^~~~~-I A An m ... . - A -w I 1 ! r 5 , V -l - U , PCrCINTCV\I4TO R1T`X7TT1?AT - WWW- Wwi 0i y %.w m In w =w W wRWai Volume 11 January 1.997 Number 1 -A- SYMPOSIUM ISSUE ON How INTERNATIONAL EXCHANGE, TECHNOLOGY, AND INSTITUTIONS AFFECT WORKERS .How International Ekchange, Technology and Institutions Affect Workers. An introauction Ishac Diwan and Michael Walton Learning by Trading and the Returns to HumanmCapital in Developing Countries Christopher A. Pissarides' Openness and Wage Inequality in Developing Countries:: I ne Latin Anmcrican %_nanienge io Easr Asian uonventiona wisAdom- Adrian Wood _ ~~~~~~~~~~~~~~-,ad Fir Si -W....-. a Technohogy-and Firm Size Wage Differentials in Colombia, ' 'Mexico, and Taiwan (China) -. _''Hong Tan and Geeta^Batra I neBeiiects oi'ruuOic aecor:t0 ir- ing ana compensation roacies on the Egyptian Labor Market -The.Impact of Labor Market- Regulations- - Lyn Squire andn ethaput-Suthiwiart-Narueput _ .-, ... . L~~~~1c0numlic Mauro ud LaruLuo U uiuus ..- A General-Equilibrium Analysis,Applied to Bangladesh and Indonesia.-, Does the Type;of Political Regime Matter for Trade and Labor Market Policies? -, :- ~:: - - '-- ' ,,- Ar'up Banerji and Hafez G- FF 17r-D f|I6fFM v7 HL9bLu I FrrurWTST Tlf xr r_ _ * ' Yr _r 1 T W UKLLI bAIN K ECONOMIC REVIEW EDITOR MseSyrquin CONSULTING EDITOR Sandra Gain EDITORIAL BOARD Kaushik Basu, Cornell University and University of Delhi David Dollar Francois Bourguignon, DELTA, Paris Gregorv K Ingram Willem H. Buiter, University of Cambridge John Page Gerald K. Helleiner, University of Toronto Lant H. Pritchett Mark R. Rosenzweig, University of Pennsylvania Jacques van der Gaag L. Alan Winters The World Bank Economic Revieiv is a professional journal for the dissemination of World Bank- sponsored research that informs policy analyses and choices. ir is directed to an internationai readership among economists and social scientists in government, business, and international agencies, as well as in universities and development research institutions. The Review emphasizes policy relevance and opera- tional aspects of ecnmis rather than p'rimar!ly theoretical and methodological Issues. It:s intended for readers familiar with economic theory and analysis but not necessarily proficient in advanced mathematical or econometric techniques. Articles will illustrate how professional research can shed light on policy choices. Inconsistency with Bank policy will not be grounds for reiection of an article. Articles will be drawn primarily from work conducted by World Bank staff and consultants. Before being accepted for publication by the Editorial Board, all articles are reviewed by two referees who are not mem- bers of the Bank's staff and one World Bank staff member; articles must also be recommended by at least one external member of the Editorial Board. The Review may on occasion publish articles on specified topics by non-Bank contributors. Any reader interested in preparing such an article is invited to submit a proposal of not more than two pages in length to theE'I t -tor. The views and interpretations expressed in this journal are those of the authors and do not necessarily represent the views and policies of the World Bank or of its Executive Directors or the countries they rep- resent. The World Baok dne.s notr unaranree the accuracy of the data i-nc-lded ti publication and accepts no responsibility whatsoever for any consequences of their use. When maps are used, the bound- aries, denominations, and other information do not imply on the part of the World Bank Group any judg- ment on the legal status of any territory or the endorsement or acceptance of such boundaries. Comments or brief notes responding to Review articles are welcome and will be considered for publica- tion to the extent that space permits. Please direct all editorial correspondence to the Editor, The World Bank Economic Review, The World Bank, Washington, D.C. 20433, U.S.A. The World Bank Economic Review is published three times a year (January, May, and September) by the World Bank. Single copies may be purchased at $12.95. Subscription rates are as follows: Individoals Institutions I-year subscription US$30 US$50 2-year subscription US$5S TISt9C 3-year subscription US$70 US$130 Orders should be sent to: World Bank Publications. Box 7247-7956 Philadelohia. PA 19170-7956. U.S.A. Subscriptions are available without charge to readers with mailing addfesses in developing countries and in socialist economies in transition. Written request is required every two years to renew such subscriptions. (C 1997 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, N.W., Washington, D.C. 20433, U.S.A. All1 rights reserve' Manufactured in the United States of America IsBN 0-8213-3781-5; ISSN 0258-6770 Material in this journal is copyrighted. Requests to reproduce portions of it should be sent to the Office of the Publisher at the address in the copyright notice above. The World Bank encourages disse'mination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to make photocopies is granted through the Copyright Clear- ance Center,Suite 910,222 Rosewood Drive, Danvers, MA 01923, U.S.A. This journal is indexed regularly in Current Contents/Social & Behavioral Sciences, Index to Interna- tional Statistics, journal of Economic Literatuire, Public Affairs Information Service, and Social Sciences Citation IndexO. It is available in microform through University Microfilms, Inc., 300 North Zeeb Road, Ann Arbor, MI 48106, U.S.A. THE WORLD BANK ECONOMIC REVIEW Volume 11 January 1997 Number 1 A SYMPOSIUM ISSUE ON How INTERNATIONAL EXCHANGE, TECHNOLOGY, AND INSTITUTIONS AFFECT WORKERS THIS SYMPOSIUM ISSUE draws on papers originally prepared as background mate- rial for World Development Report 1995: Workers in an Integrating World. The articles in the symposium were refereed in the usual way. The Editorial Board invited Ishac Diwan, who was a member of the team that prepared the report, and Michael Walton, who led the team, to write an introduction to the symposium. How International Exchange, Technology, and Institutions 1 Affect Workers: An Introduction Ishac Diwan and Michael Walton Learning by Trading and the Returns to Human Capital 17 in Developing Countries Christopher A. Pissarides Openness and Wage Inequality in Developing Countries: 33 The Latin American Challenge to East Asian Conventional Wisdom Adrian Wood Technology and Firm Size-Wage Differentials in Colombia, 59 Mexico, and Taiwan (China) Hong Tan and Geeta Batra The Effects of Public Sector Hiring and Compensation Policies 85 on the Egyptian Labor Market Ragui Assaad The Impact of Labor Market Regulations 119 Lyn Squire and Sethaput Suthiwart-Narueput Economic Reform and Labor Unions: A General-Equilibrium 145 Analysis Applied to Bangladesh and Indonesia Shantayanan Devarajan, Hafez Ghanem, and Karen Thierfelder Does the Type of Political Regime Matter for Trade and Labor 171 Market Policies? Arup Banerji and Hafez Ghanem How International Exchange, Technology, and Institutions Affect Workers: An Introduction Ishac Diwan and Michael Walton In a world of deepening trade links, rapid technological change, and weakening institutions, workers in rich and poor countries alike are concerned about their in- comes and the security of their work. In contrast to the substantial quantity of analysis on industrial countries, relatively little careful work has been done on these issues in developing countries, especially in the context of the recent globalization of eco- nomic relations. Empirical work suggests that disequalizing trends in some develop- ing countries may have been caused by the entry of low-income countries such as China into world production or by the greater quality and technological require- ments of contemporary trade in goods. Whatever the source, these results raise ques- tions about the viability of institutional mechanisms for supporting the income secu- rity and working conditions of workers. Many labor market regulations are already ineffective owing to weak enforcement capabilities. However, in most developing countries, there will be a rising fraction of workers in formal labor contracts, rising demands for formal mechanisms for dealing with income insecurity, and a poten- tially larger role for unions in an important segment of the workforce. Thus, it is of increasing importance to set the policy and institutional framework in a fashion that is both consistent with competitive pressures and supportive of workers' participa- tion and security. Rising wage disparities in the United States, persistent unemployment in Eu- rope, and "hollowing out" (relocation abroad of production jobs) of employ- ment structures in Japan reinforce workers' concerns about international economic relations and domestic institutional changes. Institutional support systems for workers-from European-style welfare states to Japanese-style firm- based security-are being forced to adapt to international competition and the shifting labor demands of new technologies. These workers' concerns have par- allels in developing countries, where state-directed employment creation- whether through protection for industrial jobs or central planning-has been ineffectual. New concerns over employment in developing countries resonate with two long-standing employment debates: the debate about the consequences of open- Ishac Diwan is with the Economic Development Institute, and Michael Walton is with the East Asia and Pacific Region, both at the World Bank. The authors wish to thank Martin Rama and Adrian Wood for their comments. This symposium issue is made up of articles that were originally prepared as background papers for World Development Report 1995: Workers in an Integrating World. D 1997 The International Bank for Reconstruction and Development /THE WORLD BANK 1 2 THE WORLD BANK ECONOMIC REVIEW, VOL. II, NO. I ness for workers and the debate about the kind of regulatory and institutional framework that can support the security and working conditions of workers in a labor market usually dominated by rural and informal workers outside the ambit of formal rules. These old issues have acquired fresh topicality for several reasons. First, al- though some talk of globalization is overblown, the circumstances of the 1990s differ from those of previous decades for the developing world. With the fall in protectionism and the demise of central planning, the proportion of the world's workers who are shielded from the influence of international economic forces likely will drop from around two-thirds in the late-1970s to about one-tenth by the end of the 1990s (see World Bank 1995b). Second, evidence from some coun- tries, especially in Latin America, indicates that wage inequalities increased over the past decade, coinciding with rising openness. Third, workers in many coun- tries perceive as a threat the entry of China and other low-income countries into international trade. Fourth, increasingly, people in rich countries worry about the conditions of work in poor countries and are afraid that international ex- change will force all workers into insecure work in unhealthy conditions. And fifth, within countries, workers and policymakers debate about the role of labor regulations, unions, and public employment policy in a world of rising intercon- nections and apparently rapid structural change. The employment debates are rarely informed by careful analysis. The articles in this issue seek to address some of the open questions about the influence of international and institutional developments on labor incomes. The articles have two points of departure. One focuses on technology, international transactions, and workers' skills. The other highlights the effects of domestic institutions- regulations, government employment policies, unions, and political structures- on the labor market. This introductory article surveys some of the important issues for both cur- rent policy and future work. Section I discusses issues relating to international transactions, technology, and skills. Section II discusses the role of domestic institutions and regulations in an internationally integrated environment. Sec- tion III concludes. I. INEQUALITY, TECHNOLOGY, AND INTERNATIONAL TRANSACTIONS The fear that development goes hand in hand with increased inequality-the so-called Kuznets curve-has until recently consistently been rejected by careful empirical evidence (see, for example, Little 1982; Deininger and Squire 1996; and Bruno, Ravallion, and Squire 1996). Indeed, it has become more common in development circles to argue that development can be equalizing. For example, World Bank (1993) shows a positive link between growth and more equal in- come distribution in East Asia. In addition, others have also argued that and have presented empirical evidence supporting the claim that (a reasonable de- gree of) equality is good for development (Alesina and Rodrik 1994; Perrson Diwan and Walton 3 and Tabellini 1995). We are concerned here with an alternative view: that some dimensions of inequality are rising and that this may be associated with aspects of international integration-through trade in goods, flows of capital, or migra- tion-and with technological change. Many economists believe that as countries become more open to trade, in- equality falls in relatively poor countries with abundant unskilled labor and rises in their richer trade partners. This view-developed by Krueger (1978 and 1983) and more recently by Wood (1994)-makes perfect Heckscher-Ohlin sense. It also seems consistent with a set of stylized observations: (a) that as they became more open to external exchange, most industrial countries have suffered an ad- verse shock to the relative demand for unskilled labor in the past twenty years (Wood 1994); (b) that over the same period, inequality fell in the fast-growing, export-oriented Asian countries (Wood 1994); and (c) that during the globaliza- tion episode that also took place in the late nineteenth century, inequality fell in the labor-abundant countries of the Old World and rose in the resource-rich countries of the New World (Williamson 1996). International migration and flows of capital potentially also affect wage in- equality in poorer economies. Outmigration tends to tighten local labor mar- kets, and the inflow of capital raises the real wage and hastens the labor market transition to industrial and service-based work. Although the poorest workers generally do not migrate, and even capital associated with unskilled work may directly employ only workers with basic education, virtuous cycles tend to de- velop, with international flows helping to equalize growth. With respect to the late nineteenth century, Williamson (1996) finds that migration was more im- portant than trade in explaining rising inequality in the New World and falling inequality in those parts of the Old World that were rapidly integrating. In trying to understand the empirical evidence, it is important to take into consideration changes in the relative supply of different categories of workers, since this also influences labor market outcomes. To the extent development is associated with increased education, the return to skills should fall on this ac- count. For societies starting with low levels of education, the expansion of skills plays a powerful equalizing role. Indeed, in the United States, the rising relative supply of educated workers underpinned the "great compression" of relative wages that occurred in the 1940s and masked the disequalizing shifts in demand until about the mid-1970s (Goldin and Margo 1991; Katz and Murphy 1992). In a recent study of the Korean experience (Kim and Topel 1995), a large expan- sion in the relative supply of education was associated with a large decline in wage inequality. The study also reveals the mechanisms whereby sharply chang- ing structures of labor demand translated into declining inequality in an initially agrarian society. The growth of manufacturing was accompanied by a whole- sale shift of employment out of agriculture. The manufacturing sector attracted new entrants to the labor force: young educated urbanites, who stayed in this sector throughout their careers while older workers moved directly from rural areas and agriculture into cities and service employment. 4 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 New Views Recent literature has raised the possibility that international interactions gen- erate rising inequality, using frames of reference other than the Heckscher- Ohlin model. Most of the work in this new literature focuses on experiences in industrial countries. In the United States and the United Kingdom, the already large inequality in earnings has grown throughout the 1980s and 1990s (OECD 1996). By contrast, most other industrial countries have had small changes, with slightly equalizing trends in the 1990s in some, including Canada, Germany, and Japan. In most industrial countries, unemployment rates among unskilled workers have doubled or tripled since the 1970s. Many analysts have argued that where relative wages have not been allowed to adjust, as in Europe, adverse shifts in demand have caused particularly severe rises in unskilled unemployment owing to sclerotic labor markets (Krugman 1995; Freeman 1995). But Nickell (1996), for example, points out that Norway, with a sclerotic labor market, has less unskilled unemployment than the United States, with a flexible one, and that the United Kingdom, with a flexible labor market, has higher unskilled unemploy- ment than most other European economies. He concludes that labor market institutions explain only part of the differential unemployment performance and suggests that countries with low unemployment have better-educated unskilled workers. Although the reality in rich countries is complex, most observers start from the view that there have been adverse relative demand shifts against unskilled labor. One debate about the reasons behind this has been couched in terms of trade versus technology. The most influential writings on this score are perhaps those of Wood (1994), Krugman and Lawrence (1993), Lawrence and Slaughter (1993), and Leamer (1995). Another debate focuses on free versus controlled migration; here, the more convincing arguments are perhaps those of Bhagwati (1991) and Borjas and Freeman (1993). Most analysts accept that trade, tech- nology, and migration all matter, with important country-specific variations (see World Bank 1995b). New evidence from some developing countries also indicates a rise in in- equality. A series of recent studies has shown that income and wage inequality has risen in Mexico in the last decade (Feenstra and Hanson 1994; Revenga 1994). Pissarides (this issue) and especially Wood (this issue) survey recent evi- dence from several countries. As Wood stresses, these recent results are in sharp contrast with earlier studies that show, mainly in the Asian context, equalizing growth strongly related to the opening of trade and expanding supply of skills in Hong Kong, Korea, Singapore, and Taiwan (China) in the 1960s and 1970s. Either rising wage inequality or rising income inequality has been observed in the mid-1970s to early 1980s in Argentina and Chile; in the mid-1980s to early 1990s in Colombia, Costa Rica, and Mexico; by the end of the 1980s,in Egypt and Jordan; and recently in China, Hong Kong, and Thailand. The evidence of Diwan and Walton S rising inequality is based largely on household surveys; the Mexican studies, using enterprise surveys, also find rising inequality between production and non- production workers. Almost all developing economies have experienced rapid changes in the rela- tive supply of skills. Robbins (1996a) attempts to disentangle supply and de- mand shifts, building on the work in the United States. In a series of nine case studies (Argentina, Chile, Colombia, Costa Rica, Malaysia, Mexico, the Philip- pines, Taiwan (China), and Uruguay), he concludes that underlying shifts in demand almost always increase inequality, although in some cases these are masked by large equalizing shifts in relative supply. Robbins further argues that the disequalizing changes in demand are often associated with trade liberaliza- tion and, in particular, the skill-intensive requirements of contemporary trade. Tan and Batra (this issue) present a microeconomic story that is consistent with these trends. They use evidence from firms in Colombia, Mexico, and Tai- wan (China) to study firms' decisions with respect to capital investment, train- ing, and exports. They also investigate the effect of those decisions on labor productivity, wages, and wage differentials. The results are both striking and intuitively appealing. Productivity and wages rise with all of these decisions, but, although all workers gain from in-firm training (workers appropriate about two-thirds of the productivity gains), white-collar workers and other workers already possessing high skills gain much more than blue-collar workers. The premium that skilled workers receive in firms undergoing technological change compared with the premium they receive in firms not investing in training are surprisingly large: 42 percent in Colombia, 54 percent in Mexico, and 32 per- cent in Taiwan (China). The comparable premiums for unskilled workers are positive but small (22, 11, and 7 percent, respectively). Unskilled workers ap- pear to gain relatively little from the introduction of new technologies, while the productivity of skilled workers rises quite dramatically. This is strong evidence that technological change favors skilled workers. Trade, Migration, and Technology As Wood (this issue) argues, the new observations are consistent with some of the old explanations. The trade and migration arguments are likely to apply to the middle-income developing countries where disequalizing trends have been observed recently. The technology arguments, in contrast, are likely to apply to all countries. In his view the trade argument relates not only to changing patterns of pro- tection, but also to increased competition from new entrants to the global mar- kets-China and India, which have huge reservoirs of cheap unskilled labor. With the takeoff of these countries' exports since the early 1980s, unskilled wages could have been depressed globally, especially in middle-income coun- tries that produce competing goods. In a general equilibrium model that disag- gregates skill levels, we found that middle-income countries, especially those in Latin America, were hurt by the entry of low-wage countries into world exports 6 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I (World Bank 1995a). In addition, evidence links the deterioration in the com- modity terms of trade, which has hurt African countries in the past three de- cades, to efficiency-improving methods introduced in East Asian countries. International labor mobility can also cause rising inequalities in middle- income countries. Inequality is much larger internationally than in any one country-the international Gini coefficient is close to 0.70 but is substan- tially less than that in national settings-and has been rising steadily (Berry, Bourguignon, and Morisson 1989). Even though labor mobility has not risen in recent times in global terms, its nature has changed, and it now increas- ingly involves unskilled workers moving into middle-income and rich econo- mies (Borjas and Freeman 1993). For example, Borjas and Freeman estimate that migration displaces unskilled workers in the United States at least as much as do imports from developing countries. Indeed, competition for jobs from migrants is more direct. The trade and technology arguments are, at some level of abstraction, tough to disentangle because technical change is itself partly driven by increased com- petition (see Wood 1994). Increased openness to trade itself exacerbates the influence of new technologies, because trade liberalization induces more imita- tion, makes reverse engineering easier, and makes complementary imports cheaper. Increased openness also increases competition, thereby increasing the incentives to imitate. Pissarides (this issue) reviews the evidence on this. But unlike the trade explanation, the technology story may well affect poore5 countries, which tend to copy new technologies, even when they are not appro- priate for their needs. As the richer North leaps ahead in education in order to escape competition from cheap labor, new technologies could increase inequal- ity in poor countries, and the process of globalization could leave behind an increasingly large part of the world. There also seems to be something new about the current globalization. Re- cent years have witnessed an acceleration of change. The recent demise of com- munism and end of the cold war, the conclusion of the Uruguay Round, the initiation of peace in South Africa and the Middle East, faster integration in Europe, Asia, and the Americas, and the seeming acceleration in technological change in the communication industry all have led to faster change in the pat- terns of production and trade. In an environment of global change, Pissarides argues that the returns to skills must rise because skills facilitate the adaptation to newness. This argument applies well to Eastern Europe, China, India, and even Latin America, where the recent wave of liberalization was sudden com- pared with the relatively slow and deliberate changes that occurred in East Asia. In periods of change, skills become especially valuable, because they are needed to acquire and put in place new technology. Micro-evidence presented by Foster, Rosenzweig, and the Rural Indian Economic Growth Research Group (1996) for India shows how the returns to schooling rose significantly during the green revolution. The new, high-yielding seed varieties were potentially much more Diwan and Walton 7 productive than the traditional seeds, but significantly more sensitive to the use of such inputs as water and fertilizer. Because farming profitability depended so critically on the allocation of inputs, the ability to decode information became more valuable. In the states most affected by the green revolution (such as the Punjab), the profit differential between educated and uneducated farmers-with a focus on primary education-was as high as 40 percent. Accepting such a view on the relationship between change and inequality, the recent rise in inequality in a number of economies could be temporary. If change slows-for example, as the consequences of opening work their way through an economy-inequality will fall. The Role of Capital Mobility Capital movements from rich to poor economies pull down wages in the former and pull up wages in the latter. But these effects are small relative to total invest- ment levels in industrial countries and have clearly not offset global trends of divergence between countries (Pritchett 1995; World Bank 1995b). Capital move- ments amplify existing international and national trends. Countries with good domestic investment environments attract mobile international capital. Foreign direct investment frequently follows where other firms have previously gone- as found in one study of Japanese foreign direct investment (Kinoshita and Mody forthcoming). Within East Asia, for example, Japanese foreign direct invest- ment and increased flows of investment from Hong Kong, Korea, Singapore, and Taiwan (China) have facilitated huge shifts in the international structure of production in the past two decades. Within countries, the effects of capital inflows may equalize or disequalize wages, depending on the nature of shifting production structures. Inflows of foreign direct investment into labor-intensive production appear to have accel- erated, leading to equalizing absorption of surplus unskilled labor into industry in countries such as Bangladesh, Malaysia, and Mauritius. But foreign capital can be relatively intensive in skills. Feenstra and Hanson (1994) argue that, with a range of production technologies, the marginal activity that moves from, say, the United States to Mexico will have factor proportions that are intensive in unskilled labor, compared with relative labor supplies in the United States, but intensive in skilled labor, compared with relative labor supplies in Mexico. These effects could increase inequality in both places. Policies That Directly Affect Wage Differentials In some countries, in addition to the effects of derived demand and relative supplies, direct policies affect wage differentials. A potentially important force leading to rising wage or income differentials is the recent demise of state efforts to influence the workings of labor markets. Whether through central planning or regulation, state intervention in setting wages and work conditions has been strong in large parts of the world but is now receding everywhere. Typically, such intervention, coming from a socialist and egalitarian background, has led 8 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I to compressed wage schedules. The reversal of such policies restores returns to education to a normal level.' In some cases, a policy of subsidized education as the primary means of social mobility has exacerbated labor market policies. Assaad (this issue) describes this phenomenon in Egypt. In this case, and in other countries in the Middle East and North Africa, inequality increasingly results from falling demand for a certain type of skill. Reforms put new kinds of skills at an advantage and can effectively reverse past educational policies. Similarly, labor market reforms in economies that successfully held up the wages of low-paid workers can lead to wage decompression, as appears to have occurred in Chile in the 1970s and Italy in the 1990s (see Robbins 1996b on Chile and OECD 1996 on Italy). In the end, whatever the deep reasons for the current wave of inequality, how societies respond to change will determine its effects. An optimistic view is that new technologies are cheaper than the old ones, capital is increasingly mobile, and the possibility of leap-frogging into modernity is increasingly open to flex- ible societies ready to attract and apply new technologies to production. There- fore, temporary inequality could buy a lot of growth. But this result depends on both the skills and the institutions of the society and, in particular, on how the education system, training in new technologies, and the interface between labor and firms adapt to new circumstances. Whether the new inequality will persist and rise, or fall rapidly, will ultimately depend on how societies organize their labor market, education, and production-related institutions. II. INSTITUTIONS AND OUTCOMES FOR WORKERS Labor market institutions affect the determination of wages and inequality. Will they matter less as reliance on domestic and international markets rises, capital becomes mobile, and technological change requires more "flexible" and shifting labor? We argue that institutions are crucial determinants of the conse- quences of market-driven change for workers. Institutions can influence both the extent to which economies participate in global advance and the extent to which different groups of workers within a country participate in national ad- vance. However, many existing institutional arrangements in developing and transition economies are moribund. The issue is institutional reform, not insti- tutional destruction. Formal institutions affecting the labor market evolved historically in response to two pressures. First, the development of formal labor contracts in modern firms led to the creation of a legal and institutional framework for employer- worker relations.2 Second, societies and workers wanted certain workplace rights, 1. A normal level would be a ratio of average wages of workers with secondary education to average wages of workers with primary education of 4-5. A ratio of 2-3 would reflect wage compression. 2. We treat the formal sector as including the portion of employment with some form of formal labor contract. The usual proxy for informal employment is nonwage employment (including family workers) plus wage employment in very small establishments. Clearly, there will be some purely casual employment in large firms and some effectively formal relationships in small firms. Diwan and Walton 9 conditions of work, and income security. Some of these demands also apply to informal contracts, and there is widespread evidence of complex contractual arrangements, in rural labor markets in particular, that evolved in response to the problems of managing risk and providing incentives for work when moni- toring is imperfect. Here we are concerned with formal institutions. Job Creation in Protected or Public Sectors In the past few decades, the predominant employment model in both cen- trally planned and most developing countries has been the creation of formal sector jobs in industry, supplemented by government jobs to varying degrees. In developing economies with a protectionist labor market, this model provided "good" jobs to only a few workers. It failed to develop a modern sector respon- sive to the needs of the market economy and fostered labor market dualism. The core mechanism involved the creation of rents in protected or public sectors for a limited number of workers who obtained access to good jobs. Implicit or di- rect taxation of other sectors financed the rents. Egypt provides a clear example of the power of institutions to affect labor demand and supply, with potentially severe long-term effects. The policy mix for most of the past thirty years involved a combination of protectionist indus- trialization, educational expansion, and public sector employment expansion- with a job guarantee for secondary and tertiary graduates. As Assaad (this issue) shows, this led to severe job rationing and queuing. Even with the sharp drop in public sector wages since the 1980s, there is still an implicit premium on govern- ment jobs-especially for women, who suffer less discrimination in the public than in the private sector. The premium on government jobs interacted with the education system to produce an excess supply of the wrong kinds of skills, with significant compression of wage differentials for skills. The employment strategy also failed under central planning, but with a very different dynamic. In Eastern Europe and the former Soviet Union, massive creation of "good" jobs, especially in industry, was accompanied by widespread job security provisions. But this turned out to be a chimera. Many good jobs quickly became bad when subject to the market test, and the legacy of cradle-to-grave income protection remains a major fiscal problem. More- over, even when the jobs were there, they were neither noted for healthy working conditions nor for industrial democracy. By contrast, central plan- ning in China did not lead to the hugely accelerated employment shifts of the Soviet Bloc. In 1978, at the inception of China's transition from central plan- ning, some 80 percent of the employment was still in agriculture. This has greatly eased employment aspects of the subsequent transition (Sachs and Woo 1994; World Bank 1996b). However, the "modern" jobs that were cre- ated in government and the state enterprise sector receive a much higher degree of income protection and social benefits than in market developing economies. These are only beginning to be disentangled in the mid-1990s in the current phase of reform. 10 TH-IE WORLD BANK ECONOMIC REVIEW, VOL. II, NO. I Effects of Protective Labor Market Policy Many analysts believe that protective labor market policy contributes to poor economic and employment performance, in both rich and poor economies al- though actual evidence is sparse (see OECD 1994 for industrial countries). Freeman's (1993) review did not find general evidence in support of the view that labor regulations had significant economic costs, nor did a set of case stud- ies of labor in the adjustment process (Horton, Kanbur, and Mazumdar 1994). A cross-country analysis of Latin America by Rama (1995a) did find that a synthetic index of labor market distortions reduced growth by about half a per- centage point (after controlling for other factors), but most of this appears to be associated with unionization and perhaps with public employment as opposed to regulations. One interpretation of these results is that labor market regulations that go against the market fail. Squire and Suthiwart-Narueput (this issue) model the incentives to evade and avoid distorting labor market regulations, using minimum-wage law as an example. They show that where regulations are binding, compliance is charac- teristically low, especially (but not only) in developing countries. They also show that once compliance is endogenized, there is likely to be an upper limit to efficiency losses. And such binding regulations tend to be in protected sectors-public and industrial. Here the better conditions are a form of rent-sharing. In the public sector, a labor cost-raising regulation may actually bring choices closer to an efficient solu- tion, to the extent that it offsets a tendency toward excess employment. Where unions have some degree of independence, their influence often ap- pears to be negative for overall economic and employment performance. Free- man and Medoff (1984) discuss the contrast between the monopoly aspects of union behavior and the potentially productivity-raising gains from unions giv- ing a voice to workers. Rama (1995b) and Standing (1992) provide evidence that unionized sectors experience relatively slow employment growth. Rama estimates that in Jamaica, employment in highly unionized sectors grew 2 to 5 percentage points slower than other sectors in 1986-93. Standing estimates that in Malaysia, employment in sectors with industrial unions grew 5 percentage points slower than sectors without unions or with plant-level unions. Unions have often been resistant to needed economic reforms-whether in Indian trade liberalization or in Latin American public sector reform. Devarajan, Ghanem, and Thierfelder (this issue) illustrate the potential importance of un- derstanding union behavior in modeling economic change. They use Bangladesh as an example of a small but powerful unionized sector in protected industries. Exploring the effects of alternative assumptions on union behavior, including the option of unions bargaining simultaneously over employment and wages, as in the McDonald-Solow model, they find potential additional benefits of disprotection through trade liberalization when unions are present. Their results show that unions can matter but that the additional gains from liberalization are based on reducing the rents to union members. Diwan and Walton 11 Labor Policy and Institutions At first glance, recent history looks gloomy for labor policy and institutions. One conclusion of the failure of employment strategy in countries with high levels of protection or central planning might be that labor policy needs to be deregulated. A superficial interpretation of East Asia's extraordinary success in both growth and modern sector employment creation is that it shows the advantages of flexible, unregulated labor markets. We argue the opposite here. Indeed, rising international integration adds to the case for reform of some labor regulations. Overall, however, three trends in international labor markets highlight the growing need for effective institutional arrangements for labor in much of the world: * The secular rise of modern wage work in all regions except industrial coun- tries and the former Soviet Bloc * Increased domestic and international demands for stronger workplace rights, including freedom of association and better working conditions * The declining efficacy of informal mechanisms for dealing with working conditions and income security as populations become more urbanized and work becomes more formalized. Politics will continue to matter crucially for institutional choices. But the view that repression of labor institutions is necessary to get favorable results for em- ployment and wage outcomes is debatable. Banerji and Ghanem (this issue) ex- plore the empirical relationships between measures of political and civil liberties and both trade and labor market outcomes. They find that on average more liberal regimes are associated with more open trading and less labor market dualism. Authoritarian East Asian countries are the exception. It is also a mistake to interpret East Asia's success in terms of a flexible, regulation- and union-free, neoclassical economist's ideal. In fact, countries in East Asia have had diverse labor arrangements. Unions have been severely re- stricted in Korea (especially in the period preceding democracy) but have exten- sive rights in Hong Kong and Japan (see Freeman 1993; World Bank 1996a). Korea has highly pervasive labor regulations, notably on job security, and an extensive government presence in areas that in many countries would be subject to collective bargaining. State enterprise workers in China have unparalleled levels of protection and are extraordinarily immobile by international standards. East Asian countries are often characterized as having a cultural tradition of reliance on the family (see, for example, Rohwer 1995). But the elderly in Japan are living less with their families than before, and the richer newly industrialized countries, including Korea, Malaysia, and Singapore, all have significant ex- panding forms of formal social insurance. An important rationale for unions is that flexible arrangements for produc- tion organization and information flow yield potential productivity benefits that could be on the rise with more sophisticated technologies. However, these benefits-as opposed to the potential monopoly effects of unions-are more 12 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I likely to be reaped with a predominantly firm-based form of industrial relations within competitive product markets (see World Bank 1995b; Pencavel 1991 and 1995). Unions also have other important benefits, including protecting individual workers from abuse, monitoring health and safety regulations, and reducing gender and ethnic discrimination. For example, MacIsaac and Rama (1996) find that the significant discrimination experienced by women and indigenous employees in Ecuador disappears for union members; Patrinos (1994) finds com- parable results for Mexico. Women tend to be over-represented (compared to the private sector) in the well-protected public sectors of Egypt and Jordan (Assaad, this issue; World Bank 1995b). Unfortunately, carefully documented research on both productivity and other effects is still rare in developing coun- tries. Thus, policy will have to proceed with some uncertainty, creating condi- tions for unions to operate, reforming areas where monopolistic practices are probable (especially in the public sector), and not expecting unions to be able to solve all labor's problems. Where does this leave labor regulations? Despite the weak evidence, a highly differentiated view is desirable at this juncture. Guarantees of rights for unions are desirable, but not monopoly privileges. Measures to protect the vulnerable are important, but not widespread mandates on working conditions. Public ac- tion to reduce income insecurity is clearly justified, but the design of interven- tions is crucial. In addition, the functioning of education and training systems will have a powerful complementary effect on labor market performance. Al- though government intervention remains of central importance in this sector, here, too, the form of this intervention is now the subject of many debates. Although it goes beyond the scope of this introduction to review this area, solv- ing the quality problems that affect many developing-country school systems will be critical to producing school leavers with the wherewithal to acquire chang- ing skills during the course of their working lives. III. CONCLUSION Increased trade linkages, mobile capital, and technological change are rais- ing new questions over the attainment of old employment objectives. Debates in rich countries on insecurity of work and unemployment of unskilled work- ers find strong echoes in recent evidence from some developing and transition countries. As the articles in this issue illustrate, many of the debates are still current. They involve questions of how to achieve reasonable degrees of eq- uity, satisfactory workplace conditions, and income security in the face of ap- parently rising market pressures. These debates also raise fundamental ques- tions concerning the role of institutions that affect labor market outcomes, although evidence on the effects of alternative institutional arrangements is particularly scarce. The answer is not institutional destruction. Societies that successfully put in place effective institutions are more likely to capture international gains. Three Diwan and Walton 13 areas of institutional innovation are likely to be central in the coming years: workplace organizational forms, formal mechanisms for the management of household risks, and education and training systems. These innovations are likely to be complementary: the greatest payoff will come from action in all three areas together. And it is likely that future research will have a rich payoff in this central domain for households and economies. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Alesina, Alberto, and Dani Rodrik. 1994. "Distributive Policies and Economic Growth." Quarterly Journal of Economics 109(2, May):465-90. Berry, Albert, Francois Bourguignon, and Christian Morisson. 1989. "The World Dis- tribution of Income: Evolution over the Recent Period and Effects of Population Growth." Paper prepared for the conference The Consequences of Rapid Population Growth, United Nations, New York. Processed. Bhagwati,Jagdish. 1991. "Free Traders and Free Immigrationists: Strangers or Friends?" Russell Sage Foundation Working Paper 20. Russell Sage Foundation, New York. Processed. Borjas, George J., and Richard B. Freeman. 1993. Immigration and the Work Force: Economic Consequences for the United States and Source Areas. Chicago, Ill.: Uni- versity of Chicago Press. Bruno, Michael, Martin Ravallion, and Lyn Squire. 1996. "Equity and Growth in De- veloping Countries." Policy Research Working Paper 1563. Policy Research Depart- ment, World Bank, Washington, D.C. Processed. Deininger, Klaus, and Lyn Squire. 1996. "A New Data Set Measuring Income Inequal- ity." The World Bank Economic Review 10(3, September):565-93. Feenstra, Robert C., and Gordon H. Hanson. 1994. "Foreign Direct Investment and Relative Wages: Evidence from Mexico's Maquiladoras." Department of Econom- ics, University of Texas at Austin. Processed. Foster, Andrew D., Mark R. Rosenzweig, and the Rural Indian Economic Growth Research Group. 1996. "Technical Change and Human Capital Returns and Invest- ments: Consequences of the Green Revolution." University of Pennsylvania, Philadel- phia; National Council of Applied Economic Research, Washington, D.C. Processed. Freeman, Richard B. 1993. "Labor Market Institutions and Policies: Help or Hindrance to Economic Development?" World Bank Annual Conference on Development Eco- nomics 1992. Washington, D.C.: World Bank. . 1995. "Will Globalization Dominate U.S. Labor Market Outcomes?" Paper prepared for the conference Imports, Exports, and the American Worker, Brookings Institution, Washington, D.C. Processed. Freeman, Richard B., and James L. Medoff. 1984. What Do Unions Do? New York: Basic Books. Goldin, Claudia, and Robert A. Margo. 1991. "The Great Compression: The Wage Structure in the United States at Mid-Century." NBER Working Paper 3817. National Bureau of Economic Research, Cambridge, Mass. Processed. 14 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I Horton, Susan, Ravi Kanbur, and Dipak Mazumdar. 1994. Labor Markets in an Era of Adjustment, 2 vols. EDI Development Studies. Washington, D.C.: World Bank. Katz, Lawrence, and Kevin M. Murphy. 1992. "Changes in Relative Wages, 1963-87: Supply and Demand Factors." Quarterly Journal of Economics 107(February):35- 78. Kim, Dae-l1, and Robert H. Topel. 1995. "Labor Markets and Economic Growth: Les- sons from Korea's Industrialization, 1970-1990." In Richard B. Freeman and Lawrence F. Katz, eds., Differences and Changes in Wage Structures. Chicago: Uni- versity of Chicago Press. Kinoshita, Yuko, and Ashoka Mody. Forthcoming. "Private and Public Information for Foreign Investment Decisions." Policy Research Working Paper, East Asia and Pa- cific Region, World Bank, Washington, D.C. Processed. Krueger, Anne 0. 1978. "Foreign Trade Regimes and Economic Development: Liberal- ization Attempts and Consequences." NBER Working Paper. National Bureau of Eco- nomic Research, Cambridge, Mass. Processed. . 1983. Trade and Employment in Developing Countries: Synthesis and Conclu- sions. Chicago: University of Chicago Press. Krugman, Paul. 1995. "Past and Prospective Causes of High Unemployment: Current Issues and Policy Options." Proceedings of a symposium in Jackson Hole, Wyo- ming. Federal Reserve Bank of Kansas City, Missouri. Processed. Krugman, Paul, and Robert Lawrence. 1993. "Trade, Jobs, and Wages." NBER Work- ing Paper 4478. National Bureau of Economic Research, Cambridge, Mass. Pro- cessed. Lawrence, Robert, and Matthew Slaughter. 1993. "International Trade and American Wages in the 1980s: Giant Sucking Sound or Small Hiccup?" Brookings Papers on Economic Activity: Microeconomics 2:161-226. Leamer, Edward E. 1995. "A Trade Economist's View of U.S. Wages and Globaliza- tion." In Suzan Collins, ed., Imports, Exports, and the American Worker. Washing- ton, D.C.: Brookings Institution. Little, I. M. D. 1982. Economic Development. New York: Basic Books. MacIsaac, Donna, and Martin Rama. 1996. "Determinants of Hourly Earnings in Ecuador: The Role of Labor Market Regulations." Policy Research Department, World Bank, Washington D.C. Processed. Nickell, Stephen. 1996. "Unemployment and Wages in Europe and North America." Institute of Economics and Statistics, University of Oxford. Processed. OECD (Organization for Economic Cooperation and Development). 1994. The Jobs Study-Facts, Analysis, Strategies. Paris. . 1996. Employment Outlook. Paris. Patrinos, Harry Anthony. 1994. "The Costs of Discrimination in Latin America." Edu- cation and Social Policy Department, World Bank, Washington, D.C. Processed. Pencavel, John. 1991. Labor Markets under Trade Unionism: Employment, Wages, and Hours. Cambridge, Mass.: Basil Blackwell. .1995. "The Role of Labor Unions in Fostering Economic Development." Back- ground paper to World Development Report 1995: Workers in an Integrating World, World Bank, Washington, D.C. Processed. Diwan and Walton 15 Perrson, Torsten, and Guido Tabellini. 1995. "Is Inequality Harmful for Growth?" American Economic Review 84(3, June):600-21. Pritchett, Lant. 1995. "Divergence, Big Time." World Bank Policy Research Paper 1522. Policy Research Department, World Bank, Washington D.C. Processed. Rama, Martin. 1995a. "Do Labour Market Policies and Institutions Matter? The Ad- justment Experience in Latin America and the Caribbean." Review of Labour Eco- nomics and Industrial Relations (special issue). - . 1995b. "Unions and Employment Growth: Evidence from Jamaica." World Bank, Washington D.C. Processed. Revenga, Ana. 1994. "Employment and Wage Effects of Trade Liberalization: The Case of Mexican Manufacturing." Paper presented at the Labor Markets Workshop, Policy Research Department, World Bank, Washington, D.C. Processed. Robbins, Donald. 1996a. "HOS Hits Facts: Facts Win-Evidence on Trade and Wages in the Developing World." Harvard Institute for International Development, Har- vard University, Cambridge, Mass. Processed. .1996b. "Trade Liberalization, Modernization, and Earnings Inequality in Chile." Harvard Institute for International Development, Cambridge, Mass. Processed. Rohwer, Jim. 1995. Asia Rising: Why America Will Prosper as Asia's Economies Boom. New York: Simon and Schuster. Sachs, Jeffrey D., and Wing Thye Woo. 1994. "Experiences in the Transition to a Mar- ket Economy." Journal of Comparative Economics 18(June):271-75. Standing, Guy. 1992. "Do Unions Impede or Accelerate Structural Adjustment? Indus- trial versus Company Unions in an Industrializing Labour Market." Cambridge Jour- nal of Economics 16:327-54. Williamson, Jeffrey, G. 1996. "Globalization and Inequality, Past and Present." De- partment of Economics, Harvard University, Cambridge, Mass. Processed. Wood, Adrian. 1994. North-South Trade, Employment, and Inequality: Changing For- tunes in a Skill-Driven World. Oxford: Clarendon Press. World Bank. 1993. The East Asia Miracle: Economic Growth and Public Policy. New York: Oxford University Press. 1995a. Will Arab Workers Prosper or Be Left out in the Twenty-First Century? Regional Perspectives on World Development Report 1995: Workers in an Integrat- ing World. Washington D.C.: World Bank. . 1995b. World Development Report 1995: Workers in an Integrating World. New York: Oxford University Press. . 1996a. Involving Workers in East Asia's Growth. Regional Perspectives on World Development Report 1995: Workers in an Integrating World. Washington D.C.: World Bank. . 1996b. World Development Report 1996: From Plan to Market. New York: Oxford University Press. Learning by Trading and the Returns to Human Capital in Developing Countries Christopher A. Pissarides Recent evidence shows that the returns to labor and the skill premium both increase in developing countries after trade liberalization, despite the low skill content of their exports. The author explains this apparent puzzle by arguing that trade in- creases technology transfers from industrial to developing countries and that the transfer technology is biased in favor of skilled labor. The relative demand for skilled labor increases during the transition following liberalization, and so the gains en- joyed by skilled labor are temporary, even in the absence of supply responses. The gains become longer lasting when the transferred technology is also skill-biased. Several recent studies have found that trade liberalization in developing coun- tries is often associated with a large increase in the returns to labor and with an increase in wage inequality. The increase in the returns to labor should come as no surprise: labor should receive at least some of the benefits from trade. The increased differential between the wages of skilled and unskilled labor, how- ever, is more puzzling. According to the conventional and largely substantiated view of North-South trade, the industrial "North" exports high-quality goods to the developing "South," which exports primary products or lower-technol- ogy goods to the North. Trade liberalization in a developing economy should therefore be associated with an increase in the relative demand for unskilled labor and a narrowing of wage differentials. This article discusses a channel through which the effects of trade liberal- ization are transmitted to the labor market and explains why both labor returns and wage inequality can increase after liberalization. According to this view, developing economies (the South) advance by learning from the technology of industrial economies (the North). Learning is faster when trade Christopher A. Pissarides is with the Centre for Economic Performance at the London School of Economics. This article is a revised version of part of his background paper for World Development Report 1995: Workers in an Integrating World, which circulated under the title "Trade Liberalization and the Returns to Human Capital in Developing Countries." The author is grateful to Michael Walton and the staff of the World Development Report office for pointing out the puzzle, for offering their comments, and for making available unpublished information on labor returns in developing countries. The author also benefited from discussions with Tony Venables, Mary Amiti, Donald Robbins, Andrew Oswald (who suggested the title), and seminar participants at the Centre for Economic Performance, Warwick University; University of California, Los Angeles; the World Bank; and the Organization for Economic Cooperation and Development. C) 1997 The International Bank for Reconstruction and Development /THE WORLD BANK 17 18 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I links the economies of North and South, so trade liberalization in a develop- ing country leads to more technology transfers from the North to the South. Technology transfers can take place either through the production in the South of capital goods that are already in use in the North or through the importation of capital goods from the North. Results are independent of the method of technology transfer. The key assumption is that the transfer of technology requires skilled labor. When a developing economy liberalizes trade, it experiences more technology transfers than before. Learning about the new technology and putting it to use in the South increase the demand for skilled labor, whose wages rise over and above any rise accruing to all kinds of labor from more production. But as the South learns the new technology, the pace of transfer slows, and the benefits to labor that remain are derived entirely from the production technology. If the transferred production technology is neutral, the relative advantage that skilled labor enjoyed during the learning process ceases. Thus, with a neutral produc- tion technology, a temporary increase in the relative demand for skilled labor causes the relative advantage of skilled over unskilled labor. With a skill-biased transferred production technology, the relative increase in the demand for skilled labor can be permanent. Some recent literature supports the idea that more trade brings about more technology transfers. Coe, Helpman, and Hoffmaister (1995) find that the re- search and development (R&D) spillovers from the industrial countries to the developing countries are substantial. Total factor productivity in developing countries was positively associated with R&D expenditure abroad. More im- portant here, however, the spillovers were linked to trade flows between the industrial and developing countries. The spillover from an industrial country to a developing country was proportional to the share of the industrial country's imports in the developing country's gross domestic product. The empirical papers cited in section I, in particular Robbins (1994, 1995b) and Hanson and Harrison (1994), identify technology transfers as the most likely reason for the increase in wage inequality. Also, Wood (1995), in a recent evalu- ation of the widening of wage differentials, concludes that the transfer of tech- nology is only one of two plausible explanations for the widening (the other one being the increase in the supply of low-technology goods in world markets be- cause of the expansion of Chinese exports). Section I discusses the recent evidence and explains why the widening of dif- ferentials is a puzzle in light of the recent widening of wage differentials in the labor markets of the industrial world. Section II explains the main idea of learn- ing and technology transfers. Section III formalizes the learning process. Section IV describes and solves the full model. Section V considers the adjustments that follow trade liberalization and presents the main findings of this article. Section VI briefly considers other factors that might play a role in the relation between trade liberalization and wage differentials. Section VII contains some conclud- ing remarks. Pissarides 1 9 I. THE RETURNS TO HUMAN CAPITAL AFTER TRADE LIBERALIZATION The most convincing evidence supporting the widening of wage differentials after trade liberalization comes from middle-income countries in Latin America and East Asia. In each case reviewed here, other reforms took place alongside trade liberalization, so other factors might have caused the widening of differ- entials. But in all cases the widening of differentials followed a large increase in trade flows. As the number of cases where trade liberalization and widening of differentials are observed concurrently increases, the link between the two will gain more support. Robbins (1994, 1995b) examines data for Chile and Colombia. For Chile, he examines household survey data for 1957-92 to see if there were any marked changes in the earnings structure after trade liberalization took place beginning in 1975. He finds that although the skill composition of imports exceeded that of exports, skilled labor did not suffer a relative drop in earnings after trade liberal- ization. The returns to skilled labor increased by more than conventional trade theory would predict and also by more than changes in labor supply would predict. He concludes that the most likely explanation for the observed changes was the importation of capital that was complementary to skilled labor. Robbins finds a similar situation in Colombia. Using household data for Bogoti for 1976-89, he examines the response of wage differentials to the large rise in exports after the 1984 devaluation. He finds that wage inequality increased and attributes this to changes in the composition of the demand for labor that are induced by trade. Pessino (1995) finds similar results for Argentina after 1990, also using house- hold data. However, trade does not play an important part in her analysis, so it is not clear whether the widening of differentials could be attributed to trade. In more recent work, Robbins (1995a) examines household data for Argentina, Costa Rica, the Philippines, and Taiwan (China) and also finds similar results. Trade liberalization increased the relative demand for skilled labor in virtually all cases. Hanson and Harrison (1994) examine plant-level data for 1984-90 in Mexico, where trade liberalization took place beginning in 1985. They find that wage inequality increased after liberalization, despite the relatively low skill content of Mexico's exports. They conclude that the most likely cause of the rise in wage inequality was the importation of skill-biased technology from abroad. Tan and Batra (1995) present another set of related data. They do not explic- itly look at the implications of trade liberalization; rather they calculate the wage premium paid by firms that engaged in R&D, worker training, and export activities in Colombia, Mexico, and Taiwan (China). Using firm-level data and controlling for other firm characteristics, they find that firms that engaged in technology-advancing activities paid all their workers a premium over and above the wages paid by other firms but that the premium paid to skilled workers far exceeded that paid to unskilled workers. Although the direct contribution of exports was less important than that of R&D and worker training, Tan and 20 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Batra claim that trade liberalization increased the R&D activities of firms and the demand for skilled labor. They do not explore the links between trade in- volvement and technology-advancing activities at the level of the firm. The experience of developing countries with wage inequality reflects the ex- perience of industrial countries, which started somewhat earlier. A well- documented fact is that beginning sometime in the mid-1970s, the industrial world experienced a negative shock at the unskilled end of the labor market, which intensified in the 1980s. Despite growth in the economy overall, the wage and employment prospects of unskilled labor suffered relative to those of skilled labor and in some cases absolutely as well. The shock in labor markets of indus- trial countries did not have the same implications everywhere. It caused real wage reductions for unskilled labor in the United States and high unemploy- ment in the major European economies, but what is more important here is the source of the shock. One view claims that trade with low-wage economies caused the shock in industrial countries. If this is true, however, the relative wages of unskilled labor in developing economies that opened up to trade would have risen, not fallen. The data summarized in this section contradict this view.' In another view, skill-biased technological progress caused the shock in industrial countries. Recent technological advances, such as computerization, require more skilled labor than the technologies they are replacing. As new investment crowds out unskilled labor, its wages relative to those of skilled labor fall. The technology transfers modeled in this article cause more wage inequality in developing countries because the transfer technology is biased in favor of skilled labor. If the transferred production technology is also biased in favor of skilled labor, the inequality is reinforced. The skill-biased technology that comes into operation because of trade causes more wage inequality. So the implication for the debate of "trade versus technology" is that neither trade nor technology alone causes inequality; rather both act together. My approach does not contra- dict the existence of Hecksher-Ohlin effects on wage differentials. Instead, it disputes the view that the Hecksher-Ohlin effects on wages dominate in the tran- sition from steady state in a semiclosed economy to steady state in an open one. II. TRADE AND LEARNING In this section, I follow recent literature and assume that technological im- provements result from activities that use skilled labor as an input. I refer to these activities as R&D (or, when the reference is explicitly to developing coun- tries, as imitation). I interpret R&D broadly to include all activities that lead to new discoveries, as well as imitation and assimilation activities that lead to the adoption of products or techniques known elsewhere. 1. Wood (1995), who in earlier work claimed that the widening of wage differentials in the industrial North was associated with trade, claims that the widening of wage differentials associated with trade growth in developing countries is a recent phenomenon. In the early 1970s there was some evidence that wage differentials narrowed in southern European countries that liberalized trade. Pissarides 21 A firm producing with the most advanced known technology can improve its productivity only by undertaking R&D aimed at new discovery. This is likely to be the cause of technological change in the more advanced countries. But pro- ducers in less advanced countries are not likely to be using the most advanced known technology. For them, it is cheaper to copy other firms' technologies than to attempt to make new discoveries. Imitation has a higher probability of success than R&D aimed at an original discovery: learning from the mistakes of others is cheaper than learning from one's own. (See Tan and Batra 1995 for a summary of some case studies that find that technological progress in develop- ing countries is achieved through imitation.) Given the low-wage advantage that developing countries enjoy, if a firm in a developing country succeeds in imitat- ing a production process implemented by firms in a richer trading partner, it is also likely to succeed in selling its product to the more advanced country. Trade induces more imitation in a developing country by giving its producers several types of incentives to learn fast the technology of the North. First, trade between two countries exposes producers in each to the capital equipment and techniques used in the other. Trade increases the probability that imitation in the South will succeed because the traded goods embody the transfer of infor- mation about new products and techniques. Nelson (1970) niakes a related dis- tinction between "experience" and "inspection" goods. In general, producers in developing countries might be aware that producers in other countries use supe- rior techniques or that producers elsewhere produce better varieties of goods. But producers in developing countries that are exposed to the superior goods through trade are more likely to imitate those products and techniques. Second, even with a given probability of successful imitation, the competition from trading partners increases the firm's incentives to innovate. The trading partners enter the market with superior goods, which can drive domestically produced goods out of the market. This occurs regardless of whether firms in the South move from monopolistic to competitive conditions, which may or may not increase R&D activities. Producers in the South have more incentives to imitate the superior technology of the North, in order to be able to compete with newcomers, retain their domestic market share, and export to markets in the North. Third, after trade liberalization, the developing country can import more capi- tal and other intermediate goods from the North. In this case, the R&D technol- ogy should be interpreted as a learning-how-to-use technology, rather than as a learning-how-to-produce technology. This interpretation is consistent with the plausible assumption that skilled labor is needed to show the rest of the work- force how to put imported capital goods into productive use. These ideas imply that when a developing country liberalizes trade, it devotes more resources to imitating the technology of its trading partners and partici- pates in more technological transfers. But how well does the available interna- tional R&D data for developing countries measure this broad concept of R&D? It is a well-documented fact that R&D activities are heavily concentrated in a 22 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I few industrial countries and that developing countries undertake very few of them (see, for example, Coe and Helpman 1995 and Coe, Helpman, and Hoffmaister 1995). Nevertheless, a first potential test of the ideas presented here is whether trade liberalization is associated with more R&D, however nar- row its empirical definition. In an empirical analysis, imitation in the context discussed here should be interpreted broadly to include all new production ac- tivities in the South that are influenced by the exposure of its economy to that of the North. III. A FORMALIZATION OF THE LEARNING TECHNOLOGY In this section, I formalize the idea of a technological transfer by assum- ing that the speed of learning depends on the difference between the knowl- edge of producers in a country and the knowledge of its trading partners. Trade liberalization leads to more technology transfers because it enlarges the set of technological innovations that become known to producers in de- veloping countries. I measure the state of technology by the number of capital varieties in exist- ence (Romer 1990). Capital varieties are differentiated but not necessarily supe- rior to one another. Here, I look for a symmetric equilibrium with the same quantity of each known variety of capital. More capital varieties lead to more total factor productivity because each variety has diminishing returns to scale. This assumption gives the same results for the returns to human capital as the alternative assumption of improvements in the quality of capital goods, which provides another appealing way of measuring the state of technology (Grossman and Helpman 1991b). Borrowing ideas from Grossman and Helpman (1991a), I assume that the level of technology in the North improves through original expensive research and that a lower level of technology in the South improves through cheaper imitation of the North's technology. To simplify the analysis, I assume that the level of technology in the North is independent of the activities of firms in the South. In addition, because the economy of the North is not of interest in this analysis, I simplify further by assuming that the technology of the North im- proves at exogenous rate g. In the South, producers engage in R&D to discover ways either to imitate the innovations used elsewhere or to learn how to use imported machinery. In order to write an equation for the learning technology, I denote the number of varie- ties of known capital goods in the North by A. Of these, the South already uses several varieties, B. B can never exceed A, and normally it is strictly less than A. Following Rivera-Batiz and Romer (1991), I assume that the production tech- nology uses three factors of production-unskilled labor, denoted L; human capital (or skill), denoted H; and all the varieties of the capital goods known in the economy. In contrast, the more skill-intensive imitation technology activity uses only human capital as an input. Pissarides 23 The number of varieties of capital goods increases according to the learning technology, which depends on the input of human capital and on the number of varieties of capital at home and abroad. I write the learning technology in the South as (1) B = XHB'F(B, A - B) where B has already been defined as the number of varieties of capital goods already copied (the stock of knowledge in the developing economy), X is a con- stant, HB is the stock of human capital employed in the imitation process, and 41(B, A - B) is a homogeneous function, increasing in both its arguments, and concave. The function 'D(.) is the main new element in the model and the key to the results. Two features are important. First, imitating the technology of the North is a skill-biased activity in that it uses skilled labor (human capital) but not unskilled labor. The fact that no unskilled labor or capital is used in imitation is a simplification. What is important is that, relative to the pro- duction technology, more human capital and less unskilled labor are used in R&D. Second, the function D(.) captures the idea that learning is easier when there is already a lot of knowledge and also when there is more knowledge to acquire. Thus, holding constant the stock of knowledge in the developing country, expo- sure to more knowledge abroad raises the speed of learning at home for given inputs into R&D. For convenience, I assume that at A = B; that is, when there are no more capital varieties to-be copied, 4D = B. This and the other technical assumptions on 4D(.) imply that (2) 4(B,A - B) = 4D(1,A/B - 1) B =p (AIB)B. 0(.) increases in its argument at a decreasing rate and satisfies 0(1) = 1. Using equation 2, equation 1 becomes B (3) - = XO(AIB)HB. B Equation 3 gives the rate of technological progress in the economy of the devel- oping country, which, with fixed labor supply, in the steady state is also the rate of growth of aggregate output. The function o(AIB) represents the advantage of the imitator over the inventor. If A = B, 4 = 1 and equation 3 can be interpreted as a process of discovery. But with A > B, which will always be the case if the economy of the South is less technologically advanced than the economy of the North, 0 > 1 and imitation is cheaper (that is, more productive for a given hu- man capital input) than original discovery. 24 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO 1 IV. GROWTH EQUILIBRIUM Equilibrium is described by the evolution of capital varieties, B, given an arbi- trary initial value, market-clearing prices for the three factors of production, and the allocation of human capital between production and imitation. I assume a constant supply of each type of labor in order to isolate the effects of trade liberalization that work through the demand for labor. (The implications of variable supply are discussed briefly in section VI.) Factors are fully employed, and the allocation of human capital between its two uses is at the point where marginal returns in each use are equal. The consumption side of the model is not specified, so the model does not need to include trade flows. I assume instead that output is divided between final consumption and capital and obtain equilibrium output and capital stock from the market-clearing conditions in factor markets. In order to do that, I next specify the production technology. The production technology uses the three factors of production-the known varieties of capital, human capital, and unskilled labor-to produce output that is subsequently sold in competitive markets at a unit price (the numeraire). To avoid integer problems, suppose that the capital varieties lie on a continuous line and that the representative production unit employs x(i) units of each capi- tal variety i. The production function is (4) Y = HVLP x(i)1-lxdi 0 where Y denotes output, Hy denotes the human capital used in the production of goods, L denotes the unskilled labor input, and a and 3 are positive parameters that sum to a number less than 1. Aggregate output Y is divided between consumption and investment in new capital goods. For exposition purposes, suppose that the property rights to each variety of capital are owned by a patent holder or by the single importer who first succeeded in imitating (or importing and adapting) the variety. The patent holder, a monopolist, then licenses this product to a manufacturer, who supplies it at a unit price to the patent holder. The patent holder decides how much of the capital variety to order on the basis of the price for selling it back to manufac- turers. That price, the price of the capital variety, is denoted p(i). Assuming for simplicity that capital fully depreciates after one period, p(i) equals the static marginal product of x(i), (5) p(i) = a = (1 - a - P)H0LPx(i)-'P. (5) p(i) a~~~~x(i) Because the patent holder has to pay one unit price for each unit of capital goods, the per period profit accruing from the sale of the capital variety is p(i)x(i) - x(i). The monopolist chooses the quantity of x(i) supplied to maximize Pissarides 25 profit, so the equilibrium quantity of each capital variety supplied maximizes the expression (6) p(i)x(i) - x(i) = (1 - a - P)H cL x(i)1-(x- - x(i). The maximization condition is (7) (1- aX - )HO'Lx(i)-`P -1 = 0. Equation 7 gives the equilibrium condition in the market for each capital variety i, given the inputs of the other factors of production. Substitution of x(i) from equation 7 into equation 5 gives the monopolist's price for the capital good: (8) p(i)= > 1. The difference between the equilibrium price and the cost of capital, p(i) - 1, is the monopolist's profit from owning the patent for the capital variety. Because thex(i) that solves equation 7 and thep(i) in equation 8 are indepen- dent of i, I can drop i and write x for the equilibrium quantity of each capital variety and p for the equilibrium price. It follows that the aggregate capital input into the production technology, the integral of all B varieties of x(i), is simply Bx. The production function then becomes (9) Y = BHyLYx`. Equilibrium in the market for unskilled labor, given the fixed labor supply L, gives the solution for the unskilled wage. The marginal product of labor is ob- tained from equation 9 and is equated to the unskilled wage, WL, to give (1 0) BHxLP-x-a-> = WL. The allocation of human capital between the two activities-imitation and production-takes place such that the rewards to human capital in each are equalized. The rewards to human capital in production are equal to the mar- ginal product of human capital, aY/ DHy. In the imitation technology, the profit from employing one more unit of human capital is the present discounted value of profit from the discovery, given by [(px - x)XO(A/B)B]/r for a constant dis- count rate r. Equating this expression with the marginal product of human capi- tal in the production technology and making use of equation 7 to substitute out x give the solution for Hy: (a + )(1x- - P5)X4(AIB) The only remaining equation for full characterization of the solution is the one for the wages of skilled labor. This is obtained by differentiating the produc- 26 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I tion function with respect to human capital and equating the result to the wage rate: (12) oBHy lLPxlIa- = WH. Equations 3, 7, 10, 11, and 12 are uniquely solved for the paths of B, x, HY, WL, and WH, given an initial value for B, labor supply L, and human capital supply H, where HB = H - HY. The path of output can then be obtained from equation 9. The steady state of the system is found by looking for a steady state for B. Because 0(.) is not a linear function, if A is growing at exogenous rate g, equa- tion 3 implies that in the steady state B must grow at the same rate. Therefore, in the long run, the economies of both North and South must grow at the same rate (a result known since Krugman's 1979 pioneering paper), though obviously their income levels might differ. Substitution of the steady-state growth rate into equation 3 gives the steady-state relation between the level and rate of growth of technology and human capital: (13) X(A/B) (H - Hy) - g = O. I illustrate the solution with a simple diagram that is useful in the analysis of trade liberalization (section V). Equations 13 and 11 give unique solutions for HY and AIB, as illustrated in figure 1. The line HH is a representation of equation 11 and shows the condition under which the returns to human capital are equalized in the two activities in which human capital takes part. HH has a negative slope because a higher AIB makes imitation more rewarding, so the input of human capital into production falls. The line SS shows the steady-state condition of equal growth rates in the two economies (equation 13). SS slopes up because, for a given growth rate abroad, a higher AIB in- creases the domestic growth rate because it facilitates imitation. To main- tain a constant growth rate requires switching human capital from imitation to production. Under the,simplifying assumption that human capital is perfectly (and instan- taneously) mobile between production and imitation, the economy is always at some point on the HH line. That point is determined by the value of the exog- enous variable A and the value of the predetermined variable B. Because equa- tion 3 is a stable differential equation in B, the economy is always tending to the intersection of the two curves in figure 1 along HH, as shown by the arrows (ignore point T for the moment). With knowledge of B and HY from figure 1, it is possible to obtain a unique solution for the path of x from equation 7, a unique solution for the path of output from equation 9, and unique solutions for the paths of wages from equa- tions 10 and 12. Pissarides 27 Figure 1. Equilibrium with Technological Transfers Ratio of capital varieties known in the North to capital varieties H S known in the South, A/B T S /\ Human capital used in the production of goods, H,, V. TRADE LIBERALIZATION I now consider the influence of trade on the relative returns to skilled and unskilled labor. I first solve explicitly for the relative returns to skilled and un- skilled labor by dividing equation 12 by equation 10 to get (14) WH a L WL I3Hy Relative returns are independent of capital (or the state of technology) because, according to the Cobb-Douglas assumption, technology is neutral. Substitution of Hy from equation 11 into equation 14 gives the relative returns to skilled and unskilled labor as (15) WH L(a. + p)(1 - a - P)7t O4(Afl). WL P)r The key result in equation 15 is that the higher the fraction of the technol- ogy of the North that the South has imitated, the lower the relative rate of return to human capital. This result occurs because the imitation technology uses human capital but not unskilled labor. Human capital enjoys an advan- 28 THE WORLD BANK ECONOMIC REVIEW, VOL I1, NO 1 tage as long as there is still a lot of technology to imitate. But the more imitation that has already taken place, the less the advantage enjoyed by skilled labor. As equation 15 shows, the only unknown influencing the relation between the returns to skilled and unskilled labor is the ratio AIB. I therefore analyze the influence of trade liberalization on relative returns with reference to figure 1. Trade influences equilibrium in this model by enlarging the technology set that becomes known in the South. In the formal model this can be shown by an enlargement in set A. Looking at the steady-state equilibrium in figure 1, it is clear that setA does not influence relative equilibrium returns. It influences only the set B that is imitated in the South, because the solution for the ratio AIB is independent of A. Thus, in the steady state a country-that trades has a higher level of technology and higher absolute returns to labor (from equations 10 and 12) than a country that trades less but has the same growth rate and the same relative returns to labor. The results are different during the adjustment following liberalization. Sup- pose that trade liberalization is accompanied by a once-and-for-all rise in the set -A that is known in the South. (Obviously, the increased knowledge will take place over time as foreign goods are experienced in the South, giving rise to more sluggish dynamic effects than the ones described here.) The effect of an increase in A is an increase in the rate of return to imitation. Producers in the South shift human resources out of production and into R&D. Because unskilled labor is not useful in the R&D technology, the relative returns to skilled labor rise immediately after trade liberalization. In figure 1, an increase in A for a given initial level of B shifts the (temporary) equilibrium to a point such as T. The relative return to human capital is higher at T because there has been an increase in the relative demand for it by firms engaged in R&D. Both the increase in set A and the shift of human capital to R&D lead to the introduction of more capital varieties in the South and to an increase in the rate of growth of its economy. Eventually, the number of capital varieties copied increases sufficiently to restore the ratio of known varieties, AIB, to its pretrade level. At this point the advantage of skilled labor over unskilled labor evapo- rates, and human capital moves back to production, as the demand for it drops in the R&D sector. The rate of growth of the economy of the South at this point drops down to the rate of growth of the economy of the North. In the diagram, for as long as the economy is above point E, both the rate of growth of the economy of the South is higher than the rate of growth of the economy of the North and the relative demand for human capital is higher than it was previ- ously. But when the economy reaches E, both values drop to their previous steady-state levels. Thus, the relative gain that skilled labor receives from trade liberalization is temporary, even for a fixed supply of human capital and unskilled labor. Human capital in this formulation has a comparative advantage only during the transition to a higher steady state, which takes place after trade liberal- Pissarides 29 ization. When the transition is complete, both human capital and unskilled labor gain from the opening up to trade, but relativities do not change. VI. CAVEATS AND EXTENSIONS The increase in the relative demand for human capital after trade liberaliza- tion is temporary because the imported production technology is neutral; that is, the new varieties of capital copied in the South are not better complements for human capital than for unskilled labor. I briefly consider here three cases in which richer results could be obtained. In the first case, the imported technology is not neutral but is biased in favor of skilled labor. This is a plausible assump- tion to make, given the increase in the rate of return to human capital-largely attributed to new technology-that has taken place in industrial countries. If the capital varieties invented in the North are more complementary to skilled than to unskilled labor, then it is natural that the varieties copied in the South will bias productivities in favor of skilled labor. If technology is biased in favor of skilled labor, the number of capital varie- ties B will enter the expression for relative wages, equation 14, with a positive sign.2 Then, for given supplies, the increase in the technology set of the South that results from trade liberalization will lead to a permanent and ever-increas- ing gain in the relative rate of return to human capital. Such gains will mirror any gains to skilled labor in industrial countries. Such a situation, however, cannot be a long-run steady state. Eventually, the relative supplies of factors will adjust endogenously in response to the higher rate of return to human capital. A new steady state will then be reached when the rise in the supply of human capital offsets exactly the increase in the relative demand for human capital that results from the biased technology. How long the gain to human capital will persist when the technology transfers are not neutral depends on the speed with which the supply of human capital catches up with the shift in relative demands. In the second case, human capital might enjoy a more long-lasting increase in relative demand if the production of capital goods is more human capital-intensive than the production of consumption goods (see Keller 1994). In the model pre- sented in this article, both consumption and capital goods are produced by the same technology. But if capital goods embody more up-to-date technology, it might be more reasonable to assume that there are two separate production sec- tors in the economy: one producing consumption goods and one producing capi- tal goods. If the capital goods sector requires more human capital than the con- sumption sector, then the fact that foreign capital varieties are copied at a faster rate after trade liberalization is another reason why the relative demand for hu- man capital will increase. Producers will shift human capital to the capital goods 2. I do not attempt to formalize this idea here, which can be done by replacing the Cobb-Douglas production function by a nested constant elasticity of substitution function, because of the additional complexity. 30 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I sector of the economy. Relative wages for skilled labor will then rise either until it becomes too expensive to copy any more capital varieties or until the domestic supply of human capital increases to offset the gains in relative wages. The first two cases considered in this section share a common property. They both imply that the technology transferred to the South makes the aggregate production function more skill-biased than it was before the liberalization of trade. The correction mechanism is an increase in the relative supply of skilled labor. Another plausible supply response, however, is likely to reinforce the in- crease in the relative returns to human capital in the period immediately follow- ing trade liberalization, even if the liberalization shock is neutral. It is common to think of developing countries as having a large potential supply of unskilled labor, which is either underemployed in agriculture or dis- couraged because of the lack of jobs and is waiting for an opportunity to enter formal employment. The response of this supply to a positive shock can be fast, certainly faster than the response of skilled labor, which needs to be trained. It follows that when trade liberalization increases the demand for both kinds of labor, the wages of skilled labor increase more than the wages of unskilled la- bor, because the initial rise in unskilled wages is checked by the response of the supply of unskilled labor. Eventually, however, the supply of educated labor catches up, restoring the wage differentials. Thus, the third case for an increase in the relative wages of skilled labor oc- curs even without biased technology transfers because the supplies of the two types of labor respond differently to an increase in aggregate labor demand. As in the other cases considered in this section, the situation just described is not likely to be a long-run equilibrium. Eventually, labor will train in response to the higher rate of return to skill and the relative advantage of skilled labor will erode away. The length of time that skilled labor can maintain an advantage over unskilled labor for the reasons outlined here depends on the speed with which unskilled labor "migrates" to the skilled sector, through education and training, relative to the speed with which it responds to the higher demand for unskilled labor. VII. CONCLUSIONS The widening of the skilled-unskilled wage differential that has been observed in developing countries that have liberalized trade is not a puzzle. Trade acts as a channel for the transfer of technology from industrial to developing countries. In the formal model described and solved in this article, the technology im- ported from the North into a developing country is neutral. It gives a temporary advantage to skilled labor over unskilled labor because the importation and assimilation process needs the services of skilled labor. Trade liberalization moves the economy of the developing country on to a permanently higher level of tech- nology, although its rate of growth in the new equilibrium is not likely to be higher than it was before liberalization. Pissarides 31 Skilled labor enjoys a relative advantage during the transition to the higher- level technology, which is reflected in higher relative wages for the duration of the transition. Of course, if the imported technology is biased in favor of skilled labor, skilled labor gains a permanent increase in relative demand. The correc- tion mechanism then is the increase in the supply of skilled labor to meet the higher demand. The response of the relative supply of skilled and unskilled labor to the trade liberalization might also cause a temporary widening in wage differentials. The supply of unskilled labor in a developing country is likely to be more elastic in the short run than the supply of skilled labor because of the existence of under- employed workers in agriculture or discouraged rent seekers waiting for new job opportunities. If this is the case, any overall increase in labor demand asso- ciated with trade liberalization will have a larger short-run impact on the wages of skilled labor than on the wages of unskilled labor, until the supply of skilled labor also increases to match the higher demand. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Coe, David T., and Elhanan Helpman. 1995. "International R&D Spillovers." Euro- pean Economic Review 39(May):859-87. Coe, David T., Elhanan Helpman, and Alexander W. Hoffmaister. 1995. "North-South R&D Spillovers." CEPR Discussion Paper 1133. Centre for Economic Policy Research, London. Processed. Grossman, Gene M., and Elhanan Helpman. 1991a. "Endogenous Product Cycles." Economic Journal 101(September):1214-29. . 1991b. Innovation and Growth in the Global Economy. Cambridge, Mass.: MIT Press. Hanson, Gordon H., and Ann Harrison. 1994. "Trade, Technology, and Wage In- equality: Evidence from Mexico." University of Texas at Austin. Processed. Keller, Wolfgang. 1994. "Absorptive Capacity: Understanding the Creation and Acqui- sition of Technology in Development." Yale University, New Haven, Conn. Pro- cessed. Krugman, Paul R. 1979. "A Model of Innovation, Technology Transfer, and Trade." Journal of Political Economy 87:(April):253-66. Nelson, Phillip. 1970. "Information and Consumer Behavior." Journal of Political Economy 78(March-April):311-29. Pessino, Carola. 1995. "The Labor Market during the Transition in Argentina." CEMA. Processed. Rivera-Batiz, Luis, and Paul M. Romer. 1991. "Economic Integration and Endogenous Growth." Quarterly Journal of Economics 106(May):531-56. Robbins, Donald J. 1994. "Earnings Dispersion in Chile after Trade Liberalization." Institute for International Development, Harvard University, Cambridge, Mass. Processed. 32 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO I .1995a. "Trade, Trade Liberalization, and Inequality in Latin America and East Asia: Synthesis of Seven Country Studies." Institute for International Development, Harvard University, Cambridge, Mass. Processed. . 1995b. "Wage Dispersion and Trade in Colombia: An Analysis of Greater Bogota: 1976-1989." Institute for International Development, Harvard University, Cambridge, Mass. Processed. Romer, Paul M. 1990. "Endogenous Technological Change." Journal of Political Economy 98(part 2, October):S71-S102. Tan, Hong, and Geeta Batra. 1995. "Technology and Industry Wage Differentials: Evi- dence from Three Developing Countries." Private Sector Development Department, World Bank, Washington, D.C. Processed. Wood, Adrian. 1995. "Does Trade Reduce Wage Inequality in Developing Countries?" Institute for Development Studies, University of Sussex. Processed. Openness and Wage Inequality in Developing Countries: The Latin American Challenge to East Asian Conventional Wisdom Adrian Wood The experience of East Asia in the 1960s and 1970s supports the theory that greater openness to trade tends to narrow the wage gap between skilled and unskilled work- ers in developing countries. In Latin America since the mid-1980s, however, in- creased openness has widened wage differentials. This conflict of evidence is prob- ably not the result of differences between East Asia and Latin America. Instead, the conflict is probably the result of differences between the 1960s and the 1980s, spe- cifically, the entry of China into the world market and, perhaps, the advent of new technology biased against unskilled workers. According to conventional wisdom, greater openness to trade in developing coun- tries not only increases efficiency but also reduces wage inequality. Openness boosts the relative demand for unskilled workers and hence narrows the gap in wages (and in unemployment rates) between unskilled and skilled workers. The experience of Latin America since the mid-1980s, however, has challenged this optimistic view. Greater openness to trade has been accompanied by rising rather than falling wage inequality. In contrast, the debate over trade and inequality in developed countries is now over the magnitude of the effects, with their direc- tion-adverse to unskilled workers-being largely agreed (Wood 1995). This article attempts to resolve the conflict of evidence in developing coun- tries. Section I outlines the theory underlying the conventional wisdom. Section II provides an overview of the empirical evidence, both in favor of and against the conventional wisdom. Section III considers explanations of the conflict of evidence based on differences between East Asia and Latin America. Section IV Adrian Wood is with the Institute of Development Studies at the University of Sussex. Financial support for the preparation of this article was provided by the U.K. Overseas Development Administration through its support of research at the Institute of Development Studies and by the World Bank. This article was originally prepared as a background paper for World Development Report 1995: Workers in an Integrating World. The author extends special thanks to Don Robbins, whose work motivated this article and who provided many unpublished papers and comments. The author acknowledges the valuable comments of Shahen Abrahamian, Riccardo Faini, Richard Freeman, Edward Leamer, J6rg Mayer, Patricio Meller, Jeffrey Sachs, Wolfgang Stolper, three anonymous referees, and participants in seminars at Harvard and the Massachusetts Institute of Technology, at conferences at the universities of Oxford, Reading, and the Auvergne (CERDI), and at a Centre for Economic Policy Research conference in La Corunia. C) 1997 The International Bank for Reconstruction and Development / THE WORLD BANK 33 34 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I considers explanations based on differences between the 1960s and the 1980s. Section V sums up. I. HECKSCHER-OHLIN THEORY The belief that increased openness reduces wage inequality in developing coun- tries rests on an apparently indisputable fact-that the supply of unskilled la- bor, relative to the supply of skilled labor, is larger in developing than in devel- oped countries (Wood 1994: table 3.1)-and on an often-disputed but widely used theory of trade-that of Heckscher and Ohlin. The Simplest Model Heckscher-Ohlin theory asserts that countries export goods that use inten- sively those factors of production that are relatively abundant at home and im- port goods that use intensively factors that are relatively scarce. Trade thus increases the demand for abundant factors, because of the expansion of export sectors, and reduces the demand for scarce factors, because of the contraction of import-competing sectors, with corresponding effects on factor prices. In developing countries, where unskilled labor is abundant and skilled labor is scarce, trade tends to raise unskilled wages and to lower skilled wages and hence to narrow the gap between them. To explain the effect on relative wages more precisely, consider a simple model with two countries (developed and developing), two factors (skilled and unskilled labor), and two goods (skill-intensive machinery and labor-intensive clothing). The developing country has a relatively large supply of unskilled labor, giving it a com- parative advantage in clothing, while the developed country has a relatively large supply of skilled labor, giving it a comparative advantage in machinery. Barriers to trade (transport costs and tariffs, for example) drive wedges be- tween the prices of goods in the two countries and may even result in no trade or autarky. In particular, the barriers keep the price of clothing lower in the devel- oping country than in the developed country and have the opposite effect for machinery. A reduction in barriers, and the resulting expansion of trade, would thus raise the price of clothing and lower the price of machinery in the develop- ing country. Such a change in relative domestic producer prices would raise the wage of unskilled workers relative to that of skilled workers. This link, known as the Stolper-Samuelson theorem, exists because Heckscher-Ohlin theory assumes that technofogy (that is, the production function for each good) is given. In other words, it assumes a fixed functional relationship between outputs of goods and inputs of factors, which (with no excess profits) implies a similarly fixed rela- tionship between the prices of goods and the wages of factors. The outcome can be illustrated in a type of supply-and-demand-curve dia- gram adapted from Leamer (1995). In figure 1, the downward-sloping line, dd, is the demand curve for unskilled labor that would prevail in a state of autarky. Wood 35 Figure 1. Effects of Openness on Relative Wages: Two Traded Goods Unskilled/skilled wages d D D Open economy demand ...... w0 for unskilled labor d Closed economy demand S3 B S1 52 for unskilled labor Unskilled/skilled workers Figure 2. Effects of Openness on Relative Wages: Many Traded Goods Unskilled/skilled wages d D Open economy demand for unskilled labor l_________ L________ - id Closed economy demand 53 B Si 52 for unskilled labor Unskilled/skilled workers 36 THE WORLD BANK ECONOMIC REVIEW, VOL. II, NO. 1 In the absence of trade, wages would be determined by the intersection of this demand curve with a supply curve (assumed for simplicity to be completely inelastic), whose position depends on the country's endowment of skilled and unskilled labor. With supply S2, say, as in a country with many unskilled work- ers, the relative wage of unskilled labor would be at the low level, wo. The demand curve in a country open to trade is the line DD. It crosses dd at B on the horizontal axis: if it had this skill supply ratio, even an open country would not trade. The developing country, which has a relatively large supply of unskilled labor and hence is a net exporter of clothing, must lie to the right of B (and the developed country must lie to the left of B). So for a developing coun- try, opening to trade shifts the demand curve in favor of unskilled labor (DD lies above dd) and narrows the gap in wages. With a skill supply ratio S2, the relative wage of unskilled labor would rise from wo to w2. The open-economy demand curve DD is an odd shape, with two downward- sloping segments separated by a flat segment in the middle: even to the right of B, there are two distinct segments in the developing-country range. The flat segment covers the range of skill supplies in which a trading economy would be diversified, in the sense of continuing to produce both clothing and machinery (albeit in different proportions than under autarky), as for example in a devel- oping country with skill supply S,. But a country with a high proportion of unskilled workers, as at S2, would not produce machinery; rather it would spe- cialize in clothing (a country with very few unskilled workers, as at S3, would specialize in machinery). Such specialization puts a country on a segment of the demand curve that slopes downward because increases in the relative supply of unskilled labor have to be absorbed by relative-wage-induced changes in the technique chosen to produce the single good. Trade raises the relative wage of unskilled workers, whether the outcome is diversified or specialized. But the effects on wages of subsequent changes in the relative domestic supply of labor differ. In a diversified country (as at Sj), rela- tive wages are fixed by relative world prices, at the level wl. Changes in the domestic labor supply, unless they are big enough to affect world prices, do not change relative wages: they alter only the composition of output and trade. By contrast, in a specialized country on a downward-sloping segment of DD, as at S2, changes in domestic labor supply do affect relative wages. For instance, an increase in the relative number of skilled workers would raise the relative wage of unskilled labor. More Goods, Countries, and Factors This model can be extended to include many goods (differentiated by skill intensity) and many countries (differentiated by skill supply) without funda- mentally altering the conventional theory that a reduction in trade barriers has a beneficial effect on the relative wages of unskilled workers in developing coun- tries. However, the inclusion of nontraded goods and additional factors can yield contrary results in special cases. Wood 37 Consider a rise in the number of traded goods. Figure 2 is drawn on the same principles as figure 1, but with six rather than two goods (and at least six coun- tries). The open-economy demand curve, DD, instead of having one flat seg- ment, has five, which alternate with downward-sloping segments. Countries whose relative skill supplies put them on a flat segment produce two goods, adjacent in skill intensity, while those on a downward-sloping segment produce only one good. All countries are specialized, because none produces all of the goods, but in those that produce two goods, as in a fully diversified country, small changes in labor supplies do not alter relative wages. However, larger changes in labor supplies do alter relative wages by moving the country to a different segment of DD. The effects of trade on relative wages remain the same as in figure 1. In devel- oping countries, to the right of B, a reduction in trade barriers shifts demand in favor of unskilled labor and narrows the skill differential in wages. The impact on wages is bigger for countries with a relatively large supply of unskilled labor (that is, the further the country lies to the right of B). For countries with inter- mediate skill supplies, in the vicinity of B, trade has a smaller effect on wages and could either increase or decrease them. This is because trade causes the most and the least skill-intensive sectors to contract: such countries export goods of medium skill intensity and import goods of high and low skill intensity from countries with higher and lower relative supplies of skilled labor. In reality, the number of traded goods of differing skill intensity is very large. It is thus reasonable, as well as convenient, to approximate DD by a continuous line (shown with dashes in figure 2). Dornbusch, Fischer, and Samuelson (1980) and Feenstra and Hanson (1995) present formal models that assume an infinity or continuum of traded goods. This multiple-goods formulation emphasizes two important points. First, in an open economy, the demand curve is more elastic than in a closed economy, so that changes in factor supplies have smaller effects on relative wages. Second, and contrary to the impression conveyed by the two- goods model in figure 1, even in an open economy, changes in factor supplies are generally likely to have some effect on relative wages. Realism also requires the inclusion of nontraded goods in the model. The low ratios of trade to output in many "open" economies indicate that transport costs are often more important barriers to trade than tariffs and quotas. Nontraded goods do not necessarily alter the conventional theory of the effects of greater openness on relative wages, but they do create the possibility of "per- verse" outcomes, arising from particular patterns of substitution in consump- tion between traded and nontraded goods.' Robinson and Thierfelder (1996) present a formal model in which the effect of changes in world prices on factor prices can be either normal or perverse, depending on the elasticity of substitu- tion in consumption. Consider, for example, a country with an abundant supply of unskilled labor in which labor-intensive nontraded goods are close substitutes for the more skill- 1. Edward Leamer suggested this possibility. 38 TFiE WORLD BANK ECONOMIC REVIEW, VOL II, NO. I intensive of two traded goods. Greater openness reduces the price of the more skill-intensive traded good, inducing consumers to buy more of it and less of the nontraded good. The resulting drop in demand for unskilled labor in the nontraded sector might more than offset the rise in demand for unskilled labor in the traded sector, leading to a fall rather than a rise in the relative wage of unskilled workers. Perverse outcomes are also possible (though by no means inevitable) when the model is extended to include more factors than skilled and unskilled labor. Consider, for example, infrastructure, a third factor that complements skilled labor in production. In principle a country might have a low ratio of skilled to unskilled labor, but a large supply of infrastructure and hence a comparative advantage in infrastructure-intensive traded goods. In such a country, greater openness to trade would increase the output of infrastructure-intensive goods, which also require a high ratio of skilled to unskilled labor and could thus raise the demand for (and the wages of) skilled relative to unskilled workers. II. OVERVIEW OF THE EMPIRICAL EVIDENCE This section summarizes the factor content of trade and time-series studies that support the conventional wisdom about the effects of greater openness on relative wages in developing countries. Two other types of supporting evidence are neglected here (but are covered in Wood 1994: 220-27): cross-country stud- ies, which are scarce, and simulation studies, which are vulnerable to disagree- ments about specification. This section also summarizes the recent time-series studies that challenge this conventional wisdom. The conflict of evidence thereby exposed provides the focus for the rest of the article. Factor Content of Trade Studies Studies of the factor content of trade calculate the amounts of skilled and unskilled labor used to produce a country's exports and compare these with the amounts of skilled and unskilled labor that would be required to produce do- mestically the goods that the country imports. If the ratio of skilled to unskilled labor is lower for exports than for imports, then increased openness to trade- more exports, more imports, and less import-competing production-should raise the relative demand for (and so the relative wages of) unskilled workers. Moreover, most calculations of factor content underestimate the impact of trade on the relative demand for skilled and unskilled labor, because they fail to allow for noncompeting imports (Wood 1994: 72-74, 87-91, 121-26). Table 1 encapsulates some relevant results from the classic study by Krueger and others (1981), which calculated the factor content of trade in broadly de- fined manufactures in a range of developing countries in the early 1970s. The table shows the ratio of the average skill intensity of exporting sectors to that of import-competing sectors, using two different measures of skill intensity. In ev- ery case, the ratio is less than unity, implying that exporting sectors are less skill-intensive than import-competing sectors, usually by a wide margin. Wood 39 Table 1. Skill Intensity of Trade in Manufactures in Selected Countries Ratio of average skill intensity of exporting sectors to that of import-competing sectors Numbers Wages Country Year measurea measureb Brazil 1971-72 - 0.92' Chiled 1966-68 - 0.26' Colombia 1973 0.53' 0.60O C6te d'lvoired 1972 0.62c - Hong Kongd 1973 0.51c - Indonesia 1971 0.55c 0.45' Tunisia 1972 Less than lc 0.65c Uruguay 1968 0.49f - Unweighted average 0.54 0.58 - Not available. a. Based on the ratio of skilled to unskilled employment. Skilled workers are white-collar, unskilled blue-collar, except in Hong Kong (where skilled workers are defined as professional) and C6te d'lvoire and Tunisia (where some blue-collar workers are in the skilled category). For Tunisia, the source does not give the precise number of skilled workers. b. Based on the average wage per worker; the precise nature of the calculation varies among the country studies. c. Trade with all partners. d. Includes home (nontraded) goods' indirect labor use. e. Trade with industrial countries only. f. Exports to industrial countries only, imports from all partners. Source: Krueger and others (1981), Krueger (1983); for further details, see Wood (1994: table 3.2). Fischer and Spinanger (1986, retabulated in Wood 1994: table 3.6) find the same result for twenty-one developing countries trading in manufactures with developed countries in 1965, 1973, and 1983: in sixty-one of the sixty-three cases, exports are less skill-intensive than imports. Similar results emerge for Taiwan around 1970 in Lee and Liang (1982: table 10.20), for India in the 1980s in Nambiar and Tadas (1994: table 10), and for several developing coun- tries in Bourguignon and Morrisson (1989: 282). All these studies are limited to manufacturing, but Kim and Vorasopontaviporn (1989) show that, for Thai- land, more trade would also increase the demand for low-wage agricultural labor. Applying a somewhat different approach to three Latin American countries in 1970-85, Londero and Teitel (1996) compare the skill intensity of fast- growing manufactured exports with that of the manufacturing sector as a whole, using sectoral wage levels as their measure of skill intensity. As shown in table 2, all three countries export manufactures of low, medium, and high skill intensity. In Argentina and Colombia, products of medium and low skill intensity account for three-quarters or more of the exports (discounting the value shares for Co- lombia, which are heavily influenced by two specific products). But in Venezu- ela, products of high skill intensity account for 40-60 percent of the total. The fact that almost all calculations of factor content in developing countries show exports to be less skill-intensive than imports is generally seen to support 40 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 Table 2. Skill Intensity of Manufactured Exports in Argentina, Colombia, and Venezuela, 1970-85 Country and Percentage based on skill intensity Number of products Value of exports Argentina High 25.0 18.0 Medium 52.1 53.6 Low 22.9 28.4 Colombia High 13.5 38.4 Medium 32.7 27.2 Low 53.8 34.4 Venezuela High 39.0 58.6 Medium 39.0 9.9 Low 22.0 31.5 Source: Londero and Teitel (1996: tables 5 and 6). the conventional view that greater openness benefits unskilled workers in par- ticular. However, analysts devote insufficient attention to variations among developing countries in the skill intensity of exports, which theory implies should be higher in better-educated (or middle-income) countries. And few analysts divide the imports of middle-income countries between those from higher-skilled trading partners and those from lower-skilled partners, which in theory should have different effects on the skill structure of domestic labor demand. The Krueger study is an honorable exception, but the pattern of world trade has changed significantly since the period to which it refers. Studies that attempt to relate the movement of relative wages over time in particular countries to changes in trade barriers-sometimes casually, some- times econometrically-provide an entirely different sort of evidence. Most such work has been done on countries in East Asia and Latin America; the results for the two regions are reviewed separately here, following a few remarks on meth- odology. The few studies of countries in other regions are severely limited by a lack of relevant data, for example, Bourguignon and Morrisson (1989) on Malawi and Morocco and Fontana (1994) on Ghana and Sri Lanka. There are also sev- eral studies of the impact of trade liberalization on unemployment (reviewed and extended by Edwards and Edwards 1995), but they rarely distinguish be- tween skilled and unskilled workers and are mainly concerned with transitional dislocation rather than with enduring changes in the labor market. Methodology All time-series studies encounter serious problems in the measurement of skill differentials in wages. The common division between white-collar (nonproduc- tion) and blue-collar (production) workers is treacherous, because the skill com- position of both groups changes over time (for example, fewer clerks and more engineers in the white-collar category). Data on wages by level of education are Wood 41 more satisfactory, but not always available. Nor is it easy to obtain accurate measures of changes in openness: information on trade barriers, particularly on the restrictiveness of nontariff barriers, is often inadequate, and changes in the ratio of exports or imports to gross domestic product (GDP) are an unreliable proxy. To measure the impact of changes in openness, it is also necessary to control for changes in domestic influences on relative wages, especially changes in the relative supply of skilled and unskilled labor and changes in labor market institutions that increase or reduce wage flexibility (but also changes in the sup- ply of factors that complement skilled or unskilled labor, in the composition of demand, and in cyclical forces). CHANGES IN THE RELATIVE SUPPLY OF SKILLED AND UNSKILLED LABOR. As is clear from figures 1 and 2, the impact on relative wages of increased openness (a shift of demand from dd to DD) varies with the level of supply of unskilled labor (the position of the vertical line S). The observed change in wages does not accurately measure the impact if increased openness coincides with a shift in supply. This measurement problem would be less serious if DD were infinitely elastic over the relevant range: the observed change in wages would be an accurate measure of the impact of increased openness at the initial level of supply (though not at the final level of supply, where the closed-economy wage would have been different). But if DD is less than infinitely elastic, the change in wages does not accurately measure the impact of increased openness even at the initial level of supply. Robbins (for example, 1995a) uses two methods to isolate the effects of sup- ply shifts in developing countries. His first method, the inner product test, asks whether relative wages and relative employment in the various skill categories moved in opposite directions (implying that supply shifts dominated) or in the same direction (implying that demand shifts dominated). In addition, to test the Heckscher-Ohlin prediction that changes in openness affect relative wages by altering the composition of output, Robbins decomposes changes in the skill structure of employment into between-industry and within-industry (between- occupation) components. However, this decomposition is vulnerable to the high level of aggregation of industries in most of his data because changes in product mix can occur within statistically defined industries. Robbins's second method of controlling for changes in supply is to estimate time series of implied shifts in the relative demand for skilled and unskilled labor, using information on relative wages and an equation of the form A A A (1) (d1/d2) = a(w1/w2) + (S1/S2) where d and s are demand and supply, the subscripts 1 and 2 denote skilled and unskilled labor, a is the elasticity of the relative demand curve, and A over a variable indicates a logarithm. Equation 1 is fundamentally an accounting iden- tity, but, if an assumption is made about the value of a, it can be used to deduce shifts in demand from observed changes in supply and wages. Attempts to use 42 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I this method are complicated by the fact that the absolute value of a is likely to become larger (and perhaps infinite) as a result of increased openness. Robbins (199Sb) thus experiments with different values of ca. He also shows that even in apparently open economies, relative wages are inversely related to relative sup- ply (with the significant exception of Taiwan after 1978), which supports the view that DD is usually less than infinitely elastic. CHANGES IN LABOR MARKET INSTITUTIONS THAT INCREASE OR REDUCE WAGE FLEXIBILITY. Changes in government policies might affect relative wages by altering legal minimum wages, the rights and powers of unions, or the extent of em- ployment in the public sector (where wage differentials tend to be narrower than in the private sector). For example, even if demand and supply tend to narrow the wage gap between skilled and unskilled workers, the gap might widen because of a decline in the minimum wage. Thus, changes in wage flexibility affect the accuracy of calculations based on equation 1 and more generally make it harder to measure the impact of changes in openness. In principle, using data on unemployment rates by skill level can solve this problem (because the combination of changes in relative wages and relative unemployment rates should indicate the direction of demand and supply shifts), but in practice such data are rarely available. East Asia Most of the time-series evidence from this region refers to the four tigers (Hong Kong, the Republic of Korea, Singapore, and Taiwan). Analysts always cite the experience of the four tigers as confirmation that greater openness raises the relative demand for unskilled workers. In fact, this evidence, reviewed in more detail in Wood (1994: 228-43), is by no means as clear-cut as commonly supposed. The data on relative wages contain gaps and deficiencies. In addition, few analyses have attempted to control for internal influences on the movement of relative wages. Nonetheless, most of the evidence supports the conventional view that the adoption of more outward-oriented policies increases the demand for workers with only a basic general education relative to the demand for workers with more education and skills. The evidence is also consistent with the theoretical prediction that a switch in trade regime causes a step (or once-and-for-all) change in the composition of demand, whose effects on skill differentials in wages ap- pear to be spread over a period of about ten years. In three of the four tigers, the gap in wages between skilled and unskilled workers narrowed during the decade in question (the 1960s in Korea and Taiwan, the 1970s in Singapore). In Hong Kong in the 19SOs (where the change in trade regime was rather different), the wage gap widened, but this was probably the result of a simultaneous, large increase in the relative supply of unskilled labor. Expansion of post-basic education is another possible explanation for the narrowing of skill differentials in wages. However, in all three of the cases in Wood 43 which differentials narrowed, the change in trade regime appears to have been at least partly responsible-either because the narrowing was faster than in ad- jacent decades or because the expansion of higher education was deliberately restrained during the period concerned. After the initial decades of export- oriented industrialization, the expansion of higher education further compressed wage differentials in all four economies (albeit with periods of widening wage differentials, which occurred for a variety of reasons). For more details on the experiences of the four tigers in the 1970s and 1980s, see also Fields (1994) and Robbins and Zveglich (1995). There is less evidence for other countries in East Asia. Robbins (1994a) finds persistent compression of wage differentials by level of education in Malaysia from 1973 to 1989, particularly between university graduates and less-educated workers. This continued in the early 1990s, with skilled and semiskilled blue- collar workers in manufacturing gaining relative to other groups. Robbins, ar- guing that this was not a case of increased openness, ascribes this compression mainly to the relative growth in the number of highly educated workers (the inner product test shows that supply shifts dominated demand shifts), perhaps augmented by the policy of discrimination toward indigenous Malays. How- ever, his data show that demand shifted in favor of less-skilled workers within industries, perhaps because of trade-related changes in product mix-the ex- pansion of export-oriented labor-intensive activities within the textiles and ma- chinery sectors. These two sectors account for a big and rising share of all em- ployment-up from 13 percent in 1984 to 16 percent in 1989 (Robbins 1994a: table 22). Moreover, between 1984 and 1989, the share of university-educated workers declined in textiles and was static in machinery, despite growth in the relative supply of such workers (Robbins 1994a: tables 1-3, 15; see also the information on Malaysia in Bourguignon and Morrisson.1989: 116-17). Robbins (1994b) also analyzes movements in wages by education level in the Philippines between 1978 and 1988, a period of modest trade liberalization ac- companied by a severe recession in 1982-86. Skill differentials in wages show no clear trend: they widened during the recession but then narrowed again. The initial, remarkably large relative supply of university graduates increased some- what, but inner product tests yield conflicting results about the relative impor- tance of demand and supply pressures on relative wages. Latin America Other studies by Robbins cover Argentina, Chile, Colombia, Costa Rica, and Uruguay. Table 3 summarizes the results for seven periods in which efforts were made to increase openness by lowering tariffs, easing quantitative import re- strictions, or devaluing the currency (which, in the presence of quantitative re- strictions, reduces antiexport bias by lowering the "quota premia" on-or tariff equivalents of-these restrictions). In Argentina in 1989-93, there was little if any net increase in openness, because the reduction of barriers was offset by appreciation of the exchange rate. It also seems unlikely that openness increased 44 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I Table 3. Effects of Increased Openness in Five Latin American Countries Predominant Relative Skill influence demand Changes in differentials (inner prod- for skill Country and years trade regime in wages tct test) (time series) Argentina (Buenos Aires) 1976-82 Barrier reduction Widened Demand Rising with appreciation 1989-93 Barrier reduction Narrowed Supply Falling with appreciation Chile (Santiago) 1974-79 Barrier reduction Widened Demand Rising with devaluation 1984-92 Devaluation Fluctuated Demand Rising Colombia Devaluation to 1989, Widened Demand Rising (seven cities) barrier reduction in 1985-94 1990-92 Costa Rica Barrier reduction Widened Supply, Rising 1985-93 and devaluation except in 1988-90 Uruguay (Montevideo) Barrier reduction Widened Demand Rising 1990-95 Source: For Argentina, Robbins, Gonzales, and Menendez (1995); for Chile, Robbins (1995a); for Colombia, Robbins (1996a); for Costa Rica, Robbins and Gindling (1997); for Uruguay and various other countries, Robbins (1995b, 1996b). in Chile during 1984-92, despite a large devaluation, because by that time al- most no quantitative restrictions were placed on imports. In all the five remaining periods (one in each country) in which openness did increase, skill differentials in wages (by level of education) widened, contrary to the conventional wisdom. The widening occurred from the mid-1970s to the early 1980s in Argentina and Chile and between the mid-1980s and the mid- 1990s in Colombia, Costa Rica, and Uruguay. In all five cases, the relative num- ber of skilled workers was rising;and thus the dominant influence on the change in wages was a rise in demand (as indicated by the results of the inner product tests, except for part of the period in Costa Rica). The time-series calculations (based on equation 1) confirm that the relative demand for skilled labor rose in all five of these episodes. Mexico is another Latin American country where skill differentials in wages widened after the mid-1980s in parallel with radical liberalization of the trade regime. The wage gap between nonproduction and production workers in manu- facturing widened between 1984 and 1990 (Feenstra and Hanson 1995; Hanson and Harrison 1995; Revenga and Montenegro 1995). Robbins (1996b) corrobo- rates this finding for wage differentials by level of education over the period 1987-93 and shows, by controlling for changes in relative supply, that the driv- ing force was a shift in relative demand. In some of these countries, liberalization of labor market institutions might explain the widening of skill differentials, but this possibility is rejected by the Wood 45 authors of these studies. Robbins (1996b) finds no correlation in Chile, Colom- bia, or Costa Rica between movements in relative wages (or in his relative de- mand time series) and changes in the legal minimum wage. (By contrast, Bell 1995 finds that minimum wages do have an effect in Colombia, but her study focuses on average rather than relative wages and ends earlier.) Robbins also argues that reduced union power is unlikely to have been the cause, with the possible exception of Argentina. In Mexico, the widening of skill differentials during 1984-90 coincided with a steep decline in the real minimum wage, but Feenstra and Hanson (1995) and Hanson and Harrison (1995) reject this expla- nation, citing Bell (1995), who finds that even in 1984 the average production worker was paid more than the minimum in 97 percent of the manufacturing plants in the data set. These rejections are not entirely convincing. In particular, the widening of wage differentials in Chile in the late 1970s, following the military overthrow of the Allende government, did coincide with a severe curtailment of union activity (Edwards and Edwards 1995) and, moreover, simply restored differentials to the levels prevailing in the late 1960s (Robbins 1995a: figure 5). The widening of wage differentials in Argentina in the late 1970s also occurred after a military coup and in parallel, as Robbins points out, with a decline in the union move- ment. The counter-evidence for Mexico also needs to be treated with caution: because wages vary among production workers, the legal minimum wage could affect the average wage even if it set only the lowest wage (well below the aver- age). Moreover, Bell (1995) finds that the legal minimum wage did affect the wages of many workers in the informal sector. Assessment of the Time-Series Evidence All the time-series studies-both those that support and those that contradict the conventional wisdom that increased openness reduces wage inequality in developing countries-are open to doubts of one sort or another. But it is hard to avoid the conclusion that there is a genuine conflict of evidence: in some countries and periods, increased openness does appear to have caused a narrow- ing of skill differentials, but in others the opposite seems to have happened (even allowing for influences other than trade). The conflict is particularly sharp between, on the one hand, the evidence from East Asia in the 1960s (Korea and Taiwan) and 1970s (Singapore and probably Malaysia) and, on the other hand, the evidence from Latin America in the late 1980s and early 1990s. Increased openness was also associated with widening wage differentials in Argentina and Chile in the late 1970s, but in both these cases labor market liberalization under military rule is a plausible alterna- tive explanation. It is harder to explain away the rise in wage inequality in Co- lombia, Costa Rica, Mexico, and Uruguay since the mid-1980s. For the rest of this article, I assume that there was a genuine difference be- tween the effects of greater openness on wage inequality in East Asia in the 1960s and 1970s and its effects in Latin America in the late 1980s and early 46 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I 1990s. Nonetheless, I acknowledge the various residual doubts and the need for further research on the countries concerned, which might erase the apparent conflict. My assumption not only allows the following pages to focus on pos- sible reasons for the difference but also suggests a convenient framework for the discussion. The conflict involves two regions in two periods: the next section thus asks whether it might be caused by differences between East Asia and Latin America, and the following section asks whether it might be caused by differ- ences between the 1960s-70s and the 1980s-90s. III. DIFFERENCES BETWEEN EAST ASIA AND LATIN AMERICA Some important differences between the two regions have already been taken into account in the preceding review of evidence: the much faster growth in the supply of skilled labor in East Asia and the more regulated nature of labor mar- kets in most Latin American countries. This section examines two other possi- bly important areas of differences: the availability of natural resources and the choice of policy instruments for increasing openness. Natural Resource Endowments In some respects, East Asia and Latin America are similarly endowed: both are well educated by comparison with Africa and South Asia, and both have fairly well-developed infrastructure. However, Latin America is far better en- dowed with natural resources than East Asia and consequently has a compara- tive advantage in primary products (including processed products, such as canned fish). By contrast, East Asia's comparative advantage lies in narrowly defined manufactures (such as clothing) that require few inputs of local natural resources. The difference in comparative advantage is clear from data on land per worker in the two regions and on the composition of their exports (Owens and Wood 1995: figures 6 and 7). The difference is also clear from the sectoral pattern of the output response to increases in openness. In East Asia, the growth of exports was concentrated in manufacturing. In Latin America, the gains were mainly in primary and processed primary exports, with other (nonprimary processing) manufacturing often shrinking, except in the parts of Mexico adjacent to the United States. As explained in section I, the introduction of a third factor, such as land, into the Heckscher-Ohlin model could in principle cause greater openness to widen wage differentials even in a country with a relatively low, ratio of skilled, to unskilled labor. For Latin America, the mechanism would have to be that skill and natural resources were complementary inputs, so that the expansion of primary output raised the relative demand for skilled labor. In practice, however, primary production is not generally skill-intensive, in com- parison with import-competing manufacturing. Mining and oil refining do require a highly skilled labor force but account for only a small share of total employment. Agriculture, a far larger employer, has a low ratio of skilled to Wood 47 unskilled workers.2 And most primary processing is not skill-intensive. Ex- cept for Hong Kong, most of the manufactured exports in Krueger's factor content study (summarized in table 1) were processed primary products (as defined in Owens and Wood 1995). Londero and Teitel (1996: table 5) show that, in Argentina and Colombia, only 13 percent of manufactures intensive in natural resources are of high skill intensity, and in Venezuela, only 30 percent are. The cross-country regressions in Owens and Wood (1995) sug- gest that primary processing needs a literate labor force but uses a smaller proportion of highly skilled workers than other sorts of manufacturing. It remains possible that some more complicated process, involving nontraded sectors as well as natural resources, could explain the difference in wage out- comes between the two regions.3 For instance, it might be the case that all manu- factures were import substitutes in Latin America, but only skill-intensive manu- factures were import substitutes in East Asia. In that case, nontraded sectors (of a given absolute skill intensity) might be more skill-intensive than import- competing sectors in Latin America and less skill-intensive than import- competing sectors in East Asia. Hence, if greater openness (through substitution in consumption) caused nontraded as well as export sectors to expand (and import-competing sectors to contract), the net effect might be to increase the relative demand for skilled labor in Latin America, but to decrease it in East Asia. This specific chain of causation may seem far-fetched, but the correlation between the regional differences in natural resources and in wage outcomes is strong enough to warrant further empirical research. It might also be worth checking whether the wage data for Latin America are hiding increased demand for unskilled workers because of gaps in the coverage of primary sectors and rural areas. The data for Argentina, Chile, Colombia, and Uruguay cover only cities, and most of the data for Mexico are limited to manufacturing. However, the Costa Rican data, which show much the same wage pattern, are nationwide. Moreover, even in the other countries, the data include factory-based primary processing. And if labor markets work properly, changes in relative wages in sectors that are not covered by the surveys should be transmitted into the sec- tors that are covered. Trade Policy Instruments The opening to trade in East Asia, particularly in Korea and Taiwan in the 1960s, was achieved mainly by increasing incentives for exporters, while keep- ing fairly high levels of protection against imports in most sectors. Most of the opening to trade in Latin America, by contrast, involved large reductions in barriers to imports. 2. However, a referee suggested that in Latin America the least skill-intensive agricultural subsector, staple food, may be an import substitute, so that increased openness may raise the relative demand for skilled labor within agriculture. 3. Jeffrey Sachs suggested this. 48 THE WORLD BANK ECONOMIC REVIEW, VOL. II, NO. 1 From a macroeconomic viewpoint, the two approaches are equivalent: it does not matter whether bias against exporting is reduced by giving subsidies to ex- ports or by lowering barriers to imports. However, because neither the subsidies nor the barriers are uniform across sectors, their effects on the sectoral structure of output, and so on the skill composition of the demand for labor, may be different. Labor-intensive manufacturing subsectors such as clothing and foot- wear were initially highly protected in most Latin American countries (as in developed countries), so that making barriers against manufactured imports lower and more uniform tended to hurt unskilled workers more than skilled workers. For example, in Chile in 1974, textiles and footwear were the second and third most protected manufacturing sectors, with effective rates of protection over 200 percent (Edwards and Edwards 1995: table 2). And in Mexico, Revenga and Montenegro (1995: table 2) show that the nominal tariff on apparel and footwear in 1985 was among the highest in manufacturing, although it stands out less in the calculations of Hanson and Harrison (1995: table 2). The differences in trade regimes deserve further investigation (and compari- son with the liberalization of imports in Korea and Taiwan after the mid-1980s), but this is not a promising explanation of the difference in wage outcomes be- tween the two regions. If Latin America had a comparative advantage in un- skilled labor-intensive goods, manufactured or primary, then the effects of im- port liberalization in sectors such as clothing and footwear should have been more than offset by the growth of exports in (these or other) labor-intensive sectors. And if in fact the outcome of increased openness was a rise in the rela- tive demand for skilled workers because Latin America did not have a compara- tive advantage in unskilled labor-intensive sectors, the explanation must be sought somewhere other than in the details of its trade regime. IV. DIFFERENCES BETWEEN THE 1960s-70s AND THE 1980s-90s This section examines the possibility that the observed difference in wage outcomes occurred because the world changed in some respect between the 1960s and 1970s, the period covered by the East Asian time-series studies, and the late 1980s and 1990s, the period covered by the most convincing Latin American evidence. One such change was the entry of the largest low-income countries into world markets in the 1980s. Another was perhaps a shift in the skill bias of technology. These two changes seem to be the most promising explanations in this category: others that seem less promising are the debt crisis and greater international mobility of labor and capital. However, Feenstra and Hanson (1995) present a model in which capital flows could explain the widening of wage dif- ferentials in Mexico. Entry of Large Low-Income Exporters Half the world's population, and an even higher proportion of the world's unskilled workers, live in five low-income Asian countries: Bangladesh, China, Wood 49 India, Indonesia, and Pakistan. In the 1960s and 1970s, all five countries were largely closed to trade, and thus their workers did not form part of the effective world labor supply. By the mid-1980s, these countries were all opening to trade, led by Indonesia and China, with the South Asian countries also making some progress. Their manufactured exports grew rapidly: in the six years between 1987 and 1993, developed countries' imports of manufactures from low-income countries rose nearly four times, from $28 billion to $110 billion (about one- third of all manufactured imports from developing countries), while imports from middle-income countries rose less than 50 percent.4 This opening of the low-income half of the world is likely to have altered the comparative advantage of middle-income countries, whose ratio of skilled to unskilled workers is above the global average, though below that of developed countries. The economic world of the 1960s and 1970s consisted effectively only of developed and middle-income countries, and thus the middle-income countries had a comparative advantage in goods of low skill intensity. In the 1980s, when low-income Asia started to realize its own comparative advantage in goods of low skill intensity, the comparative advantage of middle-income countries shifted to goods of intermediate skill intensity. Rapid accumulation of skills in the East Asian economies that opened in the 1960s also helped to shift their comparative advantage toward goods of intermediate skill intensity. The effect on relative wages in a middle-income country of opening to trade is thus likely to have changed over time. In the 1960s, increased openness would have raised the relative wage of unskilled workers, because it would have caused sectors of above-average skill intensity to shrink and sectors of below-average skill intensity to expand. In the 1980s, however, greater openness had conflict- ing effects on relative wages. Greater openness caused the contraction of sectors both of high skill intensity (replaced by imports from developed countries) and of low skill intensity (replaced by imports from low-income countries). The net effect might be in either direction, but greater openness could widen the wage gap between skilled and unskilled workers, as happened in the middle-income countries of Latin America in this period. In theoretical terms, an increase in the effective relative world supply of un- skilled labor lowers the relative world market price of unskilled-labor-intensive goods and thus alters the position of the open-economy demand curve. Figure 3 illustrates this process in the two-goods model introduced in figure 1: the lower relative world price of clothing is shown (with a dashed line) as a downward shift of the horizontal segment of DD. Figure 4, the case of a many-goods model (as in figure 2), approximates the open-economy demand curve by a continuous downward-sloping line. The north- west quadrant depicts the world of the 1960s (with diagonal ddl), consisting of developed countries (with unskilled to skilled labor ratios in the range OA) and 4. Data are from World Bank (1989: indicators table 17; 1995: indicators table 16, with Indonesia moved into the low-income group for consistency). The figure for the middle-income group is slightly inflated by the reclassification of countries in Eastern Europe and the former Soviet Union. 50 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Figure 3. Entry of Low-Income Asia: Two Traded Goods Unskilled/skilled wages d D D Open economy demand for unskilled labor S________ S______________________ - d Closed economy demand S3 SI 52 for unskilled labor Unskilled/skilled workers Figure 4. Entry of Low-Income Asia: Many Traded Goods Unskilled/skilled wages d D Open economy / demand for unskilled D2 labor, 1960s Closed economy demand for unskilled labor,_________s D2 Open economy demand for unskilled labor, 1980s 0 A B d2 Closed economy demand for unskilled labor, 1980s Unskilled/skilled workers Wood Si Figure 5. Price of Developing-Country Manufactured Exports, Relative to Developed-Country Exports of Machinery, Transport Equipment, and Services, 1975-95 (index: 1980 = 100) Price of developing-country manufactured exports 120 - 110-- 100 90 80 70 60 - i l l l l :l l E . . l E 1975 1980 1985 1990 1995 Note: Manufactured export prices are unit value indexes from U.N. trade data. Service price index refers to the United Kingdom only and has a weight of 0.27 in the combined developed-country index. Source: Update of Minford, Riley, and Nowell (1995: figure 6). middle-income countries (in the range AB), but excluding the low-income Asian countries (with unskilled to skilled labor ratios beyond B). The open-economy demand curve in the 1960s is the line DD1. The whole figure depicts the world of the 1980s. The closed-economy demand curve extends to d2, and the open- economy demand curve pivots down to DD2 (drawn for simplicity to cross dd2 at B). For the middle-income countries, with skill supply ratios between A and B, DD, is above ddl, but DD2 is below it, so that opening to trade would have reduced wage inequality in the 1960s but increased it in the 1980s. The coincidence in the late 1980s of expanding exports from low-income Asia with the opening to trade of Latin American countries provides a plausible explanation for the widening of wage differentials in the latter countries. This expansion of exports did apparently cause a downward shift in the open-economy demand curve from the mid-1980s to the mid-1990s, when, as shown in figure 5, the world price of unskilled-labor-intensive goods dropped relative to that of more skill-intensive goods.' There is also anecdotal and case study evidence that labor-intensive Latin American manufacturing sectors suffered as a result of S. Patrick Minford kindly provided these data. For a discussion of earlier movements of a similar terms-of-trade index, see Sarkar and Singer (1991) and the October 1993 issue of World Development. 52 THE WORLD BANK ECONOMIC REVIEW, VOL. II, NO. I competition from Asian imports both domestically and in third markets such as the United States (for example, see Kaplinsky 1993). A downward shift of the open-economy demand curve should also have af- fected relative wages in other countries that were already open to trade (through the Stolper-Samuelson linkage between product and factor prices), but only in those that produced both labor-intensive and other goods. Figure 3, in the simple two-goods case, thus shows that the downward shift of DD reduces (from w, to w1 ,) the relative wage of unskilled workers in the country with skill supply Si, which produces clothing and machinery, but does not change relative wages in countries that specialize in one of the goods (S2 in clothing, S3 in machinery), since, with only one good produced, there can be no change in domestic relative producer prices. (However, the change in world prices does alter relative con- sumer prices and thus the real wages of all workers in the specialized countries, downward in S2 and upward in S3.) The distinction between diversified and spe- cialized countries is blurred in figure 4, but however the diagrams are drawn, open middle-income economies-with intermediate unskilled to skilled labor ratios-are the ones most likely to produce both labor-intensive and more skill- intensive goods and so to have experienced external pressure for wider wage differentials. Thus, an indirect test of the hypothesis that wage differentials widened in Latin American countries that opened to trade because of the simultaneous ex- pansion of exports from low-income Asia would be to see whether wage differ- entials also widened during the same period in other, already open, middle- income countries. The time-series studies reviewed earlier reveal a mixed picture for such countries: in Hong Kong, wage differentials widened throughout the 1980s; in Taiwan and Singapore, they widened in the early 1980s but narrowed in the late 1980s; in Korea and Malaysia, they narrowed throughout the decade; and in Chile, they fluctuated (Robbins 1995b: figure 3; Wood 1994: 228-43). The estimated relative demand for skilled labor also moved in different direc- tions in the three countries for which data are available-Chile, Malaysia, and Taiwan (Robbins 1995b: figure 6). The test is thus inconclusive on this small set of fairly open middle-income countries. The hypothesis that the difference between the experience of East Asia in the 1960s and Latin America in the 1980s was caused by the entry of low- income Asian countries into world markets clearly requires further research. Such research should include a larger sample of open middle-income coun- tries and should also examine the opening low-income countries themselves (whose wage differentials should have narrowed). Even more vitally, further research would need to look closely at the consistency of this hypothesis with the experience of the middle-income countries that have opened since the mid-1980s (again with a larger sample). In particular, it should use fac- tor content analysis to check whether the reduction of trade barriers caused sectors of intermediate skill intensity to expand and sectors of low skill in- tensity to contract. Wood 53 Skill-Biased Technical Progress Another possible reason why opening to trade had different effects on wage inequality in the 1980s than in the 1960s is that world technology changed be- tween these periods in a way that raised the relative demand for skilled labor. This explanation is put forward by Robbins (for example, 1995a), who calls it "skill-enhancing trade." He argues that increased openness in a developing coun- try affects the skill structure of labor demand in two ways: by altering the sectoral composition of production (as in Heckscher-Ohlin theory) and by changing the production technology available-through increased imports of advanced capi- tal goods, for example, or through opportunities for exporters to learn from foreign buyers and be exposed to foreign markets. The net impact depends on the size of these two effects and on the exact nature of the difference between domestic and world technology, but the effect on relative wages could be con- trary to the predictions of standard Heckscher-Ohlin theory, which assumes that all countries always have access to the same technology. Moreover, the real wage of unskilled workers could rise, even if their relative wage falls, whereas under strict Heckscher-Ohlin assumptions, including constant returns to scale, real and relative wages should move in the same direction. The skill-enhancing trade explanation is depicted in figure 6, in which there are two pairs of closed-economy and open-economy demand curves: dldl and D1Dj are based on the use of 1960s technology, d2d2 and D2D2 are based on the use of 1980s technology. In the figure, technology is assumed to have changed over time in a biased way, requiring a generally lower ratio of unskilled to skilled labor. Another assumption is that the new technology is available only to open economies, so that a country remaining closed to trade over this period would have continued to use 1960s technology. For most developing countries (to the right of B), opening to trade would thus have had different effects on wage inequality in the 1960s (decreasing it, by moving from dldl to D1Dj) than in the 1980s (increasing it, by moving from dldl to D2D2). Even in the 1980s, the Heckscher-Ohlin tendency for openness to reduce inequality remains-rever- sion to a closed economy, with 1980s technology and demand curve d2d2, would aggravate the increase in wage inequality-but it is dominated by the effect of the change in technology. The skill-enhancing trade explanation has considerable plausibility. It is likely that increased openness does alter the availability of technology. It is also widely believed that technical progress over the past couple of decades has in fact been biased against unskilled workers and that this is why wage inequality has in- creased in mostdeveloped countries. Moreover, Robbins (1996b) provides econo- metric support for this explanation: pooling the time series for six of the devel- oping countries in his sample, he finds the relative demand for skilled labor to be positively correlated with the rate of growth of GDP and with the ratio of the imported capital stock to GDP, interpreting both variables as proxies for access to new technology. 54 THE WORLD BANK ECONOMIC REVIEW, VOL. II, NO. I Figure 6. Effects of Openness and Tecbnology Transfer Unskilled/skilled wages di D2 N.~~~~~~~~~~N N.~~~~~~~~~~~~~~~~D N.~ ~ ~ ~ ~ ~ ~~~~~~~~D B d2 d Unskilled/skilled workers Note d,d, and d2d2 are the closed economy demand curves for unskilled labor based on 1960s and 1980s technology, respectively. D,D, and D2D2 are the open economy demand curves for unskilled labor based on 1960s and 1980s technology, respectively. However, there is also considerable room for doubt about this explana- tion. The opening countries were not completely cut off from new technol- ogy. There is little hard evidence of an autonomous skill-using bias in recent technical progress, and there is another plausible explanation-increased openness to trade with developing countries-for the rise in inequality in developed countries (Wood 1995). And, as mentioned earlier, there is no consistent evidence of a rise in wage inequality, or in the relative demand for skilled labor, in open middle-income economies. In addition, Robbins's econo- metric results are open to alternative interpretations. Again, further research is needed, both to determine whether the open-economy demand curve has shifted down over time-an assumption common to the technological bias and the entry of low-income Asia hypotheses-and to establish which hy- pothesis provides a better explanation of this shift (including examination of movements in the real wages of unskilled workers, concerning which the two hypotheses make different predictions). Wood 55 V. SUMMARY AND CONCLUSIONS A substantial amount of empirical evidence supports the conventional wis- dom that increased openness to trade in developing countries tends to raise the demand for unskilled, relative to skilled, labor and thus to reduce wage inequal- ity. However, some recent evidence contradicts the conventional wisdom. In particular, there is a conflict in the time-series evidence between the experience of East Asia in the 1960s and 1970s, which is consistent with the conventional wisdom, and the experience of Latin America in the late 1980s and early 1990s, where increased openness appears to have widened rather than narrowed skill differentials in wages. Possible explanations fall into two classes: differences between East Asia and Latin America and differences between the 1960s-70s and the 1980s- 90s. Preliminary assessment of more specific explanations within these classes suggests that the reason is more likely to be the difference between the two time periods than the difference between the two regions. East Asia and Latin America differ in several respects, not least their endowment of natural re- sources, but it is difficult to establish a convincing causal link between any of these differences and the difference in the impact of increased openness on wage inequality. By contrast, there are two plausible reasons why the effects of increased open- ness on wage inequality might have differed between the earlier and the later periods. First, the entry of China and other large low-income Asian countries into the world market for labor-intensive manufactures in the 1980s shifted the comparative advantage of middle-income countries into goods of medium skill intensity. As a result, increased openness in middle-income countries reduced the relative demand for unskilled workers by causing sectors of low skill inten- sity to contract. Second, technical progress between the 1960s and the 1980s was biased against unskilled workers. The available evidence does not permit any strong conclusion as to which of these two period-difference explanations is the right one. More generally, the slen- der empirical basis of this whole exercise should be emphasized. The sample of developing countries for which there is recent evidence on trade and wage inequal- ity is extremely limited-it not only is small but is also confined to East Asia and Latin America, covers no low-income countries, and covers no countries from Af- rica or South Asia. Within each country, too, there are gaps and other problems with the data and analysis. More research is needed to confirm whether the impact of increased openness on wage inequality has changed in recent years, as well as to determine how and why such change might have occurred. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. 56 THE WORLD BANK ECONOMIC REVIEW, VOL. t1, NO. I Bell, Linda A. 1995. "The Impact of Minimum Wages in Mexico and Colombia." Policy Research Working Paper 1514. Policy Research Department, World Bank, Wash- ington, D.C. Processed. Bourguignon, Francois, and Christian Morrisson. 1989. External Trade and Income Distribution. Paris: Organization for Economic Cooperation and Development. Dornbusch, Rudiger, Stanley Fischer, and Paul Samuelson. 1980. "Heckscher-Ohlin Trade Theory with a Continuum of Goods." Quarterly Journal of Economics 95(2):203-24. Edwards, Alejandra Cox, and Sebastian Edwards. 1995. "Trade Liberalization and Unemployment: Evidence from Chile." California State University and University of California, Los Angeles. Processed. Feenstra, Robert, and Gordon Hanson. 1995. "Foreign Investment, Outsourcing, and Relative Wages." In Robert Feenstra, Gene Grossman, and Douglas Irwin, eds., Po- litical Economy of Trade Policy: Essays in Honor of Jagdish Bhagwati. Cambridge, Mass.: MIT Press. Fields, Gary. 1994. "Changing Labor Market Conditions and Economic Development in Hong Kong, the Republic of Korea, Singapore, and Taiwan, China." The World Bank Economic Review 8(3):395-414. Fischer, Bernhard, and Dean Spinanger. 1986. "Factor Market Distortions and Export Performance: An Eclectic Review of the Evidence." Kiel Working Paper 259. Institut fur Weltwirtschaft an der Universitat Kiel. Processed. Fontana, Marzia. 1994. "Trade Liberalisation and Income Distribution in Developing Countries." M. Phil. diss., Institute of Development Studies, University of Sussex. Processed. Hanson, Gordon H., and Ann Harrison. 1995. "Trade, Technology, and Wage In- equality." NBER Working Paper 5110. National Bureau of Economic Research, Cam- bridge, Mass. Processed. Kaplinsky, Raphael. 1993. "Export Processing Zones in the Dominican Republic: Trans- forming Manufactures into Commodities." World Development 21(11):1851-65. Kim, K. S., and Pompen Vorasopontaviporn.' 1989. "Foreign Trade and the Distribu- tion of Income in Thailand." Working Paper 124. Helen Kellogg Institute for Inter- national Studies, University of Notre Dame, Notre Dame, Ind. Processed. Krueger, Anne 0. 1983. Trade and Employment in Developing Countries. Vol. 3: Syn- thesis and Conclusions. Chicago: University of Chicago Press. Krueger, Anne O., Hal B. Lary, Terry D. Monson, and Narongchai Akrasanee, eds. 1981. Trade and Employment in Developing Countries. Vol. 1: Individual Studies. Chicago: University of Chicago Press. Leamer, Edward E. 1995. "A Trade Economist's View of U.S. Wages and Globaliza- tion." In Susan Collins, ed., Imports, Exports, and the American Worker. Washing- ton, D.C.: Brookings Institution. Lee, T. H., and K. S. Liang. 1982. "Taiwan." In Bela Balassa, ed., Development Strategies in Semi-Industrial Economies. Baltimore, Md.: Johns Hopkins University Press. Londero, Elio, and Simon Teitel. 1996. "Industrialisation and the Factor Content of Latin American Exports of Manufactures."Journal of Development Studies 32(4):581-601. Minford, Patrick, Jonathan Riley, and Eric Nowell. 1995. "The Elixir of Growth: Trade, Non-Traded Goods, and Development." CEPR Discussion Paper 1165. Centre for Economic Policy Research, London. Processed. Wqod 57 Nambiar, R. G., and Gopal Tadas. 1994. "Is Trade Deindustrialising India?" Eco- nomic and Political Weekly 15(October):2741-46. Owens, Trudy, and Adrian Wood. 1995. "Export-Oriented Industrialisationlthrough Primary Processing?" IDS Working Paper 19. Institute of Development Studies, Uni- versity of Sussex. Processed. Revenga, Ana, and Claudio Montenegro. 1995. "North American Integration and Fac- tor Price Equalization: Is There Evidence of Wage Convergence between Mexico and the United States?" In Susan Collins, ed., Imports, Exports, and the American Worker. Washington, D.C.: Brookings Institution. Robinson, Sherman, and Karen Thierfelder. 1996. "The Trade-Wage Debate in a Model with Nontraded Goods: Making Room for Labor Economists in Trade Theory."TMD Discussion Paper 9. Trade and Macroeconomics Division, International Food Policy Research Institute, Washington, D.C. Processed. Robbins, Donald. 1994a. "Malaysian Wage Structure and Its Causes." Harvard Insti- tute for International Development, Cambridge, Mass. Processed. . 1994b. "Philippine Wage and Employment Structure, 1978-93." Harvard In- stitute for International Development, Cambridge, Mass. Processed. . 1995a. "Earnings Dispersion in Chile after Trade Liberalization." Harvard Institute for International Development, Cambridge, Mass. Processed. . 1995b. "Trade, Trade Liberalization, and Inequality in Latin America and East Asia: Synthesis of Seven Country Studies." Harvard Institute for International De- velopment, Cambridge, Mass. Processed. .1996a. "Stolper-Samuelson (Lost) in the Tropics: Trade Liberalization and Wages in Colombia 1976-94." Harvard Institute for International Development, Cambridge, Mass. Processed. .1996b. "HOS Hits Facts: Facts Win. Evidence on Trade and Wages in the Devel- oping World." Harvard Institute for International Development, Cambridge, Mass. Processed. Robbins, Donald, and Thomas Gindling. 1997. "Educational Expansion, Trade Liberalisation, and Distribution in Costa Rica." In Albert Berry, ed., Poverty, Eco- nomic Reform and Income Distribution in Latin America. Boulder, Colo.: Lynne Rienner Publishers. Robbins, Donald, Martin Gonzales, and Alicia Menendez. 1995. "Wage Dispersion in Argentina, 1976-93: Trade Liberalization amidst Inflation, Stabilization, and Overvalu- ation." Harvard Institute for International Development, Cambridge, Mass. Processed. Robbins, Donald, and Joseph Zveglich. 1995. "Skill-Bias in Recent Taiwanese Growth." Harvard Institute for International Development, Cambridge, Mass. Processed. Sarkar, Prabirjit, and Hans Singer. 1991. "Manufactured Exports of Developing Coun- tries and Their Terms of Trade since 1965." World Development 19(4):333-40. Wood, Adrian. 1994. North-South Trade, Employment, and Inequality: Changing For- tunes in a Skill-Driven World. Oxford: Clarendon Press. . 1995. "How Trade Hurt Unskilled Workers." Journal of Economic Perspec- tives 9(3):57-80. World Bank. 1989. World Development Report 1989: Financial Systems and Develop- ment. New York: Oxford University Press. . 1995. World Development Report 1995: Workers in an Integrating World. New York: Oxford University Press. Technology and Firm Size-Wage Differentials in Colombia, Mexico, and Taiwan (China) Hong Tan and Geeta Batra In many economies, studies have found large wage differentials not accounted for by workforce characteristics, collective bargaining, or market power. Researchers attribute these differentials to either unobserved worker quality or pay incentives designed to elicit worker effort. This article finds empirical support for an alterna- tive explanation: These wage differentials result from firms' technology-generating activities. Using firm-level data from Colombia, 'Mexico, and Taiwan (China), the article compares the effects of research and development, worker training, and ex- ports by employers on the wages of skilled and unskilled workers. The results sug- gest that technology investments lead to large wage premiums for skilled workers but not for unskilled workers: These wage premiums are primarily the result of investments in research and development and in training, while exporting is rela- tively less important except in Colombia. Technology plays a vital role in shaping the interfirm structure of wages in de- veloping economies. How do the technology-generating activities of firms- broadly defined to include investments in research and development (R&D), for- eign technology and know-how licenses, worker training, and exporting-affect the size structure of wages? Are the effects of these firm investments symmetri- cal with regard to all workers, or are they biased toward skilled workers as suggested by the skill-biased hypothesis of technical change? Which of these investment activities has the largest effects on wages? Large differentials in wages paid to ostensibly similar workers characterize labor markets in many industrial and developing economies. For example, stud- ies using household- or individual-level surveys have found sizable wage differ- entials, even after accounting for a wide range of human capital attributes such as age, sex, and education. These wage differentials appear to vary systemati- cally by employer size and industry, to persist over long periods of time, and to be correlated across countries (Katz and Summers 1989). The enduring nature of these significant, and largely unexplained, wage differentials raises thorny questions about whether labor and product markets are efficient-especially in Hong Tan and Geeta Batra are with the Private Sector Development Department at the World Bank. This article was originally prepared as a background paper for World Development Report 1995: Workers in an Integrating World. The project was funded by RPO 678-39, from the World Bank Research Support Budget. {D 1997 The International Bank for Reconstruction and Development /THE WORLD BANK 59 60 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 developing economies-and whether policy interventions to improve the opera- tion of these markets are indicated. Groshen (1991) advanced four hypotheses to explain these wage differen- tials, each with very different policy implications. Unexplained differentials may (a) reflect quasi-rents from imperfectly competitive markets; (b) compensate for differences in the quality of working conditions; (c) represent wage incentives designed to elicit greater worker effort and reduce shirking; and (d) reflect un- measured worker quality and skills. To date most studies have found little em- pirical support for the first two hypotheses, as measured by the degree of union- ization, market power of firms, or working conditions in jobs. The last two hypotheses are more difficult to test because of the paucity of relevant worker and firm-level data, and their efficacy in explaining these firm-size and industry wage differentials remains unclear. A growing body of evidence, primarily from industrial economies, indicates that technological change influences wage inequality. Rising wage inequality in the United States between skilled and unskilled workers has been attributed to skill-biased technology by Katz and Murphy (1992), using time-series labor force surveys, and by Davis and Haltiwanger (1991), using panel firm data. Several recent U.S. studies using direct firm-level measures of technology use or R&D have found corroborating evidence for this hypothesis. Dunne and Schmitz (1995) find that use of advanced manufacturing technology is associated with signifi- cant wage differentials, with technology-wage premiums of equal magnitude as firm size-wage premiums. Bernard and Jensen (1994) jointly investigate the roles of exporting and R&D by employers and find evidence that both activities are associated with wage differentials. Similar research on these hypotheses in de- veloping economies is limited; however, using firm-level data from the 1986 Taiwan (China) census, Aw and Batra (1995) find that the average wage premi- ums associated with exporting were 24 percent. Research has focused on the linkages between technology and worker training and their consequences for wage growth and wage inequality. Using worker-level data with self-reported training measures, several studies find evidence in indus- trial economies that the propensity to train is greater for more educated workers, larger firms, and employers in R&D-intensive industries. Furthermore, these stud- ies show higher returns to training in technologically progressive industries (Mincer 1991; Lillard and Tan 1992; Tan and others 1992). Several firm-level studies, exploiting linkages with employee files, have found similar correlations between technology use by employers and wages and worker characteristics (Doms, Dunne, and Troske 1994; Hellerstein, Neumark, and Troske 1994). Because the firm-level studies had no data on worker training, the possibility remains that part of the technology-wage premiums that they estimated-for different firm-worker char- acteristics-may actually reflect important, but largely unmeasured, employers' investments in worker training and skill upgrading. In this article, we bring together unique firm-level data from three developing economies-Colombia, Mexico, and Taiwan (China)-to investigate the rela- Tan and Batra 61 tionships between technology and the interfirm structure of wages in manufac- turing. The data for Taiwan (China) are for 1986, and those for Mexico and Colombia are for 1992. Each firm reported information on several key groups of variables: mean wages by level of worker skill, firm and industry characteris- tics, and the firm's technology inputs. We hypothesize that firms acquire techno- logical knowledge through three channels: exports, an informal channel to tech- nology and know-how from abroad; investments in new technology, either through foreign licenses or through own R&D efforts; and worker training, which is a critical input in the adoption, modification, and effective use of new technologies. For each economy, we estimate semiparametric models to characterize the size-wage distributions of two groups of firms differentiated by their technol- ogy orientation: firms that invest in one or more of the three channels of tech- nology (investors) and firms that do not invest in any of the three (noninvestors). We compare the firm size-wage distributions of skilled and unskilled workers in investor firms and noninvestor firms to test the hypothesis of skill-biased technical change. It is possible that some sources of technology have a larger impact on wages than others. To identify the relative wage impacts of each of these technology investments, we also estimate separate models for each tech- nology source while controlling for investments in the other two technology sources. Section I discusses the motivation for the analysis, drawing upon salient find- ings from our parallel research on training and its effects on firm-level produc- tivity and wages using data from these developing economies. Section II de- scribes the data, provides some descriptive statistics on the firms in our samples, and discusses the econometric model to be estimated. Section III reports the principal findings. Section IV summarizes the cross-economy results and draws out their policy implications. I. TECHNOLOGY AND WAGE DIFFERENTIALS Our analytic approach is motivated by a model in which firms develop tech- nological capabilities through conscious investments in knowledge-generating activities. Firms can develop these capabilities-including own R&D and pur- chases of foreign technology and know-how, worker training and skill upgrad- ing, as well as exports-in different ways, and each investment strategy may have different productivity and wage outcomes. We focus on estimating the wage outcomes of these technology-generating activities, both jointly and sepa- rately by source, and on how these activities shape the size-wage distributions in Colombia, Mexico, and Taiwan (China). Our definition of technological capability follows Bell and Pavitt (1993). They distinguish between production capacity and technological capability. Produc- tion capacity measures the capacity of firms to produce output at given levels of efficiency with existing inputs of capital, labor, and technology. Technological 62 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 capability incorporates the additional and distinct resources needed to generate and manage technological change, including specialized managerial and techni- cal skills, knowledge and experience, and linkages with other firms. Employers whose workers have these specific skills and knowledge have a productivity advantage over less capable firms. The firm-specificity of these skills and knowl- edge creates a bilateral monopoly situation and incentive problems of motivat- ing workers to acquire and exercise specific skills. The bilateral monopoly prob- lem is resolved through firm-worker sharing of the higher productivity (rents) from these technology investments (Becker 1975; Tan and others 1992; Lindbeck and Snower 1989). We assume that there is a prior distribution of technological capabilities in a given economy at any point in time. The heterogeneity in capabilities and its correlation with firm size have been modeled by Lucas (1978), Jovanovic (1982), and Dunne and Schmitz (1995), among others, in the context of advanced tech- nology use. The Dunne and Schmitz study finds evidence that larger firms are more likely to use advanced technology, that technology usage is related to sig- nificant wage differences, and that technology-wage premiums are roughly of the same magnitude as size-wage premiums. In addition, employer incentives to invest in new technology may vary across industries because of interindustry differences in technological possibilities. In science-based industries, such as chemicals and electronics, the potential for product and process innovations is typically greater than in traditional industries such as leather and furniture. These technology differentials can give rise to a distribution of wages across employ- ers, varying the distribution by size and industry-differentials that can persist because of the firm-specificity of technology investments. If technology is the driving force, then similar interindustry differentials are likely to be found across countries. Tabulations of the data for Colombia, Mexico, and Taiwan (China) indicate that larger firms are typically more likely to train their workers, license foreign know-how, or invest in R&D, while smaller firms tend not to train and to rely on older and less skill-intensive kinds of technology. The data also show systematic differences across industries in the proportion of firms that invest in R&D and training, with the propensity to train workers, conduct R&D, or purchase know- how being higher in the science-intensive industries than in the more traditional, labor-intensive industries. I . Technological capabilities can be developed in several ways. First, employers can invest in their own R&D or purchase technology and know-how through licensing agreements with foreign firms. The evidence from developing coun- tries suggests that reverse engineering, imitation, and modification of foreign technology have been more critical to developing capabilities than own invest- ments in basic research and innovation (Pack 1992b). Second, firms can acquire relevant and best-practice technology through their links with foreign buyers of exported products as well as with foreign firms operating in the local markets (Westphal, Rhee, and Pursell 1984; Pack 1992a). Finally, whether it is through Tan and Batra 63 importing foreign technology or using, adapting, and redesigning technology through deliberate investments in R&D, building technological capacity depends fundamentally on the education and training of the workforce. As technologies evolve, a continuous process of job-specific training and retraining is needed to supply the technical and managerial skills required by new process and product innovations. There is a body of evidence from both industrial and developing countries that technologically progressive industries are more likely to train their work- ers and that these investments give rise to higher wages (Lillard and Tan 1992; Tan and others 1992). Using census data from Taiwan (China), Aw and Tan (1993) show that worker training has a large positive impact on firm-level productivity in all industries studied and that this effect is larger when training is accompanied by complementary investments in formal R&D and foreign know-how. Our related research on the productivity effects of enterprise training sup- ports the focus here on elements that are important in developing technological capabilities (see Tan and Batra 1995). First, within a production function frame- work, our cross-national analyses showed significant firm-level productivity gains from investments in each of the three sources of technology. Our findings of positive productivity effects from R&D and exporting are consistent with those reported in the productivity literature (see Bregman, Fuss, and Regev 1991). The productivity effects of training, however, are new. The results showed that formal training of skilled workers had a significant, positive effect on firm-level productivity, while the productivity effects of unskilled worker training were generally insignificant. Second, we found that some of the productivity gains from training were shared with workers in the form of higher wages. The results of joint estimation of a production function and wage model showed that the training effects in the wage model were approximately two-thirds those of training in the production function. The implied sharing of productivity gains-one-third to the employer, two-thirds to workers-was of roughly the same magnitude across the three economies. Firm-worker sharing of the costs and returns to training are consis- tent with the presence of firm-specific training (Becker 1975). These two findings-the link between investments in technology and firm- level productivity, on the one hand, and the link between productivity and wages, on the other-constitute the empirical basis for this article. We build on these findings to investigate the firm size-wage distribution across different groups of firms varying in technological orientation, by skill level, and across economies. II. DATA AND METHODOLOGY The data required to investigate the hypothesized technology-wage relation- ships are available for a cross-section of firms in the three economies-Colom- 64 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I bia, with 500 firms for 1992; Mexico, with 5,072 firms for 1992; and Taiwan (China), with 8,408 firms for 1986.' The three surveys of manufacturing firms contain broadly comparable, firm-level information on the key variables of in- terest. These include characteristics of the establishment, including year estab- lished, single-plant or multiplant status, two-digit industry classification, and foreign ownership; data on production and inputs, including capital assets, em- ployment, intermediate inputs, and energy use; number of workers and mean wages by broad occupations; and information on exports, training, and expen- ditures on R&D and foreign technology licenses. Variable Definitions "Training" is defined to include only formal structured training provided by the employer, whether by in-house trainers or by external providers, but ex- cludes informal on-the-job training provided either by coworkers or by supervi- sors. A more inclusive training variable would have little discriminatory power because most employers provide some informal on-the-job training, especially to new hires. The decision to exclude informal on-the-job training was informed by the findings of existing training studies (for example, see Lillard and Tan 1992) and by our own analyses of the productivity effects of different types of formal and informal training (see Tan and Batra 1995). The training variable is based on the number of employees trained by each source, except in Taiwan (China) where it is defined by positive expenditures on (presumably formal) training. Training in the Taiwan (China) sample, in all likelihood, is understated by our reliance on reported training expenditures, which are often not adequately recorded, rather than on the number of workers trained where better data are available. Thus more training firms in Taiwan (China) are probably misclassified as nontrainers than in the other two countries. For the Taiwan (China) analysis, this is likely to cause a downward bias in the estimated effects of training on wages, which strengthens our test for the presence of training-wage premiums. The variable "investment in R&D and know-how" is defined as positive ex- penditures on either research and development or on the purchase of foreign technology and know-how licenses. We note that R&D is defined more broadly in Mexico and Colombia than in Taiwan (China) and includes less-formal, adap- tive R&D and engineering modifications to product and process technologies. The variable "exports" is defined simply as having positive foreign exports. We will refer to these three variables as sources of technology and use them to separate firms into two groups, investors and noninvestors. Investors invest in 1. We exclude maquiladora firms from the Mexico sample because of the their unique, assembly-for- export characteristics. These firms, although typically large, conduct virtually no R&D and generally pay lower wages to both skilled and unskilled labor. Although they were not explicitly identified in our sample, we defined maquiladora firms as those that export 67 percent or more of their total output and import 75 percent or more of their raw materials. Thus, we dropped 305 firms from the analysis. We acknowledge Amy Hwang (Academia Sinica) for providing the data for Taiwan (China); the Secretariat of Labor (Mexico) for providing the data for Mexico; and SENA, the National Training Agency, for the data for Colombia. Tan and Batra 65 any one, or more, of the three sources of technology. Noninvestors do not invest in any of them. We chose this simple stratification to facilitate comparisons between investors and noninvestors and between the three technology sources and to ensure that the sample size for different groups was adequate for analy- sis. In the simple stratification, a firm is defined as an investor by one of seven combinations of investments in any one or more of the three dichotomous tech- nology indicators. The number of investor permutations rises quickly when each indicator variable is stratified by intensity, making the analysis intractable be- cause of small sample sizes in each investor category. However, we do report some findings for a more disaggregated stratification of firms by intensity of their technology investments. For each economy, we construct comparable definitions of skilled and un- skilled workers using information on the broad occupational breakdowns of the workforce in the firm. For Taiwan (China), the group of skilled workers in- cludes managers, engineers, and technicians; for Mexico, it includes directors, professionals, engineers, supervisors, and technicians; for Colombia, it includes administrators, technicians, and supervisors. Unskilled workers in all three econo- mies comprise unskilled production workers and employees in the residual "other administrative" group. This breakdown of skills is an improvement over the more commonly used breakdown between production and nonproduction work- ers. For Colombia and Mexico, but not Taiwan (China), we also have informa- tion on the educational distribution of workers in the firm. We have not used these data here in the interest of cross-national comparability, but education is partly reflected in our refined measures of skills. Our wage variable is defined as monthly wages and salaries. It excludes fringe benefits and other nonwage payments because these data, while reported, are for the workforce as a whole and are not broken down by occupational group. In Taiwan (China), nonwage payments for the investor sample are on average 40 percent higher than those for noninvestors. In -Mexico, the different bonuses for which we have data-for example, productivity, quality, and punctuality bonuses-are also on average higher among technology investors than among noninvestors and rise with firm size. Thus, inclusion of fringe benefits, bonuses, and other nonwage payments should result in higher technology-wage premi- ums in these economies and reinforce our findings. Overview of the Data In all three surveys, large firms were oversampled relative to their true weight in the population of industrial enterprises. The population of smaller firms, microenterprises in particular, is not known with any precision in Colombia and Mexico. A further complication is that the Taiwan (China) sample was selected on the basis of sales. Although crude population weights exist for Colombia and Mexico, we have opted to report only unweighted figures until more comparable weights can be developed for all three countries. 66 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 Table 1. Mean Characteristics of Sample Firms in Colombia, Mexico, and Taiwan (China), by Technology Source Percentage of firms Number of Number of Formal R&D Economy firms employees training2 licenses Exporting Colombia 500 186.00 65.60 76.60 40.00 Mexico 5,072 297.81 56.77 50.43 18.23 Taiwan (China) 8,408 145.47 15.84 25.89 51.42 a. For Colombia and Mexico, it includes firms that report formal internal or external training. For Taiwan (China), it includes firms that report training expenditures. Source: Authors' calculations. Table 1 indicates that the average firm size is about 298 employees in Mexico, 186 employees in Colombia, and 145 employees in Taiwan (China). Given the over-representation of large firms, these statistics cannot (and should not) be used to draw inferences about the relative level of technology orientation in the three economies. With this caveat, the remaining columns of table 1 show the percentage of firms in each economy reporting formal training, investments in R&D and technology licenses, and exports. Table 2 shows the mean characteristics of firms by their technology orienta- tion. In all three economies, in comparison with noninvestors, firms that invest in any of the sources of technology tend to be larger (in terms of employment) and older (they have been in operation longer), have multiple plants and a higher proportion of foreign capital, and pay higher wages to both their skilled and unskilled employees. Given the stylized finding of a firm size-wage relationship in most economies, part of the higher wages paid by investors in technology (or any source of technology) reflects their larger average size. In the analysis that follows, this firm size-wage relationship is explicitly modeled in estimating the technology-wage premium; interindustry differences are accommodated by a vector of two-digit industry dummy variables, but these are not the focus here. Empirical Methodology The impact of technology on industrial wage differentials is investigated in several ways. First, we compare the wage distributions of the two groups of firms that differ in their technology orientation and test the hypothesis that, other things being equal, technology investors pay higher wages. Employer size is a useful way to characterize this distribution of wages because technological capability is thought to be a positive function of firm size. Second, the wage comparisons are done separately for skilled and unskilled workers. If techno- logical change is highly skill-demanding, as suggested by the literature, then employer investments in technology should have a larger wage impact on skilled workers relative to unskilled workers. Finally, we estimate the separate wage effects of employer investments in exporting, worker training, and R&D to see which technology strategy has the largest impact on wages and whether there are major differences across economies in the role of each technology source. Table 2. Mean Characteristics of Sample Firms in Colombia, Mexico, and Taiwan (China), by Technology Orientation Ratio of wages paid by Ratio of wages Foreign investors to those paid paid to skilled Multiple plants ownership Percentage by noninvestors workers to those Economy and Number of Years in (percentage (percentage of skilled Skilled Unskilled paid to unskilled technology orientation employees operation of firms) of firns) labor labor labor workers Colombia Investors 203 20.82 20.31 - 0.23 2.36 2.69 2.50 Noninvestors 43 14.52 3.85 - 0.24 n.a. n.a. 2.13 Mexico Investors 349 23.99 22.74 18.97 0.31 1.30 1.26 3.79 Noninvestors 130 18.74 11.72 6.67 0.29 n.a. n.a. 3.47 Taiwan (China) Investors 227 12.68 33.69 11.03 0.49 1.34 1.21 1.81 Noninvestors 31 10.76 13.06 1.52 0.55 n.a. n.a. 1.86 - Not available. n.a. Not applicable. Note: Investors are defined as firms reporting investments in R&D or formal training (internal or external) or export. Source: Authors' calculations. 68 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I These wage comparisons are based on nonparametric and semiparametric regression methods (Robinson 1988; Hardle 1990). In initial data exploration, we found the firm-level wage data to have a lot of noise and the size-wage dis- tributions to be nonlinear. In preliminary analyses, we compared wage models estimated by ordinary least squares (OLS) and nonparametric methods in each of the three economies and concluded that the size distribution of wages was better estimated by nonlinear methods. Although the OLS parameter estimates are broadly comparable to the nonparametric estimates for wage intercept differ- ences between investors and noninvestors, and for other explanatory variables such as industry and firm characteristics, they differ markedly in estimates of the size-wage distribution. The OLS quadratic size specification yields poor re- sults because of its sensitivity to noise, nonlinearities, and outliers. Thus we adopted a more flexible approach. Unlike the more familiar linear regression technique, the wage models we estimate impose no specific func- tional form on the data, such as a quadratic firm-size specification, in effect allowing the data to speak for themselves. Deaton (1989) discusses these and other advantages of nonparametric techniques over alternative, more familiar econometric techniques such as cross-tabulations, which are not very flexible and do not convey information as transparently, and linear regressions, which tend to oversummarize and rarely do justice to the amount of information avail- able. The semiparametric model extends this methodology by combining a para- metric component with the nonparametric one. We let the nonparametric com- ponent be related to firm size and the parametric component be a function of other firm attributes, such as age of the firm, multiplant status, foreign capital ownership, and industry. The semiparametric approach thus enables us to com- pare the within-industry, firm size-wage distributions of technology investors and noninvestors net of the wage effects of other firm-specific factors and cross- industry wage-level differentials. The semiparametric wage model has the following form: (1) y = x'1 + 0(z) + u where y represents the dependent variable, log(wage), ,B is assumed to be a lin- ear function of x, a matrix of firm attributes including foreign ownership, firm age, multiplant status, and industry. The nonparametric component, firm size, is represented by z, measured as log (employment), and u, the error term, is as- sumed to be independently and identically distributed with finite variance. The estimation is done in two stages. In the first stage, we obtain the coeffi- cients of x, that is, P, using a kernel nonparametric method and a smoothing parameter designed to minimize the sum of squared residuals (described in greater detail in the appendix). In the second stage, to obtain the relationship between wages and firm size net of the wage effects of other firm and industry character- istics, we estimate the following nonparametric model (2) y - x', = 0(z) Tan and Batra 69 where, as before, y is log (wage), x'p is a vector of predicted values obtained from the first-stage estimation, z is firm size as measured by log (employment), and 0 is the function relating wages to firm size. We estimate a semiparametric wage model for each group of firms, the tech- nology investors and noninvestors samples, and, within each firm sample, sepa- rate regressions for skilled workers and for unskilled workers. Because we have already controlled for the wage effects of x',B, differentials in the size distribu- tion of wages between the two groups of firms are attributed to employer in- vestments in exports, training, or R&D and know-how. Wage differentials for the two skill groups can also be compared to test the hypothesis of skill-biased technological change. We extend this semiparametric methodology to estimate the separate wage effects of each technology source. Recall that firms are classified as being in the investor sample if they do one or more of the following: exports, training, or R&D and know-how. For any given technology source, we estimate the semiparametric wage model for that group of investors, controlling for the wage effects of contemporaneous investments in the -other two technology sources. For example, to estimate the wage effects of training, we expand x'3 to include two dummy variables-for exporting and for R&D and know-how-so that 0(z) provides an estimate of the pure size-wage effects from training. (This proce- dure ignores second-order, and possibly important, interaction effects between training and the other two technology sources.) III. EMPIRICAL FINDINGS The first set of regressions compares the wages paid by technology investors and by noninvestor firms. We first report the parametric component of wages- firm characteristics and industry effects. We then present graphical comparisons of the nonparametric wage component, the firm size-wage distributions by skill group, for the investor and noninvestor samples. The second set of regressions disaggregates these results by technology source-exports, worker training, and investments in R&D and know-how. For brevity, the discussion here focuses only on the size-wage distributions net of the parametric wage component.2 Technology Investors and Noninvestors Table 3 reports the semiparametric coefficient estimates for the wage effects of different firm-level characteristics, separately by the technology orientation of firms and by skilled and unskilled worker groups. These estimates hold con- stant interindustry wage-level differentials; these are discussed subsequently. Table 3 highlights what appear to be common correlates of firm-level produc- tivity across all three economies. First, firm age is generally associated with a 2. The parametric estimates by source of technology resemble those found for the aggregate measure of technology orientation. The detailed results are available from the authors. 70 THE WORLD BANK ECONOMIC REVIEW, VOL. I, NO. I Table 3. Estimates of the Impact of Firm Characteristics on Wages in Colombia, Mexico, and Taiwan (China) Skilled labor Unskilled labor Economy and Years in Multiple Foreign Years in Multiple Foreign technology orientation operation plants ownership operation plants ownership Colombia Investors 0.002 0.103* - 0.002- 0.052 (0.002) (0.060) (0.001) (0.043) Noninvestors 0.003 0.188 - 0.001 0.265** (0.004) (0.225) (0.002) (0.116) Mexico Investors 0.067*** 0.189*** -0.037 0.004*** 0.102*** 0.175*** (0.030) (0.034) (0.100) (0.001) (0.027) (0.029) Noninvestors 0.003* 0.096 0.227** 0.003* 0.096 0.131 (0.002) (0.085) (0.112) (0.002) (0.071) (0.093) Taiwan (China) Investors 0.006*** 0.022 0.227*** 0.006*** 0.024 0.093*** (0.001) (0.013) (0.020) (0.001) (0.017) (0.026) Noninvestors 0.007*** 0.094*** 0.379*** 0.005*** 0.011 - (0.001) (0.026) (0.068) (0.002) (0.035) - Not available. Note: Values are semiparametric coefficient estimates. The dependent variable is the log of the wage of labor. Numbers in parentheses are standard errors. * Significant at 10 percent. * * Significant at 5 percent. * Significant at 1 percent. Source: Authors' calculations. positive and statistically significant effect on wages, except in Colombia. This finding is consistent with the interpretation that longevity is correlated with firm-level productivity, because less efficient firms have a greater likelihood of exit (Jovanovic 1982). Second, foreign ownership is associated with significantly higher wages in Mexico and Taiwan (China) (this variable was not collected for Colombia), possibly because of their higher technological capabilities. (See Tan and Batra 1995 for cross-country findings on the productivity differentials asso- ciated with foreign ownership of firms.) Finally, multiplant status is associated with higher wages, but its statistical significance is mixed. Several U.S. estab- lishment-level studies find large multiplant wage effects; the mixed results re- ported here may simply reflect the fact that our unit of observation is the firm, not the establishment, and that firm-size effects (for which multiplant status is a proxy) are captured in the nonparametric wage component. Industry dummy variables were also included in the semiparametric wage models to control for interindustry wage differentials. Though not reported here, industry intercepts estimated in each regression suggest that two findings are common to all three economies. First, there appear to be two distinct groups of industries: industries that pay high wages (typically chemicals, pharmaceuticals, general and electrical machinery, basic metals, and transportation equipment) and industries that pay low wages (usually food and beverages, clothing and Tan and Batra 71 textiles, leather, and wood and furniture products). Second, the magnitude and statistical significance of interindustry wage differentials are usually greater for the sample of technology investors, especially in the skilled-worker regressions. In contrast, interindustry wage levels are relatively undifferentiated in the case of unskilled workers. The similarities of these two findings across economies are consistent with the hypotheses of interindustry differences in technological possibilities and their associated differential demand for skilled workers. The findings raise questions about the efficacy of more traditional explanations that relate industry wage differentials to market structure, openness to trade, and unionization. We did not explicitly address these competing explanations, given the limited two-digit industry variation in our data, but we argue that cross-national variations in these factors could not give rise to the common patterns of interindustry wage differentials observed. The Size Distribution of Wages Our estimates of the nonparametric wage components show the size-wage distributions net of the parametric wage effects reported above. At any given firm size, the gap between the two curves for technology investors and noninvestors is a measure of the percentage wage differential between the two groups of firms. Because investors and noninvestors differ only in whether they invest in any technology, this wage gap is an estimate of the technology-wage premium. To facilitate comparisons by firm size, we adopt the following firm-size no- menclature for all economies. We define firms with less than 16 employees as micro, firms with 16 to 100 employees as small, firms with 101 to 250 employ- ees as medium, and firms with more than 250 employees as large. We use these definitions to report average wage premiums by discrete firm-size categories. Figure 1 presents our estimates of the size-wage distributions for Mexico. The wages paid to skilled workers are everywhere higher in firms that invest in technology than in those that do not; furthermore, the wage differential is larg- est for technology investors in the smaller firm-size range. Compared with the wage premiums paid by noninvestors, those paid by micro, small, medium, and large firms that invest in technology are 88, 33, 32, and 21 percent, respectively. The corresponding size-wage distributions for unskilled workers are not well differentiated by the technology orientation of firms. For the smallest and big- gest firms, wages paid to unskilled workers by technology investors are slightly higher than those paid by noninvestors. The reverse is true for medium-size and larger firms-investors pay slightly lower wages to unskilled workers as com- pared with noninvestors. By discrete size categories, the technology-wage pre- miums paid by investors relative to noninvestors are 13, -4, -1, and 1 percent for micro, small, medium, and large firms, respectively. Figures 2 and 3 show the corresponding size-wage distributions for Tai- wan (China) and Colombia, respectively. For skilled workers, the size-wage 72 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 Figure 1. Estimates of the Size-Wage Distributionsfor Skilled and Unskilled Labor, Mexico Skilled labor Mean wage (log) 8.5 - 8.0 0 0 7.5 - 0 7.0 - 0 0 6.5 - 0 6.5- 0 o Noninvestors 0 6.0 O Investors 5.5- I I I I 0 2 4 6 8 10 12 Unskilled labor 8.5 - 8.0 - 0 7.5 -0 7.0 - 6.5- 0 o Noninvestors 6.0 0 Investors 5.5- I l l 0 2 4 6 8 10 12 Mean number of employees (log) Source Authors' calculations Tan and Batra 73 Figure 2. Estimates of the Size-Wage Distributions for Skilled and Unskilled Labor, Taiwan (China) Skilled labor Mean wage (log) 12.2 00 12.0 8 o 11.8 - 11.6 -X a OO t 11.4 O Noninvestors 0 11.2 0 Investors 11.0 I I I 0 2 4 6 8 10 12 Unskilled labor 12.2- 12.0- 11.8 11.6 e 0e 11.4 - co 0 0 Noninvestors 11.2 Ooo0 o ,9 Investors 11.0- I I I I 0 2 4 6 8 10 12 Mean number of employees (log) Source: Authors' calculations. 74 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Figure 3. Estimates of the Size-Wage Distributions for Skilled and Unskilled Labor, Colombia Skilled labor Mean wage (log) 13.5 - 13.0 - 12.5- 0000 0 12.0 - o 0 0 > :/ %4C a Noninvestors 11.5 - O O 0 Investors 11.0 I I I I 0 2 4 6 8 10 12 Unskilled labor 13.5 - 13.0 12.5 lb 12.0 - a 0 00 0 0 03 Noninvestors 11.5 - a0 0 Investors 11.0 I I I I I 0 2 4 6 8 10 12 Mean number of employees (log) Source. Authors' calculations. Tan and Batra 75 distributions in Taiwan (China) and Colombia resemble those in Mexico in being higher for technology investors than for noninvestors. However, each of these economies has slightly different patterns of technology-wage premi- ums by firm size. For Taiwan (China) (figure 2), the technology premium for skilled workers is highest in the small size category, and the premiums de- cline with firm size as in Mexico. The technology-wage premiums for skilled workers in Taiwan (China) are 8, 21, 15, and 14 percent in micro, small, medium, and large firms, respectively. For Colombia (figure 3), by contrast, the technology-wage premiums for skilled workers tend to rise with size (ex- cept for the largest firms); the corresponding wage premiums by size are 20, 34, 48, and -22 percent, respectively. For unskilled workers, technology investors in Taiwan (China) and Colom- bia-as their Mexican counterparts-pay smaller wage premiums for unskilled workers than for skilled workers. In Taiwan (China), there is no systematic pat- tern by firm size-the mean wage premiums by size are -7, 3, -3, and 0 percent for micro, small, medium, and large firms, respectively. In Colombia the firm- size pattern of technology premiums for unskilled workers resembles that for skilled workers. The corresponding premiums by size are 19, 22, 26, and -13 percent for micro, small, medium, and large firms, respectively. Technology-wage premiums are averages across all investor firms, irrespec- tive of the intensity of their investments. To verify that the wage effects of tech- nology increase in proportion to investment intensity, we estimated pooled OLS wage models with indicator variables for two intensity levels-moderate and high-measured relative to noninvestors. This OLS wage model has the same variable specification as the nonparametric model except for the inclusion of a quadratic firm-size measure and indicator variables for moderate and high- intensity investors. (This wage model was estimated for Mexico and Taiwan (China), but not Colombia, where small sample size precluded this disaggrega- tion.) A firm is defined to be a high-intensity investor if it invests more than the median intensity in any one of the three technology sources; otherwise it is a moderate investor. The OLS estimates suggest that high-intensity investors pay significantly higher wages than moderate-intensity investors, and both investor groups pay higher wages compared with noninvestors. The results also suggest that technology-wage differentials are larger among skilled workers than un- skilled workers. Both results are consistent with our nonparametric findings and reinforce our interpretation of these wage differentials. To summarize, the results for all three economies provide evidence of large technology-wage premiums for skilled workers, and small or nonexistent tech- nology-wage premiums for unskilled workers (table 4). (We tested, and rejected, the null hypothesis that the wages of skilled workers are the same in investor and noninvestor firms across all size categories.) In Mexico the overall (pooling across firm sizes) technology-wage premium for skilled workers is 54 percent, which is substantial, but the premium for unskilled workers is just 11 percent. In Taiwan (China) the corresponding overall premiums for skilled and unskilled Table 4. The Technology-Wage Premium by Labor Type in Colombia, Mexico, and Taiwan (China) (percent) Skilled labor Unskilled labor Any Any Economy Training R&D Exports investment Training R&D Exports investment Colombia 39.93 24.23 61.12 41.76 20.20 8.65 29.18 22.51 Mexico 76.12 74.54 51.29 53.88 17.82 12.75 -2.37 10.96 Taiwan (China) 53.88 51.13 22.38 31.92 14.91 16.77 1.92 7.47 Note: Values are based on semiparametric regression results. Technology premium is defined as the overall difference between the average wage of firms investing in any source of technology and the average wage paid by noninvestors. The wage premiums for each technology source (training, R&D, exports) are estimated controlling for investments in the other sources of technology. Source: Authors' calculations. Tan and Batra 77 workers are 32 and 7 percent, respectively; in Colombia the skilled and un- skilled technology-wage premiums are 42 and 23 percent, respectively. We interpret these results as evidence of skill-biased technological change. We note that the same findings are also consistent with skill-neutrality if two assumptions hold: that technology raises the productivity of skilled and unskilled workers to the same degree and that skilled workers are not mobile across firms (so wage premiums can persist), while unskilled workers are mobile and readily substitutable for each other (thus competing away wage differentials).3 How- ever, we discount this competing explanation on the basis of two findings from our related research (Tan and Batra 1995). First, there is strong cross-national evidence that training for skilled workers has a large productivity impact, but training for unskilled workers does not. Second, although skilled workers are generally less mobile than unskilled workers, the evidence suggests that techni- cally efficient firms, which are highly correlated with our investor firms, have lower job turnover rates among employees than inefficient firms, which make few technology investments. Size- Wage Distributions by Technology Source Which technology source has the largest impact on the size distribution of wages? To address this issue, we estimate semiparametric wage regressions sepa- rately for samples of firms investing in each of the three technology sources, controlling for their contemporaneous investments in the two remaining tech- nology sources. The sample size used for each technology source varies depend- ing on the specific source under consideration. The resulting firm size-wage dis- tributions are compared with those of noninvestors, yielding estimates of the technology-wage premiums attributable to specific sources of technology. Figure 4 summarizes the wage premiums for each economy by technology source and by firm size for skilled and unskilled labor. At the broadest level-the case of overall technology wage premiums-the wage premiums estimated for each technology source replicate the principal results from the aggregate mea- sure of technology orientation. Even disaggregating by technology source, wage premiums are greater for skilled workers than for unskilled workers in all three economies. Furthermore, economy-specific patterns of technology-wage premi- ums by firm size remain-wage premiums are largest for the smallest size cat- egory in Mexico; they tend to rise with size (except for the largest firms) in Colombia; and they are highest in the small size category in Taiwan (China), though these size-related wage differentials are not striking. Finally, figure 4 shows that the relative importance of the three technology sources does not vary with size. Several new results emerge when the wage premiums from each source are considered separately. First, for Mexico and Taiwan (China), table 4 shows that firm investments in R&D and worker training are the principal drivers of the 3. We acknowledge an anonymous referee for drawing our attention to this possibility. 78 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Figure 4. Wage Premiums by Technology Source and Firm Size, Colombia, Mexico, and Taiwan (China) Skilled labor Unskilled labor Colombia r 120 120 8 10 80 8 2 40 40 c~~~~~~~~0 ~~~~~~0 : -40 -40 Mexico 120 120 V __80 80 2 40- 40- C . * I I i n _ __ -40 --40 Taiwan (China) 120 120 V ~.80- 80- 2 40 - 40 - 0. 0 C-40 -i40 Micro Small Mediuim Large Micro Small Medium Large Firm size (number of employees) Firm size (number of employees) *Training * R&D E Exports Note: Micro firms have fewer than 16 employees, small firms have 16 to 100, medium firms have 101 to 250, and large firms have more than 250. Souirce: Authors' calculations. Tan and Batra 79 wage premiums paid to skilled workers by technology investors; exporting, as an informal source of technology, is relatively less important. In Mexico, com- pared with wages paid by noninvestors, employer investments in training are associated with a skilled worker wage premium of 76 percent, R&D and know- how investments with a premium of 74 percent, and exporting with a premium of 51 percent. In Taiwan (China) the corresponding skill wage premiums for worker training, R&D, and exports are lower-54, 51, and 22 percent, respec- tively-possibly because of higher levels of human capital and knowledge in- vestments in Taiwan (China) than in Mexico. For unskilled workers the corre- sponding average wage premiums for worker training, R&D, and exports are 18, 13, and -2 percent, respectively, for Mexico, and 15, 17, and 2 percent, respectively, for Taiwan (China). In Colombia, by contrast, the skill wage premiums for training or for R&D and know-how'-40 and 24 percent, respectively-are dwarfed by the skill wage premiums from exporting of 61 percent. This relative ranking of the wage pre- miums by source persists for the unskilled worker group as well-20, 9, and 29 percent for training, R&D, and exports, respectively-though clearly the size of these premiums is smaller. We suspect that part of the explanation for the rela- tive importance of exports lies in the high protection afforded Colombian indus- try, at least until recently, as compared with the relatively more open economies of Taiwan (China) and Mexico. In such a protected environment, a firm's con- tacts with foreign buyers and suppliers may be a more important source of best- practice technology than own R&D or training. IV. CONCLUSIONS AND POLICY IMPLICATIONS Unique firm-level data were used to investigate the structure of industrial wages and the role of technology in giving rise to wage differentials across firms in Colombia, Mexico, and Taiwan (China). Semiparametric methods provided a flexible tool for estimating the size distribution of wages and for disentangling the separate contributions of R&D, training, and exports on the size-wage rela- tionship. The following findings and policy implications emerged from our cross- economy analyses. Our cross-economy comparisons yielded two principal results. First, control- ling for firm characteristics and industry, we found evidence in all three econo- mies that firm investments in technology have a large impact on the size-wage distributions for skilled workers and a relatively smaller impact on wages paid to unskilled workers. This asymmetric wage impact of technology is consistent with the hypothesis of skill-biased technical change. Second, a decomposition of wage effects by source of technology revealed that the wage premiums paid to skilled workers are driven primarily by firm investments in R&D and training; exporting is relatively less important. The results suggest that the large firm-size wage differentials corimonly ob- served in many economies are primarily the outcomes of employer investments 80 THE WORLD BANK ECONOMIC REVIEW, VOL 11, NO. I in technology. In the extant literature, many analysts have speculated that the size and persistence of these wage differentials reflect monopoly rents from market power and imperfections in the labor market, unobserved worker attributes, or incentive wage schemes. Our findings indicate that these wage differentials re- flect the returns to firm investments in technology; as such, they do not call for remedial policies to improve the working of labor markets. Though not the focus of this article, we also found large interindustry wage differentials that may reflect industry differences in technological change; this is a fruitful area for future research. The findings also highlight striking similarities across the three economies in the wage effects of technology investments-sizable wage premiums for skilled workers, but not for unskilled workers. Compared with noninvestor firms, the overall wage premiums for skilled workers are 32 percent in Taiwan (China), 54 percent in Mexico, and 42 percent in Colombia (table 4). In contrast, the corre- sponding wage premiums for unskilled workers are just 7, 11, and 22 percent, respectively. This result by skill group continues to hold when the separate sources of technology-R&D, training, and exports-are considered. These differential wage effects by skill level mirror the productivity effects of training estimated within a production function framework for the same sample of firms in the three economies. We conclude that technological change is skill biased. The im- plication is that sustained future economic growth and technological change will depend critically upon an increased availability of educated and skilled workers. This technology-skill complementarity also has implications for income in- equality. For many developing economies, several recent trends-the accelerat- ing pace of best practice in information and industrial technology, the increased inflows of capital and technology, and the growing integration into world mar- kets-could create strong demand for skilled workers far outstripping the sup- ply capacity of their educational and training institutions. Without appropriate responses from the private sector and governments, the outcome is likely to be growing income inequality between skilled and unskilled workers and between those with more and less education. APPENDIX. SEMIPARAMETRIC AND KERNEL NONPARAMETRIC ESTIMATION Semiparametric Regression Estimation The semiparametric regression model has the following functional form: (A-1) y = X' + 0(Z) + u where y is an n x 1 vector representing the dependent variable, and n indexes firms in the sample; ,B is assumed to be a linear function of x, an n x k matrix of k firm attributes. The nonparametric component is represented by an n x 1 vector, z. We assume that the error term u is independently and identically dis- Tan and Batra 81 tributed with finite variance. In addition, E (u/z, x) = 0 and O is an unknown function of z. Since E(y/z) = E(x/z)', + O(z), equation A-1 can be rewritten as (A-2) y = E (y/z) + [x - E (x/z)]' 1 + u where the deterministic part of y is decomposed into two parts: one is the effect on y of z, E (y/z), and the other is the effect on y of x net of z, [x - E(x/z)]' P. To obtain the coefficients of x, that is, ,B, we estimate E (y/z) and E (x/z) using a kernel nonparametric method and a smoothing parameter designed to mini- mize the sum of squared residuals as described below. Kernel Nonparametric Estimation Given the data base [(Y,,Zi)/i = 1, .. . , n], the nonparametric estimate of m(z) is calculated as a weighted average of g(Y,), where the heavier weights are given to the observations with the z, closest to z. That is, m(z) is estimated by ,, (A-3) in(zj) = ,g(Yi)W(z, z,) i=l where [W(z,z,), i = 1, . . . , n] is a sequence of weights that sums to 1. The idea is that the observations, the g( Y,)s, with the z; close to z, contain more information on m(z) than observations far away from z. The weights can be expressed as (A4) W(z,z,)=K(z-z,)Ih)/X(K(z-z,)/h). The weighting function expressed in equation A-4 will sum to 1 for all z,. K(z - z,)/h is called the Kernel function given by N - (z,h). h is a positive scalar bandwidth number or smoothing parameter that determines the weights to be assigned to observations in the neighborhood of z (Hardle 1990; Delgado and Robinson 1992). The choice of the smoothing parameter, h, plays an important role in non- parametric regression estimations, because it affects the magnitude of the weights assigned to observations in the neighborhood of z. For example, if h is too large, the observations far from z will have a large impact on the E(y/z). Although it is common practice to assume an exogenous smoothing parameter, it is important that the smoothing parameter depends on the data with a view to reflecting sample size and scale of measurement. We determine the optimal smoothing parameter using least squares cross-validation techniques to determine the opti- mal bandwidth that gives the best fit of the nonparametrically estimated regres- sion curve to the actual data (Hardle 1990). REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. 82 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Aw, B. Y., and Geeta Batra. 1995. "Wages, Firm Size, and Wage Inequality: How Much Do Exports Matter?" Working Paper. Department of Economics, Pennsylvania State University, University Park. Processed. Aw, B. Y., and Hong Tan. 1993. "Training, Technology, and Firm-Level Productivity in Taiwanese Manufacturing." Private Sector Development Working Paper. Private Sector Development Department, World Bank, Washington, D.C. Processed. Becker, Gary. 1975. Human Capital: A Theoretical and Empirical Analysis with Spe- cial Reference to Education. New York: Columbia University Press. Bell, Martin, and Keith Pavitt. 1993. "Accumulating Technological Capability in Devel- oping Countries." Proceedings of the 1992 Annual World Bank Conference on De- velopment Economics. Washington, D.C.: World Bank. Bernard, A. B., and J. B. Jensen. 1994. "U.S. Manufacturing Exports, Jobs, and Wages." CES Working Paper. Center for Economic Studies, U.S. Bureau of the Census, Wash- ington, D.C. Processed. Bregman, Ariel, Melvyn Fuss, and Haim Regev. 1991. "High Tech and Productivity: Evidence from Israeli Industrial Firms." European Economic Review 35:1199-21. Davis, Steve J., and John Haltiwanger. 1991. "Wage Dispersion between and within U.S. Manufacturing Plants, 1963-86." NBER Working Paper Series 3722. National Bureau of Economic Research, Cambridge, Mass. Processed. Deaton, Angus. 1989. "Rice Prices and Income Distribution in Thailand: A Non- Parametric Analysis." Economic Journal 99:1-37. Delgado, Miguel A., and Peter M. Robinson. 1992. "Nonparametric and Semiparametric Methods for Economic Research." Journal of Economic Surveys 6(3):201-49. Doms, Mark, Timothy Dunne, and Kenneth Troske. 1994. "Workers, Wages, and Tech- nology." CES Working Paper. Center for Economic Studies, U.S. Bureau of the Cen- sus, Washington, D.C. Processed. Dunne, Timothy, and J. A. Schmitz, Jr. 1995. "Wages, Employment Structure, and Employer Size-Wage Premia: Their Relationship to Advanced Technology Usage at U.S. Manufacturing Establishments." Economica 62:89-107. Groshen, Erica. 1991. "Sources of Intra-Industry Wage Dispersion: How Much Do Employers Matter?" Quarterly Journal of Economics 106:869-84. Hardle, Wolfgang. 1990. Applied Nonparametric Regression. New York: Cambridge University Press. Hellerstein, J., David Neumark, and Kenneth Troske. 1994. "Wages, Productivity, and Worker Characteristics." CES Working Paper. Center for Economic Studies, U.S. Bureau of the Census, Washington, D.C. Processed. Jovanovic, Boyan. 1982. "Selection and Evolution of Industry." Econometrica 50:649- 70. Katz, Lawrence F., and Kevin M. Murphy. 1992. "Change in Relative Wages, 1963- 1987: Supply and Demand Factors." Quarterly Journal of Economics 107:35-78. Katz, Lawrence F., and Lawrence Summers. 1989. "Industry Rents: Evidence and Im- plications." Brookings Papers on Economic Activity, Microeconomics. Lillard, Lee A., and Hong Tan. 1992. "Private Sector Training: Who Gets It and What Are Its Effects?" In R. G. Ehrenberg, ed., Research in Labor Economics, vol. 13. Greenwich, Conn.: JAI Press. Lindbeck, Assar, and Dennis J. Snower. 1989. The Insider-Outsider Theory of Em- ployment and Unemployment. Cambridge, Mass.: MIT Press. Tan and Batra 83 Lucas, Robert. 1978. "On the Size Distribution of Business Firms." The Bell Journal of Economics 9(2):508-23. Mincer, Jacob. 1991. "Human Capital, Technology, and the Wage Structure: What Do Time-Series Show?" NBER Working Paper 3581. National Economic Bureau of Eco- nomic Research, Cambridge, Mass. Processed. Pack, Howard. 1992a. "Learning and Productivity Change in Developing Countries." In G. K. Helleiner, ed., Trade Policy, Industrialization, and Development: New Per- spectives. Oxford: Clarendon Press. . 1992b. "New Perspectives on Industrial Growth in Taiwan." In Gustav Ranis, ed., Taiwan: From Developing to Mature Economy. Boulder, Colo.: Westview Press. Robinson, P. M. 1988. "Semiparametric Econometrics." Journal of Applied Economet- rics 3:35-51. Tan, Hong, and Geeta Batra. 1995. "Enterprise Training in Developing Countries: In- cidence, Productivity Effects, and Policy Implications." Private Sector Development Department, World Bank, Washington, D.C. Processed. Tan, Hong, Bruce Chapman, Christine Peterson, and Allison Booth. 1992. "Youth Train- ing in the United States, Great Britain, and Australia." In R. G. Ehrenberg, ed., Research in Labor Economics, vol. 13. Greenwich, Conn.: JAI Press. Westphal, Larry, Yung W. Rhee, and Garry Pursell. 1984. "Sources of Technological Capability in South Korea." In Martin Fransman and K. King, eds., Technological Capability in the Third World. London: Macmillan. The Effects of Public Sector Hiring and Compensation Policies on the Egyptian Labor Market Ragui Assaad This article examines the combined impact of the employment guarantee for gradu- ates and public sector compensation policies on the Egyptian labor market. Besides contributing to an unsustainable rate of growth in the government labor force, these policies have encouraged queuing for government jobs, contributed to high graduate unemployment rates, and reduced the employment of graduates in the private sector. Despite substantial wage erosion in the public sector in recent years, government wages, when appropriately corrected for observed heterogeneity and sample selec- tion, are on a par with, or higher than, private sector wages, especially for graduates. When combined with the more desirable nonwage aspects of government jobs, these compensation levels explain the attractiveness of public sector employment to gradu- ates. Government pay scales are especially advantageous to female secondary school graduates, who appear to face considerable discrimination in the private sector. The public sector's role in shaping labor market outcomes through its hiring and compensation policies takes on special importance in Egypt because of the country's long-standing policy of guaranteeing employment in the public sector for all graduates of secondary and postsecondary institutions. When the policy was first instituted in the early 1960s, its impact was relatively limited because of the small number of eligible graduates, but, over time, it has had major con- sequences for the Egyptian labor market and economy. The employment guarantee fueled the growth in demand for secondary and university education, which has, in turn, led to rapid growth in public sector employment and in the number of aspirants to such employment. When the growth of the public sector wage bill became unsustainable in the early 1980s, the government responded by eroding real public sector wages and extending the waiting period for government jobs. Despite these measures, long queues for government jobs indicate that rents received by public sector workers have not dissipated. Although the lifetime job security and many of the generous benefits Ragui Assaad is with the Humphrey Institute of Public Affairs at the University of Minnesota. This article was originally prepared as a background paper for World Development Report 1995: Workers in an Integrating World. A previous version was circulated as Working Paper 9517 of the Economic Research Forum for the Arab Countries, Iran, and Turkey. The author gratefully acknowledges research assistance by Paul Stern and thanks Insan Tunali, Ishac Diwan, and three anonymous referees for their invaluable comments and suggestions. C) 1997 The International Bank for Reconstruction and Development / THE WORLD BANK 85 86 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I associated with these jobs are in theory guaranteed by law, they are difficult to obtain in the private sector because of widespread noncompliance with labor regulations. Anecdotal evidence also suggests that expectations of effort and performance and tolerance of moonlighting in the public sector have adjusted to counteract at least partially the erosion of public sector wages. For example, analyzing data from C6te d'Ivoire and Peru, van der Gaag, Stelcner, and Vijverberg (1989) link a greater incidence of moonlighting with lower public sector wages. Because of the employment guarantee for graduates and the resulting domi- nance of the public sector in the market for educated labor, public sector com- pensation policies have had a major impact on the employment and wages of graduates elsewhere in the economy. In this article, I present a Harris-Todaro- type model that predicts the major features of the market for educated labor in Egypt, namely queuing for public sector jobs, which leads to unemployment rates that are three to five times higher among graduates than nongraduates, and reduced employment of graduates in the private sector. Although graduates are overremunerated in this economy, in a static sense, the resulting excess de- mand for secondary and higher education results in an oversupply of educated labor and comparatively low rates of return to education by international stan- dards. Psacharopoulos (1983) and others find that rates of return to education are generally higher in the competitive sector than in the uncompetitive sector because of the presumed equalization of pay scales in the latter. In contrast, I find that rates of return to secondary education in Egypt are higher in the public sector than in the private sector. Section I describes the history and operation of the public sector hiring and compensation policies in Egypt. Section II examines the effects of the employ- ment guarantee on the size and composition of the public sector labor force and on open unemployment in the economy. Section III develops a model of the market for educated labor in Egypt and presents the main findings on the effects of public sector employment and compensation policies on the wage structure and returns to education in the Egyptian labor market. Section IV concludes. I. PUBLIC SECTOR HIRING AND COMPENSATION POLICIES This section reviews the history and mode of operation of the graduate em- ployment guarantee scheme, highlighting changes in the size of the program and the resulting job queue that it engendered. It also describes the government's wage-setting policy. Historical Background of the Employment Guarantee for Graduates As part of the extensive nationalization drive of 1961-62, the Egyptian gov- ernment initiated a major public employment drive that included an employ- ment guarantee to university graduates. In 1964 the guarantee was extended to the graduates of vocational secondary schools and technical institutes, and the Assaad 87 policy was formalized in Law 14 of 1964. It was later made permanent in Law 85 of 1973, which extended the employment guarantee to demobilized military conscripts of all educational levels, a provision that was abrogated in 1976 (Hansen and Radwan 1982). The employment guarantee followed a period of rapid expansion in the num- ber of graduates from all levels of education. (See Sanyal and others 1982 and World Bank 1989 for a detailed description of the Egyptian education system.) Primary education expanded from 1 million students in 1952 to nearly 3.5 mil- lion in 1965-66, a rate just under 9 percent a year (Richards 1992). Preparatory and secondary enrollment expanded even faster, multiplying sixfold and three- fold, respectively, from 1956 to 1961. The expansion underscored the commit- ment of the Nasser regime to making educational opportunities more accessible to the mass of the population. However, despite the rapid growth of enrollment, the education system was starved of resources because of major increases in military spending and the demands of the state-led import-substitution drive. In an attempt to substitute for the large number of expatriate technicians who left the country after the Suez Crisis in 1956, the government also greatly expanded access to higher education. Enrollment increased from 50,000 in 1952- 53 to 97,000 in 1961-62. Combined with the abolition of fees for higher educa- tion institutions in 1963, the employment guarantee for graduates boosted the demand for education. The employment guarantee greatly enhanced the private benefits of university education, and the abolition of fees significantly lowered private costs (Richards 1992). Operation of the Employment Guarantee Scheme With the exception of medical graduates and teachers, whose assignment was the responsibility of the relevant ministries, the law extended to graduates the right to apply to the Ministry of Manpower and Vocational Training for a pub- lic sector job. In addition to receiving the graduates' applications, the ministry solicited requests for graduates from government agencies and state-owned en- terprises. There was little incentive for these agencies to limit their requests because appointees came with a budgetary allocation to cover their salaries. Nonetheless, the requests consistently fell short of the number of applicants, presumably because the budgetary allocations did not cover the full costs of employment. The ratio of applicants to requests was 1.2 in 1977 and rose to 5.1 in 1981 (Fergany 1991a). Excess supplies of applicants were usually allocated to local authorities, which were allowed to use them at their discretion (Hansen and Radwan 1982). Until 1978, with few exceptions, public agencies and enterprises were not allowed to hire permanent workers outside this centralized system of labor force allocation. However, many agencies circumvented this restriction by hiring workers on temporary contracts. In 1978 public enterprises were allowed to opt out of the centralized labor force allocation scheme, to set their own hiring levels, and to select their own workers. Thus, the brunt of the employment guar- Table 1. Number of Graduates and the Queue for Government Employment, Egypt, 1995 Structure Average Percentage of (percentage annual graduates still seeking Number of graduates of all growth rate, government appointments Area of specialization (by year of graduation) graduates) 1983-91 (by year of graduation) and type of institute 1983 1991 1983-91 1983-91 (percent) 1983 1984-90 1991 Area of specialization Health sciences 9,998 7,230 74,185 2.1 -4.1 0.2 0.7 0.8 Agriculture 8,194 8,147 72,380 2.1 -1.5 1.3 36.3 48.5 Commerce 28,441 26,293 282,434 8.1 -1.8 1.0 21.5 32.5 Engineering 8,575 7,614 78,386 2.3 -2.6 1.1 8.2 14.1 Sciences 4,370 3,625 37,792 1.1 -2.2 1.5 25.9 38.5 Law 9,562 17,848 124,308 3.6 6.9 1.4 18.3 26.7 Teacher training 15,688 22,446 203,677 5.9 2.9 0.3 0.5 4.6 Other bachelor's degrees 20,444 27,276 226,886 6.5 1.5 1.1 16.7 31.1 All universities 105,272 120,479 1,100,048 31.7 0.5 0.9 15.0 24.3 Type of institute Commercial technical institute 17,709 25,993 218,902 6.3 4.7 18.4 35.2 46.3 Industrial technical institute 4,622 13,015 79,022 2.3 19.2 23.5 46.8 38.7 Other technical institute 2,115 5,241 29,202 0.8 10.3 3.9 12.1 10.8 All technical institutes 24,446 44,249 327,126 9.4 8.3 17.9 34.3 39.3 Agricultural vocational secondary 22,267 35,103 251,107 7.2 5.0 60.6 51.6 38.0 Commercial vocational secondary 107,672 112,898 1,015,129 29.2 0.9 42.0 50.0 48.0 Industrial vocational secondary 58,412 131,472 779,987 22.5 11.4 58.0 48.3 29.2 All vocational secondary 188,351 279,473 2,046,223 58.9 5.4 49.2 49.1 37.9 All graduates 318,069 444,201 3,473,397 100.0 4.1 30.8 36.5 34.4 Source: Government of Egypt, Ministry of Manpower and Vocational Training. Assaad 89 antee fell on the government sector, which includes the central government min- istries, local government authorities, and the service and economic authorities. During the twenty-year period from the early 1960s, when the employment guarantee scheme was opened to secondary school graduates, to the early 1980s, when severe budgetary pressures resulting from a decline in oil revenues began to manifest themselves, the burden of the employment guarantee scheme on the public sector increased significantly. The number of graduates grew at about 12 percent a year from 1963 to 1983 (CAPMAS 1965, 1985), compared with an overall rate of growth of the labor force of about 2 percent a year. The rate of growth in the number of graduates has since 1983 slowed to about 4 percent a year, but graduates now constitute the bulk of new entrants to the labor market. As shown in table 1, there were about 444,000 new graduates in 1991, which compares with a net increment to the overall labor force of 360,000, at the prevailing growth rate of about 2.5 percent a year. Although the latter figure represents the net growth of the labor force rather than the number of new entrants, it leaves no doubt that graduates now constitute the majority of new entrants. These figures show why the public sector employment guarantee to gradu- ates has become untenable. In its attempt to respond to the growing burden, the government has avoided the outright abolition of the program. Instead it has established two policies. The first is to curtail the supply of graduates through direct rationing of enrollment by the Ministry of Education. The second is to increase the waiting period for a government appointment in the hope that some graduates will drop out of the queue. The first policy-to limit enrollments-has significantly reduced the rate of growth in the number of graduates and shifted their composition toward spe- cializations that are in greater demand in the private sector. The number of university graduates peaked at about 130,000 in 1986 and has been declining or stable ever since, with the decline occurring across all specializations except for law and teacher training.' The share of university graduates among all gradu- ates declined from 33 percent in 1983 to 27 percent in 1991. In contrast, the number of vocational secondary school graduates grew about 5.4 percent a year over the same period, with their share increasing from 59 to 63 percent. The mix of specializations at the vocational level has shifted strongly away from the commercial track and toward the industrial track, which has grown nearly 20 percent a year. This shift resulted from the concerted efforts of the Ministry of Education to adjust the mix of graduates in favor of the industrial track by building more of the substantially more expensive industrial secondary schools. (World Bank 1989 provides a detailed account of the changes in policy toward technical education in Egypt.) Because postsecondary technical insti- tutes are a relatively recent addition to the Egyptian education system, their enrollment grew rapidly, at a rate of 12 percent a year, until 1987 but then 1. The statistics in this section are based on data provided by the Ministry of Manpower and Vocational Training, some of which are summarized in table 1. 90 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. 1 slowed to 3 percent a year from 1987 to 1991. The mix of technical institute graduates also shifted significantly in favor of industrial specializations. The second policy response-to increase the waiting period for a government job-has resulted in waiting periods of up to thirteen years. By law, university graduates must wait at least two years and secondary school and technical insti- tute graduates must wait three years after graduation before they can apply to the Ministry of Manpower for a government job. The waiting period was de- signed to allow male graduates to complete their military service. With the ex- ception of the period from 1967 to 1973, when conscripts could serve indefi- nitely, the duration of military service is normally one year longer for secondary and technical institute graduates than for university graduates. By 1984 the pe- riod between graduation and appointment had been extended to three and a half years for university graduates and to four years for vocational secondary and technical institute graduates; by 1987 it had been extended to five and six years, respectively (Handoussa 1989). Hiring through the centralized labor force allocation system of the Ministry of Manpower has been suspended but has not been formally abolished. As of 1995 the last cohorts of graduates who were offered appointments were the 1983 university graduates and the 1982 voca- tional secondary and technical institute graduates. The Ministry of Manpower maintains a registry of applicants to public sector jobs by year of graduation. Although graduates are removed from the registry as a matter of course if they obtain a public sector job, many do not give up their position in the queue if they get a private job, thus reserving their right to a public sector job when their turn comes up. An attempt to remove graduates from the registry when they obtain a position in a formal private sector firm resulted in a mass resignation of graduates from these firms (see Assaad 1996b and "Escapees to the Public Sector" 1992). This episode, which received a lot of attention in the Egyptian press, indicates the value that graduates continue to attach to public employment. As shown in table 1, 36.5 percent of individuals who graduated between 1984 and 1990 were still queuing for public sector jobs as of mid-1995. The queuing rates vary substantially by level of education and area of specialization. Because appointments have already been made to virtually all of the 1983 university graduates, their queuing rate is only 0.9 percent. The queuing rate increases to 24 percent for 1991 graduates. Health sciences and teacher training graduates have the lowest queuing rates because their appointments are handled directly by the relevant ministries. Some categories of university graduates, such as holders of agriculture degrees, have queuing rates well in excess of 30 percent. Overall, however, queuing rates among university graduates are substantially lower than those of vocational secondary school graduates, which approach 50 percent in some specializations. Unlike university graduates, queuing rates among voca- tional secondary graduates have been falling with year of graduation. This may indicate that some recent graduates are no longer bothering to register with the Ministry of Manpower because the probability of obtaining a government ap- Assaad 91 pointment has dwindled. The drop in queuing rates is especially noticeable for graduates from the industrial track, who are likely to be employable in the pri- vate sector. Although the Ministry of Manpower data on graduates are not disaggregated by gender, there is considerable indirect evidence to indicate that women are much less likely to drop out of the queue than men. This can be ascertained by the fact that unemployment rates are typically much higher for female gradu- ates than for their male counterparts (Fergany 1991b). The lower dropout rate can also be ascertained from the increasing feminization of the civil service work force (see section II). Wage Setting in the Government Sector Wage setting in the Egyptian government was driven by nonmarket consider- ations well before the institution of the graduate employment guarantee. The Law of the Price-List of Educational Certificates of 1951 stipulated a fixed ini- tial salary for each educational certificate and a system of periodic increments according to seniority and level of education, regardless of the position occu- pied. Basic wages in the government are still being set in this manner. Compen- sation levels are tied to job grades, with lower and upper bounds for each grade set according to a unified schedule. Each year the worker is entitled to a periodic increase so long as the resulting wage does not exceed the maximum for the grade. In addition to basic wages, workers can receive allowances for hazardous work, accommodation, and various other aspects of the job. The employing agency can also pay incentives for good performance. The sum total of allow- ances and incentives is limited to 100 percent of the basic wage. The public compensation system is laid out in articles 40 and 41 of the Civil Servants Law (Law 47 of 1978) and amended in Law 115 of 1983. II. EFFECTS OF THE EMPLOYMENT GUARANTEE SCHEME The need to absorb an ever larger number of workers every year has had major implications for the public sector's finances, performance, and size rela- tive to the rest of the economy. Despite the abandonment of the socialist devel- opment path in 1973, public sector employment growth continued unabated in the 1970s (see El-Issawy 1983). This growth was fueled by the continuation of the employment guarantee to graduates as well as its extension to demobilized military conscripts from 1973 to 1976 (Hansen and Radwan 1982). The expan- sion was financed in part by the substantial increase in oil receipts flowing into public coffers after 1973. Employment and Wages in the Government Although it is difficult to reconcile the various sources of data on the exact size of government employment, the general trend is one of very rapid growth in the 1960s and 1970s, followed by a slowdown in the mid-1980s, when the oil 92 THE WORLD BANK ECONOMIC REVIEW, VOL. II, NO. 1 windfall of the 1970s disappeared (see El-Issawy 1983 for 1952-78 and Handoussa 1989 for the trend up to 1986). Since 1981, the Central Agency for Public Mobilization and Statistics (CAPMAS) has reported the distribution of employment by institutional sector as measured by the Labor Force Sample Sur- vey (LFSS, see table 2). The LFSS confirms that government employment grew rapidly from 1981 to 1984 (8.4 percent a year) and then slowed markedly be- tween 1984 and 1995 (4 percent a year), when it was still growing faster than overall employment. Real government wages rose during the boom period of the 1970s and early 1980s but then dropped precipitously thereafter, as the government attempted to bring its finances under control. According to Zaytoun (1991), real govern- ment wages rose to a peak of 19 percent above their 1974 level in 1981 but by 1986 had fallen to nearly 60 percent of their 1981 level. Although more recent data on government wages are not readily available, real wages have probably continued to decline because cost-of-living adjustments have been kept well be- low the inflation rate by agreement with the International Monetary Fund. Employment in Public Enterprises No comprehensive series of public enterprise employment is available for the period 1960 to 1975, but Handoussa (1989) estimates that public enterprise employment grew 8.7 percent a year from 1960 to 1966 and 2.9 percent from 1966 to 1976. Data from the Public Enterprise Information Center suggest that employment grew about 5 percent a year from 1974 to 1979. In 1979 the policy allowing public enterprises to opt out of the centralized labor force allocation scheme became effective. Employment in public enterprises subsequently slowed to 1.4 percent a year from 1979 to 1986-87 (Zaytoun 1991). The more recent Table 2. Distribution of Employment by Institutional Sector, Egypt, Selected Years Percentage of total employment Public Total employmentb Trend (index, Year Government enterprises Private Othera (thousands) 1984 = 100) 1981 19.9 13.4 66.4 0.2 9,946 84 1984 21.4 10.8 64.9 3.0 11,819 100 1988 18.0 8.1 72.8 1.2 16,263 138 1990 22.9 10.5 65.6 1.0 14,361 122 1991 22.9 9.8 66.1 1.2 14,489 123 1992 24.2 9.8 64.9 1.1 14,399 122 1995 25.9 9.0 63.9 1.2 15,208 129 Note: Data are for the population ages twelve to sixty-four, based on the May rounds of the CAPMAS Labor Force Sample Surveys in 1981, 1984, and 1995; the October round in 1988; the December round in 1990; and the combined annual results in 1991 and 1992. a. Includes joint ventures and foreign-owned enterprises. b. Total employment in 1981 and 1988 is not comparable to other years because of significant methodological differences in the way female employment was measured. Source: CAPMAS Labor Force Sample Survey (various years). Assaad 93 data from the LFSS, in table 2, show a slowly declining share of public enterprise employment in total employment. Composition of Public Sector Employment In addition to its effect on the overall size of government employment and wages, the guaranteed employment scheme has skewed the composition of pub- lic sector employment toward more graduates, most of whom are white-collar workers. It has also contributed to an increasing feminization of the govern- ment labor force because of the high likelihood that female graduates will re- main in the job queue. As shown in table 3, nearly 57 percent of male and 95 percent of female government employees are graduates (of secondary, postsecondary, or univer- sity education). The proportion in public enterprises is lower than in govern- ment, but nonetheless considerably higher than in the private sector. Graduates constitute a tiny fraction of employment in private agriculture and less than 20 percent of employment in the private nonagricultural sector. Moreover, the government sector is the largest employer of graduates, especially for females. More than two-thirds of employed females work for the government, and nearly 80 percent work for the public sector broadly defined. The increasing feminization of the government labor force can be readily seen from table 4, which is based on a comparison of census data from 1976 and 1986. Female employment in the government has grown at a rate of 8.7 percent a year compared with 2.4 percent for males. The government sector is increas- ingly becoming the dominant employer of female labor, accounting for more than half of female employment in 1986. Finally, the proportion of females in the government has increased from 15 to 25 percent, considerably more than it has in other sectors. Open Unemployment For a country where it is virtually impossible to get unemployment benefits, Egypt's open unemployment rate of 7 to 10 percent is relatively high. The defi- nition of open unemployment used in Egypt is a person who did not work at all, but was able and desiring to work and was searching for it, during the reference period, which is one week in the LFSS and one day in the census. In practice, graduates waiting for government jobs are considered to be actively searching for work. Most of the open unemployment is comprised of graduates and can be attributed to queuing for government jobs. As shown in table 5, graduates, par- ticularly secondary school graduates, have made up the vast majority of the unemployed since the mid-1980s, and their share has increased over time. The sharp increase in graduate unemployment rates between 1976 and 1986 can be attributed to the slowdown in the government hiring of graduates that occurred in the 1980s, when the waiting period was extended from between two and three years to between five and six years. The increase in graduate unem- ployment rates in 1987 can be discounted because of differences in definitions Table 3. Employment by Sector and Educational Attainment for Males and Females, Egypt, 1988 Males Females Below Intermediate University Below Intermediate University Sector intermediate and above and above All intermediate and above and above All Government 883,340 649,646 541,024 2,074,010 46,150 547,987 274,606 868,743 Public enterprise 668,155 305,737 175,429 1,149,321 36,790 9s,506 40,190 172,486 Private agriculture 3,640,616 193,065 37,770 3,871,451 4,051,966 49,221 598 4,101,785 Private nonagriculture 3,154,515 513,859 276,796 3,945,170 810,798 108,298 60,874 979,970 Other 54,631 62,384 43,731 160,746 5,456 7,803 16,642 29,901 All sectors 8,401,257 1,724,691 1,074,750 11,200,698 4,951,160 808,815 392,910 6,152,885 Note: Data are for individuals ages twelve to sixty-four. Below intermediate denotes individuals who did not earn a diploma from a secondary school. Intermediate and above denotes individuals who earned a diploma from a secondary school or a two-year postsecondary technical institute. University and above denotes individuals who earned a bachelor's or master's degree at a university or other higher education institution. Source: Calculated from CAPMAS Labor Force Sample Survey, October 1988 round. Assaad 9S Table 4. Distribution of Employment by Institutional Sector and Gender, Egypt, 1976 and 1986 (percent) Females As a Males percentage Structure Growth rate Structure Growth rate of total Sector 1976 1986 1976-86 1976 1986 1976-86 1976 1986 Government 16.0 18.2 2.4 40.7 55.9 8.7 15.4 25.5 Public enterprise 9.4 10.3 2.0 10.9 10.2 4.9 7.6 10.0 Agriculture 50.1 41.7 -0.7 23.0 10.6 -2.2 3.2 2.7 Private nonagriculture 24.4 29.9 3.1 25.4 23.3 4.7 6.9 8.0 Total 100.0 100.0 1.1 100.0 100.0 5.5 6.7 10.0 Total (thousands) 9,430 10,568 n.a. 676 1,176 n.a. n.a. n.a. n.a. Not applicable. Source: CAPMAS Population Census for 1976 and 1986. and data collection methodologies between the population census in 1986 and the LFSS in 1987. Still, the data from 1987 to 1995 indicate an increasing trend in the rate of unemployment, accounted for primarily by the growing proportion of secondary graduates among the unemployed. These data confirm the pattern observed in section I, which indicates that vocational secondary school gradu- ates have the highest queuing rates for government jobs. Because eligibility for the employment guarantee starts at the secondary level, the greatest labor mar- ket distortion should occur at that threshold level. Disaggregated by gender, unemployment data reveal that female gradu- ates are disproportionately represented among the unemployed. In 1995 women made up 50 percent of unemployed graduates but only 25 percent of employed graduates. Among secondary school graduates, the vast majority of whom graduated from vocational schools, the female unemployment rate was 50 percent compared with 23 percent for their male counterparts. The rate for female university graduates was lower (16 percent) but still consid- erably higher than that for male university graduates (9 percent). These re- sults confirm that female vocational secondary school graduates, most of whom pursue the commercial track, are the least able to find employment outside the government. III. A MODEL OF SECTOR SELECTION AND WAGE DETERMINATION IN EGYPT In this section, I present a simplified Harris-Todaro-type model to analyze the public sector employment guarantee and compensation policies in Egypt. I then lay out the econometric model and empirical estimates of sector-specific wage equations that take into account the simultaneous selection process into nonagricultural wage work and into the public and private sectors. Table 5. Distribution of Unemployed Workers and the Unemployment Rate by Educational Attainment, Egypt, Selected Years (percent) Indicator Ability to read Below Above Number and year Illiterate and write intermediate Intermediate intermediate University All (thousands) Unemployed workers 1976a 27.1 12.6 9.0 38.0 2.2 11.0 100.0 513 1986a 14.6 6.5 4.5 52.2 5.0 17.2 100.0 1,574 1987 2.7 2.8 2.2 60.5 7.9 22.6 100.0 1,354 1990 12.9 4.1 8.4 53.3 6.8 14.6 100.0 1,345 1991 3.7 1.7 3.5 66.8 8.2 16.0 100.0 1,397 1992 1.2 1.0 1.2 70.4 9.1 17.0 100.0 1,416 1995 1.6 1.4 1.4 74.6 8.4 12.6 100.0 1,774 Unemployment rate 1976a 2.5 2.5 7.0 20.6 13.5 10.9 5.0 n.a. 1986a 4.1 4.4 11.0 28.8 27.2 25.9 12.3 n.a. 1987 0.4 1.5 1.5 31.8 19.8 18.8 7.4 n.a. 1990 2.6 1.8 12.7 23.2 15.4 12.3 8.5 n.a. 1991 0.8 0.8 5.6 30.2 19.3 14.0 8.8 n.a. 1992 0.3 0.5 2.1 30.2 20.1 14.6 9.0 n.a. 1995 0.5 0.7 2.6 31.9 18.6 11.2 10.4 n.a. n.a. Not applicable. Note: Data are for individuals ages twelve to sixty-four. Data for 1987 and 1990 are from the December round of the Labor Force Sample Survey, data for 1991 and 1992 are from the combined annual results, and data for 1995 are from the May round. All are based on a one-week reference period. The ability to read and write denotes literate individuals who did not complete primary school. Below intermediate denotes individuals who earned a diploma after five years of primary school or three years of preparatory school. Intermediate denotes individuals who earned a secondary school diploma. Above intermediate denotes individuals who earned a diploma from a two-year postsecondary technical institute. University denotes individuals who earned a bachelor's or master's degree at a universliy or other higher education institution. a. The data for 1976 and 1986 refer to individuals ages fifteen and above and are based on a one-day reference period. Source: CAPMAS Population Census for 1976 and 1986. CAPMAs Labor Force Sample Survey for 1987,1990, 1991, 1992, and 1995. Assaad 97 A Simple Supply and Demand Model for Educated Labor The effect of the employment guarantee and public sector compensation poli- cies in Egypt can be readily analyzed using a simplified Harris-Todaro-type model as follows. Consider the model shown in figure 1; Let SS be the supply curve of educated labor without the policy and DD be the private sector demand for this labor. Government employment adds a fixed amount to labor demand, shifting the demand curve to DD'. If both the public and private sectors pay the market- clearing wage, the wage would be w*, private employment would be LO, and total employment would be L;0. There would be no graduate unemployment. Assume instead that government wage-setting practices are such that wages are set at-w above the market-clearing wage w*. Assume that each graduate has a probability p of receiving the government wage w. The prevailing wage in the private sector will be the expected wage we = piw. The probability of p employ- ment is assumed to depend on the number of unemployed graduates Lu (or the length of the queue for government jobs) as follows: p = f(L'), where f(O) = 1 and f'(Lu) < 0; 0 S p < 1. we will fall on the intersection of a rectangular hyperbola (the curve EE) and the supply curve (see Basu 1984). At wage we > w*, private sector employment of graduates will decline to L'R and unemployment will be L' . Public sector employment is assumed to be invariant to the wage. Figure 1. Supply of and Demandfor Graduate Labor in the Presence of a Public Sector Wage Floor Wage, w D' S D -w -____ - E ull~~~~~~~L LI LOI Irdat ab r I I I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~,L 98 THE WORLD BANK ECONOMIC REVIEW, VOL 11, NO. I If the increased demand for education at the secondary and university levels prompts the government to shift resources to these levels, the supply curve will shift outward. In that case the market clearing will be w'* < w*, and the new prevailing wage will be we. Because it is not clear whether we will be above or below the original market-clearing wage w*, the policy might increase or reduce graduate wages in the long run. As discussed in earlier sections, the Egyptian labor market exhibits all the features predicted by this simple model: a queue of unemployed graduates wait- ing for a government appointment and limited private sector employment of graduates. However, a direct comparison of public and private wages does not support the idea of a public sector wage floor. Average wages in the public sector appear to be substantially lower than those in the private sector and have lagged behind inflation in recent years. However, once wages are corrected for observed heterogeneity and for selectivity into the two sectors, and when the difference in nonpecuniary benefits across the two sectors is taken into account, the attraction of public sector jobs to graduates becomes readily apparent. Econometric Model The wage equations for male and female nonagricultural wage workers are based on a standard Mincerian human capital model. Variations in wages are assumed to be associated with differences in educational attainment and on-the- job experience. Regional variables are included to control for differences in cost- of-living and labor market conditions. Because public sector wages are not nec- essarily set to equal marginal productivity as assumed in the Mincerian approach, I distinguish between the two sectors by allowing each to have its own wage determination process. The public sector is defined to include the government (or civil service) and the public enterprise sector. Everything else is included in the private sector. Several other studies have examined public-private wage differentials in develop- ing countries (Lindauer and Sabot 1983; Corbo and Stelcner 1983; Psacharopoulos 1983; Al-Qudsi 1989). These studies do not take into consideration the endogeneity of the sectoral allocation process as is done in some of the more recent studies (van der Gaag and Vijverberg 1988; Stelcner, van der Gaag, and Vijverberg 1989; van der Gaag, Stelcner, and Vijverberg 1989; Terrell 1993). Because nonagricultural wage workers are themselves not a random sample of the population, I go a step further in this study by accounting for the endogeneity of participation in wage work in a context where self-employment is common. Tansel (1994) corrects for selectivity into wage work but does not disaggregate by sector. Vijverberg (1993) adjusts for selection into wage work (as well as other labor market states) and selection by region but does not address public-private differentials. I abstract from the issue of selection into the two components of the public sector. I assume that the wage determination process in public enterprises is identical to that in the govern- ment except for a shift parameter. Terrell (1993) takes into account nonrandom selection into the two components of the public sector. I do not take into account Assaad 99 endogenous selection into different levels of education. See Willis and Rosen (1979) for a discussion of the issues involved. I use a double selection framework that extends the single selection proce- dures developed by Heckman (1976, 1979) and Lee (1976) to account for two selection rules jointly determining inclusion in a particular subsample. The double selection model used here, which is based on the bivariate probit technique, was developed by Tunali (1986) to analyze two consecutive migration decisions, each with two possible outcomes. Vijverberg (1993) has developed a model to deal with a similar double selection process, but in his model each decision is allowed to have more than two outcomes. Using the dichotomous variables D, and D2, the two selection rules for participation in nonagricultural wage work and sector selection can be expressed as: D ( 1 participate in nonagricultural wage work if I;> 0 l0 otherwise (if I; < 0) D= 1 become a nonagricultural wage worker in the public sector if 12 > 0 10 become a nonagricultural wage worker in the private sector if 12 < 0 where l; -and 12 are latent variables indicating the difference in the worker's utility between nonagricultural wage work and other labor market states and between public and private nonagricultural wage work, respectively. The deci- sions are not necessarily sequential. The methodology used allows for the two decisions to be simultaneous. The worker's utility in each sector takes into ac- count both pecuniary and nonpecuniary aspects of participation in each sector as well as noncompetitive barriers to entry, which can take the form of waiting queues or other costs of entry. Omitting subscripts indicating a particular individual, the latent variables can be written as a linear function of observable characteristics and an error term as follows: (1) I;= lZ+ o (2) 12=72Z+e2 where Z is a vector of individual characteristics; y, and y2 are vectors of un- known parameters; and el and E2 are zero-mean, constant-variance disturbance terms. Because l; and IŽ are not observed, but sector selection is, equations 1 and 2 can be estimated as a system of two dichotomous dependent-variable equations. The wage equations in the public and private sectors are specified according to the standard Mincerian human capital model, where log wages y are assumed to depend on human capital characteristics and regional labor market and cost- of-living differences as follows: (3) yp Pop + PXP + VP (4) YR = OR + fRXR + VR- 100 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Xp and XR are vectors of characteristics for workers in the public and private sectors, respectively; Pop and iOR are the two intercepts; and fip and PR are vectors of unknown parameters other than the intercept. The vector of disturbances U = (e,Ie2,VPVR) is assumed to have a multivariate normal distribution with zero means and covariance matrix I. With the usual standardization of the unidentifiable variances of the dichotomous dependent-variable models, I is given by: 1 P12 P1pyp PIR R 1 S = Pl2 1 P2POP P2ROR P1pGyp P2POP

w and define units in such a way that w and w77 incorporate nonwage benefits as well. As mentioned above, we assume the probability of inspection to be higher for larger firms and let ,B be strictly increasing in O6. Although O6 itself is directly unobserv- able, in this simplified setting there is a one-to-one correspondence between firm size and Oi. We can therefore represent an inspection policy that targets larger firms as a function 1 that is increasing in 0,. 3. There are several implications in allowing the penalty, r, and cost of legal avoidance, A, to be functions of employment. Consider, for example, A(L) = aL and F(L) = yL. In this case, there is a knife- edge situation depending on parameter values. If ivw - w > a, then firms prefer to avoid rather than comply, but if iv - w < a, then firms prefer to comply rather than avoid. Because this holds for any given value of 0, we only observe evading and complying or avoiding firms, but not all three. In the case of r(L), the penalty could be imposed on the number of workers employed when the firm evades. Labor demand under evasion becomes L(O,, w + fry). Inequality 3, which indicates when firms prefer to evade rather than comply, becomes 1t(0,, w + !3y) - t (8,, w) > PyL(0,, w + fry) / (1 - ,3). If w + fy = w, then fyL(O,, wv) > i(O,, w + fy) - t (6,, wi3) = 0, and firms prefer to comply rather than evade. Because this is true for any 0,, if y > [(wiv - w) 1 li], we would never observe any evading firms. 130 THE WORLD BANK ECONOMIC REVIEW, VOL. II, NO. I From equation 1, firms will evade rather than comply if (3) An=- n(0, w) - n(0, ) > XI(00) Firms will comply rather than avoid if (4) n(0j,w)- n(0,jT-) < A. We know from equation 2 and the envelope theorem that the derivative of the left side of the above inequalities with respect to O, is positive for iw > w. This is not altogether surprising because higher 0 firms will be larger and will obtain greater surplus from hiring at a lower wage than will smaller firms. We consider the situation when I increases in O, sufficiently sharply that the right side of inequality 3 intersects the left side from below as depicted in figure 1. We also separately considered the case when the right side of inequality 3 intersects the left side from above; for example, when 0(0,) = P. This yields a different pattern of firm compliance in which lower O0 firms comply, intermediate O firms evade, and larger 0, firms avoid. The main qualitative results, however, carry through whether the intersection is from above or below. We denote the values of 0, that satisfy inequalities 3 and 4 with equality as O and OA, respectively. Given the direction of inequalities 3 and 4, it follows Figure 1. Pattern of Firm Compliance Ast; I3()r/[1-P(0)]; A A oC OA Productivity parameter, 0H Squire and Suthiwart-Narueput 131 that we only observe firms (an interval of 0,) seeking to comply if Oc < O' or (5) A > P(c)r 1 - p(o ) If the transaction cost of avoidance A is not sufficiently large, there will be no compliance, because firms either evade or avoid. Firms will evade if An(O,) < A / [1 (0,) - r]; and they will avoid otherwise. For the remainder of the article, we assume that inequality 5 holds because it is the more interest- ing case. We now consider the pattern of compliance. For those firms with Oi < e c, A > Air(0,) > 1i(0i)T / [1 - P(0i)]. The first inequality implies that these firms prefer to comply rather than to avoid, while the second implies that they prefer to evade rather than to comply. Firms in this range therefore evade. For those firms with O, E (0C, OA), A > An(0,) still, but (0,)o / [1 - P(0,)] > An(O,). The first inequality again implies that firms prefer to comply rather than to avoid, but the second now implies that they prefer to comply rather than to evade. Firms in this second range therefore comply. Lastly, for firms with O > OA, p(Oj)r1 / [1 - P(0,)] > An(0j) still, but An(0i) > A. The first inequal- ity implies that firms prefer to comply rather than to evade, while the second now implies that they prefer to avoid rather than to comply. Firms in this last range avoid. We therefore get the result that smaller, less productive firms with O0 < Oc evade, those with oc < 0 < OA comply, and larger, more productive firms with Oi > OA avoid. Regulator Options Inequality 3 shows that the regulator can influence the degree of evasion through two variables: the probability of inspection and the severity of the pen- alty. Other things being equal, the regulator can minimize noncompliance by increasing these two variables. The effective penalty may reflect not just explicit statutory penalties (which as indicated above may be quite minimal) but also any reputation or other cost associated with being found in violation of labor regulations. The regulator can also influence the extent of avoidance through the precision and comprehensiveness of the relevant labor legislation that would work through the fixed transaction cost, A (see inequality 4). As loopholes in the legislation are progressively closed, the transaction costs associated with avoidance are likely to increase. Taking stock of these results, we conclude the following. First, increases in the probability of inspection 1 lower Oc. The proportion of firms complying increases, while the proportion of those evading decreases. Second, increases in the per firm penalty, r, have a similar effect. Third, increases in the specificity of the labor legislation raise OA. The proportion of firms avoiding decreases, and the proportion of firms complying increases. 132 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Although the regulator has sufficient instruments to eliminate evasion and avoidance, the empirical evidence makes clear that regulators choose not to do this. The regulator's choice reflects both the objectives of the regulator and the costs of enforcement. Depending on the objectives, the real costs associated with avoidance may also influence the regulator's view of the relative merits of evasion and avoidance. Labor Market Outcomes Does the likelihood of noncompliance increase as the (binding) minimum- wage increases and as the elasticity of labor demand increases? First, we estab- lish that this is indeed so. Second, we derive the circumstances in which an increase in a binding minimum wage results in increased demand for labor. And third, we discuss the implications of these results for the efficiency losses arising from labor market regulation. Consider the left side of inequality 3, denoted An, in figure 1. An increase in w shifts up At and therefore increases Oc. To determine the impact of changes in elasticity, we conduct a second-order Taylor series expansion on Anr around (w, OC). With some manipulation, this yields c 1 L(w, Oc) (W_W2 (6) L(w,O )(w-w) (w)- w) e 2 w where e < 0 is labor elasticity. Because expression 6 increases with the absolute value of e, an increase in the absolute value of e shifts up Ax and increases Oc while decreasing OA, which decreases the proportion of complying firms. Thus, the likelihood of noncompliance does increase with both the minimum wage and the elasticity of demand for labor, which makes intuitive sense. Both of these factors increase the costs of the minimum wage and therefore encourage more noncompliance. We now examine the effect of an increase in the minimum wage on labor demand. We conduct a partial equilibrium analysis and take the market wage as given. This could occur, for example, in a small formal sector in which the wage level is determined outside the market, although in a more general setting, changes in the level of the minimum wage could affect the market wage, w. Total labor demand LD(.) is given by (7) LD(.)=-1J c(O,,w)dG(O)dO- J c r(e1,ii)dG(o)do - IOA ir (0e, ,w)dG(O)dO where the three integrals correspond to labor demand from evading, complying, and avoiding firms, respectively. After some manipulation and simplification, the effect on labor demand of increasing w- is given by c C _OL(A6 (A -j f) (8) o6[L(Oc,w) - L(8c,i)] i[L(-ALw) -L(Aw)]+ oc (w " dGdO where °L is the partial of 0' with respect to w-. Squire and Suthiwart-Narueput 133 The result fits with intuition. Recall that Oc increases (and OA decreases) with w- and that L(6i, w) - L (0i, wi) is positive for w-> w. Each of the first two products in expression 8 is therefore positive and represents the increase in the proportion of noncomplying firms (AP) multiplied by the increase in labor de- mand as (each) firm switchs from-compliance to noncompliance. The in- tegral is negative and captures the decrease in labor demand from complying firms-those in the interval (Oc, OA)-caused by higher minimum wages. Be- cause the integral term is a second-order effect compared to the two product terms, expression 8 could well be positive, and an increase in the minimum wage could increase labor demand as a result of lower compliance. However, this depends on the size of the gap between the minimum and market wages. The expression is more likely to be positive the greater the gap between mini- mum and market wages, because L(0I, w) - L(O', w-) is larger and the proportion of complying firms is smaller. If the minimum wage is close to the market wage, however, then the proportion of complying firms is large, and the expression is likely to be negative because the effect of decreased labor demand on compliers will dominate. We therefore have a Laffer-curve type of relationship between the minimum wage and labor demand caused by incomplete compliance. This relationship contrasts markedly with the outcome in a world of complete com- pliance, where raising the minimum wage unambiguously lowers labor demand and increases unemployment. What are the implications of these results for the efficiency losses usually associated with labor regulations? There are at least three effects of an increase in minimum wages, assuming that enforcement costs remain fixed: an increase in the distortion for complying firms, a reduction in the costs of distortions as firms switch from being compliers to being noncompliers, and an increase in transaction costs as firms switch from being evaders to being avoiders. The first two effects arise immediately from expression 8, and the third derives from expression 4 defined with equality. As with labor demand, these results imply that efficiency losses could increase, stay constant, or fall as the minimum wage is increased. Moreover, the likelihood of a reduction in efficiency losses will increase with the size of the existing distortion, which, as we have seen, is one of the factors determining the likelihood of an increase in labor demand. This analy- sis points, therefore, to the existence of natural limits on the magnitude of effi- ciency losses. These results have two implications for policy. First, although the efficiency losses from compliance may arguably be justified on redistributive grounds be- cause compliance results in a transfer to workers, governments are unlikely to be willing to incur the costs (enforcement costs and efficiency losses) associated with full compliance. (Compliance does not necessarily lead to a more equitable outcome; although rent is transferred to employed workers, the number of em- ployed workers is reduced.) There is therefore likely to be an optimal degree of noncompliance. Second, because there is no redistributive transfer to workers under avoidance relative to evasion, and yet there are real resource costs associ- 134 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. 1 ated with the former, it may be preferable to reduce legal avoidance rather than evasion. Another factor is that the avoiding firms are larger than the evading ones in our setup and will therefore incur higher distortionary costs from com- pliance. Tightening legislation will also make it costlier for those that continue to avoid. The policy response should be to focus on tightening legislation in an effort to cut back on legal avoidance rather than to increase inspections and penalties per se. This result holds even if real costs are associated with evasion, as long as those costs are less than those incurred when avoiding. The results also bear on the recent debate in the literature regarding the im- pact of minimum-wage legislation on employment in the United States. Card and Krueger (1995) summarize an array of statistical evidence and conclude that increases in the minimum wage have not reduced employment and may even have increased it. Their empirical evidence is hotly disputed. We are not in a position to judge the empirical validity of their conclusion. We note, however, that our results and, in particular, the change in compliance following an in- crease in the minimum wage could explain what they claim to observe. More- over, our explanation does not require invoking the untested assumption of monopsony power or informational imperfections; rather it rests on a standard competitive model with incomplete compliance. Although our model may have more relevance for developing countries where enforcement is especially weak, in some of the "natural experiments" considered by Card and Krueger, non- compliance did increase as predicted by our model. IV. PUBLIC ENTERPRISES In this section we examine the impact of labor market regulations on public enterprises. We assume that public enterprises operate according to a different objective function and effectively face a looser budget constraint than private enterprises. Large and Growing Employment and Wage Bills Public enterprise employment and wage bills have often grown very rapidly in developing countries and have remained large. After a strong increase in pub- lic enterprise employment during the 1960s and 1970s, the share of public en- terprise employment in total employment remained relatively constant at be- tween S and 6 percent worldwide between 1978 and 1991 (see table 3). But regional differences in public enterprise employment remain large. Between 1986 and 1991 the share of public enterprises in total employment in Africa was 18.1 percent, significantly larger than the corresponding share in Latin America (3.7 percent) and Asia (4.7 percent). This share was as high as 46.4 percent in Niger, 45.3 percent in Senegal, and 37.2 percent in Zambia (Galal 1994). Even in those countries where real wages in the public sector have declined, these reductions have often been more than offset by an increase in public sector employment. In her study on public sector pay and employment policies, Nunberg Squire and Suthiwart-Narueput 135 Table 3. Weighted Average Share of Public Enterprises in Employment, Selected Countries, 1978-91 (percent) Region 1978-85 1986-91 1978-91 Africa 17.55 18.07 17.77 Asia 4.92 4.68 4.82 Latin America and the Caribbean 4.07 3.67 3.90 World 5.55 5.28 5.43 Note: Only a subset of countries within each region was used for deriving world averages. The Latin American and Caribbean regional average excludes Argentina, Brazil, Ecuador, Grenada, and Trinidad and Tobago. The African regional average excludes Burundi, C6te d'lvoire, Egypt, Kenya, Malawi, Niger, Senegal, Seychelles, and Tanzania. The Asian regional average excludes Indonesia, Myanmar, Sri Lanka, and Vietnam. Three-year moving average estimates were used to complete our time series in case of missing values when deriving regional averages. Regional weighted averages were obtained using GDP in current U.S. dollars as weights. Source: Galal (1994). (1988) found an excessive wage bill to be a major problem for the majority of countries where the World Bank pursues government pay and employment in- terventions. In Egypt, for example, the wage bill grew from 22 percent of gov- ernment expenditures in the mid-1970s to 33 percent in the late 1980s (Banerji and Sabot 1993). Noneconomic Objectives Public enterprises are often assigned employment goals, are required to act as model employers, and may be required to serve as employers of last resort. Although governments seek to protect real wages for workers, when this is not possible they may attempt to protect employment by cutting spending on nonwage items, compressing the wage structure, and cutting real wages. Public enterprises act as model employers and seek to protect workers against declines in the wage rate. In Morocco overall productivity in formal manufactur- ing fell between 1985 and 1990. However, although the productivity decline was larger in the public sector, real wages in the public sector rose 0.4 percent a year, while falling 2.3 percent in the private sector, even though they were higher than real wages in the private sector to begin with (Harrison 1993). In Bangladesh five out of seven public enterprises sampled had increases in average labor costs that outpaced increases in productivity. For a public enterprise producing steel billets and plates, average labor costs increased 55 percent between 1985 and 1992, but the value of output per worker actually fell 18.6 percent (World Bank 1994). When operating under tight financial constraints, however, governments may seek to protect employment by reducing spending on material. The ratio of wage to nonwage public expenditures has been increasing in many countries with fiscal constraints, which indicates growing distortions and imbalances in the input mix of public sector production. In Liberia, for example, the wage bill in relation to revenues increased from 36 percent in 1977 to 66 percent in 1981 (Lindauer 1988). 136 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 The government may also compress the wage structure to protect employ- ment of lower paid workers, often reducing wages for their more skilled employees below those offered for equivalent skills in the private sector. Lindauer, Meesook, and Suebsaeng (1987) find significant salary compres- sion for a number of African countries. In Ethiopia, for example, the lowest real salaries in the public sector fell 37 percent between 1975 and 1985, but real wages fell nearly 70 percent for the highest salaries (ILO 1990). In Zam- bia the salary ratio between the highest- and lowest-rank civil service em- ployee fell from 19.2 percent in 1970 to 6.9 percent in 1983 (see Lindauer 1988). As a consequence, the wage difference for skilled labor between the private and public sector has become particularly high. In Thailand in 1982 the differential for top positions in both sectors was as high as 350 percent (Lindauer 1988). Because there is a negative tradeoff between employment and the wage bill when governments are financially constrained, governments may resort to re- ducing real wages to protect employment. There has been a considerable reduc- tion in real wages in Africa. Public sector wages for the lowest salary groups fell 45 percent between 1975 and 1985, whereas the highest salaries in the public sector fell more than 60 percent on average during the same period. In Somalia, for instance, the real value of the lowest salary and of the highest salary in the civil service in 1985 was only 5.2 and 4.0 percent of their 1975 value, respec- tively (ILO 1990). Public enterprises may also be required to act as employers of last resort. To reduce unemployment, governments often guarantee jobs and provide funds to hire graduates, which make these employees costless for public enterprises. In Egypt such a policy has resulted in significant overstaffing (Gelb, Knight, and Sabot 1991). Similar employment policies are found in Central African Repub- lic, Guinea, Mali, Senegal, Sudan, and Togo, where university graduates are automatically hired by public institutions. Collectively, these factors have resulted in considerable overstaffing among public enterprises in many countries. A comparison of public and private manufacturing enterprises in Turkey has shown the existence of significant labor and capital surplus among public manufacturing enterprises: the ratio of public to private labor and of capital inputs per unit of output was 1.97 and 1.66, respectively (Gelb, Knight, and Sabot 1991: 1187). Overstaffing in the public sector in Egypt was as high as 40 percent in 1976. In part, this reflected the provision of funds to hire graduates, which made these employ- ees costless for public enterprises (Gelb, Knight, and Sabot 1991). As a con- sequence, whereas employment in Egypt's private sector grew only 7.8 per- cent, it rose 24.5 percent in public enterprises during 1976-86 (Banerji and Sabot 1993). Overstaffing in particular sectors is estimated to be as high as 93 percent in the spinning industry of Egypt, 54 percent in the transport sector of Ghana, 91 percent in the Bombay port of India, and 40 percent in the railway sector of Turkey (Banerji and Sabot 1993). Squire and Suthiwart-Narueput 137 Negative Profits Public enterprises are frequently not expected to make a profit. Noneconomic objectives such as employment or equity appear to outweigh efficiency consid- erations, making public enterprises a significant burden on the government bud- get. Gross operating profits of public enterprises in nearly all developing coun- tries were significantly negative between 1978 and 1991 (Galal 1994). Negative gross operating profits were between 2 and 3 percent of gross domestic product (GDP) in Africa and Asia for the period 1978-91. There was little or no improve- ment during the 1980s in Asia or Africa. Only in Latin America did gross oper- ating profits improve significantly between 1978 and 1991, from negative gross operating profits of more than 1 percent of GDP between 1978 and 1985 to a small surplus between 1986 and 1991. V. THE MODEL FOR PUBLIC ENTERPRISES In contrast to the pure profit-maximizing behavior of the private sector, pub- lic enterprises are assumed to maximize a modified measure of profits: (9) 7T(°i;Xw) = maxpOiF(K,L)- XwL- rK. L Here x is an exogenously given measure reflecting the net influence of a concern with the wage bill (employment generation and wages), on the one hand, and recognition of the need for fiscal discipline, on the other. If X equals 1, the gov- ernment has no special employment objective or the fiscal constraint dominates or both. In this case, public enterprises are required to maximize profits. With a more pressing national employment objective and a more relaxed fiscal situa- tion, x may approach 0, at which point labor becomes costless to the enterprise. This implies overstaffing and a drain on the central budget. In exceptional cir- cumstances, X could be negative. Alternative specifications of the objective function of public enterprises are possible and would yield different implications. Public enterprises may have an employment objective (to alleviate unemployment) or a high wage objective. Our specification reflects both these considerations as well as the stylized fact of high wage bills. Compliance with Minimum-Wage Laws Unlike private firms, the choices open to public enterprises with respect to labor market regulations may be more severely circumscribed. We can distin- guish three different institutional arrangements. First, public enterprises may have little control over their wage and employment decisions. There may be a central wage-fixing and wage-paying body that applies to all public enterprises. In this case, public enterprises will comply with minimum-wage laws because they have no choice. Or there may be a virtual prohibition on firing workers in public enterprises, in which case increases in the minimum wage will reduce 138 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 profits or increase losses. Second, each public enterprise may have the capacity to clhoose between the minimum wage and the market wage, but the effective penalty or probability of inspection may be higher for public enterprises. If it is easier for government regulatory authorities to ensure payment of fines by pub- lic than by private firms, for example, the effective penalty will be higher for public firms. Recourse to nonpecuniary measures-for example, firing of public managers-could also result in higher effective penalties for public than for pri- vate firms. Alternatively, the probability of inspection ,B could differ. Labor may be more aware of its rights in a public enterprise or have a higher expectation that its grievances will receive attention. If so, the frequency of complaints, and hence of inspections, may be greater for public enterprises. These factors will increase the likelihood of compliance. A third possibility is that the determination of wage payments is left to the enterprise as before but that the effective penalty and probability of inspection are the same for public and private firms. Because this case offers the most immediate comparison with private firms, we pursue it here to see whether pub- lic enterprises are more likely to comply than their private counterparts. If the answer is "yes," then we know that public enterprises are more likely to comply than private firms for all institutional arrangements because public enterprise compliance will be lowest for the particular one under study. To examine this issue, we note that the general form of the results for public enterprises is identical to that for private firms. In particular, we can derive two inequalities similar to inequalities 3 and 4 to determine whether public enter- prises will comply, evade, or avoid. The only difference is that the left side of these inequalities will reflect the modified objective function: (10) ;(0i;XW) - 7r(ej; Uw). In both inequalities the right side is the same for public as for private firms, and so, as before, we will observe evaders among public enterprises if inequality 5 holds. In the following, we assume that it does hold. These similarities (on the right side) and differences (on the left side) allow us to shed light on whether public enterprises are more or less likely to comply than private firms. To see this, note that the relevant margin is the compliance/eva- sion one and from equation 9, for k = 1, the private and public firm problems are identical. Therefore, if expression 10 is decreasing in X, then we know that pub- lic enterprises-for which in general X < 1-have less incentive to evade and more incentive to comply than private firms. This result holds, all things being equal, obviously, in particular, for a given 0;. Let L(w, 0j; X) solve equation 9. Then, expression 10, which reflects the in- centive to evade, can be written as (11) -X[wL(w,.) - iwL(iw, . )] + pO i {F[K, L(w,.)] - F[K, L(iw*,.)]1) To assess how the incentive to evade changes with the degree of fiscal discipline and the concern with employment, differentiate expression 11 with respect to X. Squire and Suthtwart-Narueput 139 Using the envelope theorem, this yields w-L(iw7,.) - w L(w,.). With inelastic labor demand, this expression is positive, and the incentive to evade decreases as the social concern with employment increases, fiscal discipline is relaxed and x is reduced, or both occur. With elastic labor demand, the expression is negative, and the incentive to evade increases as the social concern with employment in- creases, fiscal discipline is relaxed and x is reduced, or both occur. Thus, public enterprises are more likely to comply with minimum-wage legislation than their private sector counterparts in situations where the demand for labor is inelastic. This makes intuitive sense. For any positive X, an inelastic demand curve implies that compliance results in a higher wage bill, which in turn lowers the objective function for both private and public firms. From equation 9, however, any given increase in the wage bill will lower the value of the objective function of public enterprises less than that for private firms because X < 1, which reflects the fact that the wage bill is less "costly" in the public enterprise objective function be- cause of relaxed fiscal discipline. Empirical estimates of the wage elasticity of labor demand in manufacturing typically produce absolute values well below 1 (see, for example, the estimates provided in Bell 1994: tables 7 and 8, or in Revenga 1994: table 10). These values indicate that in general public enterprises are in- deed more likely than private firms to comply with labor market regulations even when confronted with the same regulatory environment. Public Enterprise Compliance and Efficiency If public enterprises are more likely than private firms to comply with mini- mum-wage legislation, then are efficiency losses greater for public enterprises? Here we show that the factor that leads to the result on compliance also has implications for the efficiency impact. To see this, consider the relationship be- tween public enterprise compliance and labor demand. The first-order condi- tion for a complying public enterprise is given simply by OipFL(K, L) = X iw, while that for an equivalent noncomplying private firm is given by OipFL(K, L) = w. De- note the latter solution as L-. The complying public enterprise therefore hires less labor than L- if x 2 w / w- (and hires more labor if x < w / w-).4 This outcome reflects the fact that X and w exert opposing forces on labor demand-? in- creases labor demand, but w reduces it. Evaluated purely from the standpoint of productive efficiency, it appears rea- sonable to assume that efficiency losses increase the farther away L is from L-. Evaluating this from a welfare standpoint requires an assessment of the alterna- tive public enterprise objective function. It follows that the introduction of a minimum wage could reduce efficiency losses for complying public enterprises as they move away from a situation of excess labor demand and closer to the free-market equilibrium. In effect, public enterprises have too many workers, so a minimum wage that cuts employment is beneficial. Thus, ?, < w / w- is suffi- 4. Since w > w > 0, w / wiJ'e (0, 1) while XE [1, - w), e know that there exists a critical value X. =wl/wi. 140 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I cient but not necessary for compliance by public enterprises to reduce efficiency losses.5 By contrast, in private firms, compliance increases efficiency losses. This contrast between public and private firms has noteworthy implications for policy. It suggests that from the standpoint of improving compliance, it is preferable to focus on public than on private enterprises because with public enterprises there are less likely to be increased efficiency losses that offset the gains in equity from greater compliance. VI. ANALYZING LABOR MARKET REGULATIONS: A CHECKLIST Based on the preceding analysis, the following checklist provides a means for heuristically evaluating the likely distortionary impact of minimum-wage regu- lations. Where the preliminary evaluation suggests significant distortionary costs, further research and analysis are needed. * Check whether the legislated minimum wage is potentially binding. As empirical studies demonstrate, minimum wages are not likely to have sig- nificant employment or other effects if they do not "bite." One helpful exercise would be to consider the relationship of the minimum wage to the wage distribution. Wage histograms that spike at or near the minimum wage rather than significantly to the right indicate situations where the minimum wages have more bite. Such histograms for Morocco, differenti- ated by firm size, indicate, for example, that minimum wages are consider- ably less binding for larger than for smaller firms. Similarly, the fact that only 6 and 24 percent of firms had average wages below 1.5 and 2 times the minimum, respectively, in Mexico, but that 27 and 71 percent did so in Colombia indicates that minimum wages potentially had more bite in Co- lombia (data are for 1989 for Mexico and 1987 for Colombia, from Bell 1994). Although this check will give a preliminary indication, in practice there may be several complications. For example, several legislated mini- mum wages may correspond to different sectors or classes of firms. Fur- thermore, because the total return to labor includes both wage and nonwage elements, simple comparisons of average to legislated minimum wages may not capture the true extent to which the minimum wages are binding. * If the legislated minimum wage is potentially effective, check the extent of noncompliance. To arrive at a quick assessment of the extent of noncom- pliance, it would be useful to interview the relevant staff in the country's department of labor. Their assessment of the situation could be supple- mented with data on the size of the enforcement budget, the number of inspectors and prosecutions, and the severity of fines. It would also be 5. Public enterprise compliance does not always reduce efficiency losses because such firms could end up hiring so few workers that efficiency losses are greater under compliance rhan under noncompliance. Recall that for x = 1 public enterprise behavior is identical to private firm behavior. Because compliance results in efficiency losses for private firms, public enterprise compliance at x = 1 results in efficiency losses. Squire and Suthiwart-Narueput 141 useful to check the relevant legislation to establish the scope for easy and legal avoidance through such means as use of temporary workers, appren- tices, or subcontracting. As a cross-check, a small survey of selected enter- prises in different segments (private/public, formal/informal) would be de- sirable. In addition to government administrative capacity, strong labor unions and judicial systems may also play an important role in determining the effectiveness of enforcement. In Ecuador and Morocco, for example, unions reportedly act as important channels for transmitting complaints regarding violations of labor regulations. In the United States, many plain- tiffs take cases of violations directly to court without involving the Depart- ment of Labor. If the legislated minimum wage is binding and enforced, check the relative size of the public and private sectors. If the private sector predominates, then the evidence already compiled provides a basis for concluding that distortionary costs are potentially significant and that further analysis is required. If, instead, the public sector predominates, one more test is re- quired because in this case the efficiency impact depends on the net strength of the employment objective and the fiscal constraint. One indication of the net strength of these factors is the extent of net financial flows from government to public enterprises. High figures could indicate a looser fis- cal constraint. For example, as a proportion of GDP in 1978-91, these fig- ures were 2.5 percent for Argentina, but -12.1 percent for Chile, and 16.6 percent for Algeria, but -6 percent for Egypt (Galal 1994). In the event that public enterprises appear to operate like profit-maximizing private firms, then, as with private firms, the evidence would point to potentially significant efficiency losses and would call for further investigation. To conclude, we return to the opening quote from Freeman. In our checklist we have set out three conditions that have to be fulfilled before we can expect to see significant distortions associated with legislated minimum wages. It is our presumption that in many developing countries these conditions are unlikely to be fulfilled. If this is true, then Freeman's surprise at the lack of evidence regard- ing the distortionary costs of minimum wages is explained, at least as far as broad, cross-country comparisons are concerned. Such costs could, however, still be significant in some countries at some times. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Anker, Richard, Theopiste Butare, and Andre Marinakis. 1992. "Minimum Wages in Developing Countries: Trends and Determinants." International Labour Office, Geneva. Processed. Ashenfelter, Orley, and Robert Smith. 1979. "Compliance with the Minimum Wage Law." Journal of Political Economy 87(2):333-50. 142 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Banerji, Arup, and Richard H. Sabot. 1993. "Wage Distortions, Overmanning, and Reform in Developing Country Public Enterprises." Policy Research Department, World Bank, Washington, D.C. Processed. Bell, Linda. 1994. "The Impact of Minimum Wages in Mexico and Colombia." Policy Research Working Paper 1514. World Bank, Washington, D.C. Processed. Card, David. 1991. "Do Minimum Wages Reduce Employment? A Case Study of Cali- fornia, 1978-89." NBER Working Paper 3710. National Bureau of Economic Re- search, Cambridge, Mass. Processed. Card, David, and Alan B. Krueger. 1995. Myth and Measurement: The New Econom- ics of the Minimum Wage. Princeton, N.J.: Princeton University Press. Freeman, Richard B. 1993. "Labor Market Institutions and Policies: Help or Hindrance to Economic Development?" Proceedings of the World Bank Annual Conference on Development Economics 1992. Washington, D.C.: World Bank. Freeman, Richard B., and Alida Castillo Freeman. 1991. "Minimum Wages in Puerto Rico: Textbook Case of a Wage Floor?" NBER Working Paper 3759. National Bu- reau of Economic Research, Cambridge, Mass. Processed. Galal, Ahmed. 1994. "Public Enterprises in Developing Countries after a Decade of Divestiture." Policy Research Department, World Bank, Washington, D.C. Pro- cessed. Garvie, Devon, and Andrew Keeler. 1993. "Incomplete Enforcement with Endogenous Regulatory Choice." Discussion Paper 873. Institute for Economic Research, Queens University, Kingston, Canada. Processed. Gelb, Alan, John B. Knight, and Richard Sabot. 1991. "Public Sector Employment, Rent Seeking, and Economic Growth." Economic Journal 101(September):1186-99. Harrison, Ann. 1993. "Morocco Private Sector Assessment: The Labor Market, 1993." Policy Research Department, World Bank, Washington, D.C. Processed. Heckman, James J., and Kenneth I. Wolpin. 1976. "Does the Contract Compliance Program Work? An Analysis of Chicago Data." Industrial and Labor Relations Re- view 29(4, July):544-64. ILO (International Labour Organisation). 1990. "Wages, Labour Costs, and Their Im- pact on Adjustment, Employment, and Growth." Geneva. Processed. Leonard, Jonathan S. 1985. "Affirmative Action as Earnings Redistribution: The Tar- geting of Compliance Reviews." Journal of Labor Economics 3(3):47-63. . 1990. "The Impact of Affirmative Action Regulation and Equal Employment Law on Black Employment." Journal of Economic Perspectives 4(4):47-63. Lindauer, David L. 1988. "Government Pay and Employment Policies and Government Performance in Developing Economies." Policy Research Working Paper 42. Policy Research Department, World Bank, Washington, D.C. Processed. Lindauer, David, Oey Meesook, and Parita Astra Suebsaeng. 1987. "Government Wage Policy in Africa: Some Findings and Policy Issues." Country Policy Department Dis- cussion Paper 8624. Country Policy Department, World Bank, Washington, D.C. Processed. Morrisson, Christian. 1993. "Le probleme des bas salaires et du salaire minimum dans les pays en developpement." Organisation for Economic Co-operation and Develop- ment, Paris. Processed. Squire and Suthiwart-Narueput 143 Nunberg, Barbara. 1988. "Public Sector Pay and Employment Reform." Working Pa- per Series 113. Country Economics Department, World Bank, Washington, D.C. Processed. Polinsky, A. Mitchell, and Steven Shavell. 1990. "Enforcement Costs and the Optimal Magnitude and Probability of Fines." NBER Working Paper 3429. National Bureau of Economic Research, Cambridge, Mass. Processed. Revenga, Ana. 1994. "Employment and Wage Effects of Trade Liberalization: The Case of Mexican Manufacturing." Policy Research Working Paper 1524. Policy Research Department, World Bank, Washington, D.C. Processed. Standing, Guy, and Victor E. Tokman. 1991. "Towards Social Adjustment: Labour Market Issues in Structural Adjustment." International Labour Office, Geneva. Pro- cessed. World Bank. 1994. "Bangladesh: Labor Market Policies for Higher Employment." South Asia Country Department, Washington, D.C. Processed. Economic Reform and Labor Unions: A General-Equilibrium Analysis Applied to Bangladesh and Indonesia Shantayanan Devarajan, Hafez Ghanem, and Karen Thierfelder Noting the trend toward more independent trade unions in developing countries, this article examines whether the presence of unions strengthens or weakens the benefits to be gained from economic policy reform. We show that the presence of "passive" unions-ones that choose their wage-employment contract given the firm's cost-minimizing strategy-increases the welfare gains from trade liberalization, be- cause trade reform lowers the wage premium enjoyed by the unionized sector, reduc- ing a distortion in the labor market. These gains are amplified when the unions are "active," namely, when they negotiate a contract with the firm that is off its labor demand curve. Such a contract results in featherbedding-paying workers more than their marginal product-and trade reform reduces the amount of featherbedding. The policy implication for Bangladesh-a country with strong trade unions and a protected unionized sector-is that the benefits of further trade liberalization may be greater than otherwise predicted. In Indonesia, where both unionization and im- port tariffs are low, allowing greater independence to unions. may preserve flexibil- ity and reward workers better than the current minimum-wage policy. The world is-moving toward more openness, political freedom, and civil liber- ties. Freedom of association and organization is now accepted as a basic human right in most societies. In many developing countries, workers have increased their demands for the right to associate, organize freely, and bargain collec- tively. Independent trade unions are now appearing in countries where in the recent past only state-sponsored unions were allowed. The transitions toward democracy and independent trade unions in Chile and Korea in the late 1980s are dramatic examples of this process. While the share of unionized labor in the workforce has been declining in many industrial countries, at least partly as a consequence of the increasing share of services in total employment, it has been rising in developing countries because of greater industrialization and democratization. Shantayanan Devarajan is with the Policy Research Department at the World Bank, Hafez Ghanem is with the Europe and Central Asia Country Department IV at the World Bank, and Karen Thierfelder is with the Department of Economics at the U.S. Naval Academy. This article was originally prepared as a background paper for World Development Report 1995: Workers in an Integrating World. The authors are grateful to three anonymous referees for comments on an earlier version. C) 1997 The International Bank for Reconstruction and Development /THE WORLD BANK 145 146 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I The political and social effects of independent trade unions are generally thought to be positive. Unions fought for democracy in Poland, equality and the end of racial segregation in South Africa, and national independence in Bangladesh, India, and Sri Lanka. The impacts of labor unions on efficiency, equity, and growth, however, are not clear. (For a detailed discussion of the economic effects of unions see Freeman and Medoff 1984.) On the one hand, unions can act as monopolists, increasing their members' wages and discourag- ing investment and job creation. The wage premia that unions obtain for their members are often at the cost of slower growth and lower wages and employ- ment for unorganized workers (see Blanchflower and Freeman 1990, Moll 1993, Panagides and Patrinos 1994, Park 1991, Standing 1992, and Teal 1994). On the other hand, Freeman (1980), Lee and Nam (1994), and Panagides and Patrinos (1994) present evidence that unions improve the distribution of income and reduce discrimination against women and ethnic minorities. Other evidence (Standing 1992) indicates that unions can help raise productivity and improve competitiveness. This article does not address the question of whether unions are "good" or "bad" for development. Taking as given the fact that independent trade unions are going to play a more important role than they did in the past, we explore the implications of this fact for economic policy. This focus is relevant for analytical and policy reasons. On the analytical side, economic reform in developing coun- tries has typically been evaluated assuming that labor markets are perfectly com- petitive (see, for example, Devarajan, Lewis, and Robinson forthcoming). This assumption is becoming increasingly heroic. How would the results change when union behavior is explicitly incorporated into the analysis? We develop a styl- ized, two-sector general-equilibrium model and graft onto it different types of unions. "Passive" unions take the firm's labor demand curve as given and deter- mine the optimal wage markup, similar to unions in de Melo and Tarr (1992). "Active" unions negotiate a wage and employment contract that captures some of the inframarginal rents, as described by McDonald and Solow (1981) and Johnson (1990). In active unions, firms are off their labor demand curves, and featherbedding occurs as excess labor is hired and workers are paid more than their marginal product. We test whether the outcome of economic reform is different under each of these specifications of union behavior. Although the analytical motivation of the exercise is clear, the policy rel- evance requires some explanation. After all, most workers in developing coun- tries are in the rural and informal sectors where unions do not exist. Formal sector employment is only around 17 and 57 percent of total employment in low- and middle-income countries, respectively. Nonetheless, there are two rea- sons to pay attention to the role of unions. The first is particularly relevant to low-income countries in South Asia and Africa, where the strategic importance of the unionized sector is much greater than is indicated by its share in total employment or gross domestic product (GDP). For example, Bangladesh's manu- facturing sector represents only around 10 percent of GDP, but manufacturing is Devarajan, Ghanem, and Thierfelder 147 by far the most important foreign-exchange earner, responsible for some 70 percent of exports. Trade unions in Bangladesh are particularly important in manufacturing, especially textiles (cotton and jute). Jute, which represents around one-third of net exports, has been in a crisis caused by low competitiveness and high unit labor costs, and the government has embarked on a massive restruc- turing program. (See World Bank 1995 for estimates of unit labor costs in Bangladesh's jute industry and comparisons with other Asian countries.) Under- standing the structure of the labor market, including the role of unions, and its links with the rest of the economy is key to the success of this effort. Other examples of strategic sectors in less-developed countries where labor unions play an important role include the textiles industry in Egypt, coal mines in India, and oil extraction in Nigeria. The second reason is particularly relevant for fast-growing middle-income countries in East Asia and Latin America. Those countries exploited their com- parative advantage in labor-intensive goods to achieve rapid growth in exports and overall growth. Often they did not allow independent unions, which some argue helped to keep wages low and maintain competitiveness. There is increas- ing evidence that labor repression was not necessary for export expansion and economic growth in East Asia. (For example, see Fields 1994 or Freeman 1993.) But workers in many of these countries have become increasingly vocal in de- manding free unions, and the international community is pressuring countries to respect workers' rights. Indonesia is a good example (as are China, Malaysia, and Singapore). Until recently, labor legislation in Indonesia limited the right to form unions, and only one large government-supported union, SPSi (Serikat Pekerja Seluruh Indonesia, the All-Indonesia Workers' Union), is operating. The Indo- nesian government is under domestic and international pressure to liberalize labor legislation. Understanding the implications of allowing new, more inde- pendent unions is key to enhancing worker welfare without jeopardizing Indonesia's tremendous economic achievements. Building on the insights of a stylized model, we analyze the interaction be- tween union behavior and economic policy in both Bangladesh and Indonesia using more detailed, country-specific models. In Bangladesh, unions are free, but markets are heavily protected. Although only about 3 percent of the labor force is unionized, unions are powerful players in the labor market. They have strong ties to political parties and are concentrated in urban centers where they can exert pressure on government. Particularly strong in highly protected or subsidized sectors, unions have been increasing their pressure on the govern- ment to ensure continued protection and support. This pressure has taken the form of widely publicized nationwide strikes and hartals (lockouts with civil protest), which have given industrial relations in Bangladesh a bad name. How does this situation affect the economic reform process? The existence of mili- tant unions clearly raises the political cost of economic liberalization policies. In contrast, Indonesia has adopted more market-friendly policies and low levels of protection. This has led to a boom in manufacturing exports and for- 148 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 eign direct investment and has put Indonesia on the path to rapid export- oriented, labor-intensive growth. The government has resisted pressure to liber- alize labor laws and has tried to ensure workers' political support by increasing legislated benefits. Minimum wages were raised substantially every year be- tween 1989 and 1994, almost tripling (in nominal terms) during this period. The government also introduced a Workers' Social Security Law including life insur- ance, retirement benefits, free health care, and workers' compensation, raising labor costs 12 percent. Yet, labor unrest continues to rise; the number of work- days lost because of (mostly illegal) strikes increased thirty-fivefold between 1989 and 1992. In section I we describe the changes in the labor market when there are either passive or active unions. We present a stylized model with each type of union behavior. In sections II and III we describe the computable general equilibrium (CGE) models and simulation results for Bangladesh and Indonesia, respectively. Section IV concludes. I. STYLIZED MODEL To focus on the effects of various types of union behavior, we begin with a simple, two-factor, two-sector general equilibrium model. The two sectors are agriculture (A) and manufacturing (M), with manufacturing being the protected sector. Each sector produces output by a production function using the two factors, labor (L) and capital (K): (1) XA = f(LA, KA) (2) XM =g(LM, KMJ. Capital is fixed and sector-specific. Factor supplies are held constant to fo- cus on the static efficiency gains from trade liberalization. Allowing factor supplies to be endogenous requires an explicit intertemporal model. Devarajan and Go (forthcoming) show that in such a model the welfare effects are not very different. That capital is sector-specific is also not crucial (Devarajan and Offerdal 1989). Here labor is the only mobile factor. The supply of labor is fixed: (3) LA+ LM = L*. Although labor is mobile, we observe a difference in the returns across sectors: (4) WA = PA [LA (5) WM = PM gLM. Devaralan, Ghanem, and Thierfelder 149 Whether we attribute this difference to productivity differences or union premia will play a crucial role in the analysis. We can define the average wage, w, as the wage that solves: (6) WALA + wMLM = wL . We can now define the wage premium, 4, in each sector as the sector's wage relative to the average wage: (7) A- W (8) m- W In what follows, we assume that the Os are exogenous for the benchmark (no-union) case. This implies that the observed differences in labor produc- tivity will remain even after labor is reallocated between sectors. We assume that these differences are caused by sectoral attributes (such as differences in technology or capital stock) rather than by individual characteristics. Con- sequently, there will be welfare costs, or second-best effects, if labor moves from a high-productivity to a low-productivity sector following a policy shock. This specification of the labor market allows us to maintain the factor pay- ment differentials observed in the base data, which are an important feature of the actual country models. For the union cases, we assume that OM is the outcome of the union's optimizing behavior (for passive unions) or the out- come of a bargaining solution between unions and management (for active unions). In these cases, the premium can change following a policy change, giving rise to different welfare effects. The demand side of the stylized model is standard. Imports and domestically produced goods in the same sector are imperfect-substitutes (see, for example, Devarajan, Lewis, and Robinson forthcoming). The single, representative con- sumer allocates expenditure in fixed shares over the "composite commodities" (a constant elasticity of substitution-CEs-aggregate of imports and domestic goods). The small country faces exogenous world prices in both exports and imports. There is a tariff on imports of manufactured goods. All income accrues to the representative consumer, including the tariff revenue collected, which is rebated in a lump-sum fashion. Passive Unions To incorporate union behavior, we attribute all differences in payments to labor in the manufacturing sector to a union markup, OM > 1. The union chooses this markup to maximize its utility subject to the firm's labor demand curve. The payment differential to labor in agriculture is constant (OA <1)- The passive union takes the derived demand for labor as given and chooses the wage premium that maximizes its utility over wages and employment. Fol- ISO THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I lowing de Melo and Tarr (1992), we assume that the union has a Cobb-Douglas utility function:' (9) U = (uO M - w)l (L - Ti )(1-l) where U is the union's utility over wages and employment; w is the minimum acceptable wage, which we take to be the economywide average wage; OM is the wage premium paid to union workers; L is the number of workers hired; and L, is the minimum acceptable employment for union type i, which equals P if pas- sive and A if active. Manufacturing output, XM, is produced using a CES function over two inputs, labor (L) and capital (K): (10) XM = q[cxLP + (1 -c)K-P ](P1. The elasticity of substitution between labor and capital in production is: (11) a 1 (1 + p) Given this production technology, the derived demand for labor is: ( 12 ) ( # M ) ( A )MW where PM is the output price of manufacturing. Substituting the labor demand curve into the union's utility function and choosing the wage premium to maxi- mize utility, we find that the optimal wage premium OM solves: (13) ~~(OM -1) (L - LP) (13) X~~~~~~~m (1 - g)a5 L The endogenous wage differential creates an additional channel for welfare changes following a policy shock such as trade liberalization. For example, as the demand for labor in the union sector declines, the wage differential declines, reducing the labor market distortion. This contributes to the welfare gains from the policy shock. Active Unions McDonald and Solow (1981) demonstrate that the wage and employment level chosen with a passive union is not efficient. Consider a union with the utility 1. de Melo and Tarr (1992) use passive rather than active behavior to describe unions in the U.S. auto and steel industries because they find little evidence that U.S. producers in these sectors are off the derived demand for labor. Devarajan, Chanem, and Thierfelder 151 function described in equation 9; a passive union will maximize its utility at point A in figure 1 2 However, the union can be made better off, and the producer no worse off, with a different combination of wages and employment. For example, consider the isoprofit curve-the locus of wage-employment combinations that yields the same profit to the firm-through A. There are points on this isoprofit curve that are preferred by the union. In equilibrium, the union's indifference curve will be tangent to the producer's isoprofit curve (point B). Note that the equilibrium wage and employment are not on the derived demand curve for labor. The set of equilibrium points-where the producer's isoprofit curves and the union's indifference curves are mutually tangent-defines the contract curve. We main- tain sector-specific capital so profit is the return to capital. To derive the equilib- rium wage markup (OM), first consider the producer's profit function (where t equals profit) treating manufacturing as the unionized sector: (14) it = PMf(L, K) - OMWL. The slope of the profit function is: (15) dW/dL = (PMfL -HMW) L where W = oMw. From the union's utility function in equation 9, the slope of the indifference curve is: (16) dW/dL - -1) (Mw- w) Ii (L- LA) Equating equations 15 and 16, we define the contract curve as the locus of points (OMw,L) satisfying: (17) (PMfL - Mw) v - 1) (OMw-W) L It (L -LA) The union bargains with the producer to hire more labor at a given wage than the firm would otherwise have at that wage. This excess labor can be described as featherbedding. Note that there is "excess" labor relative to the case with no unions at the same wage. If the reference point is the passive union, the producer may well be gaining with an active union. In terms of figure 1, excess labor is the horizontal distance between point B on the contract curve and point C on the derived demand for labor, at the agreed upon wage. As shown in figure 1, the contract curve has a positive slope. Under certain conditions, it can have a nega- tive slope. However, because the contract curve is to the right of the labor de- mand curve, featherbedding is always positive. 2. Our utility function differs slightly from that in McDonald and Solow (1981). We include a minimum acceptable employment level as well as the minimum acceptable wage level that they include. 152 THE WORLD BANK ECONOMIC REVIEW, VOL. I11 NO. I Figure 1. Active Union Behavior Wage, (,'J LD Contract cu rve isoprofit CLrves w _________ ____________________ _ tLabor, L LA N The point on the contract curve at which the firm and union agree depends on the relative bargaining strengths of the two. McDonald and Solow (1981) de- scribe two extreme cases: a dominant union and a dominant firm. In the case of the dominant union, equilibrium is reached at the point on the contract curve associated with zero profit for the producer. When the firm is dominant and the union has no bargaining strength, the firm faces w, the union's minimum ac- ceptable wage, and maximizes profit at 1. This is the level of employment and wage at which the labor market is competitive, which we take to be the point at which the union has no bargaining power. To summarize, we describe the union's bargaining strength as 0, the ratio of the wage it received from the bargaining process to the wage it would get (Wmax) if all revenues accrued to labor: (18) 3= Ow Wmax where (19) Wma - MM where PMXM equals revenue from sales; asO approaches 1, the union dominates. Because both the contract curve and the derived demand for labor shift fol- lowing a policy shock, we cannot determine, a priori, the change in featherbedders following a policy shock that changes output in the unionized sector. In the case Devarajan, Ghanem, and Thierfelder 153 Table 1. The Structure of the Stylized Model Value Item Agriculture Manufacturing Tariff 0.00 0.30 Output share 0.50 0.50 Share of sectoral value added paid to capital 0.75 0.41 Share of sectoral value added paid to labor 0.25 0.59 Share of capital employed by sector 0.40 0.60 Share of labor employed by sector 0.65 0.35 Wage differential 0.40 1.75 with active unions, the producer shares the return to capital (profit) with union labor.3 Table 1 provides a description of the structure of the model. Simulation Results We simulate trade liberalization (removing the 30 percent tariff protecting the unionized sector) in the stylized model with various assumptions about union behavior. In the case with no unions, imports of the formerly protected good increase 31.3 percent and domestic production declines 1.1 percent (see table 2). Exports of both goods increase to maintain the exogenous trade balance. Despite the elimination of the trade distortion, there are no welfare gains. We define welfare as equivalent variation, the amount the consumer would be will- ing to pay to consume at the original prices before the policy shock. With this specification, a negative number indicates a welfare gain. However, for report- ing purposes, we change the sign so that a positive number indicates a welfare gain. As the economy adjusts to the policy shock, labor moves into low- productivity agriculture and out of high-productivity manufacturing, counter- acting the welfare gain created by removing the trade distortion. By coincidence, in table 2 the labor market distortion effects exactly offset the welfare gain from trade liberalization. With passive unions, output and trade change in the same direction as in the model with no unions (table 2). (We assume that the slope of the labor demand curve falls in absolute value more slowly than that of the union's indifference curve, so that second-order conditions are satisfied.) Trade liberalization shifts the derived demand for labor in the union sector downward, reducing the labor market distortion and generating welfare gains (see equivalent variation in table 2). The decline in manufacturing output is smaller, compared with the model with exogenous wage differentials, because labor costs decline as the union pre- mium declines (see wage differentials). Likewise, the decline in the demand for labor in manufacturing is smaller than it is in the model with exogenous wage differentials. 3. The calibration programs for the two different union models are available from the authors on request. Table 2. The Impact of Removing the 30 Percent Tariff in the Manufacturing Sector under Alternative Specifications of the Labor Market Exogenous wage Passive union Active union differentials (no unions) = 0.3 u = 0.9 Active union and bargaining Item Agriculture Manufacturing Agriculture Manufacturing Agriculture Manufacturing Agriculture Manufacturing Agriculture Manufacturing Percentage change from base level Employment 1.01 -1.88 0.82 -1.53 0.07 -0.13 0.49 -0.51 -0.29 9.44 Real market-clearing wage 1.10 1.10 1.39 1.39 2.56 2.56 1.98 1.98 4.64 4.64 Featherbedders 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.82 0.00 -18.16 Wage differential 0.00 0.00 0.00 -0.58 0.00 -2.87 0.00 -1.99 0.00 -11.58 Real sectoral wage 1.10 1.10 1.39 0.80 2.56 -0.38 1.98 -0.05 4.64 -7.47 Real sectoral return to capital 2.63 -1.74 2.64 -1.50 2.67 -0.58 2.73 -0.81 4.18 5.93 Output 0.25 -1.11 0.20 -0.90 0.02 -0.08 0.12 -0.30 -0.07 5.39 Exports 15.90 19.71 15.67 20.15 14.75 21.90 15.11 21.40 11.08 32.94 Imports -16.39 31.31 -16.27 31.32 -15.77 31.35 -15.90 31.38 -12.39 32.55 Equivalent variation, 0.01 0.17 0.81 0.70 6.19 a. Measured in units of the numeraire good. Equivalent variation is a weighted average over households, and a positive value indicates a welfare gain. Source: Authors' calculations. Devarajan, Ghanem, and Thierfelder 155 Figures 2 and 3 show the difference between the no-union and union cases. In figure 2, M's labor demand curve slopes from left to right, while A's labor de- mand curve slopes from right to left. The initial equilibrium is portrayed as (WM, WA). With no unions, the difference between WM and WA remains constant, so that for a given downward shift in M's labor demand (because of trade liberal- ization), the new wages are wM and wA. With a passive union, the initial equilibrium (also WM, WA) is determined by the tangency between the union's indifference curve and Lm3 (figure 3). When the labor demand curve shifts downward, the new point of tangency with the union's indifference curve is at a lower manufacturing wage w' . The wage differential has narrowed, implying that manufacturing does not contract by as much as in the no-union case. Interestingly, when there is a high weight on wages in the union's utility func- tion, there is a bigger decline in the wage differential than when there is a low weight on wages-a decline of 2.9 percent rather than 0.6 percent. The smaller wage differential translates into a larger welfare gain when the weight on wages is high; the change in equivalent variation is 0.8 rather than 0.2 in the low- weight case. As a result of the decline in the wage markup, the payment to labor in manufacturing declines 0.4 percent when the union cares more about wages than employment. It increases 0.8 percent when the union cares more about employment than wages. Likewise, fewer union jobs are sacrificed in the high- Figure 2. The Impact of Trade Liberalization on Employment and Wages in the Case witb No Unions Wage WM - WA - - - A - \ At w'4~~~~~~~~~~~D Labor Ls > 156 THE WORLD BANK ECONOMIC REVIEW, VOL. II, NO I Figure 3. The Impact of Trade Liberalization on Employment and Wages in the Case with Unions Wage WAI~~~~~~~~~~~~~~~~~I WA, tVs1 - - - - -- \ IV - - - - - - - \ L wA LD~~~~e Labor weight version of the passive union-they decline 0.1 percent rather than 1.5 percent. This may seem like a counterintuitive result: a union that cares more about wages, as represented by a higher weight on wages in its utility function, sacri- fices the wage premium rather than employment when the sector contracts. To resolve this apparent paradox, recall that the union has both a minimum accept- able wage and employment level in its utility function. We take the minimum acceptable wage to be the economywide average wage. The unobserved param- eter is the union's minimum acceptable employment level, LP. We therefore calibrate this parameter from the observed wage and employment levels and the other parameters of the union's utility function (including the weight on wages, j). Thus, a higher ,u implies a higher LP for consistency with the observed wage and employment levels. It follows that in response to a negative shock, with a higher weight on wages, employment has less "room" to fall, so that the adjust- ment occurs in wages. This paradox appears in de Melo and Tarr (1992) and Thierfelder and Shiells (1996). With active unions and the removal of protection, the wage differential de- clines 2.0 percent. The number of productive workers in the economy increases as the number of featherbedders declines 0.8 percent. The changes in output, trade, and factor allocation resemble those in the model with no unions. Be- Devarajan, Ghanem, and Thierfelder 157 cause of the reduction in labor costs, manufacturing output declines 0.3 percent, which is less than the decline in the case with exogenous wage differentials. Similarly, the demand for labor declines 0.5 percent, less than in the case with exogenous wage differentials, in which the demand for labor in manufacturing declines 1.9 percent. Suppose that, anticipating the trade liberalization, the union makes some con- cessions in terms of its bargaining strength. This scenario produces more dra- matic results. The supply of productive labor increases as the number of featherbedders declines 18.2 percent. The demand for labor in manufacturing increases 9.4 percent. The market-clearing wage increases 4.6 percent, indicat- ing that the changes in labor demand dominate the increase in labor supply. However, given the 12.0 percent decline in the union premium, the wage for workers in the manufacturing sector declines 7.5 percent. The labor market changes dominate tariff elimination: output in the formerly protected sector actually increases 5.4 percent, while output of the other sector declines slightly (0.07 percent). This is the only scenario in which the return to capital in manu- facturing increases, reflecting the increase in manufacturing output. Two changes in the labor market affect output in the formerly protected sec- tor: (1) the wage differential declines in the union sector and (2) the supply of labor increases. To decompose the effects, we consider trade liberalization in the model with an exogenous wage differential. We reduce the wage differential to the level observed in the case with trade liberalization and union concessions. We find that output of manufacturing increases because of the decline in labor cost. Manufacturing output increases more (and agricultural output declines less) when the labor supply adjusts because of the decline in the number of featherbedders. Simulations with the stylized model lead to four general conclusions: 1. Loss of protection in the unionized sector results in additional welfare gains from trade liberalization. 2. With passive unions, the reduction in the union wage premium results in welfare gains, whose magnitude depends on union preferences. 3. With active unions, both the decline in the wage differential and the fall in the number of featherbedders contribute to the welfare gain. 4. If unions make bargaining concessions, the labor market changes can domi- nate the effect of tariff elimination on output. II. BANGLADESH The structure of the labor market and the role of unions in Bangladesh are representative of those in other South Asian countries. Bangladesh has three types of labor markets: formal, rural informal, and urban informal. The formal labor market is characterized by a contractual employment relationship and governed by labor laws and regulations that aim to protect workers, such as minimum wages and allowances and limitations on the employer's ability to fire 158 THE WORLD BANK ECONOMIIC REVIEW, VOL. 11, NO. I employees. The rural and urban informal labor markets are not covered by pro- tective labor regulations. The informal sector dominates the labor market. In 1991, 47.2 percent of the labor force were classified as unpaid family workers, 15.4 percent were classified as self-employed, 13.9 percent were classified as casual workers (day laborers), and only 11.7 percent had regular full-time wage employment.4 Despite the small size of the formal sector, Bangladesh has a large number of labor unions. In 1992, there were 4,065 registered unions, with a total membership of 1,648,783, which represents some 3 percent of the labor force. The unions are organized into roughly 700 union federations. Almost all federations have links with political parties, with the three largest federations acting as the labor front of the three major parties. (For a more detailed description of the labor movement in Bangladesh, see Rahman 1994.) Unions and union federations are very active in Bangladesh's political and economic life. They were key partners in the coalition that ended the rule of General Ershad and restored democracy to Bangladesh in 1991. On the economic front, they entered into agreements with government in 1984, 1991, and 1992 to raise legislated labor benefits and protections. For most of Bangladesh's existence, the government has provided high pro- tection to domestic industries. In the late 1980s the average manufacturer sell- ing its products domestically enjoyed more than an 80 percent effective protec- tion rate from tariffs alone. Some industries-like textiles, soap and detergents, and iron and steel-were even more highly protected by both tariffs and quanti- tative restrictions on imports. The situation started to change in the 1990s. In 1992 import discriminating taxes other than customs duties were replaced by the largely trade-neutral value added tax and supplemental duties; with some exceptions, the maximum tariff rate was reduced steadily from 400 percent to around 50 percent; and many quantitative restrictions were removed. More trade liberalization is being discussed in Bangladesh, including further reduction and compression of customs duties, removal of remaining trade-related quantitative restrictions, and reforms to facilitate the process of import clearance, As can be expected, there is a lot of political resistance to trade liberalization in Bangladesh, much of it from labor unions. The existence of strong labor unions in Bangladesh clearly increases the political cost of trade liberalization. How does it affect the economic benefits to be gained from liberalization? To answer this question, we turn to a more detailed general equilibrium model of Bangladesh, which has many of the core features of the stylized model in section I. Model Description Naturally, some of the features of the Bangladesh model are more complex, making the effects of a union less clear than in the case of the stylized model. Of the thirty-five sectors in the model, nine are unionized. The unionized sectors 4. Of the labor force, 11.4 percent are classified as "employers" and 0.4 percent are "not reported." The data references in this section come from Bangladesh Bureau of Statistics (1995). Devaralan, Ghanem, and Thierfelder 159 are cotton yarn, paper, leather, fertilizers, cement, basic metals, metal products, petroleum, and electricity and gas. (See Devarajan, Ghanem, and Thierfelder 1995 and Dahl and Mitra 1991 for a more detailed description of the Bangladesh model.) The model contains a Cobb-Douglas production function over an ag- gregate of intermediate goods and a CES aggregate of four inputs: rural labor, urban informal labor, urban formal labor, and capital. Capital is sector-specific, representing the producer's profit that is shared with the active unions. The assumption of sector-specific capital, which reflects the segmented nature of capital markets in Bangladesh, is not crucial (see Devarajan and Offerdal 1989). The original data for Bangladesh do not indicate capital use by sector. We construct the input capital by allocating some of the value added by the other labor inputs to it. Each of the three households in the base data receives income from one labor category. To maintain the house- holds' structure, we restore the capital income (net of payment to featherbedders where applicable) to the appropriate household, based on the share of value added that each labor category contributed to the construction of capital input. When there are no active unions, the income to each household is identical to the case with no capital inputs. Because there are intermediate goods, there is the potential for offsetting changes in demand following a policy shock such as trade liberalization. Pro- duction may not always decline in sectors where protection is eliminated. It can sometimes rise in response to increased demand from sectors that use the for- merly protected good as an intermediate input. The structure of protection also affects the distribution of output among the unionized sectors. In Bangladesh, not all unionized sectors receive protection (see table 3). For example, fertilizer and electricity and gas have zero tariffs. Likewise, some nonunion sectors receive protection. Unlike the stylized model, tariff elimination does not exclusively target production in the unionized sector. In contrast to the stylized model, there are additional (nonunionized) labor groups in Bangladesh. There is potential for factor substitution in response to Table 3. Characteristics of Unionized Sectors in Bangladesh Share of urban Share of sectoral value Share of formal employment added paid to urban Sector Tariff output by sector formal workers Cotton yarn 2.30 0.01 0.01 0.22 Paper 1.31 0.01 0.01 0.15 Leather 1.09 0.01 0.01 0.25 Fertilizers 0.00 0.01 0.03 0.36 Cement 0.41 0.00 0.01 0.33 Basic metals 0.56 0.01 0.04 0.38 Metal products 0.43 0.04 0.25 0.31 Petroleum 0.41 0.02 0.08 0.23 Electricity and gas 0.00 0.01 0.04 0.23 Source: Authors' calculations. 160 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Table 4. Macroeconomic Effects of Trade Liberalization in Bangladesh under Alternative Specifications of the Labor Market Active union Exogenous wage No Low High Item differentials (no unions) concessions concessions concessions Real market-clearing factor return (percentage change from base level) Rural 6.09 6.09 6.31 6.51 Urban informal 2.07 2.09 2.25 2.52 Urban formal -0.56 -0.16 1.26 2.26 Capital 10.37 10.49 10.10 9.77 Real average factor return (percentage change from base level) Rural 6.08 6.07 6.30 6.50 Urban informal 2.75 2.75 2.94 3.22 Urban formal -1.29 -1.42 -3.81 -5.88 Capital 5.47 5.46 5.74 6.00 Featherbedders 0.00 -1.52 -33.91 -63.75 Total exports 78.24 77.57 76.91 76.76 Total imports 30.91 30.73 30.56 30.54 Unemployment (share of urban formal labor supply) 0.00 0.00 0.00 0.00 Equivalent variationa 20.08 20.07 21.34 22.45 a. Measured in units of the numeraire good. Equivalent variation is a weighted average over households, and a positive number indicates a welfare gain. Source: Authors' calculations. changes in the cost of unionized labor. This implies stronger spillover effects on the returns to other labor groups. Simulation Results Tariffs in Bangladesh mainly protect the urban formal workers. In a model with no unions (the observed factor payment differentials are treated as exog- enous), trade liberalization causes the real average return to urban formal work- ers to decline 1.3 percent, the real average return to rural labor to increase 6.1 percent, and the real average return to urban informal workers to increase 2.8 percent (see table 4).5 The contracting sectors-such as cement, basic metals, and metal products, which had tariffs of 41, 56, and 43 percent, respectively- use urban formal workers relatively intensively. If we assume that active unions in Bangladesh explain the observed wage differentials, the income distribution following trade liberalization changes slightly. The real return to urban formal workers decreases 1.4 percent rather 5. These numbers differ in magnitude but not direction compared with our earlier analysis of Bangladesh (Devarajan, Ghanem, and Thierfelder 1995). This model differs from the earlier model in that we introduce sector-specific capital and factor payment differentials for urban formal workers. Devarajan, Ghanem, and Thierfelder 161 than 1.3 percent. The change is caused by the increased supply of productive urban formal workers, as the number of featherbedders declines 1.5 percent. For the same reason, the factor allocation and production changes are not as extreme as when there are no unions. For example, the wage differentials for cement, basic metals, and metal products decline 4.7, 5.5, and 2.9 percent, re- spectively, in the model with active unions (see table 5). The less extreme changes dampen the decline in demand for urban formal labor in those sectors. Like- wise, output for those sectors declines less than when there are no unions be- cause of the decline in the cost of urban formal workers. Table 5. Effects of Trade Liberalization in Bangladesh under Alternative Specifications of the Labor Market for Unionized Sectors Active union Exogenous wage No Low High Sector differentials (no unions) concessions concessions concessions Wage differential for urban formal workers (percentage change from base level) Cotton yarn 0.00 7.23 -1.42 -9.48 Paper 0.00 -2.43 -8.26 -13.59 Leather 0.00 1S.24 5.00 -4.71 Fertilizers 0.00 2.73 -8.68 -19.10 Cement 0.00 -4.67 -12.75 -20.37 Basic metals 0.00 -5.49 -16.40 -6.51 Metal products 0.00 -2.87 -10.66 -18.00 Petroleum 0.00 3.05 -4.65 -11.95 Electricity and gas 0.00 -3.99 -10.86 -16.97 Demand for urban formal labor (percentage change from base level) Cotton yarn 12.28 7.18 11.32 16.11 Paper -5.11 -3.82 -1.04 1.96 Leather 30.82 15.53 22.27 30.43 Fertilizers 5.55 3.44 10.86 19.44 Cement -11.31 -8.73 -4.32 0.63 Basic metals -12.19 -9.70 -2.88 -10.83 Metal products -7.64 -6.09 -1.90 2.99 Petroleum 4.43 2.36 6.50 11.21 Electricity and gas -7.48 -5.71 -3.69 -1.57 Output (percentage change from base level) Cotton yarn 6.56 6.02 6.03 6.15 Paper -0.76 -0.52 -0.03 0.50 Leather 32.75 28.47 29.21 30.36 Fertilizers 8.16 7.90 9.60 11.48 Cement -7.71 -6.79 -5.27 -3.72 Basic metals -6.70 -6.25 -5.30 -6.11 Metal products -5.40 -4.69 -2.90 -1.02 Petroleum 7.77 7.12 -8.59 10.17 Electricity and gas -3.75 -3.53 -3.60 -3.60 Source: Authors' calculations. 162 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I When the active union makes wage concessions in conjunction with trade liberalization, jobs are saved in the unionized sector, but workers receive a lower wage. Consider the scenario with high concessions. The number of featherbedders declines dramatically (63.8 percent, see table 4), increasing the labor supply. The real return to urban formal workers declines 5.9 percent because of both loss of the union premium and increase in the supply of productive workers. The spillover effects favor the other labor groups. As output expands in response to the increase in the supply of urban formal workers (because of the reduction in featherbedders), the demand for other inputs increases as well. The real return to rural workers increases 6.5 percent, and the real return to urban informal workers increases 3.2 percent. Both are slightly better off than when there is trade liberalization with no unions. To report the sensitivity of our results to the bargaining concessions, we con- sider the case in which the union makes low, rather than high, concessions. In this case, the pattern of wage change is the same, but the magnitude is smaller. Likewise, the number of featherbedders declines, but by a smaller amount (33.9 percent rather than 63.8 percent). The sectoral effects of trade liberalization for the case with active unions and high concessions are more dramatic. The wage differential for cement and metal products declines 20.4 and 18.0 percent, respectively (table 5). The change in factor cost reverses the change in factor demand-there is an increase in the demand for urban formal workers in these sectors, compared with the de- crease in demand in the model with no unions. However, the change in factor cost, coupled with the dramatic increase in productive labor as the number of featherbedders declines 63.8 percent (table 4), is not enough to reverse the change in output observed with trade liberalization. Output in cement, for example, declines 3.7 percent rather than 7.7 percent when there are no unions; likewise, output in metal products declines 1.0 percent rather than 5.4 percent. Wage changes by sector reflect the loss of the union premium. For example, urban formal workers employed in the nonunion sectors experience a real wage increase of 2.3 percent (table 4). This real wage increase accrues to urban for- mal workers who do not change jobs. Those who change sectors may receive a different sectoral wage differential, changing the real wage received. However, those in the union sector experience a wage change that accounts for changes in the union premium, which declines in most unionized sectors. This decline in the wage premium means that the real average return to unionized labor declines 5.9 percent. Given the structure -of protection and intermediate demands, certain union- ized sectors expand following trade liberalization (table 5). For example, leather production increases 32.8 percent following trade liberalization in a model with no unions. This reflects an increase in the demand for leather as an intermediate good in production. In the case with active unions and no bargaining conces- sions, the wage differential increases 15.2 percent because of the increase in Devarajan, Ghanem, and Thierfelder 163 demand for leather as an intermediate good. This cost increase does dampen the increase in output, which rises only 28.5 percent in this case. When high bargaining concessions accompany trade liberalization, the union premium declines for all sectors (table 5). In general, output increases (or does not decline as much) in this scenario, compared to the model with trade liberal- ization and no unions. What does all of this imply for the benefits to be gained from trade liberaliza- tion in Bangladesh? The existence of active unions, which refuse to make bar- gaining concessions, does not seem to have much of an effect on welfare gains. Equivalent variation is 20.1 for the cases of trade liberalization both with exog- enous price differentials and with active unions and no concessions (table 4). This result, different from that derived from the stylized model, is explained by the fact that some unionized sectors, such as leather, expand following trade liberalization. The wage differential in these sectors increases as well, offsetting the welfare gains that accrue when other unionized sectors contract. In the more likely case that liberalization pushes the unions into making bar- gaining concessions, the gains from trade are higher. In this case the bigger fall in the union premium and the huge decline in the number of featherbedders more than offset any losses due to the unionization of export sectors. III. INDONESIA In contrast to Bangladesh, Indonesia opted early on for less interventionist policies and an open trade regime, with excellent results. Real per capita GDP growth over the last fifteen years (1979-94) has averaged more than 4 percent a year. Exports of manufactured goods have been rising at more than 30 percent a year. Workers have reaped tremendous benefits from this growth. Formal sector jobs, which offer higher wages and greater job security, are being created at faster rates than ever before, much faster than in most other countries. Wage employment has been increasing at more than 4.5 percent a year, while real earnings have been rising at 9 to 10 percent a year. Yet, in the 1990s labor unrest has increased. During 1989-92, the number of strikes rose thirteenfold from 19 to 251 and the number of working hours lost because of strikes rose from about 29,000 to more than 1 million. The fact that many of these strikes occurred in export-oriented enterprises close to Jakarta has drawn international and domestic public attention. Many observers, for example Agrawal (1995), attribute the spread of labor unrest to the failure of the officially sanctioned trade union movement to represent workers' rights and to negotiate better working conditions. Indonesia is now under internal and external pressure to liberalize its labor regulations and ease restrictions on unions. Although Indonesian law clearly guarantees the right to form unions, difficulties arise in practice. There is only one legally recognized union in the private sector (the spsi), which is closely linked to the government. Government decrees essentially prevent the forma- 164 THE WORLD BANK ECONOMIC REVIEW, VOL II, NO. I tion of other independent unions that can represent workers in collective bar- gaining. Although workers can form worker organizations or company-level unions, only a registered union can bargain collectively with employers. The conditions for registration are very restrictive. As Freeman (1993) states in the context of other East Asian countries, "The seemingly innocuous procedure for registering trade unions is a powerful tool for repression." Instead of liberalizing labor regulations, the government has tried to achieve industrial peace by increasing mandated benefits, especially minimum wages. What are the efficiency effects of mandating minimum wages compared with allowing unions greater independence? Model Description To answer this question, we consider the effect of a policy shock that reduces the demand for urban formal workers in Indonesia under three assumptions about the structure of the labor market: completely free labor markets where wage differentials are exogenous, labor markets where government sets a bind- ing minimum wage, and labor markets where unions are allowed greater inde- pendence. We use Lewis's (1991) CGE model of Indonesia and modify the labor market structure. Because tariffs are quite low in Indonesia and do not protect sectors in which unions are likely to form, we choose a policy shock that is different from the one used in the Bangladesh model: a 20 percent reduction in government spending. The Indonesia model has thirty sectors. We treat seventeen of the sectors that currently use formal labor relatively intensively as potentially unionized sectors to illustrate the rigidities that union behavior may introduce when there are economic reforms. These sectors account for 54 percent of the urban formal labor force. There are six factors of production: agricultural labor, rural labor, urban informal labor, urban formal labor, capital, and land. As in the Bangladesh model, all sectors use intermediate goods. As we did in the case of Bangladesh, we compare the impact of an exogenous shock in a model with and without unions. In addition, we carry out simulations on a model with a minimum wage in order to see the impact of present policies. In the case with a minimum wage paid to urban formal workers, urban formal workers who become unemployed move into the urban informal sector, putting downward pressure on wages for workers in that sector. Unions, in contrast, provide high wages only for a portion of urban formal workers; urban formal workers who lose their job in a unionized sector remain in the urban formal labor market in another sector. Spillover effects to the other types of labor occur only through substitution in production. Simulation Results We consider a 20 percent decline in government spending. The purpose of the experiment is to change the composition of spending and therefore reduce the demand for commodities produced by unionized sectors. The government spends Devarajan, Ghanem, and Thierfelder 165 about 85 percent of its income on services, which constitutes a unionized sector. With a 20 percent reduction in government spending, government savings and therefore total savings increase. Because savings equal investment, the increase in savings means an increase in investment spending. Some investment spending also increases the demand for output of the unionized sectors, but the net effect is a decline in demand for urban formal workers. In the case with exogenous wage differentials (no unions), the real average wage for urban formal workers declines 3.1 percent (see table 6) because the government purchases services, a good that uses urban formal labor intensively. .The share of value added accruing to urban formal workers in the service sector is 33 percent (see table 7). The real average wage decline is caused by both a decline in the demand for urban formal labor and the movement of these work- Table 6. Macroeconomic Effects of a 20 Percent Reduction in Government Spending in Indonesia under Alternative Specifications of the Labor Market Exogenous wage differential Minimum Active Passive union Item (no unions) wagea union , = 0.35 , = 0.8 Real market-clearing factor return (percentage change from base level) Agricultural labor -0.67 -0.57 -0.76 -0.72 -0.80 Rural 0.27 0.22 0.30 0.25 0.24 Urban informal 0.40 -3.28 0.45 0.36 0.32 Urban formal -2.84 0.00 -2.40 -2.02 0.07 Capital 0.00 0.00 0200 0.00 0.00 Real average factor return (percentage change from base level) Agricultural labor -O.S9 -0.49 -0.67 -0.63 -0.71 Rural -0.14 -0.12 -0.19 -0.18 -0.20 Urban informal 1.90 -1.42 1.93 1.88 1.90 Urban formal -3.10 0.66 -3.33 -3.16 -3.27 Capital 0.49 0.45 0.48 0.47 0.48 Featherbedders 0.00 0.00 -2.41 0.00 0.00 Total exports 0.07 0.01 0.02 0.01 -0.06 Total imports 0.28 0.31 0.23 0.21 0.14 Unemployment (share of urban formal labor supply) 0.00 0.05 0.00 0.00 0.00 Equivalent variationb 38.1 23.15 38.01 38.92 44.60 a. There is comparable labor coverage-the number of urban formal workers affected by the policy shock-in the minimum wage and the various union scenarios. b. Measured in units of the numeraire good. Equivalent variation is a weighted average over households, and a positive number indicates a welfare gain. Source: Authors' calculations. 166 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. 1 Table 7. The Structure of Indonesia's Economy for Selected Sectors Share of urban Share of sectoral Share of formal employment value added paid to Sector Tariff output by sector urban formal workers Chemicals 0.04 0.02 0.01 0.12 Machines 0.06 0.04 0.01 0.08 Construction 0.00 0.12 0.01 0.05 Trade 0.00 0.09 0.42 0.31 Transportation 0.00 0.05 0.02 0.08 Services 0.00 0.17 0.50 0.33 Source: Authors' calculations. ers out of sectors that pay them a high wage differential. The real return to urban informal workers increases 1.9 percent (table 6). In contrast, in the case with a minimum wage for urban formal workers, producers respond to the decline in demand by shedding labor. These workers move into the urban informal labor market where the real average wage de- clines 1.4 percent (table 6). The real average return to urban formal workers increases slightly, reflecting sectoral changes among them. The presence of an active union forces the adjustment on urban formal rather than informal workers. Furthermore, the adjustment costs have more impact on unionized urban formal workers who lose their wage markup. The nonunionized urban formal workers who do not change sectors, in fact, are slightly better off because the wages paid to urban formal workers in nonunion sectors decline 2.4 percent rather than 2.8 percent in the case of exogenous wage differentials (table 6). The real return to urban formal workers, accounting for the decline in the wage premium and the reallocation of labor to sectors that pay different wage differentials, declines 3.3 percent in the case with active unions compared with 3.1 percent in the case with exogenous wage differentials..The larger decline in the case with active unions is caused by the loss of the union premium. Both demand and supply changes in the labor market affect the wage for urban for- mal workers. The labor supply increases because the number of featherbedders declines 2.4 percent. However, the demand for labor increases as the union wage premium decreases in certain sectors, such as services, which employs a large share of urban formal labor. The labor demand shift dominates, and the market-clearing wage declines less than in the model with exogenous wage differentials. In the case with passive unions in Indonesia, the adjustment is also forced on urban formal workers whose real average return declines 3.2 percent when the union puts a low weight on wages and 3.3 percent when it puts a high weight on wages (table 6). The real average return to urban informal workers increases 1.9 percent for either specification of preferences. In all union sectors there is a positive relationship between the demand for urban formal labor and the wage differential (see table 8). When the union places a low weight on wages, the Devarajan, Ghanem, and Thierfelder 167 results are quite similar to those in the model with exogenous wage differentials. When the union places a high weight on wages, there is a bigger change in the wage premium and a smaller change in employment, compared with the low- weight case. As a result, output generally declines more or increases less than when the union puts a low weight on wages. Construction, a sector that expands as investment demand increases, illus- trates the effect that different union preferences have on wages. When the union cares about wages-the high-weight case-the wage differential increases 11.3 percent as the sector expands (table 8). In contrast, when the union cares about employment-the low-weight case-the wage differential only increases 0.9 percent. Instead, employment increases 12.3 percent, as opposed to 1.1 percent in the high-weight case. How do the different labor market specifications affect the welfare gains caused by the change in policy? The welfare gain is lowest in the case with a binding minimum wage, which has an equivalent variation of 23.2 (table 6). This reflects the fact that the binding minimum wage prevents the down- Table 8. Selected Sectoral Effects of a 20 Percent Reduction in Government Spending in Indonesia under Alternative Specifications of the Labor Market Exogenous wage differential Minimum Active Passive union Sector (no unions) wagea union , = 0.35 ji = 0.8 Wage differential (percentage change from base level) Chemicals 0.00 0.00 1.26 0.16 -0.60 Machines 0.00 0.00 5.33 2.73 6.32 Construction 0.00 0.00 8.07 0.88 11.30 Transportation 0.00 0.00 1.19 0.55 -0.28 Services 0.00 0.00 -1.91 -2.01 -6.55 Demand for urban formal workers (percentage change from base level) Chemicals 2.08 -0.49 0.57 1.19 -0.07 Machines 9.47 6.44 4.35 6.13 1.06 Construction 13.97 10.48 6.39 12.26 1.14 Transportation 2.39 -0.39 0.87 1.11 -0.05 Services -4.36 -6.52 -3.27 -3.49 -1.47 Output (percentage change from base level) Chemicals -0.28 -0.19 -0.39 -0.35 -0.50 Machines 3.54 3.59 3.43 3.42 3.18 Construction 7.27 7.27 7.39 7.22 7.07 Transportation 0.06 0.26 -0.04 0.00 -0.04 Services -4.07 -4.39 -3.87 -3.90 -3.50 a. There is comparable labor coverage-the number of urban formal workers affected by the policy shock-in the minimum wage and the various union scenarios. - Source: Authors' calculations. 168 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I ward adjustment in formal wages needed for an efficient reallocation of la- bor. In contrast, the equivalent variation under the various union specifica- tions, which do allow flexibility in the economy, are all higher, ranging from 38.1 to 44.6. The union effects, compared with the exogenous wage differ- entials, follow the same pattern as in Bangladesh. In the case with active unions, the welfare analysis is the same as the no-union case because some unionized sectors expand, offsetting the gains as other unionized sectors con- tract. The case with passive unions and a high weight on wages has the larg- est welfare gain because the wage differential for services decreases the most. Note that the increase in the wage differential for construction as invest- ment spending increases is dominated by the decline in services as govern- ment spending declines. IV. CONCLUSIONS In this article, we have evaluated the difference that the presence of labor unions has made to the outcome of economic reform in developing countries. Using a stylized, two-sector model, we have shown that the welfare gains from trade liberalization are amplified when there are unions in the protected sector. The reason is that increased import competition drives down the wage premium that unions enjoy, leading to a more efficient allocation of labor in the economy. Although this is true for both passive and active unions, the effect is much greater for active unions. Trade reform reduces the active unions' ability to protect featherbedders, yielding an additional gain as these surplus workers are absorbed into the productive labor force. Finally, if the threat of import competition prompts active unions to shed featherbedders as part of a contract with man- agement, the welfare gains are astonishingly high. We have applied the insights from this stylized analysis to two contemporary policy issues. In Bangladesh, where unions are strong, we have shown that fur- ther liberalization of international trade would produce a "double dividend"- greater efficiency in both labor and product markets. In Indonesia, where there are restrictions on union operations but trade has been substantially liberalized, we have shown that allowing unions greater freedom is superior to the current minimum-wage policy. This is because the wage markup of unions is still flex- ible in the face of adverse demand shocks, whereas a minimum wage introduces an additional rigidity in the economy. Unions are also preferable on equity grounds. In the case with a minimum wage, urban informal labor bears the cost of the adjustment. In the case with a union, however, urban formal labor bears the cost of adjustment. Our work raises a host of political-economy issues that warrant further re- search. In a country like Bangladesh, unions will probably try to block reforms, such as trade liberalization, which make product markets more competitive. Governments in Indonesia or other East Asian countries may be reluctant to forgo regulation in favor of allowing more freedom to trade unions, which are Devarajan, Ghanem, and Thierfelder 169 often opposed to the regime in power. How to convince vested interests to sup- port globally beneficial policies is perhaps the central question facing develop- ing countries today. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Agrawal, Nisha. 1995. "Indonesia: Labor Market Policies and International Competi- tiveness." Policy Research Working Paper 1515. World Bank, Washington, D.C. Processed. Bangladesh Bureau of Statistics. 1995. Report on Labor Force Survey. Dhaka, Bangladesh. Blanchflower, David G., and Richard B. Freeman. 1990. "Going Different Ways: Union- ism in the U.S. and Other Advanced OECD Countries." NBER Working Paper 3342. National Bureau of Economic Research, Cambridge, Mass. Processed. Dahl, Hendrik, and Pradeep Mitra. 1991. "Applying Tax Policy Models in Country Economic Work: Bangladesh, China, and India." The World Bank Economic Re- view 5(3):5S3-72. de Melo, Jaime, and David Tarr. 1992. A General Equilibrium Analysis of U.S. Foreign Trade Policy. Cambridge, Mass.: MIT Press. Devarajan, Shantayanan, Hafez Ghanem, and Karen Thierfelder. 1995. "Labor Mar- ket Regulation, Trade Liberalization, and the Distribution of Income in Bangladesh." Policy Research Department, World Bank, Washington, D.C. Processed. Devarajan, Shantayanan, and Delfin S. Go. Forthcoming. "The Simplest Dynamic Gen- eral Equilibrium Model of an Open Economy." Journal of Policy Modeling. Devarajan, Shantayanan, Jeffrey D. Lewis, and Sherman Robinson. Forthcoming. The General Equilibrium Approach to Adjustment Policy. Cambridge, U.K.: Cambridge University Press. Devarajan, Shantayanan, and Erik Offerdal. 1989. "Capital Markets and General Equi- librium Models: Comparative Statics without Apology." Journal of Policy Modeling 11(2):191-212. Fields, Gary. 1994. "Changing Labor Market Conditions and Economic Development in Hong Kong, Korea, Singapore, and Taiwan." Department of Economics, Cornell University, Ithaca, N.Y. Processed. Freeman, Richard B. 1980. "Unionism and the Dispersion of Wages." Industrial and Labor Relations Review 34(1):3-23. - 1993. "Does the Suppression of Labor Contribute to Economic Success? Labor Relations and Markets in East Asia." Department of Economics, Harvard Univer- sity, Cambridge, Mass., and London School of Economics, London. Processed. Freeman, Richard B., and James L. Medoff. 1984. What Do Unions Do? New York: Basic Books. Johnson, George E. 1990. "Work Rules, Featherbedding, and Pareto Optimal Union- Management Bargaining." Journal of Labor Economics 8(1, part 2):S237-59. Lee, Joung-Woo, and Sang-Sup Nam. 1994. "The Effect of Labor Unions on the Wage Dispersion in Korea." Korean Economic Journal (in Korean) 41(3):251-77. 170 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Lewis, Jeffrey D. 1991. "A Computable General Equilibrium (CGE) Model of Indo- nesia." Development Discussion Paper 378. Institute for International Development, Harvard University, Cambridge, Mass. Processed. McDonald, Ian M., and Robert M. Solow. 1981. "Wage Bargaining and Employment." American Economic Review 71(5):896-908. Moll, Peter G. 1993. "Black South African Unions: Relative Wage Effects in Interna- tional Perspective." Industrial and Labor Relations Review 46(2):245-61. Panagides, Alexis, and Harry Anthony Patrinos. 1994. "Union-Nonunion Wage Differ- entials in the Developing World: A Case Study of Mexico." Policy Research Work- ing Paper 1269. Education and Social Policy Department, World Bank, Washington, D.C. Processed. Park, Young-Bum. 1991. "Union/Minimum Wage Differentials in the Korean Manu- facturing Sector." International Economic Journal 5(4):79-91. Rahman, Masihur. 1994. "Structural Adjustment, Support, and Workers: Public Issues and Choice for Bangladesh." World Bank, Washington, D.C. Processed. Standing, Guy. 1992. "Do Unions Impede or Accelerate Structural Adjustment? Indus- trial versus Company Unions in an Industrialising Labour Market." Cambridge Jour- nal of Economics 16(September):327-54. Teal, Francis. 1994. "The Size and Sources of Economic Rents in a Developing Country Manufacturing Labor Market." Center for the Study of African Economics, St. John's College, Oxford University, Oxford. Processed. Thierfelder, Karen, and Clinton R. Shiells. 1996. "Trade and Labor Market Behavior in General Equilibrium Models." In Joseph F. Francois and Kenneth A. Reinert, eds., Applied Methods for Trade Policy Analysis. Cambridge, U.K.: Cambridge Uni- versity Press. World Bank. 1995. "Bangladesh: Labor Market Policies for Higher Employment." Re- port 13799-BD. South Asia Country Department 1, Washington, D.C. Processed. Does the Type of Political Regime Matter for Trade and Labor Market Policies? Arup Banerji and Hafez Ghanem This article uses cross-country data to examine the link between a country's type of political regime and its degree of openness and labor market distortion. The analysis indicates that more authoritarian regimes are associated with higher trade protec- tion and greater labor market distortions. This supports the view that political authoritarianism may be counterproductive to development in important respects. The link between the, political system of a country and its economic growth has attracted a great deal of attention in recent years (for an extensive survey of the literature see Alesina and Perotti 1994). Empirical studies have yielded mixed results, and a survey of the literature by Przeworski and Limongi (1993) con- cluded that they did not know whether democracy fosters or hinders economic growth. Authors from Friedman (1962) to Scully (1988) have argued that politi- cally open societies grow at faster rates than societies where freedom is restricted. Others such as Kormendi and Meguire (198S) have found a negative relation- ship between civil liberty and growth, and Barro (1994) has argued that more political freedom may encourage a greater role for interest groups in the legisla- tive process, thereby retarding growth. Landell-Mills and Serageldin (1992) sum- marize this literature by concluding that benevolent dictators are rare and that democracies often resort to populist policies that are inimical to growth. The focus of this article is more narrow. We consider labor policies and, to the extent that they influence labor policies, policies affecting the openness of the economy. What, if any, is the link between such policies and the nature of the political regime? Many governments in developing countries have adopted labor policies-including high minimum wages, public sector overemployment, and job security guarantees-that benefit a small group of "insiders." But these same policies limit the opportunities of "outsiders," thereby aggravating income inequality, generating efficiency losses, and possibly discouraging investment and growth. Other countries in the developing world have resisted the tempta- Arup Banerji and Hafez Ghanem are with the Europe and Central Asia IV Department of the World Bank. This article was originally prepared as a background paper for World Development Report 1995: Workers in an Integrating World. The authors thank Claudio Montenegro for research assistance and help. They also thank Joseph Ryan of Freedom House for providing data on political and civil liberties and Michael Walton, Alejandra Cox Edwards, Ishac Diwan, David Lindauer, Ana Revenga, Michal Rutkowski, John Sheahan, and three anonymous referees for their comments on an earlier draft. (D 1997 The International Bank for Reconstruction and Development / THE WORLD BANK 171 172 TFIE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I tion to distort the labor market. Because labor policies are usually chosen for political rather than economic reasons, one would expect the nature of a country's political system and its degree of civil liberties to influence its choice of labor policies. Does the type of regime really matter? We use data for ninety developing countries to shed light on this question. The article is divided into five sections. Section I discusses possible links between political structure and labor policies. Section II describes a simple political economy model similar to that of Grossman and Helpman (1994), but with a more elaborate labor market. The underlying analytical model is presented in the appendix. Section III discusses data chosen for variables in our estimating equations. Section IV presents results of empiri- cal tests on cross-sectional data and section V summarizes our results. I. POLITICAL SYSTEMS AND LABOR POLICIES The existing literature linking economic growth to political regimes mostly avoids an exact definition of "democratic" and "authoritarian." This is under- standable, considering the range of political systems existing in the world today. Although we realize that authoritarianism and democracy exist in a continuum, our understanding of the terms is closest to that of Przeworski (1991). He notes that, "Authoritarian regimes abhor independent organizations; they either in- corporate them under centralized control or repress them by force" (p. 55, em- phasis added). His definition of democracy is pithier: "Democracy is a system in which parties lose elections" (p. 10), but he goes on to say that the key feature of democracies is that they are populated by collective organizations (p. 12). Less authoritarian (and thus more democratic) societies allow independent organiza- tions to participate more freely in the political process. At one extreme, only relatively harmless organizations may be allowed to associate freely-Przeworski gives the examples of stamp collectors societies-but only the most democratic regimes allow a free press, free trade unions, and truly contesting political parties. There are two opinions about whether efficient labor market policies-de- fined as market-determined wages and employment levels-are more or less likely to occur under more authoritarian regimes. Country examples provide apparent support for both views. One view is that labor repression, which is only possible under authoritarianism, offsets the rent-seeking behavior of po- tentially powerful trade unions. According to this notion, freer societies are sus- ceptible to inflated insider wages, strikes, and other forms of industrial unrest and consequently are susceptible to lower investment, less employment, and slower rates of growth. In this view high levels of civil liberty permit trade unions to agitate for above-market rents. With contestable elections, unions are likely to be granted these rents to nullify their political opposition. Hence supporters of this view conclude that authoritarian societies, which are better able to dis- courage the rent-seeking activities of free unions, are more compatible with labor market efficiency. Banerji and Ghanem 173 Examples from Chile, the Republic of Korea, Singapore, and Turkey in the 1970s and early 1980s support this view. At various times during these two decades, authoritarian regimes repressed trade unions and denied basic rights to workers. And, during the years of repression, Korea, Singapore, and Turkey experienced spectacular growth in the manufacturing sector and an increasing demand for labor. Rising profitability and labor demand in manufacturing in- creased the welfare of workers as a whole. Although similar results were not immediately apparent during Chile's authoritarian phase, many observers be- lieve that labor reforms undertaken during authoritarian rule aided the strong resurgence of the Chilean economy in the 1990s. Examples of democracies that adopted counterproductive labor market poli- cies lend further credence to this argument. In India, long considered to be one of the few stable democracies in the developing world, labor market distortions have been common. For example, highly unionized coal miners in Bihar, one of India's most populous and politically important states, enjoyed large jumps in real wages in the years before national elections from 1972 to 1985 (Banerji and Sabot 1994). Over the years, democracies as diverse as Trinidad and Tobago, Sri Lanka, and Senegal have awarded large economic rents to their unionized employees. The contrasting view is that the probability of governments passing ineffi- cient labor legislation to benefit insiders is higher under authoritarian than un- der democratic regimes. Authoritarian regimes are often subordinate to special interests that let them hold on to power (Ames 1987). Lacking the broader base of more democratic governments, such regimes may use labor policies to ac- quire support from powerful groups including the urban labor elite. An impor- tant difference between democratic and authoritarian regimes is the degree of outsider influence. In a well-functioning democracy outsiders vote and impose some limits on what narrower interest groups can achieve. In a dictatorship the government need only worry about those groups that wield direct power. Be- cause power is less evenly distributed than votes in this view, one should expect insider groups to have a greater say in dictatorships. Support for this view comes from the fact that labor market distortions per- sist in many countries that are not democratic. For instance, the authoritarian government of General Velasco in Peru introduced numerous inefficient labor laws. The legacy of these laws continued into the 1980s, when no workers could be dismissed or even made to retire (Sheahan 1987). Sub-Saharan Africa pro- vides many examples. Overstaffing in the public sector, high minimum wages, and restrictions on firing-policies introduced in the Congo, Kenya, Sudan, Tan- zania, and Zambia in the 1960s-were typical of postindependence authoritar- ian African governments that needed to placate urban populations to avoid po- litical unrest, but did not have to worry about millions of poor, informal, and rural workers. The phenomenon of authoritarian regimes succumbing to inter- est group pressure is not limited to Sub-Saharan Africa. In Egypt in the 1950s and 1960s, President Nasser, needing the support of the urban middle class, 174 TlHE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I guaranteed all college graduates a public sector job. In Bangladesh in the 1980s, General Ershad needed to neutralize the labor elite represented by SKOP, an or- ganization of trade unions. He negotiated with SKOP and agreed to increase pub- lic sector wages and double severance pay, allowances, and nonwage benefits. Although there are numerous democratic industrial countries with efficient labor markets, there are few long-lived democracies with efficient labor policies in the developing world. A few examples do stand out, such as Botswana and Mauritius-countries that combine relatively free polities with less distorted labor market policies. Also, countries that have made the transition from authoritarianism to democracy recently have been more likely to avoid adverse labor policies. For example, the new government in Chile moved to democracy and trade union freedom without increasing the rents for insiders. The end of repression in Korea in 1987 was initially associated with very contentious indus- trial relations. However, since 1990, collective bargaining has become an estab- lished institutional arrangement, with no negative impact on the functioning of the labor market or wage competitiveness. The paucity of examples of countries with free political systems and efficient labor markets may merely reflect the fact that political freedom is a relatively new phenomenon in the developing world. However, as more developing coun- tries adopt more participatory political systems, the question of whether such systems will improve or worsen the operation of their labor markets becomes even more pertinent. II. ANALYTICAL FRAMEWORK We use a simple analytical framework to develop empirical tests to see which of these two views gets more support from available cross-country data. The framework is based on a static two-sector political economy model similar to that of Grossman and Helpman (1994), but with a more detailed specification of the labor market. An abbreviated version of the model is presented in the appendix. It describes a developing economy with two sectors-an urban, for- mal, manufacturing sector and a rural, informal, agricultural sector. Labor market distortions, like a minimum wage, are usually introduced to benefit formal sector workers, because informal sector workers are not covered by any labor legislation. Formal sector workers and owners attempt to sway the government to enact policies favorable to them. Workers want higher minimum wages, and capital- ists want higher profits. Therefore, both groups try to limit the degree of open- ness of the economy. While in a closed economy, above-market minimum wages and above-normal profits can be passed on as higher prices to domestic consum- ers, this becomes less easy when consumers are free to buy imported substitutes. Rents in the economy are created through protection and then are divided be- tween formal sector workers and capitalists, although sometimes the govern- ment itself takes a share of the rent. Banerji and Ghanem 175 Although government policies may be influenced by dispassionate concerns about societal welfare, the government also undertakes actions aimed at main- taining its hold on power.' Rulers could have several reasons for wanting to remain in power, one being to continue appropriating "economic rent." The size of the rent the government wants to expropriate varies-the more venal a government, the more it extracts resources (which are "directly unproductive," the term coined by Jagdish Bhagwati) from the economy.2 The government has to take several factors into account in this sort of a political economy framework. First, it has to decide on its degree of venality- that is, how many resources it wants to seize. This would, in a more complete model than that presented in the appendix, depend on two factors: how much it cares about investment and future growth of the economy and how strongly it believes this venality will affect its chances to continue governing. Second, the government needs to estimate how support from each pressure group will im- prove its chances of staying in power. In a way, it has to weigh the importance of each group in the political process. For example, in a country where formal labor is fragmented and politically powerless, only capitalists have a voice in the political process. The reverse may be true in societies where organized labor is a key to mobilizing votes. How could the type of regime matter? First, it could affect the government's level of venality. If authoritarian regimes tend to be more (or less) corrupt than democratic ones, they will create more (or less) rent through protection, some of which the government will appropriate. Second, the type of regime could affect the influence of different interest groups. If authoritarian regimes tend to be more (or less) subordinate to special interests, including formal sector labor, then they will generate more (less) rent through protection and give part of it to the special interests.3 This implies that policies are influenced by political factors (including the nature of the regime and the weight it attaches to labor and capitalist groups) as well as economic factors (wages, prices, and the structure of production and consumption). From this discussion and the formal model in the appendix, we derive two reduced-form equations for the level of protection of the domestic economy and the amount of wage distortions: (1) T =f(e,1,k,R) (2) (w=f(r,e,1,k,R). 1. For another example of this type of analysis, see Hettich and Winer (1988). They develop the essential elements of tax systems as the outcome of rational behavior in a model where government maximizes expected support. 2. Lal (1984) argues, for example, that successive empires in Northern India's past were extremely venal. For a theoretical discussion of directly unproductive, profit-seeking (Dup) activities, see Bhagwati, Brecher, and Srinivasan (1984). 3. Subordinate here is used in the sense of Rodrik (1992). 176 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I The level of protection (t) depends on economic parameters (e), the relative political importance of urban workers and capitalists (I and k, respectively), and the nature of the regime (R). The wage distortion (co) is a function oft as well as e, 1, k, and R. In the case of a small economy, the economic parameters that we can expect to influence X and co include the price elasticities of consumption and production, the wage elasticity of the demand for labor, and the international price of the industrial good. (See the appendix.) A priori, we expect X to be increasing in I and k-. As interest groups become more powerful, there will be more pressure to generate protectionist rents. The impact of R, the degree of political authoritarianism, on r depends on which view about the effect of democratization on protectionism proves to be correct. We expect the wage distortion to be increasing in X and I and to be decreasing in k. As more rent is available through trade protection, more could be given to urban workers. As the importance of urban labor as an interest group rises, urban workers will be able to get a larger share of the available rent, but an increase in the political importance of capitalists will lower the share of rents going to labor in the formal sector. As described in the previous section, the relationship between co and R is the subject of debate and the focus of our em- pirical tests. III. ESTIMATING EQUATIONS AND DATA We estimated equations 1 and 2, using cross-sectional data for ninety devel- oping countries, to test whether the type of regime influences the level of trade protection and wage distortion. Our first task was to find consistent cross- country data. Because some data were missing for some countries, the actual number of observations varied from one regression to the other depending on the availability of the variables included in the regression. For a measure of trade distortion, r, we started by using the work of Dollar (1992), who estimates a measure of an economy's openness to trade for a large number of countries. This index-which we refer to as DOLLAR-rises as the level of protection increases. Because this index is available only for 1990, our analysis using DOLLAR was limited to cross-sectional work for that year. How- ever, DOLLAR, like other measures of trade distortion, has shortcomings. Pritchett (1991) shows that alternative measures of trade distortion produce very differ- ent country rankings and argues that the probability of deriving a single mea- sure that produces a correct ranking of countries is very low. (Also see Leamer 1988 and Harrison 1994.) We did not attempt to find such a measure, but we did experiment using alternatives to DOLLAR to check the robustness of our con- clusions. Three alternatives were used: the mean frequency of nontariff barriers on manufacturing goods (MNTB), the mean of total import charges on manufac- turing goods (TARIFF), and the black market premium (BMP). Table 1 shows the correlation coefficients between different measures of trade distortion. They were not highly correlated, with correlation coefficients rang- Banerji and Ghanem 177 Table 1. Correlation Coefficients between Different Indexes of Trade Distortion Mean frequency of Mean total import nontariff barriers charges on on manufactured manufactured Black market goods, goods, premium, Index DOLLARa MNTB TARIFF BMP DOLLARa 1.00 Mean frequency of nontariff barriers on manufactured goods, MNTB 0.08 1.00 Mean total import charges on manufactured goods, TARIFF -0.21 0.23 1.00 Black market premium, BMP 0.21 0.37 0.08 1.00 a. An index measuring an economy's openness to trade for a large number of countries (see Dollar 1992). Soturce: Authors' calculations. ing from a high of 0.37 forMNTB andBMP to a low of-0.21 for theDOLLAR index and TARIFF. The difficulty of choosing a single measure of trade distortion is further dem- onstrated by table 2, which shows the rankings of the average of each measure of trade distortion for seven regions of low- and middle-income countries. No two measures identified the same region as most trade distorted (1) or most open (7), and SouthAsia was ranked as most distorted by one measure (TARIFF) and most open by another (DOLLAR). We relied on DOLLAR as our primary measure and used the others for sensitiv- ity analysis, because DOLLAR provided the largest number of consistent cross- country observations for a recent year, 1990. We could not use some appealing measures of openness-like the one developed by Leamer (1988)-because of the small number of observations for low- and middle-income countries. Measuring labor market distortions and the degree of protection to insiders was also a difficult empirical task. Ideally, we would have liked to use the ratio of formal to informal or rural wages, adjusting it for different worker character- istics. Once again, however, the data were not available for a large number of countries. Hence, we used as the numerator the average wage of an unskilled worker in manufacturing (obtained from ILO 1991), assuming that unskilled workers have characteristics resembling those of informal workers. Finding a suitable denominator was more problematic, because informal sector wage rates are not available for many countries. Instead, we had to use nonmanufacturing gross domestic product (GDP) per worker as a proxy for the informal wage rate. Thus, our wage distortion index (IDX) is simply the ratio of the manufacturing 178 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO. I Table 2. Rankings of Degree of Trade Distortion in Seven Regions by Four Indexes Middle East East East Europe and Latin and North South West Index Africa Asia Central Asia America Africa Asia Africa DOLLAR' 2 6 5 4 3 7 1 Mean frequency of nontariff barriers on manufactured goods, MNTB 1 6 2 7 5 3 4 Mean total import charges on manufactured goods, TARIFF 4 7 5 2 6 1 3 Black market premium, BMP 4 6 7 5 1 2 3 Note: Rankings are from most distorted (1) to most open (7) and are based on the mean value of each index for each region. Indexes are defined in table 1. a. An index measuring an economy's openness to trade for a large number of countries (see Dollar 1992). Source: Dollar (1992) and Pritchett (1991). wage rate to the nonmanufacturing value added per worker (calculated from World Bank 1993). An increase of this index was taken to imply greater distor- tions and more privileges for insiders. For sensitivity analysis we used the proportion of workers in wage employment (PwE) as a second measure of labor market distortion (also from World Bank 1993). We expected PWE to be negatively related to wage distortion, because higher formal wages lead to a decline in output and employment in that sector, pushing more workers into agriculture and other informal activities. We used PWE rather than IDX as the dependent variable to test the robustness of our conclusions. Table 3 ranks the same seven regions according to our two measures of labor market distortion. The region with the highest mean value of IDX received the ranking of 1, the second highest, 2, and so on. Whereas the region with the lowest mean value of PWE received the ranking of 1, the second lowest, 2, and so on. As can be seen from table 3, the two measures of labor market distortion yielded almost identical rankings for the seven regions. Regarding economic parameters, in the discussion of the structural equation, we specified three elasticities-the price elasticity of consumption, the price elas- ticity of production, and the wage elasticity of labor demand-as possible sources of intercountry variance in distortions. Reliable data on these elasticities are not available for most countries, particularly developing countries. As countries develop, the mix of goods produced and consumed changes also, as do the corresponding average elasticities. This product mix would depend not Banerji and Ghanem 179 just on the income level of a country but also critically on the human capital stock of the labor force (and, perhaps, of the consumers). Therefore, we thought it reasonable to use GDP per capita and the level of schooling as proxies for the economic factors in our structural equations. Both of these variables should be positively related to each of the three elasticities. This would hold under the assumptions that richer developing countries that have a higher stock of human capital are more likely (i) to consume proportionately fewer goods that are ne- cessities and (ii) to have more diversity of production and access to nonlabor factors of production. If richer countries consume a higher proportion of luxury goods than do their poorer peers, increased GDP per capita would be associated with a higher economywide price elasticity of consumption. Although the rela- tionship with schooling is not as intuitive, there is nothing to suggest that more educated consumers would prefer more goods that are necessities. A more di- verse and flexible production process should, certainly, be associated with a workforce that is relatively better off and more educated. Richer countries should also, in general, have greater access to capital, technology, and other factors of production (including, of course, skilled labor as opposed to unskilled labor, whose wages were used here as data). Once again, the price elasticity of produc- tion should be positively associated with both GDP per capita and schooling. The same arguments would hold for the wage elasticity of labor demand. We used the secondary school enrollment rate as our measure of the level of schooling. This is because primary school enrollment rates show very little vari- ability across countries, except for some African economies. We did, however, run our regressions using the primary enrollment rate, as a test of robustness, and our conclusions remained unchanged. We obtained measures of political structure from various sources. For a mea- sure of the relative importance of urban, formal labor, we used the rate of ur- banization. Ideally, we would have liked to use a measure of labor organization, such as the degree of unionization, but the availability of these data for enough countries and their comparability across countries are serious problems. To test Table 3. Rankings of Degree of Labor Market Distortion in Seven Regions by Two Indexes Middle East East East Europe and Latin and North South West Index Africa Asia Central Asia America Africa Asia Africa Wage distortion index, IDXa 2 4 7 5 6 3 1 Proportion of workers in wage employment, PWE 3 4 7 5 6 2 1 Note: Rankings are from most distorted (1) to least distorted (7) and are based on the mean value of each index for each region. a. The ratio of the manufacturing wage rate to the nonmanufacturing value added per worker. Source: Authors' calculations, from ILO (1991) and World Bank (1993). 180 THE WORLD BANK ECONOMIC REVIEW, VOL. II, NO. 1 for robustness, we also used the proportion of nonpoor urban workers in the population as a proxy for the political importance of formal labor. As a measure of the importance of capital owners, we used the income share of the top fifth of the population, recognizing that this is not an ideal measure. We also wanted to add to equations 1 and 2 a variable that would represent the size of government and the amount of resources it is able to extract from the economy. (The need for a tax measure as an explanatory variable is made clear by the model in the appendix.) The ratio of tax to GDP, obtained from the International"Monetary Fund (IMF) statistics, was a natural choice. The type of regime was classified using the indexes for political rights and civil liberties compiled by Freedom House.4 These indexes range from 1 to 7, with 1 indicating the most rights and liberties and 7 the least. As defined by Freedom House, "Political rights enable people to participate freely in the po- litical process, [and] civil liberties are the freedoms to develop views, institu- tions, and personal autonomy apart frorm the state" (Ryan 1994, p. 671). In practice, political and civil liberties go hand in hand. Figure 1 presents regional averages for the two indexes and shows the high positive correlation between them-for the year 1990, we calculated the correlation coefficient be- tween the two indexes to be 0.92. In our regressions, we used the two indexes interchangeably, with very little difference in results. On the basis of the preceding discussion, we estimated three equations. Equa- tion 3 is derived from equation 1 and estimates trade distortions. Equation 4 is derived from equation 2 and estimates labor market distortions measured by IDX, while equation S is derived from equation 2 and estimates labor market distortions measured by PWE. (3) DOLLAR = tL0 + Ca, GDP per capita + a2 urbanization + a3 INSH + a4 education + a5 tax rate + a6 POL + Ul (4) IDX = Po + 131 GDP per capita + ,B2 urbanization + 3 INSH + 4 education + ,B5 tax rate + 06 POL + 137 DOLLAR + U2 (5) PWE = So + 81 GDP per capita + 82 urbanization + 83 INSH + 84 education + 85 tax rate + 86 POL + 87 DOLLAR + U3 where INSH is the income share of the top fifth of the population, and POL de- notes the index for political rights. The test of our two competing hypotheses of whether democratic or authori- tarian regimes are more prone to have distortionary labor policies consists of testing for the signs of a6, 136, 17, 86, and 87. If democratic regimes are more likely to achieve more efficient labor market outcomes, a6 and ,B6 should be positive and 86 negative-that is, regimes with more political freedom tend to place fewer restrictions on trade, which in turn implies a smaller spread between formal and informal workers and a relatively larger proportion of the labor 4. This index is sometimes called the Gastil index, after its originator. However, the updated indexes are no longer compiled by Gastil, but by Freedom House. Banerji and Ghanem 181 Figure 1. Measures of Political and Civil Liberties, by Region, 1990 Index 6 5 4 3 East East Europe and Latin Middle South West Africa Asia Central America East and Asia Africa Asia North Africa C,J Mean political rights index * Mean civil liberties index Note. Lower values for the indexes indicate higlher levels of rights and liberties. Source- Freedom House. force in formal employment. Our analytical model predicts that under both hypotheses the sign of P7 should be positive and 67 should be negative-that is, labor market distortions cannot persist if product markets are -competitive. Estimates of equations 3 to 5 can demonstrate whether there is a relationship between regime type and labor policy. However, they cannot shed light on why this is the case. Particularly, they do not show whether authoritarian or demo- cratic regimes are more subordinate to organized labor.5 To do so, we added an interaction term between the labor power and regime variables. IV. EMPIRICAL RESULTS We started by estimating equation 3 to look at the relationship between type of regime and level of trade protection. The parameter estimates are presented in table 4. The coefficient on the political liberties index was positive and signifi- cant in nearly all the regressions we tried, indicating that countries with less political freedom tend to have trade regimes that are more closed. When we included the indexes for both political and civil liberties in the regression, the 5. We are grateful to an anonymous referee of this journal for pointing this out and suggesting a solution. Table 4. The Impact of Regime Type and Other Variables on the Level of Trade Protection Income share Interaction term Proportion Ratio of Political of the top between labor of nonpoor Civil GDP Rate of tax to rights fifth of the power and regime urban liberties Number of Equation Constant per capita urbanization Educationa GDpb index' population variablesd workers index' observations R2 1 4.7 -0.0 * 1.6* -1.3 104.2 17.5** 0.7 -0.3 43 0.27 (0.1) (-1.8) (1.9) (-0.8) (1.1) (2.1) (0.8) (-1.4) 2 53.7 -0.0 0.6 -1.5 52.2 7.2 0.6 43 0.23 (1.03) (-1.4) (1.2) (-0.8) (0.6) (1.6) (0.7) 3 107** -0.0 0.3 50.2 9.1* -5.6 71 0.24 (4.7) (-0.6) (1.0) (0.76) (1.8) (-0.8) 4 98.2** -0.0 0.3 -2.7** 63.4 5.4** 71 0.24 (5.0) (-0.5) (1.0) (-2.3) (1.0) (2.16) S 90.3** 0.0 0.2 -2.4 7.9* 0.2 40 0.20 (3.8) (0.2) (0.3) (-0.0) (1.9) (0.3) 6 111** -0.0 -2.4** 40.2 5.1* 82 0.18 (7.6) (-0.1) (-2.1) (0.7) (2.0) 7 97** -0.0 44.1 7.1** 82 0.18 (8.2) (-1.0) (0.8) (2.9) 8 100** -0.0 7.6** 90 0.18 (8.2) (-1.0) (3.4) 9 101** -0.0 8.0** 90 0.14 (6.8) (-1.1) (3.0) 10 96.8** -0.0 0.3 -2.7** 62.5 5.8** 71 0.24 (4.8) (-0.5) (1.0) (-2.3) (1.0) (2.1) Note: The dependent variable is the DOLLAR index of openness. An increase in the DOLLAR index implies greater restrictions on trade. The last equation was estimated using instrumental variables. Numbers in parentheses are t-statistics. * Significant at the 10 percent level. ' Significant at the 5 percent level. a. The secondary school enrollment rate. b. Obtained from IMF statistics. c. Compiled by Freedom House (1994). d. Labor power is the proportion of nonpoor urban workers; the regime variables are the political indexes. e. Compiled by Freedom House (1994), also called the Gastil index. Source: Authors' calculations. Banerji and Ghanem 183 coefficient on the political index remained significant, although less than before, but the civil liberties index was insignificant. This result is not surprising, be- cause the two indexes are so highly correlated. When we dropped the civil liber- ties index, the political index became much more significant, and when we dropped the political index, civil liberties became significant. But in most of the regressions we tried, the political index outperformed the civil liberties index. We also tried to sort out possible two-way linkages. It has been argued that richer societies choose to be democratic (see, for example, Helliwell 1992). Be- cause they also tend to be more open, the direction of causation may be from openness to the political regime and not, as is suggested by our analytical frame- work, the other way around. Therefore, we reestimated our equations using lagged values of the political index. The results, reported in the last row of table 4, did not change. Of the two variables included for economic structure, only the level of educa- tion appeared significant, with countries that have a more educated labor force tending to be more open. Surprisingly, GDP per capita and the tax to GDP ratio appeared to be unrelated to openness. Although our results indicate that more democratic countries tend to have more open economies, they do not support the view that this is related to the power of interest groups. The proxy for the importance of formal labor (urban- ization) was marginally significant in only one equation, and the proxy for the importance of capitalists (the income share variable) was never significant. When we dropped the income share variable, the significance of urbanization did not improve. We also tried to interact the regime variable with the labor power variable, but the interaction term was not significant. Finally, we tried using the proportion of nonpoor urban workers as a proxy for labor power instead of urbanization, but the results were even worse. This could be because the num- ber of urban workers is not a good proxy for labor influence. However, if it is a good proxy, the insignificance of the urbanization and interaction variable would indicate that, for our sample, labor power in the economy has no effect on the degree of openness, regardless of the type of regime. The robustness of the relationship between openness and regime type was further tested by reestimating the equations presented in table 4 using MNTB, TARIFF, and BMP as dependent variables. The results for MNTB and BMP were quite similar to the ones obtained for the DOLLAR index, with the coefficient on the index for political rights being positive and significant. However, the use of TARIFF as a measure of openness changed our conclusion. The coefficient on the index for political rights in the TARIFF equation was insignificant and had a nega- tive sign. Table 5 reports results for one form of equation 3 with MNTB, TARIFF, and BMP as dependent variables. Estimates of the parameters of equation 4 are presented in table 6. As predicted by our analytical framework, there was a strong positive correlation between the DOLLAR index and the index of wage distortion. The index of political liberties became significant only when the DOLLAR index was dropped from the equation. Table 5. The Impact of Regime Type and Other Variables on Three Measures of Openness Ratio of Political GDP Rate of tax to rights Number of Dependent variable Constant per capita urbanization Education' GDP b index, observations R2 Mean frequency of nontariff 26.7 -0.00 -0.04 0.31 -46.4 8.0** 48 0.24 barriers on manufactured (1.2) (-0.5) -(0.13) (0.22) (0.6) (2.7) goods, MNTB Mean total import charges 64.3 -0.00 -0.02 0.53 -94.4* -2.6 48 0.19 on manufactured goods, (4.2) (-1.1) (-0.1) (0.5) (-1.8) (-1.20) TARIFF Black market premium, 15.9 0.01 -0.05 9.2 -1082** 47.7** 67 0.23 BMP (0.14) (0.4) (-0.0) (1.4) (-2.8) (2.7) Note: Numbers in parentheses are t-statistics. Significant at the 10 percent level. * Significant at the 5 percent level. a. The secondary school enrollment rate. b. Obtained from IMF statistics. c. Compiled by Freedom House (1994). Source: Authors' calculations. Table 6. The Impact of Regime Type and Other Variables on the Index of Wage Distortion Income Interaction share of the term between Proportion Rate of Political top fifth labor power of nonpoor GDP Rate of tax to rights of the and regime urban Number of Equation Constant per capita urbanization Educationa GDPb index' DOLLARd population variableso workers observations R2 1 3.3 0.0 0.0 -0.1 0.0 0.1 0.01 * -0.1 ** -0.0 26 0.41 (1.6) (0.3) (0.1) (-1.2) (0.0) (0.4) (1.7) (-2.1) (-0.2) 2 3.5 ** 0.0 -0.0 -0.6 -0.4 0.1 0.01** -0.1** 26 0.41 (2.5) (0.4) (-0.2) (-1.2) (-0.2) (0.6) (2.3) (-2.3) 3 3.1** 0.0 -0.04 -3.7 0.1 0.02** -0.1** -0.02 21 0.64 (2.1) (0.7) (-0.8) (-1.4) (0.6) (3.4) (-2.8) (-1.0) 4 0.6 0.00 -0.02 -0.04 -2.3 0.1 0.02** 42 0.45 (0.6) (0.6) (-1.4) (-0.8) (-0.8) (0.7) (3.0) 5 0.9 0.00 -0.02 -0.04 -2.4 0.02** 42 0.44 (1.0) (0.5) (-1.6) (-0.8) (-0.9) (3.5) 6 1.9** 0.00 -0.01 -0.1 0.4 0.21* 42 0.32 (2.0) (0.2) (-0.8) (-1.S) (0.14) (1.8) 7 -0.4 -0.00 0.02** 48 0.26 (-0.4) (-1.4) (3.3) 8 0.65 -0.00 0.36** 50 0.21 (0.9S) (-0.6) (2.7) 9 0.S -0.0 -0.0 -0.0 -2.2 0.1 0.02** 42 0.45 (0.5) (-0.7) (-1.4) (-0.7) (-0.8) (1.0) (2.8) Note: The dependent variable is the index of wage distortion. An increase in the index of wage distortion implies worse labor market policies. The last equation was estimated using instrumental variables. Numbers in parentheses are t- statistics. * Significant at the 10 percent level. * Significant at the 5 percent level. a. The secondary school enrollment rate. b. Obtained from IMF statistics. c. Compiled by Freedom House (1994). d. An index measuring an economy's openness to trade for a large number of countries (see Dollar 1992). e. Labor power is the proportion of nonpoor urban workers; the regime variables are the political indexes. Source: Authors' calculations. Table 7. The Impact of Regime Type and Other Variables on the Proportion of the Labor Force in Wage Employment Income Interaction share of the term between Proportion Rate of Political top fifth labor power of nonpoor CDP Rate of tax to rights of the and regime urban Number of Equation Constant per capita urbanization Educationa GDpb index' DOLLAR' population variablese workers observations R2 1 -30.3 -0.0 0.9** 0.3 96.4** 3.1 -0.1* 0.7* -0.0 34 0.61 (-1.3) (-1.2) (3.0) (0.5) (2.6) (1.0) (-1.8) (1.9) (-0.5) 2 -24.6 -0.0 0.8** 0.3 90.2** 1.9 -0.1* 0.6* 34 0.70 (-1.30) (-1.1) (4.2) (0.5) (2.7) (1.6) (1.7) (1.9) 3 -37.8* -0.0* 0.5 62.8* 0.3 -0.1 1.2** 0.9** 30 0.71 (-1.9) (-1.7) (0.8) (2.0) (0.2) (-0.8) (3.6) (4.9) 4 23.6* -0.00 0.7** 0.2 74.2 -0.4 -0.15** 50 0.65 (2.1) (-0.4) (4.5) (0.4) (2.1) (-0.3) (-2.3) 5 18.3* -0.00 0.7** 0.2 72.6** -0.15** 50 0.65 (1.8) (-0.4) (4.6) (0.4) (2.1) (-2.4) 6 7.5 -0.0 0.6** 0.8 52.4 -1.0 53 0.63 (0.8) (09.2) (4.0) (1.6) (1.6) (-0.8) 7 48.0** 0.01** -0.1 60 0.35 (4.9) (5.0) (-1.4) 8 50.2** 0.00* -2.8** 67 0.37 (7.7) (5.2) (-2.1) 9 18.8> -0.00 0.7** 0.2 73.3** -0.2 -0.15** 50 0.65 (1.7) (-0.4) (4.6) (0.4) (2.1) (-0.1) (-2.4) Note: The dependent variable is the proportion of the labor force in wage employment. An increase in the proportion of wage employment is taken as an indication of good labor policies. The last equation was estimated using instrumental variables. Numbers in parentheses are t-statistics. * Significant at the 10 percent level. ** Significant at the 5 percent level. a. The secondary school enrollment rate. b. Obtained from IMF statistics. c. Compiled by Freedom House (1994). d. An index measuring an economy's openness to trade for a large number of countries (see Dollar 1992). e. Labor power is the proportion of nonpoor urban workers; the regime variables are the political indexes. Source: Authors' calculations. Banerli and Ghanem 187 Our results imply a strong link between authoritarianism and protectionist poli- cies, and between protectionist policies and labor market distortions, but the direct link between political liberties and labor distortions seems less strong. This sug- gests that regime type may affect the labor market mainly through trade policies and that, for a given level of protection, the distribution of rents between different interest groups is little affected by the degree of political freedom. Our results concerning the impact of lobbies and interest groups on labor market distortions were mixed. The coefficient on the income share variable was negative and statistically significant. This fits with the predictions of our analytical framework: capitalists will try to limit labor's share in the economic rents. However, the coefficients on the proxies for labor power (urbanization and the proportion of nonpoor urban workers) were never significant-either alone or when interacted with the political variable. Assuming that these are good proxies, the results suggest that, whatever the regime type, organized la- bor has little power to affect overall wage distortions in the economy. Both results are consistent with the anecdotal evidence presented in section II. Remarkably, none of the economic structure variables appeared to have an impact on wage distortions. This contradicts the general impression that poorer countries tend to have more distorted labor markets. We further tested the robustness of our results by estimating equation 5 (see table 7). The results supported our previous conclusion concerning the link be- tween openness and labor policy. The parameter on the DOLLAR index was nega- tive and significant in nearly all of the regressions we tried, indicating that more closed economies generally have relatively smaller formal labor markets. The parameter on political liberties was also negative, but significant only in equa- tions where the DOLLAR index was dropped. The proxies for labor power (urbanization and the proportion of nonpoor urban workers) entered those regressions with very significant positive coeffi- cients, which is the opposite of what is predicted by the model and may simply reflect the fact that urbanization and the proportion of nonpoor urban workers are not predetermined with respect to the size of the formal sector. We tried to deal with this by using instrumental variables, with lagged values of the ex- planatory variables as instruments, but the results did not change (see the last row of table 7). The coefficient on the income share variable was positive and significant, as predicted by the model. This result could mean that in most coun- tries employers' political power is more important than labor's power in deter- mining the share of rents that go to labor. Or it may simply reflect the poor quality of the data for labor organization. We also experimented with dropping some variables. The results of regres- sions where we dropped all explanatory variables except GDP per capita and the DOLLAR and political liberties indexes are also presented in table 7. In both cases GDP per capita had a positive and significant coefficient. The openness index had a negative coefficient, with a t-statistic that was just below the 10 percent criti- cal value, and the political liberties index had a negative and significant coeffi- 188 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO 1 cient. Countries with less political freedom apparently have relatively smaller formal labor markets, an indication of greater distortions. We tried redoing the estimates in tables 6 and 7 using other measures of openness. The results were not as strong. Estimates with MNTB and BMP had the same sign as estimates with the DOLLAR index but were rarely statistically sig- nificant. This could be because we had fewer observations for MNTB and BMP than for DOLLAR or because MNTB and BMP are inferior to DOLLAR as measures of trade protection. Estimates using TARIFF were also insignificant but had the wrong sign-which could have been expected from the correlation coefficients pre- sented in table 1. Thus, our general conclusion that countries with less liberty tend to have more barriers to trade and more labor market distortions is quite sensitive to the choice of openness measures. V. CONCLUSIONS Our work provides evidence that authoritarianism may be associated with higher trade protection and that trade restrictions, in turn, are associated with more labor market distortions. This conclusion fits the pattern observed in many countries in Sub-Saharan Africa, the Middle East, and Latin America in the 1970s and 1980s, where authoritarian governments provided producers with protection through tariffs and quotas and ensured that workers in favored in- dustries obtained part of the rent through various labor market interventions. Those same governments often repressed labor in other sectors and only toler- ated unions subordinate to the regime. Landell-Mills and Serageldin (1992) argue that freedom of association is one of the elements of good governance that are necessary for development. Inas- much as authoritarian regimes that did not respect freedom of association tended to be linked with policies restricting trade and distorting labor markets, our work supports this view. This is not to say that inequitable or inefficient labor policies are the exclusive domain of authoritarian governments or that authoritarianism automatically implies these policies. As we discussed in the introduction, there are numerous counter-examples. Our results, based on a large cross section of developing countries, also support the work of Fields (1994) and Freeman (1993) on East Asia, which argues that labor repression was nei- ther necessary nor desirable for development. Our findings are also consistent with empirical literature linking political de- mocracy and economic growth. Works by Barro (1991), Ozler and Rodrik (1992), and others have found that having less civil liberty is strongly correlated with slower growth. And the case for openness stimulating growth is well known. Similarly, undesirable labor market policies may handicap economic growth by stifling the process of production and capital accumulation. Our conclusions associating authoritarianism with restrictive trade policies and labor market dis- tortions suggest some specific mechanisms as well as provide further empirical evidence linking the lack of political freedom with slower rates of growth. Banerli and Ghanem 189 How is the relationship between authoritarianism and trade and labor poli- cies mediated? Do authoritarian regimes tend to be more subordinate to special interests than regimes with more political and civil liberty? Data limitations have prevented us from responding adequately to those questions. Finding bet- ter proxies for the political power of labor and business groups and specifying equations that allow a deeper understanding of those links is a subject for future research. APPENDIX. AN ANALYTICAL MODEL OF LOBBIES, TRADE, AND LABOR POLICIES The influence of lobbying by organized labor and capitalist groups can be illustrated with the help of a simple, static, two-sector political economy model similar to that of Grossman and Helpman (1994) and Rama and Tabellini (1994). A more detailed exposition of the model can be found in Banerji and Ghanem (1995). Structure of the Economy, Production, and Incomes Assume that there are three groups in the economy, with the total population normalized to 1: (A-1) 0 < 1 < 1 urban manufacturing laborers (A-2) 0 < k < 1 capitalists (A-3) a = (1 - 1- k) rural agricultural laborers. The three groups are denoted by superscripts L, K, and A. Each individual in the economy consumes two goods-an agricultural good x and a manufactured consumption good c. Individual utility maximization for individual i is given by: (A-4) Max ui = xi + U(ci) subject to y'= x+ pc'. This gives us ci = c(p) for all consumers, where p is the domestic price of the manufacturing good, and yt is income of individual i. The agricultural good x is produced with labor only, at a wage normalized to 1: x = f(a) = a. With no loss of generality, migration from the agricultural to the manufacturing sector is assumed to be costless. Thus, urban workers have a floor wage of 1-the supply price of labor in the economy. In the manufacturing sector, quantity Q of the good c is produced with labor L < 1, and capital K owned by capitalists: Q = F(L, K). For tractability, K is assumed to be fixed. The government sets tariff X on imports of the manufactured good by setting the domestic price p = (p* + t), where p " is the world price of the manufactured good. Tariff revenues T are redistributed evenly to every individual in the economy (following Grossman and Helpman 1994). Because the population is normal- 190 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO. I ized to 1, each individual receives an amount T (p, w), where w denotes manufacturing wages, because imports can be shown to be a function only of prices and wages. Capitalists also receive income through manufacturing sector profits. Profit maximization is on the basis of the domestic price p (which differs from the world price p* by the amount of the tariff) and the prevailing manufacturing wage rate, w (which the government can influence through minimum wage and other legislation). Employed manufacturing workers receive w, while those left unemployed revert to the floor wage, 1. Finally, employed manufacturing workers and capitalists are forced to pay a tax v to the government out of their incomes (this is explained in greater detail below). They also pay contributions (or bribes), denoted as bi (j = L, K), to the government in an attempt to influence policy. The contribution from each of the two groups is a mapping with the prevailing policy parameters-the wage and the domestic price (affected by the tariff set by the government): b' = bl(p, w). They each optimize their political contributions until the marginal effect of the contribution on policy equals the marginal increase in their own welfare from the same policy: (A-5) uL = bp; UL= b L (A-6) u~=p bp; Uw = bK. Note the use of the stylized facts: rural/agricultural workers do not pay the income tax but do pay the tariff; however, they have no influence on the level of the tariff. The Government The government sets the policy parameters. Its objective may be not merely to act as a social planner that maximizes national welfare, but, in addition, to remain in power. This second purpose is motivated in our model by presuming that if the government is in power, it can extract rent from the formal sector's product-that is, it can charge the tax v on the incomes of employed labor and capitalists. For the time being, we assume that v is a given for the economy and measures a particular government's level of venality. The value v = 0 indicates a purely altruistic social planner, and v = 1 indicates a totally confiscatory regime. We also assume that the probability of the government maintaining its power is affected by nonpolicy actions that the government may take-for example, launching an election campaign or maintaining a military. These actions are financed by the political contributions it receives from labor and capitalists (which can, of course, also be thought of as proxies for the degree of support these groups provide to the government). The government thus maximizes its objective function F, where r = W + v, and Banerji and Ghanem 191 (A-7) WW=k uK(p, w) + IuL (p,w) +a uA (p, w) (A-8) V = 7r(B) v[pQ(p, w)] (A-9) B = sLbL (p, w) + sKbK (p, W). W represents pure aggregate welfare of individuals in the economy. The term V, by contrast, captures the expected rents to a nonaltruistic government from staying in power. As discussed, the rents amount to a tax v on the value of manufacturing production. Note that when v = 0 (the government is nonvenal), V = 0 and thus r = W-the government only maximizes aggregate welfare. tc indicates the probability that the government will stay in power in order to enjoy the rents (in this static model, it may also be thought of as the number of years that the government believes it will be in power). This probability is as- sumed to increase with larger political contributions from urban groups, that is, it '(B) > 0, as the government is assured more tangible and intangible political support from the interest groups. However, for a given government, the contri- butions from labor and capitalists may not have equal value: their relative im- portance in the particular polity is measured by the variables, SL and sK, respec- tively, with 1 2 sL > 0, and SK = 1 - sL. The term B, therefore, summarizes the subjective weight the government places on total political contributions. The government maximizes r with respect to its policy parameters p and w. Simplifying the resulting simultaneous equation system, we get the government's optimal tariff and manufacturing wage premium in equilibrium: (A-10) - = (- ){,_ 1--)[SK + (y + X)(SL -SK)]- _ v1 + y(l + Q] (A-11) (w) [SL (1 V + (l / X)(SLSK) Kw)L _1v)p (1-V)pJ where Q is the quantity produced of the manufacturing good, SI measures the government's assessment of the likelihood of staying in power due to the politi- cal contribution of group J, a indicates the corresponding weighted average of this likelihood, and SJ = (Vp / B)sJ, J = L, K. Note that when v = 0, SJ = 0, that is, a nonvenal government is not concerned about the effect of interest groups on the likelihood of its remaining in power. We also define a = (1+ kSK + ISL) (when v = 0, a = 1). Finally, we use some elasticity measures to simplify the equations: p = Bn'(B) / i measures the elasticity of the probability of remaining in power with respect to political contributions, y = pQp / Q is the price elasticity of production (that is, the supply elasticity) of the manufactured good, i1 = pcplc is the price elasticity of consumption, and X = (wLw)/L is the wage elasticity of labor demand in manufacturing. 192 THE WORLD BANK ECONOMIC REVIEW, VOL. I1, NO I These equations demonstrate a common feature of this class of models. The government makes its decision about the degree of openness in the economy independent of the decision about wages. Since Q and c are endogenous, the decision on the level of protection is a function of variables relating to the struc- ture of the economy (TI, y, and X) and political variables summarized by V, c, and S. This simple model also predicts that a government only interested in maxi- mizing welfare, that is, one that is nonvenal, will set tariffs at the level of opti- mal tariffs. However, if staying in power and collecting rents are also in its objective function, it will impose higher tariffs and create a larger wedge be- tween domestic and foreign prices. The size of the tariff will depend on the government's preferences, the power of interest groups, and economic factors (such as the price elasticity of imports) that determine the relationship between changes in tariffs and changes in revenue. Finally, the wage distortion depends directly on the trade regime. REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Alesina, Alberto, and Roberto Perotti. 1994. "The Political Economy of Growth: A Critical Survey of the Recent Literature." The World Bank Economic Review 8(3):351-71. Ames, Barry. 1987. Political Survival: Politics and Public Policy in Latin America. Ber- keley: University of California Press. Banerji, Arup, and Hafez Ghanem. 1995. "Political Regimes, Trade, and Labor Poli- cies in Developing Countries." Policy Research Working Paper 1521. Office of the Vice President for Development Economics, World Bank, Washington, D.C. Processed. Banerji, Arup, and Richard H. Sabot. 1994. "Wage Distortions and Overemployment in Developing Country Public Enterprises." Policy Research Department, World Bank, Washington, D.C. Processed. Barro, Robert J. 1991. "Economic Growth in a Cross-Section of Countries." Quarterly Journal of Economics 106(May):407-44. - . 1994. "Democracy and Growth." NBER Working Paper 4909. National Bureau of Economic Research, Cambridge, Mass. Processed. Bhagwati, Jagdish, Richard A. Brecher, and T. N. Srinivasan. 1984. "DuP Activities and Economic Theory." In David C. Colander, ed., Neoclassical Political Economy: The Analysis of Rent-Seeking and DUP Activities. Cambridge, Mass.: Ballinger Publish- ing Co. Dollar, David. 1992. "Outward-Oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDCs, 1976-1985." Economic Development and Cul- tural Change 40(3):523-44. Fields, Gary. 1994. "Changing Labor Market Conditions and Economic Development in Hong Kong, Korea, Singapore, and Taiwan." Cornell University, Ithaca, N.Y. Processed. Banerli and Ghanem 193 Freedom House. 1994. Freedom in the World: 1993-94. New York. Freeman, Richard. 1993. "Does Suppression of Labor Contribute to Economic Suc- cess? Labor Relations and Markets in East Asia." Harvard University, Cambridge, Mass., and London School of Economics, London. Processed. Friedman, Milton. 1962. Capitalism and Freedom. Chicago: University of Chicago Press. Grossman, Gene, and Elhanan Helpman. 1994. "Protection for Sale." American Eco- nomic Review 84(4):833-50. Harrison, Ann. 1994. "Openness and Growth: A Time-Series Cross-Country Analysis for Developing Countries." Working Paper Series 809. Office of the Vice President for Development Economics, World Bank, Washington, D.C. Processed. Helliwell, John F. 1992. "Empirical Linkages between Democracy and Economic Growth." NBER Working Paper 4066. National Bureau of Economic Research, Cam- bridge, Mass. Processed. Hettich, Walter, and Stanley L. Winer. 1988. "Economic and Political Foundations of Tax Structure." American Economic Review 78(4):701-12. ILO (International Labor Organization). 1991. Bulletin of Labor Statistics: October In- quiry Results. Geneva. Kormendi, Roger, and Philip Meguire. 1985. "Macroeconomic Determinants of Growth: Cross-Country Evidence." Journal of Monetary Economics 16(2):141-63. Lal, Deepak. 1984. "The Political Economy of the Predatory State." Discussion Paper DRD 105. Development Research Department, World Bank, Washington, D.C. Pro- cessed. Landell-Mills, Pierre, and Ismail Serageldin. 1992. "Governance and the External Fac- tor." In Lawrence Summers and Shekhar Shah, eds., Proceedings of the World Bank Annual Conference on Development Economics 1991. Washington, D.C.: World Bank. Leamer, Edward. 1988. "Measures of Openness." In Robert Baldwin, ed., Trade Policy Issues and Empirical Analysis. Chicago: University of Chicago Press. Ozler, Sule, and Dani Rodrik. 1992. "External Shocks, Politics, and Private Investment: Some Theory and Empirical Evidence." Journal of Development Economics 39(July):141-62. Pritchett, Lant. 1991. "Measuring Outward Orientation in Developing Countries: Can It Be Done?" Working Paper Series 566. Country Economics Department, World Bank, Washington, D.C. Processed. Przeworski, Adam. 1991. Democracy and the Market. Cambridge, U.K.: Cambridge University Press. Przeworski, Adam, and Fernando Limongi. 1993. "Political Regimes and Economic Growth." Journal of Economic Perspectives 7(summer):51-69. Rama, Martin, and Guido Tabellini. 1994. "Endogenous Product and Labor Market Distortions." Paper presented at the labor markets workshop, Policy Research De- partment, World Bank, Washington, D.C. Processed. Rodrik, Dani. 1992. "Economics and Politics Yesterday and Today: Political Economy and Development Policy." European Economic Review 36(2-3):329-36. Ryan, Joseph. 1994. "Survey Methodology: The Tabulated Ratings." In Freedom House, Freedom in the World 1993-94. New York. Scully, Gerald. 1988. "The Institutional Framework and Economic Development."Jour- nal of Political Economy 96(3):652-62. 194 THE WORLD BANK ECONOMIC REVIEW, VOL. 11, NO I Sheahan, John. 1987. Patterns of Development in Latin America: Poverty, Repression, and Economic Strategy. Princeton, N.J.: Princeton University Press. World Bank. 1993. World Development Report 1993: Investing in Health. New York: Oxford University Press. The best in development St atstics just got better! The World Bank introduces its most comprehensive, statistics-packed product to l ~~~date-World Developrnent Indicators 1997. Forrnerly the stastisfcal appendix to . ~~~~~the World Developmnent Report, these comprehensive data are now available . l ~~~~~~in their own volume and have been enlarged to include more than 80 tables with some 400 indicators. This major new publication provides an expanded view of the world economy for more than 130 countries- with chapters focusing on people, economy, environment, states and markets, and 11nom ll § global links. Concise, insightful commentary tells the story of how people live and work, and how countries are expanding and changing. April 1997 350 Pages Stock no. 13701 (ISBN 0-8213-3701-7) $60.00 I-SS 6AM This comprehensive database, which replaces World Data, contains most of the under- lying time-series data for the World DevelopmentIndicators and World BankAtlas. We've added powerful new features-now you can generate maps and charts, add more features! download your results to other software programs. Requires Windows 3..l'm April 1997 Individual Version: Stock no. 13703 (ISBN 0-8213-3703-3) $275.00 Network Version: Stock no. 13702 (ISBN 0-8213-3702-5) $550.00 world Bank Atlas 1997 __: -] e , .One of the Bank's most popular offerings, the Atlas has been redesigned as a companion to the World Development Indicators. Tables, charts, and 21 colorful maps address the development themes of people, economy, environment, and states and markeLs This easy-to-use, inexpensive book is an internaional standard in sta- dstical compilations and an ideal reference for office or classroom. Text, maps, and references appear in English, French, and Spanish. April 1997 48 Pages Stock no. 13576 (ISBN 0-8213-3 576-6) $15.00 World Bank Publications For US customers, contact The World Bank, P.O. Box 7247-8619, Philadelphia, PA 19170-8619. Phone: (202) 473-1155, Fax: (202) 522-2627. Shipping and handling: US$5.00. For airmail delivery outside the US, add US$8.00 for one item plus US$6.00 for each additional item. Payment by US$ check drawn on a US bank payable to the World Bank or by VISA, MasterCard, or American Express. Customers outside the US, please contact your World Bank distributor. 1244 Order Form CUSTOMERS IN THE UNITED STATES CUSTOMERS OUTSIDE THE UNITED STATES Complete this coupon and return to: Contact your local Bank publications distributor The World Bank for information on prices in local currency and Box 7247-8619 payment terms. If no distributor is listed for your Philadelphia, PA 19170-8619 country, use this order form and return it to the USA. U.S. address. Orders that are sent to the U.S. To have your order shipped faster, charge by address from countries with distributors will be credit card by calling (202) 473-1155 or send this returned to the customer. competed order coupon by facsimile by dialing (202) 522-2627. Quantity Title Stock# Price Total price * SHIPPING AND HANDLING charges are $5.00 Subtotal cost US$ per order. If a purchase order is used, actual ship- Shipping and handling* US$ ping will be charged. For air mail delivery outside T us$ the United Sates, add $8.00 for one item plus $6.00 ota for each additional item. CHECK METHOD OF PAYMENT Ui Enclosed is my check payable to the World Bank. z Charge my [ VISA O MasterCard U American Express Credit card account number Expiration Date Signature (required to validate all orders) 3 Bill me. (Institutional customers only. Purchase order must be included.) PLEASE PRINT CLEARLY Name Address City State Postal code Country Telephone 1244 Distributors of ~~~~COLOMBIA HONG KONG, MACAO JAMAICA NORWAY SINGAPORE, TAIWAN, TANZANIA Distributors of ~Intoenlace Ltda Asia 2000 Ltd. Ian Randle Publishers Ltd NIC Into A'S MYANMAR, BRUNEI Oaford Unrveorsihy Press WVorld Bank Carrera 6 No 51-21 Sates & Circstation Copartrment 206 Old Hope Road Book Cepartomont Asahgate Publishing Asia Maklaba Street Publications ~~~~~Apartado Aereo 34270 Seabird H-ouse, unit 1101.02 Kingston 6 P.O. Boo 6125 Enerstad Pacric Pte Ltd -PO Boo 5299 Publications ~~~Saotatd do Boget6, C C 22-28 Wyndhsam Street, Control Tel:9809-927-2085 N-0602 Osto 6 41 Kattang PuddIng Road 904-03 Oar es Saleaam Prices and credit terms Tot. (57 1) 285-2798 Haag Kong Tel: (47 22) 57-3300 Golden Whreet Burldiog Tot- (255 51) 29205 vary from country to CYPRUS ~~~~~~~~~~Tot: (852) 2530-1409 JAPAN Singapore 349316 ConsultoCoolerttartResYardS Eastern Book Service PAKISTAN Tel (65) 741-5166 THAILAND country. C nutyour CeirfrAppliod ReerhINDIA 2-13 Hoago 3-shame, Bunkyo-ku Mirza Back Agency Centrat Books Distributioo local distributor before Cyprus Cottege Allied Pablishers Ltd Tokyo 113 65, Shahrah-e-Quad-e-Azamo SLOVENIA 306 Sitam Road placing an order. 6. Ciogeoes Street, Engomni 751 Mount Road Tel. (81 2) 3818-0861 Lahrore 54000 Gosopdarski Vestnik Pablashog Groap Baogkok 10500 PO Boo 2006 Madras -600 002 Tel. (92 42) 735 3601 Ouaaa(ka costa 5 Tel, (66 2) 235-5400 ARGENTINA Nicosia Tot (91 44) 852-3938 KENYA 1000 Ljubljana OrScina dot Libro Intemoacional Tel (357 2) 44-1730 Atrica Book Service )E A)I Ltd. Ooford Usiversdy Press Tel: (386 61)13383 47:132 12 30 TRINIDAD & TOBAGO, Av Cordoba 1877 INDONESIA Guaran House, Mtanguan Street 5 Bangalore Town AND THE CARRIBBEAN 1120 Buenos Aires CZECH REPUBLIC Pi. Incfrroa Lmed P0 Boa 45245 Sharae Faoal SOUTH AFRICA, BOTSWANA Systemoatics Studies UneS Tel (54 1) 815-8354 National IntormoabonCenter Jalan Borobudur 20 Nairobi P0 Baa 13033 Forasingle tates. 9 Watsa Street prodeona, Koevktska 5 P0. BOa 181 Tel: (254 2) 223 641 Karachlm75350 Oxford Lisiversity Press Curepe AUSTRALIA, FUJI, PAPUA NEW CS - 113 57 Prague 1 Jakiaia 10320 Tel: (92 21) 446307 Southern Atrica Tronidad, West Indies GUINEA, SOLOMON ISLANDS, Tel: (42 2) 2422-9433 Tel: (62 21) 390-4290 KOREA, REPUBUC OF P0. Boo 1141 Tot: (809) 662-5654 VANUATU, AND WESTERN SAMOA Dae(on TradfIng Co. Ltd. Pak Book Corporation Cape Town 6000 CA titormatio Services DENMARK IRAN P0, Boa 34, Youkta Azn Chsambers 21 Tel (27 21) 45-7266 UGANDA 648 Whroehorse Road SamrrundsLdleratur Ketab Sara Co. Puiblishers 700 Seoun Bldg Queen's Road Gastro Lid Mitdhamo 3132 Rosenoerms AIlS 11 Khraled Esanborboi Ave, 44-6 YouidD-Dong,WYoogchrengpo-Ku Lahora For subscripion orders: PO Boa 9997 Vicoria DK-1 970 Frederflsberg C 6th Street Seoul Tel (92 42) 036 3202, 636 0088 ItetorationaJ Subscrptoon Service Madhvaoi Building Tet: (61) 39210 7777 Te: (45 31) 351542 Kusheh Delarooz No. 8 Tel. (82 2) 785-1631/4 P.O. Boa 41095 Plot 1614 Jor(a Rd PO. Boa 15745-733 PERU Craighall Kampala AUSTRIA EGYPT, ARAB REPUBLiC OF Tehran MALAYSIA Editoriat Desarrolo SA Johoannesburg 2024 Tot: (256 4t) 254 703 Gerold and Go. Al Ahram Distribution Agency To (99 21) 8717819; 8716104 Unaiversity ut Malaya Cooperotive Apartado 3024 Tel: (27 11) 880-1 448 We0hburggasse 26 Al Galaa Street Bookshep, Lnimted Lana 1 UNITED KINGDOM A-lOll Woen Cairo Kowktab Publishrers R.O, Boa 1127 Tel: (51 14) 2835380 SPAIN Micmoonto Ltd. Tel (43 t) 5t2-47-31 -0 Tel. (20 2) 578-6803 PO. Boxo19575-511 dalan Pantaj Barn Mundi-Prensa Liros, SA. P0 Baa 3 Tehran 59700 Kuala Lumopur PHILIPPINES Castello 37 Aflan, Hampsphire GU34 2PG BANGLADESH The Middle East Observer Tot: (98 21) 258-3723 Tel: (80 3) 756-500 International Booksource Center Inic. 28001 Madrid England Micra Iodustides tDevelopmeneo 41, Shoot Street 1127-A Antqpolo St. Tel: (34 1) 4.31P3399 Tot: (44 1420) 86848 Assistance Society (MIDAS) Cairo IRELAND MEXICO Bamnogay, Venezuela HouseS5, Road 16 Tot: (20 2) 303-9732 Goverrment Supplies Agency INFOTEC Matkati Cay MuodF-Prensa Burcetoe VENEZUELA Dhanonrodi 6/Area Od'ig an ISotAthair Ac Sean Fernando No. 37 Tel. (63 2) 898 6501; 6505; 6507 Cmnsell do Cent, 391 Tecnit-Cencea Libros, S A. Dhaka 1209 FINLAND 4-5 Harcaurt Road Gol. lionelo GuenTa 66009 Barcelona Contra Cukldad Coroercial Tamvanoo Tel. (880 2) 326427 Akateeminene Krrlakauppa Dubtim 2 14050 Mooice, DPF POLAND Totl (34 3) 488-3492 Nivel C2 P0. Baa 128 Tot. (353 1( 661-31 11 Tel: (52 5) 624-2800 International Pubishing Service Caracas BELGIUM FIN-00101 HeWisiki Ut Pieknra 31137 SRI LANKA, THE MALDIVES Tel. (58 2(959 5547, 5035; 001 6 Jean De Laniroy Tel (356 0) 12141 ISRAEL NEPAL 00-677 Warzawa Lube House Bookstrop Ac. do Roi 202 Yozmoro Literature Lid Everest Madea lnteirmhonat Tel- (48 2) 628-809 100, Sir Ciharoalaro Gardiner ZAMBIA 1 060 Brussets FRANCE P.O. Baa 56055 Services (2) Ltd Mawatha Uniersity Bookshop Tot: (32 2) 539-6169 Wibrtd Bank Pubrciafins 3 Yahanao Hasuncflar Street GOo Ban 5443 PORTUGAL ColorobD 2 University ot Zaobia 66, avenue dIrirra Tot Aviv 61560 Kathmsandu Lrvrana Portugal Tel: (94 1) 32105 Great East Road Camopus BRAZIL 75116 Paris Tel (972 3) 5285-397 Tel (977 1) 472 152 Apariado 2681 PO. Boa 32379 Pitlcmes Teacncas Inteoadonai Ltda. Tel (33 1) 40-69-30-56/57 RuaDo Caowa70-74 SWEDEN Lusaka Rua Peaxoto Gaomide, 209 RO0Y tntemoationral NETHERLANDS 1200 Lisbon Wennergren-Willaro AB Tot: (280 1( 252 576 01489 Sao Paulo, SR GERMANY P0 Boa 13050 Ce Lrndeboormno0r-Publikatirs Tel: (1) 347-4982 P 0 Boo 1305 Tel: (55 11) 259-6644 UNO-Vertag Tel A"e 61130 PO. Boa 202 S-171 25 Solna ZIMBABWE Poppelsodorer Allee 55 Tel. (972 3) 5461423 7480 AS Htaksbatrgen ROMANIA Tel. (48 8) 705-97-50 Longrrr Zonbabwee (Pte (Ltd CANADA 53115 Boon Tel. (31 53) 574-0004 Coropani Ce Librrodi Bucuresli S.A. Tourte Road, Aidbennie Renout Publishing Co Ltd. Tot: (49 228( 212948 Palestorian Aulthority/Middle East SIr. Lipscarni ron 26, sedtor 3 SWITZERLAND PO. Boo ST125 5369 Canotek Road ondes IntomoratbDn Services NEW ZEALAND Bucharest LibraiAe Payot Southerton Ottawa, Ontano KidJ 9,13 GREECE RO.B. 1950 Jerusalern EBSCO NZ Lid. Tot, (40 1) 613 9845 Servie institutionnel Harare Tel. (613) 745-2665 Papasotiroau S.A. Tel: (972 2) 6271219 Private Mail Bag 99014 C6tes-de-Montherono 30 Tel, (263 4) 6216617 35, Stouornra Sir New Market RUSSIAN FEDERATION 1002 Lausannee CHINA 100 82 Atheos ITALY Aucklanrd Isdatelstvo - Tel: (41 21( 341-3229 China FIanacial & Ecourwomi Tet (30 1( 384-1826 LioosasCornsissionrrn SansoniSPA Tel: (849) 524-8119 9a, Loipachniy Pereulok Publishing Hoause Via Coca Di Caiaboa, 111 Moscow 101831 ADECO f, DCieran Eciboass Tf ueooas 8, Ca Fo Si Cong die HAITI Casella PostaJv 552 NIGERIA Tel. V7095) 91178749 Ch. de Lacueez 41 Beijing Cufture CDtfuoion 50125 Firenoze University Press Lemked CHt807 Blonay Tel: (88 10) 6333-8257 5, Rue Capois Telt (55) 645-415 Three Crowns Building deficho Tet: (41 21) 943 2673 CRP257 Peivote Mart Bag 5095 Port-ao-Proc Ibadain Wot 'iVsIA 1 e-i9 Tel, 934 "'i 41-1356 Coming in the next issue of THE WORLD BANK ECONOMIC REVIEW May 1997 Volume 11, Number 2 * Financial Market Fragmentation and Reforms in Four African Countries by Ernest Aryeetey, Hemamala Hettige, Machiko Nissanke, and William Steel * Managing Price Risk in the Pakistan Wheat Market by Jonathan R. Coleman, Rashid Faruqee, and Tom Scott * Civil Liberties, Democracy, and the Performance of Government Projects by Jonathan Isham, Daniel Kaufmann, and Lant H. Pritchett * Explaining Industrial Growth in Coastal China: Economic Reforms... and What Else? by Ashok Mody and Fang-Yi Wang o Organized Labor and the Political Economy of Product Market Distortions by Martin Rama * What Can New Survey Data Tell Us about Recent Changes in Distribution and Poverty? by Martin Ravallion and Shaohua Chen * The Relative Efficiency and Implementation Costs of Alternative Methods for Pricing Irrigation Water by Yaceov Tsur and Ariel Dinar THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. Telephone: (202) 477-1234 Facsimile: (202) 477-6391 Telex: MCI 64145 WORLDBANK or MCI 248423 WORLDBANK Cable Address: INTBAFRAD WASHINGTONDC World Wide Web: http://www.worldbank.org/ E-mail: books@worldbank.org H'*I II 00 ISBN 0-8213-3781-5