short-run model, since in the long run there would be movement of immigrant labor from the export sector to the local retail sector. Retail labor demand, LDR, depends on the real retail wage, whereas labor supply, LSR, depends on the real retail wage and a real reservation wage, PR LDR ¼ l (8) WR
Does Immigration Affect Labor Demand? Model and Test
LSR ¼
f VN
WR PR
143
(9)
where WR is the nominal retail wage and VN the same reservation wage facing native workers in the export sector. The equilibrium nominal retail wage is sffiffiffiffiffiffiffiffiffiffi lV N (10) W R ¼ PR f 2.3. The Market for the Retail Product In the market for the retail product, we assume a simple demand function in which consumer demand QDR depends only upon the product’s price and aggregate consumer income Y, and the product is normal:6 Y QDR ¼ c (11) PR Retail product supply is assumed to depend upon price and the retail wage paid: PR (12) QSR ¼ p WR Retail consumers in this economy include employed natives in the retail sector, employed natives in the export sector, and employed immigrants in the export sector. Consumers’ incomes consist of wages and any retail profits distributed to them. Natives spend all their incomes locally, but immigrants are assumed to remit a fraction of their incomes elsewhere.7 We specify that immigrants spend the fraction k ðko1Þ of their incomes on local retail goods and remit the remainder elsewhere. Hence, total retail consumer income spent locally is Y ¼ W nR N nR þ W nX N nX þ k W nX I n þ y (13)
where y is distributed retail profits. Substituting the above expression for Y into Eq. (11) implies that retail product demand is: c W nR N nR þ W nX N nX þ k I n W nX þ y (14) QDR ¼ PR
144
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
Eq. (14) illustrates the linkage between the retail and export sectors of the economy. Retail product demand depends on the incomes of both export and retail workers and employment in both sectors. There is a direct link between immigration and retail demand; an increase (decrease) in immigrant employment bolsters (weakens) retail demand. The equilibrium retail price, P R, is found by substituting the equilibrium nominal retail wage from Eq. (10) into the demand Eq. (14), setting Eq. (14) equal to the supply Eq. (12), and solving: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi c W X N X þ kW X I þ y lV N =f n (15) PR ¼ p 1 cl=p The first two terms within the parentheses of the numerator in Eq. (15) reflect the influence of the export sector on retail price. When export sector wages rise (fall), all other things equal, retail prices rise (fall). Likewise, an increase (decrease) in employment in the export sector, all other things equal, will push up (down) retail prices. The retail wage is found by substituting Eq. (15) into Eq. (10): c W X N X þ kW X I þ y lV N =f n WR ¼ (16) p 1 cl=p The equilibrium retail wage depends on the equilibrium export wage, the equilibrium levels of immigrant and native employment in the export sector and the equilibrium level of retail employment. These in turn depend upon the export product’s price, the reservation wages facing immigrant and native workers, immigrants’ propensities to spend on local retail products, retail producer profits, and a remaining set of parameters that reflect various supply and demand elasticities in the three different markets. Eqs. (15) and (16) illustrate precisely how immigrant inflows influence retail labor demand; immigration affects retail prices, which in turn influence retail wages. Immigration affects retail prices and wages by: (a) expanding the pool of those retail consumers who work in the export sector; (b) directly affecting the export wage and, hence, each export worker’s spending power in the retail market; and (c) indirectly influencing the quantity supplied of retail labor and, hence, the number of retail consumers who work in the retail sector. The exact relationship between immigration and retail wages depends on the type of immigration. For example, in the case of demand-pull
Does Immigration Affect Labor Demand? Model and Test
145
immigration due to an increase in export prices the marginal effect of export price on the retail wage is unambiguously positive: h n n i @W X lV N n n @N X n @I n n þ W þ kI N þ k c X X @PX @PX @PX f @W R ¼ 40 (17) cl @PX p 1 p
The first term in the bracketed expression in the numerator is positive because, when export prices rise, export wages rise. The second term is also positive because when the export price rises, there will be more retail consumers and each consumer has more income to spend. Therefore, demandpull immigration always results in higher retail prices. A similar prediction is obtained when demand-pull immigration results from other factors, i.e. when the parameter b in the export labor demand function (Eq. (1)) rises: h n n i @W X lV N n n @N X n @I n f @W nR c @b N X þ kI þ W X @b þ k @b ¼ 40 (18) @b p 1 cl p This prediction implies that an exogenous capital shock such as, for example, the establishment of a new export manufacturing facility, will increase wages in both the export and retail sectors of the local economy. Supply-push immigration may not always stimulate the retail sector. Suppose that there is a decrease in the immigrant reservation wage. The marginal effect of the lower reservation wage on the retail wage is measured by: h n n i @W X lV N n @N X n n @I n f @W nR c @V I N X þ kI þ W X @V I þ k @V I (19) n ¼ @V I p 1 cl p The sign of this derivative is ambiguous because there are conflicting effects on the retail wage of a lower immigrant reservation wage. On the one hand, a lower immigrant reservation wage, by raising immigrant employment and the number of retail consumers, stimulates retail demand. On the other hand, because the lower reservation wage puts downward pressure on the export sector wage, retail demand softens because each export worker’s spending power in the retail market falls. The net effect of supply-push immigration on the retail wage thus depends on which of these effects dominates. If the export wage elasticity of demand is relatively low, for example, then the retail price and wage are very likely to fall. The reason is that, in this case, supplypush immigration induces a relatively large decline in the export sector wage,
146
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
generating a large negative effect on retail spending, while boosting export sector employment and the retail consumer pool by a relatively small amount. In contrast, the retail price and wage are likely to rise if the export wage elasticity of demand is relatively high. Note that precisely the same types of ambiguous predictions result from other exogenous causes of supply-push immigration that change the parameter i in the immigrant labor supply function. The model also generates the intuitive result that a decline in the fraction of immigrant wages remitted to the home country raises the retail wage, all other things equal: n n lV N n I cW X f @W R o0 ¼ (20) cl @k p 1 p
Furthermore, the greater the effect of demand-pull immigration on the retail wage the smaller is the fraction of wage income remitted abroad by immigrants, i.e. @2 W nR ð@2 PnX @kÞ and @2 W nR ð@bn0 @kÞ are both positive. In summary, our special case model shows that immigration is indeed capable of exerting a positive effect on retail labor demand. The common belief that immigration always adversely affects the destination economy’s labor market is not generally correct, therefore. However, labor demand is not always bolstered by immigration, and the precise effects of an immigrant inflow on local wages depend on the source of the inflow and on local demand and supply conditions. Our model also predicts that demand-pull immigration always raises retail wages, a result that also clashes with popular perceptions of the effects of immigration on the labor market. Furthermore, our model predicts that the effects of supply-push immigration on retail wages are ambiguous and depend specifically on how much income immigrants remit abroad and on how sensitive the export wage is to immigration. Therefore, the exact effect of immigration on a local economy is ultimately an empirical issue. Our model is very different from the single-sector labor supply-only models used by previous immigration researchers. The unique feature of this model is that it allows for an unbiased estimation of the ceteris paribus effect of immigration on labor demand using standard statistical methods. Because the explanatory variables comprising Eq. (16) are all exogenous to the retail product and labor markets, there is no need to find instrumental variables in order to obtain an unbiased estimate of the local demand effect of immigration.
Does Immigration Affect Labor Demand? Model and Test
147
3. A TEST CASE OF THE DEMAND EFFECTS OF IMMIGRATION In order to test the implications of the model and to gain some measure of the size of the local labor demand effect of immigration, we sought a test case where there is relatively a clean separation between the markets where immigrants work and where they spend. We could then conveniently bypass the simultaneity problems that are likely to occur in cases where immigrants work and spend in the same market, or where there is demand-pull immigration to the retail sector. Second, like the test cases used by previous researchers, we sought a natural experiment characterized by unusually large immigrant inflows so that the effect of immigration on the local economy would not be overwhelmed by other unrelated economic changes occurring concurrently with immigration. Immigrants to the United States are increasingly moving to smaller cities where manufacturing jobs are available, most often in the Midwest and the South. These immigrant inflows are sometimes very large relative to the populations of the destination cities. We focus here on Dawson County, Nebraska, located in the South-Central part of Nebraska about 225 miles West of Omaha along Interstate 80 and the mainline of the transcontinental Union Pacific Railroad. Dawson County experienced an extraordinary inflow of Hispanic immigrants during the 1990s, and nearly all of these immigrants took jobs at a new and very large meatpacking plant.8 As the county’s meatpacking industry was essentially an export industry serving the U.S. and overseas markets, the Hispanic immigration was pure demand-pull immigration triggered by an exogenous capital shock. The size of Dawson County’s immigrant inflow was exceptional; Hispanic immigrants made up about one-quarter of the county’s population in 2000 (nearly 6,000 Hispanic residents), compared to just 3% in 1990. During the 1980s, many rural counties on the Great Plains experienced substantial out-migration due to a severe farm and debt crisis. According to Gouveia and Stull (1997), through the 1980s Nebraska experienced a net out-migration of over 100,000 people, and 40 out of 52 rural counties in the state experienced double-digit rates of net out-migration. Gouveia and Stull report that out-migration in rural Nebraska was highest among persons aged 25–44, with some counties losing 40–60% of younger workers during the 1980s. Dawson County was hit especially hard by the 1980s farm and debt crisis. According to U.S. Census data, Dawson County lost 11% of its population and experienced a record number of farm bankruptcies and other business failures during the 1980s. On top of the farm crisis, a large
148
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
combine factory owned by Sperry New Holland in Lexington, the Dawson County seat, closed. The economic decline of the 1980s in Dawson County was halted at the end of the decade when Iowa Beef Processors (IBP), a subsidiary of Tyson Foods, the world’s largest meat-processing firm, announced that it would open a beef-packing plant on the site of the old combine factory. U.S. meatpacking was undergoing a major restructuring, moving plants away from urban areas with high concentrations of unionized labor to smaller cities located closer to cattle and feed grain supplies. In addition, counties like Dawson wooed companies such as IBP with generous tax abatements and other state and local subsidies. According to the 1997 U.S. Census of Manufacturing, during the 1990s there were 132 counties in the U.S. with meatpacking industries employing at least 1,000 people. Seventy of these counties were rural (defined as having in 2000 a total population under 50,000) and, like Dawson County, many experienced Hispanic immigration to their meatpacking industries. IBP began hiring workers for the Lexington plant in 1990 and commenced slaughtering cattle in November the same year. As detailed in Gouveia and Stull (1997), during the first 21 months of operation over 5,000 workers were hired, among whom many of them were natives of Mexico, Guatemala, and El Salvador. Nearly all of these workers were new to Dawson County, having migrated from other areas on the Great Plains or were directly from their native countries. As is typical in the meatpacking industry, there was substantial labor turnover during this period. By February 1992, the plant’s labor force had stabilized at slightly over 2,000 workers. In 1997, the plant employed approximately 2,300 workers, of which roughly 75% were reported to be Hispanics. Since 1997, the total size and the Hispanic proportion of the plant’s labor force have remained at that level. Lexington’s large, new packing plant led to sharp changes in Dawson County’s population and ethnic composition. During the 1990s, Dawson County experienced a 22% increase in population, the largest rate of increase of any Nebraska county (based on decennial census data). By 2000, nearly 75% of the county’s population of 24,365 was concentrated in three towns – Lexington (10,011 residents), Cozad (4,163), and Gothenburg (3,619). Lexington experienced the largest population increase of the three during the 1990s (52%), followed by Gothenburg (12%) and Cozad (9%). Dawson County experienced positive net migration of 2,934 persons during the 1990s, that was equal to 14.7 % of the county’s 1990 population, when compared to negative net migration of 3,400 during the 1980s, or 15.2% of
Does Immigration Affect Labor Demand? Model and Test
149
1980 population. The extraordinary level of net migration during the 1990s was largely the result of the large influx of Hispanics hired to work at IBP. According to census data, the county’s Hispanic population share rose from 3.28% in 1990 to 25.36% in 2000. Currently, about one of every two Lexington residents is Hispanic. Table 1 ranks the top 20 counties in the U.S. with meatpacking industries employing at least 1,000 people on the basis of percentage growth in the Hispanic share of the population during the 1990s. Note that all these counties are rural, and nearly half are located in Nebraska, Kansas, and Iowa. Dawson County’s change in Hispanic population share of 22.08 percentage points between 1990 and 2000 ranked just behind Seward County, Kansas (22.57 percentage points) and Colfax County, Nebraska (23.75). Like Dawson County, the large inflows of Hispanic immigrants during the 1990s to Colfax and Seward Counties were primarily related to growth in their meatpacking industries. We chose Dawson County as our test case for three important reasons. First, because the initial immigrant inflows to the county coincided almost precisely with the 1990 decennial census – it was very easy to track the performance of the county’s economy against the timing of the inflow. Second, because some of the county data needed to test our model only became available in 1980, the 1990 surge in immigration enabled us to construct a two-decade sample in which the immigrant inflow began at the midpoint of the sample. This gives us an equal number of observations before and after the surge in immigrant labor. Third, Dawson County was a clear-cut example of demand-pull immigration that was triggered by an exogenous capital shock. Because demand-pull immigration raises both wages and employment, its positive labor demand effects will be easier to detect than for the case of supply-push immigration. Table 2 shows that food manufacturing employment in Dawson County varied between 400 and 800 during the 1980s, but jumped abruptly to approximately 2,000 in 1991 and stayed within the 2,000–3,000 range thereafter. Fig. 1 shows the real food manufacturing sector wage for Dawson County from 1984 (the first year such data were available) to 2000. Since food manufacturing in the county consists of mostly meatpacking, this wage series is essentially a meatpacking wage series. The food manufacturing wage declined during the 1980s, but rose steadily shortly after IBP commenced operations in Lexington. Thus, IBP’s decision to locate in Dawson County appears to have created a substantial labor demand shock, pushing up both wages and employment. Hispanic workers were pulled into the county by very favorable demand conditions in the county’s export sector.
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
150
Table 1. Ranking
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Top 20 Counties in Meatpacking Employment and Growth in Hispanic Population Share During 1990s. Percent of State Population that was Hispanic in 1990
Percent of State Population that was Hispanic in 2000
23.75 22.57 22.08
0.99 1.97 0.99
5.5 7.0 5.5
29.9 43.3 22.6 46.7
19.95 18.04 16.53 16.50
1.34 1.97 0.99 2.37
5.2 7.0 5.5 7.5
18.39
31.2
12.81
5.1
17.1
49,063 20,411
2.54 0.8
15.1 12.5
12.56 11.7
0.47 0.46
4.7 2.8
21,139 35,935 20,832 49,040 53,534 25,357 60,161
1.0 6.09 1.3 1.93 4.32 0.17 1.54
12.7 16.7 11.2 11.7 14.0 9.7 10.08
11.7 10.61 9.9 9.77 9.68 9.53 9.26
0.29 1.97 0.50 0.47 0.99 0.29 0.47
3.2 7.0 2.9 4.7 5.5 3.2 4.7
21,681
0.2
9.4
9.2
0.42
2.1
12,183 49,329
3.67 0.7
12.6 9.6
8.93 8.90
0.46 0.47
2.8 4.7
Percent of Change Between Percent of 1990 and 2000 in County County Percent of Population Population County that was that was Hispanic in Hispanic in Population that was Hispanic 2000 1990
County
Population in 2000
Colfax, NE Seward, KS Dawson, NE Texas, OK Finney, KS Dakota, NE Franklin, WA Morgan, CO Duplin, NC Buena Vista, IA Yell, AR Lyon, KS Nobles, MN Lee, NC Hall, NE Carroll, AR Sampson, NC McDonald, MO Louisa, IA Chatham, NC
10,441 22,510 24,365
2.45 19.53 3.28
26.2 42.1 25.36
20,107 40,523 20,253 49,347
9.95 25.26 6.07 30.20
27,171
Source: 1997 Census of Manufacturing and American Factfinder (both at http://www. census.gov).
Colfax and Seward Counties also experienced large demand-pull immigration during the 1990s. However, since their meatpacking industries are older, some of the recent immigration to these counties is likely to have been of the supply-push type, as family and friends followed earlier migrants who were well-established in their adopted communities.9 Colfax and Seward Counties do not fit our model because earlier immigrants to those counties were likely to have spilled over into the retail sector, violating our assumption that immigrants only enter the export sector labor market. Dawson
151
Does Immigration Affect Labor Demand? Model and Test
Table 2. Food Manufacturing Employment in Dawson County, 1980–2000. Year
Persons Employed
Year
Persons Employed
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
520 404 577 589 611 785 505 458 338 369 386
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
1,967 2,334 2,373 2,528 2,587 2,739 2,674 Between 2,500 and 4,999a 2,763 Between 2,500–4,999a
Source: U.S. Bureau of the Census, County Business Patterns. a Due to confidentiality considerations, point estimates for employment for 1998 and 2000 were not available.
Inflation-adjusted dollrs
Food manufacturing wage in Dawson County, 1984-2000 20000 15000 10000 5000 0 1980
1985
1990
1995
2000
2005
Year
Fig. 1.
Food Manufacturing Wage in Dawson County, 1984–2000. Source: Nebraska Department of Labor.
County’s more recent immigrants began shifting into the local retail sector in significant numbers only in the early 2000s. Fig. 2 shows the behavior of the average annual real wage paid to retail workers between 1969 and 2000 in Dawson County as well as the average of the same wage during the same period for eight other rural control counties on the Great Plains. These eight counties are Keith and Red Willow Counties in Nebraska, Marshall and Wilson Counties in Kansas, Beadle and
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
152
Inflation-adjusted dollars
Retail wages in Dawson County, 1969-2000 14000 12000
Dawson
10000 8000
Comparison
6000 4000 2000 0 1965
1970
1975
1980
1985
1990
1995
2000
2005
Year
Fig. 2.
Retail Wages in Dawson County, 1969–2000. Source: Iowa State University (SETA).
Davison Counties in South Dakota, and Carroll and Hardin Counties in Iowa. These counties were selected because they have socio-economic characteristics very similar to Dawson County, while at the same time they did not experience significant levels of Hispanic immigration during the 1990s or an exogenous capital shock in the form of a new and large plant such as in the case of Dawson County.10 Furthermore, these eight counties provide insight into how Dawson County’s economy performed relative to similar Great Plains counties that did not experience Hispanic immigrant inflows. Fig. 2 shows that the behavior of retail wages in Dawson County was very similar to the behavior of wages in the comparison counties, indicating that the economies of all these counties were strongly influenced by regional and national economic forces. Wages fell significantly during the 1970s and 1980s, but less so during the early to mid-1990s. For most of the period, Dawson’s wage was lower than the wage in the comparison counties. However, around 1990 Dawson’s wage stabilized and overtook the benchmark wage thereafter. These data strongly suggest that Dawson County’s retail labor market began to outperform its Great Plains counterparts around the time of the initial large Hispanic immigrant influx.
4. ESTIMATING THE LOCAL DEMAND EFFECTS OF IMMIGRATION Dawson County allows for a very convenient test of the demand effects of immigration because immigrant inflows to its export sector during the 1990s
Does Immigration Affect Labor Demand? Model and Test
153
were exogenous to its retail sector. Because of the clean separation in Dawson County between the labor markets where immigrants work and where they spend, immigration does not create a simultaneity problem when estimating a retail wage equation for the county. Consequently, the marginal effects of immigration on retail wages can be estimated without resorting to instrumental variables estimation techniques. This is in contrast to a single labor market model, such as the one used by Altonji and Card (1991) and others, in which wages are simultaneously determined by immigrant inflows, migratory responses, and local demand effects. In such a single-market model, the researcher has to explicitly deal with potential estimation bias caused by simultaneous supply-push and demand-pull immigration, endogenous migratory responses by native workers, and the demand-augmenting effects of immigration. The single-market model requires a complex system of simultaneous equations or extensive data is (often unavailable for local economies) needed to instrument all the simultaneously determined variables. There are thus obvious econometric advantages to estimating immigration’s labor demand effect using the Dawson County test case of segmented labor markets.
4.1. Specifying the Regression Model The objective of our empirical analysis is to estimate Eq. (16), the equation for the equilibrium retail wage. Our sample includes the nine counties described above, each with 20 annual observations running from 1980 to 1999. The eight control counties are included in the sample so as to effectively control for general macroeconomic conditions that influenced Dawson County’s economy during the sample period. As Mundlak (1978) suggests, cross-section regressions are attractive in that they can easily capture the often-large differences in variables across counties. Even though the eight counties in our sample were selected due to their strong similarities to Dawson County, there are likely to be other remaining differences not captured by the explanatory variables. We therefore ran the panel regressions with fixed effects in order to capture those remaining differences between the counties. The first step in the estimation process was to obtain estimates for the equilibrium levels of immigrant labor (I) and native labor in the export sector (N X). These values had to be estimated because, unfortunately, there are no data available on the immigrant and native shares of each county’s labor force. The decennial U.S. census does provide the Hispanic share of
154
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
every county’s population, however, and Hispanic population and immigrant labor force shares are likely to be closely correlated because, until recently, there was no Hispanic population in the Great Plains counties. Furthermore, the various states provide annual data on population, births, and deaths for each county. We used these annual data for each county, along with the decennial census data, to distinguish between in- and out-migration of natives and immigrants during the year from the beginning-of-year stocks of natives and immigrants. This approach allows us to test directly for the effects of annual immigrant inflows and native outflows on local labor markets. We estimated the equilibrium levels of native and immigrant labor using the following procedure. First, we specified the equilibrium level of native population each year as equaling the initial (beginning-of-year) stock N0 minus net out-migration during the year (OM): N nX ¼ N 0
OM
(21)
Furthermore, the equilibrium level of immigrant population each year equals an initial stock (I0) plus net in-migration during the year (IM): I nX ¼ I 0 þ IM
(22)
In- and out-migration were calculated in the following manner. We used decennial census data on Hispanic population shares to proxy the size of the Hispanic labor force during each of the census years. We then used the annual county data on population, births, and deaths to extrapolate the annual stocks of native and immigrant labor for the years between the decennial census years. Net in-migration and out-migration during the year were calculated based on the simplifying assumptions that: (1) out-migration is strictly the net outflow of natives (non-Hispanics) from the county; and (2) new inmigration is strictly the net inflow of Hispanics into the county. Beginning with annual population estimates, new immigration, IMt, for the county was calculated as follows: IMt ¼ ðPopt Þn ð% Hispanict Þ
ðPopt 1 Þn ð%Hispanict 1 Þ
(23)
Note that Popt is the annual population estimate and (% Hispanict) refers to the annual Hispanic population share. The Hispanic population share was constructed for the years between the decennial census years 1980, 1990, and 2000 (United States Department of Commerce, 1982, 1993, 2003) by setting the annual adjustments during the interim years proportional to the growth of manufacturing employment. This adjusted extrapolation procedure
155
Does Immigration Affect Labor Demand? Model and Test
recognizes that immigration to Dawson County and the other counties is likely to have occurred in direct response to manufacturing employment opportunities. The annual change was added to the previous year’s total. Out-migration was estimated as: OMt ¼ Popt
1
Popt þ Birthst
Deathst þ IMt
(24)
Birthst and Deathst refer to the annual numbers of births and deaths, respectively, each year. Note that new immigration and out-migration are net variables. There is evidence that many immigrants have departed Dawson County during the 1990s (Gouveia & Stull, 1997). There was no doubt some in-migration of non-immigrants, such as return of natives working or studying elsewhere. We do not have data on such movements, however, and we are therefore forced to use net measures of new immigration and outmigration. We employ the linear regression equation for the retail wage below: WR ¼ d0 þ d1 I 0 þ d2 IM þ d3 N 0 þ d4 OM þ d5 PX þ d6 V N þ d7 V I þ a0 X 0 þ f0 Z 0 þ 0
0
ð25Þ
in which X is a vector of county fixed effect controls, Z a vector of other control variables, a0 and f0 the coefficient vectors and e an error term. Since in- and out-migration are exogenous to the retail wage, instrumental variables are not called for. Therefore, a single-equation estimate of d2 provides an unbiased estimate of the marginal effect of immigrant inflows to the export sector on the retail wage.
4.2. Variable Descriptions and Data Sources Data on county immigrant populations and the data used to calculate in- and out-migration were obtained from several sources. Population data are from the website of the Office of Social and Economic Trend Analysis (SETA) at Iowa State University (www.seta.iastate.edu). Birth rates and death rates are from the state health departments, while the number of Hispanic immigrants are from the U.S. Census bureau’s website (www. census.gov). The export price (PX) variable is proxied by U.S. real GDP. While the price of meat is the dominant export price for Dawson County, it is not relevant to the eight comparison counties where the meatpacking industry is not a major employer. The advantage of U.S. real GDP is that it is correlated with the average level of prices for goods exported from all the
156
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
counties. Data for U.S. real GDP were obtained from the FRED database at the Federal Reserve Bank of St. Louis website (www.stls.frb.org). We used the average wages in Kansas City (Missouri), Des Moines (Iowa), and Sioux Falls (South Dakota) to proxy the native reservation wage (VN) as well as the reservation wage faced by earlier immigrants who migrated to the county from another region in the U.S. It is presumed that attractive alternative employment opportunities for workers in the two South Dakota counties lie in Sioux Falls and that workers in the two Iowa counties face relatively attractive alternative opportunities in the Des Moines labor market. Workers in the Kansas and Nebraska counties are presumed to have the best alternative employment opportunities in the Kansas City labor market. Data on these urban wages are from the Bureau of Economic Analysis’s website (http://www.bea.gov). The immigrant reservation wage (VI) is proxied by the annual real Mexican GDP, which was obtained from International Financial Statistics (International Monetary Fund, 2002). We included two control variables – the real U.S. minimum wage and a time variable. The former is included to capture any secular trends in wages, and the latter is included to capture any secular trends in technological progress. Minimum wage data were obtained from the U.S. Department of labor website (http://www.dol.gov). Finally, we included two dummies for two potentially important shocks that affected the Great Plains states. Specifically, we included a dummy for the ‘‘farm crisis’’ from 1985 through 1987, a period of exceptionally low farm incomes, and a dummy for the shift in government assistance to farmers after the ‘‘Freedom to farm Act’’ of 1996. The sharp changes in farm income during these periods are likely to have affected retail demand in the nine rural counties. Table 3 shows population-related descriptive statistics for the nine counties in our sample. Consistent with general demographic trends in rural areas on the Great Plains, all counties except Davison, South Dakota, lost population between the 1980s and 1990s. Dawson County’s population fell only very slightly due to the large influx of Hispanic immigrants (5,247) offsetting continual native out-migration. All counties except Dawson experienced little or no change in Hispanic population. All the counties, to varying degrees, experienced at least some in- and out-migration.
4.3. Regression Results We report results for a slightly modified version of Eq. (25). This version includes the sum of initial immigrant population and immigrant inflow
157
Does Immigration Affect Labor Demand? Model and Test
Table 3.
Selected Descriptive Statistics for Individual Counties.
Description
County
Mean
1980 Population (final)
Dawson Keith Red Willow Marshall Wilson Beadle Davison Carroll Hardin
23,307 9,360 12,646 12,761 12,042 19,190 17,826 22,964 21,763
1999 Population (final)
Dawson Keith Red Willow Marshall Wilson Beadle Davison Carroll Hardin Dawson
23,277 8,877 11,304 10,908 10,399 16,637 17,858 21,518 18,159 5,247
Change in number of Hispanic residents between 1990 and 1999
In-migration
Out-migration
Keith Red Willow Marshall Wilson Beadle Davison Carroll Hardin Dawson Keith Red Willow Beadle Davison Wilson Marshall Carroll Hardin Dawson Keith Red Willow Beadle Davison
13 11 6 2 52 2 0 23 268.37 2.02 6.24 5.22 4.41 4.44 2.36 1.50 18.04 323.97 51.07 106.19 190.02 89.56
158
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
Table 3. (Continued ) Description
County Wilson Marshall Carroll Hardin
Mean 66.89 89.26 157.7 209.34
Source: Iowa State University, U.S. Census Bureau, and Federal Reserve Bank of St. Louis.
during the year (the sum equaling the estimated value of I) as a regressor. This modification was necessary because we found that the calculated values for initial immigrant population and within-period immigrant inflow were very highly correlated. When initial immigrant population and in-migration were included as separate regressors, their estimated coefficients varied extremely from one regression to another, suggesting substantial multicollinearity. Thus, to correct for multicollinearity, we summed the two immigration variables.11 We then used the regression coefficient on our proxy for I X as an estimate of the ceteris paribus effect of immigration on the retail wage. To avoid potential problems with heteroskedasticity, we estimated the regressions using generalized least squares (GLS). Coefficient estimates from the modified version of Eq. (25) are shown in Table 4. We find strong evidence of local demand effects of immigrant inflows: The estimated coefficient for immigrant population is positive and significant at the 95% level. The coefficient value implies that, all other things equal, each immigrant raises the annual retail wage of the county population by about $0.17. Thus, the arrival of nearly 6,000 immigrants in Dawson County is predicted to have raised annual wage income in the retail sector by nearly $1,000, after taking into consideration all other influences on county wage income. The estimated coefficients for the remaining variables in Table 4 are mostly as expected. A higher reservation wage for natives and earlier immigrants (proxied by the regional urban wage) increases retail wages. The stock of native labor increases the retail wage, but out-migration was not found to have a significant effect. These results are consistent with particular scenarios predicted by our theoretical model: out-migration reduces both the consumer base and the labor supply. Thus, the two effects could offset each other. The regression results suggest that the overall stock of native population has a predominantly positive effect on wages, but, at the margin, the wage effect of the reduction in native labor just offsets the wage effect of
159
Does Immigration Affect Labor Demand? Model and Test
Table 4.
Estimated County Retail Wage Equation.
Independent Variable
Coeffient (t-value)
U.S. minimum wage
587.44 (1.40) 0.1650 (2.14) 0.1692 (1.77) 0.0318 ( 0.14) 0.5863 (1.08) 0.1850 (2.51) 0.2724 (0.14) 128.80 (0.84) 21240 (0.86) 18,963 (0.87) 17,281 (0.93) 14,117 (0.92) 11,051 (0.90) 10,007 (1.07) 7,076 (1.12) 2,092 (0.68) 115.83 (0.97) 142.63 ( 0.45) 0.97 180
Immigrant population Native population stock Out-migration U.S. real GDP Regional urban wage Mexican GDP Time County dummy 1 County dummy 2 County dummy 3 County dummy 4 County dummy 5 County dummy 6 County dummy 7 County dummy 8 Farm crisis dummy Freedom to farm dummy R2 Sample size Note: (t-statistics in parentheses; 1%, 5% and 10% noted with Dependent Variable: County Retail Wage. Significant at 10% Significant at 5% Significant at 1%
, and , respectively)
160
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
the reduction in the supply of labor. Mexican GDP, which proxies the reservation wage for migrants who came directly from across the border, is not statistically significant. In fact, we tried a number of other proxies for the reservation wage facing new immigrants, but none performed as expected. Most likely, the decision to immigrate is based on a complex set of factors that are difficult to capture in a single variable. Finally, the fixed effect dummies plus the two dummies for shocks to farm income were not statistically significant; we still feel justified using the fixed effects model rather than the simple panel model because the set of dummies is still likely to contribute toward reducing omitted variable bias across the nine counties and 20 years of agricultural policies. We also performed a Hausman test for simultaneity. According to Eq. (16) and our procedures for estimating the annual stocks of immigrant and native labor above, the various functional relationships embedded in the equilibrium retail wage can also be expressed in terms of variables that are clearly exogenous to the local economy. That is,
W nR ¼ f N nX ðPX ; N 0 ; I 0 ; V I ; V N Þ; I n ðPX ; N 0 ; I 0 ; V I ; V N Þ (26)
We used this set of exogenous variables, plus the time trend and U.S. minimum wage as the exogenous (instrumental) variables for the Hausman test. The test statistic of 1.46 means that we cannot reject the null hypothesis of no simultaneity, indicating that our model avoided simultaneity as we expected and we were justified in using GLS to estimate the demand effect of immigration using the single regression Eq. (25). The retail wage is not the only variable with which to measure the local demand effect of immigration. In order to measure the demand effect of migration to California and other West Coast states during the World War II period, Rhode (2003) used housing prices to proxy the local demand effects of migration. Housing prices may reflect people’s expectation of the long-run economic health of a region more accurately than retail wages do. Table 5 shows the regression results when we substitute median county housing prices, obtained from the decennial census of housing (1980–2000), for local retail wages.12 Note that the regression results are qualitatively similar, but not identical, to the results reported in Table 4. Specifically, an additional immigrant to the county raises the median housing price by over $2. Note the negative coefficient for the native and earlier immigrant reservation wage: housing prices are inversely related to regional urban wages. This result may reflect the general trend for the differential between rural and urban housing prices to increase over time. The result may also reflect the complex relationships among the variables in the model, which
161
Does Immigration Affect Labor Demand? Model and Test
Table 5.
Estimated County Housing Price Equation.
Independent Variable
Coeffient (t-value)
U.S. minimum wage
3492.6 (1.28) 2.2882 (4.58) 2.8463 (1.77) 1.7699 (1.22) 4.9086 (1.40) 1.9611 ( 4.11) 0.3715 ( 0.03) 416.44 (0.42) 66,977 (0.42) 97,767 (0.70) 76,522 (0.64) 33,172 (0.33) 31,645 (0.40) 40,857 (0.67) 31,040 (0.76) 7,950 (0.40) 452.48 (0.58) 110.86 (0.05) 0.89 180
Immigrant population Native population stock Out-migration U.S. real GDP Regional urban wage Mexican GDP Time County dummy 1 County dummy 2 County dummy 3 County dummy 4 County dummy 5 County dummy 6 County dummy 7 County dummy 8 Farm crisis dummy Freedom to farm dummy R2 Sample size Note: (t-statistics in parentheses; 1%, 5%, and 10% noted with Dependent Variable: Median County Housing Price. Significant at 10% Significant at 5% Significant at 1%
, and , respectively)
162
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
underscores the importance of estimating alternative regression equations to examine the robustness of our results. It is therefore reassuring that the coefficient for immigration is positive and highly significant in both Tables 4 and 5. Immigration’s labor demand effects are revealed in both retail wages and local housing prices. An interpretation of our regression results must recognize that we used a model for a very special case. The regression methodology cannot be generalized across all immigration episodes. In fact, the model will not apply to Dawson County in the long run as immigrants and especially their offspring spread into other sectors of the local labor market. The results of our special case are of general importance, however. Our ability to isolate and measure immigration’s labor demand effect in Dawson County confirms that immigration indeed has a positive labor demand effect. We find that the effect is quantitatively important, amounting to about $1,000 per capita, or almost 5% of per capita income in Dawson County. Our study thus supports the small number of previous studies, which pointed out that immigration is likely to have a positive effect on labor demand, such as Harrison (1983), Altonji and Card (1991); Hercowitz and Yashiv (2001), and Rhode (2003).
5. CONCLUSIONS As intuition would suggest, but most previous immigration researchers have ignored, immigrants increase both the supply of and the demand for labor. Immigrants are consumers as well as workers. When immigrants arrive because, for example, they are pushed to immigrate by factors exogenous to the destination region, they add to the supply of labor, but they increase the consumer base as well. Consequently, there is a positive wage effect in the receiving economy that partially offsets the negative effect predicted by a model based only on supply-push immigration. In the case of demandpull immigration, this local demand effect is potentially large because immigrants’ local spending is fueled by both higher wages and increasing numbers of immigrants. The likely reason researchers have generally omitted immigration’s labor demand effects from their models is that it is very difficult to isolate immigration’s effect on labor demand in a single labor market model. We exploited the special case of Dawson County, Nebraska, a county in which there was separation between the markets where immigrants worked and where they spent. We were thus able to obtain an unbiased estimate of the
Does Immigration Affect Labor Demand? Model and Test
163
demand effect of immigration without having to use instrumental variables estimation or other statistical procedures that may introduce new sources of bias. Our estimates for Dawson County and eight other control counties on the Great Plains show that immigration during the 1990s raised local annual wage income by about $1,000, roughly 4–5% of current average wages in those counties. While our specific model and empirical approach is limited to special cases such as the one we found in Dawson County, our finding that immigration has strong labor demand effects applies to all immigration episodes. Previous studies of the effects of immigrant inflows on local employment outcomes need to be reevaluated. For example, even though Card (2001) makes substantial progress in adjusting native wage estimates for migratory responses to immigrant inflows, his study does not in any way control for the demand-augmenting effects of immigration. In fact, our study suggests that Card’s estimates of the effects of exogenous immigrant inflows may be biased. It would also be interesting to carefully examine other studies of immigration to see how the explicit inclusion of immigration’s labor demand effects helps to explain the frequent finding, reported by Friedberg and Hunt (1995), that immigration has little effect on local wages.
NOTES 1. Quoted in Simon (1989, p. 208). 2. For example, Card (1990) studied the 1980 Mariel Boatlift of Cubans to Miami, and Hunt (1992) studied the 1962 repatriation of French colonists from Algeria to France. Carrington and de Lima (1996), Friedberg (2001), and Suen (2000) studied the repatriation of overseas Portuguese following the independence of Portugal’s African colonies in 1973, Jewish migration to Israel after the fall of the Soviet Union, and Chinese refugees to Hong Kong, respectively. 3. See, for example, Altonji and Card (1991), Borjas (1987), Card (2001), LaLonde and Topel (1991), Pischke and Velling (1997), Pedace (1998), and Schoeni (1997). 4. See also Millman (2004). 5. Several authors have studied immigrant networks and their influence on immigration flows. See,for example, Stark (1991) and Edin, Frederiksson, and Aslund (2003). 6. An example of a utility function generating such a demand function is U ¼ qRq0, where qR is the individual’s consumption of the retail good and q0 is consumption of other goods (see Henderson & Quandt, 1980, pp. 18–19). 7. Immigrant remittances from developed to developing countries are very large, although estimates of their size vary. A recent study by the Inter-American
164
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
Development Bank (IDB, 2004) states that over 90 cents out of every dollar earned by immigrants in the U.S. stays in their adopted communities, creating a huge boost to local economies (see Millman, 2004). Za´rate-Hoyos (2001) estimates that Mexican migrants living in the U.S. send between $300 and $500 home to Mexico each month. The Pew Hispanic Center (2002) estimates that the average foreign-born U.S. resident remits $1,260 to relatives abroad annually. 8. For most of the 1990s and certainly by the end of our sample period, Hispanic immigrant workers in Dawson County were concentrated in meatpacking. In the last several years, though, new Hispanic immigrants have taken jobs in the retailing sector or started their own retail businesses. In addition, some of these immigrants have obtained work in the retail sector after having been employed in meatpacking. 9. Several recent studies of immigration show that family ties often overwhelm employment opportunities and other influences in determining in which states recent immigrants to the U.S. decided to settle; see, for example, Zawodny (1997), Murayama (1991), and Kahan (1978). 10. Each of the eight comparison counties are rural, low income, and have generally the same cost of living. In 1980, Dawson County’s population was 23,307 compared to an average of 16,069 for the comparison counties and by 2000 Dawson’s population stood at 23,227, compared to an average of 14,458 for the comparison counties. While Dawson County experienced a very mild population increase between 1980 and 2000, the other eight counties experienced losses averaging approximately 10%. The population losses in the comparison counties reflect the general out-migration experienced by many rural counties on the Great Plains since 1980. The Hispanic share of the population in the comparison counties has always been very low, averaging 1.7% in 2000. In addition, earnings per job in the comparison counties averaged $21,163 in 1999, about 3% less than Dawson County’s average. Unemployment rates in the comparison counties averaged 2.73% in 2000, compared to 2.8% for Dawson County. Unadjusted for inflation, the annual manufacturing wage in Dawson County averaged $19,208 for the sample period, compared to an average of $17,122 for the eight comparison counties. Unadjusted for inflation, retail wages averaged $8,813 during the sample period in Dawson County, compared to $9,001 in the comparison counties. However, median housing prices in the Nebraska counties were significantly higher ($42,177 in Dawson and an average of $43,504 in the other two Nebraska counties, compared to $25,533 in the other states’ counties). While higher housing prices often indicate a higher cost of living, this is likely not the case, especially since retail wages and housing prices are relatively uncorrelated. Therefore, we rule out the possibility that immigration could be endogenous to living costs. The only distinct difference between Dawson County and the eight comparison counties is that Dawson has a relatively large Hispanic workforce that is concentrated in an export sector, but spends in the retail sector, whereas the comparison counties have almost no immigrants. 11. Goldberger (1991, pp. 250–251) points out that when two regressors are highly collinear, precise estimation of their separate marginal effects is clearly hampered. However, the estimate of the sum of the two is more precise than the sum of the individual estimates when there is collinearity. 12. We used the 1980 price for 1980–1989, the 1990 price for 1990–1998, and the 2000 price for 1999.
Does Immigration Affect Labor Demand? Model and Test
165
ACKNOWLEDGMENTS We thank two anonymous referees and participants at the ‘‘Immigration, Minorities and Social Exclusion’’ conference, held at Bar-Ilan University in June 2004, for very helpful comments as well as Jim Schmidt and Joshua Lewer for valuable suggestions on the econometrics used.
REFERENCES Altonji, J., & Card, D. (1991). The effects of immigration on the labor market outcomes of lessskilled natives. In: J. Abowd & R. Freeman (Eds), Immigration, trade and the labor market. Chicago: University of Chicago Press. Borjas, G. (1987). Immigrants, minorities, and labor market competition. Industrial and Labor Relations Review, 15, 382–392. Borjas, G. (2003). The labor demand curve is downward sloping: Reexamining the impact of immigration on the labor market. Quarterly Journal of Economics, 118(November), 1135–1174. Borjas, G., Freeman, R., & Katz, L. (1996). Searching for the effect of immigration on the labor market. American Economic Review Papers and Proceedings, 86(May), 246–251. Card, D. (1990). The impact of the Mariel boatlift on the Miami labor market. Industrial and Labor Relations Review, 43(2), 245–257. Card, D. (2001). Immigrant inflows, native outflows, and the local labor market impacts of higher immigration. Journal of Labor Economics, 19(1), 22–64. Carrington, W., & de Lima, P. (1996). The impact of 1970s repatriates from Africa on the Portugese labor market. Industrial and Labor Relations Review, 49(1), 330–347. Edin, P. A., Frederiksson, P., & Aslund, O. (2003). Ethnic enclaves and the economic success of immigrants – Evidence from a natural experiment. Quarterly Journal of Economics, 118(1), 329–357. Friedberg, R. M. (2001). The impact of mass migration on the Israeli labor market. Quarterly Journal of Economics, 116(4), 1373–1408. Friedberg, R. M., & Hunt, J. (1995). The impact of immigrants on host country wages, employment and growth. Journal of Economic Perspectives, 9(Winter), 23–44. Goldberger, A. S. (1991). A course in econometrics. Cambridge, MA: Harvard University Press. Gouveia, L., & Stull, D. D. (1997). Latino immigrants, meatpacking, and rural communities: A case study of Lexington, Nebraska. Julian Samora Research Institute Research Report # 26 (August), Michigan State University. Harrison, D. S. (1983). The impact of recent immigration on the South Australian labour market. Report to Committee for the Economic Development of Australia, May. Henderson, J., & Quandt, R. (1980). Microeconomic theory: A mathematical approach. New York: McGraw-Hill. Hercowitz, Z., & Yashiv, E. (2001). A macroeconomic experiment in mass immigration. Center for Economic Policy Research (CEPR) Discussion paper # 2983, September (www.cepr.org/pubs/dps/DP2983.asp).
166
O¨RN B. BODVARSSON AND HENDRIK VAN DEN BERG
Hunt, J. (1992). The impact of the 1962 repatriates from Algeria on the French labor market. Industrial and Labor Relations Review, 45, 556–572. Inter-American Development Bank (IDB). (2004). Sending money home. May 17. International Monetary Fund. (2002). International Financial Statistics (CD-ROM version), Washington, DC, (Fall). Kahan, A. (1978). Economic opportunities and some pilgrims’ progress: Jewish immigrants from Eastern Europe in the U.S., 1890–1914. Journal of Economic History, 38(1), 235–251. LaLonde, R. J., & Topel, R. H. (1991). Labor market adjustments to increased immigration. In: J. M. Abowd & R. B. Freeman (Eds), Immigration, trade, and the labor market. Chicago: University of Chicago Press. Millman, J. (2004). Immigrants spend earnings in U.S.. Wall Street Journal, 263(May), 17. Mishan, E. J., & Needleman, L. (1966). Immigration, excess aggregate demand and the balance of payments. Economica, 33(May), 129–147. Mishan, E. J., & Needleman, L. (1968). Immigration: Some long term economic consequences. Economia Internazionale, 21(May), 281–300. Mundlak, Y. (1978). On the pooling of time series and cross section data. Econometrica, 46(1), 69–85. Murayama, Y. (1991). Information and immigrants: Interprefectual differences of Japanese emigration to the Pacific Northwest, 1880–1915. Journal of Economic History, 51(1), 125–147. Pedace, R. (1998). The impact of immigration on the labor market for native-born workers: Incorporating the dynamics of internal migration. Eastern Economic Journal, 24(4), 449–462. Pew Hispanic Center. (2002). Billions in motion: Latino immigrants, remittances and banking. Los Angeles: Pew Hispanic Center; downloaded from the Pew Foundation website on November 11. Pischke, J.-S., & Velling, J. (1997). Employment effects of immigration to Germany: An analysis based on local labor markets. Review of Economics and Statistics, 79, 594–604. Rhode, P. W. (2003). After the war boom: Reconversion on the U.S. Pacific Coast, 1943–1949. NBER Working paper No. w9854, July 2003. Schoeni, R. F. (1997). The effects of immigrants on the employment and wages of native workers: Evidence from the 1970s and 1980s. The RAND Corporation, March. Simon, J. L. (1989). The economic consequences of immigration. Washington, DC: Cato Institute. Stark, O. (1991). The migration of labor. Oxford, UK: Blackwell. Suen, W. (2000). Estimating the effects of immigration in one city. Journal of Population Economics, 13(1), 99–112. United States Department of Commerce. (1982). Bureau of the census, 1980 decennial census of population and housing. Washington, DC: U.S. Government Printing Office. United States Department of Commerce. (1993). Bureau of the census, 1990 decennial census of population and housing. Washington, DC: U.S. Government Printing Office. United States Department of Commerce. (2003). Bureau of the census, 2000 decennial census of population and housing. Washington, DC: U.S. Government Printing Office. Za´rate-Hoyos, G. A. (2001). The case for a remittance policy in Mexico. Paper prepared for the Pacific Coast Council on Latin American Studies Conference Building Bridges, April 5–7. Tijuana, Mexico; downloaded from the Federal Reserve Bank of Chicago website, May 11, 2003. Zawodny, M. (1997). Welfare and the locational choice of new immigrants. Economic Review, Federal Reserve Bank of Dallas, 2ndQuarter, 2–10.
LENIENT POLICY PROPOSAL FOR THE STRUGGLE AGAINST ILLEGAL IMMIGRATION Nava Kahana and Tikva Lecker ABSTRACT Policy strategies directed against illegal immigration have largely concentrated on border and domestic enforcement. This paper suggests that host countries could consider a more lenient approach: contributing to the deportation cost of self-reporting illegal immigrants. The increase in illegal immigration that such a reward might bring about is shown to be more than offset by the rise in the number of self-reporting illegal immigrants leaving the rich country, with a concomitant decrease in the number of remaining illegal immigrants. An added advantage of this policy is that the self-reporting immigrants would be predominantly the relatively lower socio-economic group. When adopting this policy, the rich country must choose the appropriate mix of two policy means: funds allocated to strengthening its border and domestic control; and rewards to self-reporting illegal immigrants.
Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 167–175 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24005-0
167
168
NAVA KAHANA AND TIKVA LECKER
INTRODUCTION Rich countries often face sizeable illegal immigration. According to an estimate of the Immigration and Naturalization Service (INS), five million illegal aliens live in the United States, with an additional 400,000 settling there annually (Camarota, 1998). Most immigrants choose ‘‘traditional’’ destinations. More than half go to the United States, Canada and Australia. However, many other countries also have relatively large immigrant flows. Even Japan, which is regarded as being very homogeneous and geographically immune to immigrants, reports major problems with about 300,000 illegal aliens, primarily low-skilled workers from various parts of Asia (Borjas, 1994). In every country, there are people who object to the presence of illegal immigrants since they fear that these immigrants will ultimately steal their jobs and increase crime rates. On the other hand, there are people keen on hosting more illegal immigrants, since they need them desperately to fill vacant positions in agriculture, in service sectors such as housecleaning, in care for the aged and in gardening. However, the majority of voters in most countries are opposed to the mass of illegal immigrants. Rich countries use diverse policy strategies to contend with this phenomenon due to its negative impact on the wellbeing of the majority of voters, as discussed by Djajic (1997, 1999), among others. Edgardo, Campos, and Lien (1995) proposed helping source countries maintain their political stability, thereby moderating the flow of illegal labor. Myers and Papageorgiou (2000), Kahana and Lecker (2005), among others, claimed that transferring financial aid to the source countries could eliminate the welfare differential between destination and sending countries, which is said to be at the root of immigration, thereby reducing immigration pressure along the border. Lecker (2000) suggested imposing fines on the source countries of illegal immigrants. Ethier (1986) and Chiswick (1988), among others, suggested border and domestic enforcement. Current United States policy concentrates on this type of enforcement, which aims at making the cost of crossing the border prohibitive. For example, since the late 1980s, the United States has doubled the force protecting the United States–Mexico border (Hanson & Spilimbergo, 1999). However, illegal immigrants apprehended by the INS, who agree to voluntary deportation, are not processed by the United States judiciary system. They spend only a short time in custody and are not restricted from re-entering the country legally. In Malaysia, due to harsh legislation introduced in 2002, a foreigner caught working without a permit may face a fine of $2,600, a mandatory jail term of five years and six lashes
The Struggle Against Illegal Immigration
169
with a rattan cane. Immigration laws introduced in March 2002 gave illegal immigrants an amnesty, which lasted until August 2002, to afford them time to leave the country without fear of arrest. This resulted in 318,272 illegal immigrants leaving this country. In July 1996, the United Arab Emirates (UAE) launched a three-month amnesty, during which illegal aliens who had overstayed their visa could leave the UAE without penalty. On September 9, 1996, the free-departure amnesty was extended to include all illegal aliens. By October 31, 1996, it was estimated that at least 167,000 illegal immigrants, which are equivalent to seven percent of the population, had left the UAE. Since January 1, 1997, illegal immigrants face up to three years in jail and a fine of 30,000 dirhams ($8,147). We propose that host countries adopt a more lenient approach to selfreporting illegal immigrants than to those who do not give themselves up. More specifically, in this paper, we suggest that host countries reward selfreporting illegal aliens leaving voluntarily by partially or totally covering their deportation cost. The increase in illegal immigration that may be brought about by such a reward is shown to be more than offset by the rise in self-reporting illegal immigrants returning to their source countries, with a concomitant decrease in the number of remaining illegal immigrants. An added advantage of this policy is that the self-reporting illegal immigrants would be predominantly the relatively lower socio-economic group. The rich country can choose an appropriate mix of a given sum designated for border and domestic enforcement, using the following two policies: allocation of funds to improvement of its border and domestic control and allocation of funds for rewards given to self-reporting illegal immigrants. This can minimize the number of expected illegal immigrants subject to the budget constraint.
THE MODEL Economists view immigration as a cost–benefit decision in terms of utility (e.g., Greenwood, 1985; Chiswick, 1988; Stark, 1991). A risk-neutral individual decides to immigrate if the expected discounted difference in the stream of income between the destination and source countries (herewith, the income gap) exceeds the moving costs. Taking a population of N potential risk-neutral illegal immigrants.1 Prior to immigration, the illegal immigrants have no way of knowing the size of their income gap, y, but do know the income gaps distribution, which for
170
NAVA KAHANA AND TIKVA LECKER
simplicity is assumed to be uniform, such that 0 y 1; i.e., y is determined as a proportion of the maximum possible gap.2 Note that y 0 reflects the fact that the host country is richer than the source country. Individuals planning to immigrate are aware of the non-refundable costs, C, involved, including wages foregone during the immigration process and the costs of transport and safe passage across the border. These costs, which vary according to individual characteristics, are also measured as a proportion of the maximum possible income gap. The density function of C is also assumed to be uniform over [0, 1]. The illegal immigrants return home from time to time to visit their families. The income gap and the corresponding immigration cost refer to the time that passes between two such consecutive visits. Based on the findings of Donato, Durand, and Massey (1992) that the apprehension probability, p is not significantly correlated with observable individual characteristics, including age and previous border-crossing experience, it is assumed that all illegal immigrants face the same p.3 Under the destination countries’ current policy, each illegal immigrant that leaves the destination country either compulsorily or voluntarily bears the deportation cost, T, which is also determined as a proportion of the maximum income gap. Given this policy, the expected pre-immigration net income gap of a potential illegal immigrant is R1 ¼ 0:5ð1
pÞ
pT
(1)
An individual decides to immigrate if the net expected income gap exceeds the immigration costs, i.e., C. The breakeven level of immigration costs, C0, below which migration is worthwhile, is given by C 0 ¼ 0:5½1
pð1 þ 2T Þ
(2)
Since C is uniformly distributed over [0, 1], C0 is the proportion of the illegally immigrating individuals out of the total potential illegal immigrants. Note that C 0 40 if pð1 þ 2TÞo1: On reaching the destination country, the illegal immigrant discovers his income gap and the immigration expenses, C, have become sunk costs. At this stage, he faces a decision of whether to stay or leave. If his decision is to stay, he earns y with probability 1 p and is caught with probability p, in which case he is deported and must pay the deportation costs. However, if he decides to leave, his income gap is zero and he bears the deportation costs. Thus, the respective net expected income gaps for staying and leaving are
171
The Struggle Against Illegal Immigration
R 2 ¼ ð1
pÞy
(3)
pT
and R0 ¼
(4)
T
where T is the deportation cost. Since R2 4R0 ; even for y ¼ 0; the illegal immigrant will not decide to leave. In order to motivate the illegal immigrants to return to their source country, the receiving country should be more lenient with self-reporting than with other illegal immigrants. Under the proposed policy, a share, a, 0oa 1; of the deportation cost of self-reporting illegal immigrants would be covered by the host country. While such a lenient approach would actually encourage illegal immigration, this increase would be more than offset by the number of self-reporting, returning illegal aliens. An illegal immigrant who has the option of leaving voluntarily and only bearing the cost ð1 aÞT; would decide to stay if R2 4 ð1 aÞT: The breakeven level of the income gap, y0, below which he would report and leave voluntarily, is y0 ¼
ðp þ a 1ÞT 1 p
(5)
Since y is uniformly distributed, the probabilities of self-reporting and staying illegally are y0 and 1 y0 ; respectively. Therefore, the range of the income gap for the remaining illegal immigrants is y0 y 1; with an average income gap of 0:5ðy0 þ 1Þ: Note that y0 40 iff a41 p: Prior to immigration, the illegal immigrant would take into account his decision at stage 2, i.e., after arrival. His expected net income gap under the proposed policy is therefore given by pT y0 ð1 aÞT (6) R3 ¼ 1 y0 ð1 pÞ0:5 y0 þ 1 Substituting (5) into (6) gives R3 ¼
½1
pð1 þ T Þ2 þ ða 2ð1
1Þ2 þ 2ða pÞ
1ÞP T 2
(7)
Thus, an individual would decide to immigrate if CoR3 where the breakeven cost below which immigration is worthwhile is C 1 ¼ R3 : The marginal illegal immigrant is characterized by C ¼ C 1 : The increase in the proportion of illegally immigrating individuals out of the total population of potential illegal immigrants, as a result of rewards for selfreporting, is given by
172
NAVA KAHANA AND TIKVA LECKER
C1
C0 ¼
ðp þ a 2ð 1
1Þ2 T 2 40 pÞ
(8)
Since y0 is the rate of self-reporting illegal immigrants, the difference between this rate and the increase in the rate of illegally immigrating individuals, D, is given by D ¼ y0
ðC 1
C0Þ ¼
ðp þ a
1Þð2 pT þ ð1 2ð 1 p Þ
aÞT ÞT
(9)
Since 0 p 1; 0 T 1; and 0oa 1; we obtain D40 for p þ a41: The increase in illegal immigration is therefore more than offset by selfreporting illegal immigrants returning to their source countries, leaving behind the relatively higher socio-economic group. The cost of reducing illegal immigration, TC, is composed of the cost of border and domestic control, E(p) and the cost of rewarding self-reporting illegal immigrants, aðp þ a 1ÞNT=ð1 pÞ: The derivatives of TC with respect to a and p are @TC ðp þ 2a 1ÞNT ¼ 40 @a ð1 pÞ
(10)
@TC dE a2 NT ¼ þ 40 @p dp ð1 pÞ2
(11)
and
Thus,TC increases with a and with p. However, the expected number of remaining illegal immigrants, G, G ¼ ð1 pÞðC 1 y0 ÞN; decreases in a and p, because @G ¼ ða 1ÞT 2 þ TðpT 1Þ No0 (12) @a and @G ¼ ð pð1 þ TÞ 1Þð1 þ TÞ þ ða 1ÞT 2 T No0 (13) @p
The host country aims at minimizing the expected number of illegal immigrants subject to a budget constraint. Therefore, in addition to other policies against illegal immigration, we propose that the receiving country combines two policy means: (1) allocating funds to reinforce border and domestic control and; (2) giving rewards to self-reporting illegal immigrants and chooses an appropriate mix of these two policy means (i.e., a and p), given the sum designated for border and domestic control.
The Struggle Against Illegal Immigration
173
The reward of self-reporting immigrants must not be solely a contribution to their deportation costs. Apprehended illegal aliens in the European Union are often not given an opportunity to take necessary measures to protect their property immediately after arrest. Partial or total property losses are common in such cases. In the context of our proposed model, one may consider enabling the self-reporting illegal aliens to avoid such losses. In South Korea, illegal immigrants (foreign workers who overstayed their ‘‘training program’’) were required to pay a fine before being allowed to leave. That policy has been modified, since the authorities tried to encourage return immigration by waiving the fine. Waiving the fine of self-reporting immigrants is another possible reward. There are already countries that cover the deportation cost of illegal immigrants.4 We suggest that this policy should apply only to self-reporting immigrants, serving as an incentive to self-report and thereby saving enforcement costs.
CONCLUDING REMARKS Various policy strategies have been directed toward reducing the mass of illegal immigrants. We suggest that in addition to border and domestic control, host countries should adopt a more lenient approach to selfreporting immigrants. While such a policy would provide an incentive to increase illegal immigration, it would be more than offset by the increase in the number of self-reporting immigrants leaving voluntarily. Thus, the number of remaining illegal immigrants would decrease. Moreover, the selfreporting immigrants leaving the host countries would be predominantly the relatively lower socio-economic group. When adopting the proposed policy, the rich country must choose the appropriate mix of two policy means: allocating funds for reinforcing its border and domestic control and affording rewards to self-reporting illegal immigrants. Our model is static and thus naturally ignores the problem of repeated immigration. In a static model, in order to prevent successful illegal immigrants (or seasonal illegal immigrants) from taking advantage of the lenient policy and returning home only for a short period, the government should restrict eligibility for a reward to those who earn below a certain income. This will ensure that they do not return. In a dynamic model, the voluntary deportation should be documented in the passport and a threat, in the form of a high enough penalty, should be imposed on those caught trying to
174
NAVA KAHANA AND TIKVA LECKER
sneak back. Moreover, the suggested carrot policy can be modified to be a carrot-and-stick policy. More specifically, the stick is that by self-reporting the illegal immigrants will actually sue their employers, who will have to pay a fine, a portion of which will be assigned as a reward to the illegal immigrants – the carrot. Thus, by self-reporting the immigrant ends the employment relationship and when returning, if at all, may face difficulties finding another suitable job. The modified policy makes hiring an illegal worker a more risky business and reduces the incentive to employ foreigners without a permit.5 The demand for illegal labor will therefore decrease at a given wage, also causing a decrease in the incentive to immigrate illegally.
NOTES 1. For simplicity, we assume risk-neutral individuals, while in case of risk averse individuals our result is even strengthened. 2. In order to simplify the analysis, in the following, the relevant variables (C and T) are defined as proportions of the maximum possible income gap. They are therefore measured in the same units. 3. The probability of being caught crossing the border and the probability of being caught while working illegally are in general different, because they depend on two distinct enforcement policies. For example, the risk of being apprehended during any given attempt to enter the U.S. illegally is about 30%, whereas an illegal alien already established in the U.S. faces 1%–2% probability of being arrested (Yoshida, 2000, p. 9). However, for simplicity, and without changing the main results of the paper, we assume that they are the same. In addition, instead of assuming a different p for different immigrants one can assume a different C. Note that the cost, C, also includes the cost of lowering the probability of being caught, which might be different form person to person. 4. For example, Italy has begun airlifting immigrants back to their point of departure after facilities on the Italian island of Lampodusa were swamped by hundreds of new arrivals. A group of illegal workers who worked at construction sites in the Moscow region was deported from Russia, This group of Tajiks, comprising 115 people, was put on an Ilyushin-76 transport plane and sent back to their motherland. 5. There is asymmetry in information; the employee knows his income gap whereas the employer knows only the income gap distribution and thus the resulting probability of self-reporting.
ACKNOWLEDGMENT The authors are grateful to three anonymous referees for their valuable comments.
The Struggle Against Illegal Immigration
175
REFERENCES Borjas, G. J. (1994). The economics of immigration. Journal of Economic Literature, 32, 1667–1717. Camarota, S. A. (1998). Does immigration harm the poor? Public Interest, 133, 23–32. Chiswick, B. R. (1988). Illegal immigration and immigration control. The Journal of Economic Perspectives, 2(3), 101–115. Djajic, S. (1997). Illegal immigration and resource allocation. International Economic Review, 38(1), 97–117. Djajic, S. (1999). Dynamics of immigration control. Journal of Population Economics, 12, 45–61. Donato, K. M., Durand, J., & Massey, D. S. (1992). Stemming the tide? Assessing the deterrent effects of U.S. immigration and control. Demography, 29(2), 139–157. Edgardo, J., Campos, L., & Lien, D. (1995). Political instability and illegal immigration. Journal of Population Economics, 8, 23–33. Ethier, W. J. (1986). Illegal immigration: The host country problem. The American Economic Review, 76(1), 56–71. Greenwood, M. J. (1985). Human migration: Theory, models and empirical studies. Journal of Regional Science, 25(4), 521–545. Hanson, G. H., & Spilimbergo, A. (1999). Illegal immigration, border enforcement, and relative wages: Evidence from apprehensions at the U.S. Mexico border. The American Economic Review, 89(5), 1337–1357. Kahana, N., & Lecker, T. (2005). Competition as a track for preventing illegal immigration. Economics of Governance, 6(1), 33–39. Lecker, T. (2000). Foreign aid as a discipline on illegal immigration. International Advances in Economic Research, 6(3), 571–577. Myers, G. M., & Papageorgiou, Y. Y. (2000). Immigration control and the welfare state. Journal of Public Economics, 75, 183–207. Stark, O. (1991). The migration of labor. Cambridge: Blackwell. Yoshida, C. (2000). Illegal immigration and economic welfare. Heidelberg: Physica-Verlag.
This page intentionally left blank
176
PART II: GROUP DIFFERENCES AND ECONOMIC ACHIEVEMENT
177
This page intentionally left blank
178
THE LINGUISTIC AND ECONOMIC ADJUSTMENT OF SOVIET JEWISH IMMIGRANTS IN THE UNITED STATES, 1980–2000$ Barry R. Chiswick and Michael Wenz ABSTRACT This paper is an analysis of the English-language proficiency and labor market earnings of adult male Soviet Jewish immigrants to the United States from 1965 to 2000, using the 2000 Census of Population. Comparisons are made to similar analyses using the 1980 and 1990 Censuses. A consistent finding is that recently arrived Soviet Jewish immigrants have lower levels of English proficiency and earnings than other immigrants, other variables being the same. However, they have a steeper improvement in both proficiency and earnings with duration in the United States and the differences from the other European immigrants disappear after a few years. The Soviet Jewish immigrants have both a higher level
$
Earlier versions of this paper were presented at the Conference on Soviet and Post-Soviet Jewry, Hebrew University, Jerusalem, December 28–30, 2003, the Conference on Immigration, Minorities, and Social Exclusion, Bar-Ilan University, June 27–28, 2004, and the Fourteenth World Congress of Jewish Studies, Jerusalem, August 2005.
Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 179–216 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24006-2
179
180
BARRY R. CHISWICK AND MICHAEL WENZ
of schooling and a larger effect of schooling on earnings than other immigrants, even other European immigrants. The lower initial English proficiency and earnings, the steeper improvement with duration, and the rapid attainment of parity is consistent with the ‘‘refugee’’ nature of their migration, as distinct from being purely economic migrants. That the same pattern exists across three censuses suggests that the low English proficiency and earnings of those recently arrived in the 2000 Census data reflects a refugee assimilation process, and not a decline in the unmeasured dimensions of the earnings potential of recent cohorts of Soviet Jewish immigrants. The very high level of schooling and the larger effect of schooling on earnings among Soviet Jewish immigrants are similar to the patterns found among Jews born in the United States. Soviet Jewish immigrants appear to have made a very successful linguistic and labor market adjustment, regardless of their period of entry into the United States.
1. INTRODUCTION This study constitutes an extension of earlier work by one of the authors on the economic status of turn-of-the-20th century Russian Jewish immigrants, as well as work on Soviet Jewish immigrants to the United States in the late 20th century (Chiswick, 1991, 1992, 1993, 1997, 1999). The specific purpose of this paper is to continue this line of research on the linguistic and labor market adaptation of adult male Soviet Jewish immigrants in the United States in the post-1965 period.1 Linguistic adaptation, that is, the acquisition of English language proficiency, is important for many reasons, including increasing access to US schooling and job training and success in the labor market, whether measured by employment or earnings. Moreover, it is important for acquiring US citizenship and thereby expanding job opportunities and increasing political influence. Labor market success is an important element in a family’s economic well being and determines current consumption, as well as having an influence on marital formation and stability, fertility, and parental investments in the human capital of their children. The data under study are from the 2000 Census of Population of the United States, Public Use Microdata Sample (Census, 2003), 5 percent random sample of the population, as well as comparable data from the 1980 and 1990 Censuses.2
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
181
2. MIGRATION FROM THE FORMER SOVIET UNION 2.1. The Extent of Migration With the impending and actual collapse of the Soviet Union in 1989 a massive exodus began of the Jewish population. Between 1989 and 2003, 1.6 million Jews and their non-Jewish relatives left the former Soviet Union (FSU), 200,000 each in 1990 and 1991 alone, with the numbers declining thereafter to only 35,000 in 2003 (Tolts, 2004a,b). The primary destination was, of course, Israel, which received over 950,000, or 61 percent of the emigrants. The emigration data suggest that about 315,000 Jews and their non-Jewish relatives left the FSU for the United States, or about 20 percent of the emigrants. Another 160,000 (10 percent) went to Germany and about 20,000 went to Canada, with the remainder settling in a wide range of destinations.3 From the start of official record keeping in the United States in 1820, to the present, approximately, 4.0 million people are recorded as having immigrated (permanent resident aliens) to the United States from the Russian Empire or the FSU (Table 1). The peak decade was 1901–1910 when 1.6 million immigrants were recorded, followed by 1911–1920 with 0.9 million immigrants (Table 2). Immigration from the Soviet Union declined sharply thereafter, with less than 600 recorded in the 1940s, rising to nearly 700 in the 1950s, 2,500 in the 1960s, 39,000 in the 1970s, 58,000 in the 1980s, and nearly 463,000 in the 1990s (1991–2000), for a total of 560,000 over the period 1965–2000. Because of these trends, the analysis is limited to those who first came to the United States to stay in 1965 or later. The 2000 Census suggests that there were about 700,000 people living in the United States who were born in the former Soviet Union. They may have entered with permanent resident alien visas or under other visas and provisions of immigration law, and some of these subsequently became permanent resident aliens. A large proportion entered as refugees or asylees (Table 2). 2.2. The Refugee Experience Many who sought to leave the Soviet Union would not have had an incentive to leave if not for the anti-smitism and generalized repression. Many were motivated, at least in part, by these factors and not simply conventional economic incentives. There had been a pent up demand for
182
Table 1. Time Period
BARRY R. CHISWICK AND MICHAEL WENZ
Immigration to the United States from Russia and the Soviet Union, 1820–2002a. Number of Immigrants
1820–1830 1831–1840 1841–1850 1851–1860 1861–1870 1871–1880 1881–1890 1891–1900 1901–1910 1911–1920 1921–1930 1931–1940 1941–1950 1951–1960 1961–1970 1971–1980 1981–1990 1991–2000 2001 2002
89 277 551 457 2,512 39,284 213,282 505,290 1,597,306 921,201 61,742 1,370 571 671 2,465 38,961 57,677 462,874 55,099 55,464
Total
4,017,143
Source: US Department of Justice, 1993 Statistical Yearbook of the Immigration and Naturalization Service, Washington, DC., September 1994; US Department of Justice, 2001 Statistical Yearbook of the Immigration and Naturalization Service, Washington, DC., February 2003; and US Department of Homeland Security, 2002 Yearbook of Immigration Statistics, Washington, DC., October 2003. a Individuals granted permanent resident alien status. Includes all constituent units of the Russian Empire and of the FSU.
emigration from the Soviet Union, but there had been little expectation that it could be realized. Most emigrants had a limited ability to prepare for the move because of the seemingly arbitrary nature of the Soviet bureaucracy and the apparent randomness as to whose application for an exit visa would be approved, or when it would be approved. Many who sought to leave before the collapse of the Soviet Union experienced various degrees of reprisals and persecution, including loss of their jobs and imprisonment or internal exile. The unexpected and sudden opening for emigration with the collapse of the Soviet Union was accompanied by fears that the door could close at any time accompanied by a resurgence of anti-semitism and
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
Table 2. Year 1961–69 1970 1971 1972 1973 1974 1975 1976 TQ 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Total
183
Soviet Refugee and Asylee Arrivals and Admissions, FY 1961– 2002. Dept of Justicea
Dept of Stateb
456 209 88 228 591 2,221 3,209 5,882 1,208 5,296 9,931 27,135 28,692 11,244 2,838 1,449 791 674 833 3,728 18,880 39,831 53,130 57,587 66,026 51,983 NA NA NA NA NA NA NA NA NA
8,191 10,688 24,449 28,444 13,444 2,756 1,409 715 640 787 3,694 20,421 39,553 50,716 38,661 61,298 48,627 43,470 35,716 29,536 27,072 23,349 17,220 15,103 15,749 23,150
394,140
598,519
6,211 7,450
Source: US Department of Justice, 2001 Statistical Yearbook of the Immigration and Naturalization Service, Washington, DC, February 2003, Table 24. US Department of Homeland Security, Yearbook of Immigration Statistics, 2002, Washington, DC, October, 2003. a Soviet refugee and asylee approvals, fiscal year 1961–1993. TQ1976 means transition quarter when fiscal year was adjusted to start October 1 rather than July 1. b Refugee admissions from the Soviet Union, 1976–2002, including all republics from the FSU.
184
BARRY R. CHISWICK AND MICHAEL WENZ
repression. Thus, the Soviet Jewish migrants to the United States are more appropriately characterized as refugees than as economic migrants. Refugees have a different adjustment in the destination than do economic migrants (Chiswick, 1978, 1979). They have more skills specific to the origin and fewer skills that are destination specific or internationally transferable. As a result, at arrival they would be expected to have lower levels of human capital specific to the destination, including language skills, and hence lower earnings than economic migrants with similar measured characteristics. As they make implicit and explicit investments in the destination to increase the transferability of previously acquired skills and to create new skills, it would be expected that they would exhibit a more rapid improvement in language skills and earnings than economic migrants. Yet, because refugees are likely to be less favorably selected for economic success in the destination than otherwise similar economic migrants, it would be expected that the gap between them and economic migrants would narrow, but never close (Chiswick, 2000). Moreover, because of the lesser degree of the transferability of the skills acquired in the origin in school and on the job (labor market experience) among refugees, the effects of these variables on their earnings in the US would differ from that of economic migrants. In particular, refugees would be expected to have a smaller effect of schooling and pre-migration experience on earnings than would be the case for economic migrants. While Soviet Jewish immigrants would reflect these refugee characteristics, these might be offset by the different labor market characteristics that have been exhibited by Jews in the US, whether immigrants or native born. American Jews have had high rates of occupation and earnings mobility, have a larger effect of schooling on earnings, and have obtained higher earnings, compared with observationally similar non-Jews (Chiswick, 1999). As a result the linguistic and labor market progress of Soviet Jewish immigrants in the United States, in comparison to other (non-Jewish) economic migrants would be expected to reflect both their refugee and Jewish experiences and backgrounds.
3. WHO IS A SOVIET JEW? The first step in an analysis of ‘‘Soviet Jews’’ in the United States is to define each of the two terms. For the purpose of this study, persons born in any of the constituent republics of the FSU are referred to as ‘‘Soviet immigrants’’.
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
185
Thus, the analysis is not to be limited to those born in ‘‘Russia’’ loosely defined or in the Russian Federation. Defining Jews is more problematic. The Census of the United States, unlike censuses in some other countries, such as Australia, Canada, and Israel, has never asked religion. In the 2000 Census microdata file anyone who responds to the question on ethnic ancestry by revealing a religion is assigned the same ancestry code (998) as all other religious responses. Any response indicating Jewishness, even if the response is ‘‘secular Jew’’, is combined with and thereby masked with other religious responses. Yet, clearly, not all respondents from the FSU are Jews. Those who report an Armenian ancestry or who report that they speak Armenian or Ukrainian at home are not likely to be Jewish. Thus, for a first approximation for the purposes of this paper, persons born in the FSU who do not report an Armenian ancestry, or Armenian or Ukrainian as a language spoken at home are the subject of this analysis and for simplicity of exposition are considered ‘‘Soviet Jews’’.4 (Chiswick, 1993, 1997). This study is limited to the analysis of adult (aged 25–64) males. For younger and older persons school enrollment and retirement decisions have a major impact on labor supply and choice of jobs, and hence earnings. Similarly, the labor market attachment of women is strongly influenced by marital status and child care responsibilities. Analyses of these labor supply decisions are beyond the scope of this study.
4. DESCRIPTIVE STATISTICS Table 3 reports the means and standard deviations of selected variables relevant for the analysis. The Soviet Jewish immigrants, as defined here, are less proficient in English than either European or Asian immigrants. Among the Soviet Jews, 73 percent reported that they speak only English at home or speak another language, but speak English ‘‘very well’’ or ‘‘well’’ (Tables 3 and 4). Twenty-seven percent reported that they spoke English ‘‘not well’’ or ‘‘not at all’’. In contrast, 89 percent of the European immigrants and 82 percent of the Asian immigrants satisfy this definition of English proficiency. Among those with earnings, the Soviet Jews earned nearly $37,600 in 1999, considerably less than the earnings of other European ($50,900) and Asian ($42,400) immigrants, but substantially more than Latin American immigrants ($23,000). The Soviet Jewish immigrants have some characteristics that would enhance their language proficiency and earnings potential, but other
186
Table 3.
BARRY R. CHISWICK AND MICHAEL WENZ
Selected Characteristics of Adult Males Who Immigrated Since 1965 by Region of Birth, 2000.
Variablea
FSUb
Europe (Excluding FSU)
42.4 (10.8) 14.8 (3.2) 37,555 (48,691) 9.35 (2.30) 46.5 (11.1)
42.4 (10.4) 13.6 (3.8) 50,889 (61,390) 9.95 (1.98) 47.6 (9.9)
Asia
Latin America
Totalc
(A) Means and S. D.a Age Education (years) Earnings ($) Log of earnings Weeks worked
41.1 38.3 39.6 (10.3) (9.6) (10.0) 14.1 9.3 11.5 (4.0) (4.7) (4.9) 42,370 22,966 32,704 (53,979) (29,191) (45,018) 9.65 9.19 9.43 (2.08) (1.90) (1.99) 46.8 45.3 46.1 (10.7) (11.6) (11.2)
(B) Percents Period of Immigration 1995–2000 1990–1994 1985–1989 1980–1984 1975–1979 1970–1974 1965–1969 Total Married Speaks Englishd With children at home Rural (non-metropolitan area) residence Southern states Unemployede Sample size
31.2 38.3 12.7 5.9 9.6 1.7 0.7 100.0 73.8 72.7 50.6 0.5
22.6 14.5 13.2 11.9 11.0 12.8 14.1 100.0 68.9 88.8 44.3 0.8
19.1 18.3 17.0 18.4 15.1 8.0 4.2 100.0 67.4 81.9 54.3 0.6
17.8 17.0 21.2 17.4 11.9 9.4 5.3 100.0 56.3 57.3 67.6 1.6
19.4 17.5 18.8 16.8 12.6 9.1 5.8 100.0 61.1 69.6 60.2 1.2
10.1 4.3 9,384
20.5 2.8 42,911
19.9 3.1 1,25,487
32.6 4.9 2,50,828
27.6 4.1 4,51,844
Source: 2000 Census of Population, Public Use Microdata Sample, 5 percent sample. a Mean values. Standard Deviations within parentheses. Percents with specific characteristics. b FSU excludes persons of Armenian ancestry or who speak Armenian or Ukrainian at home. c Total includes groups not shown separately (23,234 observations), primarily from Canada and Oceania. d Speaks only English at home or speaks another language but speaks English very well or well. e Unemployed as a percent of the labor force.
187
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
characteristics that would have a negative impact. Their educational level is very high, an average of 14.8 years of schooling, far greater than even the 14.1 years among Asian immigrants, the 13.6 years among other European immigrants, or the 11.5 years among all (including Soviet) immigrants. On the other hand, the Soviet immigrants had a very short period of residence in the US In 2000, among those who immigrated in 1965 or later, 70 percent of the Soviet Jews had been in the US 10 or fewer years, in contrast to 37 percent overall. The two measures of employment tell a similar story. Among those who worked, the weeks worked in 1999 were lower for Soviet Jews (46.5 weeks) than for European (47.6 weeks) or Asian (46.8 weeks) immigrants, although greater than among Latin American immigrants (45.3 weeks). Among those in the labor force in the reference week, the last week in March 2000, 4.3 percent of the Soviet Jewish immigrants were unemployed, in contrast to 2.8 percent and 3.1 percent for European and Asian immigrants, respectively. Table 4 provides greater detail on the English language proficiency of immigrants. The Soviet Jews are least likely to speak only English at home (4.5 percent compared to 13.4 percent for all immigrants) and are more likely (26.8 percent) than European and Asian immigrants to report that they speak English ‘‘not well’’ or ‘‘not at all (11.2 and 17.9 percent, respectively). Only
Table 4. Fluency in English Among Adult Male Immigrants Who Immigrated Since 1965 by Region of Origin (percent)a. English Fluency Speaks only English at home
FSUb Europe (Excluding FSU) 4.5
32.3
Asia
Latin America
All
7.4
10.7
13.4
Speaks another language at home and speaks English: Very well Well Not well Not at all Total Sample size
30.1 37.8 22.5 4.3
36.6 19.9 9.6 1.6
45.3 29.4 15.4 2.5
22.0 24.6 28.2 14.5
31.1 25.1 21.4 9.0
100.0 8,373
100.0 42,590
100.0 1,24,735
100.0 2,50,826
100.0 4,51,844
Source: 2000 Census of Population, Public Use Microdata Sample, 5 percent sample. Note: Detail may not add to total due to rounding. a All immigrants include groups not shown separately. b FSU excludes persons of Armenian ancestry and persons who speak Armenian or Ukrainian at home.
188
BARRY R. CHISWICK AND MICHAEL WENZ
the Latin American immigrants have a greater proportion (42.7 percent) in these two least proficient categories. Appendix Tables A1–A3 report the ethnic ancestry, language spoken at home if it is not exclusively English and the republic of birth for the sample of Soviet Jews under study by sub-period of immigration to the US since 1965. There appears to be relatively little variation in these characteristics across the sub-periods.5
5. METHODOLOGY FOR THE STATISTICAL ANALYSIS A multivariate statistical analysis (ordinary least-squares regression analysis, OLS) is used to compare Soviet Jewish immigrants to other immigrants, when other measured variables are held constant. That is, controlling for factors such as age, schooling, marital status, and duration in the United States, do Soviet Jews differ in English language proficiency and earnings from other immigrants?6 The statistical analysis uses the adult (aged 25–64) male respondents in the 2000 Census Public Use Microdata Sample, 5 percent sample of the population, as the unit of observation. The means and standard deviations for the dependent and explanatory variables are reported in Table 3. Language skills are measured by a dichotomous variable defined to equal unity for those who speak only English at home or if they speak another language they speak English ‘‘very well’’ or ‘‘well’’. It is zero for those who speak English ‘‘not well’’ or ‘‘not at all’’ (see Table 4). The earnings variable is the natural logarithm of annual earnings in 1999, where earnings are the sum of wage, salary, and self-employment income. Those who reported zero earnings or did not work in 1999 are deleted from the analysis. Those who reported earnings of less than $100, including the negligible number reporting negative earnings, were assigned a value of $100 since the natural logarithm is not defined for zero or negative values.7 The econometric model for the analysis of language proficiency is based on earlier research that specifies three fundamental concepts (Chiswick & Miller, 1998). These are exposure to the destination language, efficiency in destination language acquisition, and economic incentives for learning the destination language. In the empirical application the measurable variables reflecting these concepts include two continuous variables, years of schooling and years of age, and a set of dichotomous variables. The dichotomous variables include marital status (whether married, with spouse present),
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
189
whether there are children under age 18 currently living in the household, and whether the respondent lives in a rural area or a southern state (the swath of 17 states from Texas to the Atlantic Ocean, from Maryland to Florida, including Washington, DC).8 The Census asks, when did this person come to the United States to stay? The census does not ask the type of visa used to enter the United States or whether permanent resident status was obtained. Given that many Soviet Jews entered the United States as asylees only to become permanent resident aliens (immigrants) at a later date, the census question is more appropriate for this analysis than would be the year the respondent obtained permanent resident alien or immigrant status. Since few Soviet Jews subsequently left the United States to return to the FSU or go to a third country, such as Israel, the emigration from the United States of Soviet immigrants does not pose a selectivity problem (Ahmed & Robinson, 1994; Mulder, 2003).9 Variables for duration in the United States are central to the analysis and they are entered as period of arrival dichotomous variables.10 This specification was chosen to permit a finer determination of non-linearities than would a quadratic specification of a continuous duration variable. Moreover, it increases comparability with earlier research on Soviet Jews in the United States. When duration is held constant, the age variable reflects the effect of age at migration on English language proficiency. Another key variable is country of birth. A person born in any of the republics that constituted the FSU (other than those who reported Armenian ancestry or language or the Ukrainian language) is considered to be a Soviet Jewish immigrant (FSU).11 Data are not available on when the person left the FSU or on country of last permanent residence, so it is not possible to identify whether there was a destination prior to coming to the US In this analysis, the country categories Europe and Asia constitute all of Europe and Asia, other than the designated parts of the FSU. Other countries of origin groups are Canada, Latin America (including the Caribbean), and other countries (Africa, Oceania, etc.). Europe other than the FSU serves as the benchmark. The econometric analysis of earnings is based on the human capital earnings function, modified for immigrant adjustment (Chiswick, 1978). The natural logarithm of annual earnings in 1999 is regressed on years of schooling completed, years of potential labor market experience (age minus schooling minus 5 years), and its square, the natural logarithm of weeks worked, and dichotomous variables as defined above for being proficient in English, married spouse present, living in a rural area and living in a southern state. The same dichotomous variables are used, as defined above, for
190
BARRY R. CHISWICK AND MICHAEL WENZ
period of arrival and country of origin. Controlling for period of arrival, the labor market experience variable measures the effect on earnings in 1999 of experience in the country of origin.
6. ECONOMETRIC ANALYSIS 6.1. Language – Soviet and Other Immigrants The results of the multiple regression analysis for adult males for the dependent variable, proficient in English, are reported in Tables 5 and 6. The variable is unity for those who speak only English at home or who speak another language but speak English very well or well, otherwise the English fluency variable is zero. Table 5 reports the equation for all immigrants by sub-period and for the whole period 1965–2000. As shown in column (1), consistent with what has been found elsewhere for immigrants, English language proficiency increases with years of schooling (3.5 percentage points more are proficient for each extra year of schooling). Proficiency is lower for those who immigrated at an older age. Five years older at immigration is equivalent to about one fewer year of schooling. Men who are married are more proficient (by 4 percentage points), but children at home detracts from their proficiency (by 2.6 percentage points per child). Duration in the United States has a major impact on English language proficiency. The coefficients are highly statistically significant and show a consistent gradient of increased proficiency with duration in the US, with the effect of an extra year in the United States becoming smaller the longer the duration of residence. With those who immigrated in 1980–1984 as the benchmark, other variables the same, the most recent immigrants (1996– 2000) were 24 percentage points less proficient in 2000, or the equivalent of the effect seven years of schooling. The earliest cohort, 1965–1969 immigrants, was 12 percentage points more proficient than the 1996–2000 cohort or the equivalent of 3.5 years of schooling. Other variables the same, Soviet immigrants are about 10.4 percentage points less likely to be proficient in English than other European immigrants. They are even less proficient than Asian immigrants (Asians are at a 7.7 percentage points disadvantage compared to European immigrants), but less disadvantaged than those from Latin America (17.3 percentage point differential compared to European immigrants).
Regression Analysis of Fluency in English among Adult Males Who Immigrated Since 1965: 2000. Dependent Variable ¼ ZENGSPK 2000 Census
Immigration Period Variable CONSTANT EDUCYRS AGE IM95_00 IM90_94 IM85_89 IM75_79 IM70_74 IM65_69 IM95_00FSUJEW IM90_94FSUJEW
(1)
(2)
0.7238 0.7241 (169.12) (169.07) 0.0349 0.0349 (255.53) (255.37) 0.0067 0.0067 ( 99.89) ( 100.05) 0.2405 0.2380 ( 120.21) ( 118.47) 0.1396 0.1415 ( 68.98) ( 69.46) 0.0631 0.0637 ( 31.99) ( 32.24) 0.0573 0.0571 (26.28) (26.08) 0.0958 0.0959 (39.38) (39.38) 0.1230 0.1232 (42.54) (42.57) 0.0908 ( 7.64) 0.0502 (4.46) 0.0574 (3.70)
1965–1979 (1) 0.7980 (114.50) 0.0335 (150.77) 0.0068 ( 61.96)
1980–1989 (2)
0.7980 (114.50) 0.0335 (150.74) 0.0068 ( 61.94)
(1)
1990–2000
(2)
(1)
0.7618 .07615 0.6048 (94.72) (94.69) (89.67) 0.0363 0.0363 0.0342 (147.89) (147.87) (139.85) 0.0079 0.0079 0.0056 ( 62.49) ( 62.50) ( 49.66) 0.1041 ( 50.05) a
(2) 0.6063 (89.89) 0.0342 (139.83) 0.0057 ( 49.82) 0.1006 ( 47.84) a
0.0614 0.0617 ( 28.54) ( 28.60) 0.0382 ( 17.28) a
0.0381 (14.19)
0.0381 ( 17.20) a
0.0382 (14.21) 0.1118 ( 10.15)
0.0494 (1.89)
191
IM85_89FSUJEW
1965–2000
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
Table 5.
192
Table 5. (Continued ) Dependent Variable ¼ ZENGSPK 2000 Census Immigration Period
1965–2000
Variable
(1)
IM75_79FSUJEW IM70_74FSUJEW
MARRSP RURAL SOUTH CHILD ( FSU ( ASIA ( LATAMER ( CANADA OTHER
0.0409 (29.96) 0.0082 (1.55) 0.0081 (6.08) 0.0257 19.19) 0.1043 23.85) 0.0766 35.26) 0.1725 79.66) 0.0705 (14.45) 0.0584 (16.59)
0.0182 (1.10) 0.0021 ( 0.06) 0.0066 (0.12) 0.0410 (30.09) 0.0082 (1.56) 0.0080 (6.02) 0.0255 ( 19.05) 0.1050 ( 12.44) 0.0764 ( 34.58) 0.1723 ( 78.95) 0.0704 (14.42) 0.0584 (16.53)
(1)
1980–1989 (2)
(1)
1990–2000
(2)
(1)
(2)
0.0537 (23.08) .0075 (0.86) 0.0041 (1.83) 0.0400 17.64) 0.1733 29.65) 0.0804 21.72) 0.2562 66.94) 0.1263 (15.99) 0.0882 (15.62)
0.0539 (23.15) 0.0075 (0.85) 0.0041 (1.80) 0.0398 17.52) 0.1354 19.53) 0.0839 22.58) 0.2588 67.49) 0.1234 (15.62) 0.0854 (15.11)
0.0202 ( 0.89) a
0.0220 (9.53) 0.0139 (1.52) 0.0094 (4.25) 0.0086 ( 3.94) 0.0276 ( 2.69) 0.0658 ( 20.19) 0.0911 ( 29.53) 0.0015 ( 0.21) 0.0052 ( 0.86)
0.0436 ( 0.87) 0.0220 (9.53) 0.0139 (1.52) 0.0094 (4.25) 0.0086 ( 3.94) 0.0118 ( 0.61) 0.0660 ( 20.19) 0.0913 ( 29.54) 0.0017 ( 0.23) 0.0054 ( 0.89)
0.0264 0.0264 (10.53) (10.53) 0.0142 0.0142 (1.47) (1.48) 0.0166 0.0166 (6.83) (6.82) 0.0201 0.0201 ( 7.83) ( 7.84) 0.0390 0.0555 ( 3.07) ( 3.61) 0.0907 0.0902 ( 20.10) ( 19.94) 0.1622 0.1618 ( 35.95) ( 35.81) 0.0397 0.0401 (3.57) (3.61) 0.0195 0.0199 (2.79) (2.85)
( ( ( (
( ( ( (
BARRY R. CHISWICK AND MICHAEL WENZ
IM65_69FSUJEW
(2)
1965–1979
4,51,843 4,51,843 1,24,512 0.3925 0.3924 0.3344 0.2723 0.2722
0.2726 0.2725
0.2586 0.2585
1,24,512 1,40,887 1,40,887 1,66,684 0.3344 0.3983 0.3983 0.4178 0.2586 0.2585
0.2299 0.2298
Note: t-ratios in parentheses. Source: 2000 US Census of Population, Public Use Microdata Sample, 5 percent Sample. a Omitted as benchmark; 1980–1984 and Europe are benchmarks unless otherwise noted.
0.2299 0.2298
0.2739 0.2739
1,66,684 0.4176 0.2744 0.2743
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
SAMPLE SIZE STANDARD ERROR R2 ADJUSTED R2
193
194
Table 6.
BARRY R. CHISWICK AND MICHAEL WENZ
Regression Analysis of English Fluency among Adult Males Who Immigrated Since 1965: 2000, 1990, 1980.
Dependent Variable ¼ ENGSPK Immigration Period 2000 Census 1965–2000 Variable CONSTANT EDUCYRS AGE IM96_00 IM91_95 IM87_90 IM85_86 IM75_79 IM70_74 IM65_69
(1) 0.7177 (74.75) 0.0347 (113.15) 0.0066 ( 44.38) 0.2492 ( 52.88) 0.1626 ( 35.26) 0.0876 ( 18.99) 0.0347 ( 6.07) 0.0544 (11.07) 0.1069 (19.64) 0.1222 (19.04)
IM96_00FSUJEW IM91_95FSUJEW IM87_90FSUJEW IM85_86FSUJEW
IM75_79FSUJEW IM70_74FSUJEW IM65_69FSUJEW MARRSP RURAL SOUTH CHILD
0.0397 (12.98) 0.0188 (1.58) 0.0083 (2.79) 0.0149 ( 4.98)
(2) 0.7188 (74.87) 0.0347 (113.12) 0.0067 ( 44.49) 0.2443 ( 51.50) 0.1639 ( 35.27) 0.0891 ( 19.23) 0.0345 ( 6.04) 0.0546 (11.06) 0.1073 (19.68) 0.1222 (19.02) 0.1580 ( 6.36) 0.0147 ( 0.53) 0.03281 (1.00) 0.0036 (0.04) 0.0386 ( 0.99) 0.0421 ( 0.58) 0.0514 (0.57) 0.0399 (13.05) 0.0189 (1.59) 0.0083 (2.79) 0.0147 ( 4.88)
1990 Census 1965–1989 (1)
(2)
1980 Census 1965–1979 (1)
0.6031 (114.99) 0.0356 (192.66) 0.0067 ( 73.07)
0.6027 (114.94) 0.0355 (192.59) 0.0067 ( 73.30)
0.5243 (41.60) 0.0388 (89.92) 0.0050 ( 20.26)
0.1387 ( 51.87) 0.0698 ( 23.07) 0.0796 (33.12) 0.1312 (50.16) 0.1690 (58.01)
0.1341 ( 49.77) 0.0693 ( 22.85) 0.0789 (32.64) 0.1316 (50.21) 0.1695 (58.13)
0.0956 ( 19.20)
0.0574 (11.18)
0.2527 ( 10.31) 0.0456 ( 0.85)
0.0404 (20.53) 0.0177 (4.55) 0.0174 (8.99) 0.0267 ( 14.16)
0.0249 ( 1.00) 0.0346 ( 0.87) 0.0263 (0.45) 0.0409 (20.78) 0.0177 (4.55) 0.0175 (9.03) 0.260 ( 13.80)
0.0134 (2.17) 0.0102 (1.16) 0.0030 (0.60) 0.0093 (1.98)
195
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
Table 6. (Continued ) Dependent Variable ¼ ENGSPK Immigration Period 2000 Census 1965–2000 Variable
(1)
(2)
1990 Census 1965–1989 (1)
(2)
FORMARa FSU ASIA LATAMER CANADA OTHER Sample size Standard error R2 Adjusted R2
0.0979 0.0492 0.1374 0.0273 ( 10.00) ( 2.18) ( 16.05) ( 1.31) 0.0739 0.0754 0.0632 0.0631 ( 15.05) ( 15.20) ( 22.43) ( 22.43) 0.1768 0.1777 0.1514 0.1514 ( 36.33) ( 36.38) ( 54.87) ( 54.88) 0.0825 0.0812 0.0739 0.0739 (7.73) (7.60) (11.05) (11.06) 0.0621 0.0601 0.0228 0.0227 (7.91) (7.73) ( 6.10) ( 6.08) 90,383 90,383 2,27,554 2,27,554 0.39307 0.3929 0.3879 0.3877 0.2711 0.2718 0.2649 0.2656 0.2710 0.2716 0.2649 0.2655
1980 Census 1965–1979 (1) 0.0431 (8.32) 0.1384 ( 7.47) 0.0431 ( 6.97) 0.1445 ( 25.13) 0.1265 (9.26) 0.0202 (2.41) 35,915 0.3790 0.3047 0.3044
Note: I ratios in parentheses. Source: 2000 Census of Population, Public Use Microdata Sample, 5 percent Sample; 1990 Census of Population, Public Use Microdata Sample, 5 percent Sample; 1980 Census of Population, Public Use Sample, B and C Sample Files Combined, 2 percent Sample. a Variable cannot be reconstructed for 1990, 2000 Census. Omitted as benchmark; benchmark is 1980–1984 and Europe unless otherwise noted.
It is possible to test whether the effect of duration in the US on proficiency in English differs between Soviet and other immigrants. The statistical analysis (Table 5, column 2) shows that during the first four years the negative effect on proficiency of being an immigrant is much greater for Soviet Jews than it is for other immigrants. Compared to other recent European immigrants, Soviet Jews who arrived in 1995–2000 are 20 percentage points less proficient ( 0:1050 0:0908 ¼ 0:20). Soviet Jews experience a steeper improvement in proficiency with duration in the US so that the disadvantage is only 5.5 percentage points 0:1050 þ 0:0502 ¼ 0:055) for those who immigrated in 1990–1994 (6–10 years in the US), and 4.8 percentage points ( 0:1050 þ 0:0574 ¼ 0:048) for those who immigrated 1985–1989 (11–15 years in the US). Indeed, the very large proportion of Soviet immigrants in the US a short period of time and the very low English proficiency of this group are very important determinants of the overall low proficiency among Soviet immigrants.
196
BARRY R. CHISWICK AND MICHAEL WENZ
The analysis was also performed for sub-periods within the 1965–2000 period (Table 5). For each of these sub-periods the effects of schooling, age at immigration, marital status, and children are quite similar.12 That is, their partial effects on proficiency in 2000 do not appear to vary by period of immigration. The effects of duration do vary by period of immigration. One fewer year in the US has a larger negative effect on proficiency the more recently the immigrant cohort arrived in the US, which is consistent with the non-linear effect of duration on proficiency. The results reported here for the 2000 Census can be compared with analyses reported previously for Soviet Jews and other immigrants who came to the US in 1965 or later using the microdata files from the 1980 and 1990 Censuses (Chiswick, 1993, 1997) (see Table 6). The effects on English language proficiency of schooling, age, marital status, and rural residence are virtually identical across the three censuses, although the positive effect of being married was much smaller in the 1980 Census and the positive effects of living in the South is smaller in 2000 than in 1990.13 The negative effects of children in the household are also smaller in absolute value in 2000 than in 1990, but it was not significant in 1980. The strong positive effect of duration in the US on proficiency is also observed in these earlier censuses. The 10 percentage point disadvantage of being from the FSU compared to another part of Europe in the 2000 data is somewhat smaller than the 14 percentage points in the 1980 and 1990 Censuses. When the interaction terms of Soviet origin with duration are added, the Soviet intercept is a highly significant—5 percentage points in 2000, compared to a nonsignificant+3 percentage points in 1990. The negative effect of being in a particular immigrant cohort compared to an earlier arrival cohort diminishes from the 1980 to the 2000 Census as the cohorts are in the US a longer period of time. Most striking is that in 1990, the only Soviet-duration of residence interaction term whose coefficient was large or statistically different from the benchmark (1980–1984) was the most recent cohort, 1987–1990 (coefficient of—25 percentage points). Ten years later, compared to the same benchmark, the 1987–1990 interaction term has a coefficient of only 3 percentage points and it is not statistically significant. These results suggest that the sharp gradient of English language proficiency with duration in the US is not a consequence of declining proficiency among more recent cohorts. Rather it appears to be reflecting a longitudinal or adjustment effect, that is, the acquisition of English language proficiency as a cohort has a longer duration in the US Moreover, this initial deficiency and speed of adjustment (improvement) appear to be more intense for
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
197
Soviet Jews than for other immigrants. This may reflect their refugee motivated migration, the limited ability to prepare for the emigration because of the arbitrary nature of the Soviet bureaucracy, and the unexpected and sudden opening for emigration from the Soviet Union, with uncertainty as to how long emigration would be possible.
6.2. Earnings – Soviet and Other Immigrants The analysis of earnings (Table 7, column 1) indicates that an extra year of schooling raises the earnings of immigrants by about 4.6 percent, that earnings increase at a decreasing rate with an increase in total labor market experience, that earnings rise by about 0.85 percent for each one percent increase in weeks worked (about one half of a week), and that earnings increase with duration of residence in the US Indeed, compared to those who immigrated in 1980–1984, those who recently arrived (immigrated 1996–2000) have about 16 percent lower weekly earnings, while those who immigrated in 1965–1969 had about 11 percent higher weekly earnings. The effects of country of origin are quite large. Compared to European immigrants, those from the Soviet Union had weekly earnings that were nearly 20 percent lower, other measured variables being the same. Only Latin American immigrants had a larger earnings disadvantage (about 32 percent) compared to those from Europe, while Canadian immigrants showed a large earnings advantage over Europeans (about 13 percent). Other factors that resulted in higher earnings are being proficient in English (about 17 percent), being married (21 percent), living in an urban area (8 percent), and living outside the south (3 percent). Other variables the same, as shown in Table 7, column 2, an extra year of schooling is associated with 7.0 percent higher earnings for the Soviet Jewish immigrants, in contrast to the 4.6 percent for other immigrants, and the difference is highly statistically significant (t ¼ 11:0). Also, other things the same, the earnings of Soviet Jewish immigrants are much lower (and the difference is highly significant) than those of other immigrants who came in the same time period during the first few years in the US (immigrated 1996–2000 or 1991–1995). The magnitude diminishes but does not disappear for those who have been in the United States for 10 or more years in 2000. Thus, the earnings gap between Soviet and other immigrants varies with duration in the US and level of schooling. At the mean level of schooling of Soviet immigrants (14.8 years), those who immigrated in
Regression Analysis of Earnings among Adult Males Who Immigrated Since 1965, 2000.
Dependent Variable ¼ LNEARN2000 Census Immigration Period 1965–2000 Variable CONSTANT EDUCYRS EXP
LNWW IM95_00 IM90_94 IM85_89 IM75_79 IM70_74 IM65_69 IM95_00FSUJEW IM90_94FSUJEW IM85_89FSUJEW
6.248 (451.41) 0.0461 (130.43) 0.0108 (22.35) 0.00017 ( 19.61) 0.8461 (314.94) 0.1554 ( 35.35) 0.1041 ( 24.20) 0.0434 ( 10.46) 0.0562 (12.25) 0.0965 (18.68) 0.1184 (19.18)
(2) 6.248 (451.29) 0.0458 (129.27) 0.0108 (22.40) 0.00017 ( 19.78) 0.8459 (314.93) 0.1503 ( 34.06) 0.1027 ( 23.71) 0.0434 ( 10.44) 0.0561 (12.18) 0.0986 (19.04) 0.1214 (19.64) 0.3447 ( 9.29) 0.2299 ( 6.36) 0.1048 ( 2.49)
(1) 6.030 (214.00) 0.0547 (75.35) 0.0113 (11.34) 0.00014 ( 8.20) 0.8656 (151.51)
0.0388 ( 6.82) a
0.0270 (3.87)
1980–1989 (2) 6.030 (213.99) 0.0546 (75.03) 0.0113 (11.37) 0.00014 ( 8.26) 0.8656 (151.51)
(1)
1990–2000 (2)
6.581 (280.50) 0.0438 (77.98) 0.0127 (14.85) 0.00022 ( 14.06) 0.7757 (170.96)
6.578 (280.29) 0.0437 (77.63) 0.0127 (14.84) 0.00022 ( 14.09) 0.7757 (170.96)
0.0459 ( 11.26)
0.0457 ( 11.16)
(1) 6.143 (291.94) 0.0415 (70.28) 0.0084 (10.62) 0.00016 ( 10.21) 0.8879 (218.39) 0.0530 ( 12.56)
(2) 6.142 (291.77) 0.0412 (69.59) 0.0084 (10.50) 0.00016 ( 10.19) 0.8876 (218.35) 0.0483 ( 11.29)
0.0392 ( 6.88) a
0.0277 (3.96) 0.1395 ( 5.85)
0.0306 ( 0.68)
BARRY R. CHISWICK AND MICHAEL WENZ
EXPSQ
(1)
1965–1979
198
Table 7.
IM70_74FSUJEW IM65_69FSUJEW ENGSPK MARRSP RURAL SOUTH FSU ASIA LATAMER CANADA OTHER
0.1742 (54.05) 0.2115 (78.13) 0.0804 ( 7.10) 0.0288 ( 10.25) 0.1850 ( 19.46) 0.1673 ( 36.48) 0.3247 ( 70.91) 0.1327 ( 13.09) 0.2071 ( 27.88)
FSUEDUCYRS Sample size Standard error R2 Adjusted R2
3,98,520 0.7833 0.3580 0.3580
0.1148 ( 2.61) 0.2233 ( 2.80) 0.2416 ( 2.00) 0.1738 (53.92) 0.2121 (78.35) 0.0807 ( 7.12) 0.0292 ( 10.37) 0.3106 ( 15.30) 0.1608 ( 34.46) 0.3211 ( 69.62) 0.1371 (13.50) 0.2025 ( 27.16) 0.0244 (11.02) 3,98,520 0.7831 0.3583 0.3583
0.0101 (0.13) a
0.1371 (18.18) 0.2661 (47.16) 0.1349 ( 5.64) 0.0551 ( 9.67) 0.0309 ( 1.17) 0.0521 ( 6.23) 0.2137 ( 26.89) 0.0316 (1.76) 0.0870 ( 5.62)
1,10,840 0.8121 0.3225 0.3224
0.1297 ( 0.94) 0.1373 (18.21) 0.2659 (47.14) 0.1348 ( 5.64) 0.0552 ( 9.68) 0.1983 ( 3.17) 0.0492 ( 5.84) 0.2122 ( 26.65) 0.0333 (1.85) 0.0849 ( 5.47) 0.0129 (2.29) 1,10,840 0.8120 0.3226 0.3225
0.1643 (32.43) 0.2168 (49.57) 0.0536 ( 2.91) 0.0230 ( 5.02) 0.1178 ( 5.76) 0.2302 ( 27.39) 0.3838 ( 45.74) 0.1309 (6.31) 0.2181 ( 16.73)
1,45,315 0.7621 0.3176 0.3175
0.1645 (32.48) 0.2168 (49.58) 0.0536 ( 2.91) 0.0231 ( 5.03) 0.3136 ( 7.36) 0.2232 ( 26.24) 0.3789 ( 44.86) 0.1366 (6.57) 0.2125 (16.25) 0.0174 (5.01) 1,45,315 0.7769 0.3217 0.3213
0.2018 (40.27) 0.1699 (38.95) 0.0740 ( 4.16) 0.0158 ( 3.49) 0.2545 ( 21.26) 0.2280 ( 30.89) 0.3927 ( 51.24) 0.2070 (13.47) 0.3113 ( 27.26)
1,42,363 0.7774 0.3886 0.3885
0.2007 (40.04) 0.1707 (39.13) 0.0741 ( 4.17) 0.0162 ( 3.58) 0.3448 ( 14.39) 0.2198 ( 28.93) 0.3881 ( 50.10) 0.2131 (13.82) 0.3054 ( 26.93) 0.0112 (6.31) 1,42,363 0.7772 0.3888 0.3887
199
Note: t ratios in parentheses. Source: 2000 Census of Population, Public Use Microdata Sample, 5 percent Sample; 1990 Census of Population, Public Use Microdata Sample, 5 percent Sample; 1980 Census of Population, Public Use Sample, B and C Sample Files Combined, 2 percent Sample. Includes only immigrants who worked and had non zero earnings in 1999. a Omitted as benchmark; benchmark is 1980–1984 and Europe unless otherwise noted.
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
IM75_79FSUJEW
200
BARRY R. CHISWICK AND MICHAEL WENZ
1980–1984 (16–20 years in the US) had about 5 percent higher weekly earnings than other European immigrants (the partial effect is: 0:3106 þ ð14:8Þð0:0244Þ ¼ 0:051). The comparison of these results with the 1990 and 1980 Census analyses is striking (Chiswick, 1997) (Table 8). In 1990, the effect of schooling on earnings was larger for Soviet Jewish immigrants by 1.9 percentage points, in 1980 by 2.8 percentage points, and in 2000 by 2.6 percentage points, all of which were significantly different from zero, but not from each other. In 1990, the Soviet immigrant duration of residence interaction term for the most recent arrivals was large and highly significant compared to the benchmark (1980–1984 cohort), as was the case in 1980 (1970–1974 benchmark), but the differential shrank with duration. Although only in the US 6–10 years at the time of the 1990 Census, at the mean level of schooling for Soviet immigrants (14.9 years), the earnings of the 1980–1984 cohort of Soviet Jews was only 1 percent lower than that of other European immigrants. As in the 2000 Census, the larger return from schooling narrowed the earnings gap between Soviet Jews and other immigrants in spite of a larger initial earnings disadvantage. Among the Soviet immigrants (Table 8), the 31 percent greater earnings disadvantage of the 1987–1990 cohort compared to the 1984–1985 cohort in 1990, shrank to a marginally significant (t ¼ 1:6) 14 percent disadvantage 10 years later in 2000. This too suggests that what is being observed is an immigrant assimilation process rather than a change (deterioration) in the earnings potential of more recent cohorts. For most of the other explanatory variables their partial effects on earnings did not change by much across the three censuses. Perhaps the most dramatic change is the increase in the negative effects of living in a rural area. This may be due to the change in the definition of rural from the old census definition of rural (farm and non-farm) to only those living on a farm. Moreover, the lower initial earnings and the steeper rise in earnings with duration of residence in the US of the Soviet Jewish immigrants, compared with other immigrant groups, is a phenomenon to be expected among refugee populations. Since their motives for migrating are not strictly economic, refugees tend to be less prepared for the move, especially Soviet migrants, and to have skills that are less readily transferable to the destination. 6.3. Language and Earnings – Soviet Jewish Immigrants Parallel analyses to those reported above were performed separately for just the Soviet Jewish immigrants (Appendix Tables A5 and A6). The statistical
201
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
Table 8.
Regression Analysis of Earnings among Adult Males Who Immigrated Since 1965: 2000, 1990, 1980.
Dependent Variable ¼ LNEARN 2000 Census Immigration Period Variable CONSTANT EDUCYRS EXP EXPSQ LNWW IM96_00 IM91_95 IM87_90 IM85_86 IM75_79 IM70_74 IM65_69 IM96_00FSUJEW IM91_95FSUJEW IM87_90FSUJEW IM85_86FSUJEW IM75_79FSUJEW IM70_74FSUJEW IM65_69FSUJEW
1965–2000 (1) 6.198 (200.12) 0.04507 (57.14) 0.0102 (9.40) 0.0002 ( 8.72) 0.8679 (145.45) 0.1598 ( 15.34) 0.1229 ( 12.53) 0.0692 ( 7.13) 0.0448 ( 3.75) 0.0561 (5.42) 0.0913 (7.92) 0.1124 (8.17) 0.5036 ( 6.39) 0.3033 ( 3.97) 0.1352 ( 1.60) 0.3156 ( 1.73) 0.0911 ( 0.95) 0.0100 ( 0.06) 0.1193 ( 0.56)
(2) 6.201 (200.29) 0.0447 (56.49) 0.0103 (9.43) 0.0002 ( 8.83) 0.8675 (145.44) 0.1501 ( 14.37) 0.1205 ( 12.19) 0.0705 ( 7.23) 0.0429 ( 3.59) 0.0557 (5.360) 0.0931 (8.06) 0.1152 (8.36)
1990 Census
1980 Census
1965–1979
1980–1989
(1)
(2)
5.204 (303.00) 0.0480 (103.10) 0.0268 (42.81) 0.0004 ( 34.33) 0.9534 (270.03)
5.208 (303.01) 0.0479 (102.62) 0.0267 (42.71) 0.0004 ( 34.26) 0.9526 (269.66)
0.0949 ( 16.32) 0.0708 ( 11.46) 0.1062 (21.85) 0.1787 (33.67) 0.1996 (33.64)
0.0910 ( 15.57) 0.0698 ( 11.28) 0.1069 (21.90) 0.1797 (33.78) 0.2006 (33.76)
0.3090 ( 5.65) 0.1979 ( 1.79) 0.1458 ( 2.85) 0.1137 ( 1.40) 0.0029 (0.02)
(1) 4.360 (102.18) 0.0462 (40.89) 0.0300 (19.77) 0.0005 ( 16.88) 1.048 (114.66)
0.1345 a
( 13.12) 0.0804 (7.60)
202
BARRY R. CHISWICK AND MICHAEL WENZ
Table 8. (Continued ) Dependent Variable ¼ LNEARN 2000 Census Immigration Period Variable ENGSPK MARRSP RURAL SOUTH FSU ASIA LATAMER CANADA OTHER
1965–2000 (1) 0.1755 (24.40) 0.2022 (33.38) 0.1159 ( 4.51) 0.0386 ( 6.13) 0.1937 ( 9.11) 0.1592 ( 15.35) 0.3224 ( 31.39) 0.1760 (7.92) 0.2180 ( 13.14)
FSUEDUCYRS Sample size Standard error R2 Adjusted R2
79,582 0.7830 0.3646 0.3645
(2)
(1)
1990 Census
1980 Census
1965–1979
1980–1989 (2)
0.1743 0.1723 0.1717 (24.24) (39.38) (39.25) 0.2030 0.2093 0.2099 (33.52) (57.12) (57.26) 0.1163 0.0183 0.0186 ( 4.53) ( 2.36) ( 2.40) 0.0389 0.0925 0.0925 ( 6.19) ( 23.52) ( 23.51) 0.2899 0.1759 0.3021 ( 5.11) ( 9.36) ( 3.42) 0.1559 0.1955 0.1953 ( 14.89) ( 34.34) ( 34.32) 0.3216 0.3227 0.3231 ( 31.20) ( 57.83) ( 57.90) 0.1777 0.0936 0.0937 (8.00) (6.99) (7.00) 0.2159 0.2511 0.2511 ( 12.99) ( 32.60) ( 32.60) 0.0256 0.0194 (5.41) (3.79) 79,582 2,02,113 2,02,113 0.7827 0.7456 0.7455 0.3652 0.4267 0.4268 0.3650 0.4266 0.4268
(1) 0.1632 (14.84) 0.1718 (16.11) 0.0190 ( 1.03) 0.0312 ( 2.97) 0.0895 ( 0.66) 0.1862 ( 14.44) 0.2612 ( 21.63) 0.1375 (4.83) 0.2276 ( 13.03) 0.0280 ( 3.08) 35,915 0.7898 0.3895 0.3892
Note: t-ratios in parentheses. Source: 2000 Census of Population, Public Use Microdata Sample, 5 percent Sample; 1990 Census of Population, Public Use Microdata Sample, 5 percent Sample; 1980 Census of Population, Public Use Sample, B and C Sample Files Combined, 2 percent Sample. a Omitted as benchmark; 1980–1984 and Europe are benchmarks unless otherwise noted.
significance of many of the variables is reduced because of the much smaller sample size. Of particular interest is whether there are differences among Soviet immigrants depending on their reported ethnic ancestries. Excluding those of Armenian ancestry or language and Ukrainian language, four groups are defined, Russian (53 percent of the sample), Ukrainian (18 percent), a response that revealed a person’s religion (12 percent), and all other responses (17 percent). Those of Russian ancestry serve as the
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
203
benchmark. The coefficients and significance levels of the other variables do not change when the ethnic ancestry variables are entered into the equation. In the language analysis, other variables are the same, few differences are found in English language proficiency by ancestry (Appendix Table A5). Those of Ukrainian origin are 2 percentage points less proficient in English than those of Russian ancestry, but the difference is at the margin of being significant (t ¼ 1:7). Those of ‘‘other ancestries’’ are one percentage point less proficient than the Russians, but this is not statistically significant (t ¼ 0:7). There is no difference from those of Russian ancestry among those who gave a response indicating their religion (the coefficient indicates a 2.5 percentage point higher proficiency with t ¼ 1:8). The analysis of earnings, other variables being the same, presents a similar picture (Appendix Table A6). For the post-1965 immigrants, there is no difference in earnings between the Russian, Ukrainian, and religious revealing ancestries. Compared to the Russians, the Ukrainians had 2.1 percent lower earnings, but a t ¼ 0:7; while those who gave a religious response had 4.5 percent higher earnings, but a t ¼ 1:3: Only the heterogeneous group of ‘‘other ancestries’’ showed an earnings differential, a marginally significant (t ¼ 1:9) 5.5 percent higher earnings. The coefficient on the education variable in the earnings analysis limited to Soviet Jewish immigrants is about 7.3 percent, whether or not the Soviet ancestry variables are held constant. This is a very large coefficient for an immigrant population in the United States and is significantly greater than for other immigrants. That it does not change when ancestry is held constant suggests that it holds across the ancestry groups that in this study are used to identify Soviet Jews.
7. SUMMARY AND CONCLUSION This paper is been concerned with the English language proficiency and labor market earnings of adult (aged 25–64 years) male Soviet Jews who immigrated to the United States since 1965. The data for the empirical analysis are from the 2000 Census of Population, Public Use Microdata Sample, and is for a five percent sample of the population. Comparisons are made to earlier parallel analyses using the 1980 and 1990 Censuses. Because of the absence of direct information on who is Jewish or of Jewish ancestry, the empirical analysis is based on persons born in the FSU who are not of Armenian ancestry and do not speak Armenian or Ukrainian at home. This
204
BARRY R. CHISWICK AND MICHAEL WENZ
definition should capture most Soviet Jews but include few non-Jewish immigrants from the FSU. The Soviet Jews were less proficient in English than other European and Asian immigrants. Under the definition of proficiency used in this study, 72 percent of the Soviet Jews were proficient, compared to 89 percent for European immigrants, 82 percent for Asian immigrants, and 57 percent for those from Latin America. Their earnings (at $37,600 in 1999) were considerably less than the earnings of other European ($50,900) and Asian immigrants ($42,400), but were greater than the earnings of Latin American immigrants ($23,000). The much higher level of schooling of the Soviet immigrants would tend to enhance their English proficiency and earnings; 14.8 years for the Soviet Jews, compared to 14.1 for Asian immigrants, 13.6 years for European immigrants, and 9.3 years for Latin American immigrants. On the other hand, the refugee motivations for their move and their recency of arrival would tend to lower their English language skills and earnings. Among those who immigrated since 1965, 70 percent of the Soviet Jewish migrants were in the United States 10 or fewer years, compared to only 37 percent of those from Europe, 37 percent of the Asians, and 35 percent of the Latin Americans. Multiple regression analysis is used to examine the effects of being a Soviet Jewish immigrant compared to coming from another region, when all other measured variables are held constant. It is found that recently arrived Soviet immigrants have a lower level of English proficiency than other European immigrants, but they have a faster rate of improvement with duration in the US As a result, the difference virtually disappears for those in the United States from 16 to 20 years. The 1980 and 1990 Census data analyses show a similar pattern for recent immigrants. This appears to be a longitudinal phenomenon reflecting their refugee experience, rather than inherently poorer English proficiency that will persist among the most recent cohorts. Thus, the low level of English proficiency among Soviet immigrants is due to the low proficiency among recent arrivals and the large proportion that recently arrived. It is a temporary and not a permanent phenomenon. The analysis of earnings, other measured variables the same, also shows much lower earnings among recent Soviet Jewish immigrants, but a steeper improvement with duration in the United States. The Soviet immigrants have a much larger positive effect of schooling on earnings compared to other immigrants. An extra year of schooling raises the earnings of Soviet Jewish immigrants by about 7.3 percent, compared to only 4.6 percent for
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
205
other immigrants. As a result there is an earnings catch-up coming sooner the higher the level of schooling. Similar patterns were found in the analyses for the 1980 and 1990 Censuses. Again, this suggests that the earnings disadvantage of Soviet Jewish immigrants as a group is short-lived and is due to the low earnings of recent arrivals and the disproportionate number of recent arrivals in the 2000 Census. Analyses of English language proficiency and earnings were also performed for those classified here as Soviet Jewish immigrants by the ancestry they reported in the 2000 Census: Russian, Ukrainian, an ancestry response that reveals one’s religion, and all other ancestry responses. In the language analysis, there was essentially no difference in English proficiency, other variables the same, between those of Russian and ‘‘other ancestries’’, although those who indicated Ukrainian had slightly lower proficiency while those who indicated a religion were marginally more proficient. In the earnings analysis, other variables the same, there were no significant differences among these three groups, although the heterogeneous group of other ancestries showed a marginally significant 5 percent earnings advantage. The addition of ancestry variables to the language and earnings equations does not alter the effect of schooling. Overall, it appears that Soviet Jewish immigrants adjust very well in the United States compared to other European immigrants. Their initial disadvantages in English language skills and earnings may be due to the refugee motivations for migration.14 With the passage of time this disadvantage disappears. For earnings it disappears most rapidly for those with higher levels of schooling. This very high level of schooling and the greater effect of schooling on earnings among Soviet Jewish immigrants compared to other immigrants parallels patterns found among Jews and non-Jews born in the US (Chiswick, 1999). Thus, the Soviet Jews appear to be reflecting patterns that are specific to both refugees and Jews in the United States.
NOTES 1. Analyses using a similar methodology have been conducted for the Hebrew language proficiency and labor market earnings of Jewish immigrants in Israel. See Chiswick (1988) and Chiswick and Repetto (2001) for analyses of the 1972 and 1983 Censuses of Israel. Unfortunately, the 1995 Census did not include any questions on language usage or language proficiency. The U.S. and Israel studies are not strictly comparable because of differences in the Census questionnaires, the nature of
206
BARRY R. CHISWICK AND MICHAEL WENZ
immigration into these two countries, the relative magnitudes of the immigration flows after the collapse of the Soviet Union (small for the U.S., large for Israel), and the differences in the local (native) populations. Israel policy regarding intensive efforts to promote Hebrew language usage among immigrants was relaxed with regards to the Russian-speaking immigrants who arrived following the collapse of the FSU. For a discussion of this implicit change in policy see Glinert (1995). 2. In principle, data from the recently released National Jewish Population Survey (NJPS) 2000/2001 can be used to study the economic status of Soviet Jewish immigrants. The NJPS 2000/2001, however, provides a relatively smaller sample of Soviet Jews. Of the 5,148 respondents, both male and female aged 18 and over, only 281 were born in the FSU. 3. With the demise of the Soviet Union and the reunification with East Germany, Germany instituted a special immigration program to attract Soviet Jews to rebuild the German Jewish community (see Tress, 1995). In 2005, the German government was taking steps to effectively close this program (Bernstein, 2005). 4. According to the 2000 Census, the ethnic origins (ancestry) of the adult (aged 25–64) males born in the Soviet Union who immigrated in 1965 or later were 41 percent Russian, 20 percent Ukrainian, 11 percent Armenian, 10 percent response indicating a religion, 6 percent no ancestry reported, and 13 percent other responses. By languages spoken in the home, ‘‘only English’’ was reported by 4 percent, Russian 72 percent, Armenian 9 percent, Ukrainian 7 percent, Yiddish 0.2 percent, and all other languages 8 percent. There was little variation in the reported ancestry or language by sub-period of immigration. See Appendix Tables A1and A2. 5. The very low proportion reporting Yiddish reflects the very rapid decline in the use of Yiddish by Russian/Soviet Jews during the 20th century. By the 1970s, ‘‘for the great majority of contemporary Soviet Jews (80 percent of our respondents), Russian is the native language’’, with the proportion being greater for younger Jews. Yiddish was spoken primarily by older Jews or when younger Jews were speaking with their parents (Karklins, 1987, p. 29). 6. The schooling data cannot be decomposed into pre- and post-migration schooling, although given the age at migration there is likely to be little postmigration schooling among Soviet Jews. 7. The schooling data cannot be decomposed into pre- and post-migration schooling, although given the age at migration there is likely to be little postmigration schooling among Soviet Jews. 8. ‘‘Rural residence’’ is defined as living on a farm in the 2000 Census analysis and living in a rural area (farm or non-farm) in the 1980 and 1990 Census analyses. 9. Tolts (2004a) also finds a very low re-migration rate of Soviet Jewish immigrants who arrive in Israel. 10. The period of arrival categories used here are: 1996.–2000, 1991–1995, 1987–1990, 1985–1986, 1980–1984, 1975–1979, 1970–1974, and 1965–1969. For the proportion of the sample who arrived in each interval, see Appendix Table A4. 11. In the 2000 Census, unlike previous censuses, there is republic of birth codes for each of the 15 republics in the FSU, as well as a generic ‘‘USSR’’ code. Excluding those reporting Armenian by ancestry or language or that they speak Ukrainian at home, 46 percent reported the Russian Republic, 29 percent the Ukraine, 6 percent the USSR, 5 percent Belarus, and 14 percent reported having been born in the other
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
207
12 republics (Appendix Table A3). In the post-World War II censuses until 2000 only the three Baltic Republics (Estonia, Latvia, and Lithuania) were separately identified from the rest of the Soviet Union because the U.S. State Department did not recognize their incorporation into the Soviet Union. 12. A notable exception is the much larger positive effect of being married in the most recent cohort, 1990–2000. 13. For a discussion of the regional distribution of immigrants and their language skills, see Chiswick and Miller (2005). 14. Lower initial English proficiency and earnings and a speedier improvement appear to be a general refugee phenomenon, although not the larger payoff from schooling (see Chiswick, 1978, 1979;Chiswick & Miller, 1998).
ACKNOWLEDGMENTS Chiswick acknowledges the research support of the Institute of Government and Public Affairs, University of Illinois. Comments on earlier version from Carmel U. Chiswick, Allen Glicksman, and Mark Tolts are appreciated.
REFERENCES Ahmed, B., & Robinson, J.G. (1994). Estimates of emigration of the foreign-born population: 1980–1990. US Bureau of the Census, Population Division Technical Working Paper no. 9. Bernstein, R. (2005). Policy shifts in Germany trims Jewish migration. New York Times, February 20. Chiswick, B. R. (1978). The effect of Americanization on the earnings of foreign-born men. Journal of Political Economy, 86(5), 897–922. Chiswick, B. R. (1979). The economic progress of immigrants: Some apparently universal patterns. In: W. Fellner (Ed.), Contemporary economic problems (pp. 357–399). Washington: American Enterprise Institute. Chiswick, B. R. (1988). Hebrew language usage: Determinants and effects on earnings among immigrants in Israel. Journal of Population Economics, 11(2), 253–271. Chiswick, B. R. (1991). Jewish immigrant skill and occupational status at the turn of the century. Explorations in Economic History, 28(1), 64–86. Chiswick, B. R. (1992). Jewish immigrant wages in America in 1909: An analysis of the Dillingham commission data. Exploration in Economic History, 29(3), 274–289. Chiswick, B. R. (1993). Soviet Jews in the United States: An analysis of their linguistic and economic adjustment. International Migration Review, 27(2), 260–286. Chiswick, B. R. (1997). Soviet Jews in the United States: Language and labor market adjustments revisited. In: N. L. Epstein, Y. Ro’i & P. Ritterband (Eds), Russian Jews on three continents: Migration and resettlement (pp. 233–260). London: Frank Cass Publishers.
208
BARRY R. CHISWICK AND MICHAEL WENZ
Chiswick, B. R. (1999). The occupational attainment and earnings of American Jewry, 1890–1990. Contemporary Jewry, 20, 68–98. Chiswick, B. R. (2000). Are immigrants favorably self-selected: An economic analysis. In: C. D. Brettell & J. F. Hollifield (Eds), Migration theory: Talking across disciplines (pp. 61–76). New York: Routledge. Chiswick, B. R., & Miller, P. W. (1998). English language fluency among immigrants in the United States. Research in Labor Economics, 17, 151–200. Chiswick, B. R., & Miller, P. W. (2005). Do enclaves matter in immigrant adjustment. City and Community, 4(1), 5–35. Chiswick, B. R., & Repetto, G. (2001). Immigrant adjustment in Israel: Literacy and fluency in Hebrew and earnings. In: S. Djajic (Ed.), International migration: Trends, policy and economic impact (pp. 204–228). New York: Routledge. Glinert, L. H. (1995). Inside the language planner’s head: Tactical responses to a mass immigration. Journal of Multilingual and Multicultural Development, 16(5), 351–372. Karklins, R. (1987). Determinants of ethnic identification in the USSR: The Soviet Jewish case. Ethnic and Racial Studies, 10(1), 27–47. Mulder, T. J. (2003). Foreign-born emigration from the United States: 1990 to 2000. Paper presented at the Population Association of America, Annual Meeting, Minneapolis, May. Tolts, M. (2004a). Demographic trends among the Jews of the FSU. Paper presented at the International Conference on Soviet and Post-Soviet Jewry in Honor of Professor Mordechai Altshuler, Hebrew University, Jerusalem, December 28–30, 2003. (Revised January 12). Tolts, M. (2004b). The post-Soviet Jewish population in Russia and the world. Jews in Russia and Eastern Europe, 1(52), 37–63. Tress, M. (1995). Soviet Jews in the Federal Republic of Germany: The rebuilding of a community. The Jewish Journal of Sociology, 37(1), 39–54. U.S. Bureau of the Census. (2003). 2000 Census of population and housing, public use microdata sample, United States, Technical Documentation, Washington, DC.
Data of the 2000 Census analysis on immigrants to the US are given in Tables A1–A6 Table A1.
Ancestry or Ethnic Origin of Adult Male Soviet Immigrants Who Immigrated Since 1965, 2000 (percent).
Ethnic Ancestry
Period of Immigration 1965–2000
Russian Religionb Armenian Ukrainian Not reported Soviet Union, n.e.c.c Lithuanian Latvian Polish All other Total
1965–1979
1980–1989
1990–2000
All
Excl. Armenian/ Ukrainiana
All
Excl. Armenian/ Ukrainiana
All
Excl. Armenian/ Ukrainiana
All
Excl. Armenian/ Ukrainiana
41.1 9.6 10.8 19.9 5.8 4.7
52.6 11.7 – 18.1 5.6 2.6
36.7 10.9 12.7 18.9 6.7 1.3
49.9 10.3 – 19.8 8.4 0.4
39.1 9.6 17.1 18.3 7.3 1.1
51.9 13.7 – 14.9 6.9 2.3
41.4 9.7 9.8 20.9 5.4 5.4
53.2 11.5 – 18.5 4.8 3.1
1.5 0.8
1.5 0.8
0.8 1.6
1.1 1.1
1.3 1.5
1.8 1.0
1.5 0.5
1.5 0.7
0.3
0.5
1.3
1.3
0.0
0.7
0.1
0.2
5.5 100.0
6.7 100.0
9.1 100.0
7.5 100.0
4.7 100.0
6.7 100.0
5.3 100.0
6.5 100.0
209
Note: Detail may not add to total due to rounding. Source: 2000 Census of Population, Public Use Microdata Sample, 5 percent sample. a Excludes persons of Armenian ancestry or who speak Armenian or Ukrainian at home. b Response to ancestry question indicating the person’s religion or religious origin, ancestry code 998. c Includes Azerbaijani, Belorussian, Estonian, Ossetian, Moldavian, Tatar, Turkestani, Uzbek, Georgian, Tajik, and those who reported Soviet Union.
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
APPENDIX
210
BARRY R. CHISWICK AND MICHAEL WENZ
Table A2. Language Spoken in the Home by Adult Males Who Immigrated from the FSU Since 1965, 2000 (percent)a. Period of Immigration Language English only Russian Armenian Ukrainian Yiddish Other Total
1965–2000
1965–1979
1980–1989
1990–2000
4.0 71.6 9.4 7.2 0.2 7.6 100.0
9.2 63.3 12.7 2.5 0.4 11.9 100.0
5.0 69.4 17.3 2.5 0.0 5.8 100.0
3.1 73.4 8.0 8.4 0.1 7.0 100.0
Note: Detail may not add to total due to rounding. Source: 2000 Census of Population, Public Use Microdata Sample, 5 percent sample. a Language currently spoken in the home other than or in addition to English.
Republic of Birth
Estonia Latvia Lithuania Armenia Azerbaijan Belarus Georgia Moldova Russia Ukraine USSRb Kazakhstan Kyrgyzstan Tajikstan Turkmenistan Uzbekistan Total
Republic of Birth of Immigrants from the FSU, Adult Males, by Period of Immigration Who Immigrated Since 1965, 2000. 1965–2000
1965–1979
1980–1989
1990–2000
NonArmenian/ Ukrainiana
All
NonArmenian/ Ukrainiana
All
NonArmenian/ Ukrainiana
All
NonArmenian/ Ukrainiana
All
4 120 155 17 114 417 106 228 3,540 2,180 454 22 5 8 1 250 7,621
4 120 155 919 168 421 121 236 3,610 2,848 488 23 5 8 2 256 9384
2 24 18 3 7 15 7 25 428 319 100 0 0 0 0 15 963
2 24 18 133 7 15 8 25 434 344 102 0 0 0 0 15 1127
1 40 35 8 15 87 14 24 636 361 115 2 0 3 0 26 1,367
1 40 35 298 16 87 15 24 647 425 121 3 0 3 0 26 1741
1 56 102 6 92 315 85 179 2,476 1,500 239 20 5 5 1 209 5,291
1 56 102 488 145 319 98 187 2,529 2,079 265 20 5 5 2 215 6,516
Source: 2000 US Census of Population, Public Use Microdata Sample, 5 percent sample. Excludes persons reporting Armenian ancestry, or who speak Armenian or Ukrainian at home. b Persons reporting USSR rather than a specific republic.
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
Table A3.
a
211
212
Table A4.
BARRY R. CHISWICK AND MICHAEL WENZ
Period of Immigration for All Adult Male Immigrants Born in the FSU, Including Armenians, 2000 (percent).
Period of Immigration 1995–2000 1990–1994 1985–1989 1980–1984 1975–1979 1970–1974 1965–1969 1960–1964 1950–1959 Before 1950 Total
All Years
Since 1965
37.4 26.3 12.7 8.2 6.1 2.7 2.1 1.9 1.5 0.9
38.7 27.4 13.2 8.6 6.4 2.8 2.2 –– –– ––
100.0
100.0
Note: Detail may not add to total due to rounding. Source: 2000 Census of Population, Public Use Microdata Sample, 5 percent sample.
Regression Analysis of Fluency in English Among Adult Soviet Jewish Males Who Immigrated Since 1965. Dependent Variable ¼ ENGSPK 2000 Census
Immigration Period Variable CONSTANT EDUCYRS AGE IM95_00 IM90_94 IM85_89 IM75_79 IM70_74 IM65_69
(1) 0.7773 (22.19) 0.0391 (26.09) 0.0104 ( 22.69) 0.3547 ( 18.16) 0.1057 ( 5.62) 0.0408 ( 1.93) 0.0386 (1.77) 0.0662 (1.81) 0.1089 (2.01) 0.0123 (1.08)
(2)
1965–1979 (1)
(2)
0.7793 0.7989 0.7897 (21.83) (13.55) (13.12) 0.0388 0.0172 0.0173 (25.57) (6.05) (6.06) 0.0104 0.0027 0.0027 ( 22.74) ( 3.52) ( 3.57) 0.3515 ( 17.97) 0.1031 ( 5.47) 0.0388 ( 1.83) 0.0428 0.0051 0.0081 (1.96) (0.22) (0.35) 0.0678 (1.85) 0.1063 0.0033 0.0021 (1.95) (0.08) ( 0.05) 0.0128 0.0078 0.0063 (1.12) ( 0.39) ( 0.31)
1980–1989 (1)
1990–2000
(2)
0.8356 (14.06) 0.0285 (9.50) 0.0081 ( 9.27)
0.8364 (13.76) 0.0282 (9.25) 0.0081 ( 9.32)
0.0378 ( 2.15)
0.0373 ( 2.11)
a
0.0141 (0.65)
(1)
(2)
0.6814 (17.25) 0.0452 (23.39) 0.0129 ( 21.03) 0.2552 ( 21.93)
0.6921 (17.07) 0.0446 (22.81) 0.0129 ( 21.08) 0.2544 ( 21.83)
0.0161 (1.06)
0.0160 (1.06)
a
0.0142 (0.66)
213
MARRSP
1965–2000
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
Table A5.
214
Table A5.
(Continued )
Dependent Variable ¼ ENGSPK 2000 Census Immigration Period Variable RURAL
CHILD UKRAINE RELIG OTHANCS
Sample size Standard error R2 Adjusted R2
(1) 0.0247 (0.38) 0.0186 (1.40) 0.0074 ( 0.77)
(2)
1965–1979 (2)
(1)
0.0255 0.1399 0.1328 (0.39) (0.87) (0.83) 0.0187 0.0021 0.0025 (1.40) ( 0.08) ( 0.10) 0.0070 0.0252 0.0252 ( 0.72) (1.44) (1.44) 0.0198 0.0047 ( 1.68) (0.23) 0.0245 0.0121 (1.75) (0.45) 0.0081 0.0197 (0.67) (0.90)
0.0421 ( 0.33) 0.0210 (0.73) 0.0108 ( 0.59)
6,492 6,492 0.3440 0.3438 0.2347 0.2356 0.2333 0.2338
(1)
1980–1989
856 0.2235 0.0651 0.0563
(2) 0.0395 ( 0.31) 0.0218 (0.75) 0.0110 ( 0.60) 0.0053 (0.22) 0.0312 (1.29) 0.0061 (0.28)
1990–2000 (1) 0.0126 (0.15) 0.0198 (1.17) 0.0085 ( 0.66)
(2) 0.0136 (0.17) 0.0197 (1.16) 0.0075 ( 0.58) 0.0315 ( 2.04) 0.0258 (1.37) 0.0041 (0.25)
856 1,240 1,240 4,394 4,394 0.2238 0.2833 0.2835 0.3733 0.3731 0.0661 0.1196 0.1208 0.2422 0.2436 0.0539 0.1146 0.1136 0.2410 0.2419
Note: t-ratios in parentheses. Source: 2000 Census of Population, Public Use Microdata Sample, 5% Sample. a Omitted as benchmark; 1980–1984 is benchmark unless otherwise noted.
BARRY R. CHISWICK AND MICHAEL WENZ
SOUTH
1965–2000
Regression Analysis of Earnings Among Adult Soviet Jewish Males Who Immigrated Since 1965. Dependent Variable ¼ LNEARN 2000 Census
Immigration Period Variable CONSTANT EDUCYRS EXP EXPSQ LNWW IM95_00 IM90_94 IM85_89 IM75_79 IM70_74
(1) 5.024 (40.48) 0.0732 (19.17) 0.0082 (2.02) -0.00022 ( 2.74) 1.045 (42.21) 0.3272 ( 6.76) 0.2093 ( 4.60) 0.0228 ( 0.45) 0.0736 (1.40) 0.0270 (0.31) 0.0592 (0.45)
1965–1979 (2)
(1)
(2)
5.010 4.229 4.176 (39.92) (10.11) (9.91) 0.0736 0.0791 0.0789 (19.00) (6.36) (6.34) 0.0081 0.0246 0.0236 (2.00) (1.85) (1.78) -0.00023 -0.00062 -0.00061 ( 2.74) ( 2.37) ( 2.33) 1.044 1.175 1.179 (42.17) (13.51) (13.55) 0.3215 ( 6.64) 0.2024 ( 4.44) 0.0175 ( 0.34) 0.0836 0.0549 0.0650 (1.58) (0.58) (0.67) 0.0259 (0.29) 0.0399 0.0711 0.0430 (0.30) (0.43) (0.26)
1980–1989
1990–2000
(1)
(2)
(1)
(2)
4.798 (14.20) 0.0885 (9.74) 0.0107 (1.08) -0.00024 ( 1.17) 1.021 (12.96)
4.833 (14.10) 0.0880 (9.52) 0.0111 (1.11) -0.00024 ( 1.19) 1.017 (12.85)
5.029 (38.91) 0.0683 (15.32) 0.0042 (0.89) -0.00014 ( 1.44) 1.028 (38.19) 0.1233 ( 4.63)
5.016 (38.32) 0.0689 (15.26) 0.0043 (0.90) -0.00014 ( 1.45) 1.027 (38.14) 0.1257 ( 4.71)
0.0147 ( 0.28)
0.0102 ( 0.19)
215
IM65_69
1965–2000
Linguistic and Economic Adjustment of Soviet Jewish Immigrants
Table A6.
(Continued )
216
Table A6.
Dependent Variable ¼ LNEARN 2000 Census Immigration Period Variable ENGSPK
RURAL SOUTH
(1)
(2)
(1)
0.3133 (10.43) 0.1413 (5.68) 0.1490 ( 0.95) 0.0358 ( 1.11)
0.3112 (10.35) 0.1441 (5.78) 0.1531 ( 0.97) 0.0364 ( 1.13) 0.0206 ( 0.73) 0.0451 (1.33) 0.0547 (1.88)
0.3109 (2.16) 0.2947 (3.69) 0.3456 (0.52) 0.0263 ( 0.26)
6492 0.8292 0.3442 0.3425
856 0.9326 0.2736 0.2650
UKRAINE RELIG OTHANCS
Sample size Standard error R2 Adjusted R2
1965–1979
6492 0.8294 0.3436 0.3422
(2)
1980–1989
1990–2000
(1)
(2)
(1)
(2)
0.3028 (2.11) 0.2962 (3.69) 0.3252 (0.49) 0.0173 ( 0.17) 0.0795 (0.94 0.2817 (2.53) 0.1122 (1.23)
0.2654 (3.11) 0.2275 (3.86) 0.1459 (0.39) 0.0229 ( 0.27)
0.2660 (3.12) 0.2304 (3.89) 0.1520 (0.40) 0.0212 0.25) 0.0933 1.33) 0.0199 0.28) 0.0174 0.26)
0.3258 (9.98) 0.0879 (3.03) 0.2357 ( 1.34) 0.0403 ( 1.11)
0.3241 (9.93) 0.0903 (3.11) 0.2430 ( 1.38) 0.0422 ( 1.16) 0.0224 ( 0.67) 0.0196 (0.48) 0.0677 (1.96)
856 0.9305 0.2796 0.2685
1240 0.8392 0.2319 0.2262
1240 0.8396 0.2330 0.2255
4394 0.8038 0.3622 0.3609
4394 0.8036 0.3630 0.3612
( ( ( (
Note: t-ratios in parentheses. Includes only immigrants who worked and had non-zero earnings in 1999. Source: 2000 Census of Population, Public Use Microdata Sample, 5 percent Sample.
BARRY R. CHISWICK AND MICHAEL WENZ
MARRSP
1965–2000
MULTI-GENERATION MODEL OF IMMIGRANT EARNINGS: THEORY AND APPLICATION Joseph Deutsch, Gil S. Epstein and Tikva Lecker ABSTRACT This paper presents a three-generation migrant analysis, comparing the relative economic performance of various migrant generations to the native population. We develop a theoretical model to explain the relationship between the different earnings of the migrants over three generations and relate the model to the results in the literature. The empirical analysis explores the suitability of the theoretical implications based on data from the 1995 Israeli Census. We show that assimilation of the third generation into the local population is far from clear.
1. INTRODUCTION Most studies to date comparing the economic performance of immigrants, among other aspects, with that of the native-born population mainly focused on the first rather than the second generation of immigrants. This motivated researchers to carry out more extensive research on the Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 217–234 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24007-4
217
218
JOSEPH DEUTSCH ET AL.
diverse aspects of absorption in various host countries among the second generation of immigrants, as compared to their parents and the native population. Card, DiNardo, and Estes (2000) present a comparative perspective on the economic performance of immigrants and their children, utilizing data from the 1940 to 1970 Censuses, and from recent Current Population Surveys (1994–1996). They show important intergenerational links between the economic status of immigrant fathers and the economic status and marriage patterns of their native born sons and daughters. The authors distinguish between three mutually exclusive groups: immigrants; individuals born in the US of immigrants fathers (the second generation); and others. They refer to the latter group as the ‘‘third and higher generation’’ and restrict the definition of the second generation to native born individuals whose mother and father were immigrants. The authors show that controlling for age and region the relative wages of immigrants declined substantially between 1940 and the mid-1990s. Moreover, the secondgeneration men earned more than the first generation of migrants while the second-generation men earned 9 percent higher wages than the third generation men in 1940, 6 percent higher wages in 1970 and 6 percent higher wages in 1994–1996.1 A related finding is that the children of immigrants tend to have noticeable higher education and wages than the third generation controlling for parental background. The results presented by Card et al. (2000) may suggest a better performance of the second generation, on an average, relative to both the first and third generations.2 One should note however that Card et al. (2000) define third generation as third and higher generation using this group as natives. Our analysis is rather different as we specifically study the third generation immigrants and use a fourth group as the natives. Therefore, the results that second generation, on an average, do better than the first and third generations cannot be directly deduced. Several studies on changes in the relative earnings and employment patterns of the second generation have been carried out in various countries. Chiswick (1977, 1978), for example, in his early work, examined the effect of foreign parentage in the United States in 1969 on earnings of native-born white male workers in the 25–64 age range. He showed that earnings among second-generation immigrants were similar or slightly higher than among native white-born male Americans. Earnings were higher among immigrants with foreign- rather than native-born parents. Thus, according to Chiswick, any discrimination against second-generation Americans is apparently overcome by other factors.
Multi-Generation Model of Immigrant Earnings
219
Gang and Zimmermann (2000) and Gang (1999) showed that ethnicity did not affect the educational achievements of second-generation immigrants, compared to those of natives in the same age cohort, in a large German data set. While parental schooling did not play a role in the educational choices of children of foreign-born parents, contrary to the general findings in the literature, there is a statistically significant difference in favor of the father’s over the mother’s education in children of nativeborn German parents. Similar studies among Jewish immigrants of various ethnic origins in Israel have been carried out by Amir (1988), Benski et al. (1991), Lecker (1993), Mark (1994), among others. The intergenerational mobility in earnings and immigrant workers assimilation in the labor market was studied by Kossoudji (1989), Behrman and Toubman (1990), Borjas (1992), Solon (1992), among others, in the United States; by Lillard (2000) in Germany and the United States; and by Corak and Heisz (1997) in Canada. Schultz (1984) in the United States and Binder (1998) in Mexico, among others, conducted research on schooling and educational achievements of such populations. Borjas (1994) investigates whether the ethnic skill differentials introduced into the United States by the inflow of very dissimilar immigrant groups during the Great Migration of 1880–1910 have disappeared during the past century. An analysis of the 1910, 1940 and 1980 Censuses and the General Social Surveys reveals that those ethnic differentials have indeed narrowed but that it might take four generations, or roughly 100 years, for them to disappear. The analysis also indicates that the economic mobility experienced by American-born blacks, especially since World War II, resembles that of the white ethnic groups that made up the Great Migration. However, since relations between immigrants and native populations in the host countries are extremely complex, it is difficult to project the wellcharacterized economic behavior of the first generation of immigrants and the relatively less well-deciphered behavior of the second generation into the third generation. Therefore, a multi-generation model comparing performance of immigrants and the native population in the host countries, particularly with respect to the labor market, is highly pertinent. In this paper, we develop a multi-generation model comparing labor market performance of immigrants and the native population, assuming that the latter is the appropriate reference group and not the home-country population. The model is based on the concept of bilateral altruism among immigrant generations, i.e., positive linkage of the father’s and son’s utilities via their earnings. Thus, if the father earns less than the native population, the son, would maximize his own utility by investing time and effort in
220
JOSEPH DEUTSCH ET AL.
increasing his earnings to compensate also for his father’s relatively low income. Thus, the second generation of immigrants would be expected to be in an advantaged position (at least with respect to the first generation). However, the third generation would revert to a disadvantaged status relative to the second generation, and possibly also to the native population. The model suggests a possible explanation to the empirical findings of Card et al. (2000) presented above. Although the model is applied to immigrants it could also be applied to other disadvantaged individuals. As such the model could be used to explain investment into education (and consequently level of earnings) of an individual who feels that his father has faced discrimination on the labor market. We examined intergenerational mobility of relative earnings among immigrants to Israel, based on the 1995 Israeli Census of Population data. A two-fold comparative analysis over three generations was carried out on two levels: (1) among three generations of immigrants from Asian-African source countries; and (2) between immigrant and native Israeli populations. In the 1995 Israeli Census of Population data, first-generation immigrants showed relatively lower earnings than the second generation, but this fell again in the third generation. This supports the hypothesis behind our multigeneration immigrant performance model. In addition, separating the wage differential into human capital and market evaluation components throws new light on the effects of the relative investment in education in these three generations. By following immigrant economic behavior over three generations, both in theoretical and empirical terms, our model enhances understanding of economic behavior among immigrants in Israel. Since the model is formulated in terms of incomes of immigrants relative to natives, the model may be relevant even in countries with a large public sector and high social benefits because it assumes that individuals care about relative position in the society. We find that assimilation does not necessarily occur in the third generation, indicating that the two migrant-generation case cannot be generalized to all further generations. A bilateral-altruistic two-generation model of immigrant earnings is presented, which is then explored on 1995 Israeli Census data, and followed by a short summary and concluding section.
2. THE MODEL Consider a bilateral-altruistic two-generation, model of immigrant earnings, in which the father’s and son’s utilities are positively linked through their
Multi-Generation Model of Immigrant Earnings
221
earnings. Since they have no intention of returning ‘‘home,’’ the immigrants’ incomes are not given in absolute terms, but relative to those among the corresponding local native population. Under a time constraint, an individual’s earnings are determined by two consecutive decisions concerning: (1) the amount of time invested in the quality of education; and (2) the amount of time devoted to work. Note that the quality of education is positively related to time invested. In addition, the intuitive justification for time invested in education being caused by lower earnings of fathers could be made by an unobserved characteristic such as motivation or ‘‘desire to succeed financially’’. To simplify, without loss of generality, we focus on the first decision: how to allocate time between education, e, and leisure, L under the time constraint, T: In other words, one could think of this in the following way: everyone starts work at the same time of their life. The time allocation problem is whether you ‘‘pay attention’’ in school, or you ‘‘goof off’’ (or maybe help out at the family store). Thus the more effort you invest at school the better is the outcome. Since we are interested in the son’s optimal choice, we assume that the father’s education level is exogenous in the model. Simplifying further, the effect of the father’s earnings on the son’s level and type of education is ignored, focusing on the time invested by the son in education, which may also be considered as invested effort (measurable in time units). Under our assumptions, the father’s utility is defined as follows: U f ¼ U f I Rf ; I Rs ðeÞ (1) where, I Rf and I Rs ðeÞ are the father’s and son’s incomes, relative to those of the natives, respectively. It is assumed that @U f ðÞ @U f ðÞ @2 U f ðÞ @2 U f ðÞ o0; o0 40; 40; @I Rf @I Rs @I 2Rf @I 2Rs and @2 U f ðÞ o0 @I Rf @I Rs A possible justification for the negative sign of the last expression lies on the assumptions that (a) the utility of the father and son are positively related (b) the higher the level of income the higher the level of utility and (c) decreasing marginal utility. If the son cares about his father a unit increase in the utility of a high income (utility) father will increase the son’s
222
JOSEPH DEUTSCH ET AL.
utility by a smaller amount than the expected increase in the son’s utility resulting from a unit increase in the utility of a low income (utility) father. The son’s utility, which is a function of his father’s, may be expressed as follows: U s ¼ U s L; I Rs ðeÞ; U f (2) where it is assumed that
@U s ðÞ @U s ðÞ @U s ðÞ @2 U s ðÞ 40; 40; 40; o0, @L @I Rs @U f @L2 @2 U s ð Þ @2 U s ðÞ @2 U s ð Þ o0; o0 and o0. @U f @I Rf @I 2Rs @U 2f In addition, we assume a direct relationship between the time invested in educationand the son’s relative (expected) income, I Rs ðeÞ; i.e., @I Rs =@e40 and @2 I Rs @e2 o0: The son aims to maximize his utility by optimizing his level of investment in education, e, such that max U s L; I Rs ðeÞ; U f (3) e
s:t:
Lþe¼T
The first-order condition is given by dU s L; I R s ðeÞ; U f @U s ðÞ @U s ðÞ @I Rs @U s ðÞ @U f ðÞ @I Rs ¼ þ þ ¼ 0 (4) de @L @I Rs ðeÞ @e @U f ðÞ @I Rs @e or, alternatively @U s ðÞ ¼ @L
@U s ðÞ @U s ðÞ @U f ðÞ @I Rs þ @I Rs ðeÞ @U f ðÞ @I Rs @e
(5)
The son’s utility function is assumed to satisfy the second-order condition: d2 U s ðÞ @2 U s ðÞ @2 U s ðÞ @I Rs 2 @U s ðÞ @2 I Rs þ þ ¼ @e de2 @I Rs @e2 @L2 @I 2Rs @2 U s ðÞ @U f ðÞ @I Rs 2 @U s ðÞ @2 U f ðÞ @I Rs 2 þ þ @U f ðÞ @I 2Rs @e @U f ðÞ2 @I Rs @e þ
@U s ðÞ @U f ðÞ @2 I Rs o0 @U f ðÞ @I Rs @e2
ð6Þ
Multi-Generation Model of Immigrant Earnings
223
We now examine how changes in the father’s relative earnings, I Rf ; affect his son’s optimal effort, en (satisfying condition (5)). By differentiating the first-order conditions given in (4) we obtain: @en ¼ @I Rf
d2 U s ðÞ dedI Rf 2
d U s ðÞ de2
Since
d2 U s ðÞ o0 it follows that: de2
2 @en d U s ðÞ Sign ¼ Sign dedI Rf @I Rf
(7)
2 d2 U s ðÞ @I Rs @ U s ðÞ @U f ðÞ @U s ðÞ @2 U f ðÞ þ ¼ dedI Rf @U f ðÞ @I Rs @I Rf @e @U f ðÞ@I Rf @I Rs
(8)
From (4)
As stated above, @2 U f ðÞ @U f ðÞ @U s ðÞ @2 U s ðÞ o0; 40; 40; o0 @I Rf @I Rs @I Rs @U f ðÞ @U f ðÞ@I Rf and
@I Rs d2 U s ðÞ 40 thus o0. dedI Rf @e
Therefore, we conclude: @e o0 @I Rf
(9)
Note that the time invested, and not the type of education, were considered (clearly, the more the father earns, the higher the son’s level of education). Given the direct relationship between the son’s relative earnings, I Rs ; and the optimal time devoted to education, en ; i.e., @I Rs ðe Þ @e 40; we obtain: @I nRs ðen Þ o0 @I Rf
(10)
This result is summarized in the following proposition: Proposition. The less the father earns, the more time and effort the son invests in increasing his earnings. Recall that we do not consider the type of education but the time the son invests in education and at the work place with the aim of increasing his income (which is equivalent to an investment in effort, which is also
224
JOSEPH DEUTSCH ET AL.
measured in time units). Similar analysis would also apply to the son’s investment in promotion and raising his income at work. In order to graphically explain the proposition let us consider Fig. 1a and b. Fig. 1a describes the negative relationship between the father’s earnings relative to the natives and the son’s earnings relative to the natives. The 451 line enables to project the earnings relative to the natives from one generation to another. Fig. 1b displays in the horizontal axis the different generations (1st ¼ grandfather, 2nd ¼ father, 3rd ¼ son) and in the vertical axis the individual’s earnings relative to the natives (re). To gain better understanding of this proposition, let us consider the following: as a first-generation migrant, the father is disadvantaged in labor market relative to the native population, due to discrimination, asymmetric information, linguistic problems, etc. The son, affected by his father’s low income, invests time and effort in increasing his own earnings, and, thereby, in turn, his father’s utility, to compensate for his relatively low income. Thus, the immigrant second generation would be expected to be in an advantaged position. In this case, the father’s lower earnings motivates his son to invest more time and effort in education and the work place. Thus, the father’s and son’s earnings are inversely related, as described by the downward sloping curve AB in Fig. 1a, in which earnings relative to natives (re), are measured on both axis. The father’s (first generation’s) relatively
(a) 45°
(b)
IRs(IRf) A
re2nd
A
B
re3rd
B
re1st re
Fig. 1.
re2nd
re3rd
re1st
1st
2nd
3rd
Generation
Relative Earnings (re) by Generation. (a) re: Earnings Relative to Natives, Father; (b) re: Earnings Relative to Natives, Son.
Multi-Generation Model of Immigrant Earnings
225
low earnings, re1st, (on the horizontal axis), and the son’s (second generation) relative earnings are given by point A, re2nd (on the vertical axis). Since the son’s (second generation) relative earnings (re2nd-values) are above the 451-line, they are higher than the first generation’s relative earnings, and the second generation is in an advantaged position. However, since the second-generation migrant is relatively advantaged, the thirdgeneration migrant, who no longer needs to compensate for his father’s low utility by investing more effort in education and the work place, reverts to a disadvantaged status relative to the native population. The 451-line, the son’s (second generation’s) earnings are projected onto the horizontal axis and thus, the grandson’s earnings, at point B (re3rd), are less than son’s (second generation) relative earnings (re2nd). Of course, his income would remain higher than his grandfather’s (the first-generation migrant), but lower than his father’s (second-generation migrant). These results are summarized in Fig. 1b, where relative earnings are on the vertical axis and migrant generations on the horizontal axis. Thus, the intergenerational mobility in earnings follows an inverse U-shaped curve: the first generation has the lowest relative earnings, the second generation has the highest, and the third generation’s are higher than the first but lower than the second. It is not clear however what the relationship is between the wages in the first and third generation. However, what is important for our argument is that the wages in the first generation are lower than that of the second generation and the wages in the third generation are lower than those of the second generation. Convergence is not guaranteed and even if we assume convergence it need not be on the 451 line. Of course, the convergence may be on any other percent of the son’s wage such as the 90 percent. Convergence is obtained if @I nRs ðen Þ @I Rf 4 1: In this case we would see in Fig. 1a a cob-web adjustment over generations.
3. THE STATISTICAL ANALYSIS 3.1. Data The model is applied to the mass immigration to Israel after establishment of the state in 1948. Mass political immigration more than doubled the population of Israel between 1948 and 1952 – from 650,000 to 1.5 million. About 50% of these immigrants were from Islamic countries and the other 50% from Europe. However, since most of the absorbing (native)
226
JOSEPH DEUTSCH ET AL.
population in Israel at that time was from Europe, we focused on immigrants from the Islamic countries in Asia and Africa. The data for the empirical analysis were derived from the 1995 Israeli Census of Population (20% questionnaire), focusing on the male population. Three generations were defined according to their ages at migration and their ages in the 1995 Census. Thus, the first generation are Jews who were older than 10 when they immigrated to Israel between 1948 and 1952 from Asian-African countries. The second generation were immigrants aged 10 or younger who came during the same period, and Israelis aged between 33 and 53 in 1995, with immigrant fathers. The third generation are Israelis younger than 33 in 1995 with immigrant fathers whose age on immigration was 10 or younger.3 The native Israeli population is defined as those born in Israel to Israeli-born fathers. The age ranges for the first, second and third generations are 53 or older, between 33 and 53 and 33 or younger. Ideally, the analysis should be based on own families but unfortunately the Census data does not provide a way to identify individual families. This model is explored by examining the wage differentials between the Jewish immigrants to Israel from Asian-African countries (A) and the Israeli native population (N) in the three generations defined above. Table 1 presents the characteristics of groups N and A in the three generations in terms of education, years of schooling and six categories of the highest certificate, age and wages. Note that there may well be selfselection in both groups at this stage. The average ages of both groups are very similar in all the generations (see Table 1). The data are expressed as relative levels or percentages of the native (N) and immigrant (A) groups. However, to keep interpretation consistent, indicators for education at the lower levels (without elementary or high school certificates) are calculated as the ratio between A and N, and at higher levels, as ratio between N and A. Thus, ratios greater than 1 (lower than) mean that group N(A) is more advantaged than group A(N). Moreover, in first- and the third-generation migrants, the average wage and years of schooling are higher among the Israeli native (N) group than among the immigrants (A) whereas, in the second generation, the opposite was found. In Fig. 2, the education ratio is greater than 1 for the first generation at all the levels of education, i.e., immigrants are less well educated than natives. The education ratios for the second generation is less than 1 (except for B.A., M.A. and Ph.D. degrees in which the difference between N and A was greatly reduced relative to the first generation) i.e., in the immigrant second generation, the gap between their own and the native education levels closed. As in the first generation,
Variable
Years of schooling Years of experience (Years of experience)2 No certificate (%) Elementary school (%) High school (%) Post-high school (%) B.A. (%) M.A./Ph.D. (%) Age Wage Sample size
Male Sample Characteristics (%), 1995a.
First Generation
Second Generation
Third Generation
N
A
Ratio
N
A
Ratio
N
A
Ratio
12.0 (4.6) 40.4 (6.6) 1683 (599) 15.6 19.3 25.8 13.2 12.5 13.6 58.5 7,771 (3,313) 4,041
9.2 (4.2) 37.8 (8.3) 1498 (675) 22.2 30.0 29.0 10.9 5.4 2.5 53.0 6,346 (3,156) 5,406
1.30
11.4 (4.2) 23.3 (7.1) 596 (359) 14.3 30.0 24.1 11.8 13.3 6.5 40.7 5,690 (2,876) 8,489
11.8 (3.0) 21.9 (5.19) 508 (241) 6.7 25.2 44.6 13.3 7.6 2.6 42.8 6,810 (3,050) 9,489
0.97
13.7 (2.6) 8.34 (3.6) 82.6 (68.0) 2.9 6.5 48.2 13.2 24.0 5.5 28.0 6,293 (2,876) 2,817
12.6 (2.7) 10.27 (3.0) 121.1 (88.4) 5.0 11.7 49.7 17.0 9.8 6.8 28.8 5,360 (2,275) 11,348
1.09
1.07 1.12 1.42 1.55 1.12 1.21 2.31 5.44 1.18 1.23
1.06 1.17 0.5 0.8 0.54 0.89 1.75 2.5 0.95 0.84
0.8 0.6 1.72 1.80 1.03 1.29 2.45 0.81 0.97 1.17
Multi-Generation Model of Immigrant Earnings
Table 1.
Note: Ratios41 are in favor of the natives (N) and ratioso 1 are in favor of the immigrants (A). Figures within parentheses are the standard deviations. Wages are in Israeli Shekels at the May 2000 rates. Source: Israeli Census of Population (1995). a N is Israeli natives and A is Asia-Africa immigrants.
227
228
JOSEPH DEUTSCH ET AL. 6 5
first g. second g. third g.
4 3 2 1 0 no certificate
element. sch.
high sch.
post-high sch.
B.A.
M.A./Phd.
Fig. 2. Ratios of Education of Natives Immigrants: Comparison of Three Generations. Note: A Ratio Higher Than 1 is in Favor of the Natives (N) and Lower than 1 is in Favor of the Immigrants (A).
the education levels were lower in the immigrant third generation than among the natives (except at the Ph.D. level).4 These descriptive data coincide with the theoretical model. On an average, the second generation invests more time and effort in education than the first and third migrant generations. This trend is broken in immigrant third generation relative to the first. These findings indicate an increased investment in education by the migrant second generation, relative to the first, with a decrease in the third generation (note that the third generation’s performance is inferior to the second’s, but superior to that of the first).
3.2. The Empirical Analysis The empirical analysis explores the hypothesis behind the model: that the immigrant second generation’s labor market performance is better than either their parent’s or son’s, and even exceeds that of the absorbing native population. Toward this end, the wages for 1995 were compared in two groups, Asian-African (A) immigrants and Israeli-born natives (N), over three generations in Israel.
Multi-Generation Model of Immigrant Earnings
229
Statistical analysis is carried out in two stages. (1) Wage equations are estimated in each immigrant generation and the native population. (2) The wage differentials are divided into two components over the three generations in Israel, related to gaps in the human capital levels and differences in market evaluation of individual characteristics. The first component is then further decomposed into sub-components, according to observed individual characteristics, namely, education level and labor market experience. The wage decomposition is carried out according to established methods (see, for example, Oaxaca, 1973; Blinder, 1973; Cotton, 1988; Oaxaca & Ransom, 1994). In this study, the human capital variables include years of education level indicators and experience in the labor market, which is measured as age minus years of schooling minus six years. Table 2 presents the results of the wage equations for the Israeli born natives and Asian-African immigrants. The results show significant coefficients for almost all the variables, the effect of experience on the wage equation is of inverse U shape and there is a positive effect of the level education on wage returns. Table 3 presents the wage differentials and their decomposition in the three generations, based on the wage equations in Table 2. The figures in these tables clearly show that for the first- and third generations, the wage differentials between the Israeli native population and the immigrants are positive whereas, in the second generation the opposite holds. According to the decomposition of wage differentials, the human-capital component of the wage differentials markedly decreases with the immigrant generations. In the first generation, about 70% of the gap in favor of the native population can be explained by the differences in the observed characteristics and the other 30% by market evaluations. In the second and the third generations, the entire wage differential is attributable to market evaluation, and is in favor of the immigrants in the second generation but of the native population in the third generation. If we focus on the effect of market evaluation, the results of this decomposition may be interpreted in line with the model. The main conclusions remain unchanged regardless of the coefficients utilized in the Oaxaca methodology (Israeli natives coefficients or immigrants coefficients). As with the descriptive data, the wage differential analysis is also consistent with the theoretical results. Since descriptive data relating to years of schooling is embedded in the wage decomposition, the relationship between immigrant generations and their earnings and wages
230
Table 2. Variable
First Generation
Second Generation
Third Generation
N
A
N
A
N
8.2831(26.9) 0.0081(0.6) 0.0003( 2.1) 0.1673(4.3) 0.5627(14.6) 0.7796(16.9) 0.9941(19.8) 1.1319(20.5) 0.3131 0.3119 262.6 4,041
7.5284(46.7) 0.0552(6.9) 0.0009( 9.3) 0.0278(1.1) 0.2491(9.6) 0.5333(15.8) 0.8033(18.3) 0.9315(15.6) 0.2364 0.2354 283.7 5,406
7.4969(103.1) 0.0322(5.8) 0.0005( 4.9) 0.098(4.5) 0.4674(19.3) 0.6939(24.6) 0.9239(32.3) 1.1295(31.9) 0.2422 0.2416 387.3 8,489
7.7379(85.3) 0.0307(4.0) 0.0005( 3.0) 0.0994(4.1) 0.3681(15.6) 0.6257(22.9) 0.8901(28.4) 1.008(23.3) 0.1608 0.1602 259.5 9,489
7.1381(96.0) 0.1222(11.9) 0.0037( 6.8) 0.1626(2.2) 0.4854(7.6) 0.6609(9.7) 0.9004(13.6) 1.0519(13.7) 0.2218 0.2198 114.4 2,817
Note: Figures within parentheses are t-statistics. N is Israeli natives and A is Asia-Africa immigrants. The dependent variable is the natural logarithm of monthly gross wage.
a
A 7.4182(232.8) 0.0946(20.7) 0.0035( 16.4) 0.1147(4.2) 0.2762(11.0) 0.404(14.7) 0.6384(21.4) 0.6372(20.2) 0.0532 0.0526 91.0 11,348
JOSEPH DEUTSCH ET AL.
Intercept Year of experience (Year of experience)2 Elementary school High school Post-high school B.A. M.A. and Ph.D. R2 Adjusted R2 F value Sample size
Wage Equations for Male Employees, 1995a.
Wage Differentials Decomposition (Standard Oaxaca), 1995a.
Wage differentials based on natives coefficients Total wage (ln) differential Wage differential components: 1. Human capital differences: 1.1 Schooling 1.2 Experience 2. Market evaluation Wage differentials based on immigrants coefficients Total wage (ln) differential Wage differential components 1. Human capital differences 1.1 Schooling 1.2 Experience 2. Market evaluation
First generation
Second generation
0.203 (22.5%)
0.180 ( 16.5%)
0.161 (17.5%)
0.144 0.179 0.034 0.058
0.005 0.003 0.001 0.175
0.018 0.073 0.091 0.178
(71.2%) (88.1%) ( 16.9%) (28.6%)
(2.5%) (1.7%) (0.8%) (97.3%)
Third generation
( 11.0%) (45.5%) ( 56.5%) (111.0%)
0.203 (22.5%)
0.180 ( 16.5%)
0.161 (17.5%)
0.140 0.162 0.022 0.062
0.013 0.012 0.001 0.192
0.009 0.057 0.048 0.152
(69.1%) (79.9%) ( 10.7%) (30.7%)
( 7.0%) ( 6.4%) ( 0.6%) (106.9%)
Multi-Generation Model of Immigrant Earnings
Table 3.
(5.6%) (35.3%) ( 29.7%) (94.4%)
a
The schooling component was calculated from the coefficients of the five types of school and the experience component from the two coefficients relating to years of experience. Thus, positive values indicate higher wage differential for the Israeli native group relative to the Asian-African origin group. The values in brackets (%) are the differences in the average estimated wages of the two groups and the relative shares of the components of the wage differentials, respectively (thus, the sum of the latter is 100%).
231
232
JOSEPH DEUTSCH ET AL.
relative to one another and to the native population is an inversely U-shaped curve.
4. CONCLUSIONS AND POLICY IMPLICATION The literature on earnings assimilation of first and second generation immigrants, starting in the 1970s with Chiswick (1977, 1978) to Gang and Zimmermann (2000) and Gang (1999), focuses on earnings and economic performance relative to one another and to the native population. However, the migrant-third generation has so far been neglected. Thus, we pose the following questions: (1) Is the two-generation relationship generalized to further generations? (2) Does the migrant-third generation assimilate into the general population? One may find an answer to these questions in Card et al. (2000) who show that on an average the second generation does better than the first- and third generations. To address these issues, we developed a three-generation migrant model. A bilateral-altruistic two-generation model of immigrant earnings was proposed, in which the father’s and son’s utilities are positively linked through their earnings. According to this, performance of the second generation is improved over that of the first, while that of third generation falls below the second generation’s but is above that of the first. This finding coincides with Card et al.’s (2000) empirical findings. To explore this hypothesis empirically, we analyzed the 1995 Israeli Census of Population, covering three generations of migrants to Israel. We showed that the empirical analysis coincides with the theoretical findings. Inverse U-shaped relationships were found between migrant-generation characteristics and their education level and intergenerational earnings mobility. The earnings were relatively the lowest in the first generation, highest in the second generation, but higher than in first but lower than in the second. For the first and third generation immigrants, market valuation works in favor of natives, whereas for the second generation it works in favor of the second generation. Therefore, generalizations from the twogeneration migrant model may not apply to the migrant third generation. These data illustrate a case in which the third generation does not assimilate into the local population. The present model shows that this phenomenon starts with well known disadvantage of first-generation immigrants. Thus, public policy aimed at change should be directed at the first-migrant generation. Thus, countries absorbing immigrants should reduce the inequality between natives and
Multi-Generation Model of Immigrant Earnings
233
migrants. Possible ways to implement this are: (1) subsidizing the immigrants’ income; (2) providing (professional and language) training programs on a basic salary during the absorption period. Such public policy implemented in many different countries amplifies this effect, as well as serving other purposes.
NOTES 1. The pattern of wage gaps for women is generally similar. 2. Borjas (2001) shows that there exists a strong link between the skills of the first and the third generations. 3. Of course third generation would also be people whose parents where Israeli born but grandparents immigrated should be counted. However, this data is not available and by not using it may cause some bias. 4. Notice that the number of observations in the third generation are greater than that of the natives. The reason for this is that the Israeli population is build of migrants and thus the number of natives of third-generation migrants decreases rapidly over time.
REFERENCES Amir, S. (1988). Trends in wage differentials between Jewish males of different ethnic origin during the 1970s. Economic Review, 60, 52–75. Behrman, J., & Toubman, P. (1990). The intergenerational correlation between children’s earnings and their parents’ income: Results from the Michigan panel survey of income dynamics. Review of Income and Wealth, 36(2), 115–127. Benski, T., Don, Y., Lecker, T., & Krausz, E. (1991). Iraqi Jews in Israel – social and economic integration, Ramat-Gan and Institute for Research on Iraqi Jews, Or-Yehuda: Bar-Ilan Press. Binder, M. (1998). Family background, gender and schooling in Mexico. Journal of Development Studies, 35(2), 54–71. Blinder, A. S. (1973). Wage discrimination: Reduced form and structural estimates. Journal of Human Resources, 8, 436–455. Borjas, G. J. (1992). Ethnic capital and intergenerational mobility. Quarterly Journal of Economics, 107(1), 123–150. Borjas, G. J. (1994). Long-run convergence of ethnic skill differentials: The children and grandchildren of the great migration. Industrial and Labor Relations Review, 47(4), 553– 573. Borjas, G. J. (2001). Long-run convergence of ethnic skill differences, revisited. Demography, 38(3), 357–361. Card, D., DiNardo, J., & Estes, E. (2000). The more things change: Immigrants and the children of immigrants in the 1940s, the 1970s, and the 1990s. In: G. J. Borjas (Ed.),
234
JOSEPH DEUTSCH ET AL.
Issues in the economics of immigration (pp. 227–269). Chicago and London: University of Chicago Press. Chiswick, B. R. (1977). Sons of immigrants: Are they at an earnings disadvantage? American Economic Review. Papers and Proceedings,, 67(1), 376–380. Chiswick, B. R. (1978). The effects of Americanization on the earnings of foreign-born men. Journal of Political Economy, 86(5), 897–921. Corak, M., & Heisz, A. (1997). Unto the sons: The intergenerational income mobility of Canadian men. Mimeo, Statistics Canada, Ottawa. Cotton, J. (1988). On the decomposition of wage differentials. The Review of Economics and Statistics, 70, 337–345. Gang, I. N. (1999). Schooling parents and country. Quarterly Journal of Economics Research, 66, 180–186. Gang, I. N., & Zimmermann, K. F. (2000). Is child like parent? Educational attainment and ethnic origin. Journal of Human Resources, 35, 550–569. Kossoudji, S. A. (1989). Immigrant worker assimilation, is it a labor market phenomenon. The Journal of Human Resources, 24, 494–527. Lecker, T. (1993). Intergeneration employment changes among immigrants in Israel. Jewish Population Studies, 25, 258–265. Lillard, D. R. (2000). Cross-national estimates of the intergenerational mobility in earnings. Mimeo. Mark, N. (1994). Ethnic gaps in Israel over time. Discussion paper no. 5–94, The Pinhas Sapir Center for Development, Tel-Aviv University. Oaxaca, R. L. (1973). Male-female differentials in urban labor markets. International Economic Review, 14, 693–709. Oaxaca, R., & Ransom, M. (1994). On discrimination and the decomposition of wage differentials. Journal of Econometrics, 61, 5–21. Schultz, T. P. (1984). The schooling and health of children of US immigrants and natives. In: T. P. Schultz & K. J. Wolpin (Eds), Research in population economics (pp. 255–288). Greenwich, CT: JAI Press. Solon, G. (1992). Intergenerational income mobility in the United-States. American Economic Review, 82(3), 393–408.
ETHNIC ORIGIN AND MULTIDIMENSIONAL RELATIVE POVERTY IN ISRAEL: A STUDY BASED ON THE 1995 ISRAELI CENSUS$ Joseph Deutsch and Jacques Silber ABSTRACT Looking at the Jewish population in Israel in 1995 this paper compares three multidimensional approaches to poverty measurement and checks to what extent they identify the same households as poor. Logit regressions are then estimated to understand which variables have an impact on poverty. Finally, the so-called Shapley decomposition is introduced to estimate the exact marginal impact of these determinants of poverty. Of particular interest to this study was the combined effect of the generation to which the head of the household belongs and his/her place of birth. It turns out that the ethnic origin has a significant impact on multidimensional poverty in Israel insofar as being a head of household born
$
A previous version of this paper was presented at the conference in memory of Tikva Lecker on immigration, minorities and social exclusion Bar-Ilan University, June 27–28, 2004.
Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 235–264 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24008-6
235
236
JOSEPH DEUTSCH AND JACQUES SILBER
in Asia or Africa, whatever the generation to which one belongs, increases, ceteris paribus, the probability of being poor.
1. INTRODUCTION This study, as well as others, has found that the income and level of education of those born in Asia or Africa are lower than those of the other Jewish population subgroups. As a consequence if the level of education of parents and their standard of living has an effect on the achievements of their children, these findings tend to show that there is a vicious circle in which the lower level of schooling and standard of living of those born in Asia or Africa lead to a lower level of education and standard of living of their children and so on.1 (Lecker, 2001).
Tikva Lecker’s lifelong research was largely devoted to the scientific study of the integration of the various waves of immigrants in Israel, mainly of those born in Asia and Africa. This paper attempts, in a way, to continue her research project, since its focus is on the extent of poverty among various subgroups of the Jewish population in Israel.2 To measure poverty we adopt a multidimensional approach3 and a ‘‘relative poverty’’ perspective, as stressed in the title of this paper. Such a relative approach is in fact the one commonly taken in Europe. An alternative view (see, for example, Fields, 2001, pp. 91–94, for arguments defending such a point of view4) would have adopted an absolute poverty approach, one that is generally used in the field of development economics. The present paper’s aim is in fact twofold. First, it offers a systematic examination of three multidimensional approaches to poverty measurement, on the basis of the same data set, by answering the following questions: (a) Do the three approaches lead to different estimates of the extent of poverty in the Jewish population in Israel in general and in various population subgroups in particular? (b) To what extent are the same households identified as poor by the three approaches? (c) Are there differences between the three approaches in the determinants of household poverty? (d) Which explanatory variables have the greatest marginal impact as determinants of poverty? The main goal of this paper however is to find out whether, ceteris paribus, the ethnic origin, measured by the combination of information on the
Ethnic Origin and Multidimensional Relative Poverty in Israel
237
generation of the head of the household and the place of his/her birth, has an impact on the probability for a household to be considered as poor. If almost 40 years after the birth of the State of Israel,5 this ethnic origin still plays a significant role, even when such variables as the size of the household, the age of the head of the household, his/her gender, level of schooling, marital status and status at work and the type of locality of residence of the household are kept constant, one may wonder whether the integration of Jewish immigrants from all over the world was as successful as is often argued. The paper is organized as follows. We first quickly review (Section 2) the relevant theoretical literature on multidimensional poverty, describing three multidimensional approaches to poverty measurement: the ‘‘Fuzzy’’ approach, an approach derived from Information Theory and the more recent axiomatic approaches to poverty measurement.6 Then we give (Section 3) the informational basis of our analysis (the variables that were selected and their definition). In Section 4 we give some general information on the ownership of durable goods among first, second and third generation Jews living in Israel, since information on the ownership of durable goods was at the basis of our computation of multidimensional poverty indices. We then present in Section 5 estimates of the extent of poverty in the various population subgroups that were distinguished, on the basis of the three multidimensional approaches that were selected, and check to what extent the different approaches identify the same households as poor. In Section 6 we analyze, on the basis of Logit regressions, the determinants of poverty and focus our attention on the impact of the ethnic origin of the head of the household on the probability for a household to be considered as poor. Finally, in Section 7, using the so-called Shapley decomposition procedure, we attempt to determine the marginal impact on poverty of the various categories of explanatory variables that were introduced in the Logit regressions. Concluding comments are given in Section 8.
2. THEORETICAL BACKGROUND In a recent comparative study of multidimensional poverty analysis Deutsch and Silber (2005) present a detailed survey of four approaches, the so-called fuzzy approach, the information theory, the axiomatic and the efficiency analysis approaches. In this paper, we will use the first three techniques and they will be only succinctly summarized in the following sections. We refer
238
JOSEPH DEUTSCH AND JACQUES SILBER
the reader to Deutsch and Silber (2005) for more details as well as to Appendix B, which gives the mathematical definitions of the various poverty indices.
2.1. The ‘‘Fuzzy Set’’ Approach to Poverty Analysis The theory of ‘‘Fuzzy Sets’’ was developed by Zadeh (1965) on the basis of the idea that there are cases where one is unable to determine which elements belong to a given set and which ones do not. When applied to the concept of poverty this idea implies that in some cases an individual is in such a state of deprivation that he certainly should be considered as poor while in others his level of welfare is such that he certainly should not be classified as poor. There are, however, instances where it is not clear whether a given person is poor or not. This is especially true when one takes a multidimensional approach to poverty measurement, because according to some criteria one would certainly define him as poor whereas according to others one should not regard him as poor. Such a fuzzy approach to the study of poverty has taken various forms in the literature among which are the Totally Fuzzy Approach (TFA) proposed by Cerioli and Zani (1990) and the Totally Fuzzy and Relative (TFR) Approach.7 This TFR approach proposed by Cheli, Ghellini, Lemmi, and Pannuzi (1994) and Cheli and Lemmi (1995) takes a relative approach to poverty. First a value has to be determined for each poverty indicator and this value will depend on where the individual is located in the cumulative distribution of this indicator. Then these authors, following in fact Cerioli and Zani (1990), suggest that the weight given for example to the ownership of a durable good should be related to the proportion owning this durable good. The idea in fact is that if owning a washing machine is much more common than owning a dryer a greater weight should be given to the former indicator so that if an individual does not own a washing machine, this relatively rare occurrence will be taken much more into account in computing the overall degree of poverty than if some individual does not own a dryer, a case which is assumed to be more frequent. Finally after computing for each individual an index determining his ‘‘degree of belonging to the set of poor’’, the TFR approach defines the average degree of poverty in the population as the mean over all individuals of these individual poverty indices.8
Ethnic Origin and Multidimensional Relative Poverty in Israel
239
2.2. The Information Theory Approach The application of information theory to Economics was introduced by Theil (1967).9 It was then extended by Maasoumi (1986) who used information theory to measure the distance or the divergence between distributions but also to propose indices of multidimensional inequality. He suggested proceeding in two steps. First, a procedure is defined that allows aggregating the various indicators of welfare to be taken into account. Second, an inequality index is selected to estimate the degree of multidimensional inequality. Maasoumi’s idea is to replace the information on the values of the m different indicators for the various n individuals by a composite index that will be a vector of n components, one for each individual. In other words the vector of m elements corresponding to each individual i will be replaced by a scalar which may be considered either as representing the utility that individual i derives from the various indicators or as an estimate of the welfare of individual i, as an external social evaluator sees it. The question then is to select an ‘‘aggregation function’’ that allows deriving such a composite welfare indicator. Maasoumi (1986) suggested finding a vector xc that would be closest to the various m vectors giving the welfare level the various individuals derive from these m indicators. To define such a ‘‘proximity’’ Maasoumi proposes a multivariate generalization of what is called the generalized entropy index (see, Deutsch & Silber, 2005, for more details). The minimization of the ‘‘proximity’’ defines a composite index xci, which turns out to be a weighted average of the different indicators. In the general case it is a weighted harmonic mean; in specific cases it becomes either a geometric or an arithmetic weighted mean of the various indicators. Having derived a composite index xci for each individual i, one may then measure inequality by applying generalized entropy inequality indices that were defined by Shorrocks (1980) and applied to the multidimensional case by Maasoumi (1986). Information Theory has been applied several times to the analysis of multidimensional inequality but only rarely to the study of multidimensional poverty (see, however, Miceli, 1997). Miceli suggested deriving the measurement of multidimensional poverty from the distribution of the composite index xc which was previously defined. Such a choice implies deciding which weight to give to the various indicators and which type of mean to select. We decided to give an equal weight (1/m) to all the indicators j (where m refers to the total number of indicators) and we adopted the arithmetic mean.
240
JOSEPH DEUTSCH AND JACQUES SILBER
Once the composite indicator xc is defined, one still has to choose a procedure to identify the poor. Here again we will follow Miceli (1997) and adopt the so-called ‘‘relative approach’’ commonly used in the unidimensional analysis of poverty where the ‘‘poverty line’’ is defined as being equal to some percentage of the median value of the composite indicator xc. We have chosen as cutting point a ‘‘poverty line’’ equal to 70% of the median value of the distribution of the composite index xc. Thus, any household i for which the composite index is smaller than the ‘‘poverty line’’ will be identified as poor. 2.3. Axiomatic Derivations of Multidimensional Poverty Indices Only few studies attempted to derive axiomatically multidimensional indices of poverty. Chakravarty, Mukherjee, and Ranade (1998) were probably the first to publish an article on the axiomatic derivation of multidimensional poverty indices. Tsui (2002) also made such an attempt, following his earlier work on axiomatic derivations of multidimensional inequality indices (see, Tsui, 1995,1999). Chakravarty et al. (1998) as well as Tsui (2002) view a multidimensional index of poverty as an aggregation of shortfalls of all the individuals where the shortfall with respect to a given need reflects the fact that the individual does not have even the minimum level of the basic need. In other words a ‘‘poverty line’’ is defined with respect to every one of the indicators taken into account when deriving a multidimensional poverty index. The first index derived axiomatically by Chakravarty et al. (1998) is a multidimensional extension of the subgroup decomposable index suggested by Chakravarty (1983). The second index Chakravarty et al. (1998) derived is a multidimensional generalization of the Foster, Greer, and Thorbecke (1984) subgroup decomposable index (known under the name of FGT index). In the empirical investigation that will be reported we assumed that for each indicator the ‘‘poverty line’’ was equal to half the mean value of the indicator. We also decided to give an equal weight to all the indicators.
3. THE INFORMATION BASIS FOR THE DERIVATION OF MULTIDIMENSIONAL POVERTY INDICES AMONG JEWS IN ISRAEL IN 1995 The database we used was the 1995 Israeli census. This census however, provides only information on the ownership of durable goods and does not
Ethnic Origin and Multidimensional Relative Poverty in Israel
241
include any subjective information on, for example, the satisfaction of the household members with respect to their standard of living or their health. In addition we do not know whether the absence of a given durable good was a choice made by the household (who, for religious reasons, for example, may decide not to own a television set) or the consequence of the lack of adequate resources. This suggests that it might have been wiser to use income to derive poverty measures. In other words direct measures such as consumption measures or the possession of durables should be useful when income figures are poor, as it is the case in most developing countries, but since income figures are usually adequate in Israel one may wonder why use a multidimensional approach to poverty based on the ownership of durable goods. There were in fact two reasons for basing our analysis on the ownership of durable goods. First, in the Israeli census information on income is requested only from a sample of households and not from every household and we wanted to use as much information as possible from the 1995 census, in particular given the detailed information we requested to define ethnic groups and generations. Second, our exercise gives us precisely the opportunity to check whether there are good reasons to use information on durable goods when data on income are not available, which is in fact the case of censuses in many Western countries (e.g. France, Switzerland, etc.).10 The information available, concerning the ownership of durable goods, varied from one good to the other. In many cases, we knew only whether the household owned a given durable good or not but there were some cases in which the variables were polytomous. This was the case of the variables indicating the period in which the apartment or house was built, the number of rooms in the dwelling and whether there was a bath or a shower in the dwelling (see, Appendix A for the exact listing of the categories distinguished). There was also a purely quantitative variable, that giving the number of cars available for household use. As far as the number of rooms or cars in the household is concerned we used as variable the number of rooms or cars per individual. Note that the ownership of the dwelling is defined as a dichotomous variable, taking the value 1 if the apartment (or house) is owned by the household. Of special interest to this study is evidently the impact of the continent of origin of the head of the household and the generation to which he/she belongs. The definitions we adopted were those proposed by Lecker (2001) and may be summarized as follows. Three generations were distinguished and for each generation a further distinction was made between those born in Israel and the immigrants.
242
JOSEPH DEUTSCH AND JACQUES SILBER
Thus the first generation includes two subpopulations. The first one refers to the individuals born in Israel and who at the time of the 1995 census were at least 53 years old. The second one corresponds to those who immigrated between 1948 and 1952 and who were at least 10 years old at the time of their arrival in Israel. The first condition defining an individual belonging to the second generation refers to his/her age. If he is a man he has to be at least 33 years old but less than 53 years old at the time of the census. For women the minimal age11 is 32. For the second condition a distinction has to be made between those born in Israel and the immigrants. For those born in Israel, Lecker (2001) required that their father also be born in Israel. The second condition among the immigrants stipulates that either they must have immigrated between 1948 and 1952 and were less than 10 years old at the time of their arrival in Israel, or that they were born in Israel but that their father was born abroad. The first requirement in defining the third generation is that it includes individuals who were less than 33 years old in 1995 (actually 33 for men and 32 for women). This third generation also includes two subpopulations. For those born in Israel, Lecker (2001) stipulated that their father was also born in Israel. The subpopulation of ‘‘immigrants’’ included, in fact, individuals born in Israel but whose father was born abroad. Finally it should be stressed that for each generation a distinction has been made, among those defined as immigrants, between those born in Asia or Africa (AA) and those born in Europe or America (EA).12 To analyze the impact on multidimensional poverty of variables such as the gender, the household size, the age, the marital status, the continent of origin of the head of the household and the generation to which he/she belongs, his/her level of schooling, status at work or place of residence, we have estimated logit-type regressions. As will be explained in Section 6 in these regressions the dependent variable is the probability of being poor while the variables previously mentioned are the explanatory variables. In order to use the information on the ownership of durables in a compact fashion, we had to summarize the information available in the case of polytomous variables, which are categorical variables that may take many values (e.g. the period of construction of the dwelling). In order to do so, we have borrowed a technique used by Cheli and Lemmi (1995) in their work on the fuzzy approach to poverty measurement that was mentioned previously.13 It is interesting to note that this indicator of ownership of a durable good adopted in the case of a polytomous variable is consistent with the very intuitive indicator of ownership that would give us in the
Ethnic Origin and Multidimensional Relative Poverty in Israel
243
case of a binary variable the proportion of households owning the durable good. This way of defining in a compact way the ownership of durable goods should be kept in mind when examining Table 1 which gives information on the ownership of durable goods among first- second- and third-generation Jews in Israel in 1995.
4. THE OWNERSHIP OF DURABLE GOODS AMONG FIRST, SECOND- AND THIRD- GENERATION JEWS IN ISRAEL IN 1995 Table 1 gives information on the ownership of durable goods for each generation and continent of origin. Since the ownership of these goods is certainly a function of the age of the head of the household, the analysis should be limited to comparisons within the generation . In addition, it should be clear that the ownership of some of the durable goods may not be related mainly to wealth but rather depend on tastes. This is, for example, certainly the case of televisions that can be found in 97% of the households so that it is likely that those who do not have a TV made a voluntary choice for religious or other reasons. The indicators that seem the most relevant for getting some insight on the wealth of households are probably the number of rooms per individual, the year of construction of the dwelling, the ownership of an apartment or house, the presence of an air conditioner and the number of cars per individual. The following analysis will indeed be limited to those cases for the cross-table analysis. In the logit regressions the other indicators have been taken into account because the inclusion of various explanatory variables should guarantee that the ownership of durable goods reflects mainly information on the wealth of the households. If we first look at the first generation, it appears that in most cases the indicators for those households classified as being born in Asia or Africa have generally a lower value, except for the case of the ownership of a dwelling. The results are not very different for the second generation, since here also for the five indicators of ownership of durable goods that were selected, we observe a lower level among households whose head was born in Asia or Africa. For the third generation this result holds only for three indicators: the number of rooms per individual, the ownership of an air-conditioner and the number of cars per individual. The distinction, within each generation, between those born in Israel and those born in Europe or America is less clear-cut. In the first generation no
First Generation Born in Israel
First Generation Born in Europe or America
First Generation Born in Asia or Africa
Second Generation Born in Israel
Second Generation Born in Europe or America
Second Generation Born in Asia or Africa
Third Generation Born in Israel
Third Generation Born in Europe or America
Third Generation Born in Asia or Africa
Others
Total
4.8
5.6
6.3
2.3
8.9
10.9
3.6
3.2
6.9
47.5
100.0
61.6
66.5
54.1
36.8
41.8
28.1
49.1
48.2
36.2
46.5
45.6
33.5
26.8
29.1
50.6
51.6
48.1
41.1
43.1
45.2
42.3
42.0
90.3
91.7
88.8
79.1
85.8
80.2
41.7
49.9
52.6
68.9
72.7
85.2
84.0
64.8
88.0
91.4
78.5
80.6
83.2
71.5
76.1
78.2
99.2 96.6 64.2 93.5
99.0 96.8 40.1 89.1
98.0 94.8 47.4 89.9
99.4 93.2 74.7 96.1
99.6 95.1 77.2 96.7
98.7 94.3 66.8 96.7
98.2 82.8 46.3 76.0
98.8 85.7 52.3 82.2
97.0 89.0 54.7 88.6
58.0 43.4 25.8
31.9 24.6 7.9
36.5 20.7 7.9
69.9 46.3 56.4
73.5 56.5 63.1
63.9 33.0 41.8
50.3 13.6 34.1
54.6 17.6 36.3
54.9 11.0 21.0
59.8 72.4
64.0 69.3
37.1 85.0
53.2 79.6
67.3 78.4
41.2 85.1
39.5 65.2
44.1 68.3
30.5 74.7
25.1 53.9
8.4 29.2
7.4 25.9
48.9 53.8
50.0 59.1
32.5 43.4
23.4 50.2
26.4 52.1
22.2 42.5
8294.0
9774.0
10860.0
3995.0
15439.0
18997.0
6202.0
5565.0
12031.0
JOSEPH DEUTSCH AND JACQUES SILBER
Share in total population Number of rooms per individual Year of construction of dwelling Ownership of dwelling Bath or shower in dwelling Phone Television Videotape Washing machine Microwave oven Dishwasher Personal computer Air-conditioning Solar heating system Dryer Number of cars per individual Total number of observations
244
Ownership of Durable Goods by Continent of Origin of the Head of the Household.
Table 1.
Ethnic Origin and Multidimensional Relative Poverty in Israel
245
clear conclusion may be drawn. For the second generation on the contrary we observe a higher level among those born in Europe or America than those born in Israel, this being true for all the five indicators that were selected. For the third generation the level is also higher among those born in Europe or America than those born in Israel but this is true only for four of the five indicators. Let us now turn to an analysis of the incidence of poverty in the various subgroups that were distinguished.
5. THE EXTENT OF POVERTY AMONG FIRST, SECOND AND THIRD GENERATION JEWS IN ISRAEL IN 1995 Table 2 gives information on the incidence of poverty among the nine subgroups that are at the center of our analysis and are defined on the basis of the generation to which the head of the household belongs and of the place of his/her birth. These poverty rates were computed on the basis of three multidimensional approaches: that based on the theory of fuzzy sets (the so-called TFR approach) that derived from information theory, that using axiomatically derived multidimensional poverty indices. While the percentages of poor households may vary a lot, depending on the multidimensional index that is selected (see, Deutsch & Silber, 2005), the differences between the nine groups that were defined are much smaller once an adjustment is made for the overall poverty rate (see the numbers in parenthesis in Table 2). One observes, for example, that the lowest levels of poverty are generally found among those belonging to the second generation, whatever their place of birth, as well as among those who belong to the first generation and were born in Israel. The highest levels of poverty, on the contrary, are generally observed among individuals belonging to the third generation14 or among the other Jews (who were not classified as belonging to one of the three generations), although some indicators also give a higher level of poverty among those who belong to the first generation and were born in Asia or Africa.15 We have also attempted to determine whether, the multidimensional poverty indices that were selected identify the same households as poor. More precisely to check the degree of overlapping between the various multidimensional poverty indices we have assumed that 25% of the individuals were poor, whatever the index that was selected. We then checked to which degree two indices identified the same households as poor. The results of this analysis are given in Table 3.
246
Table 2.
JOSEPH DEUTSCH AND JACQUES SILBER
Incidence of Multidimensional Poverty among First, Second and Third Generation Jews in Israel in 1995a.
First generation born in Israel First generation born in Europe or America First generation born in Asia or Africa Second generation born in Israel Second generation born in Europe or America Second generation born in Asia or Africa Third generation born in Israel Third generation born in Europe or America Third generation born in Asia or Africa Others Jews The whole Jewish population
Percentage of Poor Households when the TFR Index is Selected as Multidimensional Poverty Index
Percentage of Poor Households when the Multidimensional Poverty Index is Derived from Information Theory (Equal Weights)
Percentage of Poor Households when the Multidimensional Poverty Index is Derived the Axiomatic Approach (Generalized FGT Index)
11.5 (0.48)
4.3 (0.43)
27.1 (0.81)
17.6 (0.74)
6.9 (0.68)
36.1 (1.08)
21.2 (0.89)
11.0 (1.09)
36.8 (1.10)
13.0 (0.55)
3.5 (0.35)
21.7 (0.65)
8.8 (0.37)
1.8 (0.18)
18.1 (0.54)
15.1 (0.63)
5.6 (0.55)
26.5 (0.79)
43.6 (1.83)
11.7 (1.16)
37.9 (1.13)
35.6 (1.50)
8.8 (0.87)
34.5 (1.03)
34.5 (1.45)
12.4 (1.23)
36.4 (1.09)
27.6 (1.16) 23.8
13.5 (1.34) 10.1
37.3 (1.12) 33.4
a
The numbers in parenthesis give the ratio of the percentage of poor in a given subpopulation over the percentage of poor in the whole Jewish population.
It appears that the axiomatically derived and the information theorybased indices identify almost the same households as poor (23.7% out of 25%). There is also a high degree of overlapping between the households identified as poor by the axiomatically derived and the fuzzy set theorybased indices (21.7% out of 25%). The degree of overlapping is much
247
Ethnic Origin and Multidimensional Relative Poverty in Israel
Table 3. Degree of Overlapping between the Various Multidimensional Poverty Indices in the Whole Jewish Population (percentage of households defined as poor by two multidimensional indices, assuming 25% of the households are poor). Comparison
TFR index and information theory (equal weights) based index TFR index and generalization of FGT index Information theory based index (equal weights) and generalization of FGT index
Percentage of Poor Common to Two Indices 18.4 21.7 23.7
smaller when comparing the fuzzy set theory based and the information theory related indices. In the next section, an attempt is made, for each of the three approaches, to determine the impact of the various explanatory variables on the probability that a household is considered as poor. This is evidently the only way to determine the specific impact on poverty of the generation to which the head of the household belongs and of the place of his/her birth.
6. RESULTS OF THE LOGIT REGRESSIONS The following exogenous variables have been taken into account: the level of schooling of the head of the household, the size of the household and its square, the age of the head of the household and its square, the gender of the head of the household, his/her marital status, the type of locality where he/ she lives, his/her participation in the labor force (working or not), a set of dummy variables defining both the generation and the place of birth of the head of the household. In addition, dummy variables were introduced to measure the interaction between the gender and the marital status as well as between the gender and the participation in the labor force. In each logit regression, the dependent variable is the probability that an individual is considered as poor (the variable is equal to 1 if he/she is poor, to 0 otherwise). The results of these estimations are given in Table 4, giving in each case the coefficients of the regression obtained on the basis of the three multidimensional approaches to poverty measurement: the TFR, the information theory and the axiomatic approach (generalization of the FGT index).
248
Table 4. Explanatory Variables
TFR: t-Values
Information Theory (Equal Weights): Coefficient
Information Theory t-val.
FGT: Coefficient
FGT: t-Values
5.16587 0.06165
61.68 38.53
2.62958 0.08484
25.20 41.25
6.27424 0.08497
73.76 53.79
0.81125 0.09261
48.66 47.62
0.40591 0.03838
19.86 16.28
0.83844 0.08195
52.50 43.71
0.13471
49.50
0.09310
27.59
0.13962
50.63
0.00104
42.65
0.00072
24.13
0.00116
47.14
0.07118
1.88
0.14241
3.03
0.17631
4.69
0.10146
3.24
0.01370
0.33
0.28612
9.76
0.66872
21.65
0.81049
21.20
0.78691
25.71
0.82360
22.87
1.04437
23.43
0.92103
25.39
0.04556
0.99
0.14424
2.45
0.18908
4.25
0.38878
6.70
0.28167
3.99
0.19341
3.31
0.03725
0.68
0.26225
3.93
0.04600
0.84
0.22590
9.52
0.07532
2.16
0.35136
15.12
0.14146
7.16
0.13037
5.00
0.02137
1.09
JOSEPH DEUTSCH AND JACQUES SILBER
Constant Number of years of schooling Household size Square of household size Age of head of household Square of age of head of household Head of household is male Head of household is married Head of household is divorced Head of household is single Interaction married/ male Interaction divorced/ male Interaction single/ male Household located in Jerusalem Household located in Tel-Aviv
TFR: Coefficient
Results of the Logit Regressions.
6.34
0.30290
9.62
0.16580
7.00
0.80379
31.88
0.93442
28.36
0.95097
38.95
0.06914
2.33
0.10843
2.72
0.03208
1.11
0.86965
23.10
1.05781
18.43
1.09436
32.56
0.83702
27.99
1.03611
24.09
0.84952
32.54
0.56569
20.32
0.62233
17.56
0.65308
25.28
0.74473
14.28
1.03937
11.47
1.13804
20.83
0.98294
30.30
1.54407
24.16
1.34173
40.21
0.52629
21.38
0.56427
15.61
0.74303
30.11
0.39794
11.01
0.90071
17.76
0.90455
24.05
0.53289
14.15
1.02802
18.16
0.99430
25.28
249
0.15476
Ethnic Origin and Multidimensional Relative Poverty in Israel
Household located in Haifa Head of household is working Interaction head of household working/male Head of household is first generation, born in Israel Head of household is first generation, born in Europe or America Head of household is first generation, born in Asia or Africa Head of household is second generation, born in Israel Head of household is second generation, born in Europe or America Head of household is second generation, born in Asia or Africa Head of household is third generation, born in Israel Head of household is third generation,
Explanatory Variables
0.30673
0.15893 173668
TFR: t-Values
10.67
Information Theory (Equal Weights): Coefficient
0.40737
0.14393 173668.00
Information Theory t-val.
10.56
FGT: Coefficient
0.59999
FGT: t-Values
20.33
0.22896 173668.00
JOSEPH DEUTSCH AND JACQUES SILBER
born in Europe or America Head of household is third generation, born in Asia or Africa Likelihood ratio LRI Number of observations
TFR: Coefficient
250
Table 4. (Continued )
Ethnic Origin and Multidimensional Relative Poverty in Israel
251
To have an idea of the goodness of fit of the logit regressions we used a criterion that is similar to the R2 used in linear regressions. The idea is to compute the maximal value of the log-likelihood (ln L) and compare it with the log likelihood obtained when only a constant term is introduced (ln L0 ). The likelihood ratio LRI is then defined as LRI ¼ 1 – (ln L/ln L0) The bounds of this measure are 0 and 1 (see, Greene, 1993, pp. 651–653). The value of the likelihood ratio LRI is also given in Table 4. Most of the results confirm those obtained in a previous study (see, Deutsch & Silber, 2005). One may thus observe that the probability of being poor decreases with the level of schooling one has received, has a U-shaped relationship with both the size of the household and the age of the head of the household. The probability of being poor is also lower among married heads of households but higher among divorced, separated or single heads of households (the reference group here is that of the widow(er)s). The probability of being poor is also higher (ceteris paribus) in the three big cities (Jerusalem, Tel-Aviv and Haifa, but especially Jerusalem) than in the other localities and clearly lower when the head of the household works. One should be careful in analyzing the impact of the gender of the head of the household. After introducing an interaction variable between the gender and the labor force participation of the head of the household, one observes generally that there is no difference between the genders among the heads of household who work. Among those who do not work, the probability of being poor seems to be higher when the head of the household is a man. Let us now take a look at the specific impact of the generation to which the head of the household belongs and of the place of his/her birth. Although the ranking of the nine groups we distinguished varies somehow, depending on the multidimensional poverty index that was selected, the following conclusions may be drawn: (1) Those heads of households that belong to the second generation and were born in Europe and America have the lowest probability of being poor, whatever the index that is selected. (2) A second group that is also well ranked (that also has a relatively low probability of being poor) is that of those heads of households that belong to the first generation and were born in Israel. (3) Those who belong to the second generation and were born in Israel have also a low probability of being poor. (4) On the contrary the group that has the highest probability of being poor, whatever the index selected, is that of the heads of the households that
252
JOSEPH DEUTSCH AND JACQUES SILBER
belong to the third generation and were born in Asia or Africa. Note that this result holds, other things constant, among which is the age of the head of the household. (5) Two other groups also have quite a high probability of being poor: those who belong to the second generation and were born in Asia or Africa as well as those who belong to the first generation and were born in Asia or Africa. (6) In the middle of the ranking one finds finally those who belong to the first generation and were born in Europe or America, as well as those who belong to the third generation and were born either in Israel or in Europe or America. It appears therefore that a distinction should be made, when comparing the generations of immigrants. Among those born in Europe or America, the second generation succeeds the most (has the lowest probability of being poor), then the first and then the third generation. Among those born in Israel the first generation ranks best, then the second and then the third. Finally, among those born in Asia or Africa the first generation ranks better than the second and the second better than the third. But these three groups (born in Asia or Africa) rank all worse than those born in Israel or in Europe or America. To better understand the specific impact of the explanatory variables on the probability of being poor we apply in the next section the so-called Shapley decomposition procedure, a technique that will allow us determining the exact marginal impact on the probability of being poor of each of the eight categories of explanatory variables: the size of the household and the age, gender, educational level, marital status, work status, place of residence and generation/place of birth of the head of the household.
7. THE SHAPLEY APPROACH TO INDEX DECOMPOSITION AND ITS IMPLICATIONS FOR MULTIDIMENSIONAL POVERTY ANALYSIS 7.1. The Concept of Shapley Decomposition (see, Shorrocks, 1999) The idea is to estimate all the regressions that may be defined on the basis of the explanatory variables selected and then to derive the marginal contribution of each explanatory variable to the measure of the goodness of fit of
Ethnic Origin and Multidimensional Relative Poverty in Israel
253
the logit regression that was defined previously. More details are given in Appendix C.
7.2. Determining the Marginal Impact of the Different (Categories of) Explanatory Variables in the Logit Regression This Shapley decomposition has been applied to the various Logit regressions that were estimated. To simplify the computations, we did not compute the marginal impact of each variable but the marginal impact of each category of explanatory variables (eight as a whole): household size as well as the age, gender, marital status, work status, place of residence, schooling and ethnic origin of the head of the household.16 As indicated before, the likelihood ratio LRI that was defined previously will serve as indicator of the goodness of fit of the logit regressions. The marginal impact of each category of variables that was estimated using the Shapley decomposition procedure will then give their (marginal) contribution to this likelihood ratio and the sums of these contributions will be equal, as was just mentioned, to the likelihood ratio itself.
7.3. The Empirical Investigation Table 5 reports, for each approach, the marginal impact of each of the eight categories of explanatory variables on the likelihood ratio LRI that was defined previously. These categories are respectively the education level of the head of the household, his/her age, the gender of the head of the household and his/her marital status, the size of the household and the type of locality where the household lives, the labor force participation of the head of the household and the generation/place of birth of the head of the household. The results of this Shapley decomposition are given in percentage terms in Table 5. The results, as a whole, do not depend too much on the index selected. The marital status of the head of the household has the highest marginal contribution (between 20% and 25%). Another variable that plays an important role is the size of the household (marginal contribution of 20% in all three cases). The generation/place of birth also has an important impact (15%). The age of the head of the household (marginal contribution of 10–17%, depending on the index selected), his/her degree of labor force participation (marginal contribution of 13–17%) and his/her level of education (marginal contribution of 7–12%) also play a
254
JOSEPH DEUTSCH AND JACQUES SILBER
Table 5.
Shapley Decompositions for the Logit Regressions.
Marginal Impacts
Marginal impact of education Marginal impact of the size of household Marginal impact of age Marginal impact of gender Marginal impact of marital status Marginal impact of status at work Marginal impact of area of residence Marginal impact of continent of origin
Index Based on Information Theory
TFR Index
Multidimensional
7.4
9.2
12.0
20.9
20.6
19.6
17.2 2.4
13.6 4.3
9.6 3.0
22.0
19.7
25.2
13.3
16.7
14.1
1.1
0.7
0.8
15.6
15.2
15.6
Note: Marginal impact (in percentage terms) of the eight categories of explanatory variables on the likelihood ratio LRI.
significant role. The gender of the head of the household and the type of locality in which the household lives, however, do not play an important role. These findings strengthen therefore the conclusions that were drawn on the basis of the Logit analysis and stressed the significant differences that exist between the generations as well as the places of birth, as far as the probability of being poor is concerned. Not only are the coefficients of these generation/place of birth variables significant, these variables, taken as a whole, also turn out to have an important marginal impact on the probability of being poor since they explain close to 15% of an indicator that measures somehow the dispersion of the dichotomous dependent variable.
8. CONCLUDING COMMENTS This paper had several goals. First, we wanted to compare three multidimensional approaches (the TFR Approach, an approach based on information Theory and one that used axiomatically derived multidimensional poverty indices) and check to what extent they identified the same households as poor. Second, we planned to understand better the determinants of
Ethnic Origin and Multidimensional Relative Poverty in Israel
255
poverty by estimating Logit regressions with the following categories of explanatory variables: size of the household, age of the head of the household, his/her gender, level of schooling, marital status and status at work, the type of locality of residence of the household and finally the generation/ place of birth of the head of the household. Third, we wished to introduce a decomposition procedure introduced recently in the literature, the so-called Shapley decomposition, in order to determine the exact marginal impact of each of the categories of explanatory variables. Our empirical analysis was based on the 1995 census of the Israeli population and included only the Jewish population. Although, there are some important differences between the indices selected as far as the incidence of poverty is concerned, the degree of overlapping between these approaches in the identification of poor households is quite high. The similarity is even higher when one looks at the impact of the different explanatory variables that were introduced in the logit regressions whose dependent variable was the probability for a household to be poor. Whatever the index selected, the size of the household and the age of its head have a U-shaped impact on the probability for a household to be considered as poor. This probability decreases with the level of schooling of the head of the household, is lower among married and higher among single or divorced heads and is also higher when the head of the household does not work. The place of residence has a significant impact on this probability whereas the impact of the gender of the head of the household is not very important, once his/her status at work is taken into account. Finally, the combined effect of the generation to which the head of the household belongs and his/her place of birth plays a significant role in determining the probability of being poor, this result being also confirmed by the Shapleytype decomposition that was conducted at the end of this study. In fact as far as the impact of the ethnic origin on multidimensional poverty in Israel is concerned, we have to conclude that even when the size of the household, the age, level of schooling, gender, marital status, status at work and place of residence of its head are kept constant, the combined effect of the generation to which he/she belongs and his/her place of birth still plays an important role in determining the probability for a household to be considered as poor. In this respect it turns out that being the head of a household born in Asia or Africa, whatever the generation to which one belongs, increases, ceteris paribus, the probability of being poor, a conclusion that seems largely to confirm the findings of many studies that had been conducted by the late Tikva Lecker in whose memory this paper is written.
256
JOSEPH DEUTSCH AND JACQUES SILBER
NOTES 1. This paragraph is a translation, by the authors of this paper, of the last sentence of Lecker (2001). 2. See, Deutsch and Silber (2005), for a study of multidimensional poverty in Israel that is not limited to the Jewish population. 3. For an introduction to this topic as well as for a justification of our preference for such an approach, see, Deutsch and Silber (2005). Our approach is also a cardinal approach. An ordinal approach would imply extending the unidimensional case of the stochastic dominance approach to multidimensional inequality and poverty measurement. Such extensions have, for example, been proposed by Kolm (1977) and Atkinson and Bourguignon (1982) and surveyed in Maasoumi (1999). Such an ordinal approach however seems to be practically relevant only when the number of indicators is very small. 4. We thank an anonymous referee for drawing our attention to the need to justify our choice and stress it in the title. 5. The State of Israel became independent in May 1948 and the last census took place in November 1995. 6. In another paper, Deutsch and Silber (2005) have also used the so-called efficiency analysis approach, but we decided not to include it in this paper because of space constraints. 7. For more details, see Deutsch and Silber (2005). 8. See Appendix B for a summary of the equations defining multidimensional poverty indices according to the TFR approach. 9. For more details on information theory, see Theil (1967), as well as Deutsch and Silber (2005). 10. When the data are available the differences between these two types of data may actually not be very important. Thus in a study based on the fuzzy approach to poverty and based on the 1992–1993 Consumption Expenditures Survey Silber and Sorin (2005) found that according to the TFR approach 14.6% of the individuals were poor while the headcount ratio, based on this same survey, was equal to 14.8%, assuming the poverty line was equal to 50% of the median income. 11. This difference between the minimal ages of men and women is due to the fact that women have to serve one year less as soldiers. It should also be stressed that the reason why Lecker (2001) defined the age of 32 or 33 as the minimal age is that such a definition implies that these individuals were at least 21 (20 for women) at the time of the previous census that took place in 1983. This allowed Lecker (2001) to make comparisons between the achievements of this generation in 1983 and in 1995. Finally note that the definition of this age group guarantees also that the third generation is old enough to participate in the labor force in 1995. 12. The latter category also includes individuals born in Australia, New Zealand or South Africa. 13. For more details, see Cheli and Lemmi (1995) and Deutsch and Silber (2005). 14. Given that our definition of poverty is based on the ownership of durable goods, one should not be surprised to find higher levels of poverty among younger generations since the ownership of durable goods clearly depends on age, as it will indeed appear in the logit regressions.
Ethnic Origin and Multidimensional Relative Poverty in Israel
257
15. These findings seem quite similar to those of Epstein and Lecker (2001) who concluded that the second generation earns relatively more than the first and third generations. Epstein and Lecker also found that the third generation earns less than the second, but more than the first. Since we compare at this stage only overall levels of poverty, we will have to look at the results of the logit regressions to be able to conclude that our findings are quite similar to theirs. 16. With eight factors there are 8! ¼ 40; 320 combinations for each factor. However many combinations are repeated so that we end up with 128 different combinations. To find the Shapley contribution of each factor a specific combination is run twice, with the factor and without it. We have therefore estimated 6144 ¼ 128 2 8 3 logit regressions (8 refers to the number of factors and 3 to the three multidimensional approaches).
REFERENCES Atkinson, A., & Bourguignon, F. (1982). The comparison of multi-dimensional distributions of economic status. Review of Economic Studies, 49, 183–201. Cerioli, A., & Zani, S. (1990). A fuzzy approach to the measurement of poverty. In: C. Dagum & M. Zenga (Eds), Income and wealth distribution, inequality and poverty, Studies in Contemporary Economics (pp. 272–284). Berlin: Springer. Chakravarty, S. (1983). A new index of poverty. Mathematical Social Sciences, 6(3), 307–313. Chakravarty, S. R., Mukherjee, D., & Ranade, R. R. (1998). On the family of subgroup and factor decomposable measures of multidimensional poverty. In: D. J. Slottje (Ed.), Research on economic inequality (Vol. 8). Stamford, CT: JAI Press. Cheli, B., Ghellini, A., Lemmi, A., & Pannuzi, N. (1994). Measuring poverty in the countries in transition via TFR method: The case of Poland in 1990–1991. Statistics in Transition, 1(5), 585–636. Cheli, B., & Lemmi, A. (1995). Totally fuzzy and relative approach to the multidimensional analysis of poverty. Economics Notes by Monte dei Paschi di Siena, 24(1), 115–134. Deutsch, J., & Silber, J. (2005). Measuring multidimensional poverty: An empirical comparison of various approaches. Review of Income and Wealth, 51(1), 145–174. Epstein, G. S., & Lecker, T. (2001). Multi-Generation model of immigrants’ earnings: Theory and applications. Mimeo. Fields, G. S. (2001). Distribution and development. A new look at the developing world. Cambridge, MA: The Russel Sage Foundation and MIT Press. Foster, J., Greer, J., & Thorbecke, E. (1984). A class of decomposable poverty measures. Econometrica, 52(3), 761–765. Greene, W. H. (1993). Econometric analysis (3rd ed.). NJ, USA: Prentice-Hall. Kolm, S. C. (1977). Multidimensional egalitarianism. Quarterly Journal of Economics, 91, 1–13. Lecker, T. (2001). Immigrants from Asia and Africa in Israel. Mimeo. Maasoumi, E. (1986). The measurement and decomposition of multi-dimensional inequality. Econometrica, 54, 991–997. Maasoumi, E. (1999). Multidimensional approaches to welfare analysis. In: J. Silber (Ed.), Handbook on income inequality analysis. Dordrecht and Boston: Kluwer Academic Publishers.
258
JOSEPH DEUTSCH AND JACQUES SILBER
Miceli, D. (1997). Mesure de la pauvrete´. The´orie et Application a` la Suisse. The`se de doctorat e`s sciences e´conomiques et sociale, Universite´ de Gene`ve. Shorrocks, A. F. (1980). The class of additively decomposable inequality measures. Econometrica, 48(3), 613–625. Shorrocks, A. F. (1999). Decomposition procedures for distributional analysis: A unified framework based on the Shapley value. Mimeo, University of Essex. Silber, J., & Sorin, M. (2005). Fuzzy set approaches to the measurement of multidimensional poverty: A comparison based on Israeli data. Mimeo, Bar-Ilan University, Israel. Theil, H. (1967). Economics and information theory. Amsterdam: North Holland. Tsui, K.-Y. (1995). Multidimensional generalizations of the relative and absolute inequality indices: The Atkinson–Kolm–Sen approach. Journal of Economic Theory, 67, 251–265. Tsui, K.-Y. (1999). Multidimensional inequality and multidimensional generalized entropy measures: An axiomatic derivation. Social Choice and Welfare, 16(1), 145–157. Tsui, K.-Y. (2002). Multidimensional poverty indices. Social Choice and Welfare, 19(1), 69–93. Zadeh, L. A. (1965). Fuzzy sets. Information and Control, 8, 338–353.
APPENDIX A. INFORMATION AVAILABLE IN THE 1995 ISRAELI CENSUS ON DURABLE GOODS List of Variables Number of rooms 1: 2: 3: 4: 5: 6: 7: 8: 9: 10: 11:
1 room 1.5 rooms 2 rooms 2.5 rooms 3 rooms 3.5 rooms 4 rooms 4.5 rooms 5 rooms 5.5 rooms 6 or more rooms
Year of construction of dwelling 1: 2: 3: 4: 5: 6:
Before 1947 1948–1954 1955–1964 1965–1974 1975–1984 1985–1989
Ethnic Origin and Multidimensional Relative Poverty in Israel
7: 8: 9: 10: 11: 12:
1990 1991 1992 1993 1994 1995
Ownership of dwelling 1: Family owned 2: Rented Bath/Shower 1: Bath (with/without shower) 2: Shower only 3: No bath or shower Telephone 1: Yes 2: No Television 1: Yes 2: No Videotape 1: Yes 2: No Washing Machine 1: Yes 2: No Microwave Oven 1: Yes 2: No Dishwasher 1: Yes 2: No
259
260
JOSEPH DEUTSCH AND JACQUES SILBER
Computer 1: Yes 2: No Air-Conditioner 1: Yes 2: No Solar Heating System 1: Yes 2: No Drying Machine 1: Yes 2: No Availability of cars for household use 1: 2: 3: 4:
No car one car 2 cars 3 cars or more
APPENDIX B. THE EQUATIONS DEFINING POVERTY ACCORDING TO EACH APPROACH The Totally Fuzzy and Relative Approach Let xj be the set of polytomous variables x1j , y , xnj which measure the state of deprivation of the various n individuals with respect to indicator j and let Fj be the cumulative distribution of this variable. Let xj(m) with m ¼ 1 to s refer to the various values, ordered by increasing risk of poverty, which the variable xj may take. The TFR approach proposes then to define the degree of poverty of individual (household) i as mXj ðiÞ ¼ 0 if xij ¼ xj ð1Þ
261
Ethnic Origin and Multidimensional Relative Poverty in Israel
and mXj ðiÞ ¼ mXj xjðm
1Þ
þ F j x j ðm Þ
F j x j ðm
1Þ
1
F j xj ð1Þ
(B.1)
if xij ¼ xjðmÞ ; m41 How are the various deprivation indicators aggregated? Let mXj (i) refer as before to the value taken by the membership function for indicator j and individual i, with j ¼ 1 to k and i ¼ 1 to n. Let wj represent the weight one wishes to give to indicator j. The overall (over all indicators j) membership function mP (i) for individual i is then defined as X Wj mXj ðiÞ (B.2) m p ði Þ ¼ j¼1 to k with the weights wj defined as .P .P wj ¼ 1n 1 mbXj j¼1 to k 1n 1 mbXj ¼ 1n mbXj j¼1 to k 1n mbXj
(B.3)
P
where mbXj ¼ 1=n i¼1 to n mXj ðiÞ represents the fuzzy proportion of poor individuals (households) according to the deprivation indicator xj. Finally, the average value P of the membership function in the population is given by X m P ði Þ (B.4) P ¼ 1=n i¼1 to n
The Information Theory Approach To define the degree of ‘‘proximity’’ between an aggregate vector xc and the various m vectors xij giving the welfare level the various individuals Maasoumi proposed a multivariate generalization of the generalized entropy index that is expressed as nX X g o Dg ðxc ; X ; aÞ ¼ 1=ðgðg þ 1ÞÞ x a x x 1 j ci ci ij j¼1 to m i¼1 to n
(B.5)
with ga0; 1 , and where aj represents the weight to be given to indicator j. It can be shown that when g ! 21; one obtains the following indicator hX X i D 1 ðxc ; X ; aÞ ¼ xci a x log x (B.6) j ij ij j¼1 to m i¼1 to n
262
JOSEPH DEUTSCH AND JACQUES SILBER
The minimization of the ‘‘proximity’’ defines then a composite index xci which turns out to be equal to hX i xci / d xij (B.7) j j¼1 to m
In expression P (B.7) dj is defined as the normalized weight of indicator j, that is dj ¼ aj = j¼1 to m aj : The composite indicator xc is thus a weighted arithmetic mean of the different indicators. As far as the selection of the weights dj is concerned we have given an equal weight (1/m) to all the indicators j (where m refers to the total number of indicators). Finally the ‘‘poverty line’’ was taken as being equal to 70% of the median value of the composite indicator xc. In other words any household i for which the composite index xci is smaller than the ‘‘poverty line’’ will be identified as poor. The Axiomatic Approach Let z ¼ ðz1 ; . . . ; zk Þ be the k-vector of the minimum levels of the k basic needs and xi ¼ ðxi1 ; . . . ; xik Þ the vector of the k basic needs of the ith person. Let X be the matrix of the quantities xij which denote the amount of the jth attribute accruing to individual i. Chakravarty et al. (1998) derived then axiomatically two families of multidimensional poverty indices. The first family of indices may be expressed as X X e PðX ; zÞ ¼ 1=n zj (B.8) a 1 x j ij j¼1 to k i2Sj where Sj is the set of poor people with respect to attribute j. This index is a multidimensional extension of the subgroup decomposable index suggested by Chakravarty (1983). When e ¼ 1 we get X X X PðX ; zÞ ¼ 1=n a zj xij zj ¼ aHI j¼1 to k i2Sj j j¼1 to k j j j (B.9)
where H j ¼ ðqj =nÞ and Ij arerespectively i and the pov. ratio P h the head-count erty-gap ratio for attribute j I j ¼ i2Sj zj xij qj z j : The second family of indices is expressed as X X a zj Pa ðX ; zÞ ¼ 1=n (B.10) a 1 x j ij j¼1 to k i2Sj
Ethnic Origin and Multidimensional Relative Poverty in Israel
263
This index is a multidimensional generalization of the Foster, Greer, and Thorbecke (1984) subgroup decomposable index (known under the name of FGT index). In our empirical investigation we used this multidimensional generalization of the FGT index with the parameter a equal to 2. We assumed that for each indicator the ‘‘poverty line’’ was equal to half the mean value of the indicator. We also decided to give an equal weight to all the indicators. Finally, when using this generalization of was individual 2 Pthe FGTindex, an considered as poor when the expression j¼1 to k aj 1 xij zj was greater than the value of this expression for the 75th percentile (in other words we assumed that 25% of the individuals were poor).
APPENDIX C. THE CONCEPT OF SHAPLEY DECOMPOSITION Let an index I be a function of n variables and let ITOT be the value of I when all the n variables are used to compute I. I could for example be the R2 of a regression using n explanatory variables, any inequality index depending on n income sources or on n population subgroups. Let now I k=k ðiÞbe the value of the index I when k variables have been dropped so that there are only (n k) explanatory variables and k is also the rank of variable i among the n possible ranks that variable i may have in the n! sequences corresponding to the n! possible ways of ordering n numbers. We will call I k=ðk 1Þ ðiÞ the value of the index when only (k 1) variables have been dropped and k is the rank of the variable (i). Thus I 1=1 ðiÞ gives the value of the index I when this variable is the first one to be dropped. Obviously there are (n 1)! possibilities corresponding to such a case. I 1=0 ðiÞ gives then the value of the index I, when the variable i has the first rank and no variable has been dropped. This is clearly the case when all the variables are included in the computation of the index I. Similarly I 2=2 ðiÞ corresponds to the (n 1)! cases where the variable i is the second one to be dropped and two variables as a whole have been dropped. Clearly I 2=2 ðiÞ can also take (n 1)! possible values. I 2=2 ðiÞ gives then the value of the index I when only one variable has been dropped and the variable i has the second rank. Here also there are (n 1)! possible cases. Obviously I/(n 1)n (i) corresponds to the (n 1)! cases where the variable i is dropped last and is the only one to be taken into account. If I is an inequality index, it will evidently be equal to zero in such a case. But if it is
264
JOSEPH DEUTSCH AND JACQUES SILBER
for example the R2 of a regression it would give us the R2 when there is only one explanatory variable, the variable i. Obviously I n=n ðiÞ gives the value of the index I when variable i has rank n and n variables have been dropped, a case where I will always be equal to zero by definition since no variable is left. Let us now compute the contribution Cj (i) of variable i to the index I, assuming this variable i is dropped when it has rank j. Using the previous notations we define Cj (i) as h ih X j j C j ðiÞ ¼ 1=n! I ð i Þ I ð i Þ =j h¼1 to ðn 1Þ! =ðj 1Þ
where the superscript h refers to one of the (n-1)! cases where the variable i has rank j. The overall contribution of variable i to the index I may then be defined as X C ðiÞ ¼ 1=n! C ði Þ k¼1 to n k It is then easy to prove that
X C ði Þ I ¼ 1=n! i¼1 to n
IMMIGRANTS IN THE ISRAELI HI-TECH INDUSTRY: COMPARISON TO NATIVES AND THE EFFECT OF TRAINING Sarit Cohen-Goldner ABSTRACT During the 1990s, the Israeli economy experienced two major events. First, starting in the fall of 1989, a large wave of relatively highly skilled immigrants arrived from the former Soviet Union (CIS) increasing the population and the labor force by considerable magnitude. Second, the hitech sector has grown substantially and reached a peak in growth and level in 2000. This paper provides a descriptive analysis of the integration of immigrants from the CIS in the Israeli labor market and, specifically, in the hitech sector. Based on a unique panel data that follows immigrants for up to 12 years in Israel we find a significant positive correlation between immigrants’ participation in Israeli government-provided training programs and the propensity to work as professionals in the hi-tech industry and to work in white-collar occupations in other sectors. However, this correlation diminishes with ‘time since participation’ such that recent participants face a higher probability to work in hi-tech and white-collar jobs than those who participated in training earlier. Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 265–292 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24009-8
265
266
SARIT COHEN-GOLDNER
1. INTRODUCTION During the decade that started in 1990, the Israeli economy experienced two major events. First, starting in the fall of 1989, a large wave of relatively highly skilled immigrants arrived from the former Soviet Union (CIS), increasing the population and the labor force by considerable magnitude. Second, the hi-tech (HT) industry had grown substantially and reached a peak in growth and level in 2000. In this study, we try to explore the interaction between the growth of the HT industry and the integration of CIS immigrants in the Israeli economy and, specifically, in the HT industry. In particular, we compare the integration of immigrants in the HT sector to their integration in other occupations as well as to the presence of natives in this industry. Furthermore, since the immigrants were offered governmentsponsored vocational training, we study the determinants of training attendance and the role of training in the process of labor market employment and occupational dynamics of the immigrants. The analysis is based on two sources of data that provide the best description of the Israeli labor market dynamics. First, we use the annual cross-sectional Labor Force Survey (LFS) of 1990–2000. The LFS is conducted by the Central Bureau of Statistics (CBS) and provides a very good aggregate description of the labor force by occupation, industry, years since immigration, country of birth, gender and age. In addition, we are the first to use a unique panel data that tracks a sample of immigrants for a period of up to 12 years since their arrival. All immigrants in this sample studied engineering in the CIS (hereafter engineers survey). The two databases, therefore, differ in their structure (cross section vs. panel) and in the investigated population (general population of CIS immigrants in the LFS vs. immigrant engineers in the engineers survey). In both surveys the analysis is restricted to CIS immigrants who arrived in Israel between 1989 and 1994. The cross-sectional data is used mainly for the comparison between immigrants and native Israelis, while the engineers panel data is used to describe the patterns of training attendance among immigrants and the transitions of immigrants between various labor market states. In particular, the engineers survey enables us to study the role of training in the occupational choice of immigrant engineers. Based on the cross-sectional data, we find that the integration of immigrants in the HT industry is similar to that of native Israelis. Furthermore, the share of immigrants who work in high-skill occupations in the HT industry is roughly the same as that of natives. One of the explanations to this phenomenon is that the Israeli HT industry grew parallel to the arrival
Immigrants in the Israeli Hi-Tech Industry
267
of immigrants in the 1990s, such that it absorbed new workers both among natives and immigrants. It is not clear, however, that this observation implies that immigrants’ integration in the HT sector was due to a good match between their imported human capital and the skills demanded by this specific industry. One of the novel aspects of the engineers survey is the detailed information on the participation of immigrants in government-sponsored vocational training programs. These programs were offered by The Ministry of Labor and The Ministry of Absorption and many programs were especially designed for the specific population of highly skilled immigrants. The courses were offered in software engineering, programming, electronic engineering, computers, etc. The average length of these programs was approximately 6 months, which is substantially longer than the average 3 months length of classroom training in the USA. Approximately 40% of the males in the engineers survey have participated in government-subsidized vocational training since their arrival. These participants are, on average, younger at arrival and a higher share of them worked in WC occupations in CIS. Females’ participation rate in training was 38%, and similar to males, female participants were, on average, younger at arrival and most of them worked in the CIS in WC occupations or as professionals in the HT industry. To explore the participation of immigrants in training programs and, in particular, the timing of participation, we estimate a hazard regression for time in Israel until participation in training (duration to training). We find that there is no significant difference in the duration to training of males and females. For males, the duration to training increases with age at arrival, while the knowledge of Hebrew and work in WC occupations before migration lead to a shorter duration. For females, the duration to training also decreases with the knowledge of Hebrew and with the number of children aged less than 18. The most striking effects on the duration to training are the changes in previous labor market states. For example, immigrants who moved from employment to unemployment during the 6 months prior to the training program face a substantial shorter duration to training. These findings suggest that the participation of immigrants in training is not motivated solely by considerations of investment in local human capital, but rather immigrants take advantage of the opportunity to attend training after spells of unemployment. We further estimate a Cox hazard regression for time until work as a professional in the HT industry and find that females face a significantly longer duration. Immigrants who are proficient in English face a shorter
268
SARIT COHEN-GOLDNER
duration to work as professionals in the HT industry, reflecting the international orientation of the Israeli HT industry. Participation in training leads to a longer duration to HT, as the length of the Israeli training programs is usually around 6 months. Imported education (years of schooling) does not have a significant effect on the duration to work as a professional in the HT industry, possibly due to the low variance of this variable in the engineers survey. A Multinomial-Logit analysis for the occupational choice of immigrant engineers shows that participation in training significantly (at 10% level) increases the propensity to work as a professional in HT and to work in WC occupations in other sectors, but this effect declines with ‘time since participation’, such that recent participants face a higher probability than participants who participated in training earlier. This result suggests that the accumulated knowledge in the training programs is subject to depreciation and, therefore, standard ‘before–after’ estimates for the impact of training are sensitive to the time intervals chosen. The rest of the paper is organized as follows. In the next section, we provide the cross-sectional descriptive analysis and in the third section we analyze the panel data. Section 4 concludes.
2. IMMIGRANTS AND NATIVES IN THE LABOR MARKET: CROSS-SECTIONAL DATA In this section, we describe the labor market characteristics of CIS immigrants who arrived in Israel in 1989–1994 and of native Israelis, with focus on the HT industry.1 The analysis is based on the Israeli national crosssection labor force survey, which is conducted annually by the Israeli CBS among approximately 25,000 households.2 Eckstein and Weiss (2002, 2004) provide an extensive analysis of the integration process of the immigrants and a comparison to native Israelis. In this section, we build on the above papers and extend the analysis to the integration of immigrants and native Israelis in the HT industry while in the next section we explore the panel data and incorporate training into the analysis. In order to describe the labor market dynamics of immigrants and natives we define three occupational categories which are mutually exclusive: whitecollar occupations (WC) which include professionals and managers such as engineers, physicians, teachers, nurses, etc.; Blue-collar occupations (BC) include sales agents, electricians, etc. and HT occupations in the HT industry
Immigrants in the Israeli Hi-Tech Industry
269
(see Appendix (B) for the list of economic branches which comprise the HT industry and the list of three-digit occupations, which are considered as HT occupations within the HT industry).3 Fig. 1 presents the aggregate proportion of immigrants’ employment in WC, HT and BC occupations and of immigrants’ unemployment.4 The figure demonstrates the rapid decline in unemployment and the fast increase in employment in BC occupations. Furthermore, after five years in Israel there is a gradual shift from employment in BC jobs to WC occupations. These transitions are explained by Cohen-Goldner and Eckstein (2002, 2004), who showed that as immigrants accumulate Israeli human capital via language acquisition, vocational training and on the job learning, they are able to shift to better jobs. Fig. 1 also indicates that there is no substantial difference in the integration of male and female immigrants in WC occupations. After 10 years in Israel 32% (34%) of male (female) immigrants work in WC occupations. The integration of immigrants in WC occupations is a gradual process and the share of immigrants in these occupations was substantially lower than that of native Israelis during the 1990s (Eckstein & Weiss, 2004). CohenGoldner and Eckstein (2002, 2004) found that the slow and low entrance of immigrants to WC is due to low availability of job offers in these occupations. Participation in training, however, accelerated these offer probabilities, considerably5. In the next section we investigate the role of training in the integration of immigrant engineers in the HT industry. The rapid growth of the HT industry occurred almost parallel to the arrival of immigrants from the CIS. During 1995–2000 two main factors led to the growth of employment in the HT industry: new entrants to the labor market (mainly native Israelis who graduated from college) and immigrants. As most of these immigrants were college graduates and a sizable share of them worked in the CIS as engineers, it is challenging to study the integration of this population in the growing HT industry. Fig. 2 presents the share of employed in the HT industry (in all occupations) among the general population (natives and immigrants) and among natives and immigrants, separately. The figure demonstrates that there are only minor differences between immigrants and natives, such that employment in the HT industry is an identical share from the reference groups. Furthermore, Fig. 3 shows that the share of male immigrants in HT industry is roughly the same as that of native males, while the share of female immigrants is higher than that of native females.6 Since not all workers in the HT industry work in high-skilled occupations, we divide workers within the HT industry to workers in HT occupations
270
SARIT COHEN-GOLDNER
a. Males 0.8
Hi-Tech Occupations White Collar Occupations
Blue Collar Occupations Unemployed
0.7
percentage
0.6 0.5 0.4 0.3 0.2 0.1 0
0
1
b. Females 0.8
2
3
4 5 6 years since arrival
Hi-Tech Occupations White Collar Occupations
7
8
9
10
9
10
Blue Collar Occupations Unemployed
0.7
percentage
0.6 0.5 0.4 0.3 0.2 0.1 0
Fig. 1.
0
1
2
3
4 5 6 years since arrival
7
8
Labor Force Employment and Unemployment of Immigrants. Source: LFS 1991–2000. (a) Males (b) Females.
(e.g., physicists, system analysts, electronic engineers, etc.) and in non-HT occupations (see Appendix (B) for further details on HT-related occupations). The majority of workers in the HT industry work in non-HT occupations. However, as Fig. 4 shows, the total share of workers in HT
271
Immigrants in the Israeli Hi-Tech Industry
12.99 %
0.14
percentage
12.86 % 0.12
12.08 %
0.1
11.87 %
0.08 0.06 0.04 0.02
all population natives all immigrants immigrants after year 1989
0 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
year Fig. 2.
Employment in the Hi-Tech Industry (Present). Source: LFS 1991–2000.
0.25
percentage
0.2
male natives male immigrants female natives female immigrants
15.53 % 14.97 %
0.15
10.45 %
0.1
7.52 % 0.05
0 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
year Fig. 3.
Employment in the Hi-Tech Industry by Gender and Origin (Persent). Source: LFS 1991–2000.
occupations grew from 35% to 45% through the 1990s. Furthermore, Fig. 4 shows that the share and trend of immigrants who work in HT occupations within the industry is roughly the same as the share and trend among natives. The overall composition of employment in the HT industry by
272
SARIT COHEN-GOLDNER 0.6 0.55
percentage
0.5
48.31 % 45.63 %
0.45 0.4 0.35 0.3 HT occupations - natives HT occupations - immigrants
0.25 0.2 1991
Fig. 4.
1992
1993
1994
1995 1996 year
1997
1998
1999
2000
Employment in the Hi-Tech Industry by Occupation. Source: LFS 1991– 2000.
0.6 0.5
percentage
43.46 %
non HT occupations - immigrants HT occupations - immigrants non HT occupations - natives HT occupations - natives
0.4
36.47 %
0.3 0.2 10.37 % 0.1
9.69 %
0 1991
Fig. 5.
1992
1993
1994
1995 1996 year
1997
1998
1999
2000
Composition of the Hi-Tech Industry by Occupations and Origin. Source: LFS 1991–2000.
occupation and origin is presented in Fig. 5. We see that immigrants comprise about 20% of the industry labor force in 2000 (10% work in HT occupations and 10% work in non-HT occupations). As previously pointed out in Fig. 4, the labor composition in the industry has changed, such that
Immigrants in the Israeli Hi-Tech Industry
273
the share of workers in HT occupations increases, both among nativeIsraelis and immigrants. We can summarize that unlike the slow integration of immigrants compared to natives in WC occupations, documented by Eckstein and Weiss (2004), we find that the integration of immigrants in the HT industry as well as their occupational distribution within the industry is very similar to that of native Israelis. One of the explanations to this phenomenon is that the Israeli HT industry grew parallel to the arrival of immigrants in the 1990s, such that it absorbed new workers both among natives and immigrants. Yet, it is not clear whether this observation implies that immigrants’ integration in the HT industry was due to a good match between their human capital and the skills demanded by the HT industry. Given the fact that immigrants were about 50% of the new entrants to the Israeli labor market during 1990–1992, and considering the rapid growth of the HT industry, one could expect a larger share of immigrants in the HT industry. On the other hand, the immigrants’ relatively higher rate of integration (compared to natives) into the HT industry and, in particular, to HT occupations indicates that there was a positive match between the immigrants and the sector demand for labor. The question we face now is whether government-sponsored training had any influence on these observations.
3. OCCUPATIONAL DISTRIBUTION AND TRAINING ATTENDANCE AMONG IMMIGRANT ENGINEERS: PANEL DATA In order to investigate the role of training in the integration of immigrants in HT, we use a unique panel data on immigrant engineers.7 The engineers survey is based on two interviews. The Brookdale Institute conducted the first interview in 1995, and the second interview of the same sample was conducted in 2001–2002.8 The surveys included men and women who had immigrated during 1989–1994 from the CIS, and reported that they have an engineering diploma when entering Israel. In all, 1432 individuals (824 males and 608 females) were interviewed in 1995, and 773 individuals of the original sample (453 males and 320 females) were interviewed in 2001–2002. The current study focuses on 446 men and 304 women aged between 24 and 60 years at arrival in Israel who actively looked for a job at some stage since arrival.
274
SARIT COHEN-GOLDNER
The surveys include information on: age, years of schooling, country of origin, occupation in the country of origin, knowledge of Hebrew before immigration, marital status, size of family, etc. The two interviews enable us to construct a work-history profile of immigrants from time of arrival in Israel until the date of the second interview (2001–2002), that is, a maximum period of 12 years since arrival. For each job that the immigrant worked in Israel, information is available on wages, starting and ending dates, the weekly working hours, occupation and industry.9 The surveys also provide detailed information on immigrants’ participation in Israeli governmentsponsored vocational classroom training and on their participation in Hebrew classes (‘‘Ulpan’’ in Hebrew). Each immigrant who arrived from the CIS during the 1990s received an ‘absorption package’ that included a set of monetary and non-monetary benefits. One of these benefits was the eligibility to participate in a government-sponsored vocational training program. These programs were offered by The Ministry of Labor and The Ministry of Absorption and many of them were especially designed for the specific population of highly skilled immigrants. The courses were offered in software engineering, programming, electronic engineering, computers, etc. Most of the male and female immigrants who participated in training, attended courses which lasted for four months or more. The average weekly hours of these programs was about 22 hours. Despite the long duration of the Israeli training programs, less than 5% of the participants dropped out from training. In the previous section (cross-sectional data) we made a distinction between workers in HT and in non-HT occupations within the HT industry. In this section, however, we distinguish between professionals and non-professionals in the HT industry, where professionals refer to workers who work in WC occupations in the HT industry. The change in definition is due to the lack of three-digit occupations classification in the engineers surveys.10 Non-professionals in HT are classified in this section as BC workers, such that BC, WC (non-HT) and HT-professionals are mutually exclusive employment states. Table 1 provides summary statistics of the engineers survey for males and females. The average age at arrival of males (Panel A, col. 1) is almost 42 and the average years of imported schooling is 16.3. Since all the immigrants in this sample studied engineering, the variance of education is very low. About 39% of the males worked in WC jobs prior to immigration and 16% were employed as professionals in the HT industry in the CIS. The mean duration to first job in Israel is 7.5 months. Only 0.2% of the male engineers were unemployed through the entire sample period, and all of them have
275
Immigrants in the Israeli Hi-Tech Industry
Table 1.
Summary Statistics – Engineers Surveya.
Variables
Entire Sample
Participated in Training Course
Non-Participants in Training Course
446 41.7 (8.9) 16.3 (1.7) 39.0 16.6
179 40.1 (8.2) 16.4 (1.6) 41.9 16.2
267 42.7 (9.1) 16.3 (1.7) 37.1 16.9
(A) Males Number of observations Age on arrival in Israel (years) Education (years) Worked in WC in CIS (%) Worked as a professional in HT in CIS (%) Hebrew knowledge before immigration (%) Number of children Married (%) Time in Israel at latest survey (months) Number of jobs in Israel since arrival Unemployed throughout entire sample period (%) Time from arrival to first job (months) Time from arrival to start of training course (months)
0.89
1.1
0.75
0.65 (0.92) 91.0 122.2 (16.5) 2.9 (1.5) 0.2
0.72 (0.92) 89.4 122.4 (16.8) 3.5 (1.5) 0.6
0.60 (0.92) 92.1 122.0 (16.3) 2.5 (1.3) 0.0
7.5 (7.5) ––
8.2 (9.0) 53.2 (34.3)
7.1 (6.3) ––
304 41.3 (8.6) 16.0 (1.4) 38.8 24.0
116 39.8 (8.0) 15.9 (1.4) 41.4 28.4
188 42.2 (8.8) 16.1 (1.4) 37.2 21.3
0.66
0.86
0.53
0.46 (0.7) 74.3 120.9 (16.5) 2.4
0.59 (0.8) 76.7 120.7 (17.2) 3.0
0.38 (0.7) 72.9 121.0 (16.2) 2.0
(B) Females Number of observations Age on arrival in Israel (years) Education (years) Worked in WC in CIS (%) Worked as a professional in HT in CIS (%) Hebrew knowledge before immigration (%) Number of children Married (%) Time in Israel at latest survey (months)
276
SARIT COHEN-GOLDNER
Table 1. (Continued ) Variables
Entire Sample
Participated in Training Course
Non-Participants in Training Course
(1.4)
(1.5)
(1.1)
2.0
0.0
3.2
15.2 (15.2) ––
15.1 (14.2) 43.7 (33.9)
15.3 (15.9) ––
Number of jobs in Israel since arrival Unemployed throughout entire sample period (%) Time from arrival to first job (months) Time from arrival to start of training course (months) Source: Engineers survey. Standard deviation in parentheses.
a
participated in training at some point since arrival. Around 40% ( ¼ 179/ 446) of all males have participated in government-subsidized vocational training since their arrival. These participants are, on average, younger at arrival and a higher share of them worked in WC in CIS (Panel A, col. 2–3). The average age on arrival of females (Panel B) is 41.3 and the average years of schooling is 16.1. Almost 39% of the females were employed in WC occupations in the CIS and 24% were employed as professionals in HT prior to migration. The average duration of females to the first job in Israel is 15 months, which is twice the duration of males. Females’ participation rate in training is 38%. Similar to males, female participants are, on average, younger at arrival and most of them worked in the CIS in WC occupations or as professionals in HT.11 Fig. 6 describes the dynamics of immigrants’ employment by occupations, unemployment and participation in training for males (Fig. 6a) and females (Fig. 6b). The general trends derived from the engineers panel data are not much different from those obtained from the cross-sectional LFS (Fig. 1), though the levels are different. Unemployment among immigrant engineers declines sharply and employment in BC occupations increases rapidly during the first two years since arrival. The transition to WC occupations is gradual and steady since arrival and up to 10 years later. The shift of male engineers to WC jobs is more rapid and occurs earlier than the shift of female engineers, while according to the cross-sectional data there is no substantial difference in the integration of male and female immigrants in WC occupations. The share of male (female) engineers who work as professionals in HT grows slowly and reaches 7% (5%) after 10 years.
277
Immigrants in the Israeli Hi-Tech Industry A. Males 100% Professionals in HT White-Collar Blue-Collar
90% 80%
training unemployed
Percentage
70% 60% 50% 40% 30% 20% 10% 0% 1
7
13 19 25 31 37 43 49 55 61 67 73 79 85 91 97 103 109 115 121 127 months since arrival
B. Females 100% 90% 80%
Professionals in HT White-Collar Blue-Collar training unemployed
Percentage
70% 60% 50% 40% 30% 20% 10% 0% 1
Fig. 6.
9
17
25
33
41
49 57 65 73 81 months since arrival
89
97
105 113 121 129
Labor Force Composition of Immigrant Engineers. Source: Engineers surveys. (A) Males (B) Females.
278
SARIT COHEN-GOLDNER
The occupational integration of engineers in the labor market is faster and different than that of the general population of CIS immigrants of the same cohort. After 10 years in Israel 50% (35%) of engineer males (females) work in WC occupations, compared to 32% (34%) among the general population of male (female) immigrants. On the other hand, the unemployment rate of engineers is higher than the unemployment rate of the immigrants’ population, mainly among females.12 This finding may indicate that immigrant engineers are more selective than the general population of immigrants and invest more in job-search while unemployed. CohenGoldner and Eckstein (2002, 2004) find a similar result for highly educated immigrants. Fig. 6 also shows that there is no trend in the participation of male engineers in training programs. However, the patterns of training attendance of females is consistent with the theory of investment in human capital, as their participation rate in training is higher close to their arrival and it declines with time spent in Israel.
3.1. Participation in Training and Its Timing Table 2 presents the number of participants in the Israeli training programs by occupation in the CIS and by occupation in the training programs. Based on the occupation that the immigrant studied in the program, we distinguish between HT-related training and non-HT-related training (see Appendix (C) for the list of training programs which are HT-related). The table shows that both among males and females, training attendance does not vary considerably with respect to the occupation held in the CIS. However, among immigrants who worked in BC occupations before migration, there is a higher tendency to attend non-HT training programs. In Table 3, we present the number of transition participants by their occupation in the last job prior to training and their occupation in the first job following training. It is interesting to note that for both males and females, everyone who was unemployed before training, found a job after attending the training course. However, most of these jobs were in BC occupations. Among the training participants, the share of males employed in BC occupations declined from 56.4% before training to 49.7% in the first job after training, while the share of employed in WC occupations grew from 28.5% to 44%. Among female participants, the share of employed in BC occupations did not change due to participation, but the share of employed in WC jobs has doubled from 15.5% before participation to 32.8%
279
Immigrants in the Israeli Hi-Tech Industry
Table 2.
Participation in Training by Type and Occupation in the CISa.
Type of Occupation in CIS
Type of Course
Did not Participate in Training Course
Total
Hi-tech related
Not-Hi-tech related
15 (20.27) 22 (22.0) 41 (15.36) 0 (0.0)
14 (18.92) 24 (24.0) 62 (23.22) 1 (20.0)
45 (60.81) 54 (54.0) 164 (61.42) 4 (80.0)
74 (100) 100 (100) 267 (100) 5 (100)
17 (23.29) 4 (8.89) 21 (11.41) 0 (0.0)
16 (21.92) 11 (24.44) 46 (25.0) 1 (50.0)
40 (54.79) 30 (66.67) 117 (63.59) 1 (50.0)
73 (100) 45 (100) 184 (100) 2 (100)
(A) Males Hi-tech professional White collar (non-HT) Blue collar Did not work in CIS (B) Females Hi-tech professional White collar (non-HT) Blue collar Did not work in CIS
Source: Engineers survey. a Values within parentheses are percent of the total in the row.
afterwards. This increase is mainly due to the reduction of unemployed females after training attendance. In order to describe the dynamic decision to participate in training, conditional of observed state variables, we run a Cox hazard rate regression for the duration to training. The hazard function has the form: HðtÞ ¼ H 0 ðtÞ expðx0 bÞ where the dependent variable H(t) is duration in Israel until training (in months), H0(t) the baseline hazard and x a vector of the state variables. This regression corrects for right censoring, as not all the immigrants have participated in training during the sample period. The results of the regressions for male and female immigrants separately and jointly are presented in
280
SARIT COHEN-GOLDNER
Table 3.
Occupational Transitions after Traininga. Occupation in First-Job after Training
Unemployed after Training
Professional in HT
White collar (not in HT)
Blue collar
0 (0.00) 2 (3.92) 1 (0.99) 1 (4.55) 4
3 (60.00) 45 (88.24) 28 (27.72) 3 (13.64) 79
1 (20.00) 3 (5.88) 68 (67.33) 17 (77.26) 89
1 (20.00) 1 (1.96) 4 (3.96) 1 (4.55) 7
0 (0.00) 0 (0.00) 4 (5.97) 1 (3.45) 5
1 (50.00) 12 (66.66) 15 (22.39) 10 (34.48) 38
1 (50.00) 5 (27.78) 45 (67.16) 18 (62.07) 69
0 (0.00) 1 (5.56) 3 (4.48) 0 (0.00) 4
Total
Occupation in last job prior to training (A) Males Professional in HT White collar (non HT) Blue collar Unemployed before training Total
5 51 101 22 179
(B) Females Professional in HT White collar (non HT) Blue collar Unemployed before training Total
2 18 67 29 116
Source: Engineers survey. Actual numbers. Values within parentheses are percent of the total in the row.
a
Table 4.13 The panel structure of our data allows us to study the interactions between the dynamic decisions if and when to attend training and the dynamic employment decisions. To capture the possibility that the timing of training attendance is closely related to several labor market transitions between states, we include in x indicators related to transitions between different labor market states prior to training. The variable Employed–Unemployed equals 1 if the immigrant has moved from employment to unemployment during the 6 months prior to training, and 0 otherwise. Similarly, the variable Unemployed–Employed
281
Immigrants in the Israeli Hi-Tech Industry
Table 4.
Cox Hazard Regression for Time until Participation in Training.
Dummy for females Hebrew Number of children Years of schooling Age on arrival Worked in WC in CIS Experience in WC Experience as a professional in HT Experience in BC Employed–unemployed Unemployed–employed Unemployed–unemployed Log likelihood
Males
Females
1.325 (0.153)
1.582 (0.269) 1.367 (0.212) 0.936 (0.076) 1.018 (0.018) 0.852 (0.238) 1.001 (0.009) 1.001 (0.015) 1.008 (0.007) 4.908 (1.596) 0.606 (0.313) 1.721 (0.544)
1.037 (0.047) 0.984 (0.009) 1.432 (0.247) 1.002 (0.008) 0.999 (0.012) 1.008 (0.007) 6.265 (1.601) 0.698 (0.339) 3.185 (0.972) 990.454
572.115
All 0.842 (0.105) 1.408 (0.130) 1.012 (0.039) 0.988 (0.008) 1.187 (0.169) 1.005 (0.005) 1.004 (0.009) 1.011 (0.005) 5.901 (1.193) 0.728 (0.248) 2.689 (0.584) 1794.189
Source: Engineers survey. Significant at 95% level. Significant at 90% level.
equals 1 if the immigrant moved from unemployment to employment during the 6 months prior to training, and 0 otherwise. The variable Unemployed– Unemployed indicates that the immigrant was unemployed all through the 6 months prior to training. Hence, the reference group consists of immigrants who were employed all through the 6 months.14 According to the joint regression there is no significant difference between the duration to training of males and females. For males, the duration to training increases with age at arrival while the knowledge of Hebrew and work in WC occupation before migration lead to a shorter duration to training. For females, the duration to training decreases with the knowledge of Hebrew and with the number of children aged less than 18 years of age. Previous occupation-specific accumulated experience does not significantly affect duration to training. The most striking effect on the duration to training is the impact of previous labor market transitions. Immigrants (males and females) who moved from employment to unemployment during the six months prior to the program have a substantial shorter duration to training. The same holds for immigrants who were constantly unemployed during the six months prior to the program. On the other hand, immigrants who moved from unemployment to employment during this period have a longer duration to training, though this effect is not significant. The finding
282
SARIT COHEN-GOLDNER
that changes in the labor market state provide the strongest predictors for the timing of training implies that the participation of immigrants in training is not motivated solely by considerations of investment in local human capital, but rather immigrants take advantage of the opportunity to attend training and avoid temporarily unemployment. 3.2. The Entry of Immigrants to Hi-Tech Table 5 presents the estimates from a Cox hazard regression for the duration until the first entry to the HT industry as a professional. The dependent variable is time in Israel until work as professional in HT (in months) and the regression corrects for right censoring. The hazard model is estimated for male and female immigrants separately and jointly. From the joint regression it turns that females face a significantly longer duration to HT professional. English proficiency leads to a shorter duration to HT, while the knowledge of Hebrew has no significant effect. These results reflect the fact that many Israeli HT companies have an international orientation and some of them work closely with HT companies in the USA. Years of schooling do not affect the duration to HT significantly. Surprisingly, immigrants who were younger at arrival and immigrants who worked in WC occupations in the CIS have a longer duration to enter to HT as professionals. This might result from the fact that these immigrants (i.e., younger and those who worked in WC in the CIS) tend to first participate in training and only later are integrated in high-skilled WC occupations or as Table 5.
Cox Hazard Regression for Time until Work as a Professional in Hi-Tech.
Dummy for females Hebrew English Years of schooling Age on arrival Worked in WC in CIS Worked as a professional in HT in CIS Participated in training Log likelihood
Source: Engineers survey. Significant at 95% level. Significant at 90% level.
Males
Females
All
1.233 (0.407) 1.777 (0.385) 1.041 (0.094) 0.934 (0.022) 0.255 (0.188) 1.542 (0.608)
2.168 (1.135) 0.890 (0.239) 1.078 (0.216) 0.963 (0.031) 0.526 (0.586) 2.451 (1.304)
0.596 (0.168) 1.542 (0.450) 1.387 (0.211) 1.036 (0.087) 0.939 (0.017) 0.260 (0.158) 1.904 (0.574)
2.069 (1.124) 180.994
12.406 (8.572) 80.846
4.294 (1.656) 300.868
Immigrants in the Israeli Hi-Tech Industry
283
professionals in the HT industry. Work as a professional in HT in the CIS and participation in training shortens the duration to HT professional. 3.3. Training and the Occupational Choice of Immigrant Engineers To study the role of imported and local human capital, and specifically of training on the labor market absorption of immigrant engineers, we run multinomial-logit regressions for the three occupational employment states (BC, WC and HT professionals) and unemployment as a function of various human capital variables. As the decision to participate in training is interrelated with the occupational choice, the multinomial-logit regressions provide correlation between the explanatory variable (i.e., training) and the occupational choices of the immigrants. Hence, one should not interpret the results as causal effects. In order to study the causal effect of training on occupational choices one needs to specify a model for the decision to participate in training and its affect on the occupation chosen. Previous work by Cohen-Goldner and Eckstein (2002, 2004) suggests that the simple correlations obtained from a simple multinomial-logit regression retained after controlling for the selectivity to training and occupations. Table 6 presents the estimates obtained from two specifications of the explanatory variables. The regressions are pooled over time, such that each immigrant appears in each regression the number of months she/he appears in the sample.15 The comparison group is employment in BC occupations. Both specifications include indicator for females, years of schooling and age at arrival. The levels of Hebrew and English proficiency both range from 1 (no knowledge) to 4 (perfect knowledge). To capture the possibility that the impact of training on the four potential outcomes changes over time, the regressions include the indicator for training participation as well as the interaction of training participation with ‘time since training participation’ (in years). In the first regression, we also include variables for the accumulated experience in WC occupations and as a professional in HT, but as these variables are likely to be endogenous, we replace them in the second specification with the variable ‘time in Israel’. According to the first specification, the propensity to work as a professional in HT decreases significantly with age at arrival and increases with previous accumulated experience in WC jobs or as a professional in HT. There is no significant difference in the propensity of females to work as a professional in HT compared to that of males. Hebrew proficiency does not appear to have a significant effect on the probability to work as a professional in HT, but proficiency in English does increase this probability significantly.
284
SARIT COHEN-GOLDNER
Table 6.
Occupational Choice and Unemployment of Immigrant Engineers. Hi-Tech Professionals
(A) First specification (Log likelihood: Dummy for females Hebrew English Age on arrival Years of schooling Participated in training Years since training Experience as a professional in HT Experience as a professional in HT-squared Experience in WC Experience in WC-squared Constant
White-Collar Occupations
Unemployed
0.248 (0.316) 0.182 (0.252) 0.36 (0.153) 0.1 (0.025) 0.074 (0.107) 1.299 (0.798) 0.433 (0.182) 0.292 (0.041)
0.163 0.358 0.062 0.016 0.037 1.353 0.406 0.109
0.631 (0.122) 0.17 (0.09) 0.021 (0.066) 0.001 (0.009) 0.017 (0.041) 0.178 (0.191) 0.07 (0.048) 0.076 (0.035)
0.002 (0.0003)
0.001 (0.0003)
0.001 (0.0002)
0.156 (0.027) 0.001 (0.0002) 3.495 (2.041)
0.263 (0.022) 0.002 ( 0.0002) 2.514 (0.866)
0.089 (0.02) 0.001 (0.0001) 0.946 (0.797)
0.312 0.443 0.243 0.089 0.081 0.197 0.031 0.011 3.148
0.447 (0.163) 0.455 (0.119) 0.146 (0.081) 0.039 (0.01) 0.005 (0.049) 0.446 (0.175) 0.016 (0.037) 0.012 (0.001) 1.37 (0.954)
0.607 (0.123) 0.15 (0.088) 0.038 (0.068) 0.0002 (0.009) 0.024 (0.042) 0.243 (0.201) 0.034 (0.053) 0.008 (0.002) 0.59 (0.823)
59410.455) (0.147) (0.122) (0.074) (0.01) (0.045) (0.247) (0.072) (0.038)
(B) Second specification (Log likelihood: 93965.019) Dummy for females Hebrew English Age on arrival Years of schooling Participated in training Years since training Time in Israel Constant
(0.36) (0.327) (0.2) (0.025) (0.105) (0.447) (0.092) (0.003) (1.852)
The reference group consists of immigrants who work in BC occupations. Source: Engineers survey.
Training significantly (at 10% level) increases the propensity to work as a professional in HT, but this effect declines with ‘time since training participation’, such that recent participants face a higher probability than participants who participated training years ago. This result suggests that the knowledge that is accumulated in the training programs is subject to depreciation and, therefore, before–after estimates for the impact of training are sensitive to the choice of ‘time since participation’. The probability to work in WC jobs also increases with accumulated experience in WC jobs or as a professional in HT. Hebrew knowledge also has a positive significant effect on work in WC jobs. Like in HT, training
Immigrants in the Israeli Hi-Tech Industry
285
probability to work as a professional in HT
has a positive impact on work in WC and this impact decreases with ‘time since training participation. Last, the probability of being unemployed also increases with accumulated experience in WC jobs or as a professional in HT, which suggest that immigrants with higher levels of human capital invest more in search from unemployment. Proficiency in Hebrew lowers the probability to be unemployed (compared to work in BC jobs). Females seem to have a significantly higher propensity than males to be unemployed. As was documented in previous studies on Russian immigration to Israel (Eckstein & Weiss, 2004; Cohen-Goldner & Eckstein, 2002, 2004; Weiss, Sauer, & Gotlibovski, 2003), we find that conditional on local accumulated human capital, imported schooling has no significant impact on local labor market activities of Russian immigrants. To illustrate the impact of training on the probability to work as a professional in HT, we present in Fig. 7 this probability conditional on the participation in training in the first year in Israel, and as a function of experience as a professional in HT for the average male/female immigrant engineer.16 The result is that the impact of training on the probability to work as a professional in HT is substantial for both male and female immigrants. The probability to work in HT is doubled for those who participated in training during the last 12 months. The results from the second specification of the multinomial-logit regression suggest that only age at arrival and time in Israel significantly affect the 0.7 average male who participated in training (immediately after arrival) average female who participated in training (immediately after arrival) average male who did not participate in training (immediately after arrival) average female who did not participate in training (immediately after arrival)
0.6 0.5 0.4 0.3 0.2 0.1 0
Fig. 7.
0
1
2
3
4
5
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Experience in Hi-Tech (in months)
Probability to Works as a Professional in the Hi-Tech Industry by Gender.
286
SARIT COHEN-GOLDNER
probability to work as a professional in HT, such that this probability decreases with age at arrival, as in the first specification, and it increases with time spent in Israel. The probability to work in WC is significantly lower for females than for males. Knowledge of Hebrew and English, and time in Israel increase the probability to work in WC. Training also significantly increases the probability to work in WC jobs and its effect is independent of ‘time since training participation’. Similar to in the first specification, females face a higher probability to be unemployed, while Hebrew proficiency leads to a lower probability of unemployment. As expected, the probability to be unemployed, compared to work in BC jobs, significantly declines with time in Israel. Training also has a negative impact on the probability to be unemployed, though the effect is not significant.
3.4. Transitions between Labor Market States In Table 7, we present the annual transitions between the three employment states (BC, WC and HT professionals), training and unemployment. Since the frequency of these transitions may change over time, we present the transitions that occurred during the first five years in Israel and during the subsequent five years (5–10), separately. The entries in the table are the number of individuals’ transitions between month ‘t’ and month ‘t+12’. The table reveals substantial differences between the two sub-periods, mainly with respect to the transitions from unemployment. During the first five years in Israel, 48% (50%) of unemployed males (females) move to BC jobs, while 32% (24%) move to WC jobs and 14% (21%) are unemployed 12 months later. The transitions from unemployment to HT professionals are minor for both males and females. In the second sub-period (5th to the 10th year), however, most of the unemployed immigrants remain unemployed. This finding suggests that during the first five years unemployment among immigrants is more transitory. That is, immigrants are unemployed between jobs. However, in the next five years, unemployment is more persistent, such that 54% (57%) of the male (female) immigrants who were unemployed in a given month are unemployed a year later. During the two periods, we find a high persistence in employment in WC jobs and a lower persistence in BC and in HT jobs. Among the three employment states, professionals in HT is the less stable occupational category with persistence rate of 47% (59%) among males (females) in the first period and, respectively, 55% (68%) in the second period. The transitions
287
Immigrants in the Israeli Hi-Tech Industry
Table 7.
Number of Annual Transitions between Labor Market Statesa. TO
HT Professional
White Collar
Blue Collar
Training Unemployed
28 (47) 5 (1) 39 (2) 6 (5) 71 (3)
17 (29) 405 (85) 612 (27) 57 (51) 771 (32)
9 (15) 29 (6) 1280 (56) 47 (42) 1144 (48)
0 (0) 9 (2) 59 (3) 0 (0) 71 (3)
5 (8) 26 (5) 312 (14) 2 (2) 348 (14)
71 (55) 29 (2) 56 (3) 13 (12) 15 (4)
26 (20) 1183 (87) 409 (25) 39 (36) 87 (24)
23 (18) 84 (6) 975 (61) 51 (47) 63 (17)
2 (2) 23 (2) 26 (2) 0 (0) 5 (1)
7 (5) 45 (3) 145 (9) 5 (5) 198 (54)
13 (59) 0 (0) 2 (0) 6 (5) 80 (3)
3 (14) 111 (81) 170 (20) 59 (45) 596 (24)
6 (27) 20 (15) 492 (59) 52 (40) 1272 (50)
0 (0) 0 (0) 7 (1) 0 (0) 34 (1)
0 (0) 6 (4) 166 (20) 14 (11) 539 (21)
FROM (A) Males – first five years HT professional (%) White collar (%) Blue collar (%) Training (%) Unemployed (%) (A) Males – second five years HT professional (%) White collar (%) Blue collar (%) Training (%) Unemployed (%) (B) Females – first five years HT professional (%) White collar (%) Blue collar (%) Training (%) Unemployed (%)
288
SARIT COHEN-GOLDNER
Table 7. (Continued ) TO
HT Professional
White Collar
Blue Collar
51 (68) 19 (3) 21 (2) 3 (8) 9 (2)
23 (31) 545 (85) 204 (15) 13 (35) 59 (14)
0 (0) 26 (4) 857 (65) 17 (46) 117 (27)
Training Unemployed
(B) Females – second five years HT professional (%) White collar (%) Blue collar (%) Training (%) Unemployed (%)
0 (0) 16 (2) 12 (1) 0 (0) 0 (0)
1 (1) 36 (6) 224 (17) 4 (11) 247 (57)
Source: Author’s calculations-Engineers survey. Actual number of individuals’ transitions between month ‘t’ and month ‘t+12’ and row percentage. a
from HT professionals to WC in the two periods are non-negligible and reflect the fluctuational nature of the HT industry during the 1990s. The transitions from training to the three employment states are substantial, while the transitions to unemployment are low. During the first period 51% (45%) of the males (females) who attended training moved to WC jobs, while 42% (40%) moved to BC jobs and 5% (5%) moved to HT professionals. Only 2% (11%) of the males (females) moved from training to unemployment during the first period. In the second period, the transitions from training to WC declines to 36% (35%) among males (females) and the transitions to BC increases to 47% (46%). About 12% (8%) of the males (females) moved from training to HT professionals during this period and 5% (11%) moved to unemployment. Overall, it seems that the transitions to low-skill BC jobs occur mainly from unemployment, while the transitions to high-skill WC and HT jobs occur also through training and employment in BC jobs.
4. CONCLUSION This paper provides a descriptive analysis of the integration of CIS immigrants in the Israeli labor market in comparison to natives and with
Immigrants in the Israeli Hi-Tech Industry
289
emphasis on their training attendance and employment in the HT industry. This description is informative for the researcher who is interested in evaluating the potential benefit for the growth of the Israeli HT industry from the imported and locally accumulated human capital of the recent immigrants. However, to further understand the link between the skills of the immigrants and the growth of the industry, one would need to formulate and estimate a model that controls for the dynamic selection of choices made by workers and firms. This challenging research is left for the future.
NOTES 1. We define immigrant as a person who was born in the CIS and arrived in Israel after 1989 at the age of 14 or above. Native is defined as a person who was born in Israel or immigrated to Israel before age 14 prior to 1989. 2. See Appendix (A) for further description of the LFS. 3. The definitions of HT occupations in the HT industry are based on Feldman and Abouganem (2002). 4. The average ‘years of schooling’ of the immigrants in Fig. 1 is 14 for both males and females. 5. Cohen-Goldner and Eckstein (2002, 2004) found that upon arrival, unemployed immigrant with no work experience in Israel has a very low probability to receive a job offer in WC jobs. The estimated probability to receive such an offer was approximately 12% per quarter for the average male immigrant and 6% per quarter for the average female immigrant. Training increased these probabilities by 50– 100%. 6. Immigrants in the HT industry are, on average, older than natives and have more years of schooling. 7. No data on training are available from the cross-sectional LFS. Therefore, the analysis in this section is based only on the engineers surveys. 8. The 2001–2002 survey was conducted by the PORI survey company under the supervision of Sarit Cohen-Goldner and Zvi Eckstein. 9. The first survey does not provide wage data. Not all individuals reported their wages in the second survey. 10. The integration of immigrant engineers in the HT industry is similar to the integration of the general population of immigrants. However, within the HT industry, there is a substantial difference between the occupational distribution of engineers and the general population of immigrants. In particular, the vast majority of engineers work as professionals in the industry. For example, in 2000, more than 80% of immigrant engineers (both among males and female) in the HT industry worked as professionals. In contrast, among the general population of CIS immigrants 60% of the males and 40% of the females in the HT industry worked as professionals (LFS, 2000). 11. Logit estimates for participation in training confirm these effects, although only age at arrival has a significant negative effect on participation.
290
SARIT COHEN-GOLDNER
12. Unemployment in the LFS (Fig. 1) includes also individuals who attend training programs. 13. Table 4 reports hazard ratios such that a coefficient greater than 1 indicates that the variable shortens the duration to training and a coefficient lower than 1 means that the variable leads to a longer duration to training. 14. Heckman and Smith (1999) studied the role of these variables as predictors in the static decision to participate in training, while we study their effect on the decision when to participate. 15. Standard errors are clustered for each individual. 16. The characteristics of the average male/female immigrant are taken from Table 1.
ACKNOWLEDGMENTS Support for the paper from the Science Technology and the Economy Program (STE), at the Samuel Neaman Institute for Advanced Studies in Science and Technology is gratefully acknowledged. I wish to thank Zvi Eckstein for valuable comments. I also thank two anonymous referees and the participants at the STE program at The Samuel Neaman Institute at the Technion. Marina Agranove, Yaniv Idid-Levi and Tali Larom provided excellent research assistance.
REFERENCES Cohen-Goldner, S., & Eckstein, Z. (2002). Labour mobility of immigrants: Training, experience, language and opportunities. CEPR Discussion paper series no. 3412. Cohen-Goldner, S., & Eckstein, Z. (2004). Estimating the return to training and occupational experience: The case of females. IZA DP no. 1225. Eckstein, Z., & Weiss, Y. (2002). The integration of immigrants from the former Soviet Union in the Israeli labor market. In: B.-B. Avi (Ed.), The Israeli economy, 1985–1998: From government intervention to market economics, essays in memory of Prof. Michael Bruno. Cambridge, MA: MIT Press. Eckstein, Z., & Weiss, Y. (2004). On the wage growth of immigrants: Israel 1990–2000. Journal of European Economic Association, 2(4), 665–695. Feldman, M., & Abouganem, M. (2002). Development of the high-tech industry in Israel 1995– 1999: Labour force and wages. Central Bureau of Statistics (Israel) Working Paper no. 1, April. Heckman, J., & Smith, J. (1999). The pre-program earning dip and the determinants of participation in social program: Implications for simple program evaluation strategies. NBER Working Paper no. 6983. Weiss, Y., Sauer, R. M., & Gotlibovski, M. (2003). Immigration, search, and loss of skill. Journal of Labor Economics, 21(3), 557–591.
Immigrants in the Israeli Hi-Tech Industry
291
APPENDIX: DATA DEFINITIONS (A). The Labor Force Survey The LFS is an annual household survey, which is conducted by the Israeli Central Bureau of Statistics (CBS). The data is collected from roughly 25,000 households that are interviewed four times over a period of 18 months. Each household is interviewed for two consecutive quarters, followed by a break for two quarters, and is interviewed again for two consecutive quarters. The LFS provides information on labor market participation, occupation, education, country of origin, year of immigration and other demographic variables as well as details on workplace.
(B). Definitions of the HT-Industry and HT-Occupations The definition of the HT-industry in this paper is similar to the definition of American Electronic Association (AEA). We based our definition on the following standard industrial classification of all economic activities (CBS, 1993): 30 – Manufacture of office and accounting machinery and computers 32 – Manufacture of electronical components 33 – Manufacture of electronic communication equipment 34 – Manufacture of industrial equipment for control and supervision, medical and scientific equipment 66 – Telecommunications 72 – Computer and related services 73 – Research and development The definition of HT-occupations is based on the following standard classification of occupations (CBS, 1994): 001 002 010 011 012 013 015
– – – – – – –
Biologists and related professionals Pharmacologists Chemists Physicists and astronomers Geologists and geophysicists Mathematicians and actuaries System analysts and related computer professionals
292
023 024 027 101 121 122 130 225
SARIT COHEN-GOLDNER
– – – – – – – –
Electrical and electronics engineers Mechanical engineers Computer engineers Physical engineering technicians Electronic engineering technicians Mechanical engineering Computer technicians and programmers Computer services managers
(C). Definitions of HT-Related Training HT-related training is defined based on the answer of the following question: ‘‘Which profession did you study in this course? (please provide as many details as possible)’’. 3 – Autocad 5 – ‘‘Year 2000’’ communications 6 – Optical fiber 7 – Electronics/Electricity and control/Electrician 14 – Technology and computer software/Software engineer/Computers/ Programming 15 – Signal processing 31 – Computer technician/Computer maintenance 33 – Physics 34 – Practical computer engineer 51 – Programming languages 52 – Communication networks 64 – Chemistry/Chemistry retraining
WHAT DO WAGE DIFFERENTIALS TELL ABOUT LABOR MARKET DISCRIMINATION?$ June E. O’Neill and Dave M. O’Neill 1. INTRODUCTION With the signing of the Civil Rights Act of 1964, discrimination in employment with respect to the hiring, promotion and pay of minorities and women became illegal in the United States.1 Yet, 40 years later, earnings differentials still persist between certain minorities and white non-Hispanics and between women and men. For example, although the ratio of black men’s earnings to those of white men and of black women’s to white women’s have increased considerably over the past 50 years, the black–white ratio was still only 78% in 2003 among men and 87% among women (Fig. 1). Hispanic–white wage differentials are larger than the black–white differential among both men and women (Figs. 2 and 3). And despite a significant narrowing in the gender gap, the ratio of women’s earnings to men’s was about 76% in 2003 (Fig. 4).2 Differentials such as these raise questions in the media and stir the ire of advocacy groups. Yet, the presence or absence of a wage gap in itself is not evidence of the presence of discrimination against a particular group in the
$
Paper prepared at the conference in memory of Tikva Darvish Lecker at Bar-Ilan University, June 27–28, 2004.
Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 293–357 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24010-4
293
294
100 95 90 85
Percent
Women
80 75 70 Men
60 55
[]
[]
50
1955
1958
1961
1964
1967
1970
1973
1976
1979
1982
1985
1988
1991
1994
1997
2000
2003
Fig. 1. Black/White Ratios of Median Annual Earnings of Full-time, Year-round Workers, by Sex, 1955–2002. Source: U.S. Bureau of the Census, Current Population Survey (CPS), Historical Income Tables. The data for 1955–1966 refer to median annual income of full-time, year-round workers instead of median annual earnings.
JUNE E. O’NEILL AND DAVE M. O’NEILL
65
110
Asian/White
Percent
100
90
80
Black/White
70
60
Hispanic/White
50 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Fig. 2. Ratios of Hourly Earnings of Asian, Black and Hispanic Men Relative to those of Non-Hispanic White Men, Ages 25–54, 1982–2003. Note: Median hourly earnings are derived from CPS microdata by dividing annual total earnings by the product of weeks worked during the year and hours usually worked per week. Earnings tabulations are restricted to those working at least 20 h a week and 8 weeks a year.
What do Wage Differentials Tell About Labor Market Discrimination?
120
295
296
120
110
Asian/White
Percent
100
90
Black/White
80
Hispanic/White 60
50 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Fig. 3. Ratios of Hourly Earnings of Asian, Black and Hispanic Relative to those of Non-Hispanic White Women, Ages 25– 54, 1982–2003. Note: Median hourly earnings are derived from CPS microdata by dividing annual total earnings by the product of weeks worked during the year and hours usually worked per week. Earnings tabulations are restricted to those working at least 20 h a week and 8 weeks a year.
JUNE E. O’NEILL AND DAVE M. O’NEILL
70
95 90 85
Percent
80 75 70 65 60
1955
[]
50
[]
55
1958
1961
1964
1967
1970
1973
1976
1979
1982
1985
1988
1991
1994
1997
2000
2003
297
Fig. 4. Female/Male Ratios of Median Annual Earnings of Full-time, Year-round Workers, 1955–2003. Source: U.S. Bureau of the Census, Current Population Survey (CPS), Historical Income Tables. The data for 1955–1959 refer to median annual income of full-time, year-round workers instead of median annual earnings.
What do Wage Differentials Tell About Labor Market Discrimination?
100
298
JUNE E. O’NEILL AND DAVE M. O’NEILL
labor market. Groups differ in the extent to which they have been subjected historically to overt discrimination. But groups also differ significantly in their work-related skills, and that difference alone would create wage differentials, even in the absence of labor market discrimination.3 Our short answer to the question posed in the title of this paper is ‘‘not very much’’. We base that conclusion on a detailed empirical analysis of the extent to which differences in skills and other productivity-related characteristics can explain observed wage gaps between racial or ethnic minorities and whites and between women and men. We find that differences in productivity-related factors account for most of the observed (unadjusted) wage differentials. This is an important finding because an unwarranted belief that employment discrimination is the primary source of wage differentials can lead to inappropriate policy decisions. Our results may well be at the low end of estimates of ‘‘unexplained’’ racial and gender wage gaps. An unexplained residual, however, is not a meaningful indicator of labor market discrimination when important aspects of productivity are not measured by the explanatory variables. We find small residual wage gaps when we utilize data from the National Longitudinal Survey of Youth (NLSY), a survey that provides superior measures of skill. The unexplained gap is larger, however, when we use the 2000 Census of Population, a data source that provides limited measures of skill. For example, years of schooling alone are a crude indicator of a person’s human capital when school quality and family background vary. The test score data available in the NLSY greatly improve our ability to measure human capital differences. The inclusion of individual test scores as an explanatory variable substantially narrows the unexplained gap by race and ethnicity. Similarly, the unexplained gender gap is significantly reduced in the NLSY analysis when detailed data are included on actual lifetime work experience and occupational and job choice. Other recent studies that utilize such superior measures of productivity also find smaller unexplained gaps.4 In this paper, we present the results of our analysis of the sources of racial, ethnic and gender wage gaps. We start with a brief discussion of economic concepts of labor market discrimination and their implications for earnings differences between groups.
2. ECONOMIC CONCEPTS OF DISCRIMINATION In his seminal work on the economic theory of discrimination, Gary Becker (1957) analyses the effects of employer prejudice on the wages of minorities.
What do Wage Differentials Tell About Labor Market Discrimination?
299
An important implication of Becker’s theory is that competitive markets can impose a penalty on a firm in the form of lower money profits when the firm discriminates against workers on the basis of anything other than productivity differences. Central to the theory is that a prejudiced employer – in Becker’s terminology, an employer with a ‘‘taste’’ for discrimination – would only be willing to hire a minority worker at a wage that is less than that of an equally productive non-minority worker. At any given wage rate for minority workers, non-discriminating firms will have lower real costs of production than discriminating firms. The ‘‘taste for discrimination’’ acts like a tax that firms practicing discrimination must pay when they hire a minority worker. Non-discriminating firms do not pay this ‘‘tax’’ and therefore employ larger numbers of minority workers. Although initially they will be able to employ minorities at wages below the value of their productivity, they will be willing to pay higher wages (up to the workers’ productivity level). In competitive markets, the demand for minority workers by employers with no taste for discrimination can mitigate and eventually even eliminate any earnings effects on minorities. The extent to which minority wages are ultimately reduced by labor market discrimination depends on the intensity and distribution of tastes for discrimination among employers and the interaction of those taste factors with market structure and production conditions. In situations where a large majority of employers are not prejudiced, the minority worker population may be able to avoid discrimination. Moreover, if non-discriminating firms were subject to production conditions that allow constant or increasing returns to scale, their ability to expand would enable them to drive out discriminating employers and hire more minority workers. But if non-prejudiced employers (or potential employers) were a minor presence in the market relative to the size of the minority population, their impact on discrimination in the overall market would be minimal; and if non-discriminating firms faced decreasing returns to scale, their potential impact on reducing the effect of discrimination would be further minimized. Different minorities likely vary in the extent to which they are subject to the effects of discrimination intensity and its interaction with market/production factors. At one extreme, the black population at one time was exposed to widespread labor market discrimination. In the pre-Civil Rights era, the vast majority of blacks lived in the South where discriminatory attitudes were prevalent and intense enough to be codified in Jim Crow laws that restricted the access of the black population to a wide array of public services, including education, as well as jobs (Donohue & Heckman, 1991; Smith and Welch, 1989; US Commission on Civil Rights, 1986). Other
300
JUNE E. O’NEILL AND DAVE M. O’NEILL
minorities (for example, Jews and Asians) may have been able to substantially avoid the effects of labor market discrimination because they belong to relatively small groups and a sufficient number of employers harbored no discriminatory feelings toward them. Becker’s model and those models that have been developed out of applications of his basic ideas all focus on the effects of prejudice in the labor market (for example, Black, 1995; Kahn, 1991). However, another class of models of discriminatory outcomes are based on the premise that employers lack information about the abilities of individual minority and non-minority workers and assume that individuals will have the average characteristics of the group to which they belong (Arrow, 1973; Aigner & Cain, 1977; Lundberg & Startz, 1983; Cain, 1986). Models of ‘‘statistical discrimination’’ suggest that individual minorities who are more skilled or productive than the group average can be discriminated against even if employers are not prejudiced against individual minority members. (Conversely, below-average majority workers would gain if their group on an average was viewed as highly productive.) Thus a firm might find that the quit rate among its women employees, on an average, was greater than that of men hired for the same job. Faced with the choice between hiring an individual woman or man of apparently equal qualifications (such as the same education) it might choose the man based on the premise that the probability of a woman quitting is higher than that of a man. However, statistical discrimination is likely to diminish as firms find it in their interest to invest in obtaining more information about the individual workers that they hire (e.g., checking references on prior employment). Moreover, once workers accumulate a track record at a firm, employers obtain direct information about individuals on which to base personnel decisions concerning pay and promotion. Statistical discrimination, like discrimination derived from prejudice, is prohibited by civil rights legislation. However, in practice it could be difficult to distinguish between the two.
3. MEASURING DISCRIMINATION It is difficult to unravel the role that labor market discrimination plays in causing earnings differentials. Direct measures of discrimination are unattainable for national samples of the population. Individual charges of employer discrimination that are brought to court provide little information about the extent of employer discrimination.5 Studies based on audit pair experiments attempt to document instances of discriminatory employer
What do Wage Differentials Tell About Labor Market Discrimination?
301
behavior. The relevance of their findings is questionable.6 Although individual instances of discrimination surely exist, they cannot be used as evidence about the extent of labor market discrimination or its effects on wages. In the absence of direct measures of discrimination, researchers investigating the effect of discrimination on race and gender differences in earnings typically have addressed a question more amenable to measurement, at least in principle, namely: to what extent can differences in productivity explain observed wage differentials? This is by no means an easy question to answer because productivity seldom can be observed directly. Instead, the recourse is to develop measures of characteristics that can serve as proxies for productivity. Survey data vary considerably in the adequacy of information they provide on the skills of workers, leaving open the possibility that important aspects of productivity may be omitted from the analysis. Certain measures of human capital, such as years of school completed, are now routinely available. But differences in years of schooling are a crude proxy for academic achievement, because schools vary considerably in quality as well as in standards for promotion, diplomas and degrees. It is difficult to obtain direct measures of cognitive skills, such as actual measures of cognitive achievement as revealed in test scores, or of skills developed through years of work experience. Among groups with a significant proportion of immigrants, ability to speak English is important. Rough measures of English language proficiency can be obtained from recent census surveys, but other aspects of acculturation are more difficult to assess. The measurement of gender differences in productivity presents a particular challenge. Women and men differ in their allocation of time between home responsibilities (particularly care of children) and market work, resulting in significant differences in lifetime work experience (Mincer, 1962). A credible analysis of the gender gap in wages, therefore, requires data on lifetime work experience, and such data are not routinely included in the major US surveys of work and earnings such as the Current Population Survey (CPS) or the decennial Census of Population.7 In addition, women’s anticipation of family responsibilities can influence choice of occupation made at an early age. Actual responsibilities can also lead to preferences for jobs that allow for more flexibility and less commitment of time and effort (Becker, 1985). Men and women therefore, may make different trade-offs between pay and job amenities. Do more women than men choose to stay home, or limit their hours of work to care for their children because of gender discrimination in the labor force? At one time discrimination
302
JUNE E. O’NEILL AND DAVE M. O’NEILL
undoubtedly played a role, although it was unlikely to have been the sole or even the main factor. Today, that explanation seems much less plausible. In this paper we utilize the NLSY, which provides a more adequate measure of cognitive skills than years of school completed (in particular, test scores for sample participants) and also provides detailed information on lifetime work experiences. We use the NLSY to conduct detailed analyses of the gender wage gap as well as the wage gap between white non-Hispanics and blacks (or Hispanics). However, the NLSY samples are not large enough to compare wage gaps for smaller minority groups and for that purpose we use the 2000 Census of Population. In this paper, we first examine wage differentials among a large crosssection of racial and ethnic groups, separately by sex, using the 2000 Census. We then turn to a series of analyses based on the NLSY, which, as noted, provides measures of important aspects of work-related skills that are unavailable in the Census data.
4. RACIAL AND ETHNIC WAGE DIFFERENTIALS: RESULTS FROM THE 2000 CENSUS We start with an overview of the factors influencing the relative wages of various racial and ethnic groups compared to those of whites using data from the 2000 Census. The analysis is confined to wage and salary workers in the age group of 25–54 years. The racial/ethnic groups identified are black non-Hispanics, American Indians, seven groups of Asians (differentiated by national origin) and seven groups of Hispanics (differentiated by national origin). Here and throughout the paper whites are always non-Hispanic whites. 4.1. Racial and Ethnic Wage Differentials among Men We use microdata from the 2000 Census to conduct OLS log wage regressions controlling for different sets of explanatory variables. Table 1 shows the log hourly wage differential between each group and the reference group of white men (given by the partial regression coefficients on the dummy variables indicating the race/national origin of each group). The unadjusted wage differentials (Model 1) vary considerably among the groups. Japanese, Asian Indian and Korean men earn about 15–25% more than white non-Hispanic men. Filipino and Chinese men earn 4–10% more
Model 1 Coefficient
Model 2 t-stat
Coefficient
Model 3 t-stat
Coefficient
Model 4 t-stat
Coefficient
t-stat
Race/Ethnicity indicators 0.253 0.273
24.78 80.64
0.212 0.273
21.75 82.97
0.131 0.182
14.59 59.31
0.125 0.181
13.96 59.29
Chinese Japanese Asian Indian Korean Vietnamese Filipino Other Asian
0.037 0.241 0.227 0.143 0.034 0.099 0.166
3.57 13.44 20.27 6.29 1.71 5.56 12.85
0.104 0.131 0.166 0.089 0.045 0.053 0.226
10.46 7.62 15.49 4.10 2.36 3.13 18.32
0.198 0.011 0.039 0.038 0.005 0.013 0.190
21.44 0.71 3.93 1.91 0.31 0.80 16.64
0.101 0.068 0.037 0.003 0.064 0.016 0.125
10.48 4.30 3.52 0.17 3.63 1.00 10.71
Mexican Puerto Rican Cuban Dominican Other Central American South American Other Hispanic
0.448 0.220 0.181 0.418 0.489 0.241 0.318
122.36 22.24 12.48 21.47 44.74 18.17 41.70
0.439 0.262 0.148 0.504 0.510 0.301 0.304
118.82 27.57 10.69 27.08 48.76 23.74 41.45
0.200 0.135 0.110 0.310 0.256 0.243 0.145
53.56 15.37 8.60 18.03 26.14 20.76 21.32
0.102 0.087 0.016 0.190 0.132 0.132 0.081
22.50 9.62 1.21 10.88 12.85 10.92 11.44
303
American Indian Black non Hispanic
What do Wage Differentials Tell About Labor Market Discrimination?
Table 1. Log Hourly Wage Differentials Between Men of Detailed Race/Ethnicity and White Non-Hispanic Men, Ages 25–54, in 1999, Controlling for Different Sets of Explanatory Variables (2000 Census).
304
Table 1. (Continued ) Model 1 Coefficient Control variables Age Region, MSA, central city Schooling Works part-time (20–34 hours a week) Class of worker Years since migration to US English-speaking ability
Model 2 t-stat
Coefficient
X X
Model 3 t-stat
Coefficient
X X X X X
Model 4 t-stat
Coefficient
t-stat
X X X X X X X
JUNE E. O’NEILL AND DAVE M. O’NEILL
Note: The log wage differentials are the partial regression coefficients of the dummy variables (0,1) for each of the racial/ethnic groups listed above, from a series of mutiple regressions shown as Models 1–4. The other variables controlled for are also listed above for each model. The sample, excluding active military and unincorporated self-employed persons, is restricted to wage and salary workers who worked 20 h or more a week and 26 weeks or more a year. Hourly wages are obtained by dividing annual earnings by the product of weeks and hours worked during the year. Source: Census 2000, Public Use Microdata Sample (PUMS), 1%.
What do Wage Differentials Tell About Labor Market Discrimination?
305
than white men, while the group ‘‘other Asian’’ (including Thai, Hmong, Pakistani and Cambodian groups) earn about 15% less than white men. All of the Hispanic groups earn less than white men and less than the Asian groups as well. Mexicans, Dominicans and other Central Americans have the lowest earnings of any group shown – about half of those of white men. Cubans and Puerto Ricans have the highest earnings among Hispanics, but still earn about 20% less than white men. Black and American Indian men earn 25% less than white men. Adjusting for geographic division and metropolitan/ central city location and age (Model 2) reduces some of the relative advantage of Asian groups because they live in high-wage areas. The wage differentials are substantially changed, however, when education variables are added to the equation (Model 3). Asian groups have very high levels of education. More than half of Asian men are college graduates or hold higher degrees. Their earnings advantage is eliminated once education is taken into account. Hispanic groups, on the other hand, have relatively low levels of schooling. (Almost half of Hispanic men have not completed high school and only 9% are college graduates.) Consequently, their earnings converge significantly with those of white men when education variables are added to the model. The Mexican differential is cut in half, although the change for other Hispanic groups with stronger education backgrounds is less dramatic. The black–white wage gap, and, even more so the American Indian–white differential, are also reduced when account is taken of differences in years of schooling. A relatively large proportion of Asians and Hispanics are migrants. In Model 4, we add variables indicating years since migrating to the United States and a crude indicator of English language proficiency (self-reported). The addition of these variables increases the wages of Hispanics and Asians relative to whites. At this final step, the wages of the Asian groups are mostly either slightly above or below those of white men, with some variation. Chinese and the residual group of ‘‘other Asian’’ men earn about 10% less than white men; Japanese and Vietnamese men earn about 7% more. The gap for Hispanic men is sharply reduced for all groups but still averages about 10% below that of white non-Hispanic men. But there is still considerable variation by national origin. The gap for Dominican men is the highest (19%); the gap for Cuban men is eliminated. Groups with a significant proportion of migrants present particular difficulties for analysis because cultural differences among them that influence the speed of assimilation are only partly captured by measures of schooling and crude self-reported measures of English-speaking ability. Different
306
JUNE E. O’NEILL AND DAVE M. O’NEILL
cohorts of migrants from the same country can differ because of selection factors. The second generation and earlier generations of immigrants are likely to be more assimilated. We present additional analysis of Hispanic and black men below using the superior measures of skills available in the NLSY data. 4.2. Racial and Ethnic Wage Differentials among Women Table 2 replicates for women the analysis of Table 1 and compares the wages of minority women with those of white non-Hispanic women. Although the patterns of wage differentials among the different ethnic/racial groups of women are similar to those of men, the level of the differentials are, for the most part, considerably smaller. Thus the unadjusted log wage gap between black and white men is 0.273 and between black and white women it is 0.112. The wage differentials between white non-Hispanic women and each group of Hispanic women are also much smaller than they are for men. After adjusting for schooling, migration and English-speaking skills, the differentials among women are further reduced and are mostly on the order of 5% for all groups except Dominicans and other Central Americans. The Asian-white differentials are similar for women and men. Asian women, like Asian men, typically earn more than their white counterparts because of their relatively high education levels and greater geographic concentration in high-wage cities and regions. Once we control for differences in region, schooling, immigration and language proficiency, as in Model 4, these positive wage differentials are erased and Asian women are found to earn about the same wage rate as white women.
5. NLSY RESULTS: DIFFERENTIALS BY RACE AND HISPANIC ETHNICITY We turn to the NLSY for a more intensive analysis of the black–white and Hispanic–white wage gaps among male and female workers and then in the next section, to the female–male wage gap. The NLSY cohort was first interviewed in 1979 (in the age group of 14–22 years) and was again interviewed each year through 1994 and every other year since then. Detailed information is provided on lifetime work experience, education and many other individual characteristics and behaviors of relevance to labor market outcomes.
Model 1 Coefficient
Model 2 t-stat
Coefficient
Model 3 t-stat
Coefficient
Model 4 t-stat
Coefficient
t-stat
Race/Ethnicity indicators 0.199 0.112
19.02 37.05
0.156 0.137
15.50 45.43
0.096 0.070
10.64 25.84
0.095 0.070
10.60 25.86
Chinese Japanese Asian Indian Korean Vietnamese Filipino Other Asian
0.149 0.239 0.213 0.083 0.073 0.192 0.067
14.28 13.16 15.09 3.81 3.47 11.42 4.99
0.030 0.117 0.108 0.032 0.093 0.132 0.146
2.94 6.66 7.94 1.52 4.61 8.21 11.23
0.085 0.007 0.048 0.051 0.024 0.015 0.069
9.41 0.46 3.92 2.78 1.31 1.04 5.95
0.010 0.023 0.024 0.015 0.050 0.015 0.009
1.02 1.45 1.87 0.83 2.81 1.03 0.76
Mexican Puerto Rican Cuban Dominican Other Central American South American
0.297 0.077 0.037 0.305 0.374 0.112
65.35 7.82 2.37 15.85 29.01 8.01
0.334 0.173 0.030 0.464 0.473 0.216
73.26 18.21 2.00 25.00 37.99 16.03
0.117 0.065 0.018 0.263 0.222 0.154
27.50 7.72 1.31 15.88 19.78 12.80
0.055 0.027 0.058 0.150 0.117 0.053
11.43 3.04 4.21 8.91 10.03 4.29
307
American Indian Black non-Hispanic
What do Wage Differentials Tell About Labor Market Discrimination?
Table 2. Log Hourly Wage Differentials Between Women of Detailed Race/Ethnicity and White NonHispanic Women, Ages 25–54, in 1999, Controlling for Different Sets of Explanatory Variables (2000 Census).
308
Table 2. (Continued ) Model 1 Coefficient 0.226
t-stat 28.75
Coefficient 0.241
Model 3 t-stat 31.71
X X
Coefficient 0.098
Model 4 t-stat 14.44
X X X X X
Coefficient 0.054
t-stat 7.67
X X X X X X X
Note: The log wage differentials are the partial regression coefficients of the dummy variables (0,1) for each of the racial/ethnic groups listed above, from a series of mutiple regressions shown as Models 1–4. The other variables controlled for are also listed above for each model. The sample, excluding active military and unincorporated self-employed persons, is restricted to wage and salary workers who worked 20 h or more a week and 26 weeks or more a year. Hourly wages are obtained by dividing annual earnings by the product of weeks and hours worked during the year. Source: Census 2000, Public Use Microdata Sample (PUMS), 1%.
JUNE E. O’NEILL AND DAVE M. O’NEILL
Other Hispanic Control variables Age Region, MSA, central city Schooling Works part-time (20-34 hours a week) Class of worker Years since migration to US English-speaking ability
Model 2
What do Wage Differentials Tell About Labor Market Discrimination?
309
One variable unique to the NLSY and of considerable value to the measurement of skill is the individual’s score on the Armed Forces Qualification Test (AFQT), administered to nearly all survey participants who were in the age group of 15–23 years. The AFQT is a component of a larger test of vocational aptitude first developed by the military many decades ago to determine eligibility for service and for job placement within the service. The test is a measure of verbal and mathematical skills and has been validated over the years as an effective and unbiased predictor of job performance.8 In common with any test of skill or achievement, it is influenced by genetic factors as well as by environmental factors such as the quantity and quality of schooling and the home environment from early childhood.9 For our purposes, the relative importance of these causal factors is irrelevant since our objective is to determine the relative importance of labor market discrimination in generating earnings differences. Our NLSY sample is derived from the 2000 survey when the cohort was 35–43 years of age. At this point in the life cycle, the major period of wage growth has already occurred and differences in on-the-job investment, including those potentially related to discrimination, already would have had an effect. The sample includes 5,600 wage and salary workers. Blacks and Hispanics were over-sampled, allowing adequate samples for analysis of these groups. Because the cohort sample was drawn in 1979, the 2000 survey results do not include recent immigrants. Analysis of the extent to which earnings differences between groups are explained by differences in characteristics can be executed in several ways. The wage gaps shown in Tables 1 and 2 are derived from log wage regressions in which a set of dummy (0,1) variables are used to indicate the race/ ethnicity of different groups. The partial regression coefficients on the dummy variables are interpreted as reflecting the wage differential between each group and the reference group of white men (or white women in the female regressions). The underlying assumption is that the effect of relevant characteristics (other than race/ethnicity) on wages can be approximated by the average effect for all groups included in the sample. One issue that arises, however, is the extent to which differences in the effects of explanatory variables on earnings vary in important ways among groups. For example, the effect on earnings of an additional year of schooling or of work experience may differ between blacks and whites. If it is lower for blacks, the question arises whether that difference reflects employer discrimination. To address that issue, we also conduct separate regressions for both blacks and whites and Hispanics and whites, and present the results of decomposition analysis based on both sets of partial regression coefficients.
310
JUNE E. O’NEILL AND DAVE M. O’NEILL
5.1. Racial and Ethnic Differentials among NLSY Men We first show the results of a series of multiple regressions (four models) using the dummy variable approach to identify log wage differences between groups (Table 3).10 Separate regressions were run for men of all education levels combined as well as for two education groups: those with no more than a high school education; and those who are college graduates or have post-college schooling. The highlights are as follows: Black/White differences. The unadjusted log hourly wage differential was 0.339 between black and white men in 2000 when the NLSY cohort was 35–43 years of age. The gap is smaller within education group ( 0.244 for the high school group and 0.262 for college graduates). When age and geographic location variables are included in the regression (Model 1), the gap is reduced because a much larger proportion of black than of white men live in the South where wages on an average are lower for both races. The addition to the model of detailed level of schooling reduces the gap for all men to 0.186, now similar to that of the two education groups (Model 2). As shown in Table 4, the mean percentile AFQT score for black men was 24 compared to 55 for white men, and as demonstrated below, AFQT has a large effect on wages for both blacks and whites. After adding the AFQT percentile score (Model 3), the black–white log wage difference is dramatically reduced: to 0.062 for all men, to 0.075 for the high school group and to 0.05 for college graduates (no longer significant). Our findings with respect to the explanatory power of the AFQT variable are similar to those of Neal and Johnson (1996) and O’Neill (1990) who analyzed the same NLSY cohort when the survey participants were still in their 20 s. Neal and Johnson, however, selected the younger portion of the cohort, and did not include education and differ in their measurement of AFQT scores.11 In Model 4, we add two components of work experience: total weeks of civilian employment since age 18 divided by 52 (full-year equivalents) and total weeks served in the military since 1978, also divided by 52. Close to 17% of black men were ever in the military compared to 8.5% for Hispanic men and 9.6% for white men. On an average, black men have been in the military 0.8 years compared to 0.5 years for white men and 0.4 years for Hispanic men. However, black men have accumulated fewer years of civilian employment than white or Hispanic men (close to two years less than white men and 1.4 years less than Hispanic men). Consequently, the total lifetime employment of black men is lower than that of the other two groups (Table 4). With the addition of work experience (Model 4), the black–white wage gap falls to near zero for the total sample as well as for the two
Black–White and Hispanic–White Log Hourly Wage Gap among NLSY Men, Ages 35–43 in 2000, Controlling for Different Sets of Explanatory Variables. Black–White Differential Total
Unadjusted log wage differential Log wage differential controlling for (1) Age, MSA, central city, region (2) Variables in (1) plus schooling (3) Variables in (2) plus AFQT (4) Variables in (3) plus Weeks worked in civilian job since age 18 C 52 Weeks worked in military since 1978 C 52
HS graduate or less
Hispanic–White Differential College graduate or more
Total
HS graduate or less
College graduate or more
0.339
0.244
0.262
0.198
0.086
0.059
0.277
0.192
0.227
0.205
0.094
0.040
0.186
0.190
0.193
0.089
0.068
0.040
0.062
0.075
0.050
0.021
0.003
0.019
0.009
0.019
0.029
0.031
0.001
0.014
311
Note: The log wage differentials are partial regression coefficients of dummy (0, 1) variables for black (Hispanic) from a series of OLS regressions containing the explanatory variables noted. For each racial/ethnic comparison, regressions were conducted for the following: total (all education levels); HS graduate or less; college graduate or higher. The reference group is white non-Hispanic. The analysis is restricted to wage and salary workers. The statistical significance of the black and Hispanic coefficients is indicated as follows (two-tailed test): significant at the 5% level or less. significant at the 10% level. Source: National Longitudinal Survey of Youth (NLSY79).
What do Wage Differentials Tell About Labor Market Discrimination?
Table 3.
312
Table 4.
Means and Partial Regression Coefficients of Explanatory Variablesa from Separate Log Wage Regressions for Black, White and Hispanic Men Aged 35–43 in 2000 (NLSY). Mean Black
Hispanic
M1 Coefficient
Education and skill level o10 years 10–12 years (no diploma or GED)b HS graduate (diploma) HS graduate (GED) Some college BA or equivalent degree MA or equivalent degree Ph D or professor degree
Black M2
t-stat
Coefficient
Hispanic
M1 t-stat
M2
M1
Coefficient
t-stat
Coefficient
t-stat
0.069
0.80
0.024
0.30
––
––
––
––
0.043
0.041
0.093
0.051
0.68
0.036
0.49
0.083
0.149
0.198
––
––
––
––
0.328
0.358
0.274
0.064
1.33
0.009
0.19
0.072
1.51
0.005
0.041
0.079
0.062
0.018
0.24
0.031
0.43
0.042
0.62
0.216 0.207
0.239 0.109
0.264 0.079
0.236 0.419
4.42 7.31
0.215 0.427
4.13 7.66
0.205 0.335
0.059
0.021
0.019
0.524
7.14
0.561
7.84
0.023
0.004
0.012
0.645
6.50
0.780
8.00
Coefficient
M2 t-stat
Coefficient
t-stat
0.064
0.81
0.082
1.08
––
––
––
––
0.12
0.007
0.12
0.063
1.10
0.078
1.22
0.080
0.87
0.077
-0.89
3.76 4.88
0.151 0.294
2.89 4.51
0.085 0.355
1.32 3.77
0.068 0.369
1.11 4.13
0.634
5.29
0.624
5.48
0.465
2.94
0.484
3.23
1.302
5.07
1.359
5.58
0.593
2.95
0.774
4.02
JUNE E. O’NEILL AND DAVE M. O’NEILL
White
White
Lifetime work experience (Year equivalents) Weeks worked in civilian job since age 18 C 52 Weeks worked in military since 1978 C 52 Adjusted R2 Dependent mean (log hourly wage) Sample size
5.538
2.411
3.360
17.828
15.865
0.483
0.835
0.046
0.039
6.49
17.279
0.047
0.436
0.033
0.296 2.898
1416
7.63
0.337
0.058
6.68
0.048
5.80
9.17
0.040
4.31
0.028
0.287
0.359 2.559
759
0.059
6.04
0.046
4.91
9.20
0.049
7.55
4.00
0.036
2.89
0.262
0.335 2.700
519
Source: National Longitudinal Survey of Youth (NLSY79). Model also controls for age, central city MSA and region. The analysis is restricted to wage and salary workers employed within the past month. b Reference group. a
What do Wage Differentials Tell About Labor Market Discrimination?
AFQT percentile score ( .10)
313
314
JUNE E. O’NEILL AND DAVE M. O’NEILL
education specific samples.12 (But the effect is larger for the high school sample than it is for college graduates among whom the racial gap in employment is also small.) Is it appropriate to include work experience in an analysis of the wage gap that aims to determine the role of employer discrimination? It would be inappropriate if employer discrimination was an important reason for the lower employment of black men. However, other factors appear to be much more important determinants of employment differences. The relative decline in the employment of young black men, particularly high school dropouts, that started in the 1970s and continued in the 1980s appears to have been related to the broader decline in demand for low-skilled workers (Bound & Freeman, 1992) and also to increased crime and incarcerations. Incarceration directly reduces normal work experience and also makes it harder to obtain employment when out of jail. The labor force interruptions related to incarceration may depreciate work-related skills and a job applicant with a criminal record may well be regarded as a risky hire. In our NLSY sample, as of 2000, close to 13% of black men had been interviewed in jail in at least one of the survey years (compared to 6% of Hispanics and 3% of whites), which likely accounts in part for the lower amount of work experience accumulated by blacks since age 18.13 Hispanic/White differences. In our analysis of wage differences in the 2000 Census reported above, we found that the relatively low years of schooling received by Hispanics is a major factor explaining their relatively low earnings. The importance of education differentials is also apparent in the analysis of the NLSY cohort. We again start with results from Table 3 using dummy variables to identify log wage differences. The unadjusted differential between Hispanic and white non-Hispanic men is smaller than the unadjusted black–white gap ( 0.198 overall); and within the two broad education groups it is 0.086 for those with no more than a high school diploma and only 0.059, a statistically insignificant difference, for college graduates. Adding age and geographic controls has little effect,14 but adding detailed schooling reduces the overall differential by more than half and reduces the gap for the high school group by about 2 percentage points. Hispanics, on an average, scored about 20 percentile points lower than white non-Hispanics on the AFQT (Table 4). When AFQT scores are included as explanatory variables in the regression, the log wage gap for Hispanic men is no longer either statistically or practically significant (Model 3). The addition of work experience has no effect on the outcome. The NLSY data suggest that differences in schooling and scores on the AFQT explain most of the difference in hourly pay between black and white
What do Wage Differentials Tell About Labor Market Discrimination?
315
men and all of the pay difference between Hispanic and white men. The 2000 Census data indicate larger residual wage gaps mainly because they provide no standardized measure of actual attainment of cognitive skills. Years of school completed can be a poor proxy for actual educational attainment when standards for promotion and the attainment of diplomas and degrees vary widely. The AFQT provides a standardized measure of attainment. Without the AFQT variable, the census and the NLSY indicate about the same adjusted black–white wage gap. In fact, comparing the results of models that include only age, geographic location and schooling, we find that the black–white log wage gap using 2000 Census data is 0.182 (Model 3 of Table 1) and the gap is 0.186 using NLSY data (Model 2 of Table 3).15 In sum, we find that differences in years of schooling and, more importantly, AFQT scores, explain most of the black–white wage gap among men and all of the Hispanic–white wage gap. When years of work experience are included in the regression, the black–white gap is virtually closed. The question remains, however, whether these results are reliable or instead reflect selection effects, bias in the explanatory variables, omitted variables or other problems that typically confound statistical analysis of wage differentials. We later investigate the effects of sample selection, issues related to the use of the AFQT, and endogeneity problems. Here, we begin to address the issue of tainted variables by examining the market returns to work experience, education and AFQT scores in separate log wage regressions for blacks, whites and Hispanics. Lower returns to additional years of work experience and education (and less plausibly, to higher scores on the AFQT) for minorities than for whites could be evidence of employer discrimination that might discourage investment in workrelated skills. We have conducted separate regressions by race and Hispanic origin and decompose the results using alternatively, coefficients from the minority and white regressions to weight the differences in characteristics. 5.1.1. Decomposition Results for Men Table 4 displays means and coefficients of the variables used in separate regressions for black, white and Hispanic men. Regression results are shown for two models. The first model includes only the AFQT percentile score and schooling (plus controls for age and geographic location). These are the same specifications as for Model 3 in Table 3. The second model adds cumulated civilian and military work experience (same specifications as Model 4 in Table 3).
316
JUNE E. O’NEILL AND DAVE M. O’NEILL
The differential in AFQT scores is again a key factor contributing to the black–white and Hispanic–white wage gaps. As measured by the regression coefficients shown in Table 4, the return to a 10 percentile point increase in the AFQT score is larger for black and Hispanic men than it is for white men, suggesting that employers recognize and reward skill among minority men at least to the same extent as they do among white men. Holding constant education level in 2000, a 10 percentile point increase in the AFQT score increases the wage rates of black and Hispanic men by about 6% and white men by about 5% in Model 1. In Model 2 (which also includes work experience), the return to AFQT is slightly smaller for all groups, presumably because AFQT scores are correlated with work experience. However, the same pattern by race is maintained and the coefficients remain robust and significant. At least equally strong relation between the AFQT and wage rates among blacks as for whites is good evidence that the AFQT provides an unbiased measure of skills.16 The question of bias in the AFQT, however, has also been analyzed more directly by the Department of Defense, which uses it extensively as a tool for assigning military personnel to occupational training and tasks. Such tests have concluded that the AFQT predicts black performance as well as it does white performance.17 Most men have at least a high school diploma or a GED (87% among whites, 81% among blacks, but dropping to 71% among Hispanics). The differences are more pronounced at the post-secondary level where white men are much more likely to graduate from college than black or Hispanic men. Twenty-nine percent of white men are college graduates or more compared to 13% of black men and 11% of Hispanic men. Holding AFQT constant, increases in schooling through high school do not have a significant effect on earnings for any group. However, the wage returns to college graduation and to attainment of higher degrees are large and roughly similar for all groups. White men have a higher return to college graduation while black men have higher returns to an MA and to the PhD or professional degree level. With regard to the return to work experience (Model 2 in Table 4), holding constant education and AFQT, the wage gain associated with an additional year of civilian experience is somewhat lower for blacks than for the other groups: 0.040 for black men, 0.047 for white men and 0.049 for Hispanic men. The return to a year of military service is lower than the return to a year of civilian work experience for all three groups.18 The small black–white differences in work experience coefficients may be due to discontinuities in black male employment. When we add a variable indicating jail time, the work experience coefficients converge (not shown).
What do Wage Differentials Tell About Labor Market Discrimination?
317
The regression decomposition results detailed in Table 5 are based on the characteristics and regression coefficients for black and white men displayed in Table 4. Black–white differences in the mean value of each characteristic are weighted alternatively by the black (or white) regression coefficients from Models 1 and 2 and the weighted differences are then summed to obtain the amount of the wage gap explained by the particular model and characteristic differences. The same procedure is followed for the Hispanic– white wage differential. The results are similar to the results shown in Table 3, which uses the dummy variable approach to identify the wage effect of race and Hispanic origin. Most or all of both the black/white wage gap and the Hispanic/white gap are explained by differences in the basic measures of skill included in Model 1 (AFQT and schooling, plus demographic controls – age, region, Metropolitan Statistical Area (MSA), central city). Moreover, a larger share of the gap is explained when the minority coefficients are used as the weights. The basic variables included in Model 1 explain 0.315 of the 0.339 white–black log wage gap when black coefficients are used as weights, and 0.245 of the gap when white coefficients are used. The white–Hispanic gap is over-explained with Model 1 specifications using Hispanic coefficients and almost fully explained when white coefficients are substituted. The inclusion of work experience in Model 2 raises the explained amount of the white– black log wage gap and has no effect on the white–Hispanic gap. Expressed as ratios of hourly wages (the exponentiated log wage gap), the unadjusted black/white ratio is 71%. The adjusted ratio using black coefficients is 98% under Model 1 specifications and 102% under Model 2. Using white coefficients, the adjusted black/white wage ratio is 91% based on Model 1 and 96% under Model 2. The Hispanic/white unadjusted hourly wage ratio is 82%. Adjusted using Hispanic coefficients it is 103% and using white coefficients it is 98% with no difference between the models.
5.2. Racial and Ethnic Differentials among NLSY Women Using the NLSY, we conducted similar analyses of the black–white and Hispanic–white wage gaps for women as for men and the results are displayed in Tables 6–8. Once again we first ran log wage regressions for all women (and separately for women with high school educations or less, and for women who are college graduates) and use the partial regression coefficients of dummy variables indicating black race and Hispanic origin to estimate log wage differentials between these groups and the reference group of white women.
White–Black and White–Hispanic Wage Gaps: Decomposition Results for Men (NLSY). White–Black Differential
Unadjusted log wage gap Total explained by model Unexplained log wage gap
Using white male coefficient
M1
M1
M2
Using Hispanic male coefficient
Using white male coefficient
M1
M2
M1
M2
0.0622
0.0589
0.0354
0.0334
0.0282
0.0292
0.0004
0.0079
0.1800 0.0731
0.1504 0.0714 0.0691
0.1435 0.0663
0.1204 0.0713 0.0810
0.1276 0.0709
0.1001 0.0741 0.0286
0.1000 0.0768
0.0839 0.0771 0.0275
0.3387
0.3387
0.3387
0.3387
0.1982
0.1982
0.1982
0.1982
0.3153
0.3499
0.2451
0.3061
0.2267
0.2321
0.1764
0.1805
0.0234
0.0112
0.0936
0.0326
0.0285
0.0339
0.0218
0.0177
Unadjusted minority/ white hourly wage ratio
71.3
71.3
71.3
71.3
82.0
82.0
82.0
82.0
Adjusted minority/ white hourly wage ratio
97.7
101.1
91.1
96.8
102.9
103.4
97.8
98.2
Note: Decomposition results shown are derived from results of separate regressions for men aged 35–43 race and by model using NLSY79 data from the 2000 survey. See Table 4 for variable means and coefficients. Hourly wages are the exponentiated hourly log wages. Source: National Longitudinal Survey of Youth (NLSY79).
JUNE E. O’NEILL AND DAVE M. O’NEILL
Log wage gap attributable to Age, region, central city, MSA AFQT Education Lifetime work experience
White–Hispanic Differential
Using black male coefficient M2
318
Table 5.
Black–White and Hispanic–White Log Hourly Wage Gap among NLSY Women, in 2000, Controlling for Different Sets of Explanatory Variables. Black–White Differential Total
Unadjusted log wage differential Log wage differential controlling for: (1) Age, MSA, Central City, Region (2) Variables in (1) plus schooling (3) Variables in (2) plus AFQT (4) Variables in (3) plus: age at 1st birth o30 (0,1) age at 1st birth Z30 (0,1) (5) Variables in (5) plus Labor force (L.F.) withdrawal due to family responsibilities (0.1), Weeks worked in civilian job since age 18 C 52, Weeks worked in military since 1978 C 52 Weeks part-time (PT) C total weeks worked since age 22
HS Graduate or less
Hispanic–White Differential
College Graduate or more
Total
HS Graduate or less
College Graduate or more
0.189
0.155
0.159
0.092
0.058
0.057
0.161 0.096 0.040 0.045
0.101 0.087 0.055 0.062
0.139 0.117 0.035 0.028
0.124 0.030 0.070 0.074
0.094 0.041 0.063 0.065
0.031 0.013 0.070 0.082
0.052
0.087
0.054
0.060
0.045
0.090
319
Note: The log wage differentials are partial regression coefficients of dummy (0, 1) variables for black (Hispanic) from a series of OLS regressions for women containing the explanatory variables noted. For each racial/ethnic comparison, regressions were conducted for the following: total (all education levels); HS graduate or less; college graduate or higher. The reference group is white non-Hispanic. The analysis is restricted to wage and salary workers. The statistical significance of the black and Hispanic coefficients is indicated as follows (two-tailed test): significant at the 5% level or less. significant at the 10% level. Source: National Longitudinal Survey of Youth (NLSY79).
What do Wage Differentials Tell About Labor Market Discrimination?
Table 6.
320
Table 7.
Means and Partial Regression Coefficients of Explanatory Variablesa, Log Wage Regressions for Black, White and Hispanic Women Aged 35-43 in 2000 (NLSY). Mean White
White
Black Hispanic
M1 Coefficient
Fertility related variables Age at 1st birth o30 (0,1) Age at 1st birth Z30 (0,1)
M2 t-stat
M1
Hispanic M2
Coefficient t-stat Coefficient t-stat
M1
Coefficient t-stat
M2
Coefficient t-stat
Coefficient t-stat
0.018 0.082
0.028 0.112
0.075 0.146
0.261 ––
2.51 ––
0.180
1.83
0.069 ––
0.75 ––
0.015
0.17
0.130 ––
1.55 ––
0.053
0.70
0.326
0.293
0.240
0.042
0.85
0.033
0.71
0.148
3.10
0.034
0.75
0.102
1.60
0.001
0.02
0.036
0.053
0.057
0.087
1.10
0.071
0.95
0.025
0.34
0.018
0.27
0.071
0.77
0.083
0.99
0.260 0.197
0.365 0.122
0.342 0.085
0.163 0.378
3.10 6.45
0.082 0.280
1.64 4.95
0.213 0.352
4.34 5.61
0.072 0.198
1.50 3.27
0.208 0.418
3.35 4.67
0.114 0.334
1.99 4.07
0.074
0.025
0.047
0.504
7.20
0.386
5.75
0.542
5.35
0.363
3.76
0.485
4.54
0.403
4.09
0.008
0.004
0.008
0.841
5.61
0.736
5.16
0.726
3.01
0.550
2.43
0.734
3.35
0.885
4.41
5.298
2.447
3.006
0.042
6.64
0.031
5.22
0.080
9.69
0.063
7.96
0.070
7.19
0.044
4.75
0.644
0.752
0.750
0.037
1.10
0.029
0.77
0.126
2.42
0.143
0.080
0.098
0.139
3.42
0.032
0.58
0.103
1.46
JUNE E. O’NEILL AND DAVE M. O’NEILL
Education and skill level o10 years 10–12 years (no diploma or GED)b HS graduate (diploma) HS graduate (GED) Some college BA or equivalent degree MA or equivalent degree PhD or professor degree AFQT percentile score ( .10)
Black
0.496
0.579
0.644
Lifetime work experience Weeks worked in 16.453 14.478 14.999 civilian job since age 18 C 52 Weeks worked in 0.045 0.096 0.051 military since 1978 C 52 Weeks PT C total 0.169 0.097 0.120 weeks worked since age 22 Adjusted R2 Dependent mean (log hourly wage) Sample size
0.255
3.37
0.043
1.32
0.139
3.12
0.029
8.09
0.031
8.97
0.034
7.59
0.026
1.14
0.058
3.06
0.050
1.84
0.182
2.80
0.287
2.85
0.061
0.52
0.341
0.323
0.408
0.344
0.460
2.606
2.417
2.514
1358
854
492
Source: National Longitudinal Survey of Youth (NLSY79). Model also controls for age, central city, MSA and region. b Reference group. a
0.103
What do Wage Differentials Tell About Labor Market Discrimination?
L.F. withdrawal due to family responsibilities (0,1)
321
White–Black and White–Hispanic Log Hourly Wage Gap in 2000: Decomposition Results for Women, Ages 35–43, in 2000. White–Black Differential Using black female coefficient
Unadjusted log wage gap Total explained by model Unexplained log wage gap Unadjusted minority/white hourly wage ratio Adjusted minority/white hourly wage ratio
M2
0.0472
0.0690
0.2286 0.0399
0.1786 0.0289 0.0011 0.0036
Using white female coefficient M1
0.1891 0.3156 0.1265
Using Hispanic female coefficient
Using white female coefficient
M2
M1
M2
M1
M2
0.0128
0.0263
0.0249
0.0357
0.0289
0.0316
0.1193 0.0452
0.0893 0.0365 0.0048 0.0086
0.1607 0.0573
0.1018 0.0400 0.0087 0.0207
0.0959 0.0628
0.0718 0.0437 0.0024 0.0154
0.0367 0.1891 0.3157 0.1266
White–Hispanic Differential
0.0434 0.1891 0.1774 0.0117
0.1891 0.2089 0.0198
0.0465 0.0919 0.1932 0.1013
0.0919 0.1648 0.0729
0.0337 0.0919 0.1298 0.0379
0.0919 0.1354 0.0435
82.8
82.8
82.8
82.8
91.2
91.2
91.2
91.2
113.5
113.5
98.8
102.0
110.7
107.6
103.9
104.4
Note: Decomposition results shown are derived from results of separate regressions for women aged 35–43 by race and by model using NLSY79 data from the 2000 survey. See Table 7 for variable means and coefficients. Hourly wages are the exponentiated hourly log wages. Source: National Longitudinal Survey of Youth (NLSY79).
JUNE E. O’NEILL AND DAVE M. O’NEILL
Log wage gap attributable to Age, region, central city, MSA AFQT Education Fertility related variables L.F. withdrawal due to family responsibilities Lifetime work experience
M1
322
Table 8.
What do Wage Differentials Tell About Labor Market Discrimination?
323
We present a series of models, each adding new groups of independent variables (Table 6). In addition to the variables used in our analysis of racial and ethnic differences among men, we include variables that are relevant to women and may have differential effects by race and Hispanic origin. Because the age of first birth is related to education and career formation, we include a variable indicating if the woman had a first birth before age 30 and another indicating if she was at least 30 at time of first birth. (Never had a birth is the omitted category.) We also add to the work experience variables a measure of the proportion of lifetime weeks worked that were part-time and another that indicates whether the person ever had a spell out of the labor force due to family responsibilities. Similar to the analysis of Census 2000 data, the initial unadjusted log wage gaps shown in Table 6 are generally smaller for women than for men. The unadjusted log wage gap for black women (compared to the white nonHispanic reference group) is 0.189. However, similar to the pattern for men, the gap falls by half when age, geographic location and education are included (Model 2). The inclusion of AFQT eliminates the negative gap (Model 3) and creates a small positive gap (0.04). The addition of fertility and work experience variables somewhat increases the positive wage gap (Models 4 and 5). The pattern of the racial wage gap among women with no more than a high school education resembles that for all women (second column in Table 6). Among college graduates the gap declines sharply and becomes insignificant, but remains negative in sign. The unadjusted Hispanic–white log wage gap among women is 0.092. Adding age, geographic location and schooling reduces the Hispanic–white wage gap for all education groups combined by two-thirds (Model 2 compared to the unadjusted gap). The remaining differential is statistically insignificant and of insignificant magnitude as well. The addition of AFQT scores (Model 3) reverses the Hispanic–white wage gap from negative to positive for all Hispanic education groups, including college graduates. 5.2.1. Decomposition Results for Women Results of a regression decomposition analysis are shown in Table 8 and the underlying variable means and coefficients from separate regressions for white, black and Hispanic women are provided in Table 7. The differences in basic skill characteristics among women by race and ethnicity are similar to those observed among men. Black women are almost as likely as white women to have completed at least high school (90% versus 86%) while that percentage for Hispanic women is only 78%. About 28% of white women completed college, compared to 15% for black women and 14% for
324
JUNE E. O’NEILL AND DAVE M. O’NEILL
Hispanic women. White women’s mean percentile score on the AFQT is 53% compared to 24% for black women and 30% for Hispanic women. White women worked a somewhat larger number of weeks since age 18 than black or Hispanic women but white women were much more likely to have worked part-time. White women were more likely to delay their first birth to age 30 or more, a decision that is compatible with acquiring additional education and on-the-job training. The decomposition results tell approximately the same story as the Table 6 results, which are based on dummy variables indicating race/Hispanic origin from regressions including all races. Decomposition results are given for two models, based on the regression results displayed in Table 7. (Note that Model 1 includes the same variables as Model 3 in Table 6, and Model 2 the same variables as Model 6 in Table 6.) The unadjusted white–black log wage gap among women is 0.189. Model 1, in addition to age and geographic location, includes only AFQT score and schooling. When the coefficients from the Model 1 regression for black women are used to weight the mean differences in characteristics, the model implies a higher wage for black women (a gap in favor of black women of 0.1266). The large racial difference in mean scores on the AFQT test, weighted by the black return to increases in AFQT (which is considerably larger than the white return) alone explains most of that result. When the white female regression coefficients are used, the implied wage gap does not reverse, but is negligible i.e., 0.0117. The inclusion of work experience variables in Model 2 barely changes the bottom line. However, because of the correlation of AFQT and education with work experience, the net contribution of AFQT and education declines when work experience is added. Using the Model 2 variables, AFQT still explains more of the white–black wage gap than any other variable, alone accounting for the whole gap when black coefficients are employed and half of the gap with white coefficients. In sum, expressed as hourly wage ratios, the unadjusted black/white ratio for women is 82.8%. When we control only for differences in education and AFQT (as well as age, region, MSA, central city) and weight the difference in characteristics with black women’s coefficients the ratio rises to 113.5%. The ratio rises to about 99% when we weight with white coefficients. These results are barely changed when we expand the variables to include work experience and fertility variables (birth before or after age 30). The unadjusted differential between Hispanic and white women is much smaller – less than a 10% differential. The differentials in AFQT scores and education between the two groups more than explain the wage gap, using either the white or Hispanic coefficients. The unadjusted Hispanic/white
What do Wage Differentials Tell About Labor Market Discrimination?
325
hourly wage ratio is 91.2% and rises to 110.7% when we control for AFQT, education and age and location factors using Hispanic coefficients (103.9% with white coefficients). The inclusion of work experience and fertility differences has little effect on the adjusted wage ratios. Overall, the results are quite similar to those for the white–black comparison: Hispanic women with the same measured skills as white women would earn 4–10% more than white women, depending on the model and whether Hispanic or white coefficients are used to weight the differences.
6. THE GENDER GAP IN WAGES: RESULTS FROM THE NLSY Measured as the female/male ratio of median annual earnings of all fulltime, year-round workers, the gender gap in wages narrowed considerably from the late 1970s when the ratio was just below 60%, to 2003, when it was 76% (Fig. 4). Among the NLSY cohort, the wage gap in 2000 was 79%, measured as the female/male ratio of hourly wages (a log wage difference of 0.235, Table 9). Thus, a significant gap in pay remains. Yet, the women and men in the NLSY have similar scores on the AFQT test and about the same level of schooling.19 Gender differences in wages arise for reasons other than differences in productivity linked to differences in cognitive skills. Instead, the most important source of the wage gap is the gender difference in market investments and job choices that reflect the relative importance of home and market activities in the lives of women and men. The division of labor in the family is less delineated than it once was and a majority of women with children now work in the market. Nonetheless, women on an average still assume greater responsibility for child rearing than men, and that responsibility is associated with a lower extent and continuity of market work. In addition, the expectation and assumption of home responsibilities influence choice of occupation and preferences for working conditions that facilitate a dual career, combining work at home and work in the market. A significant literature has investigated the effect of work in the home on women’s lifetime patterns of labor force participation and the effect of labor force discontinuities on wages.20 Women with children devote relatively more of their time and energy to home responsibilities than women without children and as a result earn lower wages.21 On the other hand, married men earn higher wages than other men. Although that effect may be partly endogenous – women may shun low earners as husbands – it is a plausible consequence of the division of labor in the home,
326
JUNE E. O’NEILL AND DAVE M. O’NEILL
Table 9. Gender Wage Gap among the NLSY Cohort, Ages 35–43 in 2000, Controlling for Different Sets of Explanatory Variables: Results for Total NLSY Population and Specified Sub-groups. All
By Schooling Level HS Graduate or less
Unadjusted log hourly wage gap Log wage differential controlling for (1) Age, MSA, region and race, schooling, AFQT (2) Variables in (1) plus life time work experience (3) Variables in (2) plus L.F. withdrawal due to family responsibilities (4) Variables in (3) plus class of worker (5) Variables in (4) plus occupational characteristics (6) Variables in (5) plus percent female in occupation
College Graduate or more
Never had a Child and Never Married
0.235
0.229
0.287
0.076 ns
0.231
0.230
0.244
0.019 ns
0.121
0.074
0.182
0.065 ns
0.102
0.058
0.155
0.054 ns
0.095
0.060
0.120
0.042 ns
0.084
0.073
0.078
0.013 ns
0.079
0.054
0.078
0.027 ns
Note: The log wage differentials are partial regression coefficients of a dummy (0,1) variable for ‘‘female’’ from a series of OLS log wage regressions containing the explanatory variables noted. The coefficients are significant at the 10% level or lower unless indicated with ‘‘ns’’. Separate regressions were conducted for each population group shown. For further information on the individual variables included, see the text and Table 10. Source: National Longitudinal Survey of Youth (NLSY79) merged with measures of occupational characteristics (three-digit level) from the September 2001 CPS, the CPS March and the Dictionary of Occupational Titles (U.S. Department of Labor, 1991).
which leads men to take greater responsibility for providing the family’s money income and consequently to work longer, more continuously and possibly harder.22 Differences in lifetime work patterns have received considerable attention as a source of the gender gap. However, another significant source of wage
What do Wage Differentials Tell About Labor Market Discrimination?
327
differentials of particular relevance to the gender gap are the ‘‘inequalities arising from the nature of the employments themselves.’’23 As Adam Smith observed, the ‘‘agreeableness and disagreeableness’’ of employments give rise to equalizing or compensating wage differences. These non-pecuniary characteristics of employments are likely to be evaluated differently by women and men. Occupations and individual firms differ in the extent to which they offer flexible work schedules and a less stressful work environment, characteristics that are likely to be more highly valued by women. These and other work amenities are likely to come at a price – i.e., lower wages. Disamenities, such as exposure to physical hazards, would likely require a premium, other things the same.24 In addition, men and women may differ in their attitudes toward work involving dirty or otherwise unpleasant physical conditions. Physical differences are likely to affect aptitude for certain work, for example, for jobs requiring heavy lifting, although the proportion of jobs requiring hard physical labor has declined over time. It is difficult to estimate the determinants of the gender gap because differences in standard variables such as years of schooling are not likely to be important sources of the gap. The NLSY is superior to most other data sets in that it provides more detailed information than is commonly available on lifetime patterns of work participation as well as on marriage and family. We create additional proxy variables in an effort to empirically capture gender differences in choice with respect to employment amenities. Our analysis of the gender gap follows the same procedures used in our analysis of racial and ethnic differences. As before, we start with the approach that pools observations of men and women in a single equation and follow with a decomposition analysis based on separate equations for men and women. Table 9 shows the effect on the gender gap of controlling for different sets of explanatory variables from a series of log wage regressions. The wage gap is estimated as the partial regression coefficient on a dummy variable indicating whether the worker is a woman. Results are shown for the full sample of male and female workers as well as for subsets of the sample disaggregated by education. Results are also given for men and women with a similar lack of family responsibilities, namely those who never had a child and never married. Table 10 provides variable means and regression coefficients; Table 11 shows results of a decomposition analysis. The unadjusted log wage gap for the full sample of men and women is 0.235 (Table 9). It is essentially unchanged after including education, AFQT and geographic location (Model 1). The addition of a vector of three work experience variables, however, reduces the gender gap by almost half, to 0.121 (Model 2). The work experience variables include: weeks worked
Means and Partial Regression Coefficients of Explanatory Variablesa from Separate NLSY Log Wage Regressions for Men and Women Aged 35–43 in 2000. Means Female
Female
Male
Race Hispanic (0,1) Black (0,1)
Male
M2 Coefficient
M4 t-stat
Coefficient
M2 t-stat
Coefficient
M4 t-stat
Coefficient
t-stat
0.193 0.282
0.063 0.053
2.57 2.42
0.060 0.066
2.61 3.14
0.025 0.022
1.02 0.92
0.018 0.005
0.75 0.20
0.031 0.103
0.052 0.124
0.089 ––
1.76 ––
0.078 ––
1.64 ––
0.028 ––
0.65 ––
0.025 ––
0.60 ––
0.300
0.326
0.003
0.10
0.008
0.27
0.018
0.65
0.013
0.50
0.045
0.056
0.015
0.34
0.046
1.12
0.027
0.63
0.015
0.38
0.308 0.153
0.232 0.155
0.090 0.276
2.99 7.61
0.060 0.216
2.09 6.19
0.166 0.373
5.31 10.23
0.123 0.260
4.08 7.08
0.053
0.041
0.391
8.49
0.348
7.76
0.562
10.84
0.446
8.62
0.007
0.015
0.758
7.47
0.654
6.71
0.806
10.60
0.639
8.53
AFQT percentile score ( .10)
3.981
4.238
0.042
9.92
0.032
7.84
0.042
9.92
0.029
7.04
L.F. withdrawal due to family responsibilities (0,1)
0.549
0.130
0.081
4.16
0.082
4.46
0.080
3.14
0.066
2.74
Some college BA or equivalent degree MA or equivalent degree PhD or professor degree
JUNE E. O’NEILL AND DAVE M. O’NEILL
0.182 0.316
Education and skill level o10 years 10–12 years (no diploma or GED)b HS graduate (diploma) HS graduate (GED)
328
Table 10.
Employment type Government employer (0,1) Non-profit employer (0,1)
0.030
13.85
0.023
11.13
0.038
12.54
0.034
11.39
0.573
0.046
3.53
0.040
3.22
0.025
5.15
0.020
4.46
0.050
0.203
4.24
0.084
1.81
0.779
7.90
0.540
5.70
0.215
0.144
0.030
1.50
0.027
1.13
0.100
0.049
0.056
2.13
0.121
3.20
26.961
28.773
0.001
2.44
0.003
5.43
0.013 0.004 0.080 0.092 0.033
0.084 0.043 0.307 0.215 0.188
0.327 0.293 0.005 0.011 0.120
4.66 2.27 0.18 0.37 2.56
0.131 0.075 0.019 0.049 0.000
3.97 1.72 0.83 1.99 0.01
0.557
0.415
0.157
2.19
0.045
0.49
0.143
0.139
0.497
4.62
0.258
2.22
329
Characteristics of Person’s 3-digit occupation (OCC.) SVP required in occupation (months) (DOT) Hazards (0,1) (DOT) Fumes (0,1) (DOT) Noise (0,1) (DOT) Strength (0,1) (DOT) Weather extreme (0,1) (DOT) Proportion using computers (CPS) Proportion using computer for analysis (CPS)
17.169
What do Wage Differentials Tell About Labor Market Discrimination?
Lifetime Work Experience 15.565 Weeks worked in civilian job since age 18 C 52 Weeks worked in 0.062 military since 1978 C 52 Weeks PT C total 0.137 weeks worked since age 22
Means Female
Female
Male
M4 t-stat
Coefficient
M2 t-stat
Coefficient
M4 t-stat
Coefficient
t-stat
0.345
0.236
0.255
3.19
0.007
0.06
0.772
1.092
0.022
1.11
0.023
1.91
1.046
0.789
0.144
7.30
0.073
3.57
6.348
2.695
0.005
1.08
0.019
3.55
0.392
0.464
0.403
0.467
2.529
2.764
2704
2694
Source: National Longitudinal Survey of Youth (NLSY79) merged with measures of occupational characteristics (three-digit level) from the September 2001 CPS, the March CPS, the CPS ORG, and the Dictionary of Occupational Titles (1991). a Model also controls for age, central city, MSA, region, and occupation missing. b Reference group.
JUNE E. O’NEILL AND DAVE M. O’NEILL
Adjusted R2 Dependent mean (Log hourly wage) Sample size
Male
M2 Coefficient
Proportion using computer for word processing (CPS) Relative rate of transition to unemployment Relative rate of transition to OLF % female in OCC. 0.1. (CPS ORG)
330
Table 10. (Continued )
Gender Wage Gap: Decomposition Results (NLSY, 2000). Using Male Coefficients
Log wage gap (male–female) attributable to Age, race, region, central city, MSA AFQT Education level L.F. withdrawal due to family responsibilities Lifetime work experience Nonprofit, government
Using Female Coefficients
M1
M2
M3
M4
M1
0.0044 0.0132 0.0138
0.0112 0.0107 0.0128 0.0335 0.1425
0.0089 0.0073 0.0094 0.0272 0.1135 0.0088
0.0089 0.0074 0.0096 0.0277 0.1116 0.0081
0.0040 0.0143 0.0147
0.0062 0.0122
M2
M3
M4
0.0064 0.0081 0.0054 0.0344 0.0649 0.0048
0.0064 0.0081 0.0052 0.0343 0.0655 0.0050
0.0053 0.0040
0.0020 0.0054
0.0021 0.0024
0.0167 0.0116
0.0040 0.0028 0.0721
0.0252 0.0226
0.0267 0.0259 0.0137
0.2351 0.2030 0.0321
0.2351 0.2342 0.0009
0.2351 0.1578 0.0773
0.2351 0.1526 0.0825
0.0089 0.0107 0.0068 0.0340 0.0901
Occupational characteristics Investment related SVP (Specific Vocational Preparation) Computer usage Compensating differences Disamenities (physical) Unemployment risk; labor force turnover TYP: % female in occupation Unadjusted log wage gap Total explained by model Unexplained log wage gap Unadjusted hourly wage ratio (female/male) Adjusted hourly wage ratio (female/male)
0.2351 0.0037 0.2314 79.0 79.3
0.2351 0.1851 0.0500 79.0 95.1
79.0 96.8
79.0 99.9
0.2351 0.0036 0.2315 79.0 79.3
0.2351 0.1370 0.0981 79.0 90.7
79.0 92.6
79.0 92.1
331
Note: Decomposition results shown are derived from results of separate regressions for men and women. See Table 10 for variable means and coefficients using Models 2 and 4. Wage ratios are based on the exponentiated log hourly wage. Source: National Longitudinal Survey of Youth (NLSY79) merged with measures of occupational characteristics (three-digit level) from the September 2001 CPS, the March CPS, the CPS ORG and the Dictionary of Occupational Titles (1991).
What do Wage Differentials Tell About Labor Market Discrimination?
Table 11.
332
JUNE E. O’NEILL AND DAVE M. O’NEILL
in civilian jobs since age 18 (converted to years by dividing by 52); weeks worked in the military divided by 52; and the proportion part-time of total weeks worked (Table 10). On an average, women have worked about two years less than men in military and civilian jobs combined. Moreover, close to 14% of the weeks worked by women were part-time compared to 5% for men. Weeks worked have a positive and significant effect on the hourly wage for both men and women, and part-time work has a significant negative effect for both. However, the magnitude of the effect of part-time on wages is considerably larger for men than for women (Table 10). The return to years worked, however, is similar for men and women. As a proxy for commitment to home responsibilities, we add in Model 3 a variable indicating whether the worker had ever withdrawn from the labor force citing childcare or family responsibilities as the reason. Such labor force withdrawal is associated with an 8% reduction in the wage rate for men as well as women. However, 55% of women and only 13% of men have ever withdrawn because of family responsibilities (Table 10). As shown in Table 9, the addition of this variable reduces the gender gap to 0.102. In Model 4, we add two variables indicating whether the person’s job was in government employment or in the non-profit sector. Non-profit jobs offer more part-time work and are more likely to allow for flexible schedules and a more relaxed ambience than work in the for-profit sector. As shown in Table 10, women are twice as likely to work in the non-profit sector than men and employment in the non-profit sector is associated with lower pay. The effect is significant for women and men but here again the magnitude of the effect is much larger for men than for women (twice as large). Government work is also associated with lower pay. However, the effect is weak and insignificant for either sex. The addition of the class of worker variables reduces the gender gap slightly – to 0.095. The final set of variables measure particular characteristics of the 3-digit occupation held by respondents that are expected to have an effect on wages because they are associated with on-the-job investment or particular amenities or disamenities. The occupational characteristics included in our analysis are listed in Table 10 along with the mean values for men and women separately. Measures of Specific Vocational Preparation (SVP) and other occupational characteristics were derived from the Dictionary of Occupational Characteristics (DOT, 1991 version) and from special supplements to the CPS pertaining to computer use on the job. A variable measuring the probability of leaving the labor force and another measuring the risk of unemployment in the occupation were estimated using data from the March CPS.25
What do Wage Differentials Tell About Labor Market Discrimination?
333
The gender gap narrows to 0.084 when the occupational characteristics enumerated in Table 10 are added (Model 5). In Model 6, we add a variable that measures the percent female in the respondent’s 3-digit occupation. That addition narrows the gap somewhat more (to 0.079). Although measures of occupational dissimilarity between men and women have declined since the 1970s, the occupational distributions of women and men are still very different (Cavallo & O’Neill, 2004). As shown in Table 10, the women in our NLSY sample, on an average, worked in occupations in which the percent female was 63%; men worked in occupations in which the percent female was 27%. These occupational differences are sometimes viewed as evidence of discrimination.26 However, the occupations that women choose are strongly predicted by characteristics that are compatible with women’s dual careers.27 The percent female in an occupation has only a limited effect on wages because it is highly correlated with the other occupational and personal characteristics in the regression. In fact, in the log wage regression based only on the female sample, the percent female is not statistically significant and bears a positive sign (Table 10). The variable is negative and significant only for men. Results in Table 9 are shown for specific sub-groups of the NLSY sample. The results for the high school group (those with high school diplomas or GED’s or with less schooling) are similar to those described above for all women and men. However, gender differences in work experience are more important for the high school group than for all women and men and account for two-thirds of the wage gap. (Compare the unadjusted gap with Model 2.) The results for college graduates differ somewhat from those of the other groups. The unadjusted wage gap is larger, in part because gender differences in skills among college graduates are somewhat larger. Men are more likely to receive PhDs and professional degrees and among college graduates men have higher AFQT scores than women (73rd versus 65th percentile). Although the gender difference in years worked is slight at the college graduate level, the difference in part-time work is as large as for the high school group. Moreover, women who are college graduates are less likely to work in the private sector than other women, or men at any education level. (One-third of female college graduates work in the non-profit sector and 17% work in government.) A college education appears to give women access to jobs with working conditions that either allow them to work parttime or to work full-time, but under conditions more complementary with care of family such as the long vacations of teachers. Controlling for both gender differences in class of worker and occupational characteristics
334
JUNE E. O’NEILL AND DAVE M. O’NEILL
reduces the log wage gap at the college level from 0.155 (Model 3 in Table 9) to 0.078 (Model 5). Inclusion of the percent female in the occupation (Model 6) does not affect that result. The gender gap among those with no more than a high school education is dramatically reduced when we control for work experience and is reduced somewhat more when we also include labor force withdrawal for family reasons, at which point the gap is –0.058 (Model 3). Table 9 (last column) further highlights the relative importance of family responsibilities versus labor market discrimination by examining the gender gap among a subset of men and women in apparently similar lifetime family situations – namely men and women who were never married and never had a child. Never-married men and women without children are not responsible for the financial support of a family as are most married men. But nevermarried women without children do not have the responsibility for children that mothers bear. However, never-married women without children have somewhat better credentials than never-married men without children with respect to education, AFQT scores and years of civilian work experience. Therefore, it is not surprising that the unadjusted gender gap for this group is actually positive; the women earn about 8% more than their male counterparts. When we control for differences in characteristics, the gender gap in favor of women is eliminated, but the negative coefficient is small and is not statistically significant. This observation lends support to the view that the factors underlying the gender gap in pay primarily reflect choices made by men and women, given their different societal roles, rather than labor market discrimination against women due to their sex. Of course, it is still possible that men who have never married have unobservable negative productivity factors that selected them into this group in the first place. The fact that they have never married could reflect personal characteristics that would also lower their productivity in the labor market. However, the same could be true of never-married women. We do not know the proportion of men or of women who are unmarried by choice, dedication to a career, or because of negative personality traits.
6.1. Decomposition Analysis: Gender Gap The comparison of male and female earnings and the interpretation of the gender gap in pay is further complicated by gender differences in the effects of certain variables on earnings. However, the variables involved are not
What do Wage Differentials Tell About Labor Market Discrimination?
335
those that have sometimes aroused suspicion of bias. As shown in Table 10, in separate wage regressions for women and men, the returns to standard human capital variables such as schooling, years of work experience and tenure are quite similar for women and men. However, coefficients differ considerably by sex when the variable is one that is likely to have a different meaning for women and men. For example, the variable measuring the proportion of weeks worked part-time (over the years since the worker was aged 22) is negatively associated with earnings for both men and women; but the size of the effect is much larger for men. Whether one works for a non-profit employer is negatively associated with earnings for both men and women and again the effect is much stronger for men. The variable measuring the percent female in the individual’s occupation is negatively related to earnings for both women and men; but the effect for women is weak and never statistically significant, while the effect for men is usually larger in magnitude than it is for women and is also statistically significant. (The exception is for male college graduates for whom the effect of percent female in the occupation is essentially zero.) How can these findings be explained? Women are likely to choose parttime work and non-profit work because they offer more flexibility and in the case of non-profit firms, less stress. However, it seems plausible that women working full-time within the private for-profit sector are more likely to seek job situations that also offer more flexibility although we have no easy way to detect that with the available data. In that case, the difference in work situations between part-time and full-time or between non-profit and forprofit employment may be less stark for women than for men. If we could measure all of the detailed characteristics of the full-time and part-time jobs held by women and men, the measured gender dissimilarity in occupations and jobs would likely widen. With perfect measures of job characteristics, the partial regression coefficient of the part-time variable should be the same for men and women. A smaller coefficient on the parttime variable for women could then imply that women in full-time jobs are experiencing some labor market discrimination. Without better data we cannot say with certainty whether the male or female coefficients on parttime and non-profit employment more nearly reflect the real trade-offs. Decomposition results for the gender gap using both male and female coefficients are presented in Table 11 and means and regression coefficients of key variables are given in Table 10. Because of gender differences in coefficients, such as those noted above, the results of the decomposition analyses differ depending on whether male or female coefficients are used. In Table 11, the unadjusted gap, expressed as the ratio of women’s to men’s
336
JUNE E. O’NEILL AND DAVE M. O’NEILL
hourly wage, is 79%. Using male coefficients, the ratio rises to 99% when all variables are included; using female coefficients it rises to 92%. Which are the more appropriate coefficients to use? The answer depends on issues related to the degree to which the data we use can accurately measure differentials in personal and job characteristics. Without better data all we can conclude is that labor market discrimination is unlikely to account for a differential of more than 8% and may not be present at all.
7. SAMPLE SELECTION AND OTHER METHODOLOGICAL ISSUES Any empirical analysis is subject to error and some researchers have emphasized possible difficulties that could bias results in analysis of racial and ethnic differentials (for example, Darity & Mason, 1998) as well as gender differences (Blau, 1998). We take up the following: Problems in the use of AFQT scores; sample selection; and endogeneity in the human capital variables. 7.1. Problems in Using AFQT Scores Differentials in AFQT scores reflect both differences in ability and differences in educational attainment. However, the scores provided in the NLSY data pose difficulties because the AFQT test was administered only once – in 1980 – when the respondents werein the age group of 15–23 years, at which time a majority had not completed their schooling. (AFQT scores arrayed by age and education at the time of test and by education in 2000 are displayed in Table A1.) Additional schooling is likely to raise AFQT scores, particularly for the younger groups. But we have no way of determining by how much it would affect scores because the correlation of ability and education is not known; nor is the correlation between ability and AFQT scores. How important a bias this would cause depends on the strengths of the two correlations. If the ability/score correlation dominates, then our estimates are not likely to be seriously biased. The NLSY data allow us to roughly assess the degree of bias in our estimate of discrimination by using a subset of the data restricted to those who had already completed their schooling at the time they took the AFQT test. In Appendix Table A2 (upper panel), we show the results of a series of log wage regressions on age, location, AFQT, schooling and work experience, roughly
What do Wage Differentials Tell About Labor Market Discrimination?
337
similar to those in Tables 3 and 6.28 We include dummy variables indicating whether the respondent is black or Hispanic. The table also shows, in the lower panel, results for the same analysis using all individuals in our NLSY data set, whether or not they had completed their education at the time of the test. The results show only a small difference in the coefficients, indicating that the dominant correlation is between individual ability (linked to family background and IQ) and AFQT scores rather than between educational attainment and AFQT scores.29
7.2. Sample Selection Our analysis of wage differentials is based on those respondents who were employed within the last month before the survey interview and reported a wage rate. In addition, we imposed certain restrictions on the sample to remove sources of potential measurement error and persons missing crucial data. A legitimate question is whether those omitted from the sample are significantly different from those selected to be in the sample, and whether such differences would lead to bias in our estimates of the size of the effect of labor market discrimination on earnings. As shown in Appendix Tables A3 and A4, out of the entire cohort of men, 74% of white men were included in our analysis of wage differentials compared to 68% of black men and 73% of Hispanic men. A somewhat larger proportion of women were excluded from the analysis. The proportion of women included in the analysis was 66% for white women, 68% for black women and 63% for Hispanic women. Tables A3 and A4 provide information on the characteristics of those included in the analysis and those excluded. Those who were excluded are grouped into two categories: those who reported no wage in the last two years, primarily because they were out of the labor force; and those for whom a wage was reported in the last two years but were excluded on other grounds. The other grounds for exclusion were: not employed in the last month, self-employed (our analysis is restricted to wage and salary workers); AFQT score was missing; wage was below $3.50 or above $125 per hour in 2000. Most of the excluded men fall into the second category – that is, those for whom a wage was available. However, among women, those excluded because they had no reported wage in the last two years were almost as large a group as those who reported wages. The data in Table A3 show that among men the wage rates of those who were excluded were 73% of those included in our analysis. Moreover, the
338
JUNE E. O’NEILL AND DAVE M. O’NEILL
ratio varies by race. For whites it was 77%, for blacks 61% and for Hispanics 81%. Obviously, the unadjusted wage gap would be larger if those who were excluded were included in our analysis. However, there is no obvious reason why our estimates of the wage gap attributable to non-discriminatory factors should be biased toward minimizing the role of discrimination because of selectivity bias. Unless the individuals excluded from our analysis possessed special unmeasured characteristics that would lead employers to be more prejudiced against them as opposed to included workers with the same set of measured productivity factors, there is no reason for the adjusted wage gap to be a biased measure of discrimination. It is important to distinguish between selectivity bias affecting the unadjusted racial and gender gaps, which it clearly does, and its possible effect on the relative amount of the gap attributable to discrimination. In order to get some idea of the potential effect of selection bias, we have estimated our basic regression model including all the excluded respondents for whom we had wage data within the past two years. (We still exclude those with no AFQT reported because of the key role of that variable in explaining the differential.) This analysis has important limitations because the excluded group was excluded because their reported wages are both less current and reliable. The results are shown in Table A5 for men and Table A6 for women and can be compared with our basic analysis in Tables 3 and 6. The only significant finding of the expanded analysis is that our estimate of the male black/white wage gap possibly attributable to discrimination could be raised from practically zero (the result in Table 3) to 5% (the result in Table A5). The expanded analysis does not significantly change our estimates of the Hispanic/white wage gap for men or our estimates of the black/white or Hispanic/white wage gap among women. One explanation for the larger gap for black men in the expanded analysis is that black men who were excluded from our basic analysis have had much higher incarceration rates over their lifetimes than either Hispanic or white males, and much higher incarceration rates than black men included in our basic sample. As indicated in Table A3, 29% of black men who were excluded from our basic analysis but were included in our expanded analysis were interviewed in jail or in prison in at least one of the NLSY surveys, compared to 7% of white men and 18% of Hispanic men. Among black men who reported no wage during the past two years, 44% had ever been in jail. As discussed above, criminal activity is strongly associated with reduced employment, even during periods when not in jail (Bound & Freeman, 1992; Hill & O’Neill, 1993). A history of incarceration has been shown to
What do Wage Differentials Tell About Labor Market Discrimination?
339
contribute to earnings loss and decreased wage growth over the life cycle (Western, 2002; Holzer, Offner, & Sorenson, 2004). We did not incorporate the effects of incarceration into our analysis, although this would be a good future project.30 Selection issues are often raised with respect to the gender gap because a larger percentage of women than of men are out of the labor force. Among the NLSY cohort, 34% of women and 28% of men were excluded from our basic analysis. (Compare Tables A3 and A4.) Of this group, 54% of women and 70% of men reported a wage within the past two years. The female/male wage ratio for this excluded group is slightly lower than it is for those in our basic analysis (77% versus 79%). The women excluded from the basic analysis have characteristics linked to lower wage rates – in particular, almost three years less work experience, a larger proportion reporting withdrawal from the labor force due to family responsibilities, and more part-time work. Schooling and AFQT scores are only slightly lower. We have estimated the gender wage gap using the regression sequence shown in Table 9, for an expanded sample including all those women and men who were excluded from our basic analysis but reported a wage within the past two years. The results (Table A7) are highly similar to those of Table 9 – a gender wage gap of 0.067 with the expanded sample compared to 0.079 with the basic sample. In sum, the unadjusted measure of the wage gap for the various groups we have compared would be somewhat larger in each case if we could include those without an observed wage. However, in each case, with the exception of the male black/white wage gap, the wage gap adjusted for differentials in productivity does not change significantly when we include a substantial proportion of those who were initially excluded. Thus, sample selection does not appear to be concealing evidence of discrimination. Although we did not attempt to estimate the effect of including those who had no wage rate within the past two years, an inspection of their characteristics suggests that their inclusion would likely widen the unadjusted wage gap but also might not have any significant effect on the adjusted gap.
7.3. Endogeneity Issues In any analysis of cause and effect involving natural or uncontrolled experimental situations, the question arises whether the explanatory variables in the model are themselves affected by labor market discrimination. In particular, the question is often raised whether educational attainment, test
340
JUNE E. O’NEILL AND DAVE M. O’NEILL
scores and other factors are themselves affected by labor market discrimination against minorities and women. The argument is sometimes made that the lower educational attainment and test scores of blacks compared to whites can be attributed in part to their anticipation that they will earn a lower return on their investment because of discrimination in the labor market. Thus, studies that regress earnings on educational attainment and test scores to help explain the racial earnings gap will tend to underestimate the effects of labor market discrimination on earnings because some of the lower attainment of blacks is itself due to labor market discrimination. However, the evidence from our analysis and those of others, suggests that when log wage regressions are run separately for blacks and whites the partial regression coefficients of the test score and educational attainment variables do not differ significantly, suggesting the same rate of return. It was once true that the returns to education for blacks were lower than for whites (US Commission on Civil Rights, 1986). However, the return to college education has been higher for blacks than for whites since the 1970s (Heckman, 1998). Moreover, the gap in educational attainment between blacks and whites has converged sharply over the years, suggesting that blacks have been reacting to increased incentives. As we demonstrate above, the market return to higher AFQT scores is actually somewhat higher for blacks than to whites. Moreover, as discussed above, the AFQT has been tested extensively and found to be free of cultural bias. Years of schooling and AFQT scores largely reflect characteristics and skills developed outside the market and are not likely to be affected by current labor market discrimination. In the early decades of the twentieth century most blacks lived in the South, where Jim Crow laws and regulations limited employment opportunities, and societal discrimination reinforced by state and local policies severely restricted their access to education (Donahue & Heckman, 1991; Smith & Welch, 1989; US Commission on Civil Rights, 1986). Racial discrimination in publicly provided school resources was eventually eliminated and there has been significant convergence in the black–white gap in years of schooling completed. Nonetheless, the legacy of educational deprivation may have a lingering effect on the early acquisition of skills through low parental education and income. However, differences in skills that stem from a disadvantaged family background are quite distinct from employer discrimination. Work experience obviously can be directly influenced by labor market discrimination. But it is an empirical question whether that is in fact the case. We have earlier discussed the issue in the context of the lower lifetime
What do Wage Differentials Tell About Labor Market Discrimination?
341
work experience of black men and noted that the relative decline in the employment of less-educated black men, that began in the 1970s has been shown to be related to the decline in demand for low-skilled workers and to their increased involvement in crime and resulting imprisonment. There is no good evidence that we know of suggesting that labor market discrimination is greater now than it was in the 1940s through the 1960s when the employment of black men was relatively higher than it has become.31 In the analysis of the gender gap, issues of bias in the explanatory variables have been raised with respect to both work experience and occupation. Women’s employment rates have increased so rapidly over the past several decades that allegations that labor market discrimination is reducing women’s labor force participation per se have little force. However, the issue is still raised with respect to occupational differences. We believe that gender differences in occupation and type of job more nearly reflect choice, not employer prejudices. Our analysis indicates that women choose occupations and job settings that are compatible with combining market work with family responsibilities. It would be difficult to find an explanation based on employer discrimination that could explain the observed patterns.
8. CONCLUDING COMMENTS Differences in the quantity of education as measured by years of schooling, and the amount and quality of skill developed in the home and in school, as measured by test scores, are of central importance in explaining the black/ white and Hispanic/white wage gaps among women as well as among men in the labor force. Schooling and immigrant status are particularly relevant in explaining wage differences between whites and groups such as Asians (many of whom earn more than whites) and individuals from Central and South America. Our analysis of the factors underlying the black/white wage gap leads us to concur with the conclusion reached by James Heckman (1998) that ‘‘most of the disparity in earnings between blacks and whites in the labor market of the 1990s is due to the differences in the skills they bring to the market, and not to discrimination in the market.’’ The same conclusion can be applied in 2000 to other racial and ethnic comparisons. The gender gap is more difficult to analyze because the reasons for the difference are extremely hard to measure. Gender differences in schooling and cognitive skills as measured by the AFQT are quite small and therefore explain little of the pay gap. Instead, the gender gap is largely attributable to choices made by women concerning the amount of time and energy to
342
JUNE E. O’NEILL AND DAVE M. O’NEILL
devote to a career as reflected in years of work experience, utilization of part-time work and workplace and job characteristics. There is no gender gap in wages among men and women with similar family responsibilities. Comparing the wage gap between women and men in the age group of 35–43 years, who have never married and never had a child, we find a small, unadjusted gap in favor of women, which becomes insignificant after accounting for differences in skills and job and workplace characteristics. What the average woman sacrifices in earnings from choosing jobs that allow for part-time work and flexible work conditions is presumably offset by a gain in the utility of time spent with children and family.
ACKNOWLEDGMENTS The authors thank Mei Liao and Wenhui Li for excellent research assistance and participants at the conference for helpful comments. Research support was received from the Olin Foundation.
NOTES 1. During the 1940s, many states outside the South implemented fair employment legislation. For a discussion of the effects, see Landes (1968) and Neumark and Stock (2001). 2. Figs. 1 and 4 depict long-term trends in earnings ratios based on published data from the March Current Population Survey (CPS) reports on median annual earnings of full-time, year-round workers. Figs. 2 and 3 are based on estimates of mean hourly wage rates derived from the March CPS public use tapes by dividing annual earnings by the product of weeks worked during the year and hours worked per week. 3. Minorities frequently have lower skills and earnings than white workers, but some minorities have relatively high earnings and relatively high levels of human capital (e.g., Japanese and Asian Indian workers). Thus, a positive unadjusted wage gap in itself does not constitute evidence of the absence of discrimination. 4. For example, Neal and Johnson (1996) and O’Neill (1990) explain most of the black/white wage gap among men when they incorporate scores on the AFQT in the NLSY. Brown and Corcoran (1997) include field of college major, a harbinger of occupational choice, as well as actual work experience and explain a large share of the gender gap. 5. The vast majority of such cases are not decided on the merits but on mutual agreement through a consent decree, which allows the accused firm to avoid potentially large legal and other costs by payment of a negotiated settlement. In such settlements, the employer neither admits to discrimination nor is found guilty of discrimination by the court. In those relatively few cases that have been decided on
What do Wage Differentials Tell About Labor Market Discrimination?
343
the merits, either by a judge or a jury verdict, it is the employer who has won most of the time. See O’Neill and O’Neill (2005). 6. Studies involving audit pair experiments tabulate the success of matched black and white actors set up to apply for the same job. In a critique of audit studies, Heckman (1998) casts doubt on the ability of the audit experiments to produce unbiased estimates of discrimination and questions their ultimate relevance. As Heckman points out, even if an audit study was well designed and uncovered discriminatory behavior in individual firms, that would not be evidence of discrimination in the market as a whole. The experiences of the actors in these studies are not likely to translate into the actual employment and earnings experience of real workers. Heckman also shows that the findings of the audit studies have been oversold – most studies actually find that the vast majority of results for audit pairs indicate equal treatment of blacks and whites. 7. Studies that utilize longitudinal data sources typically find that the gender gap in lifetime work experience can account for a significant share of the gender gap in earnings (for example, Mincer & Polachek, 1974; Blau, 1998; O’Neill, 2003). However, few data sources provide the needed detail on hours worked during each year and on periods out of the labor market. The quality of data also can be a problem when data are collected retrospectively rather than contemporaneously (Hill & O’Neill, 1992). 8. We use the NLSY version reported in 1989, which includes scores on reading comprehension, word knowledge, arithmetic reasoning and mathematical knowledge. For a discussion of recent validation tests conducted by the National Academy of Sciences in conjunction with the Department of Defense, see Neal and Johnson (1996). 9. Herrnstein and Murray (1994) famously, but incorrectly, assume that the AFQT is an IQ test. Many analysts, however, have demonstrated that AFQT scores are significantly affected by environmental factors. See for example, Korenman and Winship (2000). Neal and Johnson (1996) find that racial differences in parental education, occupational status and other home background characteristics account for more than 40% of the racial gap in AFQT scores among men in the NLSY. 10. The sample is restricted to civilian wage and salary workers. In estimating wage rates, we use the hourly wage reported directly by NLSY respondents for those paid by the hour. For those who are paid on another basis – day, week, month, etc., we use usual weekly earnings divided by usual weekly hours. This measure is likely to be a more accurate estimate of the hourly wage than the Census measure which is based on annual earnings during the previous calendar year divided by an estimate of annual hours (weeks worked times usual hours per week during the year). We omitted from the sample workers with an hourly wage below $3.50 or more than $125 (in 2000 dollars), a restriction that eliminated 77 men and 81 women (2% of men and 2% of women). Other restrictions included omission of those who did not take the AFQT or who were missing information on key variables or for whom a complete work experience record could not be compiled. Workers were also excluded if they had never been employed during the four-week period prior to the survey interview. We examine the effect of these exclusions below. 11. The AFQT was administered to the NLSY sample just once – in 1980 when the cohort was 15–23 years of age. Test score results are affected by age and schooling at the time of the test; although the precise effect is difficult to assess, because we do not
344
JUNE E. O’NEILL AND DAVE M. O’NEILL
have readings on the AFQT for the same individual at different stages in their lives. We hold constant age and completed education in 2000 in our analyses – an implicit adjustment. Neal and Johnson (1996) adjust scores for age, but not for education at the time of test. O’Neill (1990) holds constant both years of schooling completed at time of test and since the test. We show the results of different ways of evaluating the effect of AFQT on the wage gap in Section IV. The essential results do not change with respect to the skill-adjusted racial wage gap. 12. Antecol and Bedard (2004) also find a significant effect of differential employment on the black/white wage gap. 13. In an analysis of the determinants of low work attachment among youth in the NLSY as of 1987, Hill and O’Neill (1993) (appendix, Model 3 results) found a strong positive association between ever having been in jail and low work attachment (in years when individual was not in jail, not in school and not in the armed forces). The effect of jail on employment was significant holding constant AFQT and detailed family and zip-code characteristics. 14. The differential widens slightly after adjusting for location for the total and high school groups because Hispanics are disproportionately located in high-wage cities. 15. The large influx of immigrants between 1979, the year in which the NLSY cohort was selected and 2000, the census year, makes it difficult to compare census and NLSY results for Hispanics. 16. Similar findings on the return to AFQT by race were reported by O’Neill (1990) and Neal and Johnson (1996) when the cohort was younger. 17. Neal and Johnson (1996) discuss a large study of the relation between AFQT scores and performance in the military conducted by the National Academy of Sciences in conjunction with the Department of Defense. The study concluded that the AFQT predicted performance in the military as well for blacks and whites. 18. The lower return to military service could reflect less relevance of military skills to civilian jobs, since we exclude the active military from our wage sample. However, the subject bears further investigation into the timing of exit from the military and other circumstances of military service. For example, those who recently separated may be experiencing transitional problems. 19. Women have slightly lower scores than men on the AFQT. They are less likely to be high school dropouts, more likely to have 1–3 years of college and about as likely to have college degrees. Men are more likely to have PhDs or professional degrees, but fewer than 2% have such degrees. (See Table 10 for details.) The level of schooling attained by women increased more than that of men over the past two decades and is one of the reasons for the narrowing of the unadjusted gender gap (O’Neill & Polachek, 1993). 20. See Mincer (1962), Mincer and Polachek (1974) and Mincer and Ofek (1982). Also, see Becker (1985) on the effect of home responsibilities on energy in the market. 21. On the ‘‘motherhood wage penalty’’ see, for example, Waldfogel (1998), Anderson, Binder, and Krause (2003) and Budig and England (2001). 22. See Becker (1981) on the basic theory of the family. Also see Korenman and Neumark (1991) on the effect of marriage on men’s market productivity. 23. Quoted from Adam Smith, The Wealth of Nations, 1776, Chapter X, Book I.
What do Wage Differentials Tell About Labor Market Discrimination?
345
24. DeLeire and Levy (2004) examine the trade-off between wages and safety and find that married women are much more risk averse than men, requiring a larger compensating differential. They also find that differences in the risk of death at work can account for a significant share of occupational differences by gender. 25. The data on occupational characteristics were obtained from Cavallo and O’Neill (2004). 26. One school of thought maintains that occupational segregation is the main mechanism through which discrimination is imposed. See the well-known work on the crowding hypothesis by Barbara Bergmann (1974). 27. Cavallo and O’Neill (2004) conduct an analysis of the determinants of the percent female in an occupation across three-digit occupations and find that variables compatible with women’s constraints (such as the incidence of part-time work and of a long work week and the extent of specific training required) explain most of the variation. 28. The one difference in the models is that here we use a continuous variable for schooling and in Tables 3 and 6 we use schooling dummies. We use the continuous variable to identify those who had not increased their schooling between 1980 and 2000. However, the non-linear treatment is preferred. The two ways of treating education do produce the small differences in results between Table 3 and Table A2. 29. Appendix Table A1 displays scores for men and women by race and Hispanic origin for the NLSY cohort at ages 15–18 and 19–23 at the time of the test and by years of school completed in 1980 and by schooling in 2000. From this table, it is possible to get a rough idea of the effect of education, net of ability, for those who had at least some college by 2000, by comparing scores of the older and younger cohort at the same college and college graduate level in 2000. We know all of these people attained the same level of education by 2000. But all of the younger cohort took the test before attending college, whereas a substantial portion of the older group had already completed some or more college by 1980. Therefore, the observed test score differential between these two groups provides a rough measure of the net effect of education. The table permits this comparison to be conducted for blacks, whites and Hispanics, for men and women separately. 30. We have included a measure of lifetime work experience, which is clearly associated with incarceration. However, a full treatment would require detail on the timing and length of labor force interruptions and the effect of these interruptions on wage growth. The problem is a kin to that addressed by Mincer and Polachek (1974) in studying women’s labor force interruptions. 31. The US Commission on Civil Rights (1986) provides data showing the relatively sharp decline in black male employment after 1960 that occurred despite advances in education and earnings.
REFERENCES Aigner, D. J., & Cain, G. (1977). Statistical theories of discrimination in the labor market. Industrial and Labor Relations Review, 30(January), 175–187.
346
JUNE E. O’NEILL AND DAVE M. O’NEILL
Anderson, D., Binder, M., & Krause, K. (2003). The motherhood wage penalty revisited: Experience, heterogeneity, work effort, and work schedule flexibility. Industrial and Labor Relations Review, 56(2), 273–294. Antecol, H., & Bedard, K. (2004). The racial wage gap: The importance of labor attachment differences across Black, Mexican, and White men. Journal of Human Resources, 39(2), 564–583. Arrow, K. J. (1973). The theory of discrimination. In: O. Ashenfelter & A. Rees (Eds), Discrimination in the labor market. Princeton, NJ: Princeton University Press. Becker, G. S. (1957). The economics of discrimination. Chicago: University of Chicago Press. Becker, G. S. (1981). A treatise on the family. Cambridge, MA: Harvard University Press. Becker, G. S. (1985). Human capital, effort, and the sexual division of labor. Journal of Labor Economics, 3(1, pt. 2), s33–s58. Bergmann, B. (1974). Occupational segregation, wages and profits when employers discriminate by race or sex. Eastern Economic Journal, 82, 103–110. Black, D. A. (1995). Discrimination in an equilibrium search model. Journal of Labor Economics, 13(2), 309–334. Blau, F. D. (1998). Trends in the well-being of American women, 1970–1995. Journal of Economic Literature, 36(March), 112–165. Bound, J., & Freeman, R. B. (1992). What went wrong? The erosion of relative earnings and employment among young black men in the 1980s. Quarterly Journal of Economics, 107(February), 201–223. Brown, C., & Corcoran, M. (1997). Sex-based differences in school content and the male–female wage gap. Journal of Labor Economics, 15(3, pt. 1), 431–464. Budig, M., & England, P. (2001). The wage penalty for motherhood. American Sociological Review, 66(2), 204–225. Cain, G. G. (1986). The economic analysis of labor market discrimination: A survey. In: O. Ashenfelter & R. Layard (Eds), Handbook of labor economics, Vol. 1. Amsterdam: North-Holland. Cavallo, A., & O’Neill, J. (2004). Determinants of the gender gap in occupations and earnings. Paper presented at the meetings of the Society of Labor Economists. San Antonio, May. Darity, W. A., & Mason, P. L. (1998). Evidence on discrimination in employment: Codes of color, codes of gender. Journal of Economic Perspectives, 12(2), 63–90. DeLeire, T., & Levy, H. (2004). Worker sorting and the risk of death on the job. Journal of Labor Economics, 22(4), 925–954. Donahue, J., & Heckman, J. (1991). Continuous vs. episodic change: The impact of affirmative action and civil rights policy on the economic status of Blacks. Journal of Economic Literature, 29(4), 1603–1643. Heckman, J. (1998). Detecting discrimination. Journal of Economic Perspectives, 12(2), 101–116. Herrnstein, R. J., & Murray, C. (1994). The bell curve: Intelligence and class structure in American life. New York: Free Press. Hill, M. A., & O’Neill, J. (1992). Intercohort change in women’s labor market status. In: R. G. Ehrenberg (Ed.), Research in labor economics, Vol. 13. Greenwich, CT: JAI Press. Hill, M. A., & O’Neill, J. (1993). Underclass behaviors in the United States: Measurement and analysis of determinants, Center for the Study of Business and Government, Baruch College, CUNY (revised August, 1993).
What do Wage Differentials Tell About Labor Market Discrimination?
347
Holzer, H., Offner, P., & Sorenson, E. (2004). Declining employment among black less-educated men: The role of incarceration and child support. Unpublished paper, The Urban Institute, Washington, DC. Kahn, L. M. (1991). Customer discrimination and affirmative action. Economic Inquiry, 29(July), 555–571. Korenman, S., & Neumark, D. (1991). Does marriage really make men more productive? Journal of Human Resources, 26(2), 282–307. Korenman, S., & Winship, C. (2000). A reanalysis of the bell curve. In: K. Arrow, S. Bowles & S. Durlauf (Eds), Meritocracy and economic inequality. Princeton, NJ: Princeton University Press. Landes, W. (1968). The economics of fair employment laws. Journal of Political Economy, 76, 507–552. Lundberg, S. J., & Startz, R. (1983). Private discrimination and social intervention in competitive labor markets. American Economic Review, 73(June), 340–347. Mincer, J. (1962). Labor force participation of married women: A study of labor supply. In: C. Christ (Ed.), Aspects of labor economics. Princeton, NJ: Princeton University Press. Mincer, J., & Polachek, S. (1974). Family investments in human capital: Earnings of women. Journal of Political Economy, 82, S76–S108. Mincer, J., & Ofek, H. (1982). Interrupted work careers: Depreciation and restoration of human capital. Journal of Human Resources, 17(1), 3–24. Neal, D. A., & Johnson, W. J. (1996). The role of pre-market factors in black–white wage differences. Journal of Political Economy, 104(5), 869–895. Neumark, D., & Stock, W. (2001). The effects of race and sex discrimination laws. National Bureau of Economic Research. Working Paper No. 8215, April. O’Neill, J. (1990). The role of human capital in earnings differences between black and white men. The Journal of Economic Perspectives, 4(Fall), 25–46. O’Neill, J. (2003). The gender gap in wages, circa 2000. American Economic Review, 93(2), 309–314. O’Neill, D. M., & O’Neill, J. (2005). The federal government and job discrimination, (manuscript). O’Nell, J., & Polachek, S. (1993). Why the gender gap in wages narrowed in the 1980s. Journal of Labor Economics, 11(1), 205–228. Smith, J., & Welch, F. (1989). Black economic progress after Myrdal. Journal of Economic Literature, 27, 519–564. United States Commission on Civil Rights. (1986). The economic progress of black men in America. Washington, DC: Clearinghouse Publication 91, October. United States Department of Labor, Dictionary of Occupational Titles (DOT). (1991). United States Department of Labor, Dictionary of occupational titles (4th ed.) Washington, DC. Waldfogel, J. (1998). Understanding the ‘family cap’ in pay for women with children. Journal of Economic Perspectives, 12(1), 137–156. Western, B. (2002). The impact of incarceration on wage mobility and inequality. American Sociological Review, 67, 526–546.
APPENDIX Sample selection and other methodological issues (see Table A1–A7).
348
JUNE E. O’NEILL AND DAVE M. O’NEILL
Table A1. Mean AFQT Percentile Scores in Year of Test (1980) and 20 Years Later, by Education Completed, Age in 1980 and Sex. 15–18 Years of Age, 1980
19–23 Years of Age, 1980
Black
Hispanic
White
Black
Hispanic
White
Education, 1980 oHS HS graduate Some college College graduate
20.8 36.0 –– ––
28.3 52.4 –– ––
49.6 65.9 –– ––
12.2 23.7 48.7 66.4
16.9 40.1 59.6 94.0
29.5 54.4 79.9 87.9
Education, 2000 oHS HS graduate Some college College graduate
7.1 15.5 28.5 44.6
12.7 24.0 40.4 60.8
15.4 39.0 52.1 76.4
7.8 16.6 36.5 58.8
10.1 30.0 54.6 68.7
22.6 48.0 68.3 82.3
Education, 1980 oHS HS graduate Some college College graduate
21.2 28.9 –– ––
27.5 41.2 –– ––
48.1 59.7 –– ––
8.9 23.5 38.6 61.4
12.9 28.6 52.2 77.8
26.6 49.0 72.8 85.8
Education, 2000 oHS HS graduate Some college College graduate
6.8 14.6 26.8 38.6
9.4 23.8 33.0 50.9
22.9 38.6 50.7 70.6
7.3 17.2 29.6 50.8
10.3 23.0 35.6 62.6
23.1 42.3 59.3 76.8
Men
Women
Source: National Longitudinal Survey of Youth (NLSY79)
Men
Women Regression controls for
Sample size
Analysis restricted to respondents with no additional schooling after 1980 (ages 15–23) Black wage gap
Hispanic wage gap
Age, location, AFQT
Age, location, AFQT, schooling
Age, location, AFQT, schooling, work experience
792
Sample size
Unadjusted gap
Age, location, AFQT
Age, location, AFQT, schooling
Age, location, AFQT, schooling, work experience
0.072 (0.037) 0.033 (0.047)
0.188 (0.041) 0.095 (0.049)
0.144 (0.042) 0.110 (0.048)
0.166 (0.039) 0.099 (0.045)
0.189 (0.022) 0.092 (0.027)
0.106 (0.024) 0.099 (0.027)
0.028 (0.023) 0.072 (0.026)
0.053 (0.022) 0.076 (0.024)
785
0.259 (0.040) 0.102 (0.043)
Hispanic wage gap
Analysis includes all respondents ages 15–23 in 1980 Black wage gap
Unadjusted gap
Regression controls for
0.045 (0.044) 0.003 (0.046)
0.075 (0.045) 0.012 (0.046)
0.015 (0.043) 0.022 (0.044)
2694
2704
0.339 (0.024) 0.198 (0.027)
0.011 (0.025) 0.011 (0.027)
0.020 (0.025) 0.076 (0.026)
0.035 (0.024) 0.015 (0.025)
349
Note: The log wage differentials shown are the partial regression coefficients of dummy (0,1) variables indicating whether the person was black or Hispanic, derived from OLS regressions containing the variables noted. Source: National Longitudinal Survey of Youth (NLSY79).
What do Wage Differentials Tell About Labor Market Discrimination?
Table A2. Regression Adjusted Black–White and Hispanic–White Log Wage Gap for Respondents with no Additional Schooling after 1980 (the Year of the AFQT Test) Compared to Results for the Full Sample of Respondents.
350
Table A3.
JUNE E. O’NEILL AND DAVE M. O’NEILL
Characteristics of Men who were Included and Excluded from the Basic Wage Analysis. Total
All men Number in sample (Percent of sample) Rate of pay Hourly wage (exponent Log) Log hourly wage Characteristics % missing AFQT AFQT percentile score for those with score Years of schooling Years worked since 18 (civil & military combined) % PT of lifetime weeks worked % ever in jail Black men Number in sample (Percent of sample) Rate of pay Hourly wage (exponent Log) Log hourly wage Characteristics % missing AFQT AFQT percentile score for those with score Years of schooling Years worked since 18 (civil & military combined) % PT of lifetime weeks worked % ever in jail
3726 (100.0)
1116 (100.0)
In Basic Analysis
Excluded from Basic Analysis Had wage but not employed in last month, self-employed, missing data and others
No wage reported last 2 yearsa
2694 (72.3)
727 (19.5)
305 (8.2)
15.87
11.52
––
2.76
2.44
––
0.00 42.38
21.18 34.35
7.87 28.90
13.26 17.74
12.47 16.67
12.42 10.89
4.95
7.55
4.38
6.20
15.54
32.13
759 (68.0)
208 (18.6)
149 (13.4)
12.93
7.82
––
2.56
2.06
––
0.00 24.11
15.87 17.85
5.63 12.90
12.91 16.70
12.23 14.58
11.75 8.00
5.09
8.31
5.07
12.65
29.33
43.66
351
What do Wage Differentials Tell About Labor Market Discrimination?
Table A3. (Continued ) Total
Hispanic men Number in sample (Percent of sample) Rate of pay Hourly wage (exponent Log) Log hourly wage Characteristics % missing AFQT AFQT percentile score for those with score Years of schooling Years worked since 18 (civil & military combined) % PT of lifetime weeks worked % ever in jail White men Number in sample (Percent of sample) Rate of pay Hourly wage (exponent Log) Log hourly wage Characteristics % missing AFQT AFQT percentile score for those with score Years of schooling Years worked since 18 (civil & military combined) % PT of lifetime weeks worked % ever in jail
714 (100.0)
1903 (100.0)
In Basic Analysis
Excluded from Basic Analysis Had wage but not employed in last month, self-employed, missing data and others
No wage reported last 2 yearsa
519 (72.7)
138 (19.3)
57 (8.0)
14.87
12.01
––
2.70
2.49
––
0.00 33.60
31.16 26.05
15.79 17.98
12.59 17.71
11.40 16.24
11.53 8.22
5.09
7.27
3.88
6.17
18.12
35.09
1416 (74.4)
381 (20.0)
106 (5.6)
18.14
14.02
––
2.90
2.64
––
0.00 55.38
20.47 46.49
6.60 50.29
13.69 18.31
13.00 17.96
13.24 14.14
4.82
7.23
3.93
2.75
7.09
15.09
Source: National Longitudinal Survey of Youth (NLSY79). Excluding people in the active military service.
a
352
Table A4.
JUNE E. O’NEILL AND DAVE M. O’NEILL
Characteristics of Women who were Included and Excluded from the Basic Wage Analysis. Total
In Basic Analysis
Excluded from Basic Analysis No wage reported Had wage but not last 2 yearsa employed in last month, self-employed, missing data, and Others
All women Number in sample (Percent of sample) Rate of pay Hourly wage (exponent Log) Log hourly wage Characteristics % missing AFQT AFQT percentile score for those with score Years of schooling Years worked since 18 (civil & military combined) % PT of lifetime weeks worked L.F. withdrawal due to family responsibilities % no occupation reported % female in OCC. (for those reporting) Black women Number in sample (Percent of sample) Rate of pay Hourly wage (exponent Log) Log hourly wage Characteristics % missing AFQT AFQT percentile score for those with score Years of schooling Years worked since 18 (civil & military combined)
4085 (100.0)
1248 (100.0)
2704 (66.2)
750 (18.4)
631 (15.5)
12.54
8.89
––
2.53
2.19
––
0.00 39.81
14.40 36.12
5.71 33.81
13.47 15.63
12.98 12.85
12.71 7.93
13.72
15.91
8.93
54.88
73.20
85.10
2.11
3.73
62.28
64.85
66.91
65.89
854 (68.4)
211 (16.9)
183 (14.7)
11.21
7.12
––
2.42
1.96
––
0.00 24.47
10.90 20.51
4.92 13.78
13.30 14.57
12.78 10.94
12.08 6.07
353
What do Wage Differentials Tell About Labor Market Discrimination?
Table A4. (Continued ) Total
In Basic Analysis
Excluded from Basic Analysis No wage reported Had wage but not last 2 yearsa employed in last month, self-employed, missing data, and Others
% PT of lifetime weeks worked L.F. withdrawal due to family responsibilities % no occupation reported % female in OCC. (for those reporting) Hispanic women Number in sample (Percent of sample) Rate of pay Hourly wage (exponent Log) Log hourly wage Characteristics % missing AFQT AFQT percentile score for those with score Years of schooling Years worked since 18 (civil & military combined) % PT of lifetime weeks worked L.F. withdrawal due to family responsibilities % no occupation reported % female in OCC. (for those reporting) White women Number in sample (Percent of sample) Rate of pay Hourly wage (exponent Log) Log hourly wage Characteristics % missing AFQT
780 (100.0)
2057 (100.0)%
9.66
10.41
6.06
57.85
75.36
80.87
3.51
5.21
66.12
64.38
68.30
67.62
492 (63.1)
149 (19.1)
139 (17.8)
12.35
8.93
––
2.51
2.19
––
0.00 30.06
18.12 22.67
6.47 23.37
12.92 15.05
12.04 12.38
12.06 6.53
12.04
12.21
8.24
64.43
82.55
88.49
1.22
2.68
65.47
68.06
68.37
68.97
1358 (66.0%)
390 (19.0%)
309 (15.0%)
13.54
10.01
––
2.61
2.30
––
0.00
14.87
5.83
354
JUNE E. O’NEILL AND DAVE M. O’NEILL
Table A4. (Continued ) Total
In Basic Analysis
Excluded from Basic Analysis No wage reported Had wage but not last 2 yearsa employed in last month, self-employed, missing data, and Others
AFQT percentile score for those with score Years of schooling Years worked since 18 (civil & military combined) % PT of lifetime weeks worked L.F. withdrawal due to family responsibilities % no occupation reported % female in OCC. (for those reporting)
52.98
49.91
50.46
13.78 16.50
13.44 14.07
13.37 9.67
16.87
20.30
10.93
49.56
68.46
86.08
1.55
3.33
58.58
63.97
65.60
63.91
Source: National Longitudinal Survey of Youth (NLSY79). a Excluding people in the active military service.
What do Wage Differentials Tell About Labor Market Discrimination?
355
Table A5. Regression Results on the Black–White and Hispanic–White Wage Gap for Men, Based on Expanded Samplea, NLSY 2000. (Compare with text Table 3). Black–White Differential
Unadjusted log wage differential Log wage differential controlling for (1) Age, MSA, central city, region (2) Variables in (1) plus schooling (3) Variables in (2) plus AFQT (4) Variables in (3) plus Weeks worked in civilian job since age 18 C 52, Weeks worked in military since 1978 C 52
Hispanic–White Differential
0.376
0.205
0.324
0.212
0.230
0.100
0.109
0.034
0.050
0.034
Note: The log wage differentials are partial regression coefficients of dummy (0, 1) variables for black (Hispanic) from a series of OLS regressions containing the explanatory variables noted. The reference group is white non-Hispanic. The statistical significance of the black and Hispanic coefficients is indicated as follows (two-tailed test): significant at the 5% level or less. Source: National Longitudinal Survey of Youth (NLSY79). a The ‘‘Expanded Sample’’ includes those in the basic wage analysis as well as those who were not in the basic sample but reported pay in the last two years. Those with estimated hourly earnings less than $2.50 or more than $125 were excluded as were those missing AFQT score. (See Appendix Table A3 for characteristics of the basic sample compare to those excluded from the basic sample.)
356
JUNE E. O’NEILL AND DAVE M. O’NEILL
Table A6. Regression Results on the Black–White and Hispanic–White Wage Gap for Women, Based on Expanded Samplea, NLSY 2000. (Compare with text Table 6). Black–White Differential Unadjusted log wage differential Log wage differential controlling for (1) Age, MSA, central city, region (2) Variables in (1) plus schooling (3) Variables in (2) plus AFQT (4) Variables in (3) plus: age at 1st birth o30 (0,1) age at 1st birth Z30 (0,1) (5) Variables in (4) plus L.F. withdrawal due to family responsibilities (0.1), Weeks worked in civilian job since age 18 C 52, Weeks worked in military since 1978 C 52, Weeks PT C total weeks worked since age 22
Hispanic–White Differential
0.192
0.109
0.170 0.102 0.031
0.142 0.041 0.059
0.036
0.063
0.041
0.042
Note: The log wage differentials are partial regression coefficients of dummy (0, 1) variables for black (Hispanic) from a series of OLS regressions containing the explanatory variables noted. The reference group is white non-Hispanic. The statistical significance of the black and Hispanic coefficients is indicated as follows (two-tailed test): significant at the 5% level or less. significant at the 10% level. Source: National Longitudinal Survey of Youth (NLSY79). a The ‘‘Expanded Sample’’ includes those in the basic wage analysis as well as those who were not in the basic sample but reported pay in the last two years. Those with estimated hourly earnings less than $2.50 or more than $125 were excluded as were those missing AFQT score. (See Appendix Table A4 for characteristics of the basic sample compare to those excluded from the basic sample.)
What do Wage Differentials Tell About Labor Market Discrimination?
357
Table A7. Regression Results on Gender Wage Gap, Based on Expanded Samplea, NLSY 2000. (Compare with text Table 9). Unadjusted log hourly wage gap Log wage differential controlling for: (1) Age, MSA, region and race, schooling, AFQT (2) Variables in (1) plus life time work experience (3) Variables in (2) plus L.F. withdrawal due to family responsibilities (4) Variables in (3) plus class of worker (5) Variables in (4) plus occupational characteristics (6) Variables in (5) plus percent female in occupation
0.242 0.245 0.120 0.095 0.090 0.071 0.067
Note: The log wage differentials are partial regression coefficients of a dummy (0,1) variable for ‘‘female’’ from a series of OLS log wage regressions containing the explanatory variables noted. The statistical significance of female coefficients is indicated as follows (two-tailed test): significant at the 5% level or less. Source: National Longitudinal Survey of Youth (NLSY79) merged with measures of occupational characteristics (3-digit level) from the September 2001 CPS, the CPS March and the Dictionary of Occupational Titles (1991). a The ‘‘Expanded Sample’’ includes those in the basic wage analysis as well as those who were not in the basic sample but reported pay in the last two years. Those with estimated hourly earnings less than $2.50 or more than $125 were excluded as were those missing AFQT score. (See Appendix Tables A3 and Table A4 for characteristics of the basic sample compare to those excluded from the basic sample.)
This page intentionally left blank
358
PART III: SOCIAL DIVERSITY AND INSTITUTIONS
359
This page intentionally left blank
360
CULTURAL DIVERSITY, STATUS CONCERNS AND THE ORGANIZATION OF WORK$ Chaim Fershtman, Hans K. Hvide and Yoram Weiss ABSTRACT A well-documented human tendency is to compare outcomes with others, trying to outperform them. These tendencies vary across cultures and among different individuals in a given society. To understand the implications of such diversity in status considerations on wages, contracts, sorting and output we use a standard principal agent framework in which firms consist of two workers and a principal. We find that, in equilibrium, firms mix workers with different status concerns to enhance ‘cultural trade’. Although workers may have the same productivity, equilibrium will generate a dispersion in (expected) wages, and workers with status concerns will have more high-powered incentives, work more and earn more than workers who do not care about status. Finally, we find that a more diverse workforce can increase the total output of the economy. This increase in output is a result of the higher effort exerted by the status minded workers that offsets the reduction in effort by those who do not care about status.
$
The Israeli Science Foundation, grant No 0610110951, provided support for this project.
Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 361–396 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24011-6
361
362
CHAIM FERSHTMAN ET AL.
1. INTRODUCTION One of the interesting developments in the last decades is the transformation of homogenous societies to multi-cultural, multi-ethnic societies. It is enough to visit the factories or to walk in the streets of Berlin, Amsterdam, Bangkok or Tel Aviv to understand that cultural and ethnic heterogeneity is the norm rather than the exception. The conventional wisdom is that culturally mixed societies face a higher potential for conflict, because members of such societies have different languages, goals, habits and attitudes. In this paper, we wish to argue that diversity can also be beneficial because of the increased potential for social trade among the different members of society. We shall illustrate this general point in the context of the labor market, where workers of different cultural background can trade in social status. Diversity can take many forms. Much attention has been given to observable features of diversity such as gender, race and ethnic origin.1 In this paper, we focus on diversity in preferences and in particular on status concerns at the workplace. We assume that status within the workplace is directly related to a worker’s total wage earnings.2 However, depending on the reference group this wage can be compared to the wage of different individuals. Workers who are strongly attached to a particular firm will compare their wage to other workers in the firm, and derive their status benefits from association with their co-workers.3 Workers who are less attached to their workplace may compare their wages to the wages of individuals outside the firm, such as family members, compatriots, members of the same ethnic group, or to a random member of the society. Of course, a person cares about his or her wage not only because of the possible social benefits, but also because of the direct benefits that can be obtained through the purchase of market goods. Different individuals may have different preferences over these different benefits. That is, some may care about social status, relative to private goods, more than others. Our analysis will allow for such heterogeneity too. The assumption that social status depends only on wage comparisons is restrictive, because individuals can earn distinction and esteem in ways that are unrelated or conflicting with income. In addition, wages may not be directly observable and other means, such as conspicuous consumption, are used to indicate one’s place in the income distribution. Our interest in wages is derived from our more general interest in the roles of private and social rewards in eliciting effort. A starting point is to recognize that wages play a dual role in this respect. They provide a private reward that can be used to
Cultural Diversity, Status Concerns and the Organization of Work
363
purchase market goods and a social reward that, depending on the wage of others, can be used to obtain non-market goods. To analyze the effects of diversity and status considerations in the work place, we use a standard principal agent framework in which there is one sector, and firms consist of two workers and a principal. The workers’ effort is unobserved and wages are paid based on output, which is a noisy measure of effort. The output of a worker depends on his own productivity and effort and is, in this respect, independent of other workers. However, a worker’s utility may depend on the wage of the other workers. We thus bring together two strands of the literature, incentive contracts and social interactions.4 In this respect, this paper follows earlier work of Akerlof (1982), Kandel and Lazear (1992), Lazear (1989, 1999a) and Rotemberg (1994).5 Our main purpose is to pursue the implications of heterogeneity in social concerns. We, therefore, distinguish between workers who care only about the wage of other workers inside the firm, workers who care about wages of workers outside the firm and workers who do not care about relative wages at all, and analyze several combinations of these types. Initially, we assume that all workers are equally productive. We then consider the case in which status minded individuals are also more productive. To motivate a positive correlation between productivity and status concerns, we consider the choice of workers’ productivity through investment in schooling. We show that separating equilibria exist in which workers who care about status invest in schooling, while those who do not care about status refrain from such investment. Since schooling is observable, this allows firms to sort workers with different preferences for status. We thus incorporate an important and often mentioned role of schools, to identify the individuals who are highly motivated.6 The relative nature of status comparisons together with heterogeneity of preferences are the main driving forces of our model. To see the basic intuition, consider status minded workers who are matched randomly in pairs. Each will try to outdo the other and exert extra effort. The Nash equilibrium outcome, however, is that both workers will generate the same output so that no status gap is created and, in this respect, effort is excessive. Similar coordination problems arise if the two workers have different preferences for status, because the person who does not care about status fails to internalize the impact of his effort on the worker who does. If their efforts can be coordinated, the one who cares about status can exert more effort and have higher wage (consumption), and the other would exert less effort and obtain a lower wage (consumption). In this manner, the one who cares about status gains status, while the one who does not is compensated
364
CHAIM FERSHTMAN ET AL.
for the lower wage by lower effort. We show that such an outcome can be obtained by competition among firms who offer wage contracts that lead to a coordination in the effort of the workers within the firm. Our main results can be easily summarized. First, we find that, in equilibrium, firms mix workers with different status concerns even if they have the same productivity. Second, equilibrium will generate a dispersion in (expected) wages, and workers with status concerns will have more highpowered incentives, work more and earn more than workers who do not care about status. Third, we find that a more diverse workforce can increase the total output of the economy. This increase in output is a result of the higher effort exerted by the status minded workers that offsets the reduction in effort by those who do not care about status. The paper is structured as follows. Section 2 presents the model and the main results. Section 3 extends the analysis to workers that care about global status, i.e., they make wage comparisons with workers outside the firm. Section 4 considers investments in schooling. Section 5 applies some of the theoretical findings to recent evidence on immigration, and Section 6 concludes. All proofs are located in Appendices A and B.
2. LOCAL STATUS: STATUS AT WORK We first examine a diverse society consisting of two groups of workers. One group consists of status minded individuals whose status is determined by the difference between their own wage and the wage of their co-workers. When such workers are matched in a firm, they will try to outdo each other, by exerting more effort. We refer to this pattern of behavior as status oriented or competitive behavior. The other group consists of individuals who may have no status concerns, or status concerns that are unrelated to their wage. Such workers will not act in a competitive manner.
2.1. The Model The economy consists of a large number of workers and firms. Firms offer a wage contract and workers choose in which firm to work, depending on the contracts they are offered and the characteristics of the firms. There is a free mobility of workers between firms and no entry or exit costs for firms. Production: The output y~ i of worker i depends on his productivity ti, his effort ei and a random shock ei. We let y~ i ¼ ti ei þ i ; where ei denotes his
Cultural Diversity, Status Concerns and the Organization of Work
365
effort, ti his productivity, and ei is an iid random shock with zero mean and bounded variance. We start by assuming that all workers have the same productivity and normalize by setting ti to unity. We will later consider the case in which workers may have different ti. We assume that firms have a capacity constraint and employ only two workers. Each firm’s output is the sum of the output of the two workers. The assumption of independence in production is not strictly needed but allows us to focus on the interactions that results from preferences. Preferences and Diversity: Workers are risk neutral and their utility is assumed to be linear in wages. Some of the workers are assumed to care about their local status, which is assumed to be captured by the difference in their wage from the wage of other workers in the firm.7 Effort is costly and we denote by v(ei) the cost (or disutility) of effort. The utility function is assumed to be of the form ui ¼ w~ i þ di b w~ i w~ j vðei Þ (1)
where w~ i is agent i’s realized wage and w~ j the wage of the other worker who is employed by the same firm. b represents the relative importance of local status compared with own consumption, and di indicates whether or not individual i cares about his local status. We let there be only two types of workers. Workers who care about their relative wage in the firm (local status), for whom di ¼ 1; and workers who do not care about any wage comparisons, for whom di ¼ 0: The proportion of type 1 workers in the population is x, where 0oxo1: Thus, x represents the degree of diversity, where x ¼ 1 and x ¼ 0 represent homogenous populations while x ¼ 1=2 is a fully diverse population. Wage Contracts and the Choice of Effort: We assume that effort is not observed but the output y~ i of each agent is observable and contractible. In addition, a worker’s preference type is observable.8 Given the utility function (1) it is sufficient to condition the wage of each worker on his own output and to restrict attention to linear contracts, which will achieve first best levels of effort.9 We thus set w~ i ¼ si þ ai y~ i ;
i ¼ 1; 2
(2)
where si is the salary and ai the ‘piece rate’. Given the contract, workers choose effort levels to maximize their expected utility, yielding ei ¼ ai ð1 þ di bÞ;
i ¼ 1; 2
(3)
366
CHAIM FERSHTMAN ET AL.
The implied expected profits made by the firm are, EðpÞ ¼ e1 ð1
a1 Þ þ e2 ð1
a2 Þ
s1
s2
(4)
Given the characteristics of the workers that join the firm, di, the risk neutral firm chooses the wage parameters (si, ai); i ¼ 1; 2; so that expected profits are maximized subject to each worker obtaining at least his reservation utility ri. The workers’ reservation values, ri are endogenously determined and depend on the contracts offered by other firms. We must therefore solve for an equilibrium that specifies contracts in all firms, using the condition that workers cannot benefit by switching employers, and firms cannot gain from changing the contract that they offer. At equilibrium, some firms will employ workers with identical preferences and other firms will employ workers with different preferences. We refer to these two types of firms as homogenous (or segregated) and heterogeneous (or mixed), respectively. We shall see that, because of the interdependence in preferences, firms with a different mix of workers will provide different incentives to workers of a given type. Competition requires that a contract offered by a firm must be efficient in the sense that it maximizes the payoff of each member of the coalition consisting of the firm and two workers. Otherwise another firm could have raised its profits by offering a different contract, which would attract workers. Consider, therefore, the maximization of the firm’s profit subject to the constraints that each worker receives at least his reservation utility. The associated Lagrangian is L ¼ EðpÞ þ l1 ½Eðu1 Þ
r1 þ l2 ½Eðu2 Þ
r2
(5)
The linearity assumption in preferences and the wage contracts imply transferable utility and allows us to simplify the analysis considerably. In particular, l1 and l2 are constants that depend only on the preferences of the two workers. We can, therefore, determine the incentive structure provided for workers of each type and the consequent effort level before considering the full market equilibrium, because these are independent of the reservation utility levels. For any choice of the incentive parameters (a1, a2) the maximization with respect to (s1, s2) yields li ¼
1 þ 2dj b ; 1 þ d1 b þ d2 b
i; j ¼ 1; 2
(6)
Using these conditions to eliminate the l1 and l2 from L, we see that the firm chooses the incentive parameters (a1, a2) that induce the worker to exert
Cultural Diversity, Status Concerns and the Organization of Work
367
first best levels of effort that maximize the total ‘pie’ given by W ¼ e1 þ e2
l1 vðe1 Þ
l2 vðe2 Þ
(7)
and the induced effort levels must satisfy li v0 ðei Þ ¼ 1;
i ¼ 1; 2
(8)
Unless stated otherwise, we shall assume hereafter that vðeÞ ¼ 12 e2 so that ei ¼ l1i : 2.1.1. Homogenous Firms Consider first a firm that hires two workers with the same preferences (either they both care about their local status or they both do not care about it). Since in such a case li ¼ 1; we obtain that Proposition 1. (i) The first best effort levels in homogenous (segregated) firms are independent of whether workers do or do not care about local status and given by e1 ¼ e2 ¼ 1: (ii) To implement this outcome, the firm offers different contracts to the two types, ai ¼ 1 if the two workers do not care about status and ai ¼ ð1 þ bÞ 1 ; i ¼ 1; 2; if they both care about their local status. The independence of effort from b follows from the fact that the local status concerns are purely relative and wash out when the firm hires workers with identical preferences. When both workers care about local status, incentives are slackened (i.e., ai o1). Intuitively, agents in such a case are eager to invest more effort to gain status. If such a worker would get the incentives ai ¼ 1; he would try to compete with other workers and work harder, to the point where the marginal product exceeds the marginal cost of effort (taking into account the negative effect on the other worker’s status). In such a ‘‘rat-race’’ the firm acts as a coordinator and mitigates the wasteful competition by reducing the monetary incentive for effort, compensating the workers with a fixed payment. 2.1.2. Heterogeneous Firms Consider now a firm that hires two workers, one who cares about his local status (a type 1 worker) while the other does not (a type 2 worker). Given the appropriate l1, l2 from (6), we obtain the following: Proposition 2. (i) In a heterogeneous (mixed) firm, the first best effort level of the status minded worker, e1 ¼ 1 þ b; exceeds the effort level of
368
CHAIM FERSHTMAN ET AL.
the worker who does not care about status, e2 ¼ ð1 þ bÞ=ð1 þ 2bÞ: (ii) To implement this outcome, the firm offers different contracts to the two types, with a1 ¼ 1 to the worker who cares about status and a2 ¼ 1 þ b=1 þ 2bo1 to the worker who does not. The important insight of Proposition 2 is that the status minded worker gets a stronger incentive to exert effort in a heterogeneous firm than in a homogenous firm. The worker that does not have a local status concern on the other hand is given a weaker incentive to exert effort in a heterogeneous firm than in a homogenous firm. This is because the firms internalize the status concerns and can increase the utility of both workers by reducing the effort of the non-competitive worker. The firm has an interest in doing so, because it can then attract workers at lower wages. Propositions 1 and 2 together suggest that equal workers will be treated equally; the principal will only differentiate status when workers are heterogeneous in their demand for status. A related result is obtained by Auriol and Renault (2000). They allow firms to provide, at no cost, status symbols that are independent of wages and show that the principal will use the status instrument only if workers have been proven to be of different ability.
2.2. Market Equilibrium As a benchmark, consider first the equilibrium with a homogenous workforce, i.e., x ¼ 0 or x ¼ 1: In such societies, all firms employ two workers with the same preferences. We have shown above that in such a case, workers exert the same level of effort, e ¼ 1; whether or not he cares about status. This implies that the level of effort and aggregate output are independent of whether x ¼ 0 or x ¼ 1: The independence of effort from status concerns in homogenous societies is a consequence of the role that firms play in our model. Recall that firms compete for workers by offering them a wage contract, and in doing so they have an interest to eliminate all externalities that are internal to the firm. In the absence of such coordination by the firms, homogenous societies consisting of different cultural types display different economic performance. Imagine, instead, a society in which everyone is self-employed, but individuals meet socially and compare their wages. For example, wage comparison is made with a randomly matched member of society. Then, in a homogenous society where everyone cares about status, the effort of each
Cultural Diversity, Status Concerns and the Organization of Work
369
worker will be e ¼ 1 þ b; while in a homogenous society consisting of individuals who do not care about status the effort of each worker will be e ¼ 1: Although output is higher in the first society, welfare is lower because workers engage in an inefficient rat-race, where workers try to outdo each other but the end outcome is that they gain no status. Let us now consider the equilibrium in a heterogeneous society with x 2 ð0; 1Þ: Recall that there is a large number of firms with free entry and exit, and a fixed number of workers, with a fraction x of the competitive type, and the worker type being observable. Workers choose the firm they work for and there is a free mobility of workers among firms. Since each firm employs exactly two workers, there must be some firms that employ two workers of the majority type. That is, at equilibrium, there are always segregated (homogenous) firms that employ two workers of the majority type. Since we assume free mobility, workers of the majority type must have the same utility in segregated and heterogeneous firms. From our previous analysis, two workers of type j who work in the same firm will exert the same level of effort, 1, and their cost of effort equals 12: The total expected output of the firm is equally divided and each worker has an expected wage of 1 and expected utility of 12: Thus, the reservation utility of the majority type is rj ¼ 12: We can now calculate the utility of a minority type worker in a mixed firm and compare it to what she might get in a segregated firm consisting of two minority workers. If we can show that the minority workers get a higher utility working in mixed firms then, at equilibrium, all workers of the minority type will work in mixed firms. If the minority workers get higher utility working in a homogenous firm then there are no heterogeneous firms in equilibrium, implying that some firms will hire only type 1 workers and some firms will hire only type 2 workers. Proposition 3. (Industry Structure). Heterogeneous (mixed) firms are always formed in equilibrium. When a type j is the minority, all workers of this type will be employees in heterogeneous firms. When a type j is the majority type, then there are some homogenous (segregated) firms employing two workers of type j and some heterogeneous firms employing the two different types of workers. Proof. See Appendices A and B. The intuition for Proposition 3 is as follows: If the two types of workers mix, and both types exert the same effort as in segregated firms, the expected utility of type 1 worker increases while the expected utility of type 2 and the
370
CHAIM FERSHTMAN ET AL.
firms’ expected profits remain the same. By coordinating efforts levels, raising the effort of type 1 worker and reducing the effort of the type 2 worker, the firm can further increase the expected utility of the minority type, keeping the expected utility of the majority type fixed at its reservation value, while holding expected profits constant. Thus, mixing is motivated by the modified behavior of workers with different tastes within firms that generate local status that one worker enjoys and the other is willing to provide.10 2.3. Equilibrium Wage Structure Wage differentials occur in this model as a result of the different incentives that firms offer to workers of different types and the induced differences in effort (and not because of differences in productivity). Recall that wages depend not only on the worker’s effort but also on random shocks. We, therefore, discuss the expected total wage payment that each worker receives. To economize on notation, we omit the expectation operators and denote the expected wage of a worker of type i by wi and the expected output by yi. In the absence of diversity, there is only one level of (expected) wages in our model. When some of the workers care about status, there are three levels of (expected) wages, two for the majority type and one for the minority type. Proposition 4. For any x 2 ð0; 1Þ; (i) Status minded workers in heterogeneous firms earn more than workers without status concern. (ii) The mean expected wage in heterogeneous firms exceeds the mean wage in homogenous firms. (iii) When status minded workers are in the minority, i.e., hom het xo12; whet 1 4w2 4w2 : (iv) When status minded workers are in the majority, i.e., x412; there are two possible orders, depending on the intensity of status concerns, b. hom het het hom For a low b, whet 4whet : 1 4w1 2 and for a high b, w1 4w2 4w2
Proof. See Appendices A and B. The reason that type 1 workers always receive a higher (expected) wage than type 2 workers in mixed firms is the stronger incentive to exert effort provided to the status minded workers. Thus, if type 2 workers are the
Cultural Diversity, Status Concerns and the Organization of Work
371
majority, their reduced effort must also imply a lower expected wage, since their expected utility is constant. On the other hand, the type 1 workers, who exert more effort, are compensated partially by having a higher local status and partially by increased wages. If type 2 workers are in the minority, such workers will be compensated for the association with type 1 workers and, for sufficiently strong status concern, the type 2 workers in mixed firms earn more than type 1 workers in homogenous firms. In the above analysis, wages will be dispersed even if workers are identical in terms of ability or productivity. In equilibrium, status minded workers in mixed firms get the highest wages. Workers that are not concerned with their local status and work in mixed firms get the lowest wage in the society (for sufficiently low b). Their wage would have been higher if they could work in segregated firms. Although we get wage dispersion among equally productive workers, we also have wage compression in the following sense. In homogenous firms, both workers receive expected wages equal to their expected output. In mixed firms, the expected wages of status minded workers fall short of their expected output, while workers who do not care about status receive a wage that exceed their expected output. Thus, the competitive worker ‘‘transfers’’ part of his output to the non-competitive worker, as a payment for the association and for the willingness to reduce his effort. The size of this payment depends on the relative supply of the two types and the incentives to exert effort provided through the contract.
2.4. Differences in Output Different firms provide different incentives and will have different levels of output. The total expected output of homogenous firms is 2, while heterogeneous firms have an expected output of 2 þ 2b2 =ð1 þ 2bÞ: Thus, heterogeneous firms have higher expected output than homogenous firms and the gap rises with b. In other words, the status minded worker in a mixed firm is induced to increase his effort by more than the person who does not care about status is induced to reduce his effort. This result, however, depends on the functional form of the cost of effort, v(e), and on the assumption that the two workers have the same productivity. Let us first discuss the role of the marginal cost of effort. Recall that, by (8), the effort levels of the two workers in the heterogeneous firms must satisfy li v0 ðei Þ ¼ 1; so that v0 ðe1 Þ ¼ 1 þ b and v0 ðe2 Þ ¼ ð1 þ bÞ=ð1 þ 2bÞ: Thus, if g(x) is as the inverse of the marginal cost of effort, the expected output of a
372
CHAIM FERSHTMAN ET AL.
homogenous firm is yhom ¼ 2gð1Þ and the expected output of a heterogeneous firm is yhet ¼ gð1 þ bÞ þ g 1 þ b=1 þ 2 : For b ¼ 0; output is the same in the two types of firms. A sufficient condition for yhet yhom 40 for any positive b is that gðxÞ is convex. Some degree of concavity of gðxÞ is consistent with rising output, but it is easy to construct examples in which output declines, because the increase in effort by the status minded worker is smaller than the reduction in effort by the worker who does not care about status.11 Consider now the role of differences in productivity and further assume that there is a positive correlation between status concerns and productivity such that status minded individuals are also more productive (Section 4 motivates this assumption in some detail). One can then generalize Proposition 3 and show that heterogeneous firms are formed in equilibrium (see Appendix A). Maintaining the assumption vðeÞ ¼ 12 e2 ; the effort levels of the two types of workers are ei ¼ ti in a homogenous firm and e1 ¼ ð1 þ bÞt1 and e2 ¼ ð1 þ bÞt2 =ð1 þ 2bÞ in a heterogeneous firm. The corresponding outputs are yhom ¼ 2ti and yhet ¼ t21 ð1 þ bÞ þ t22 ð1 þ bÞ=ð1 þ 2bÞ: Compared with the case of identical productivities, b has a stronger positive effect on yhet when t1 4t2 ; because the heterogeneous firm shifts the allocation of effort in the direction of the more productive worker. For the same reason, one would expect that output will rise for a broader class of specifications for the cost of effort, vðeÞ: 2.5. Aggregate Output Proposition 3 implies that mixed firms will be formed to the extent possible. Therefore, the degree of diversity, x, affects the relative number of mixed firms, which produce higher output. We can use these results to determine the impact of the degree of diversity on the total aggregate output. We continue to assume that the cost of effort is vðei Þ ¼ 12 e2i and consider first the case in which all the workers have the same productivity. In this case, Proposition 5. (Diversity and output). (i) An increase in the proportion of the status minded individuals in the population, x, raises aggregate output when xo0:5 and reduces aggregate output when x40:5: Consequently, aggregate output is maximized when the population is (almost) evenly divided between the two types of agents. (ii) An increase in degree of competitiveness, b, raises aggregate output. Proposition 5 illustrates the possible advantage of having a diverse society with individuals that have different social concerns. The diversity in our case
Cultural Diversity, Status Concerns and the Organization of Work
373
implies a higher aggregate output as well as a higher average wage. The mechanism that leads to such an outcome is ‘cultural trade’. Firms manipulate the incentives they provide, inducing the worker that care about status to exert more effort, while inducing the worker who does not care about status to reduce his/her effort. In the above discussion, all workers were assumed to be equally productive. A question is whether we get the same result if the status minded worker is more productive, i.e., t1 4t2 : In such a case, the distribution of types influences both the gains from mixing and the average workers’ productivity. The impact on aggregate output depends on the relative strength of these two opposing effects. To illustrate this issue, we shall maintain the assumption that the cost of effort is vðei Þ ¼ 12 e2i then aggregate output, Y, is i h 8 ð1þbÞ > 2n x t21 ð1 þ bÞ þ t22 ð1þ2bÞ þ ð1 2xÞt22 if xo0:5 < h i Y¼ (9) ð1þbÞ > : 2n ð1 xÞ t21 ð1 þ bÞ þ t22 ð1þ2bÞ þ ½1 2ð1 xÞt21 if x40:5
Raising b will clearly result in a higher aggregate output. Given x, the number of heterogeneous firms is fixed, but as we showed above a higher b raises output even if workers have the same productivity. But now output will be raised even further since the workers that raise their effort are the high productivity workers. An increase of x has a more complex effect. When xo0:5; the majority of workers are of type 2, and an increase in x implies also that low productivity type workers are replaced by high productivity, type 1 workers, who also exert more effort when placed in heterogeneous firms. However, if x40:5 then the new type 1 workers are placed in homogenous firms where they exert less effort, so that aggregate output will decline if b is sufficiently high. That is, when x40:5 whenever b is small, an increase in the proportion of agents who care about status, x, raises aggregate output. For b sufficiently large, aggregate output rises if xo0:5 and declines if x40:5: In such a case aggregate output is maximized when the population is (almost) evenly divided between the two types of agents. Fig. 1 illustrates Proposition 5 and describes the output per worker for t1 ¼ 1:5; t2 ¼ 1 and b ¼ 0;0.5,1. We remark again that all the above results depend also on the functional form of the cost of effort v(e). If this function is convex, it is possible that mixed firm will produce less and therefore an increase in their number may lead to a reduction in aggregate output.
374
CHAIM FERSHTMAN ET AL. β=1
2.6 2.4 β=0.5
Output
2.2 2
β=0
1.8 1.6 1.4 1.2 1 0
0.2
0.4
0.6
0.8
1
Prop. of type1workers
Fig. 1.
Output per Worker for Different Level of Status Concerns, b.
3. DIFFERENT REFERENCE GROUPS: LOCAL AND GLOBAL STATUS In the previous section, we assumed that some of the workers care about local status while the others do not care about status at all, or have a status ranking that is independent of wages. In this section, we consider a different profile of preferences, such that all the workers derive their status from wage comparisons, but differ in their social reference group. Specifically, we assume that type 1 workers care about local status at work, while type 2 workers compare themselves (i.e., their wage) to a reference group outside the firm. This reference group can be the whole population or a more specific group such as individuals of the same ethnic background or age. We will refer to such status concerns as global, since in such a case the reference group of the worker is outside the firm and therefore cannot be manipulated by the firm. Preferences are now given by 1 2 u1 ¼ wi þ bðwi wj Þ e (10) 2 i 1 2 e (11) u2 ¼ wi þ bðwi w ¯ 2Þ 2 i where w ¯ 2 is the average wage of the relevant reference group of type 2 workers. We assume for simplicity that the intensity of status preferences, b, is identical for the two groups.12
Cultural Diversity, Status Concerns and the Organization of Work
375
As in the previous section we assume linear contracts wi ¼ si þ ai yi ; i ¼ 1; 2; and consider first the case of equally productive workers. Following the same analysis as in the previous section, we obtain Proposition 6. (i) In a homogenous firm with two workers that care about their local status, the first best level of effort is lower than in a homogenous firm with two workers care about their global status. The respective levels of effort are ei ¼ 1 and ei ¼ ð1 þ bÞ: (ii) A heterogeneous firm that hires two different workers, one that cares about local status and one that cares about global status, will induce the status minded worker to exert more effort. The respective first best levels of effort are e1 ¼ 1 þ b and e2 ¼ ð1 þ bÞ2 =ð1 þ 2bÞ: Proof. See Appendix B. The difference between the two types of homogenous firms is that the firm does not coordinate the reduction of effort when the reference group is outside the firm. In contrast, a heterogeneous firm employing one worker of each type takes into account the externalities that the ‘‘global’’ worker imposes on the ‘‘local’’ worker. However, the attempt to reduce the effort of the global worker makes the association less attractive to this type and we obtain Proposition 7. In a society with two types of workers, type 1 that cares about local status and type 2 cares about his global status, only homogenous firms are formed in equilibrium. Proof. See Appendix B. The intuition for this non-mixing result is that when type 2 workers care about global status, it will be too costly for type 1 workers to compensate the type 2 workers to reduce effort, because such a reduction impinges on their global status. In equilibrium, there is no trade in status concerns within firms. The equilibrium is, therefore, characterized by a complete segregation of the two types. Individuals with local status concerns will work in separate firms and earn a (expected) wage of 1, while the individuals with global status concerns will work in different firms and earn an expected wage of 1 þ b: The situation is quite different if the two types have different productivity and t1 4t2 so that the person who cares about local status also has a higher productivity. In this case, workers in homogenous firms exert the effort levels
376
CHAIM FERSHTMAN ET AL.
¼ ¼ t21 ; whom e1 ¼ t1 and e2 ¼ t2 ð1 þ bÞ and the corresponding wages are whom 2 1 2 t2 ð1 þ bÞ: Now consider breaking the two homogenous firms and mixing the two types of workers in a heterogeneous firm. The effort levels in the heterogeneous firm are give by e1 ¼ t1 ð1 þ bÞ and e2 ¼ t2 ð1 þ bÞ2 ð1 þ 2bÞ so that the person who cares about local status raises his effort and the one who cares about global status reduces his effort. If it can be shown that even without this modification in effort the person who cares about status gains while the one who does not is indifferent, which would hold if whom 1
whom ¼ t21 2
ð1 þ bÞt22 40
(12)
then by modifying the effort in the heterogeneous firm, the two workers can be made even better off, and therefore mixing will occur. We can, therefore conclude that Proposition 8. Assume that a worker who cares about local status is also more productive than the type of worker who care about global status, t1 4t2 : Then, if the difference in productivity is sufficiently large, so that condition (12) holds, mixed firms are formed in equilibrium. In such firms, the status minded and more productive worker will exert more effort. More generally, differences in productivity are a separate source for gains from trade in status concerns that exists even if workers do not modify their behavior. If this difference is large enough, workers who care about local status are willing and able to pay for the association with workers who care about their status outside the firm.
4. UNOBSERVABLE STATUS CONCERNS So far we assumed that preferences are observed by firms. One possible justification for such a setup is that preferences are related to some observable attributes such as gender, age and ethnicity. Another possibility is that workers have available signals from which their preference for status can be inferred. Suppose now that the status concerns of workers, di, are not observed by the firms. Assume further that the workers’ productivity is determined by investment in schooling prior to entry into the labor force. For simplicity, we assume two schooling levels (0 and 1) and let the cost of acquiring 1 unit of schooling be x. The productivity level of an agent without schooling is normalized to 1, and the productivity level of an agent with schooling is labeled t1, where t1 41:
Cultural Diversity, Status Concerns and the Organization of Work
377
Consider now the following two-stage model: At the first stage, each agent decides whether or not to invest in education. The second stage is a market game in which firms offer employment contracts and workers choose the firm they work for and the wage contract. Firms offer contracts based on the education level that they observe. The wage contracts do not depend on the agents’ preferences, which is not observable by firms. There is a free entry to the market so firms will enter as long as they can offer contract that yield non-negative profits. The focus of our analysis is the existence of a separating equilibrium, in which status minded workers acquire schooling, while workers who do not care about status do not acquire schooling. Such a separating equilibrium can justify our previous assumption that t1 4t2 : We continue and refer to socially minded individuals as type 1 agents and to those who do not care about status as type 2 agents. Proposition 9. For any given b40 and t1 41; there exist an interval for the cost of schooling, x, such that a separating equilibrium exists, where only the status minded individuals invest in schooling. Proof. See Appendices A and B. The existence of separating equilibria is supported by the fact that a worker who does not care about status and acquires schooling will exert less effort than a status minded worker facing the same incentives. Therefore, his marginal benefit from schooling is lower and he will refrain from investment at same costs at which the social minded workers find it profitable to invest. The higher is the marginal utility from status, b, the larger are the differences in effort and earnings between the two types and, therefore, it will be easier for the socially minded agents to separate themselves. If socially minded workers are in the minority, they pay less to those who do not care about status for the association. Thus, a pretender (i.e., a type 2 worker who acquires schooling) will obtain a higher fixed payment. In this case, a higher cost of schooling is required to separate the two types. Because schooling raises productivity, it is clear that if the costs of schooling are sufficiently low, everyone will acquire schooling, while if the costs are high, no one will acquire schooling. In either of these cases, schooling has no signaling value. It is still possible for a separating equilibrium to exist, because firms can offer different contracts and workers will self-select based on their preferences as in Rothschild and Stiglitz (1976). However, we find signaling through schooling the more interesting case, because it appears that schools do in fact identify not only ability, as
378
CHAIM FERSHTMAN ET AL.
suggested by Spence (1973) and others, but also the response to incentives, a factor which we may refer to as motivation. It has been recognized by many observers that schooling is a source of attaining higher social status (Weiss & Fershtman, 1998). It is not surprising, therefore, that agents who care about status invest more in schooling. The more subtle issue concerns the impact of status on the monetary returns for schooling. If status is highly valued, then educated workers need not be compensated for the costs of investment, and may in fact have lower earnings, which eventually can detract from their social status. The fact that the market pays a substantial return for schooling, exceeding the return of other investments, suggests that educated workers differ in their attributes from the non-educated workers. Most of the empirical research on this problem concentrates on the role of ability, as an unmeasured attribute that explains the returns for schooling. Recent findings indicate that ability has only a small impact on the monetary returns from schooling (see Ashenfelter, Harmon, & Oosterbeek, 1999). Our analysis suggests a potential role for unobserved effort, or motivation, whereby the highly educated are compensated, in part, for additional effort. This view is consistent with the positive correlation between education and measured effort in the form of longer hours (see Blundell & MaCurdy, 1999).
5. APPLICATION Immigration is an important area in which social and economic considerations interact. Natives often express strong anti-immigration sentiments, claiming that immigrants ‘‘take jobs away’’, ‘‘increase crime rates’’ and that ‘‘immigration should be reduced’’. The strength of these sentiments vary by host country, source country and attributes of the respondents. Generally, more educated natives tend to be more favorable toward immigrants (see Bauer et al., 2000; Dustman & Preston, 2002). Immigrants and guest workers often work at low-skill jobs receiving lower wages than natives. The large wave of immigration from the former Soviet Union to Israel in the early 1990s illustrates this point. These highly educated immigrants found work very quickly but in low-skill jobs. Within five years, only 30% of the immigrants found jobs that correspond to their work and occupation in their country of origin. Immigrants were initially concentrated in particular occupations and jobs such as cleaning, personal services and gas stations, irrespective of their imported skills. With time, they gradually climbed up the occupational ladder and found jobs in highskill industries such as health and hi-tech. Consequently, their average wage
Cultural Diversity, Status Concerns and the Organization of Work
379
went up and the variance too, reflecting improved matching. Yet, immigrants continued to receive lower wages than their native co-workers. (see Weiss, 2000; Eckstein & Weiss, 2004). Remennick (2004) describes in detail the social interactions between immigrants and natives in the same work place.13 Although the Russian trained workers were older and had more formal schooling than their Israeli co-workers, they received lower wage, worked harder and also had to wait longer to be promoted. When interviewed, immigrants reported that natives free-rode on their effort, ‘‘trying to gain some short cuts for themselves.’’ The local workers felt that these differences were fair, because ‘‘this is how you start every new job, especially in a new country’’ and because formal education plays little role in the actual performance of the work. Yet, to maintain their self-respect, many veterans did in fact step up their effort and some went back to school to acquire more schooling. As expressed by one of the local workers: ‘‘I think that the largest contribution of the Russian workers to this organization is the sense of competition that made many veterans step out of the rut and seek more knowledge, take refresher courses, to start reading or even go back to the university.’’ These features are generally consistent with our model, suggesting some trading of status concerns. During their initial period in the host country, immigrants seem less concerned than natives about the implied loss of status, because their reference group includes mainly other immigrants or related individuals in the source country. They maintain their culture in forms of Russian speaking theaters, newspapers and TV channels (See Horowitz, 2005). They also report almost no social contact with native Israelis.14 These features were facilitated by the large size of the immigration wave. According to our model, immigrants with strong status concerns outside the firm will choose not to mix with local workers and work with other immigrants in segregated firms. In this respect, social segregation can lead to occupational segregation, whereby immigrants concentrate in specific industries, firms or jobs. However, with time the immigrants find jobs in the high-skill occupations that correspond to their skills and mix with native high-skill workers. Our model predicts that in such firms, the immigrants receive lower wage and exert less effort than their native’s co-workers (see Propositions 7 and 8). The empirical findings confirm the prediction on wages but, on average, newly arrived immigrants work more than comparable natives, which seems to be inconsistent with our model. However, a consideration that we have not incorporated in our analysis is that immigrants have important investment motives that may induce them to work longer hours, such as language training and acquiring familiarity with local market conditions (See Eckstein & Weiss, 2004).15
380
CHAIM FERSHTMAN ET AL.
Despite the very large entry of high-skill immigrants (the number of physicians and engineers more than doubled in a few years), educated native Israelis were more supportive of the new immigration than the less-educated natives, as in other countries with much lower entry of skilled workers (see Horowitz, 2005). In part, this is an outcome of the delay in the entry into the high-skill occupations, but also reflects the fact that immigrants entered and remained at the lower level positions in these occupations. For instance, most Russian trained physicians found jobs as physicians within 10 years in Israel, but these jobs were often in HMOs rather than hospitals. Thus, native physicians were bumped up, and gained both in wages and status (Sussman & Zakai, 1998). An important finding is that even a large and concentrated wave of immigration such as happened in Israel (1 million immigrants entered a country with a population 4.5 million within a decade) had only negligible impact on the wage and employment of natives. This lack of effect is consistent with findings in other countries and is usually ascribed to entry of capital that complements the entry of workers and mobility of native workers into different jobs or locations (see Card, 2001; Friedberg, 2001; Eckstein & Weiss, 2002). According to our model, natives who associate with immigrants in the work-place have no change in utility because, as long as they remain the majority, they continue to receive their reservation utility that reflects the option to move to alternative (homogenous) firms. However, the wage and effort of these native workers rise, as they obtain stronger incentives to exert effort. Thus, our model provides a channel by which immigrants can exert a positive effect on the wages of local workers with the same productivity, based on the difference in preferences between the two groups. Lack of evidence that natives are hurt economically by immigration suggests that the anti-immigration sentiments expressed in opinion polls may be exaggerated and unfounded. Possibly, the same bias arises in noneconomic aspects such as crime and ‘‘social values’’. Whether and why there is a systematic bias would be a fascinating topic for further research.
6. CONCLUDING REMARKS While social concerns are typically the focus of sociological studies, they cannot be ignored by economists, because of their impact on economic variables. Indeed, there is a growing interest among economists in the effects of social concerns on various economic outcomes, such as investment in education, occupational choice, growth and the distribution of income.
Cultural Diversity, Status Concerns and the Organization of Work
381
Societies and individuals may differ in the weight that they put on individual incentives and equity considerations. Thus, one question for analysis is the economic implications of such cultural differences on economic performance.16 However, many societies are also characterized by cultural diversity that reflects the heterogeneous preferences of their members. Does a society benefit from such a diversity or is homogeneity a key for political stability and economic success? This is a debated issue with important policy implications in the areas of education, immigration, taxation and welfare programs.17 As economists recognize, differences in preferences can enhance trade and for this reason diversity may be beneficial.18 In this paper, we extend the basic economic concept of ‘‘gains from trade’’ in goods to trade in social or cultural concerns and show that diversity may be beneficial when such a trade is feasible. The question is how societies may develop mechanisms that facilitate such trades. We have emphasized how the labor market may fill this role. Firms in our model offer the workers who care about status a wage contract with higher sensitivity to output but a lower fixed payment than the contract that is offered to workers who do not care about status, thereby inducing these workers to modify their behavior and creating internal differences that one worker enjoys and the other is willing to supply. In this manner, firms internalize the workers social concerns and make cultural trade feasible. The cultural trade inside firms may have important effects on the wage dispersion and aggregate output of the economy. Economists are well aware of the other important considerations that are involved in the debate on the desired (or feasible) level of heterogeneity, considerations that may conflict with the gains from trade. Some important issues are associated with potential disagreements about the provision of public goods in a multi-cultural political economy. In fact, there is some evidence that ethnic divisions and cultural fractionalization do reduce spending on public goods and weaken economic growth.19 From a policy perspective, however, it is not sufficient to examine aggregate outcomes. It is important to trace the variety of ways in which cultural differences operate, including the impact on the labor market that we have chosen to analyze.
NOTES 1. Alesina, Devleescauwer, Easterly, Kurlat, and Wacziarg (2002) and Fearon (2003) examine the economy wide implications of diversity. The management literature identifies different benefits and costs of diversity within firms. Thomas and Ely (1996), and Ely and Thomas (2001) argue that individuals of different cultural background can exchange ideas and aggregate information. Tsui, Egan and O’Reilly
382
CHAIM FERSHTMAN ET AL.
(1992) argue that diversity is associated with diminished attachment to the firm, as measured by quits and absences. Lazear (1999b) discusses the social interaction between immigrants and natives in a multi-cultural society. 2. The relationship between wages and status originates with Adam Smith. Frank (1985a) analyzes this relation within firms, and Fershtman and Weiss (1993) and Fershtman, Murphy, and Weiss (1996) analyze it across occupations. For empirical and experimental evidence on the relevance of wage comparisons see Clark and Oswald (1996) and Zizzo and Oswald (2001). Social concerns about wages go beyond status concerns and also relate to fairness and reciprocity. See, for instance, Hicks (1963), Reder (1957) and Akerlof and Yellen (1990). 3. Frank (1984a, b, 1985) was the first to recognize the importance of local status that is provided by firms. 4. Gibbons & Waldman (1999) and Prendergast (1999) provide recent surveys of agency problems within firms. Sobel (2005) provides a general survey of social interactions. Fehr and Schmidt (2000) survey the experimental and theoretical literature on reciprocity, and Charness and Rabin (2002) attempt to identify experimentally the separate roles of reciprocity, fairness and inequality aversion. 5. Auriol and Renault (2001) also analyze the impact of status concerns on incentives in a principal agent model. They argue, as we do, that status can replace money as an inducement for effort. The main difference is that by tying status to wages we make status costly to provide, while they allow firms to provide, at no cost, status symbols that are independent of wages. However, in equilibrium, they also get a positive relation between wages and status, because status and wages are complements in the workers’ utility function in their setting. 6. We wish to point out, at the outset, the social concerns that will not be discussed in this paper. We shall not discuss discrimination or xenophobia, where some individuals have low status based on their gender or ethnic origin. This is not because we dismiss such phenomena as being empirically irrelevant, but rather because we find situations in which individuals can influence their social status via some actions more interesting from a theoretical point of view. In this paper, the main instruments by which an individual can affect his or her social status are the choice of firm and the amount of effort exerted on the job. Other important channels such as investment in schooling and choice of occupation have been discussed in our previous work. See Fershtman et al. (1996). 7. A companion paper, Fershtman, Hvide, and Weiss (2003), considers a setting where firms employ only one agent, interpreted as the CEO. This agent is risk averse and compares his wage to that of other workers in the same industry (interpreted as other CEOs). That paper does not consider issues central to the present paper such as heterogeneity in preferences and sorting of workers. 8. The assumption of observable preferences is less restrictive than it appears. First, types may be correlated with gender, ethnic background, or age. If types are not (partially) observable in such a manner, they may be revealed through the workers signaling or by firms screening. Section 4 considers such possibilities in more detail. 9. Linear contracts will also be suffice to implement first best in the case where the noise term is identical to zero. 10. This mechanism differs from Frank (1984a), who obtains mixing because of inherent differences in productivity.
Cultural Diversity, Status Concerns and the Organization of Work
383
11. A border line case is when the cost of effort is vðeÞ ¼ blnða eÞ where a42b40: For this specification, the marginal cost is convex and effort cannot exceed a. For this case, yhet ¼ a
b þa 1þb
b
1 þ 2b ¼ 2a 1þb
2b
which is independent of b, implying that yhet ¼ yhom : Any convex transformation of this function yields that yhet oyhom : 12. The modification of this assumption is straightforward. 13. The work-place is a national medical facility that specializes in blood collection, processing and testing. About half of the 150 workers were immigrants from the former Soviet Union and of these, 12 Israeli born and 13 immigrants from the former Soviet Union were interviewed. 14. When a sample of Russian immigrants were asked, two years after arrival, how often they meet Israelis that are not relatives, about 64% responded ‘‘never or rarely’’, 20% responded ‘‘occasionally’’ (Damian & Rosenbaum, 2004a,b). 15. In fact, immigrants work more than Israelis mainly when placed in occupation below their qualifications. According to the Israel 1995 census, male immigrants with an academic degree from the former Soviet Union worked about half an hour less per week than comparable Israelis when they had an academic job in Israel, but worked about 2 hours more than comparable Israelis if they found a low-skill job. This suggests that the low-skill jobs serve as stepping stones for a future advance to better jobs. 16. For instance, Lipset (1993) argues that despite many similarities, Canada and the US have marked cultural differences with ensuing significant differences in welfare policy, unionization and entrepreneurship. Devroye and Freeman (2001) find that surprisingly little of the greater dispersion of wages in the US compared to (a sample of) European countries can be attributed to differences in skill distributions. This suggests a possible role for cultural and institutional differences. 17. See, for instance, Glazer (1997) for a discussion of the dilemma between the melting pot and multi-cultural approaches in schools. 18. However, too much diversity can eliminate trade altogether (see Chichilnisky, 1994). 19. See Alesina, Baqir and Easterly (1999) and Alesina et al. (2002).
ACKNOWLEDGMENTS We thank Anat Admati, Gary Becker, Edward Lazear, Canice Prendegast, Sherwin Rosen, and several seminar audiences for their helpful comments.
REFERENCES Akerlof, G. (1982). Labor contracts as partial gift exchange. Quarterly Journal of Economics, 114, 1243–1284.
384
CHAIM FERSHTMAN ET AL.
Akerlof, G. A., & Yellen, J. L. (1990). The fair wage-effort hypothesis and unemployment. Quarterly Journal of Economics, 105, 255–283. Alesina, A., Baqir, R., & Easterly, W. (1999). Ethnic goods and ethnic divisions. Quarterly Journal of Economics, 97, 543–569. Alesina, A., Devleescauwer, A., Easterly, W., Kurlat, S., & Wacziarg, R. (2002). Fractionalization. Journal of Economic Growth, 8, 155–194. Ashenfelter, O., Harmon, C., & Oosterbeek, H. (1999). A review of estimates of the schooling/ earnings relationship, with tests for publication bias. Labour Economics, 6, 453–470. Auriol, E., & Renault, R. (2001). The costs and benefits of symbolic differentiation in the work place. IDEI working paper 101/00. Blundell, R., & MaCurdy, T. (1999). Labor supply: A review of alternative approaches. In: O. Ashenfelter & D. Card (Eds), Handbook of labor economics, Vol. 3b. Amsterdam: NorthHolland. Card, D. (2001). Immigrant inflows, native outflows and the local labor market impacts of higher immigration. Journal of Labor Economics, 19, 22–64. Charness, G., & Rabin, M. (2002). Understanding social preferences with simple tests. Quarterly Journal of Economics, 117, 817–869. Chichilnisky, G. (1994). Social diversity, arbitrage, and gains from trade: A unified perspective on resource allocation. American Economic Review, 84, 427–434. Clark, A. E., & Oswald, A. J. (1996). Satisfaction and comparison income. Journal of Public Economics, 61, 359–381. Damian, N., & Rosenbaum-Tamari, Y. (2004a). Immigrants from the former Soviet Union: Immigration motives and commitment to life in Israel. Israel Ministry of Immigration Special Report no 1. (in Hebrew). Damian, N., & Rosenbaum-Tamari, Y. (2004b). Immigrants from the former Soviet Union: The first two years in Israel. Israel Ministry of Immigration Special Report no 15. (in Hebrew). Devroye, D., & Freeman, R. (2001). Does inequality in skills explain inequality of earnings across advanced countries? NBER Working Paper 8140. Dustman, C., & Preston, I. (2002). Racial and economic factors in attitudes to immigration. Mimeo: University College London. Eckstein, Z., & Weiss, Y. (2002). The integration of immigrants from the former Soviet Union in the Israeli labor market. In: A. Ben-Basat (Ed.), Structural changes in the Israeli Economy, a special volume in memory of Michael Bruno (pp. 349–377). Cambridge: MIT Press. Eckstein, Z., & Weiss, Y. (2004). On the wage growth of immigrants: Israel, 1990–2000. Journal of the European Economic Association, 2, 665–695. Ely, R., & Thomas, D. (2001). Cultural diversity at work: The effects of diversity perspectives on work group processes and outcomes. Administrative Science Quarterly, 46, 229–273. Fearon, J. (2003). Ethnic and Cultural diversity by country. Journal of Economic Growth, 8, 195–222. Fehr, E., & Schmidt, K. (2000). Theories of fairness and reciprocity. Evidence and economic applications. In: M. Dewatripont, L. P. Hansen & S. Turnovsky (Eds), Advances in economic theory. Eight World Congress of the Econometric Society. Cambridge: Cambridge University Press. (forthcoming). Fershtman, C., Hvide, H. K., & Weiss, Y. (2003). A behavioral explanation of the relative performance puzzle. Annales d’Economie et Statistique, 71–72, 349–361.
Cultural Diversity, Status Concerns and the Organization of Work
385
Fershtman, C., Murphy, K., & Weiss, Y. (1996). Education, status and growth. Journal of Political Economy, 104, 108–132. Fershtman, C., & Weiss, Y. (1993). Social status, culture and economic performance. Economic Journal, 103, 946–959. Frank, R. (1984a). Interdependent preferences and the competitive wage structure. Rand Journal of Economics, 15, 510–520. Frank, R. (1984b). Are workers paid their marginal products? American Economic Review, 74, 549–570. Frank, R. (1985). Choosing the right pond. Oxford: Oxford University Press. Friedberg, R. (2001). The impact of mass migration on the Israeli labor market. Quarterly Journal of Economics, 116, 1373–1408. Gibbons, R., & Waldman, M. (1999). Careers in organizations: Theory and evidence. In: O. Ashenfelter & D. Card (Eds), Handbook of labor economics, Vol. 3. Amsterdam: NorthHolland. Glazer, N. (1997). We are all multiculturalists now. Cambridge, MA: Harvard University Press. Hicks, J. R. (1963). The theory of wages (2nd ed.). London: MacMillan. Horowitz, T. (2005). The integration of immigrants from the former Soviet Union. Israel Affairs, 11, 117–136. Kandel, E., & Lazear, E. (1992). Peer pressure and partnerships. Journal of Political Economy, 100, 801–813. Lazear, E. (1989). Pay inequality and industrial politics. Journal of Political Economy, 97, 561– 580. Lazear, E. (1999a). Personnel economics: Past lessons and future directions. Journal of Labor Economics, 17, 199–236. Lazear, E. (1999b). Language and culture. Journal of Political Economy, 107, S95–S126. Lipset, S. (1993). Culture and economic behavior: A commentary. Journal of Labor Economics, 11(2), s330–s347. Prendergast, C. (1999). The provision of incentives in firms. Journal of Economic Literature, 37, 7–63. Reder, M. (1957). Labor in a growing economy. New York: Wiley. Remennick, L. (2004). Work relations between immigrants and old-timers in an Israeli organization: Social interactions and inter-group attitudes. International Journal of Comparative Sociology, 45, 45–71. Rotemberg, J. (1994). Human relations in the workplace. Journal of Political Economy, 102, 684–717. Rothschild, M., & Stiglitz, J. (1976). Equilibrium in competitive insurance markets: An essay on the economics of imperfect information. Quarterly Journal of Economics, 90, 629–650. Sobel, J. (2005). Interdependent preferences and reciprocity. Journal of Economic Literature, 43, 392–496. Spence, M. (1973). Job Market signaling. Quarterly Journal of Economics, 87, 355–374. Sussman, Z., & Zakai, D. (1998). The mass immigration of physicians and the steep rise in wages of Veterans in Israel: A paradox? Economic Quarterly, 45, 28–63. Thomas, D., & Ely, R. (1996). Making differences matter: A new paradigm of managing diversity. Harvard Business Review, 74, 74–90. Tsui, A. S., Egan, T. D., & O’Reilly, C. A. (1992). Being different: Relational demography and organizational attachment. Administrative Science Quarterly, 37, 549–579.
386
CHAIM FERSHTMAN ET AL.
Weiss, Y. (2000). High skill immigration: Some lessons from Israel. Swedish Economic Policy Review, 7, 127–156. Weiss, Y., & Fershtman, C. (1998). Social status and economic performance: A survey. European Economic Review, 42, 801–820. Zizzo, D., & Oswald, A. (2001). Are people willing to pay to reduce others’ incomes? Annales d’Economie et de Statistique, 63,64, 39–65.
APPENDIX A. PROOFS OF PROPOSITIONS ON MIXING We now consider mixing in general under various status concerns. The only restrictions are that utility is linear in wages and that wages are linear in output, so that the conditions for transferable utility hold. Specifically, u~ i ¼ w~ i þ bi ðw~ i
w~ j Þ þ ai ðw~ i
w ¯ iÞ
vðei Þ
(A.1)
where w~ i is the worker’s own wage, w~ j the wage of another worker in the same firm and w~ i the reference wage outside the firm, and w~ i ¼ si þ ai y~ i
(A.2)
There is no loss of generality in assuming that the wage of each worker depends only on his own output. We allow for productivity differences and a general cost of effort function v(ei) that is increasing and strictly convex. A competitive firm that consists of two workers offers contracts (si, ai) that maximize the Lagrangian L ¼ EðpÞ þ l1 ½Eðu1 Þ
r1 þ l2 ½Eðu2 Þ
r2
(A.3)
taking into account that workers choose effort level that maximize their utility, where r1 and r1 are the reservation values of the two workers,. The workers’ choices of effort depend only on the incentive parameters, a1, a2, and are independent of the fixed payments s1 and s2. Differentiating the Lagrangian with respect to s1 and s2, we have 1 þ li ð1 þ bi þ ai Þ
bj lj ¼ 0;
i ¼ 1; 2; jai
(A.4)
Substituting these values back into the Lagrangian, we see that the incentive parameters a1 an a2 and the associated levels of effort e1 and e2 must maximize the total surplus of the firm and two workers S ¼ t1 e1 þ e2 t2
l1 ðvðe1 Þ þ r1 þ a1 w~ 1 Þ
l2 ðvðe2 Þ þ r2 þ a2 w¯ 2 Þ
(A.5)
Thus, the first best levels of effort are given by li v0 ðei Þ ¼ ti
(A.6)
Cultural Diversity, Status Concerns and the Organization of Work
To induce these levels of effort the firm sets ti ti ¼ ai ¼ li ð1 þ bi þ ai Þ 1 þ lj bj
387
(A.7)
Thus, incentives are set to correct for the externalities imposed on other workers in the firm. For homogenous firms, we have that l1 ¼ l2 ¼ ð1 þ ai Þ
(A.8)
We can now calculate the levels of effort, the wage that each worker receives in a homogenous firm and define the reservation values ri for joining a heterogeneous firm to be the expected utility that each worker can receive in a homogenous firm. Let S* be the maximized value of S for a heterogeneous firm, then mixing occurs if S 40; segregation occurs if S o0 and there is indeterminacy if S ¼ 0: Another way to state this condition is to define the joint ‘‘pie’’ W as the weighted sum of the expected payoffs of the players, with a weight of 1 for the firm and weights l1 and l2 for two types of workers and require that it exceeds the weighted sum of the reservation values, which means that the reservation point is within the utility frontier. Because v(ei) is strictly convex, S is strictly concave in e1 and e2 and the maximum is unique. The choices of homogenous firms are also available to the heterogeneous firm. Therefore, a sufficient condition for mixing is that SX0; when evaluated at the levels of effort chosen by the homogenous firms. We shall focus here on the two preference profiles discussed in the text: (1) The type 1 worker cares only about his co-worker, b1 ¼ b; a1 ¼ 0; and the type 2 worker does not care about status at all b2 ¼ 0; a2 ¼ 0: (2) The type 1 worker cares only about his co-worker, b1 ¼ b; a1 ¼ 0; and the type 2 worker cares only about the wages of workers outside the firm b2 ¼ 0; a2 ¼ b: For these examples, we assume that vðeÞ ¼ 12e2 : But the method of analysis is applicable to other specifications. A.1. Type 1 Cares about Local Status, Type 2 does not Care about Status Suppose that, as in Section 3, that type 1 workers care about their local status in the firm, i.e., b1 ¼ b; a1 ¼ 0; and that type 2 workers do not care about status at all, i.e., b2 ¼ 0; a2 ¼ 0: Thus, 1 þ l1 ð1 þ bÞ ¼ 0 1 þ l2
bl1 ¼ 0
so that l1 ¼ 1=1 þ b and l2 ¼ 1 þ 2b=1 þ b
ðA:9Þ
388
CHAIM FERSHTMAN ET AL.
Workers in homogenous firms exert effort ti. Evaluating S at this point, we get a sufficient condition for mixing S ¼ t21 ð1 l1 Þ þ t22 ð1 l2 Þ b ðt2 t22 Þ40 ¼ 1þb 1
ðA:10Þ
Hence, if the workers who care about status are equal or more productive, mixing must occur. The necessary and sufficient condition for mixing is that S X0: The effort levels chosen by the heterogeneous firm are e1 ¼ ð1 þ bÞt1 ; e2 ¼
ð1 þ bÞt2 ð1 þ 2bÞ
(A.11)
Substituting these value into S, we obtain that S X0 if t21 ð2 þ 3bÞ t22 ð2 þ bÞð1 þ 2bÞ
(A.12)
which, for a positive b, can be satisfied even if the workers who care about status are less productive. A.1.1. Calculation of Wages The expected wage bill of a heterogeneous firm is given by Eðw1 Þ þ Eðw2 Þ ¼ t1 e1 þ t2 e2 ¼ ð1 þ bÞt21 þ
ð1 þ bÞt22 ð1 þ 2bÞ
(A.13)
The equilibrium division of this wage between the two workers depends on which type is in the majority. Case 1: Workers who care about status are the minority in the population ðxo12Þ: Since xo0:5; a type 2 worker, who is the majority type, must get his reservation utility, which is the utility that he would obtain in a homogenous firm. Thus, at equilibrium, E(w2) is given by 1 ð1 þ bÞt2 2 1 2 Eðw2 Þ (A.14) ¼ t2 2 ð1 þ 2bÞ 2
389
Cultural Diversity, Status Concerns and the Organization of Work
We can now solve for E(w2) and subtract it from the total wage bill to obtain the expected wage of type 1 worker in heterogeneous firms. Eðw1 Þ ¼ ð1 þ
bÞt21
1 ð1 þ bÞt2 2 2 ð1 þ 2bÞ
ð1 þ bÞt22 þ ð1 þ 2bÞ b2 t22 2ð1 þ 2bÞ2
¼ ð1 þ bÞt21
1 2 t 2 2 ðA:15Þ
Note that, for a positive b, then Eðw1 Þ4Eðw2 Þ and the wage gap rises with b. Case 2: Workers who care about status are the majority in the population ðx412Þ: We follow the same procedure as in the previous case. Because type 1 workers are the majority, there must be some homogenous firms that employ only type 1 workers. Thus, if there are heterogeneous firms in equilibrium, the type 1 workers in those firms obtain their reservation utility, which is their expected utility in a homogenous firm. Therefore, Eðw1 Þð1 þ bÞ
bEðw2 Þ
1 1 ½ð1 þ bÞt1 2 ¼ t21 2 2
(A.16)
Using this indifference condition and the zero profits condition, we obtain Eðw1 Þ ¼ t21
1 þ 2b þ 32 b2 bð1 þ bÞ þ t22 ð1 þ 2bÞ ð1 þ 2bÞ2
(A.17)
b þ 12 b2 ð1 þ bÞ2 þ t22 ð1 þ 2bÞ ð1 þ 2bÞ2
(A.18)
Eðw2 Þ ¼ t21
Again for a positive b, then Eðw1 Þ4Eðw2 Þ and the wage gap rises with b.
A.2. Type 1 Cares about Local Status, Type 2 Cares about Wages Outside the Firm Suppose now that, as in section 4, that type 1 workers have local status preferences so that b1 ¼ b; a1 ¼ 0; and that type 2 workers care only about the wages of workers outside the firm, b2 ¼ 0; a2 ¼ b: We follow the same type of analysis as in the previous section and obtain the following results. In homogenous firms2 with two workers of type 1, l1 ¼ l2 ¼ 1; imt plying e1 ¼ t1 and r1 ¼ 21 : In a homogenous firm with two type 2 workers,
390
CHAIM FERSHTMAN ET AL.
l1 ¼ l2 ¼ 1 þ b; implying e2 ¼ t2 ð1 þ bÞ and r2 ¼ t22 ð1 þ bÞ
t22 ð1 þ bÞ2 =2 þ bðt22 ð1 þ bÞ
w¯ 2 ÞÞ ¼
t22 ð1 þ bÞ2 2
bw¯ 2 (A.19)
For heterogeneous firm, we substitute the assumed ðai ; bi Þ into (A.4), which yields 1 þ l1 ð1 þ bÞ ¼ 0 1 þ l2 ð1 þ bÞ bl1 ¼ 0
ðA:20Þ
so that l1 ¼ 1=1 þ b and l2 ¼ ð1 þ 2bÞ=ð1 þ bÞ2 ; implying that e1 ¼ t1 ð1 þ bÞ and e2 ¼ t2 ð1 þ bÞ2 =ð1 þ 2bÞ: Evaluating S at this point, we get 2 2 1 t1 ð1 þ bÞ2 t21 2 2 ð1 þ bÞ þ S ¼ t1 ð1 þ bÞ þ t2 1þb 2 1 þ 2b 2 ! 2 2 2 1 þ 2b 1 t2 ð1 þ bÞ ðt2 ð1 þ bÞÞ þ ðA:21Þ 2 2 1 þ 2b 2 ð1 þ bÞ If the two types have the same productivity t1 ¼ t2 ; then the condition that S 40 is equivalent to the requirement b
ð1 þ bÞ2 þ
ð1 þ bÞ3 ð1 4bÞ 0 ð1 þ 2bÞ2
(A.22)
But this inequality cannot be satisfied for b40; which implies that there is no equilibrium with heterogeneous firms. However, for t1 4t2 a mixed equilibrium can exist. Evaluating S at the effort levels chosen by the homogenous firms e1 ¼ t1 and e2 ¼ t2 ð1 þ bÞ we get S ¼ t21 ð1
l1 Þ þ t22 ð1 þ bÞ
l2 t22 ð1 þ bÞ2 ¼
b t2 1þb 1
bt22
(A.23)
Thus, the sufficient condition for mixing holds if t21 ð1 þ bÞ4t22 : A.2.1. Calculation of Wages If the two types are equally productive, i.e., t1 ¼ t2 ¼ 1; and there are no heterogeneous firms then the wages of the two types are simply Eðw1 Þ ¼ 1 and Eðw2 Þ ¼ 1 þ b: That is, the worker who cares about global status works harder and obtains higher wages. But the expected utility of the person who cares about local status, which equals 12 exceeds the expected utility of the person who cares about global status, 1 þ b ð1 þ bÞ2 =2: This discrepancy
Cultural Diversity, Status Concerns and the Organization of Work
391
illustrates the inefficiency of the status competition when firms do not serve as coordinators that internalize the wage externalities. Assume now that the sufficient condition for mixing holds, i.e., t21 4t22 ð1 þ bÞ: Then the output of and wage bill of a mixed firm is Eðw1 Þ þ Eðw2 Þ ¼ t1 e1 þ t2 e2 ¼ ð1 þ bÞt21 þ
ð1 þ bÞ2 t22 ð1 þ 2bÞ
(A.24)
The equilibrium division of this wage between the two workers depends on which type is in the majority. Case 1: Workers who care about local status are the minority in the population ðxo12Þ: Since xo0.5, a type 2 worker, who is the majority type, must get his reservation utility, which is the utility that he would obtain in a homogenous firm. Thus, at equilibrium, E(w2) is determined by Eðw2 Þ
2 1 ðt2 ð1 þ bÞ2 2 1 þ 2b
bw ¯2 ¼
t22 ð1 þ bÞ2 2
bw ¯2
(A.25)
yielding Eðw2 Þ ¼
2 t22 ð1 þ bÞ2 1 ðt2 ð1 þ bÞ2 Þ þ 2 1 þ 2b 2
(A.26)
We can now subtract E(w2) from the expected wage bill to obtain the expected wage of the type 1 worker Eðw1 Þ ¼ ð1 þ bÞt21 þ
ð1 þ bÞ2 t22 ð1 þ 2bÞ
t22 ð1 þ bÞ2 2
2 1 t2 ð1 þ bÞ2 2 1 þ 2b
(A.27)
Under the assumption that, t21 4t22 ð1 þ bÞ; it is can be verified that Eðw1 Þ4Eðw2 Þ: Case 2: Workers who care about status are the majority in the population ðx412Þ: We follow the same procedure as in the previous case. Because type 1 workers are the majority, there must be some homogenous firms that employ only type 1 workers. Thus, if there are heterogeneous firms in equilibrium, the type 1 workers in those firms obtain their reservation utility, which is their expected utility in a homogenous firm. Therefore, Eðw1 Þð1 þ bÞ
bEðw2 Þ
1 1 ½ð1 þ bÞt1 2 ¼ t21 2 2
(A.28)
392
CHAIM FERSHTMAN ET AL.
Using this indifference condition and the zero profits condition, we obtain h i ð1þbÞ2 t2 b ð1 þ bÞt21 þ ð1þ2bÞ2 þ ½12t21 þ 12½ð1 þ bÞt1 2 (A.29) Eðw1 Þ ¼ 1 þ 2b Under the assumption that, t21 4t22 ð1 þ bÞ; it is now possible that Eðw2 Þ4Eðw1 Þ; reflecting the fact that the type 2 workers are now scarce. We can now close the model and determine the reference wage w¯ 2 for the type 2 worker. Because the wages in homogenous and heterogeneous firms where seen to be independent of w ¯ 2 ; we can choose any combination of these wages as a reference. Let us assume, for instance that each type 2 worker cares only about the average wage of his own type. Then if x40:5 and all type 2 workers work in heterogeneous firms w¯ 2 ¼ Eðw2 Þ: However, if xo0:5 and some type 2 workers work in homogenous firms then w ¯ 2 ¼ 1 x xEðw2 Þ þ 2 2 ð1 2xÞ=1 xt2 ð1 þ bÞ: In this case, Eðw2 Þot2 ð1 þ bÞ; because the type 2 workers in heterogeneous firms are induced to reduce their effort in exchange for lower wages. Therefore, an increase in x reduces the average wage of the type 2 workers, w ¯ 2:
APPENDIX B. PROOF OF PROPOSITION ON SCHOOLING We shall now present the bounds on the costs of schooling such that a separating equilibrium exists. We start out with the case xo0:5 and then consider the case x40:5: Case 1 (xo0:5): If a type 1 agent joins a homogenous firm employing two workers with no schooling (and where the other worker is of type 2) his pay schedule will be whom ¼ y: In such a case, he exerts the effort 1 þ b: His total 2 wage will be 1 þ b: His co-worker, who is of type 2, will have the same incentives but his choice of effort is 1 and his expected wage is 1. The local status of the type 1 worker will be b2 ; implying expected utility of 12ð1 þ bÞ2 b: However, the same worker may join a heterogeneous firm as an uneducated worker accepting the wage schedule whet 2
¼
1 ð1 þ bÞ 2 1 1 þ b y þ þ 2 ð1 þ 2bÞ 2 1 þ 2b
(B.1)
Cultural Diversity, Status Concerns and the Organization of Work
393
Given such a contract, the agent exerts the effort wage ð1 þ bÞ2 =ð1 þ 2bÞ: This effort level yields the total wage 1 ð1 þ bÞ 2 1 1 þ b ð1 þ bÞ2 1 þ 2b þ b2 =2 (B.2) ¼ þ þ 2 ð1 þ 2bÞ 2 1 þ 2b 1 þ 2b ð1 þ 2bÞ In such a firm, his co-worker will be an educated type 1 worker with productivity t1, and who chooses the effort level ð1 þ bÞ t1 and gets the total wage b2 2ð1 þ 2bÞ2
E mi ðw1 Þ ¼ ð1 þ bÞt21 The local status in such a case will be
1 þ 2b þ b2 =2 ð1 b ð1 þ 2bÞ ¼b
1 þ 4b þ 5b2 =b3 ð1 þ 2bÞ2
bÞt21
(B.3)
b2 þ 2ð1 þ 2bÞ2
bð1 þ bÞt21
ðB:4Þ
The agent’s expected utility in this case is 1 þ 2b þ b2 =2 1 þ 4b þ 5b2 þ b3 þb ð1 þ 2bÞ ð1 þ 2bÞ2
bð1 þ bÞt21
2 1 ð1 þ bÞ2 2 ð1 þ 2bÞ
(B.5)
Comparing the two options yields that as long as t1 41; 1 ð1 þ bÞ2 2
b4
1 þ 6b þ 11b2 þ 8b3 þ b4 2ð1 þ 2bÞ2
bð1 þ bÞt21
(B.6)
which implies that if a player type 1 chooses not to get education, he will be better off joining a homogenous firm. Intuitively, if he joins a heterogeneous firm, the agent gets lower incentives for effort and a negative local status. Although uneducated workers are paid a positive fixed amount to join heterogeneous firms, this amount is not sufficient to reverse the other effects, as seen in (A.6). Thus, the necessary condition for separating equilibrium is that a type 1 agent is better off from educating and joining a heterogeneous firm, rather than not educating and joining a homogenous firm. Specifically, 1 2b þ 3b2 ð1 þ bÞ2 t21 þ 2 2ð1 þ 2bÞ
1 x4 ð1 þ bÞ2 2
b
(B.7)
394
CHAIM FERSHTMAN ET AL.
Rearranging this condition yield the following necessary condition: ð1 þ bÞ2 t21
2x4
1
2b2 þ 2b3 ð1 þ 2bÞ
(B.8)
Now suppose that an agent of type 2 deviates and acquires schooling, then he will be considered a member of the type 1 minority and work in a heterogeneous firm as an educated worker getting the wage schedule w ¼ 2 y b =2ð1 þ 2bÞ2 : In such a case, this an effort level of 2 agent will2choose
t1, implying expected utility of 12t21 x: Thus, a necessary b =2ð1 þ 2bÞ condition for a separating equilibrium is that agent type 2 is better off not getting education. That is, 1 1 2 4 t 2 2 1
b2 2ð1 þ 2bÞ2
x
(B.9)
b2 ð1 þ 2bÞ2
(B.10)
Rearranging yields the following condition: 2x4ðt21
1Þ
Putting the two conditions (A.8) and (A.10) together yields that a separating equilibrium exists only when ð1 þ bÞ2 t21
1
2b2 þ 2b3 42x4t21 ð1 þ 2bÞ
1 þ 4b þ 5b2 ð1 þ 2bÞ2
(B.11)
It is easy to verify that the l.h.s. of (A.11) is greater than the r.h.s. as long as t1 41: Thus, the above equation define a range for x for which there exists a separating equilibrium. Case 2 (x40:5): In this case, type 2 agents are the minority. Consider now the separating equilibrium in which there are heterogeneous and homogenous firms that employ two workers of type 1. The homogenous firms offer the employment contract whom given by (27) for educated workers and 1 the heterogeneous firms offer the employment contract whet 1 given by eqs. (28) and (30) for educated workers and the contract whet given by (31) and 2 (33) for uneducated workers. In such a case, type 2 workers do not get education and accept the contract from a heterogeneous firm and type 1 workers acquire education and accept one of the two employment contracts
395
Cultural Diversity, Status Concerns and the Organization of Work
offered to them. Their utilities will be, 1 Eðu1 Þ ¼ t21 2
x;
Eðu2 Þ ¼ t21
b þ 12b2 ð1 þ bÞ2 þ ð1 þ 2bÞ 2ð1 þ 2bÞ2
(B.12)
Suppose that an agent of type 1 deviates and skips education. In that case, he can get a position as a type 2 worker in a heterogeneous firm, where his co-worker will be an educated type 1 worker. The compensation scheme that he will get in such a case is 2 whet 2 ¼ t1
2b þ b2 1þb y þ 2ð1 þ 2bÞ 1 þ 2b
(B.13)
Given such a contract, the agent exerts the effort ð1 þ bÞ2 =ð1 þ 2bÞ: This
level yields the total wage t21 2b þ b2 =2ð1 þ 2bÞ þ effort
1 þ b2 =ð1 þ 2bÞ2 : His type 1 co-worker, who got education and has the productivity t1, chooses the effort level of ð1 þ bÞt1 and gets a total wage of E mj ðw1 Þ ¼ t21
1 þ 2b þ 32b2 bð1 þ bÞ þ ð1 þ 2bÞ ð1 þ 2bÞ2
The local status in such a case will be ( 3 2 2b þ b2 ð1 bÞ3 2 1 þ 2b þ 2b b t21 þ t 1 ð1 þ 2bÞ 2ð1 þ 2bÞ ð1 þ 2bÞ2 ¼b
ð1 þ bÞð1 þ b þ b2 Þ ð1 þ 2bÞ2
b
(B.14)
bð1 bÞ ð1 þ 2bÞ2
)
1 þ b þ b2 2 t ð1 þ 2bÞ 1
ðB:15Þ
The agent’s expected utility will be in this case t21
2b þ b2 ð1 bÞ3 ð1 þ bÞð1 þ b þ b2 Þ þ þ b 2ð1 þ 2bÞ ð1 þ 2bÞ2 ð1 þ 2bÞ2 2 1 ð1 þ bÞ2 ð1 þ bÞð1 þ 3b þ b2 þ b3 Þ ¼ 2 ð1 þ 2bÞ 2ð1 þ 2bÞ2
b
1 þ b þ b2 2 t ð1 þ 2bÞ 1
b2 2 t 2 1
ðB:16Þ
Thus, a necessary condition for a separating equilibrium is that player of type 1 is better off not deviating. That is, ð1 þ bÞð1 þ 3b þ b2 þ b3 Þ 2ð1 þ 2bÞ2
b2 2 1 2 t o t 2 1 2 1
x
(B.17)
396
CHAIM FERSHTMAN ET AL.
which after simplification gives the condition 2xoð1 þ b2 Þt21
ð1 þ bÞð1 þ 3b þ b2 þ b3 Þ ð1 þ 2bÞ2
(B.18)
Now suppose that an agent of type 2 deviates and chooses to acquire schooling. In such a case, he will be considered as a type 1 worker and will be able to choose to work in a homogenous firm or in heterogeneous firm. hom ¼ Working in 2a homogenous
firm, he will get the payment schedule of w b=ð1 þ b Þ t þ y= ð 1 þ b Þ : Given these compensation, his choice of effort
1 is b=ð1 þ bÞt21 þ 2 e ¼ 1=ð21 þ bÞ t1 ; which yields the utility of x On the other hand, working in a heterogeneous firm as t1 =2ð1 þ bÞ an educated worker,
the agent will get the wage schedule w ¼ y t21 2b þ b2 =2ð1 þ 2bÞ þ bð1 þ bÞ=ð1 þ 2bÞ2 : In such a case, the agent will choose an effort expected utility of 12t21 t21
level of t1, implying
2 2 2b þ b =2ð1 þ 2bÞ þ bð1 þ bÞ=ð1 þ 2bÞ x: Comparing the expressions it yields that if player 2 is deviating he is better off working for a heterogeneous firm, provided that b is not too large. Thus, a necessary condition for a separating equilibrium is that agent type 2 is better off not getting education. That is, t21
2b þ b2 ð1 þ bÞ2 1 þ 4 t21 2 2 2ð1 þ 2bÞ 2ð1 þ 2bÞ
t21
2b þ b2 bð1 þ bÞ þ 2ð1 þ 2bÞ ð1 þ 2bÞ2
x
(B.19)
which after rearranging yields the condition 2x4t21
1
2b 2b2 ð1 þ 2bÞ
1 b2 ð1 þ 2bÞ2
(B.20)
Putting the two conditions (A.17 and 19) together yields that a separating equilibrium exists only when ð1 þ b2 Þt21
ð1
bÞð1 þ 3b þ b2 þ b3 Þ 2b 2b2 21 42x4t 1 ð1 þ 2bÞ ð1 þ 2bÞ2
1 b2 ð1 þ 2bÞ2 (B.21)
It is easy to verify that the l.h.s. of (A.20) is greater than the r.h.s. as long as . Thus the above equation defines a range for x for which there is a separating equilibrium. It is also readily seen that the lower bound of the range is higher when xo0:5: The upper bound will be higher when xo0:5 only if b is not too high, that is if 8 þ 16b 3b2 40:
ETHNIC DIVERSITY, MARKET STRUCTURE AND RISK SHARING IN DEVELOPING COUNTRIES Mohamed Jellal and Yves Zenou ABSTRACT The paper addresses mainly three questions. One, do workers tend to be employed by employers of the same ethnic group; two, what is the structure of the equilibrium wage contract; and three, do more ethnically homogeneous labor markets tend to have different labor contracts than more ethnically diversified ones. The answer to the first question is in the affirmative – in equilibrium all employers offer the same wage contract and workers are hired by employers of the closest ethnic affiliation. In terms of the equilibrium wage contract, its nature depends on the attitude toward risk of both sides of the market. Finally, the answer to the third question is also in the affirmative since the more homogenous the labor market, the more deterministic is the wage.
Those who study modern Africa commonly highlight three features: its poverty, its instability, and its ethnic diversity. y scholars reason that Africa is poor because it is unstable and that its instability derives from its ethnic complexity. Ethnicity thus lies, it is held, at the root of Africa’s development crisis. Robert H. Bates (2000)
Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 397–426 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24012-8
397
398
MOHAMED JELLAL AND YVES ZENOU
1. INTRODUCTION In recent years, there has been an increasing interest in the economic consequences of ethnic diversity, especially in terms of economic development and growth (Barro, 1991; Mauro, 1995; Easterly & Levine, 1997; Montalvo & Reynal-Querol, 2005). These studies performed at the country level suggest that ethnically diverse societies have slower economic growth and are more prone to corruption and political instability than ethnically homogenous societies as a result of political conflict across ethnic groups. As Barr and Oduro (2002) put it, ‘‘surely the time has come to place the economics of ethnicity on the agenda for policy debate’’. However, to the best of our knowledge, few researches have investigated the economic consequences of ethnicity at a more microeconomic level. Miguel and Gugerty (2005) examine the impact of ethnic diversity on local public provision (i.e. local funding of primary schools and community water wells) in sub-Saharan Africa (Western Kenya). They show that ethnic diversity is negatively related to local public goods provision in this rural African setting because social sanctions are imposed more effectively within ethnic groups than between groups. In another paper (Barr & Oduro, 2002), the consequences of ethnic fractionalization (i.e. the segmentation of a population into several groups that are distinct in terms of language and/or culture) on Ghana’s labor market is analyzed. Their main finding is to show that workers who are related to their employers earn a wage premium. The aim of the present paper is to follow this line of research by analyzing the impact of ethnic diversity on labor contracts in developing countries. The standard literature on agricultural tenancy explains the existence of different contracts, such as sharecropping, fixed-rental and fixed-wage contracts, in rural areas, without taking into account the role of ethnic diversity. In general, to account for these different contracts, three major explanations have been offered: (i) trade-off between risk sharing and transaction costs (uncertainty, risk and moral hazard problems); (ii) screening workers of different abilities (adverse selection problems); and (iii) market imperfections for inputs besides land (see in particular Binswanger & Rosenzweig, 1982; McIntosh, 1984; Eswaran & Kotwal, 1985). All of these approaches have been well developed both theoretically and empirically (see in particular Stiglitz, 1974; Cheung, 1969; McIntosh, 1984; Newberry & Stiglitz, 1979; Shaban, 1987), even though the third approach has received less attention in the literature. To the best of our knowledge, the link between ethnic diversity and wage contracts has been neglected and the aim of this paper is to provide a simple
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
399
model that sheds some light on these aspects. It is indeed well documented that, in most countries, individuals tend to work with employers of the same ethnic origin. In developed countries, foreigners and recent migrants, like for example the Cubans in Florida or the Chinese and Mexican in California, form closed-knit societies and work together in cities (see e.g. Borjas, 1999). The same is true in the U.K. for Indians, Bangladeshi or Pakistanis (see e.g. Modood et al., 1997). In developing countries, the ethnic origin is even more crucial to understand the way labor markets work both in urban and rural areas (see e.g. Assaad, 1997; Wahba & Zenou, 2005, for Egypt; Sadoulet, de Janvry, & Fukui , 1997, for Philippines; van de Walle & Gunewardena, 2001, for Viet Nam; Foster & Rosenzweig, 2001, for India and Pakistan; Barr & Oduro, 2002, and Udry & Conley, 2004, for Ghana; Krishnan & Sciubba, 2004, for Ethiopia). To be more precise, we consider a population of workers (that could be, for example, laborers) and employers (that could be, for example, landlords) who belong to different ethnic groups, so that working together implies a cost (because of language, religion and/or cultural differences). Part of the production is random (because, for example, of climate change) and not observable (ex ante) by both employers and employees. Our analysis encompasses both rural and urban labor markets. We evacuate moral hazard as well as adverse selection problems by focusing on closely knit communities. Indeed, the ignorance on the part of landlords about tenants’ abilities is quite inappropriate for most rural communities (this is already discussed in Bardhan, 1984; Eswaran & Kotwal, 1985). This is also true for urban labor markets (see e.g. Assaad, 1997; Wahba & Zenou, 2005). Most papers show that elements of altruism among kin reduce the conflict of interest between two partners and create relations of trust and confidence in which cheating is less likely to occur. For example, Pandey (2004) documents the fact that working with in-kins prevents workers to shirk (moral hazard) because of reputation effects and peer group (or family) pressures.1 Contrary to the ‘standard’ approach with moral hazard where the focus is on the tension between the interests of the workers (laborers) and the firm (landlord) in an environment where (like here) output is not perfectly observable, in our model, the ‘tension’ is between employers of different ethnic origins. Indeed, because workers and employers belong to different ethnic groups, there is a cost for workers to work for an employer with a different ethnic background. This implies that employers have market power over workers with similar ethnic origin, so that they play a Nash game with the other employers to determine the optimal wage contract. In particular, given
400
MOHAMED JELLAL AND YVES ZENOU
the competition in the labor market and the volatility of output, profitmaximizing employers set a wage in order to attract enough workers and to reduce the risk associated with the output’s uncertainty. It is not surprising that most of our results will depend on the degree of competition in the labor market (as measured by the number of landlords, the ethnic cost y) and on the degree of risk aversion of both workers and firms. In particular, if employers are very risk averse, they will be very sensitive to large variations in output and will transfer as much as risk as they can onto laborers. So our main question is how to share the risk of a random production that affects both employers’ profits and workers’ utility in a framework where all agents are ethnically differentiated and where employers imperfectly compete with each other to attract workers. Our results are the following. First, we obtain that, in equilibrium, employers tend to hire workers of similar ethnic background and employers’ co-ethnics earn more than other workers. Second, in terms of the equilibrium wage contract, its nature depends on the attitude toward risk of both sides of the market. If both employers and workers are risk averse, each side would like to shift as much risk as possible to the other side. The fact that workers bear disutility from working with ethnically different employers gives the latter market power in proportion to the average workers disutility. The fewer employers there are, the greater the cost of a worker from switching to a more distant employer, hence the more risk employers is able to shift to workers. If workers are risk neutral and employers are risk averse, workers bear all the risk in every case, and the reverse if employers are risk neutral. Finally, the more ethnically homogeneous is a labor market, the less the wage depends on random shocks.
2. THE MODEL We develop a theoretical model in which the interaction in the labor market (i.e. wage setting) between employers and workers of different ethnic origins is explicitly taken into account. In the literature, the measurement of ethnic diversity has been a very difficult task. There are six distinct characteristics of an individual that matter for ethnolinguistic classification: two of them (race and color) are inherited whereas two (culture and language) are learned. The fifth characteristic (the ethnic origin) is more difficult to define and refers to the main name by which people are known. Finally, the sixth component (nationality), in contrast to the other characteristics, can be changed. To summarize
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
401
these different aspects, two types of synthetic indices have been proposed: indices of fractionalization and indices of polarization. First, the most famous and widely used is the index of ethnolinguistic fractionalization (ELF), constructed by Taylor and Hudson (1972), which is defined as: ELF ¼ 1
N X
c2i
(1)
i¼1
where if we consider religious (or ethnic) diversity, ci is the proportion of people who professes religion i (or belongs to ethnic group i). Basically, this indicator can be interpreted as measuring the probability that two randomly selected individuals in a country will belong to different ethnolinguistic groups. As a result, the index ELF increases when the number of groups increases. Second, the polarization index has been proposed by Esteban and Ray (1994) and can be written as follows: PO ¼ k
N X X
c1þg cj jxi i
xj j
(2)
i¼1 jai
where the cs are the sizes of each group in proportion to the total population, the term jxi xj j measures the ‘distance’ between two ethnic (or religious) groups, i and j, and g and k are parameters. These two indices: ethnic (or religious) fractionalization and polarization take values between 0 and 1. The higher the indices, the more the society is ethnically diverse. Montalvo and Reynal-Querol (2005) provide a table where they give the ethnic and religious fractionalization and polarization indices for most countries in the world. For example, a country like Algeria has very low indices of religious fractionalization and polarization (not surprisingly since most individuals are Muslims), but relatively high indices of ethnic polarization (0.514) and fractionalization (0.299). On the contrary, a country like Bangladesh, has relatively high indices on religion (0.503 and 0.261 respectively), but relatively low indices on ethnicity (0.132 and 0.068 respectively). In the present paper, in order to model ethnicity, we use an approach in terms of ‘distance’ between group. We assume that there is an ethnic ‘distance’ between workers and employers of different ethnic groups and, as a result, a cost t per unit of (ethnic) ‘distance’ is borne when they interact with each other. The most natural interpretation of t is ‘language, religion and culture’. There is indeed a cost to work with individuals of different cultures, religions and different languages since, as stated by Lazear (1999), ‘common culture and common language facilitate trade between individuals’. There
402
MOHAMED JELLAL AND YVES ZENOU
are very strong evidences on this issue (see among others Chiswick, 1978; Chiswick & Miller, 1996; Dustman & Preston, 2001, for the U.S. and the U.K; and Assaad, 1997 or Barr & Ondoro, 2002, for less developed countries) showing that it is indeed costly to work with individuals of different ethnic groups because of language and cultural differences.2 It has to be clear that the way we model ethnicity is extremely simple and do not have all the rich aspects of the measures proposed above (fractionalization or polarization indices). However, we are capturing one dimension of these measures, namely the ‘distance’, which could be in terms of religion, culture, race, etc., between two different groups in a given country. We choose to represent the ethnicity space by the circumference C of a circle of length L (Salop, 1979). On this circle, n employers and a continuum of workers are uniformly distributed along its circumference. This captures the fact that ethnic diversity is pre-determined and that the ethnic distance between a worker and an employer of the same ethnic group is obviously lower than with an employer of a different ethnic group. For simplicity, we assume that employers are equally spaced along the circumference C so that L/n is the ethnic distance between two adjacent employers. Workers reside in different ‘locations’ along the circumference, which implies that they support different ethnic costs to work with different firms. In other words, we segment the population into several groups that are distinct in terms of language, religion and/or culture. Formally, the ethnic cost is given by a linear function tjx yi j of the difference between a worker of ethnicity x 2 C and an employer of ethnicity yi 2 C: To sum-up, in our framework, ethnic diversity is captured by L, the size of the ethnic space, n, the number of ethnic employers and t, the cost of interacting with other communities. Indeed, the higher (lower) L and/or the lower (higher) n, the more ethnically diversified (homogenous) is the labor market. Also, the higher t, the more costly it is to interact with members of other ethnic groups and the more frictional is the labor market. All workers are identically productive and produce q observable units of output. This means that agents are horizontally differentiated, which implies that employers do not come predominantly from one ethnic group (which is the case in most developing countries; see our discussion above and, in particular, Montalvo & Reynal-Querol, 2005). It should be clear that, because of both employers’ and workers’ ethnic diversities, the competition in the labor market is imperfect since employers have local monopsony power over workers of similar ethnic background. This is because it is more costly for a worker of a certain ethnic group to work with an employer of a different ethnic background than with a similar one.
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
403
Firms produce an homogeneous good (which is taken to be the numeraire) sold on a competitive market whose production is random. Indeed, for example, even if all the inputs that a farmer can reasonably control are properly applied, the size of the harvest is still heavily dependent on Nature and will vary. To express this uncertainty, we suppose that the production ~ where q is the observable part of the production and e level is q þ y; y is described by a random variable whose mean is chosen to be 0 (without loss of generality) while its variance is s2. As in Sandmo (1971), greater output uncertainty is measured by an increase in s2: a mean preserving spread in production. In the context of developing countries, the random part of the production is due, for example, to climate changes. In other words, all workers are assumed to produce q, but there is a common shock (uncertainty) captured by e y that is out of control of both the employer and the worker and that affects production. We would now like to define the optimal contract on which both the employer and the worker agree. As discussed in the introduction, moral hazard problems are assumed to be relatively small, so that we have chosen to ignore them. This is admittedly a simplifying assumption but help us to focus on labor heterogeneity and ethnic issues. Therefore, in this paper, we would like to focus on optimal risk sharing and on employers’ choice of method of pay in a framework where both employers and workers are ethnically differentiated. For that, each employer i ¼ 1; . . . n proposes the following revenue (contract) to the worker: R~ i ¼ ai ðy~ þ qÞ þ bi
i ¼ 1; . . . ; n
(3)
with 0 ai 1 and bi w0: It is easy to see from (3) that this contract consists of two elements: a fixed part bi that can be positive, negative or equal to zero, and a variable part, which is tied to the (random) output. In fact, the worker obtains a percentage ai of his/her production and the landlord gets a percentage 1 ai of the worker’s production. The following definition characterizes the different possible contracts. Definition 1. For workers employed by firm i, we have: A fixed-wage (or rent) contract is when the workers’ compensation is independent of what they produce, i.e. ai ¼ 0: A pure piece-rate contract is when workers are only paid according to what they produce, i.e. bi ¼ 0: A full-residual claimant contract is when workers obtain the full benefit of their work, but pay a fixed amount to the employer, i.e. ai ¼ 1 and bi o0:
404
MOHAMED JELLAL AND YVES ZENOU
A share-tenancy contract (sharecropping) is a mix of fixed and piece-rate contracts, i.e. 0oai o1 and bi 40 or bi o0: There are indeed two parts in the compensation: a fixed one, which is independent of production, and a variable one, which is a percentage of production. The most natural application of these contracts in developing countries is in rural labor markets.3 In that case, a share-tenancy contract (sharecropping) typically entitles the supplier of land services to receive from the supplier of labor a pre-arranged proportion of crop output. Sharecropping thus, differs significantly from contracts in which the rent for land or the wages for labor are fixed and do not vary with output (fixed-wage or fixedrent contract); nor it should be confused with various forms of piecework (piece-rate contract), where labor is engaged for a specific purpose, usually harvesting, and rewarded proportionally from the total crop. Finally, as pointed out by Lazear (1995), a typical example of a full-residual claimant contract is the case of taxi drivers. Indeed, the latter rent a car to a taxicab company i ðbi o0Þ and then keep everything that they make for themselves ðai ¼ 1Þ: In agriculture, the laborer can pay a fixed cost to the landlord to exploit the plot of hired land and then keep all the proceeds from the crops. One of the main originality of our framework is to consider not one (as it is usually the case) but a finite number of heterogenous employers (in terms of ethnicity) and a continuum of heterogenous workers (in terms of ethnicity). Because of this double heterogeneity, the competition in the labor market will be imperfect since employers have some monopsony power over workers that are ethnically ‘close’. The other original part of this work is that the outside option of workers is endogenous and depends on the strategies of other employers. Indeed, each employer has to decide the optimal contract by taking into account the strategies of the other firms in the market. Even if some workers are ethnically different from an employer, they may work with him/her if this employer proposes a more advantageous contract. Formally, firms choose simultaneously ai and bi (Nash equilibrium) and therefore workers’s revenues, R~ 1 ; . . . ; R~ i ; . . . ; R~ n ; before the realization of ~ but anticipating the impact of their compensation on workers’ the risk y; labor supply. Thus, given (3), the realized wage of a worker of ethnicity x working for an employer of ethnicity yi is given by: Z~ x;yi ¼ R~ i
tjx
yi j ¼ ai ðy~ þ qÞ þ bi
tjx
xi j
(4)
In this section, we assume that workers are risk averse. In order to obtain closed forms solutions, we further assume that a worker of ethnicity x
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
405
working for an employer of ethnicity yi has a mean-variance utility function given by:4 U x;yi ¼ EðZ~ x;yi Þ
a VarðZ~ x;yi Þ 2
¼ E½ai ðy~ þ qÞ þ bi ¼ Wi
tjx
tjx
yi j
a Var½ai ðy~ þ qÞ þ bi 2
xi j
tjx
yi j ð5Þ
where W i ¼ ai q þ bi a2a2i s2 is the expected utility gross of ethnic costs when working for firm i, E½ the expectation operator, Var½ the variance operator and a 0 the degree of absolute risk aversion.5 Observe that W i is not a random variable since employers commit to wages and employment before output realizations. Once each firm i proposes W i ; each worker chooses to be hired by the employer that gives the highest utility (net of ethnic costs). Since firms anticipate the choice of workers, they hire all workers who want to work at the prevailing expected utilities, ðW 1 ; . . . ; W i ; . . . ; W n Þ; because they know that these workers are ethnically quite similar. The reservation wage is assumed to be the same across workers since they are all identical in terms of productivity. Thus, without loss of generality, the reservation wage is set equal to zero. Given W i 1 and W iþ1 ; firm i’s labor pool is composed of two sub-segments whose outside boundaries are given by marginal workers x and x¯ for whom the net wage is identical between firms i 1 and i, on the one hand, and firms i and i þ 1; on the other. In other words, x is the solution of the equation: Wi
tðyi
xÞ ¼ W i
1
tðx yi 1 Þ
so that x¼
Wi
1
W i þ tðyi þ yi 1 Þ 2t
(6)
In this case, firm i attracts workers whose locations belong to the interval ½x; xi because the expected utility net of ethnic costs they obtain from firm i is higher than the one they would obtain from firm i 1: Clearly, workers belonging to the interval ½xi 1 ; x are hired by firm i 1: In a similar way, we have: x¯ ¼
Wi
W iþ1 þ tðyi þ yiþ1 Þ 2t
(7)
406
MOHAMED JELLAL AND YVES ZENOU
Consequently, firm i’s labor pool is defined by the interval ½x; x: ¯ In this context, firm i’s realized profits can thus be written as: Z x ~i ¼ ðy~ þ q Ri Þdx P x
¼ ðy~ þ q
Ri Þðx¯
xÞ ¼ ½ð1
ai Þðy~ þ qÞ
bi ðx¯
xÞ
In this section, we assume that landlords are risk averse. Here also, in order to obtain closed forms solutions, we further assume that firms have a meanvariance utility function given by:6 r ~ iÞ VarðP 2
~ iÞ V i ¼ EðP
(8)
~ i is defined where r 0 is the degree of absolute risk aversion and where P above. Hence, we can rewrite (8) as follows: V i ¼ ½ð1
ai Þq
bi ðx¯
xÞ
r ð1 2
ai Þ2 ðx¯
x Þ2 s 2
(9)
Since employers and workers are all assumed to be risk averse, the problem here is how to share the risk in the context of imperfect competition and ethnic diversity. Firms play a Nash game to determine ai and bi : We will see that distinct types of compensations will emerge depending on the values of the different parameters.
3. LABOR MARKET EQUILIBRIUM We can now derive our first result. We assume that all workers take a job in equilibrium. In this context, the outer boundaries of firm i’s labor pool are given by (6) and (7). Firm i chooses ai and bi that maximize its utility (9). We have the following result.7 Proposition 1. (Existence and uniqueness). If q43
s2 rLa 4tL þ rL þ an n
(10)
holds, there exists a unique symmetric Nash equilibrium given by: an ¼
rL rL þ an
(11)
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
407
rL tL ð1 an Þ2 q2 s2 an Þq n n 2 an s rLa tL q ¼ rL þ an rL þ an n
bn ¼ ð1
and, before ethnic costs, all workers obtain the same following positive utility: Wn ¼ q
s2 rLaðrL=2 þ anÞ ðrL þ anÞ2
tL n
whereas the equilibrium employers’ profit is equal to: 2 L r an n V ¼ tþ s2 n 2 rL þ an
(12)
(13)
Proof. See the Appendix. First, we obtain that, in equilibrium, employers tend to hire workers of similar ethnic background (in equilibrium, the maximum ‘distance’ to hire someone is L/2n) and employers’ co-ethnics earn more than other workers (indeed, even though remuneration (12) is the same for all workers, because of ethnic costs, the net remuneration decreases with the ethnic ‘distance’ to the employer). Second, this general case corresponds to an ‘impure’ piece rate (sharecropping) contract in which workers have a fixed pay equals to bw0 and a variable one which is a fraction 0oao1 of what they produce (see Definition 1). In our model, the only choice faced by workers consists in deciding which employer they want to work for (this depends on both their ethnic distance to that employer and the compensation offered). Given this choice, each employer i chooses ai and bi that maximize his/her profit by taking as given the choice of the as and bs of the other employers in the economy. Each employer also takes into account the impact of his/her compensation policy on his/her ‘natural’ workers (i.e. those who are enough close to the employer’s ethnic group). In this respect, we have:8 @W ¼q @a
aas2 40
and @W ¼ 140 @b
408
MOHAMED JELLAL AND YVES ZENOU
Indeed, when firms increase the variable part a or the fixed part b of the salary, their labor supply increases. It is interesting to observe from @W =@a that the reaction of workers negatively depends on both s2 ; the variance of the production, and a the workers’ degree of risk aversion.9 In particular, if workers are very risk averse, their utility will not increase very much following a rise in a. Third, one can verify from (11) that an increase in r; the degree of risk aversion of employers, and/or a decrease in a, the degree of risk aversion of workers, raises an : Concerning a, this is quite natural since more risk-averse workers prefer to see a reduction in a; the uncertain part of their salary. Concerning r; the decision to increase an depends on the competition in the labor market. If it is very fierce, because for example employers are not very risk averse (low r), then employers reduce an to attract more workers. The effects of r and a on b; the fixed part of the pay, are more complex, and will be analyzed in more details in the next section. We can also analyze the effect of ethnic diversification on a and b: We have the following result. Proposition 2. (Ethnic diversity). (i) The more ethnically homogenous a labor market (i.e. higher L or lower n), the lower the level of piece rate an but the higher the fixed part of the remuneration bn : In other words, the more ethnically homogeneous a labor market, the less the wage depends on random shocks. (ii) The ethnic cost t has not impact on an but has a negative effect on bn : In other words, the more ‘frictions’ or ‘conflicts’ that exist between different ethnic groups, the lower the deterministic part of the wage. Thus, in ethnically homogenous countries, where employers and employees are relatively homogenous (if L/n is quite small, then no worker will be ethnically very ‘far away’ from an employer), most of the wage is independent of the random shock. In other words, the more homogenous the society, the more firms take all the risk associated with random shocks and the more the wage is deterministic. As mentioned above, L captures the degree of ethnic diversity in the economy. Indeed, when L increases, the ethnic space is bigger, and thus workers of a certain ethnic group are even more attached to employers of similar ethnic group and more distant to employers of other groups. This implies that local employers have a higher monopsony power over workers of similar ethnic background. The variable n captures the number of employers in the economy. So, when n increases, workers are more likely to find employers of similar ethnic background, which implies that employers have less monopsony power.
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
409
Another interesting result is about t. The latter measures in some sense the cost of interacting with people of different ethnic groups. This means that we can view t as a measure of the ‘tension’ or the ‘conflict’ that exists between different ethnic groups. Result (ii) indicates that, for a given level of ethnic diversity, the more intense the conflicts in a labor market, the lower the deterministic part of the wage. Indeed, when t increases (for example the cost of learning a language is high or cultural differences are large), employers have higher monopsony power on workers of similar ethnic background since it becomes too costly for these workers to work for other employers with more diverse ethnic origin. So, when t rises, employers can decrease the fixed part of the wage. They cannot however affect the piece rate since the latter is tied to output only. Observe finally that the equilibrium wage (12), or more precisely the equilibrium utility before ethnic costs, is always below q the marginal productivity of workers. Indeed, because firms have market power, they tend to exploit workers by setting wages below their marginal productivity. The following result confirms this intuition. Proposition 3. (Perfect competition). When the number of firms becomes arbitrarily large, then lim an ðnÞ ¼ 0
n!1
and
lim bn ðnÞ ¼ q
n!1
and the equilibrium wage tends to its competitive level ðW n ¼ qÞ while profits tend to zero. This result shows that the competitive model of the labor market is indeed the limit of the spatial model. Once again a key element of our analysis is the interaction between firms to attract workers. So when n ! þ1; firms have no more market power since each worker works for an employer belonging exactly to the same ethnic group (ethnic costs are equal to zero). As a result, competition pushes the wages to workers’ marginal product.
4. FIRMS’ CHOICE OF METHOD OF PAY10 We have obtained a general result. We would now like to see under which condition firms set different types of compensations. We start with the following result. Proposition 4. (Salop, 1979). There exists a unique Nash equilibrium in which W n ¼ q tL=n and V n ¼ tL2 =n2 if
410
MOHAMED JELLAL AND YVES ZENOU
(i) either employers are risk-neutral ðr ¼ 0Þ and workers are risk-averse ða40Þ: In this case, employers set a fixed wage such that an ¼ 0 and bn ¼ q tL=n: (ii) or workers are risk-neutral ða ¼ 0Þ and employers are risk averse ðr40Þ: In this case, employers set a piece rate such that an ¼ 1 and bn ¼ tL=n: The results of Proposition 4 are obviously a particular case of Proposition 1 when either firms or workers are risk-neutral. This is the standard model of Salop (1979). Interestingly, the results strongly depend on the competition in the labor market and, therefore, on the degree of ethnic diversity of both employers and workers. Indeed, as stated above, ethnic diversity is measured here by t, L and n. So when the worker’ ethnic cost t or the degree of ethnic differentiation L increases or when the number of employers n decreases, then, workers’ utility decreases whereas employers’ utility increases. Moreover, the wage contract set by employers also hinges on ethnic diversity. Indeed, even though the two cases (i) and (ii) leads to the same utility level W n for workers and the same profit level V n for employers, the wage contract is quite different. In case (i) where employers are risk-neutral and workers are not, it is optimal for employers to set a fixed wage that do not depend on the production q of workers so that an ¼ 0 and bn 40: However, the fixed part bn decreases with ethnic diversity so that when workers and employers are more ethnically differentiated (t and L high, and n low), bn decreases because workers are more isolated from employers of different ethnic background and closer to employers of similar ethnic origin. On the contrary, when employers are risk averse and workers risk-neutral, case (ii), employers set a piece-rate contract in which an ¼ 1 and bn o0: In other words, workers are exactly paid according to what they produce ðq þ e yÞ minus a fixed part b: In this case, employers care about random production (since it affects their profit) but workers do not. Therefore, they can set a pure piece-rate system and still attract workers. Here also, bn negatively varies with ethnic diversity. Corollary 1. (No ethnic cost). When workers are risk-neutral ða ¼ 0Þ; employers are risk-averse ðr40Þ and the ethnic cost t is equal to zero, then employers set a pure piece rate such that an ¼ 1 and bn ¼ 0: The equilibrium utility and profit levels are respectively given by W n ¼ q and V n ¼ 0: This corollary reinforces our previous result on piece-rate contracts. It says that, if workers do not bear any ethnic cost to work with employer of different ethnic groups, then it is optimal for employers to set a pure piece-rate
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
411
contract in which workers are paid according to what they produce. In this case, all risk-neutral workers obtain the same maximum level of utility q, whereas all risk-averse employers get the lowest profit level 0. This is because t is a measure of employers’ market power, and thus of the degree of competition in the labor market since higher t implies more market power, and thus less intense competition. So when the ethnic cost t is equal to zero, the competition between employers to attract workers becomes fiercer and riskaverse employers do not obtain anymore the fixed compensation of their workers. Observe that this case does not imply that labor is not differentiated since L40: In order to have no heterogeneity at all in this model, one must assume that L ¼ 0: Then, without any other hypothesis, it is easy to verify that an ¼ 0 and bn ¼ q; so that W n ¼ q and V n o0: Let us now assume that employers and workers are both risk averse but have exactly the same degree of risk aversion, i.e., r ¼ a40: In this case, employers will optimally set an impure piece-rate contract to workers. Proposition 5. (Same degree of risk aversion). When both workers and employers have the same degree of risk aversion, r ¼ a40; there exists a unique Nash equilibrium given by: an ¼ n q b ¼ Lþn
L Lþn rLs2 Lþn
tL n
s2 rLðL=2 þ nÞ Lþn
tL n
n
Before ethnic costs, workers obtain: Wn ¼ q
and employers’ profits are equal to: 2 " 2 # L r n V¼ tþ s2 n 2 Lþn In this case, employers find it optimal to set an impure piece-rate contract where workers receive a fixed part bn and a variable part an : This is because employers and workers are both risk averse and thus must share the risk of uncertain production. It is thus obvious that an or bn can never be equal to zero because both parties want to avoid the randomness of production.
412
MOHAMED JELLAL AND YVES ZENOU
It is interesting to observe that, in this case, an only depends on the degree of ethnic diversity in the economy, i.e. L and n. When n the number of employers increases, then an is reduced whereas bn is augmented if n is sufficiently large.11 Indeed, more employers or less ethnic diversity implies fiercer competition since workers and employers become more and more ethnically similar, and thus employers tend to increase the fixed part and decrease the variable part of wages. On the contrary, when workers are more differentiated (L increases) and thus workers are more ethnically isolated from other ethnic groups, an increases whereas bn decreases if n is sufficiently large.12 Observe also that reducing t increases the fixed part of the wage bn : As above, this is because competition between employers become fiercer since each employer has less market power over their (local) workers. An interesting question is what happens to the model when r ¼ a ¼ 0: It is then a model without uncertainty; the utility function of a worker working for employer i is now given by W i ¼ ai q þ bi and the profit function of employer i is equal to V i ¼ ½ð1 ai Þq bi ðx¯ xÞ ¼ ðq W i Þðx¯ xÞ; where x and x¯ are still given by (6) and (7). By solving the symmetric Nash equilibrium, it is easy to verify that we obtain the following relation: ð1
aÞq
b ¼ tL=n
(14)
which implies that Wn ¼ q
tL=n
and
V n ¼ tL2 =n2
(15)
We have the following result. Proposition 6. (Risk neutrality). When both workers and employers are risk-neutral ða ¼ r ¼ 0Þ; then any pay system can emerge. However, workers’ utilities and employers’ profits are always given by (15) and they depend on the degree of isolation of workers from other communities. Indeed, there are as many values of a and b that can satisfy equation (14), given that a and b are negatively correlated. Therefore, any wage system could be implemented and each of them will always lead to (15). For example, if a ¼ 0; then b ¼ q tL=n and a pure fixed-wage emerges. If b ¼ 0; then a ¼ 1 sL=ðnqÞo1; employers set a pure piece-rate pay. If a ¼ 1=2; then b ¼ q=2 tL=n; then an impure piece-rate contract prevails. Finally, if a ¼ 1; b ¼ tL=n; we have a full-residual claimant contract in which workers receive all the benefits of their production, but pay back some money to the employer.
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
413
5. COEXISTENCE OF FIXED-WAGE AND PIECE-RATE CONTRACTS IN A REGIONAL CONTEXT Our framework can easily be extended to account for the coexistence of fixed-wage and piece-rate contracts within and between regions, a widely observed feature of labor markets in developing countries (see for example, Bardhan & Rudra, 1981; Dre`ze & Mukherjee, 1989; Baland, Dre`ze, & Leruth, 1999). Indeed, so far, we have assumed that all employers located in a region (i.e. equally spaced along the circumference of the circle) have exactly the same level of risk aversion r: Assume now that there are two types of employers with risk aversion, r1 and r2 ; and that all workers are still characterized by a. Assume also that r1 oaor2 and that, along the circumference of the circle, the location of firms alternate from a firm of type r1 to a firm of type r2 (such that there exactly n1 firms of type r1 and n2 firms of type r2 ; with n1 þ n2 ¼ n). It then easy to see that, in equilibrium, if t is sufficiently large (closed-knit societies), both fixed-wage and piece-rate contracts will coexist. Employers of type r1 set fixed-wage contracts whereas employers of type r2 set piece-rate contracts. Interestingly, some workers who have quite similar ethnic background (for example, the ones on the right of x and the ones on the left of x) will obtain different contracts. However, the general result here is that workers who are ethnically similar (i.e. belonging to the ‘natural catchment area’ of each firm; for example, workers working for firm i and residing within an ethnical ‘distance’ ½x; x ¯ from landlord i) obtain the same type of contract. Another interesting result, already stressed above, is the importance of the degree of ethnicity of the region. If the region is very diverse (i.e. large L), then differences between fixed wages and piece rates will increase. Finally, this model also show that different areas offer different wage contracts. Indeed, if some regions are characterized by employers with a high degree of risk aversion (for example, regions with small firms) and others by employers with a low degree of risk aversion (for example, regions with large firms), then this model enables us to explain the existence of different wage contracts across regions. Our results can be compared to that of Baland et al. (1999). Using a very elegant model, in which individual effort is explicitly taken into account, they show that very able laborers as well as low-ability laborers work on piece rates because they can choose their own effort level (the optimal number of tasks performed under a piece-rate contract increases with the worker’s ability). Laborers of middle ability will then be paid using
414
MOHAMED JELLAL AND YVES ZENOU
fixed-wage contracts. They also show that a monopsonistic employer always finds profitable to hire laborers under both types of contracts.
6. EMPIRICAL RELEVANCE OF THE RESULTS Let us summarize our results and see if they are empirically relevant. The first set of results (Proposition 1) indicates that, in equilibrium, all employers offer the same wage contract, workers are hired by employers of the closest ethnic affiliation and employers’ coethnics earn more than other workers. Sadoulet et al. (1994) have studied two rice-growing farmer villages in rural Thailand. In village N, located approximately 100 km northwest of Bangkok, characterized by high risk and widespread poverty, they show that, of the share of tenancy contracts, 41% are among relatives. In village Bo, located 20 km east of Chian Mai City, characterized by low risk because of non-farm activities, they show that, of the share tenancy contracts, 83% are between relatives, and these contracts are generally repeated over a long period of time. Sadoulet et al. (1994) also observe that there are gift exchanges in this village since landlords make after-harvest gifts of grains to their tenants in exchange for hard work. In another paper, Sadoulet et al. (1997) analyze a household survey that they conduct in three villages of the Philippines in 1992. They show that most sharecroppers have a kinship relationship with their landlords and that kinship ties with the landlord are a key determinant of cooperative behavior by sharecroppers and hence a key determinant of efficiency. They show that the terms of the contract affect negatively the input decisions of non-kin sharecroppers but not those of kin sharecroppers and the latter use inputs at levels similar to those of owner-operators and fixed-rent contracts. Inspection of Table 4 in Sadoulet et al. (1997) indicates that non-kin sharecropper households are somewhat less well off than the other categories. On average, they have less land assets, they own less machinery, a smaller percentage of them have off-farm income, and their off-farm income is substantially lower. Barr and Oduro (2002) focus on a labor market in the Ghanaian manufacturing sector. They use data from the fifth wave of the Ghanaian Manufacturing Enterprise Survey, where the sample of enterprises is drawn from four cities in southern Ghana. Ghana’s index of ELF (which formula is given by (1)) is 0.71, placing it close to the average for sub-Saharan Africa of 0.65, thus indicating that the population is quite segmented into several groups that are distinct in terms of language and/or culture. In a labor
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
415
market where both employers and employees come from different ethnic groups (see Tables 2 and 3 in Barr and Oduro, 2002), they show that the ethnic distributions of employers and employees are very similar. They run the following equation ln wi ¼ f0 þ f1 ri þ f2 ci þ f3 ei þ f4 pi þ f2 f i þ zi where ln wi is the log of earnings for worker i, f0 a constant term, ri a dummy variable that takes value of 1 is the worker is related to the employer, ci a dummy variable that takes value of 1 is the worker is from the same ethnic group as the employer, ei a vector of dummies, one corresponding to each ethnic group represented in the sample, pi the vector of personal characteristics of worker i, f i are employers’ fixed effects (fixed across workers not time) and zi the error term. The joint significance of f3 tells us whether there is variation across the ethnic groups, the signs and significance of specific elements of f3 tell us whether particular groups earn significantly more or less than the group chosen as a basis for comparison, and the signs and significance of differences between the elements of f3 provide us with similar information about other pairwise comparisons. Significant and positive coefficients on ri and ci indicate that relatives and coethnics, respectively, receive positive earning premium relative to other workers. They find that 11% of workers are employed by a relative and a further 23% are employed by a non-related member of the same ethnic group. More generally, worker from every ethnic group are more likely to work for a member of their own ethnic group than for a member of another Ghanaian ethnic group. Furthermore, they show that being related to the employer is associated with a 23% earnings premium. The second set of our results is also related to ethnicity. We have shown in Proposition 2 that, given that sharecropping exists, the more ethnically homogenous the labor market, the lower an (the part of the remuneration that depends of random production) and the higher bn (the independent part of the remuneration). Also, the higher the ethnic cost t (more costly to interact between different ethnic groups), the lower bn : These results are consistent with the empirical results of Sadoulet et al. (1997). Indeed, in Philippines, they show that sharetenants who have family ties with their employer are not influenced by the terms of the contract (that is a and b), while the other sharecroppers (who are further away from their employer) respond negatively to a lower output share (i.e. lower a). Wilson (1998) offers even stronger evidence of our results by providing an empirical analysis of labor transaction in the Lake Victoria13 fishing
416
MOHAMED JELLAL AND YVES ZENOU
industry. In this labor market, remuneration of both crew and owners (i.e. entrepreneurs who has rights to the boat) usually take the form of distributing shares of the catch. In our model, this would correspond to sharecropping with an 40 and bn o0: Share systems on the western part of the lake are mainly based on dividing each day’s catch among the non-gearowning crew and the owners of boats and gears. A typical example would be the boat and gear owner getting 60% of the proceeds from the catch after it is sold, and the crew getting 40% after deducing for the crew’s food, the nets, for the boat maintenance and fuel. In our model, this last part is bn o0; while the first two parts are 1 an and an (or at least a function of them), respectively. Three observations are in order. First, the fishing business is very volatile, and thus risky since the output (i.e. the catch) strongly depends on weather conditions. Risk bearing can thus explain the persistence of systems of catch sharing in fisheries rather than wages and salaries. Second, social identities (mainly kin groups, ethnic groups and local religious congregations) play a prominent role in the Lake Victoria fishing industry. Finally, even though this study is about fishing industry and not about agriculture, it is very related to our model since most workers live in rural areas, and thus the findings apply to rural labor markets. The analysis is of data gathered between March 1993 and November 1994 on nine randomly selected fish-landing sites, called throughout beaches, on the Tanzanian shore of Lake Victoria. The sample unit is the fishing boat. There results are from 102 fishing units that were currently or recently fishing. Wilson (1998) takes as the dependent variable the percentage of the catch that a non-gear-owning fisher received as payment for his labor; this corresponds to our an : Controlling for different personal characteristics, Wilson (1998) obtains the two following results (see his Table II). When a homogenous crew shares an ethnic identity with the boat owner, this has a significant negative effect on an : When a homogenous crew confronts a boat owner from a different ethnic group, this has a significant positive effect on an : These two results are exactly testing Proposition 2 since they say that the more (less) ethnically homogenous is the labor market, the lower (higher) is an the random part of the remuneration. Even if there is no evidence on bn ; it seems quite natural to assume that crewmembers sharing a common ethnic identity with the boat owner will pay lower fees (such as nets, boat maintenance and fuel) than other crew members. Finally, the last set of our results (Propositions 4, 5–6) can be summarized in Table 1, which indicates which type of labor contract emerges depending on the values of a and r.
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
Table 1.
417
Emerging Labor Contract Depending on the Values of a and r. a
r
40 ¼0
40
¼0
Sharecropping Fixed rent contract
Full-residual claimat contact Any contract
a, Degree of absolute risk aversion of workers r, Degree of absolute risk aversion of firms
Much of empirical literature on contract choice, especially in developing countries, has tried to measure workers’ and employers’ risk aversion. The usual proxies are wealth and properties. The common problem in this literature is the necessity of controlling adequately for unobserved heterogeneity (Ackerberg & Botticini, 2002; Chiappori & Salanie´, 2003). If it is not done properly, then the combination of unobserved heterogeneity and of endogenous matching of agents to contracts is bound to create selection biases. To illustrate this point, let us focus on the recent contribution of Ackerberg and Botticini (2002). They consider the choice between sharecropping and fixed rent contracts in a tenant–landlord relationship. They regress the type of contract (fixed-rent contract or sharecropping) on crop riskiness and tenant’s wealth. If wealth is taken to be a proxy for risk aversion, we would expect that richer (and presumably less risk-averse) tenants are more likely to be under a rental contract. However, wealth is only an imperfect proxy for risk-aversion, and thus the unobserved component of risk-aversion is likely to be correlated with crop riskiness. There is thus an endogeneity problem since we only observe the match between agents (endogenous matching). There are many stories that suggest matching between heterogenous employers and workers. One of them is employers with the some ethnic origin might end up matching with workers of the same ethnic origin. To remedy this endogenous matching problem, Ackerberg and Botticini instrument the crop riskiness variable, using geographical variables as instruments. Using a data set on agricultural contracts between landlords and tenants in early Renaissance Tuscany, Ackerberg and Botticini (2002) address the issue of heterogeneity across agents and find some role for risk in the choice between share contract and fixed rent contract. After correcting, for endogenous matching between landlords and tenants, they find that wealthier tenants (thus those who have lower risk aversion) are more likely to be in
418
MOHAMED JELLAL AND YVES ZENOU
fixed rent contracts. This finding suggests that risk sharing is an important determinant of contract choice and is consistent with the predictions of our model, as described in Table 1. Another paper that found results consistent with our predictions is that of Laffont and Matoussi (1995). Using Tunisian data, they show that tenants with less working capital (thus more risk averse) tend to work under sharecropping. Using a unique multi-country (Cameroon, Ghana, Kenya and Zimbabwe) panel data set for manufacturing firms in Africa, Bigsten et al. (2003) provide a test for the risk sharing model. Indeed, industrial firms in Africa are exposed to very high risks, reflecting demand shocks, price volatility, unreliable infrastructure and poor contract enforcement (Fafchamps, 1996; Collier & Gunning, 1999). At the same time in most African economies financial markets, in particular for insurance, are poorly developed. This conjunction of high risk and weak financial markets suggests that if risk sharing through (implicit) labor contracts is to be found anywhere it is in African manufacturing. Bigsten et al. (2003) find strong evidence for risk sharing and labor imperfections (when a firm experienced a negative shock, it is costly for workers to move to another firm). It is only where such costs exist that risk sharing contracts can be enforced. Risk sharing is only observed for non-production workers (production workers cover people with few skills, thus more mobile group).
7. CONCLUDING REMARKS Though the model used in this paper may seem quite stylized, we believe that it captures some basic features of optimal compensations in rural/urban labor markets in less developed countries in the context of ethnically differentiated workers and employers. It shows the role of market structure, ethnic diversity and market competition as well as of the degree of risk aversion of workers and employers in the determination of firms’ choice of methods of pay. In an uncertain production environment, our model shows that, in equilibrium, firms tend to hire workers of similar ethnic background, and employers’ co-ethnics earn more than other workers. It also shows that large language and cultural differences lead to lower wages because only workers and employers of similar ethnic origin can work together. More generally, we believe that the ethnic origin of both employers and employees is of paramount importance to understand the working of labor
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
419
markets in less developed countries and should therefore be investigated further.
NOTES 1. Sadoulet et al. (1997) also observe that kinship networks are a natural cause of cooperation and thus remove the moral hazard problem faced by landlords. To quote them: ‘‘Kinship networks are important to reduce moral hazards and provide a commitment device when intertemporal resource transfers are involved’’. 2. Co-ethnicity is defined with respect to shared ethnic identity. 3. As mentioned above, our analysis can be applied to both urban and rural labor markets. 4. It is easy to see that the case of risk neutrality is a special case of our meanvariance utility function when a ¼ 0: We will study this issue in the next section. ~ ¼ 0 and Var½y ~ ¼ s2 : 5. To derive (5), one must use our initial hypotheses: E½y 6. Again, it is easy to see that the case of risk neutrality is a special case of our mean-variance utility function when r ¼ 0: We will study this issue in the next section. 7. Since at the symmetric Nash equilibrium all firms pay the same wage, we have skipped the index i. 8. Observe that @W n =@a40 by using (A.5) in the Appendix. 9. Indeed, it is easy to check that:@2 W n =@a@s2 o0 and @2 W n =@a@ao0 10. Since all propositions in this section are a special case of Proposition 1, the existence and uniqueness of equilibrium are always guaranteed by condition (10), which is written using the parameters of each special case. 11. It is indeed easy to check that Lon is a sufficient condition to ensure that @bn =@n40: 12. Again, it is easy to verify that Lon is a sufficient condition to ensure that @bn =@Lo0: 13. Lake Victoria, shared by Kenya, Uganda and Tanzania, is the second largest freshwater lake. 14. Since Salop deals with the product market and us with the labor market, what we called here, the labor supply corresponds to the product demand in Salop’s model. Similarly, wages correspond to prices, monopsony to monopoly, etc y . 15. We skip the index i since we are at a symmetric equilibrium.
ACKNOWLEDGMENTS We would like to thank Hillel Rapoport as well as two anonymous referees for their helpful comments. Yves Zenou thanks the Marianne and Marcus Wallenberg Foundation for financial support.
420
MOHAMED JELLAL AND YVES ZENOU
REFERENCES Ackerberg, D., & Botticini, M. (2002). Endogenous matching and the empirical determinants of contract form. Journal of Political Economy, 110, 564–591. Assaad, R. (1997). Kinship ties, social networks and segmented labor markets: Evidence from the construction sector in Egypt. Journal of Development Economics, 52, 1–30. Baland, J.-M., Dre`ze, J. P., & Leruth, L. (1999). Daily wages and piece rates in agrarian economies. Journal of Development Economics, 59, 445–461. Bardhan, P. K. (1984). Land, labor and rural poverty: Essays in development economics. New York: Columbia University Press. Bardhan, P. K., & Rudra, A. (1981). Terms and conditions of labour contracts in agriculture: Results of a survey of West Bengal 1979. Oxford Bulletin of Economics and Statistics, 43, 89–111. Barr, A., & Oduro, A. (2002). Ethnic fractionalization in an African labour market. Journal of Development Economics, 68, 355–379. Barro, R. (1991). Economic growth in a growth section of countries. Quarterly Journal of Economics, 106, 407–443. Bates, R. H. (2000). Ethnicity and development in Africa: A reappraisal. American Economic Review, Papers and Proceedings, 90, 131–134. Bigsten, A., Collier, P., Dercon, S., Fafchamps, M., Gauthier, B., Gunning, J. W., Oduro, A., Oostendorp, R., Pattillo, C., So¨derbom, M., Teal, F., & Zeufack, A. (2003). Risk sharing in labour markets. World Bank Economic Review, 17, 349–366. Binswanger, H. P., & Rosenzweig, M. R. (1982). Rural labor markets in Asia: Contractual arrangements, employment and wages. New Haven: Yale University Press. Borjas, G. J. (1999). Heaven’s door. Immigration policy and the American economy. Princeton: Princeton University Press. Cheung, S. N. (1969). The theory of share tenancy. Chicago: Chicago University Press. Chiappori, P.-A., & Salanie´, B. (2003). Testing contract theory: A survey of some recent work. In: M. Dewatripont, L. Hansen & P. Turnovsky (Eds), Advances in economics and econometrics: Theory and applications Eighth World Congress, Econometric Society Monographs (pp. 115–149). Cambridge: University Press, Cambridge. Chiswick, B. R. (1978). The effect of Americanization on earnings of foreign-born men. Journal of Political Economy, 86, 897–921. Chiswick, B. R., & Miller, P. W. (1996). Ethnic networks and language proficiency among immigrants. Journal of Population Economics, 9, 19–35. Collier, P., & Gunning, J. W. (1999). Explaining African economic performance. Journal of Economic Literature, 37, 64–111. Dre`ze, J. P., & Mukherjee, A. (1989). Labour markets in rural India: Theories and evidence. In: S. Chakravarty (Ed.), The balance between industry and agriculture in economic development. London: Macmillan Press. Dustman, C., & Preston, I. (2001). Attitudes to ethnic minorities, ethnic context and location decisions. Economic Journal, 111, 353–373. Easterly, W., & Levine, R. (1997). Africa’s growth tragedy. Quarterly Journal of Economics, 112, 1203–1250. Esteban, J., & Ray, D. (1994). On the measurement of polarization. Econometrica, 62, 819–851. Eswaran, M., & Kotwal, A. (1985). A theory of contractual structure in agriculture. American Economic Review, 75, 352–367.
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
421
Fafchamps, M. (1996). The enforcement of commercial contracts in Ghana. World Development, 24, 427–448. Foster, A., & Rosenzweig, M. (2001). Imperfect commitment, altruism, and the family: Evidence from transfer behavior in low-income rural areas. Review of Economics and Statistics, 83, 389–407. Krishnan, P., & Sciubba, E. (2004). Endogenous network formation and informal institutions in village economies. Unpublished manuscript, University of Cambridge. Laffont, J.-J., & Matoussi, M. S. (1995). Moral hazard, financial constraints and sharecropping in El Oulja. Review of Economic Studies, 62, 381–399. Lazear, E. P. (1995). Personnel economics. Cambridge: The MIT Press. Lazear, E. P. (1999). Culture and language. Journal of Political Economy, 107, S95–S126. Mauro, P. (1995). Corruption and growth. Quarterly Journal of Economics, 110, 681–712. McIntosh, J. (1984). An oligopsonistic model of wage determination in agrarian societies. Economic Journal, 94, 569–579. Miguel, E., & Gugerty, M. K. (2005). Ethnic diversity, social sanctions, and public goods in Keyna. Journal of Public Economics, 89, 2325–2368. Modood, T., Berthoud, R., Lakey, J., Nazroo, J., Smith, P., Virdee, S., & Beishon, S. (1997). Ethnic minorities in Britain: Diversity and disadvantage. London: Policy Studies Institute. Montalvo, J. G., & Reynal-Querol, M. (2005). Ethnic diversity and economic development. Journal of Development Economics, 76, 293–323. Newberry, D. M., & Stiglitz, J. E. (1979). Sharecropping, risk sharing and the importance of imperfect information. In: J. A. Roumasset, J. M. Boussard & I. Singh (Eds), Risk, uncertainty and agricultural development. NewYork: Agricultural Development Council. Pandey, P. (2004). Effect of technology on incentive design of share contracts. American Economic Review, 94, 1152–1168. Sadoulet, E., de Janvry, A., & Fukui, S. (1997). The meaning of kinship in sharecropping contracts. American Journal of Agricultural Economics, 79, 394–406. Sadoulet, E., Fukui, S., & de Janvry, A. (1994). Efficient share tenancy contracts under risk: The case of two rice-growing villages in Thailand. Journal of Development Economics, 45, 225–243. Salop, S. C. (1979). Monopolistic competition with outside goods. Bell Journal of Economics, 10, 141–156. Sandmo, A. (1971). On the theory of the competitive firm under price uncertainty. American Economic Review, 61, 65–73. Shaban, R. (1987). Testing between competing models of sharecropping. Journal of Political Economy, 95, 893–920. Stiglitz, J. E. (1974). Incentives and risk sharing in sharecropping. Review of Economic Studies, 41, 219–255. Taylor, C. L., & Hudson, M. C. (1972). World handbook of political and social indicators (2nd ed. ). New Haven, CT: Yale University Press. Udry, C.R., & Conley, T.G. (2004). Social networks in Ghana. Economic growth center. Discussion Paper no. 888. Yale University. van de Walle, D., & Gunewardena, D. (2001). Sources of ethnic inequality in Viet Nam. Journal of Development Economics, 65, 177–207. Wahba, J., & Zenou, Y. (2005). Density, social networks and job search methods: Theory and application to Egypt. Journal of Development Economics, 78, 443–473. Wilson, D. C. (1998). ‘‘Markets, networks, and risk: An analysis of labor remuneration in the Lake Victoria fishing industry’’. Sociological Forum, 13, 425–456.
422
MOHAMED JELLAL AND YVES ZENOU
APPENDIX Proof of Proposition 1. Our model uses the circle model of Salop (1979). So the spirit of our model is quite the same, but the proof of existence and uniqueness is quite different since landlords strategically choose two variables a and b whereas in Salop they chose only one variable, the prices. As it is well-known in the circle model of Salop (1979), the labor supply function14 for each landlord is not differentiable and not continuous. There are in fact three regions: the monopsony region where each landlord attracts only laborers between x and x¯ and some laborers do not work; the competitive region (our case) where all laborers take a job and landlords compete to attract them and finally the supercompetitive region where one landlord, by setting a wage sufficiently high, can attract all laborers of its neighbor landlords. This labor supply is not differentiable everywhere because there is a first kink when one switches from the monopsony region to the competitive region and another one when one switches from the competitive region to the supercompetitive region (see Fig. 1, p. 143, in Salop, 1979). This labor supply is also not continuous because when one switches from the competitive region to the supercompetitive region, a landlord i that attracts the worker located at yiþ1 ; i.e. the location of the landlord i þ 1; attracts at the same time all laborers located between yiþ1 and the marginal worker who is indifferent between landlords i þ 1 and i þ 2 (see Fig. 1, p. 143, in Salop, 1979). Therefore, in order to show the existence and uniqueness of our symmetric equilibrium, we proceed further. We first restrict ourselves to the competitive region (as we did in the text) where the labor supply is continuous and differentiable everywhere (within the competitive region), and show that the profit function V ðÞ is strictly concave so that, within this region, there exists a unique maximum. We then have to check that, at this equilibrium, all laborers take a job. Furthermore, we also have to check that all possible deviations of landlord i from our symmetric equilibrium are not profitable. There are in fact only two possible deviations: one, in the supercompetitive region; two, in the monopsony region. We already know from Salop (1979) that a deviation to the supercompetitive region is never profitable because landlord i has to set a wage higher or equal than the marginal productivity of its laborers, and thus make negative or null profits. However, we have to check that the second deviation, i.e. to the monopsony region, is not profitable for landlord i.
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
423
Let us start with the following result. Lemma A.1. In the competitive region, the profit function V ðÞ is strictly concave in ai and bi : Proof. Remember first that a 2 2 as 2 i
W i ¼ a i q þ bi so @W i ¼q @ai
aai s2
@W i ¼1 @bi
and
In this context, the first order conditions yield:15 @V q aqs2 Va ¼ ½ð1 aÞq b rð1 @a t ½q
@V Vb ¼ @b
ðx¯
aÞs2 ðx¯
rð1
1 xÞ þ ½ð1 t
aÞq
b
aÞ2 s2 ðx¯
xÞðx¯
xÞ ¼ 0
rð1
aÞ2 s2 ðx¯
xÞ ðA:1Þ
xÞ ¼ 0
(A.2)
We have now to show that the Hessian matrix is negative definite, i.e. V aa o0 and V aa V bb V ab V ab 40: We have: @2 V as2 ½ð1 aÞq b V aa ¼ 2 @a t q aas2 2 ½q 2rð1 t
aÞ2 s2 ðx¯
rð1 aÞs2 ðx¯
xÞ
xÞ
q
rs2 ðx¯ x Þ2 2 aas2 ½rð1 aÞ2 s2 t
We want to show that V aa o0: Observe that at the symmetric equilibrium y¯ x¯ ¼ L=n: Therefore, for V aa o0; we will show that ðiÞ : ð1 aÞq b rð1 aÞ2 s2 L=n40; ðiiÞ q aas2 40; and ðiiiÞ q 2rð1 aÞs2 L=n40: (i) First, using (A.2), we have ð1
aÞq
b
aÞ2 s2
rð1
L tL ¼ 40 n n
(ii) Second, using (A.2), we also have:
q
rð1
L ¼ aÞs n 2
1 1
tL bþ a n
(A.3)
424
MOHAMED JELLAL AND YVES ZENOU
Using (10), it is easy to check that in (12), q4s2 rLa=ðrL þ anÞ so that b þ tL=n40; and thus
q
rð1
aÞs2
L ¼ n
1
1
tL bþ 40 a n
(A.4)
Then, by plugging (A.3) into (A.1), it is easy to verify by using (A.4) that:
q
aas2 ¼ q
rð1
(iii) Finally, condition (10) guarantees that q
aÞs2
L 40 n
2rð1
(A.5)
aÞs2 L=n40:
Now, using (i), (ii) and (iii), it is to see that V aa o0: Let us continue our demonstration of the concavity of V ðÞ: We have: @2 V 1 rð1 aÞ2 s2 2 þ V ¼ o0 bb t t @b2 @2 V @2 V V ab ¼ V ba @a@b @b@a 1 q 2rð1 aÞs2 ðx¯ ¼ t
xÞ þ ðq
rð1 aas2 Þ 1 þ
aÞ2 s2 t
After some manipulations and using the fact that, at the symmetric equilibrium, x¯ x ¼ L=n and (A.3), we obtain: rð1 aÞ2 s2 L 2 L V aa V bb V ab V ab ¼ 2 þ s t aþr n n t L 2 2L 2 þ q 2rð1 aÞs 3q 4aas þ 2rð1 aÞs n n 2 2 2rð1 aÞ s L þ ðq aas2 Þ q 2rð1 aÞs2 n t ðq By (10), q
2rð1
aas2 Þ2
aÞs2 L=n40: Moreover, using (A.5), we have: rð1
aÞ
L ¼ aa n
Ethnic Diversity, Market Structure, Risk Sharing in Developing Countries
425
so that 3q
4aas2 þ 2rð1 ¼ 3q
2rð1
aÞs2 L=n
aÞs2 L=n aÞs2 L=n
¼ q þ 2½q
rð1
¼ q þ 2ðq
aas2 Þ40
We have therefore: V aa V bb V ab V ab rð1 aÞ2 s2 L 2 L s t þ q q 2rð1 aþr ¼ 2þ n n t 2 2 2rð1 aÞ s 2 2L þ ðq aas Þ q 2rð1 aÞs n t L þ ðq aas2 Þ q 3rð1 aÞs2 n
aÞs2
L n
Since by using (10), q 2rð1 aÞs2 L=n40 and q 3rð1 aÞs2 L=n40; we have V aa V bb V ab V ab 40: Q. E. D. Because of Lemma 1, and because in the competitive region, the profit function V ðÞ is continuous in ðai 1 ; ai ; aiþ1 Þ and in ðbi 1 ; bi ; biþ1 Þ; we can guarantee that there exists a locally unique symmetric Nash equilibrium in wages. Then, by combining (A.1) and (A.2), and by equalizing the equilibrium a and b; we obtain (at the symmetric Nash equilibrium) the unique (11) and (12). Then, we deduce the equilibrium compensation W n given by (12). Furthermore, using (9), it is easy to obtain (13). We must now check that at the equilibrium candidate (12) all laborers take a job and that this wage is always positive. The equilibrium wage (12) is greater than zero, if: W n 403q4
s2 rLaðrL=2 þ anÞ tL þ n ðrL þ anÞ2
(A.6)
and the condition ensuring that there is full employment at the equilibrium candidate (12) (the worker with the worst match must have a positive utility) is given by: Wn
tL s2 rLaðrL=2 þ anÞ 3tL 403q4 þ 2n n ðrL þ anÞ2
(A.7)
426
MOHAMED JELLAL AND YVES ZENOU
Clearly, (A.7) implies (A.6) so that (A.7) guarantees that in equilibrium all laborers take a job and that the utility after ethnic costs is positive for all laborers. Now, we have to show that it is not optimal for landlord i to deviate from our symmetric equilibrium by setting the monopsony wage. It is easy to verify that the monopsony wage is equal to q/2. So we have to set a condition that rules out this possible deviation. It suffices to show that, at the monopsony wage q/2, the worker located at xi (i.e. at the location of landlord i), who thus have no ethnic cost prefer to work in landlord i þ 1 than to landlord i. This condition is given by: tL q 4 n 2 where W n is the symmetric Nash equilibrium wage given by (12). It is easy q to verify that condition (10) guarantees that both W n tL n 4 2 and (A.7). We have thus shown that, using (10), the local maximum is a global one and that our symmetric Nash equilibrium given by (11) and (12) exists and is unique. Wn
ON THE LAW OF RETURN IN RURAL–URBAN INTERACTIONS: AN ECONOMIC APPROACH TO SOLIDARITY WITH RETURN MIGRANTS$ Carine Drapier, Hubert Jayet and Hillel Rapoport ABSTRACT Community solidarity with return migrants is commonly observed in the rural areas of developing countries. In this paper, we briefly review the evidence from sociological studies on this issue and suggest a new economic approach to such solidarity. We show that an implicit institutional arrangement, whereby migrants have no obligations (e.g., no obligation to remit) but may nevertheless enjoy equal ownership rights on collective resources upon return, enhances economic efficiency via an optimal regulation of migration flows. We also address enforceability issues since, within each generation, time consistency problems may give rise to opportunistic behavior among non-migrants. $ The paper benefited from comments by Fre´de´ric Docquier, Gil Epstein, Ira Gang, an anonymous referee, and participants at the Congress of the European Society for Population Economics, Athens, June 2001, and the Journe´es de Microe´conomie Applique´e, Lille, May 2004.
Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 427–448 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24013-X
427
428
CARINE DRAPIER ET AL.
1. INTRODUCTION Return migration has recently been the focus of many economic and sociological studies. From the return migrants’ standpoint, it has been stressed that private returns to returning depend mainly on the human and financial capital accumulated abroad (which may itself depend on the migration duration), the set of possible occupations upon return, and the quality of social ties with those left behind.1 The impact of return migration on receiving communities has also been the subject of many empirical studies, the results of which are mixed. A first, important strand of the return migration literature, emphasizes the potential benefits of return migration for urban and rural development in the form of human and financial capital inflows. In particular, a large body of research demonstrates that temporary migration, of which return migration is nothing but the last sequence, helps relaxing credit constraints that limit access to entrepreneurship and self-employment at home. Evidence of this may be found for countries such as Tunisia (Mesnard, 2004; Mesnard & Ravallion, 2001), Turkey (Dustmann & Kirchkamp, 2002; Dustmann, 2003b), Mexico (Massey & Parrado, 1998; Woodruff & Zenteno, 2001), Pakistan (Ilahi, 1999) or Egypt (McCormick & Wahba, 2001 and 2003).2 These studies show that migrants opt for temporary migration with the objective of accumulating enough savings abroad to start a business upon return. Remittances may also be seen at least partly as an indirect benefit from return migration for the origin communities. Indeed, it is well recognized in the remittances literature that preparing one’s return (an option most migrants want to keep open) is central to explaining remittance behavior.3 This explains, for example, why skilled migrants, characterized by lower propensities to return, tend to remit relatively less than unskilled migrants despite their having higher earnings potentials (Adams, 2003; Faini, 2005). However, return migration also has a darker side from the perspective of receiving communities. First, it is well known that when migration costs are relatively high, emigrants tend to positively self-select (higher than average skills) out of the home country population (e.g., Chiquiar & Hanson, 2005) whereas for return migrants, there are many examples of negative self-selection. For example, in the case of immigration to the U.S., Borjas and Bratsberg (1996) find that U.S. immigrants returning to their home country compare negatively to the group of migrants remaining at destination. The same conclusion is reached by Ahlburg and Brown (1998) in the case of Samoans and Tongans working in Australia. It is important to stress that these results do not in any way imply that return migrants have lower than
On the Law of Return in Rural–Urban Interactions
429
average skills compared to the native population of the home country (that is, even if return migrants are negatively self-selected among migrants, their return can increase the average level of skills at home); however, in the case of returning Albanian immigrants, De Coulon and Piracha (2005) find that return migrants are negatively self-selected out of the home country population. And second, especially in the rural areas of developing countries, there are many instances in which return migrants come back home without carrying with them any assets or having sent remittances in the past, and yet are entitled to community solidarity no matter how sick, old or needy they are. It is on this second type of return migration, the one that represents an economic burden for the receiving communities, that we focus on in this paper. Previous research has suggested a number of possible rationales for solidarity with unsuccessful and/or unproductive return migrants.4 The common principle behind these explanations is that since migration offers social benefits in the realms of mutual insurance, risk sharing, consumption smoothing, etc. (e.g., Rosenzweig, 1988a; Stark, 1991), if present-day migrants were unable to return upon a host of possible circumstances (retirement, unemployment, etc.), tomorrow’s would be migrants would be reluctant to leave, and the community of origin would end up forfeiting the benefits of migration. In this paper, we offer a supplementary explanation based on the social benefits of migration for a village economy in a context where the marginal productivity of labor in rural activities is decreasing. In such a context indeed, and even if all migrants were unsuccessful and had to return, we show that it is in the community’s interest to provide social security for all in the form of equal access to the village’s collective resources and surplus; this is independent, in particular, of whether the members previously migrated or, for those who did, of whether they had sent remittances in the past. We show why this is the case in a simple model of rural–urban migration with overlapping generations, where migrants have no obligations after leaving and yet may claim a share of the village’s collective surplus and resources upon return. We restrict the exposition to the case of return migration for retirement but could extend it to solidarity with unsuccessful (e.g., unemployed) returnees in their working age. The remainder of this paper is organized as follows. Section 2 reviews the evidence from socio-anthropological studies on community solidarity with return migrants in traditional societies. In Section 3, we present the model for the autarkic economy and characterize equilibrium and optimal solutions once migration possibilities are introduced. More precisely, we show that the decentralized migration equilibrium is sub-optimal, derive the
430
CARINE DRAPIER ET AL.
conditions required for optimality, and propose a simple migration contract that meets such conditions. Specifically, we show that an implicit migration contract whereby community solidarity is offered to returnees without apparent compensation on their sides, enhances economic efficiency by optimally regulating migration flows. Section 4 discusses enforceability issues since, at each generation, the stayers would gain from depriving the movers of their rights. It is shown that this time-consistency problem may be overcome if a sufficient degree of intrafamilial intergenerational altruism prevails. Section 5 concludes.
2. BACKGROUND In this section we first briefly review the pessimistic view developed by leading anthropologists with respect to the developmental impact of return migration to developing countries. We then present cases where the migrants’ rights of return at home are conditioned upon their having met familial and social obligations (e.g., in the form of remittances). Finally, we report evidence from selected case studies documenting the type of solidarity with return migrants we want to analyze, namely, those systems where the migrants’ right of return is guaranteed without apparent compensation on their sides.
2.1. General Statements from the Socio-Anthropological Literature If one is to believe the conclusions from most socio-anthropological essays on return migration to developing countries, the latter is often seen as a curse for the receiving societies. For example, Kearney (1986, p. 346) asserted that: ‘‘in almost all cases, anthropologists have found that the developmental impact of return migration is negative or at best neutral.’’ This would seem particularly relevant when return migration aims at retirement in the village of origin, and returnees may claim access to collective resources and rely on community solidarity for their survival. In the words of Michael Lipton (1980, p. 14): ‘‘Reverse migration for retirement imposes care of the old as a cost on the initially poorer rural sector, increasing rural– urban inequality and reducing any net benefit from migration to the village’s potential productive surplus y making villagers bear ‘social security’ expenses that would otherwise erode the urban surplus.’’ In a similar vein, Gmelch (1980, p. 154) writes: ‘‘The industrial countries benefit from a
On the Law of Return in Rural–Urban Interactions
431
‘readymade’ workforce y at the sending societies’ expenses. And when workers are no longer productive, through illness, accident, or old age, they return home with their maintenance costs again being absorbed by the sending society.’’ These citations highlight that return migrants often represent a burden for the village economies to which they return. However, they may be generalizing too much and conceal other dimensions of the interaction between the migrants and those left behind.
2.2. Right of Return and Inheritance as a Monitoring Device to Secure Remittances Indeed, as is well known, migration and remittances are often part of an implicit migration contract whereby the family finances the migrant’s education and/or migration costs and the migrant’s repays the loan with remittances.5 Because of the temporal structure of this type of contractual arrangement, the migrant may be inclined to deviate from the contractual terms agreed upon. Altruistic preferences on the migrant’s side may prevent opportunistic behavior, though, and this may explain why implicit migration contracts generally take place within a familial context: not only because people are altruistic toward their own family, but also because the family has an informational comparative advantage in obtaining reliable information on individual members (their skills, degree of trustworthiness, etc.). Should intrafamilial altruism be insufficient to make the contract selfenforcing, however, families may sanction opportunistic behavior through inheritance procedures and social sanctions. The testable implications of this inheritance/return-based theory of remittances are that remittances should increase with the receiving household’s assets and income, the probability of inheriting (which depends on the age of the parents, the number of siblings, etc.), and the migrant’s wealth and income, and should decrease with the degree of risk aversion providing that inheritance is more risky than other available forms of savings. And indeed, many empirical studies (e.g., Hoddinott, 1994, for Kenya; de la Brie`re, Sadoulet, de Janvry, & Lambert, 2002, for the Dominican Republic; or Osili, 2004, for Nigeria) find evidence that families and communities use the threat of depriving migrants from their rights of return and/or inheritance as an enforcement device to secure remittances. An interesting implication of this approach is that since rich families only may secure remittances through inheritance, migration tends to increase
432
CARINE DRAPIER ET AL.
inter-household inequality. A limit to this approach may be the fact that many resources are collectively owned in the rural communities rather than family owned, thus limiting the scope for inheritance-seeking through remittances; however, Osili (2004) finds that the same behavior seems to apply at a community level, with migrants investing more in wealthier communities to secure their membership rights.
2.3. Right of Return without Apparent Compensation In what follows, we focus on situations where rights of return have no apparent counterpart in terms of social or financial obligations on the migrants’ side, that is, on situations where return migrants can claim right of access to the village’s common resources and right to community solidarity independently of whether they sent remittances. Such solidarity mechanisms may take a variety of forms, ranging from simple sharing rules to more sophisticated systems of implicit rights preserving the migrants’ access to land and other common resources. Access to common resources is of tremendous importance and is an essential element of poor households’ livelihood and survival strategies. Beck and Nesmith (2000) estimated that income generated by access to common resources represents about 12 percent of poor rural households’ income in India and West Africa. Interestingly, this figure is similar to estimates for eighteenth-century England (Humphries, 1990). In her study of informal solidarity in rural India, Das Gupta (1987, p. 103) indicates that ‘‘[the migrants’ rights] are sometimes encoded in the customary law y Rights accrue by birth as a member of the village and do not lapse on leaving the village, even after a generation has passed. y Consequently, there is a heavy premium on staying in one’s native village or returning to it if one migrates out.’’ Similarly, in the case of French Polynesia, it has been noted that: ‘‘The ease of return migrants’ social reintegration following lengthy absences from the community does not appear to be related to their having retained ties by repatriating remittances y It is universally acknowledged that the land rights of returnees are valid and just’’ (Lockwood, 1990, pp. 360–366).6 These same patterns based on the migrant’s rights on land are also well documented for sub-Saharan Africa.7 For Burkina Faso, Breusers (2001, p. 72) writes that: ‘‘In the course of the twentieth century, the land tenure regime developed in such a way that those who leave a place most often continue to be ensured of access to land when they eventually return. Their
On the Law of Return in Rural–Urban Interactions
433
access rights to land ‘at home’ do not lapse and give long-term social security.’’ It has also been emphasized, however, that such traditional arrangements are not immune to change, especially in a context of increased demographic pressure. This was noted in the case of rural India in a number of studies reviewed in Beck and Nesmith (2000) and, for Rwanda, by Andre´ and Platteau (1998). In other instances, when rights on land are inexistent or insufficient to achieve minimal subsistence, solidarity takes the form of hospitality and sharing rules. For example, in their recent study of two rural regions of Northern Cameroon, Gubry et al. (1996) noted that nearly half of the returnees were housed by relatives (without apparent exchange) and that most returnees (48 percent in the Western region and 84 percent in the Northern region) received a parcel of land for self-cultivation, either in full property or with free disposal. But even for those households who did receive a parcel upon return, community solidarity appeared to be an essential component of livelihood and survival strategies: the proportion of returnees for which community solidarity was the main source of food amounted to 47.5 percent in the Northern region, and 60.6 percent in the Western region studied. In summary, in the words of Platteau (2005, p. 15), ‘‘the customary land tenure system prevailing in lineage-based societies appears to be guided by both insurance and equity considerations. The insurance motive is evident from the fact that membership in a rural community automatically ensures proper access to land, be it in the form of use of rights over privately apportioned land plots or over the village commons. Thus, for instance, a member who emigrated and later decides to return to his native village because of economic difficulties or other reasons is entitled to obtain enough land to make a living.’’ These characteristics will serve partly as justification of the assumptions retained in the model below (see Section 3). However, the core of our argument is that solidarity with return migrants generates efficiency gains beyond the benefits from mutual insurance emphasized in the above literature survey.
3. THE MODEL In this section, we present a simple model of rural–urban migration with overlapping generations that accounts for the main patterns of return migration and community solidarity just described. Two critical assumptions are made that need justification: first, we assume decreasing marginal
434
CARINE DRAPIER ET AL.
productivity of labor in the rural sector, and second, we assume that there is surplus sharing taking place at the community level. The first of these assumptions may be questioned with particular reference to the work of Arthur Lewis (1954) who assumed constant marginal productivity in a context of absolute labor surplus. However, Harris and Todaro (1970) have documented the fact that the marginal productivity of labor was decreasing in traditional agriculture in various contexts (sub-Saharan Africa, South-East Asia, Latin America); in line with these findings, Harris and Todaro (1970) based their analysis of rural–urban migration on decreasing returns in rural production. More recent research on traditional agricultural systems have confirmed this conclusion but have also emphasized that it could be partly hidden by seasonal movements (e.g., Connell, Dasgupta, Laishley, & Lipton, 1976; Lucas, 1997) or increased work effort by remaining farmers after some out-migration has taken place (Stiglitz, 1988; Rosenzweig, 1988b).8 The second of these assumptions, namely, that surplus allocation is a collective decision made by older community members, is clearly a simplification. In reality, traditional solidarity systems are extremely diverse. However, with few exceptions, these systems share two central characteristics that justify the way we treat redistribution in our model. Instead of providing a series of examples, we will refer mainly to Fafchamps’ (1992) illuminating theoretical investigation of solidarity networks in pre-industrial societies. First, political power is concentrated in the hands of older people. From an economic viewpoint, this may be explained by a social organization aiming at mutual insurance, including between generations. In a system of mutual insurance where cooperation is sustained by the threat of retaliation in case of default, the threat of the young is much more credible than the threat of the old, the latter being more likely to be in need of assistance and having a lower survival probability (and, hence, fewer opportunities to retaliate in the future). Such an asymmetry in bargaining power threatens the whole solidarity edifice, and ‘‘this may explain why primitive and other preindustrial societies try to compensate by granting the old a lot of political and economic power y through direct authority over land and livestock’’ (Fafchamps, 1992, p. 151). And second, informal solidarity arrangements generally extend themselves beyond the boundaries of the enlarged family and are best described as a web of networks that encompasses the whole village community. Each person is connected to a relatively small number of other people according to familial, kinship, neighborhood or professional affiliation, and these
On the Law of Return in Rural–Urban Interactions
435
partial networks are interconnected one with another, therefore forming a global solidarity network at the community (village) level. The daily operation of this global network may be decentralized at the partial network level to save on information costs and, hence, achieve mutual insurance in a cost-effective manner. Indeed, ‘‘[partial] networks are more resilient and flexible than a global insurance pool. Births, deaths, weddings and migrations are easily accommodated without having to renegotiate and reconsider the insurance arrangements of the entire community’’ (Fafchamps, 1992, p. 159). Compared to a standard Harris-Todaro model, we assume hereafter that individuals live two periods and the expected urban wage is given; in other words, each single village has no influence on expected urban wages, which depend on the overall (country-wide) flows of rural–urban migration, of which the village economy we consider is only a fraction.9 In addition, to avoid distorting our results with additional effects well recognized in the rural–urban migration literature, we make the simplifying assumptions that there is no uncertainty neither in the rural sector nor in the urban sector, and individuals’ inter-temporal utility is linear in income. Hence, we rule out the possibility of social gains from migration through mutual insurance or consumption smoothing, and can focus on the externality argument outlined above.10 We first describe the autarkic economy and then compare equilibrium and optimal solutions once migrations are introduced.
3.1. Autarky Consider, therefore, a rural economy using homogenous labor and land to produce a given consumption good. At each period, two generations coexist, each generation consisting of a fixed number of agents, N. N1 and N2 denote the number of young (active) and old (inactive) agents, respectively. Each individual lives two periods. During the first period, he or she inelastically supplies one efficient unit of labor. Individuals are employed either in the rural sector and paid their marginal product or migrate to a high-wage destination (urban district or developed country); as explained, wages at destination are unaffected by migration. The production function in agriculture is simply given by Qt ¼ F ðN 1 Þ
(1)
where N1 is the number of efficient units of labor employed. This production function exhibits decreasing returns to scale and is assumed to conform to
436
CARINE DRAPIER ET AL.
Inada’s conditions: F ð0Þ ¼ 0; F 0 ðN 1 Þ 0; F 00 ðN 1 Þ 0; F 0 ð0Þ ¼ þ1; F 0 ðþ1Þ ¼ 0 During the second period, the individual retires. Old-age security is fully conditioned upon access to the village’s collective resources. Although in practice this may take a variety of forms, as described above, we interpret this right as a right of access to the social surplus, S, defined as S ¼ F ðN 1 Þ
N 1 F 0 ðN 1 Þ
(2)
This surplus is collected by the older members of the community and equally shared among them. Anyone who fulfilled his social obligations when young (i.e., who worked in the village) is entitled to a surplus share. We denote by R the share of the social surplus accruing to each older agent: R¼
S 1 ½F ðN 1 Þ ¼ N2 N2
N 1 F 0 ðN 1 Þ
(3)
Agents are homogenous and utility is linear in consumption: V ðC 1 ; C 2 Þ ¼ C 1 þ C 2
(4)
where C1 and C2 are the consumption levels of the first and second period. In an autarkic economy, by definition there is no migration (N ¼ N 1 ¼ N 2 ); coupled with the rest of our assumptions, this yields the representative agent’s utility, V R ðC 1 ; RÞ ¼ C 1 þ R; with R now being defined as the difference between average and marginal productivity of labor: R¼
F ðN 1 Þ N1
F 0 ðN 1 Þ
(5)
Note that @R=@N 1 may be either positive or negative, but we need no restriction on its sign for our results to hold.11
3.2. Equilibrium with Migration We now allow for possible migration of young workers in a decentralized setting, that is, when migration implies an implicit renouncement to return. Note, however, that we do not consider a fully decentralized economy, as the collection and distribution of the surplus is still assumed to be handled collectively by the older members of the community. Migrant workers are paid a fixed wage W; this wage is net of migration costs (whether monetary or psychological) and includes a pension paid at retirement. In the absence of any social arrangement whereby the migrants’
On the Law of Return in Rural–Urban Interactions
437
rights of return would be guaranteed, the migrants’ inter-temporal utility is therefore: V U ðW Þ ¼ W
(6)
From the potential migrant’s perspective, migration is worthwhile as long as utility at destination is higher than at home prior to departure. Hence, workers emigrate as long as V U 4V R : The impact of migration flows, M, on VR is given by ½F ðN 1 Þ N 1 F 0 ðN 1 Þ @V R @V R @V R N1 ¼ ¼ F 00 ðN 1 Þ 1 þ (7) @M @N 1 @N 2 N2 ðN 2 Þ2 At the steady state, the number of migrants is stationary; with N1 ¼ N2, the marginal impact of migration on the utility of the representative stayer is given by @V R 1 F ðN 1 Þ R 0 ¼ F ðN 1 Þ ¼ 40 (8) N1 N1 N1 @M showing that the inter-temporal utility of the non-migrants is monotonically increasing in the number of migrants. In equilibrium, utility levels are equalized across locations (i.e., the average productivity of labor in the rural sector is equal to the urban wage) and N e1 members of a given generation remain in the village: V U ¼ V R 3W ¼ F 0 ðN e1 Þ þ RðN e1 Þ3W ¼ F ðN e1 Þ N e1 (9)
Obviously, emigration is worthwhile from the migrants’ perspective (for if it was not, they would not migrate voluntarily) and also benefits to those left behind. This is apparent from Fig. 1, where [AB] gives the representative individual’s budget constraint and utility level (which are identical since C1 and C2 are perfect substitutes) in the autarkic economy. Note that C2(C1) is the locus of technologically compatible present and future consumption profiles for all possible levels of the active rural population (and, consequently, for all possible migration levels).12 Assume an initial situation where the autarkic economy is, say, at point I, the intersection between [AB] and C2(C1). Allowing for migration creates inequality at both periods, as the consumption profiles of the two groups of agents may now differ over time (for example, the migrants may be at point Me and the non-migrants are at point Se). However, in equilibrium, the two groups necessarily have the same utility level (represented by the indifference curve [A0 B 0 ]), with the exact inter-temporal profile depending on C2(C1)
438
CARINE DRAPIER ET AL. C2 A'' A' C2(C1)
A
Se
Mo So
I Me B
B'
C1
B''
Fig. 1. Inter-Temporal Utility and Consumption Profiles.
and on the size of the pension included in W. Clearly, free migration is pareto-improving; it is not pareto-optimal, however, as we explain below.
3.3. Social Optimum Consider now the program of a social planner for the rural community. Let us define social welfare as the sum of individual utilities; with N agents indexed by i, we have: SW ¼
N X ðC 1i þ C 2i Þ ¼ F ðN 1 Þ þ ðN i¼1
N 1 ÞW
(10)
Differentiating (10) with respect to N1, it comes that an optimal number of migrants is attained when the marginal productivity of labor is equal to the urban wage: F 0 ðN 1 Þ ¼ W
(11)
Clearly, the optimum condition (11) differs from the equilibrium condition (9): since the average productivity of labor is higher than the marginal productivity of labor, this means that there are too few migrants in equilibrium: N e1 4N o1 ; with N o1 denoting the level of migration consistent with (11). Let us now focus on average income, which is simply given by dividing (10) over the number of members, N; denoting this average income by
439
On the Law of Return in Rural–Urban Interactions
Z(N1), we have F ðN 1 Þ N1 N 1 F ðN 1 Þ þ 1 þ 1 W¼ N N N N1 N 1 F ðN 1 Þ W ¼Wþ N1 N
ZðN 1 Þ ¼ C 1 þ C 2 ¼
N1 W N ð12Þ
Plugging the optimum condition (11) into (12) gives the maximal average income: N 1 F ðN 1 Þ N1 ZðN 1 Þ ¼ W þ RðN 1 Þ (13) F 0 ðN 1 Þ ¼ W þ N1 N N Assume that each agent fully consumes his or her first-period income during the first period; for simplicity and without loss of generality assume also that the pension included in W is zero. This gives C 1 ¼ F 0 ðN o1 Þ for the nonmigrants and C 1 ¼ W for the migrants. The second-period incomes compatible with the budget constraint are obtained by the difference between total income and first-period consumption, that is: – For thenon-migrants: ZðN 1 Þ F 0 ðN 1 Þ ¼ W þ N 1 =N RðN 1 Þ F 0 ðN 1 Þ ¼ N 1 =N RðN 1 Þ: – For the migrants: ZðN 1 Þ W ¼ N 1 =N RðN 1 Þ: Hence, the following proposition: Proposition 1. An institutional arrangement ensuring that the social surplus is equally divided between all society’s older residents regardless of their previous location brings about a social optimum. If an implicit social contract as specified in Proposition 1 applies, it is clear that the migrants’ inter-temporal income is increased since they gained equal access to the social surplus without losing anything. This corresponds on Fig. 1 to the move from Me to Mo on the higher indifference curve [A00 B 00 ], Moreover, the return contract also modifies the non-migrants’ consumption profile. Their first-period income is increased via the positive effect of increased migration on the marginal productivity of labor, but their second-period income is decreased because the surplus now has to be divided between a larger number of shares. This is illustrated in Fig. 1 via the move from point Se to point So, South-East from Se on the higher indifference curve [A00 B 00 ].
440
CARINE DRAPIER ET AL.
4. CONTRACT ENFORCEABILITY Since the benefits of the above implicit migration contract accrue to the nonmigrants during their youth and to the migrants after they returned and gained access to the social surplus, one may wonder why non-migrants adopt the equal sharing rule instead of opportunistically keeping return migrants out of the sharing pool. For a better understanding of the problem, let us consider the following infinitely repeated game played at each period t by three categories of workers: the former migrants (who emigrated at date t 1), the former non-migrants (those who, upon entry into the labor force at date t 1; did not migrate), and the younger generation at the beginning of period t. There are three stages in the static game: (1) Assume that, at the beginning of period t 1; the members of that generation agreed that whoever migrates would be entitled to a share of the social surplus upon return. Then, at the beginning of period t, the former non-migrants – who collectively hold property of land – vote on whether to honor their commitment. (2) At the same time, the former non-migrants propose two contracts to the members of the younger generation. First is a job contract: those who opt for this job contract work in the village and are paid their marginal product during period t, knowing that at time t þ 1; they would collectively become owners of the common property with the counterpart of committing to equally share the surplus with the returnees of their own generation. Second is a migration contract stipulating that, would a young worker migrate at period t and return at period t þ 1; he or she would receive a share of the surplus. (3) M young workers opt for the migration contract, migrate and receive the wage W. The N 1 ¼ N M non-migrants are hired in the village, produce the output quantity F ðN 1 Þ and are paid their marginal product F 0 ðN 1 Þ: (4) Finally, we go back to stage 1: former migrants return to the village and former non-migrants decide whether to honor their commitment toward the returnees (in which case everybody receives S=N) or renege (in which case they receive S=N 1 while returning migrants receive nothing). Let us temporarily assume that, before generation t 1; former nonmigrants where honouring their commitment and that, at the first stage, the young workers of generation t expect that, at period t þ 1; non-migrants would also honor their commitment. In this case, there is an optimal number of
441
On the Law of Return in Rural–Urban Interactions
migrants (N 1 ¼ N O 1 ), who expect the same earnings when they are young, W ¼ F 0 NO N (recall that and when they are old, S N O 1 1 ; S ðN 1 Þ ¼ F ðN 1 Þ N 1 F 0 ðN 1 Þ). Therefore, both the migrants and the non-mi O N¼ grants have the same expected lifetime income, V R ¼ F 0 N O 1 þ S N1 V M ¼ W þ S NO N: Moreover, as V R V M ¼ F 0 ðN 1 Þ W is a decreas1 ing function of N1, this equilibrium is stable.13 This cooperative outcome corresponds to the upper panel of Fig. 2. However, it is not a sub-game perfect equilibrium of the static game. Indeed, at the third stage, when old, if former non-migrants honor their commitment, they receive S ðN 1 Þ=N while if they renege, they receive the higher payment S ðN 1 Þ N O 1 : With N1 next Olifetime O migrants, their generation income increases to V R ¼ F 0 N O 1 þ S ðN 1 Þ N 1 ¼ W þ S ðN 1 Þ N 1 ; while Honor Stayers Generation t
Migrants
Generation t–1 S ( N1o ) F ' ( N1o ) + N S ( N1o ) W+ N
Generation t–1 Stayers
Honor Renege
Migrants
S ( N1o ) F '(N ) + N o 1
W+
Generation t S ( N1o ) F ' ( N1o ) + N S ( N1o ) W+ N
Generation t S ( N1e ) N1o
F ' ( N1o ) +
S ( N1o ) N
W +0
Generation t – 1 when old Renege
Honor
Stayers
Generation t–1 S ( N1e ) F ' ( N1o ) + N1o W +0
Migrants
Generation t F ' ( N1e ) + W+
S ( N1o ) N
S ( N1o ) N
Generation t when old
Renege
Stayers Migrants
Fig. 2.
Generation t–1 S ( N1e ) F ' ( N1o ) + N1o W +0
The Game Tree.
Generation t S ( N1e ) F ' ( N1e ) + ) N1e W +0
442
CARINE DRAPIER ET AL.
the former migrants’ lifetime income decreases to V M ¼ W : Therefore, the best response of former stayers is to behave opportunistically and unanimously renege on their commitment and the outcome of the game is one of the two lower panels of Fig. 2. Since the young workers can rationally anticipate this at the first stage, the non-cooperative outcome (i.e., when, for all generations, only M e ¼ N N e1 young workers migrate and receive nothing upon return, so that V M ¼ W ¼ F N e1 N e1 ¼ V R ) is the unique equilibrium of the static game. Obviously, this sub-optimal equilibrium is also a possible sub-game perfect equilibrium of the infinitely repeated game. The context we describe is a context of overlapping generations. In such a context, we know from Kandori (1992a,b) that even if every individual player has a finite lifetime, an ‘‘OLG Folk Theorem’’ applies and cooperation can be sustained if the players are patient enough and their life spans are sufficiently long relatively to the overlapping periods. The key to Kandori’s result is that, as each generation may observe the behavior of the older generation before making a decision, the former can sanction opportunistic behavior by the latter. Does such a punishment mechanism apply in our context? In other words, may defection from cooperation in one generation be credibly sanctioned by the next generation? It is easy to see that the answer is negative. In our model, such a punishment mechanism could in principle arise since the members of the younger generation can observe the behavior of the older generation before making their migration decision. If they observed a defection within the previous generation, they would rationally anticipate the same opportunistic behavior to repeat itself in their own generation and, consequently, would migrate in equilibrium instead of optimal numbers. Such a move, however, punishes the members of their own generation instead of punishing the previous generation. Indeed, if N e1 4N o1 young members remain in the village, the surplus to be shared among the stayers of the older generation, N 1 RðN 1 Þ ¼ F ðN 1 Þ N 1 F 0 ðN 1 Þ; increases. Thus, instead of being sanctioned, former non-migrants of the older generation are rewarded for having defected from the social contract, which is definitely not self-enforcing. In such a setting, enforcement can only be external unless some element of social or intergenerational concern is introduced. For example, assume that each individual gives birth to one single child and attaches a given weight to the child’s welfare. Then, the representative individual’s utility function becomes: V ðC 1 ; C 2 Þ ¼ C 1 þ C 2 þ bV C
(14)
On the Law of Return in Rural–Urban Interactions
443
where Vc denotes the child’s welfare and b is a parameter of intergenerational altruism randomly distributed among the N members of the community. It is straightforward to show that, with this new utility function, the optimal outcome does not change (N o1 is not modified) and the only equilibrium of the static game is still the non-cooperative one. However, intergenerational altruism may change the outcome of the infinitely repeated game because, at the time they decide whether to renege on their commitment, former non-migrants now take into account the impact of their decision on the next generation’s welfare. In this context, the cooperative outcome may emerge as equilibrium in the infinitely repeated game. Recall that if the young workers expect that the non-migrants would accept the equal sharing rule, optimal numbers. In this case, they would migrate in C 1 ¼ W ¼ F 0 N o1 and C 2 ¼ N o1 N F N o1 N o1 W : Moreover, as the utility level is the same for all generations, V ¼V C ; so that utility at optimum, Vo, is given by V o ¼ W þ N o1 N F N o1 N o1 W þ bV o : Hence, 1 N o F N o1 Vo ¼ W (15) Wþ 1 N o1 1 b N For this outcome to be a sub-game perfect equilibrium of the infinitely repeated game, former non-migrants must accept to share the surplus with the return migrants. If they do not, the migrants of their children’s generation and of all future generations will expect to receive nothing upon return. There will therefore be only N N e1 migrants and each agent’s intertemporal consumption will be equal to C 1 þ C 2 ¼ W ; the corresponding utility level is V n ¼ W þ bV n ; which may be rewritten as Vn ¼
W 1
b
(16)
As explained, if the former non-migrants of a given generation renege on their consumption increases from their second-period commitment, N o1 N F N o1 N o1 W ; however, since the utility W to F N o1 N o1 level of their children decreases from V o to V n, the utility level associated to reneging, V r, is now given by F N o1 F N o1 bW r n (17) V ¼Wþ W þ bV ¼ þ 1 b N o1 N o1
444
CARINE DRAPIER ET AL.
The migration contract will be self-enforcing if the utility derived from not reneging, V o, is higher than the utility from reneging, V r, i.e. if: F N o1 N o1 F N o1 r o W ð1 bÞV ¼ ð1 bÞ þ bW ð1 bÞV ¼ W þ N o1 N o1 N This gives the following condition: N o1 1 b W 1 N
b
N o1 F N o1 N o1 N
(18)
from which we derive our second proposition: Proposition 2. For a given distribution of the altruistic preferences, the lower the optimal migration rate, the higher the likelihood of a cooperative outcome. Indeed, and noting that W ¼ F 0 N o1 F N o1 No1 ; condition (18) is equivalent to 1 b N o1 N 0; or b bn ¼ 1 N o1 N; with the threshold b being the minimal degree of intergenerational altruism required for an agent to vote against reneging. As there is no room for manipulation when there are only two alternatives, the collective decision will be determined by the median voter’s vote. Cooperation, therefore, is sustainable if and only if the median voter has a degree of altruism higher than the threshold b . Notice that this threshold has a very simple expression: it is equal to the optimal migration rate, as stated by Proposition 2.
5. CONCLUSION Community solidarity with unproductive return migrants is commonly observed in the rural areas of developing countries. In this paper, we briefly reviewed the evidence from sociological studies and proposed a simple economic approach to such solidarity. Specifically, we showed that an implicit institutional arrangement, whereby migrants have no obligations but may nevertheless enjoy equal ownership rights upon return, allows for internalizing the positive migration externality that arises in a context of decreasing marginal productivity of labor in rural activities. This is in addition to other potential benefits (e.g., mutual insurance) suggested by previous research.
On the Law of Return in Rural–Urban Interactions
445
Using an overlapping generations framework, we also questioned the enforceability properties of such a contract, suggesting that time consistency problems that characterize the static game may be overcome in the dynamic game providing that a sufficient degree of intergenerational altruism prevails. Interestingly, the altruistic threshold required for cooperation to prevail in the infinitely repeated game was shown to be proportional to the optimal migration rate; from a comparative statics standpoint, therefore, the main implication of our analysis is that any increase in the desirable level of rural– urban migration (e.g., higher wages at destination, demographic pressure or soil degradation at origin) also puts the institution of community solidarity with return migrants at risk. This could go part of the way toward explaining why, as documented by Platteau (2005), environmental or demographic pressures may lead to an abrupt end of such solidarity mechanisms.14
NOTES 1. On private returns to returning, see for example Co, Gang, and Yun (2000) or De Coulon and Piracha (2005). See also Stark (1991), and Dustmann (1997), on how the possibility of return affects migrants’ labor supply and savings behavior, and Dustmann (2003a) on the impact of children on parents’ return migration decisions. 2. See Mesnard (2001) and Rapoport (2002) for a theoretical perspective. 3. See Rapoport and Docquier (2005) for a comprehensive survey on migrants’ remittances. 4. Fafchamps (1992) offers a general discussion of solidarity mechanisms in traditional rural societies in a game-theoretic framework. 5. See the section on familial loan arrangements in Rapoport and Docquier (2005). 6. Lockwood (1990, p. 366) adds: ‘‘This does not necessarily reduce resentment among resident families, some of whom had achieved de facto control over, and use of, tracts of family lands to which absentees have rights.’’ 7. See Franqueville (1973), and Pontie´ (1979), on Southern and Northern Cameroon, respectively, Boutillier, Quesnel, and Vaugelade (1977), on Burkina Faso, and Delpech (1983), on Ivory Coast. 8. In his essay on solidarity networks in traditional societies, Fafchamps (1992) notes incidentally that land borrowing free of charge is very common within mutual assistance groupings in the West African semi-arid tropics, and explains this practice as an ex ante solidarity mechanism aiming at decreasing ex post claims to assistance; as noted by Fafchamps, this presents a number of advantages in terms of monitoring opportunistic behavior and, in addition, allows for an efficient use of land resources in a context characterized by decreasing returns in agriculture. 9. Extending the framework to the case where more migration reduces the expected urban wage through a decrease in the employment probability would not alter the essence of our results.
446
CARINE DRAPIER ET AL.
10. On commons as insurance, see Baland and Franc- ois (2005). 11. Since C1 is the marginal product of the working age population, @C 1 =@N 1 o0: From Eq. (5), we have that @C 2 =@N 1 ¼ @R=@N 1 ¼ ½N 1 F 0 ðN 1 Þ F ðN 1 Þ N 21 F 00 ðN 1 Þ; the sign of which is undetermined. After manipulations, 2 straightforward 00 this can be rewritten as: @C =@N ¼ ½FðN Þ N ½x 1 F ðN Þ; with xN 1 ¼ 2 1 1 1 N 1 1 @F ðN 1 Þ=@N 1 N 1 =FðN 1 Þ; the elasticity of production to labor. It is clear that the sign of @C 2 =@N 1 is positive if xN1 41 and negative otherwise. 12. The sign of the relationship between C1 and C2 depends on whether @C 2 =@N 1 is positive or negative (see the discussion on this in footnote 11 supra). We draw Fig. 1 assuming a positive slope, but could as well use a negative relationship, as exemplified by the dashed line. R 13. If there are too many non-migrants ðN 1 4N O V M ¼ F 0 ðN 1 Þ W ; is 1 Þ; V negative, thus inducing more migration. The opposite holds when there are too many migrants. 14. Rwanda would seem to be a perfect illustration of these processes. As noted by Andre´ and Platteau (1998, p. 32), ‘‘the evolution of indigenous tenure arrangements involves increasing exclusion of vulnerable categories of the population which were socially protected under erstwhile customary rules. This holds true especially for return migrants and separated or divorced women.’’
REFERENCES Adams, R. (2003). International migration, remittances and the brain drain. Policy Research Paper no. 3069. World Bank, Washington. Ahlburg, D. A., & Brown, P. C. (1998). Migrants’ intention to return home and capital transfers: A study of Tongans and Samoans in Australia. Journal of Development Studies, 35(2), 125–151. Andre´, C., & Platteau, J.-P. (1998). Land relations under unbearable stress: Rwanda caught in the Malthusian trap. Journal of Economic Behavior and Organization, 34(1), 1–47. Baland, J.-M., & Franc- ois, P. (2005). Commons as insurance and the welfare impact of privatization. Journal of Public Economics, 89(2/3), 211–231. Beck, T., & Nesmith, C. (2000). Building on poor people’s capacities: The case of common property resources in India and West Africa. World Development, 29(1), 119–133. Borjas, G. J., & Bratsberg, B. (1996). Who leaves? The outmigration of the foreign-born. Review of Economics and Statistics, 78(1), 165–176. Boutillier, J. L., Quesnel, A., & Vaugelade, J. (1977). Syste`me socio-e´conomique mossi et migrations. Cahiers des Sciences Humaines de l’ORSTOM, 14(4), 361–381. Breusers, M. (2001). Searching for livelihood security: land and mobility in Burkina Faso. Journal of Development Studies, 37(4), 49–80. Chiquiar, D., & Hanson, G. H. (2005). International migration, self-selection and the distribution of wages: Evidence from Mexico and the United States. Journal of Political Economy, 113(2), 239–281. Co, C. Y., Gang, I. N., & Yun, M.-S. (2000). Returns to returning. Journal of Population Economics, 13(1), 57–79. Connell, J., Dasgupta, B., Laishley, R., & Lipton, M. (1976). Migration from rural areas: The evidence from village studies. Delhi: Oxford University Press.
On the Law of Return in Rural–Urban Interactions
447
Das Gupta, M. (1987). Informal security mechanisms and population retention in rural India. Economic Development and Cultural Change, 36(1), 101–120. De Coulon, A., & Piracha, M. (2005). Self-selection and the performance of return migrants: the case of Albania. Journal of Population Economics, forthcoming. de la Brie`re, B., Sadoulet, E., de Janvry, A., & Lambert, S. (2002). The roles of destination, gender, and household composition in explaining remittances: An analysis for the Dominican Sierra. Journal of Development Economics, 68(2), 309–328. Delpech, B. (1983). Les nouveaux abidjanais et leurs racines. Cahiers des Sciences Humaines de l’ORSTOM, (4), 567–584. Dustmann, C. (1997). Return migration, uncertainty and precautionary savings. Journal of Development Economics, 52(2), 295–316. Dustmann, C. (2003a). Children and return migration. Journal of Population Economics, 16(4), 815–830. Dustmann, C. (2003b). Return migration, wage differentials, and the optimal migration duration. European Economic Review, 47(2), 353–367. Dustmann, C., & Kirchkamp, O. (2002). The optimal migration duration and activity choice after re-migration. Journal of Development Economics, 67, 351–372. Fafchamps, M. (1992). Solidarity networks in pre-industrial societies: Rational peasants with a moral economy. Economic Development and Cultural Change, 41(1), 147–174. Faini, R. (2005). Remittances and the brain drain. Paper presented at the Annual Meeting of the Royal Economic Society, Nottingham, March. Franqueville, A. (1973). L’e´migration rurale dans le de´partement de la Le´kie´, contribution a` l’e´tude des relations villes-campagnes dans le Sud du Cameroun. Cahiers des Sciences Humaines de l’ORSTOM, 10(2–3), 151–193. Gmelch, G. (1980). Return migration. Annual Review of Anthropology, 9, 135–159. Gubry, P., Lamlenn, S. B., Ngwe´, E., Tche´gho, J. M., Timnou, J. P., & Ve´ron, J. (1996). Le retour au village, une solution a` la crise e´conomique au Cameroun? Paris: L’Harmattan. Harris, R., & Todaro, M. P. (1970). Migration, unemployment and development: A two-sector analysis. American Economic Review, 60, 126–142. Hoddinott, J. (1994). A model of migration and remittances applied to Western Kenya. Oxford Economic Papers, 46, 450–475. Humphries, J. (1990). Enclosures, common rights, and women: The proletarianization of families in the late eighteenth century and early nineteenth century. Journal of Economic History, 50(1), 17–42. Ilahi, N. (1999). Return migration and occupational change. Review of Development Economics, 3(2), 170–186. Kandori, M. (1992a). Social norms and community enforcement. Review of Economic Studies, 59, 63–80. Kandori, M. (1992b). Repeated games played by overlapping generations of players. Review of Economic Studies, 59, 81–92. Kearney, M. (1986). From invisible hand to visible feet: Anthropological studies of migration and development. Annual Review of Anthropology, 15, 331–361. Lewis, W. A. (1954). Economic development with unlimited supplies of labor. The Manchester School, 22, 139–191. Lipton, M. (1980). Migration from rural areas of poor countries: The impact on rural productivity and income distribution. World Development, 8, 1–24.
448
CARINE DRAPIER ET AL.
Lockwood, V. S. (1990). Development and return migration to rural French Polynesia. International Migration Review, 24(2), 347–371. Lucas, R. E. B. (1997). Internal migration in developing countries. In: M. R Rosenzweig, O. Stark (Eds), Handbook of population and family economics, (Vol. 1B). Amsterdam: Elsevier North-Holland. (Chapter 13) Massey, D. S., & Parrado, E. A. (1998). International migration and business formation in Mexico. Social Science Quarterly, 79(1), 1–20. McCormick, B., & Wahba, J. (2001). Overseas work experience, savings and entrepreneurship amongst return migrants to LDCs. Scottish Journal of Political Economy, 48(2), 164–178. McCormick, B., & Wahba, J. (2003). Return international migration and geographical inequality: The case of Egypt. Journal of African Economies, 12(4), 500–532. Mesnard, A. (2001). Temporary migration and intergenerational mobility. Louvain Economic Review, 67(1), 59–88. Mesnard, A. (2004). Temporary migration and capital market imperfections. Oxford Economic Papers, 56(2), 242–262. Mesnard, A., & Ravallion, M. (2001). Wealth distribution and self-employment in a developing country. CEPR Discussion Paper DP3026. Osili, U. O. (2004). Migrants and housing investment: Theory and evidence from Nigeria. Economic Development and Cultural Change, 52(4), 821–850. Platteau, J. P. (2005). Solidarity norms and institutions in agrarian societies: Static and dynamic considerations. In: L. A. Gerard-Varet, S. C. Kolm & J. Mercier Ythier (Eds), Handbook of the economics of giving, reciprocity and altruism. Amsterdam: Elsevier-North-Holland forthcoming. Pontie´, G. (1979). La contestation par la migration, le cas des Guiziga du Nord-Cameroun. Cahiers des Sciences Humaines de l’ORSTOM, 16(1–2), 111–127. Rapoport, H. (2002). Migration, credit constraints and self-employment: A simple model of occupational choice, inequality and growth. Economics Bulletin, 15(7), 1–5. Rapoport, H., & Docquier, F. (2005). The economics of migrants’ remittances. In: L. -A. Ge´rard-Varet, S. -C. Kolm & J. Mercier Ythier (Eds), Handbook of the economics of reciprocity, giving and altruism. Amsterdam: Elsevier-North-Holland forthcoming. Rosenzweig, M. R. (1988a). Risk, implicit contracts and the family in rural areas of low-income countries. Economic Journal, 393, 1148–1170. Rosenzweig, M. R. (1988b). Labor markets in low-income countries. In: H. Chenery, T. N Srinivasan (Eds), Handbook of development economics (Vol. 1). Amsterdam: ElsevierNorth-Holland. Chapter 15. Stark, O. (1991). The migration of labor. Oxford and Cambridge, MA: Basil Blackwell. Stiglitz, J. E. (1988). Economic organization, information, and development. In: H. Chenery, T. N Srinivasan (Eds), Handbook of development economics (Vol. 1). Amsterdam: ElsevierNorth-Holland. (Chapter 5). Woodruff, C., & Zenteno, R. (2001). Remittances and micro-enterprises in Mexico. Mimeo: University of California at San Diego.
AN ECONOMIC PERSPECTIVE ON RELIGIOUS EDUCATION: COMPLEMENTS AND SUBSTITUTES IN A HUMAN CAPITAL PORTFOLIO$ Carmel U. Chiswick ABSTRACT Models the trade-offs between education in secular subjects, formal and informal, and the formation of religion-specific human capital. Explores some implications of negative externalities between religious and secular education. Develops hypotheses about religious tensions in the American public school system and means of coping with them. Discusses some implications for social cohesion in a religiously pluralistic school system.
$
An earlier version of this paper was presented in June, 2004 at Bar Ilan University to Immigration, Minorities, and Social Exclusion, an international conference in memory of Tikva Lecker. It grew out of ideas initially presented in January, 2004 at Ben-Gurion University of the Negev to a conference on Economics of Religion. The current paper has also benefited from comments received at the 2004 meetings of the Illinois Economic Association and at seminars in George Mason University (Fairfax, Virginia) and the Curtin Institute of Technology (Perth, Australia).
Research in Labor Economics: The Economics of Immigration and Social Diversity Research in Labor Economics, Volume 24, 449–467 Copyright r 2006 by Elsevier Ltd. All rights of reproduction in any form reserved ISSN: 0147-9121/doi:10.1016/S0147-9121(05)24014-1
449
450
CARMEL U. CHISWICK
1. INTRODUCTION Education is an important aspect of religious life. Most parents aspire to raise their children in their own religion, defined broadly to include as ‘‘religion’’ any belief system that speaks to spiritual needs and for which there is a community of adherents. They also provide their children with a ‘‘secular’’ education that enhances their productivity as workers, as consumers and as members of society. Individuals may be viewed as choosing a portfolio of human-capital investments, some of which are religious (i.e., specific to a particular belief system) and the rest of which are secular (i.e., general with respect to religion). Religious and secular educations are thus substitutes for each other as they compete for investment resources, primarily time and money. The goal (output) of education is the formation of human capital. The prototype educational process is formal schooling, a system especially well suited to the formation of cognitive knowledge and certain types of decisionmaking skills. In a modern economic setting, formal schooling is also an important means of acquiring occupational skills that associate positively – and often strongly – with future earning power. Another method of humancapital formation involves experience with activities that provide skills through observation, repetition or familiarity. On-the-job training is but one example. Other human capital is accumulated through social interactions that provide feedback about the desirability of various behaviors and produce memories crucial for identity formation. All three methods of education – schooling, experience and socialization – are important determinants of the human capital available to an adult decision-maker and affect his or her resource allocation decisions. The economics literature tends to focus on those types of human capital most relevant for the workplace, on investments in schooling and experience that raise the productivity of labor and therefore hourly earnings (Mincer, 1974, 1984). Other types of human capital that have received attention include investments in health (Schultz, 1980, 1993), family (marriage and children) (Becker, 1981; Schultz, 1981), language (Chiswick & Miller, 1995), culture (including the arts) (Filer, 1990; Smith, 1998), religion (Iannaccone, 1990) and consumption (Becker, 1996). Religious human capital is defined as knowledge, skill, experience and memories that enhance productivity in religious activities but have little or no effect on the productivity of resources allocated to other types of output (Iannaccone, 1990). Examples of investment in religious human capital
An Economic Perspective on Religious Education
451
include the study (formal and informal) of religious subjects, participation in religious ritual at home and in the church (or its analog for non-Christian religions), experiencing religious holidays and life cycle events, involvement in religious communal life and learning the language used for religious observance. It is not simply that the value (utility) of religious experience is sensitive to the level of religion-specific human capital applied to a given set of resources, for the content of the religious human capital embodied in a person in some sense defines that person’s religion (Chiswick, 1999). As such, it is central to religious experience and a crucial aspect of the intergenerational transmission of each religion (Chiswick & Chiswick, 2000). Education involves ideas as well as skills, and various studies suggest ways in which religious human capital raises the productivity of other types of human capital (Chiswick, 2001; Iannaccone, 1998; Lehrer & Chiswick, 1993).1 It has also been argued that religious education has a positive effect on the productivity of secular education, with positive complementarities between these two investment activities (Darnell & Sherkat, 1997; Hollander, Kahana, & Lecker, 2002). Researchers have paid less attention to the effects of secular human capital on religion, or secular education on religious education, which may be either positive or negative. A modern secular education generally increases the productivity of investments in other types of human capital and thus may be expected to have a similar effect on religious education. Yet religious education often appears to be structured around the notion that secular education leads to ideas and behaviors that are undesirable and that need to be counteracted or even destroyed (Sherkat & Darnell, 1999). Thus, while secular education may be a desirable investment overall, some of its components may have adverse effects on the productivity of religious training that offset positive complementarities and make the overall effect ambiguous. This paper focuses on how the curricula associated with secular and religious education, respectively, affect mutual complementarities in the production of these two types of human capital. Section 2 reviews the concept of religious human capital and presents a model of educational choices. Section 3 uses this model to discuss complementarity properties as they pertain to each type of education. Section 4 presents a preliminary test of the model with the experience with the American public school system of two religious groups, Protestants and Jews. Section 5 concludes with a discussion of some implications for the life of religious communities.
452
CARMEL U. CHISWICK
2. RELIGIOUS EDUCATION AND THE HUMAN CAPITAL PORTFOLIO The demand for religious education is derived from the production function for religious experience. The production function itself reflects a religious lifestyle choice, which is in turn affected by an individual’s tastes and preferences. Utility-maximizing consumers allocate their time between consumption and investment (education), and between religious and nonreligious activities. The problem can be expressed as Max UðY ; RÞ subject to LY þ LR þ LE ¼ L
(1)
where R is the religious good; Y all other consumption goods and services; LR the time spent in religious production, LY the time spent in nonreligious production, LE the total time spent in human-capital formation; and L* the total time available for all purposes. The two consumption goods, Y and R, are home-produced with production functions that depend primarily on human capital specific to each activity: Y ¼ fðhY LY Þ
(2)
R ¼ gðhR LR Þ
(3)
where hY is the level (quality) of general human capital and hR the level (quality) of religious human capital. The level of human capital thus enters the production function indirectly as a determinant of the total amount of human capital, HR hRLR or HY hYLY, which is (for simplicity) the sole input for producing the corresponding consumption good. Each type of human capital is in turn produced by an educational process with its own production function, the main input to which is the student’s time. These can be written inversely as cost functions, expressing the time cost of education as a function of the level of skill to be acquired. LE ¼ LYE þ LRE
(4)
LYE ¼ jðhY Þ j0 ; j00 40
(5)
LRE ¼ gðhR Þ þ ohY hR ;
g0 ; g00 40
(6)
where LYE is the time spent in nonreligious learning activities; LRE the time spent in religious learning activities; and the constant coefficient o indicates the degree to which the acquisition of general human capital imposes an external effect on religious education, for example, if o40 a greater level of
An Economic Perspective on Religious Education
453
general human capital (hY) would make it more costly to acquire any given level of religious education (hR), while if oo0 the opposite would be true. This problem is solved by maximizing the Lagrangian function: d ¼ UðgðhR LR Þ; fðhY LY ÞÞ
l½LR þ LY þ gðhR Þ þ jðhY Þ þ ohY hR
Ln (7)
Its first-order conditions can be solved to yield: U g g0 hR ¼ U f f 0 hY
(8)
LR =hR ¼ ½g0 þ ohY
(9)
LY =hY ¼ ½j0 þ ohR
(10)
L ¼ hR g0 þ hY j0 þ g þ j þ 3ohR hY
(11)
Eq. (8) equates the marginal rate of substitution in consumption between religious and nonreligious uses of time to –1, the slope of the time-budget line, requiring that the marginal value of time be the same in both consumption activities. Eqs. (9) and (10) equate the slopes of the human capital quantity–quality isoquants, LR/hR and LY/hY, respectively, to the marginal cost of the corresponding type of education, allocating time to each type of education up to the point where the marginal time required for an additional unit of human capital is the same as the opportunity cost of that time in consumption activities. Eq. (11), which expresses the time constraint as a function of the levels of the two types of education, is obtained by solving Eqs. (9) and (10) for LR and LY and substituting the result into the constraint in Eq. (1). Eqs. (9) and (10) may also be solved for hR and hY, the result substituted into Eq. (8) and terms rearranged to yield: 0 U g g0 L R g þ ohY ¼ 0 (12) U f f 0 LY j þ ohR The expression on the left-hand side of this equation is the marginal rate of substitution in consumption between hR and hY, the slope of an indifference curve between levels of the two types of education. The right-hand side is the slope of a production possibility frontier (PPF) that holds constant LE, the total resources devoted to education. Optimization thus requires tangency between an indifference curve and a human capital PPF determined by the allocation of time between consumption and education.
454
CARMEL U. CHISWICK
By varying the amount of time devoted to education, Eq. (12) implies an expansion path with a positive slope as long as both hY and hR are normal (in the sense that more resources devoted to education raises the demand for each type). In contrast, the boundary in Eq. (11) describes a single opportunity set where the overall constraint on time has been converted (by means of the education-production functions) into an equivalent constraint on the attainable combinations of human capital. This constraint generally has a negative slope, for which a sufficient condition is oX0 (i.e., that any external effects of general education on religious education be non-positive). The overall solution to the consumer’s problem occurs where the time constraint crosses the expansion path either at a unique combination of hR and hY or at one of its corners. These relationships are illustrated in Fig. 1. Religious education, hR, is measured on the horizontal axis and all other education, hY, on the vertical axis. A family of PPF curves depicts the maximum combinations of human capital attainable from various levels of investment in education, each curve representing a different amount of LE. If each type of investment is subject to diminishing marginal productivity, the PPF will be concave to the origin (i.e., bowed outward), and Fig. 1 illustrates the case where there are no
U2
hY
Expansion path
U1
All Other Education
A
U0
Total Time Constraint
PPF for L E fixed
hR
O Religious Education
Fig. 1.
B
Optimal Investment in Education without Externalities.
An Economic Perspective on Religious Education
455
externalities to alter this property. A family of indifference curves reflects the utility attainable (indirectly from the own-production process) from various combinations of human capital, and the points of tangency between indifference and PPF curves describe an expansion path. The heavy line with a negative slope is the time constraint from Eq. (11). The consumer’s optimum occurs where the expansion path crosses the time-constraint boundary.
3. THE EFFECTS OF EXTERNALITIES ON RELIGIOUS EDUCATION Supply-side complementarity between investments in the two types of human capital, described by the sign and magnitude of the parameter o, affects the shape of the opportunity set and the PPF curves. If secular education confers positive externalities on religious education (oo0), these curves would be bowed even further outward and the optimal investment for both types of human capital would be greater. If the two types of human capital have few complementarities and/or negative externalities (o40), the family of PPF curves would be less concave and the optimal portfolio would not only be smaller but would also display a greater tendency toward specialization in investment. Fig. 2 illustrates the implications of negative externalities for decisions about religious education. The two axes represent different types of human capital religious and nonreligious. The curve ACB is the PPF in the absence of externalities and point C, where the indifference map is tangent to the PPF, describes the optimal allocation of investment in education for this case. Negative externalities would reduce the concavity of the PPF and if large enough can even cause it to bow inward, as illustrated by the curve AEB. The optimal resource allocation in this (admittedly extreme) case is at point E. Negative externalities have two effects on the optimal allocation of resources. Because they increase the costs of education there is a scale effect, reducing the total amount of human capital attainable from fixed resources and hence the area under both the PPF and the overall time constraint. Negative externalities also induce a substitution effect, increasing the incentive to concentrate resources in one or the other type of human capital because the externalities raise the cost of combining the two. For example, an expansion path passing through the optimum at point E in Fig. 2
456
CARMEL U. CHISWICK Negative externalities expansion path
hY
No-externalities expansion path
All Other Education
A
C
E
O
B Religious Education
Fig. 2.
hR
Optimal Investment with Negative Supply-side Externalities.
(corresponding to the PPF family to which AEB belongs) would be everywhere to the left of the nonexternalities expansion path passing through point C. Since externalities have a similar effect on the shape of the overall time constraint, altering it less near the corners than in the interior, the consumer’s optimum will be lower on the expansion path the greater the negative externalities. Fig. 3 further illustrates the substitution effect of an increase in negative supply side externalities on specialization in education. As in Fig. 2, the outwardly bowed PPF (curve ACC0 B) represents the case without supply side externalities in education while the inner PPF (curve AEE0 B) represents the case where there are large negative externalities. Fig. 3 displays two indifference maps corresponding to two different people with different productivity in the production of religious experience. For example, the solid curves might apply to a person living in a weak religious community and the dashed curves for a person in a strong, vibrant one. Alternatively, the solid curves might represent a person with a talent for scientific analysis and the dashed curves a person more inclined toward the theological or mystical. While there will always be differences in the educational outcomes chosen by these two people, these differences will be far greater in the presence of negative externalities. The expansion path corresponding to point E lies to
457
An Economic Perspective on Religious Education No-externalities expansion paths
hy
All Other Education
A
C E C’
E’ O
B
hR
Religious Education
Fig. 3.
Preferences and Negative Supply-side Externalities.
the left of the expansion path for point C, while the path for E0 lies to the right of the one for C0 . Geometrically, the more outwardly bowed (convex) the PPF, the closer the expansion paths for people with different preferences. Economically, this occurs because positive externalities reduce the cost of combining the two types of education, while negative externalities have the opposite effect. Fig. 4 illustrates the effect of an exogenous educational standard imposed by law or custom, for example, a compulsory schooling law that requires all children to receive at least C* amount of secular education. By effectively eliminating the area under C* (shaded) from the opportunity set, the constraint on general education induces substitution away from religious education. In the absence of externalities, the PPF would be ACC0 B; people with the solid-line indifference curves (tangent at point C) would be unaffected but people with dashed-line indifference curves would be worse off, moving from their unconstrained optimum at C0 to C00 by substituting secular education for religious education. If there are negative externalities, the PPF would be AEE0 B and a constraint at C* would have an even more dramatic effect on the size of the feasible area, increasing both the likelihood and magnitude of adverse effects. People with the solid-line indifference curves would find a new optimum at E00 , a corner solution with somewhat
458
CARMEL U. CHISWICK
hY A All Other Education C C*
E” C”
C’
E
E’ O
B
hR
Religious Education
Fig. 4.
Discontinuity Constraints on Optimal Investment.
less utility than at the unconstrained optimum but with a much-reduced level of religious education. Those with the dashed-line indifference curves might also optimize at E00 but would experience a much greater reduction in utility and would prefer the corner solution at B. The illustrations thus far have assumed smooth production functions, yet education is often organized as a series of levels, or degrees, reflecting discontinuities in the underlying process. This characteristic suggests that one or both of the education production functions, f(hY) for general education or g(hR) for religious education, might be written as a step function. Fig. 4 can be used to illustrate the simple case where f(hY) has a threshold at C* below which there is little or no value to general education. In contrast to the compulsory-schooling situation where outcomes in the shaded area are prohibited by law, here the constraint is inherent in the production function itself and outcomes in the shaded area represent wasted investments in general education. People with the dashed-line indifference curves would optimize not at E00 but at B, opting out of general education entirely and specializing in religious schooling. This model suggests a series of hypotheses relating religious beliefs with nonreligious educational attainment. Among those groups for which
An Economic Perspective on Religious Education
459
religious beliefs are either neutral or positively complementary with a secular educational curriculum, nonreligious educational attainment should be relatively high and similar across groups, without prejudice to religious education or degrees of religiosity. Among those groups for which religious beliefs are characterized by negative complementarities with a secular curriculum there should be a greater degree of specialization. Adherents with strong religious attachments are hypothesized to have lower educational attainment, while those with high educational attainment should exhibit a greater degree of religious skepticism and lower religiosity. Finally, groups for which negative complementarities are very strong should be most likely to opt out of secular schooling entirely, concentrating in occupations for which educational requirements are minimal.
4. SOME EVIDENCE FROM THE AMERICAN SCHOOL SYSTEM The United States is a religiously pluralistic society in which public schools purport to be decidedly nonsectarian, if not completely secular. Yet inevitably some religious groups find their teachings to be less complementary than others. During the late 19th and early 20th centuries, when urban public schools were taking on their modern form (i.e., graded classrooms from kindergarten through high school), their content and culture complemented the mainstream Protestant denominations whose members dominated the educational establishment. Fundamentalist Protestant teachings were less complementary with the public school curriculum, however, and these groups were more likely to develop patterns of nonattendance, with high truancy rates and low-educational attainment. The Catholic Church set up a completely separate alternative school system for their children, while Jewish parents sent their children to public schools but invested in afterschool programs for religious education that reinforced a family-based Jewish ‘‘counterculture.’’ By the end of the 20th century, American public schools had become more self-consciously secular, in part as a response to parental pressures as upwardly mobile religious minorities sought entry into a knowledge-based economy where high-level skills were at a premium. Even so, some religious groups benefited (in the sense of lower negative complementarities) more than others (Sikkink, 1999). As a test of the hypotheses developed above, this section considers some alternative strategies chosen by religious groups
460
CARMEL U. CHISWICK
for whom negative complementarities with public schools are especially pronounced. Section 4.1 focuses on Protestants, looking at evidence that denominations differ in their behaviors according to the degree to which public schools are perceived to be compatible with their religious values. Section 4.2 considers the case of American Jewry, a very small religious minority that sends most of its children to public schools and has developed a variety of educational strategies to counteract negative complementarities.
4.1. Education and Protestant Denominations Cultural conflict in the public schools (reflecting that in American society at large) has shifted from its historical focus on the divide between Catholic and Protestant belief systems to its current focus on the divide between the ‘‘sacred’’ and ‘‘secular’’ (Sikkink, 1999). Among other questions, the Religious Identity and Influence Survey (fielded in January–March 1996) asked: ‘‘In general, do you think that public schools are hostile to your moral and spiritual values?’’2 Some 70 percent of the Pentecostals and 62 percent of the Charismatics answered yes, as did 55 percent of Evangelicals and 51 percent of Fundamentalists. In contrast, among Mainline Protestants only 39 percent answered yes, as did 31 percent of Liberal Protestants and 25 percent of those who were ‘‘not religious.’’3 This ranking of Protestant religious groups with regard to perceived negative complementarities is consistent with their rankings by educational attainment. Controlling for parents’ income, education and occupation as well as the gender and race of respondents, members of ‘‘Conservative Protestant’’ denominations have lower levels of educational attainment than do ‘‘Mainline Protestants,’’ and a similar (additional) negative effect is observed for respondents who believed that ‘‘The Bible is God’s word and all it says is true’’ (Darnell & Sherkat, 1997; Sherkat & Darnell, 1999; Lehrer, 2004).4 One study that considers the beliefs of students as well as their parents finds that Fundamentalist parents generally inhibit their children’s secular schooling but that they are less restrictive for children who themselves express strong religious beliefs (Sherkat & Darnell, 1999). This suggests that the parents’ concerns about secular schooling are indeed related to negative complementarities since they are willing to educate children with high levels of religious human capital (thus helping them move outward along an expansion path) but are reluctant to help them substitute general for religious education.
An Economic Perspective on Religious Education
461
Religious groups with the greatest negative complementarities might be expected to opt out of the public school system entirely. Some of the Protestant dissenters of an earlier era had a tradition of avoidance and (where compulsory schooling laws are binding) truancy and those religious groups were characterized by low educational attainment and consequently high rates of poverty (Sikkink, 1999). Other groups established religion-based day schools, a practice that was once dominated by the Catholic school system but has become more diversified in recent decades. There is also a growing ‘‘home schooling’’ movement, where parents take responsibility for educating their own children (Bauman, 2001). Among parents who choose home schooling for their children, the two most frequent reasons given are that their children will receive a ‘‘better education’’ at home (51 percent) and ‘‘religious reasons’’ (33 percent).5 Some of these complaints about the poor quality of public schools may also arise from religious considerations, whether from the perceived absence of religious teaching or from too much religious teaching of the ‘‘wrong’’ sort. This is explicit for the 14 percent of parents in the survey who say they ‘‘object to what school teaches,’’ and probably also for the 9 percent who say they chose home schooling in order ‘‘to develop character/morality’’ in their children.
4.2. Education and American Jewry Jews constitute less than 2 percent of the US population in 2000, down from a peak of 3.7 percent in 1940, and as such have little political influence on public schools.6 Yet Judaism has a very long history of being a minority religion, often in societies that are more or less hostile, and Jewish religious education takes this as its point of departure. Indeed, an important function of American-Jewish religious education is to help students recognize and be skeptical of incompatible religious teachings in their public schools without sacrificing the religion-neutral knowledge to which those schools give access. Jewish immigrants to America came primarily from Tsarist Russia and the countries of Eastern Europe in which religious hostility toward Judaism was deep and explicit. In such a setting, the negative complementarities between religious and secular education would have been pronounced, and Jews were likely to have specialized by emphasizing one or the other. Those who immigrated to America tended to be self-selected for people with strong preferences for secular achievement relative to religious observance, suggesting an indifference map (and therefore an expansion path) resting somewhat to the left of the old-country average. High career aspirations
462
CARMEL U. CHISWICK
would have the same effect and compulsory-schooling laws – if binding – would further favor a substitution of general for religious schooling. The empirical evidence supports these hypotheses, suggesting that American Jewish immigrants emphasized a high level of secular achievement in the public schools for their children and either neglected their formal Jewish education or provided one that was little more than perfunctory (Sarna, 2004). America was a land of democracy and religious pluralism, a society whose values were enthusiastically embraced as compatible with Jewish tradition. Judaism is nevertheless a relatively human capital-intensive religion that requires substantial investment in any environment. Many immigrants underestimated the importance of Jewish education – both formal and informal – for producing Jewish experience in the new country, taking for granted the formation of Jewish human capital attained so inexpensively in an isolated old-country community. For the community as a whole, low levels of Jewish investment by its members would generate mutually reinforcing reverse bandwagon effects that would shift their expansion paths even further to the left (Chiswick & Chiswick, 2000). One important Jewish response to this ‘‘threat’’ was to work to make American society more hospitable to Jews and thus reduce the negative externalities between Jewish and general education (Dinar, 2004). The AntiDefamation League was organized for this purpose, and many Jewish communal organizations had units focusing on political action or education of the general public, as appropriate. Interfaith activities also received support at all levels, based on the belief that anti-Semitism was born of ignorance and could be eroded by friendly relations between Jews and non-Jews (Silberman, 1985). The hypothesized effects of these activities are illustrated in Fig. 5 as a reduction in the convexity of the PPF, moving the optimum human capital portfolio from point E to F. Another popular response was to modify Judaism itself to reduce its dependence on specifically Jewish human capital. Secular Judaism has many variations, none of which require much in the way of Jewish education, and was especially popular in the early part of the 20th century. The early Reform synagogue movement emphasized Jewish ethics and universal values over ritual, discarding the use of Hebrew for prayer, the dietary laws, and much of the ‘‘parochial’’ in Jewish holiday observance. (Many of these ‘‘reforms’’ have since been reversed.) By being selective in their definition of Judaism, the adherents of these modern variants could take pride in a heritage that was fully harmonious with American ideals, confident that Jewish parochialism was an old-country aberration and therefore an embarrassment
463
An Economic Perspective on Religious Education
General Human Capital
A
F’
F E
O
B Jewish Human Capital
Fig. 5.
Optimal Investment with Adaptations in Jewish Practice.
undermining their American status (Wertheimer, 1993). The economic effect of this strategy was to alter the production function so as to reduce its reliance on specifically Jewish education, illustrated in Fig. 5 as an implicit movement from F to F 0 : A third response involved revamping the curriculum of Jewish schools so as to enhance as much as feasible the complementarity between Jewish and general education. American Jewish children would learn to read and write English in the public schools before beginning their Hebrew studies, so English literacy could be assumed in the Jewish schools. Like the public schools, Jewish education was organized by age with each grade meeting in a separate classroom and taught from its own textbook. Hebrew schools run by the Conservative synagogue movement typically held class meetings after school and on Sunday mornings, while the Orthodox ran day schools that taught the general curriculum as well as Jewish studies. (Later generations would greatly expand the Jewish day schools even for non-Orthodox Jewish children.) Textbook and curriculum development for schooling in Hebrew, Jewish history, Torah and holiday observance continued to assume basic skills acquired in the course of a student’s general schooling (Wertheimer, 1999). While negative supply side externalities invariably remained important, they could be partially offset by the integration of Jewish and general studies.
464
CARMEL U. CHISWICK
5. IMPLICATIONS FOR RELIGIOUS-GROUP BEHAVIOR PATTERNS The model developed and applied in this paper focuses on the tradeoffs between religious and general education from the perspective of a utilitymaximizing individual. The outcome of such a decision has important social implications. Negative complementarities between religious and general education can lead to social fragmentation both within and between groups. Within a religious community, they provide incentives toward polarization, encouraging a division between committed adherents with low general education (and hence earning power) on the one hand and a group with greater economic success but relatively little religious human capital on the other. This specialization can also fragment society at large, generating an inverse relationship between religiosity and income and further increasing differences among people with different religions in both behavior and attitudes. For intergenerational continuity, a religious community requires that its adherents invest in some minimum threshold of religion-specific human capital. Religious human capital is essential for intergenerational continuity in two respects. Since it is essentially what distinguishes one religion from another, the religious human capital acquired by each generation affects its ability to transmit religion to the next and there is presumably some minimum threshold below which an individual can no longer perform this function. Demographically religious human capital is, like most other forms of human capital, a homogamous marital trait (i.e., one for which ‘‘like marries like’’) (Becker, 1981).7 The less religious human capital a person brings to the marriage market, the greater the probability that religion is outweighed by other attributes of a potential spouse and the higher the probability of religious intermarriage. Whether the spouses have the same or different religions, however, a couple with less religious human capital would be relatively less efficient both as consumers of religion and as religious educators in the home. Without communal support, their children are likely to become adults with even lower levels of religious human capital. Educational choices made by individuals may thus have important implications for the group as a whole, giving rise to the quasi-public good aspect of religion that has been explored elsewhere (Iannaccone, 1992). A religious group would be especially vulnerable to outcomes in which a significant part of its membership chooses a low level of religious human capital, regardless of whether – or perhaps especially if – they form a subgroup with high secular education. The analysis in this paper suggests that it
An Economic Perspective on Religious Education
465
is not the acquisition of general human capital that would matter for religious continuity but rather the extreme reduction in religion-specific human capital. Assimilation strategies that neglect religious education may not only weaken the religious attachment of individuals, but by generating reverse bandwagon effects they may also weaken the entire religious community. Individual choices that tend to concentrate on general rather than religious education are often attributed to changes in preferences in favor of a nonreligious ‘‘secularism.’’ The analysis developed above suggests that secularism itself may be endogenous, the outcome of a time allocation problem in which religious human capital is relatively costly to acquire. Small differences in preferences that might not have much effect on educational attainment if the public school curriculum were neutral can lead to very large group differences if there are negative complementarities between religious and general education. The challenge to public education in a religiously pluralistic society is to minimize these adverse effects for as many groups as possible without sacrificing the substance of the general curriculum. In the early part of the 20th century, many Americans viewed ‘‘assimilation’’ as a high-priority goal and immigrants invested heavily in general American human capital, ‘‘Americanizing’’ their religious practices as well. By the later part of the 20th century, their grandchildren and great-grandchildren could be divided into those that felt that perhaps ‘‘assimilation’’ was bad – or, perhaps more accurately, that it was ‘‘too much of a good thing’’ – and those who were so assimilated that they didn’t care. The public discussion of this issue typically focuses on ethnic ‘‘multiculturalism,’’ but the phenomenon is clearly relevant for religious pluralism. For any religious group concerned with the transmission of adherence from one generation to the next, finding the right balance between assimilation and group cohesion is a fundamental issue. For a public school system concerned with the transmission of general knowledge from one generation to the next, finding the right balance between the sacred and the secular is similarly fundamental for long-run stability.
NOTES 1. Complementarities among various types of human capital are widely recognized as important. Health capital, for example, has been shown to raise the productivity of investments in schooling, while at the same time education raises the productivity of investments in health. See, for example, Schultz (1993).
466
CARMEL U. CHISWICK
2. The data in this paragraph are from Sikkink (1999). While the figures reported here are simple response rates, Sikkink finds that the ranking of religious groups by their hostility toward public schools remains unchanged after controlling for group differences in family income, education, and number of children. 3. Only 22 percent of churchgoing Catholics, 15 percent of nominal Catholics and 20 percent of persons with non-Christian religions responded yes to this question. Whether this is due to a ‘‘secularization’’ of these religions or simply different expectations from public schools will be discussed in the next section. 4. This question on Biblical inerrancy was presented to respondents of all religions as one of four possible responses. More than one-third responded affirmatively. 5. More than one reason could be given. The data in this paragraph are from the National Household Education Survey of 1996 and 1999, as reported by Bauman in Bauman (2001). 6. During the earlier part of the century, first- and second-generation Jewish immigrants tended to concentrate in a few areas, notably New York City, where they were able to have some influence on the public schools. Elsewhere, however, their numbers were far too small to be felt as a political force. 7. That is, people with high levels of religious human capital tend to select spouses who also have high levels for the same religion, forming family units for which the home production of religious education is more efficient (Chiswick, 1998). Conversely, people with low levels of religious human capital are inefficient producers of religion and tend to marry others who are similarly inefficient.
ACKNOWLEDGMENTS The author wishes to thank especially Barry R. Chiswick and Solomon Polachek for their encouragement and suggestions. Any remaining errors, of course, are the full responsibility of the author.
REFERENCES Bauman, K. J. (2001). Home schooling in the United States: Trends and characteristics. Washington, DC: Population Division, US Bureau of the Census. Becker, G. S. (1981). A treatise on the family. Cambridge, Mass.: Harvard University Press. Becker, G. S. (1996). Accounting for tastes. Cambridge, Mass.: Harvard University Press. Chiswick, B. R., & Chiswick, C. U. (2000). The cost of living Jewishly and Jewish continuity. Contemporary Jewry, 21, 78–90. Chiswick, B. R., & Miller, P. (1995). The endogeneity between language and earnings: International analyses. Journal of Labor Economics, 13, 246–288. Chiswick, C. U. (1998). The economics of contemporary American Jewish family life. In: P. Y. Medding (Ed.), Studies in contemporary Jewry 14: Coping with life and death: Jewish families in the twentieth century (pp. 65–80). New York: Oxford University Press. Chiswick, C. U. (1999). The economics of Jewish continuity. Contemporary Jewry, 20, 30–56.
An Economic Perspective on Religious Education
467
Chiswick, C.U. (2001). The economics of Jewish immigrants and Judaism in the United States. Papers in Jewish Demography, 1997, 331–344. Darnell, A., & Sherkat, D. E. (1997). The impact of Protestant fundamentalism on educational attainment. American Sociological Review, 62(2), 306–315. Dinar, H. (2004). Multiple outsiderness: Religious, ethnic, and racial diversity in America. Contemporary Jewry, 24, 29–50. Filer, R. (1990). Arts and academe: The effect of education on earnings of artists. Journal of Cultural Economics, 14(December), 15–38. Hollander, G., Kahana, N. & Lecker, T. (2002). Human capital and the economics of religion. In: C. Chiswick, T. Lecker & N. Kahana (Eds), The Economics of Judaism and Jewish human capital. Ramat Gan, Israel: Bar Ilan University Press (forthcoming). Iannaccone, L. R. (1990). Religious practice: A human capital approach. Journal for the Scientific Study of Religion, 29(3), 297–314. Iannaccone, L. R. (1992). Sacrifice and stigma: Reducing free-riding in cults, communes, and other collectives. Journal of Political Economy, 100(2), 271–291. Iannaccone, L. R. (1998). Introduction to the Economics of religion. Journal of Economic Literature, 36, 1465–1496. Lehrer, E. L. (2004). Religion as a determinant of economic and demographic behavior in the United States. Population and Development Review, 30, 707–726. Lehrer, E. L., & Chiswick, C. U. (1993). Religion as a determinant of marital stability. Demography, 30(3), 385–404. Mincer, J. (1974). Schooling, experience and earnings. New York: Columbia University Press for the National Bureau of Economic Research. Mincer, J. (1984). Human capital and economic growth. Economics of Education Review, 3, 195–205. Sarna, J. D. (2004). New paradigms for the study of American Jewish life. Contemporary Jewry, 28, 157–169. Schultz, T. P. (1981). Economics of population. Englewood Cliffs, NJ: Prentice Hall. Schultz, T. P. (1993). Investments in the schooling and health of women and men: Quantities and returns. Journal of Human Resources, 28(4), 694–734. Schultz, T. W. (1980). Nobel lecture: The Economics of being poor. Journal of Political Economy, 88, 639–651. Sherkat, D. E., & Darnell, A. (1999). The effect of parents’ fundamentalism on children’s educational attainment: Examining differences by gender and children’s fundamentalism. Journal for the Scientific Study of Religion, 38, 23–35. Sikkink, D. (1999). The social sources of alienation from public schools. Social Forces, 78, 51– 86. Silberman, C. E. (1985). A certain people: American Jews and their lives today. New York: Summit Books. Smith, T.M. (1998). Two essays on the Economics of the arts: The demand for art and the occupational mobility of artists. Economics. Chicago: University of Illinois Press. Wertheimer, J. (1993). A people divided: Judaism in contemporary America. New York: Basic Books. Wertheimer, J. (1999). Jewish education in the United States: Recent trends and issues. In: D. Singer (Ed.), American Jewish year book 1999 (pp. 3–115). New York: American Jewish Committee.
This page intentionally left blank
468
