Distributional Effects of Globalization in Developing ...noy/362texts/inequality.pdf · 1....

Journal of Economic LiteratureVol. XLV (March 2007), pp. 39–82

mar07_Article2 3/12/07 5:43 PM Page 39

Distributional Effects of Globalizationin Developing Countries

PINELOPI KOUJIANOU GOLDBERG AND NINA PAVCNIK∗

The authors discuss recent empirical research on how globalization has affectedincome inequality in developing countries. They begin with a discussion of conceptu-al issues regarding the measurement of globalization and inequality. Next, they pres-ent empirical evidence on the evolution of globalization and inequality in severaldeveloping countries during the 1980s and 1990s. The authors then examine the chan-nels through which globalization may have affected inequality, discussing theory andevidence in parallel. They conclude with directions for future research.

1. Introduction

One of the few uncontroversial insights oftrade theory is that changes in a coun-

try’s exposure to international trade, andworld markets more generally, affect the dis-tribution of resources within the country andcan generate substantial distributional con-flict. Hence, it comes as no surprise that theentry of many developing countries into theworld market in the last three decades coin-cides with changes in various measures ofinequality in these countries. What is moresurprising is that the distributional changeswent in the opposite direction from the onesuggested by conventional wisdom: whileglobalization was expected to help the lessskilled who are presumed to be the locallyrelatively abundant factor in developingcountries, there is overwhelming evidence

∗Goldberg: Yale University. Pavcnik: Dartmouth. Wethank the editor, Roger Gordon, and two anonymous ref-erees for many helpful comments. This research is sup-ported by funding from the National Science Foundation,Grant SES #0213459.

39

that these are generally not better off, at leastnot relative to workers with higher skill oreducation levels. What explains this apparentparadox? Is the theory underlying the con-ventional wisdom too stylized to capture thereality of the developing world? Or werethere other forces at work that may haveoverridden the effects of globalization? Whatare the mechanisms through which globaliza-tion affected inequality? Did the experiencevary across countries and, if so, why? Whatare the general lessons we can draw from theexperience of the last three decades? It isthese and other related questions that thisarticle aims to address.

To this end, we present a large amount ofevidence from several developing countriesregarding their exposure to globalizationand the parallel evolution of inequality.While the evidence is subject to severalmeasurement problems that we discussextensively in this article, two trends emergeclearly from the data analysis. First, theexposure of developing countries to interna-tional markets as measured by the degree of

40 Journal of Economic Literature, Vol. XLV (March 2007)


trade protection, the share of imports and/orexports in gross domestic product (GDP), themagnitude of capital flows—foreign directinvestment (FDI) in particular, and exchangerate fluctuations has increased substantiallyin recent years. Second, while inequality hasmany different dimensions, all existing meas-ures for inequality in developing countriesseem to point to an increase in inequality,which in some cases (e.g., pre-NAFTAMexico, Argentina in the 1990s) is severe.

We next investigate the question whetherwe can establish a causal link between theincrease in inequality and globalization. Weexamine several mechanisms through whichopenness is presumed to have affectedinequality and discuss related evidence. Ouranalysis here draws on several empiricalstudies of globalization and inequality indeveloping countries as well as existing sur-veys of related topics (Ann E. Harrison andGordon H. Hanson 1999; Adrian Wood1999; Pinelopi Koujianou Goldberg andNina Pavcnik 2004). We confine our discus-sion to the experience of developing coun-tries in the last two to three decades. Theprimary reason for this focus is that meas-ures of inequality are typically computedbased on household survey data, and suchdata did not become available until the late1970s in many developing countries. In gen-eral, the data have become more reliableover time, so that studies focusing on morerecent years tend to produce more credibleresults. The second reason we focus on thelast three decades is that, during that peri-od, many developing countries underwentsignificant trade liberalization that substan-tially increased their exposure to interna-tional markets. We argue that for manycountries, most notably Latin Americancountries in the 1980s and early 1990s andIndia in the early 1990s, trade liberalizationepisodes represent a major part of theirglobalization. Furthermore, we argue thatthe trade barrier reductions that occurredduring this period can be exploited to estab-lish a causal link between trade openness

and changes in inequality. By the mid-1990s, the economic landscape had howeverchanged, and factors other than trade liber-alization, such as increased capital flows,FDI, exposure to exchange rate fluctuationsthat in turn affected exports, immigration,etc., became increasingly more importantaspects of these countries’ integration in theworld market. Establishing a connectionbetween these phenomena and inequality ismore challenging compared to the case oftrade barrier reductions, but we discussthese aspects of globalization when relatedevidence is available.

From a methodological point of view, weexplore a variety of possible approaches toidentify the impact of globalization oninequality. A common theme across thestudies we draw upon is that they focusalmost exclusively on the experience of par-ticular developing countries within a rela-tively short time span. While our survey hasa clear comparative aspect as we rely on evi-dence from a large set of countries, weabstain from relying on cross-country regres-sions to econometrically identify the effectsof trade policy changes or conducting com-parisons of inequality measures over longertime horizons. This focus is primary dictatedby data constraints. Inconsistencies in themeasurement of inequality across countries,changes in the household survey responserates over time as incomes rise, and frequentchanges in the design of household surveyswithin the same country make inferencebased on cross-country evidence, or compar-isons of inequality measures over longerperiods of time within a specific country,potentially less reliable compared to infer-ence that relies on within-country evidenceover shorter periods of time. To delineatethe scope of this study, we should also pointout that we focus our discussion on inequal-ity alone and not poverty, as the latter is dis-cussed extensively in a recent article in thisjournal by L. Alan Winters, Neil McCulloch,and Andrew McKay (2004). Finally, weabstract from effects of globalization on

41Goldberg and Pavcnik: Distributional Effects of Golbalization


inequality that may have occurred throughthe growth channel since the evidence onthe causal link between trade openness andgrowth has been controversial and inconclu-sive to date. However, this channel is poten-tially important; the perhaps most significantbenefit of globalization is presumed to bethat it fosters economic growth, and growthitself brings about distributional changes.

Regarding our conclusions, we identify sev-eral channels that may explain why the recentexperience of developing countries did notconform to the “naive” thinking about global-ization. We argue that our understanding ofthe consequences of globalization for inequal-ity has improved as the theoretical frameworkunderlying the empirical work expanded toinclude trade in intermediate products, inter-national flows of capital, trade-induced skilledbiased technological change, short-run factorimmobility, and firm heterogeneity. We alsofind that the effect of globalization oninequality depends on many factors, severalof which are country and time specific,including a country’s trade protection patternprior to liberalization; the particular form ofliberalization and sectors it affected; the flex-ibility of domestic markets in adjusting tochanges in the economic environment, in par-ticular the degree of within-country labor andcapital mobility; and the existence of otherconcurrent trends (e.g., skill-biased techno-logical change) that may have interacted withor even partially been induced by globaliza-tion. Given that different countries experi-enced globalization in different ways and atdifferent times, it is hardly surprising that therelevant mechanisms through which inequal-ity was affected are case specific. From a pol-icy point of view, this implies that attempts toalleviate the potentially adverse distributionaleffects of globalization in the short or medi-um run need to be grounded in a carefulstudy of the nature of globalization and theindividual circumstances in each country.

The remainder of this article is organizedas follows. In section 2, we review somebasic conceptual issues regarding the

measurement of globalization and inequalityrespectively. In section 3, we present empiri-cal evidence on the evolution of globalizationand inequality in developing countries andidentify the main facts and trends thatdemand explanation. Section 4 discusses themethodological challenges one faces inattempts to causally link globalization toinequality. Section 5, the core section of thepaper, examines the channels through whichglobalization might affect inequality by pre-senting theory and evidence in parallel. Westart by focusing on the narrowest measure ofinequality—the wage gap between skilled andunskilled workers (or skill premium)—andinvestigate the main globalization-relatedexplanations for its documented increase. Wethen progressively move to discuss theimpact of openness on broader concepts ofinequality. Section 6 concludes.

2. Conceptual Issues

2.1 Measuring Globalization

Globalization is a broad concept casuallyused to describe a variety of phenomena thatreflect increased economic interdependenceof countries. Such phenomena include flowsof goods and services across borders, reduc-tions in policy and transport barriers totrade, international capital flows, multina-tional activity, foreign direct investment,outsourcing, increased exposure to exchangerate volatility, and immigration. These move-ments of goods, services, capital, firms, andpeople are believed to contribute to thespread of technology, knowledge, culture,and information across borders. Research onthe effects of globalization in economics hasconcentrated on those aspects of globaliza-tion that are easier to capture empirically.Accordingly, we confine our discussion onthe more narrowly defined components ofglobalization: trade liberalization, outsourc-ing, flows of capital across borders in theform of FDI, and exchange rate shocks.

Even when one hones in on a narrowdimension of globalization, measurement



challenges abound. The first hurdle is dataavailability. Detailed information on tradebarriers, outsourcing, or foreign directinvestment is often not readily available,especially when the analysis requires highlydisaggregate data or longer periods of timethat span periods of policy liberalization. Forexample, in their recent survey of tradecosts, James E. Anderson and Eric vanWincoop (2004) note that data on trade pol-icy barriers from UNCTAD’s TRAINS database that is systematically available for alarge set of countries only covers years from1989 onwards. In addition, in a given year, atmost 17 percent of the included countriesreport both tariff and nontariff barriers totrade and trade flows. The lack of reportingis especially pronounced in developingcounties. Consequently, researchers haveoften measured trade liberalization indirect-ly through more readily available data ontrade volumes (i.e., exports and imports).One problem with this approach is that tradevolumes are determined not only by (plausi-bly exogenous) changes in trade policy andtransportation barriers, but also by endoge-nous variables, some of which are in fact thefocus of interest in the globalization andinequality debate (i.e., wages). As a result,more recent studies have mainly relied onnational data sources to obtain trade policyinformation, as well as information on FDIand outsourcing, spanning periods of policyreforms.

Trade liberalization episodes, and in par-ticular reductions in tariff barriers, are per-haps the most commonly studiedcomponent of globalization. This focus isdetermined by practical considerations: tar-iffs are relatively easier to measure thanother forms of globalization. Because tariffsare usually imposed as ad valorem taxes onimported goods, they represent price basedforms of trade protection. As such, they aretransparent, relatively easier to measureconsistently across industries and over time,and their magnitude reflects the truerestrictiveness of the trade barrier. Of

course, tariffs are not the only policy instru-ment through which governments in devel-oping countries regulate imports. Importsinto developing countries are also subject tonontariff barriers to trade (NTBs) such asimport licenses and quotas. The informationon NTBs is often not available or not avail-able at the same level of product/industryaggregation as tariffs, especially for longertime periods surrounding trade liberaliza-tion episodes. Moreover, because manyNTBs are forms of protection that limit thequantity of imports allowed to enter a coun-try (rather than price-based measures), theyare more difficult to accurately measure.Researchers usually capture the extent ofNTBs at some level of industry aggregationby a nontariff barrier coverage ratio, whichmeasures the share of products (or totalimports, or national production) in an indus-try aggregate that is subject to NTBs. Thismeasure however does not capture the truerestrictiveness of NTBs: for example, a cer-tain industry may have the same NTB cover-age ratio in two different years, yet the NTBcould be more or less restrictive in one ofthe years because of different demand con-ditions. As a result, measurement problemsare more severe in the case of NTBs andtheir comparability across countries, indus-tries, and time is more of an issue than in thecase of tariffs. While the omission of reliableNTB measures and their changes in empiri-cal studies is a potentially serious limitation,a somewhat encouraging result is that corre-lations between tariff rates and NTB cover-age ratios (and their changes), wheneveravailable, are positive, indicating that tariffsand NTBs have been used in recent years indeveloping countries as complements andnot substitutes.1 In terms of the interpreta-tion of empirical results, such correlationsimply that the effect that is typically attrib-uted to a tariff change represents an over-estimate of the pure tariff effect as it reflects

1 See Goldberg and Pavcnik (2005), p. 89–90, for a dis-cussion of this correlation for the case of Colombia.



the combined effect of the tariff and NTBchange.

Even if one limits the analysis to tariffs,measurement concerns remain. One of themost significant ones is aggregation.National governments set tariffs at a verydisaggregate level as detailed tariff lines.Researchers however typically need toaggregate these tariffs to a higher level tomatch the level of industry aggregation atwhich the outcome of interest, such as wagesor employment, is reported. This requiresthe use of concordances between tariff linesand industries that are notoriously noisy, sothat aggregate industry tariffs are plausiblymeasured with error. In addition, aggrega-tion discards some potentially importantvariation in tariffs (or tariff changes) withinindustry groups and thus precludes theresearcher from examining some channelsthrough which individuals/firms adjust totrade liberalization within broadly definedindustries.

A further concern is that industry tariffson final goods do not capture the true extentto which an industry is affected by protec-tion (or liberalization) since they do notaccount for intermediate good linkages. Onecould, in principle, capture such linkages byconstructing effective rates of protection,which take into account not only the directprotection granted to an industry throughnominal tariffs on final products but also theindirect one that results from tariffs on inter-mediate inputs. Unfortunately, effectiverates of protection are not readily availablefor many countries over periods that spantrade liberalization episodes. In addition,effective rates of protection present addi-tional measurement/concordance problemsstemming from the use of information fromthe input output tables required in theirconstruction. Fortunately, in cases whereboth nominal and effective measures of pro-tection are available, they tend to be highlycorrelated. For example, the correlationbetween industry effective rates of protec-tion and industry tariffs in Colombia is above

0.9 in years where both of these measuresare available (Goldberg and Pavcnik 2005).

Naturally, the focus on trade policy instudying the effects of “globalization” oninequality is only useful to the extent thattrade policy is an important component of acountry’s exposure to globalization. This wasthe case in many of the countries that we dis-cuss in this article, namely Latin Americancountries such as Brazil, and Colombia, andMexico during the late 1980s/early 1990sand India during the 1990s. In other settings,most notably Mexico after the implementa-tion of NAFTA, channels other than tradepolicy, for example, immigration, foreigndirect investment, outsourcing, and the pesocrisis, have played a potentially more impor-tant role. Still, average tariff rates continueto be high in many developing countries,including some that have recently imple-mented trade reforms. India provides themost striking example. Although Indiaunderwent a drastic trade liberalizationreform starting in 1991, the average tariff inmanufacturing was over 30 percent in 1999(Petia Topalova 2004a). Thus, there remainssubstantial scope for further tariff and NTBreductions and trade policy is likely to con-tinue to be an important component of glob-alization at least in some of the lower incomedeveloping countries.

In addition to the role of trade reforms infostering trade in final goods, recent work byRobert C. Feenstra and Hanson (1996,1997, 2003) has emphasized the growingimportance of trade in intermediate inputs.This phenomenon is also referred to as “out-sourcing” or “production sharing.” Recenttrade liberalizations, coupled with theremoval of restrictions on capital flows andtechnological change, have enabled firms to“outsource” some stages of production tocost-minimizing locations abroad, eitherthrough arm’s length imports of intermedi-ate inputs or by setting up their own produc-tion facilities in a host country. A country canbe exposed to outsourcing as a purchaser ofoutsourcing activities (for example, firms in



Hong Kong have been importing relativelylabor-intensive intermediate products fromChina since the 1980s) or as a host of out-sourcing activities (for example, Mexico’smaquiladoras have been used to assembleintermediate products into final goods madefor U.S. markets since the early 1980s).

In empirical work, one would ideally like torely on a measure of exposure to outsourcingthat is related to plausibly exogenous changesin trade and capital controls. From the receiv-ing country’s perspective, this is subject to thesame data constraints we discussed in the con-text of effective rates of protection.Consequently, the literature has mainly usedthe share of imported inputs in total pur-chased intermediate inputs in an industry as ameasure of outsourcing (see Feenstra andHanson 2003; Chang-Tai Hsieh and Keong T.Woo 2005). Because direct data on importedinputs by industry are often not available, theabove outsourcing measure is constructed bycombining information from input–outputtables with information on total trade flows offinal products. As a result, it is subject to thesame endogeneity concerns as trade flows.Furthermore, this measure of outsourcingsuffers from the same measurement problemswe discussed earlier in the context of tariffsregarding the concordances between tradedata, industry data, and input–output tables.

Related to “global production sharing” isthe presence of multinational firms and for-eign direct investment in developing coun-tries. Their increased presence stems in partfrom the recent removal of controls on capitalflows in these economies. The information onaffiliates of multinational companies in devel-oping countries is usually obtained fromnational surveys of firms, such as the Censusof Manufacturers. In some countries, forexample Indonesia and Mexico, these surveysprovide information on the nationality of thecapital sources so that one can identifywhether a particular firm is partly foreign-owned. These surveys are also used to createmeasures of the presence of multinationals inan industry or region. Such measures usually

capture the intensity of multinational activityby computing the share of foreign affiliates intotal industry employment or output to cap-ture horizontal linkages, or by additionallyapplying input–output tables to this informa-tion to capture an industry’s exposure to FDIthrough vertical linkages. One concern withthis measurement approach, raised recentlyby Wolfgang Keller and Stephen R. Yeaple(2003) in the context of the United States, isthat measures of an industry’s exposure toFDI are highly sensitive to how the economicactivity of a foreign affiliate is allocated acrossthe various industries in which it is active (forexample, main line of business versus otherlines). Another more general concern withthese measures of FDI is that the decision ofa multinational to purchase an existing plantor to locate in a country/industry maydepend on unobserved wage and workercharacteristics in a firm/industry/region,which creates the potential for simultaneityand selection bias.

Finally, the removal of capital controls com-bined with a shift away from fixed and towardmore flexible exchange rate arrangements inmany developing countries has exposed thesecountries to greater exchange rate volatility.To the extent that these exchange ratechanges are partially passed through ontoprices, the increased exposure to exchangerate volatility impacts firms’ incentives toexport (or import) and, hence, presents anoth-er channel through which globalization mayhave affected inequality.2 The advantage ofusing exchange rate shocks as a measure ofglobalization is that they are easy to measure,plausibly exogenous—at least from a singleindustry’s perspective, and large in magnitude.

2 The evidence suggests that, for exchange rate shocksof the magnitude recently witnessed in several LatinAmerican and Asian economies (for example, Mexico in1994; Brazil in 1998; and Thailand, Korea, Indonesia, etc.in 1997), the price effects are larger than what is typicallyobserved for more modest shocks. In these cases, theexchange rate shocks should be more accurately charac-terized as currency crises. The fallout from currency criseshas potentially its own implications for inequality, but wedo not explore these implications in this study.



The disadvantage is that they represent aggre-gate shocks to an economy; they do not exhib-it any variation across industries or plants, sothat separating their effect from the one ofother concurrent macroeconomic shocks orpolicies can be challenging.

2.2 Measuring Inequality

The ideal measure of inequality would bebased on comparisons of individuals’ well-being over their entire lifetime. The mostappropriate variable for capturing lifetimewell-being is arguably consumption (seeAngus Deaton 1997 for a related discussion).Compared to income, consumption offersthree advantages. First, to the extent thatconsumers can intertemporally shiftresources through lending and borrowing,current consumption better captures life-time well-being. This argument may be lessrelevant for developing economies charac-terized by severe capital market imperfec-tions, yet the evidence suggests that, even inthese countries, some borrowing and lend-ing does take place (though this may occur ininformal credit markets and at exceedinglyhigh interest rates). Second, reporting prob-lems are less pronounced for consumptionthan income. Specifically, it is well docu-mented that high-income households tendto underreport their income (but not neces-sarily their consumption), while most sur-veys collect data on pretax, and not after-taxincome. Finally, many policies—trade poli-cies in particular—affect the relative pricesof consumer goods so that they impact con-sumers not only through income changes,but also through changes in the purchasingpower of their current incomes. Inequalitymeasures based on consumption data are bynature better suited to capture this effect.

Despite these advantages, consumption israrely used as the basis for measuring inequal-ity in empirical studies of the effects of glob-alization.3 The reason is that many developing

3 Guido G. Porto (2006), Alessandro Nicita (2004), andTopalova (2004a) are notable exceptions.

countries do not consistently report expendi-tures in their household surveys. The LivingStandards Measurement Surveys project ofthe World Bank aims at changing this pattern,so that research in future years may be able totake advantage of expenditure data to meas-ure inequality. To date however, most empiri-cal studies had to contend themselves withemploying income based measures ofinequality, given that some measure of incomeis always included in household surveys. Themost frequently used inequality indices (suchas the Gini coefficient or the coefficient ofvariation) are based on the second momentsof the observed income distribution.

The suitability of these indices for captur-ing true changes in inequality, especially overlonger periods of time, has been questionedrecently for a variety of reasons. First, eventhough most household surveys include somemeasure of income, the coverage of incomesources and taxes tends to vary both acrosscountries and, for a specific country, acrossyears; items such as in-kind gifts and govern-ment transfers, implicit rent from own hous-ing, and capital income and profits tend to beparticularly problematic. To avoid these prob-lems, many studies have focused on a morenarrow measure of inequality—wage inequal-ity. A second set of problems is related to thefact that high-income households are knownto have higher nonresponse rates and under-report income, so that the income distributionpresented in household surveys is a truncatedversion of the true one.4 Johan Mistiaen andMartin Ravallion (2003) and Deaton (2003)have shown that, with nonresponse ratesincreasing in income, it is possible that thevariance of the truncated distribution is lowerthan the variance of the true distribution. Inthe context of inequality measurement, thiscounterintuitive result implies that indices

4 Miguel Szekely and Marianne Hilgert (1999), forexample, report that in many Latin American householdsurveys the top ten incomes reported in a given year areabout the same as the salary of an average manager in thecountry under consideration. This suggests that the trulyrich households are missing from the surveys.



based on the second moments of the observed(truncated) income distribution may be mis-leading about changes in inequality; this isespecially the case if the comparisons involvelong periods of time during which income hassubstantially increased. On a similar note,Abhijit Banerjee and Thomas Piketty (2004,2005) document that income data based onIndian tax returns (where underreporting ispresumably less of an issue compared tohousehold survey data) indicate that the “veryrich” in India, i.e., those who were in the top0.1 percent of the population, were gettingricher faster than anyone else in the 1990s.This group seems to be missing from theIndian household survey (National SampleSurvey).5 Though tax return data provide asuperior source of information for the pur-pose of documenting income inequality, theyhave not been used in studies of the causesbehind changing trends in inequality since theconfidential nature of the data preventsresearchers from linking the income figures toother individual-specific variables of interest.

Another potential problem in inequalitystudies is that household surveys are oftenredesigned so that the wage or income dataare not comparable across years. Changes intopcoding limits, for example, can affect therange of top incomes reported in the surveys.6

In addition to these reporting problems, allinequality studies face the conceptual issue ofwhether to focus on households or individu-als. While the primary interest lies in the well-being of individuals, people usually live inhouseholds and share resources. To take thisinto account, many studies have focused onsome variant of per capita income. The sim-plest one is obtained by dividing householdincome by family size; more sophisticated

5 However, the authors point out that this group is toosmall for its absence to explain important discrepancies inthe measurement of inequality and poverty based on NSSand the national accounts data.

6 This was, for example, the case in Colombia, wherea change in the topcoding procedures used in theEncuesta Nacional de Hogares (National HouseholdSurvey) in 1994 affected the reported incomes of thericher households.

measures take into account consumptionscale economies within the household anddifferences in the needs among individuals ofdifferent gender and age to construct scale-and adult equivalent-adjusted versions of percapita income (see Deaton 1997). The prob-lem with such adjustments is that the con-structed index of well-being will ultimatelydepend on the scale and adult-equivalencyparameters, which may be poorly known.

Given the conceptual and measurementambiguities involved in measuring inequali-ty, cross-country comparisons of inequalityfigures or investigations of long-term trendsin developing countries appear problematic.Studies of the effects of trade openness oninequality have traditionally been narrowerin focus, as the majority of them have ana-lyzed concrete trade liberalization episodesor other policy changes in specific countries.Because most of these episodes unfoldedover the course of a few (2–3) years and therelated studies focus on one country at atime, many of the aforementioned measure-ment problems are less pronounced here.Furthermore, the increase in inequality doc-umented in many developing countries hasbeen associated with an increase in the so-called skill premium, i.e., the wage gapbetween skilled and unskilled workers.Motivated by this finding, a substantialamount of related work has focused on aneven more narrow measure of inequalitythan the ones discussed above: the inequalitybetween skilled and unskilled workers.

The definition of skill varies depending onthe kind of data employed. Studies that usehousehold survey or labor force survey datadefine skill based on the education of thehousehold head. Studies that exploit plant- orfirm-level data typically differentiatebetween production and nonproduction orblue-collar and white-collar workers. This lat-ter categorization is clearly unsatisfactory,especially since the skill composition of thesegroups is likely to vary over time. For manycountries however, plant-level data are morereadily available over several years; moreover,



they offer the advantage of providing infor-mation about the sector of employment at amore disaggregate level compared to house-hold surveys that in many developing coun-tries report industry information only at thetwo-digit level. Fortunately, cross-tabulationsof matched worker and employer surveys atthe plant level in the United States and theUnited Kingdom indicate a close relationshipbetween the production/nonproduction sta-tus of workers and their educational level;7

nonproduction workers have more years ofschooling and appear to be uniformly betterpaid. Although there is no direct evidence onthis issue for developing countries, these cor-relations are encouraging regarding the suit-ability of plant-level data for analyzing thedifferential impact of globalization on work-ers of different skill level. As with the incomeor wage based measures of inequality, com-parisons over short periods of time within acountry are likely to be more credible thancross-country comparisons or analyses oflong time trends.

3. Overview of the EvidenceDespite the difficulties associated with the

measurement of globalization and inequali-ty, research in the past fifteen years has triedto document their evolution by increasinglyrelying on new and better data sources. Inthis section, we summarize the existing evi-dence, focusing on the experience of a fewrepresentative countries (Mexico, Colombia,Argentina, Brazil, Chile, India, and HongKong) during the 1980s and 1990s.8 Our

7 See Eli Berman, John Bound, and Stephen Machin(1997) and Machin, Annette Ryan, and John Van Reenen(1996).

8 One obvious omission is the set of Southeast Asiancountries (South Korea, Taiwan, and Singapore) thatunderwent trade reforms in the 1960s and 1970s.Unfortunately, neither detailed data on tariffs nor microsurveys are readily available for these countries. The exist-ing evidence on these countries has been discussed in detailin Wood (1999). We also leave out China because empiri-cal work on the relationship between inequality and global-ization has just recently started to emerge for this country.Shang-Jin Wei and Yi Wu (2002) is to our knowledge theonly study that examines the link between openness andinequality in China.

choice of time periods and countries is dic-tated by the timing of trade reforms and dataconstraints. With few exceptions (Chile forexample), most developing countries did notliberalize their trade regimes and did notopen their borders to foreign direct invest-ment until the 1980s. The countries dis-cussed in this section are representative inthat sense since they all experienced drastictrade liberalization during the past twodecades. Furthermore, they all collect thedetailed micro data required to generate var-ious measures of inequality that span theperiod before, during, and after policychanges that increasingly exposed thesecountries to international markets.Consequently, these countries have servedas a testing ground for most empiricalresearch investigating the channels throughwhich globalization may have affectedinequality.

3.1 Globalization

Table 1 provides an overview of the glob-alization experience of the countries men-tioned above (changes in trade policy andother relevant measures of globalization)along with the reported changes in inequali-ty measures. The same table also lists othermajor reforms that took place during the1980s and 1990s in each of these countries.

Let us first focus on changes in globaliza-tion measures, starting with trade liberaliza-tion episodes. Table 1 indicates that,although some countries (i.e., Argentina andColombia) experimented with short-livedtrade reforms during the late 1970s, mostcountries implemented unilateral tradereforms in the mid to late 1980s and early1990s: Mexico 1985–87, Colombia 1985–91,Argentina 1989–93, Brazil 1988–94, India1991–94. Chile is an exception as it liberal-ized its trade regime early, from 1974 to1979.

An important feature of the above reformswas that they drastically reduced tariffs,which were high prior to liberalization and acrucial component of trade protection. The



TABLE 1GLOBALIZATION AND INEQUALITY IN SELECT DEVELOPING COUNTRIES

1970s 1980s 1990sMEXICO

Globalization Unilateral trade NAFTA (1994)Measures liberalization 1985–87 Peso Crisis

(WTO entry) Maquiladoras expansionDevaluation FDIMaquiladoras Immigrationliberalization (1983)FDI liberalization (1989)Immigration

InequalitySkill premium Increased Increased until mid-1990s

Stable/declined after mid-1990sIncreased between 2000–1990

Wage white collar/ Declined 1965– Increased Increased until mid-1990sWage blue collar 80 Stable after mid-1990s90-10 log N.A. Increased Increased up to 1996wage differential Stable/decline after mid-1990sGini of log wages Increased Increased up to mid-1990s

Stable/decline after mid-1990sIncome Inequality Declined Increased Stable/decline(Gini)

Other Reforms Privatization Banking CrisisLabor Market ReformDeregulation

COLOMBIA

Globalization Partial Trade Reform Gradual trade liberalization Trade liberalization 1990–91Measures starting 1979 starting 1985 Devaluation

Inequality (urban)Skill Premium Slightly Declined Increased90–10 log wage Slightly Declined Increaseddifferential 1986–90Gini of log wages Stable/ Slight Decline IncreasedIncome Inequality Declined Stable/Increased Stable(Gini)

Other Reforms Labor market reform 1990Banking reform 1993

ARGENTINA

Globalization Short Trade Reform Unilateral Trade Trade liberalization cont.Measures (1976–82) Liberalization (1989–93) Mercosur 1991

Appreciation Appreciation

Inequality (urban)Skill Premium Decreased IncreasedGini of log wages Increased IncreasedIncome Inequality Increased Increased Increased

Other Reforms Macroeconomic crisis Deregulation(1988–89) PrivatizationPrivatization Financial liberalizationDeregulation in early 1990sFinancial Liberalization Convertibility Planin the late 1980s



(continued on next page)

TABLE 1 (continued)

1970s 1980s 1990sBRAZIL

Globalization Partial unilateral trade Unilateral trade liberalizationMeasures liberalization (1988 onwards) (ends 1994)

Mercosur 1991Currency Crisis 1998

Inequality (national)Skill Premium N.A. Stable/Slight Increase IncreasedMean log deviation N.A. Stable/Increased Stableof wageGini of log wages Stable Stable/Small declineIncome Inequality Stable Increased Stable/Small decline

Other Reforms Labor market reform

CHILE

Globalization Trade DevaluationMeasures Liberalization

InequalitySkill Premium Increased Increased Declined early 1990s

Overall increased 1990–2000(national data)

Wage white collar Increased/Wage blue collar

Gini of log wages Increased Increased Decreased relative to late 1980sStable during the 1990s

Income Inequality Increased Increased Stable/Small increase late 1990s(national)

Other Reforms Structural Reforms DevaluationPrivatization Macroeconomic crisisDeregulationTex ReformLabor Market Reform

INDIA

Globalization Limited Removal of Trade Liberalization 1991Measures Import Licenses Unilateral FDI liberalization

Inequality (urban)Skill Premium Relatively stable Increased90-10 log wage Increased Increased more rapidlydifferentialIncome Inequality IncreasedConsumption Stable/Slight Increase Increasedinequality

Other Reforms Industrial delicensing Tax ReformFinancial Reform

HONG KONG

Globalization Outsourcing to China Outsourcing to ChinaMeasures

InequalitySkill Premium Slight decline Increased Increased(return to education)Wage non- Declined Increased Increasedproduction/Wageproduction workers



TABLE 1 (continued)

Notes: General sources on Latin American countries: Schady and Paramo-Sanchez (2003), Gasparini (2003),Robbins (1996). Sources for information on specific countries are: Argentina: Gasparini (2003); Brazil: Green,Dickerson, and Arbache (2001); Chile: Francisco H. G. Ferreira and Julie A. Litchfield (1999), Robbins (1996),Beyer, Rojas, and Vergara (1999), Pavcnik (2003); Mexico: Robertson (2004), Hanson (2004), Cragg andEplebaum (1996); India: Kijima (2006), Topalova (2004a), Deaton and Dreze (2002); Colombia: Attanasio,Goldberg, and Pavcnik (2004); India: Kijima (2006), Topalova (2004a), Deaton and Dreze (2002); Hong Kong:Hsieh and Woo (2005).

high tariff rates reflect the lack of participa-tion of most developing countries in the tar-iff-reducing rounds of the GATT/WTOprior to their unilateral trade reforms: somedeveloping countries were not GATT mem-bers (for example, Mexico); others (such asBrazil, Colombia, and India) were GATTmembers on paper but did not have toreciprocate tariff concessions negotiatedwith the GATT until the Uruguay Round.9

Table 2 reports the average tariffs for themanufacturing industries in the countries oftable 1 in a year before and after thereforms.10 The table illustrates that, prior tothe reforms, tariff levels were high, rangingfrom 117 percent in India to 23.5 percent inMexico. The comparison of average tariffsbefore and after the reforms suggests dras-tic tariff reductions: for example, 85 per-centage points in Chile, 73 percentagepoints in India, and 12.5 percentage pointsin Mexico. These tariff declines in develop-ing countries are in stark contrast to the lowtariff levels and rather minor tariff policychanges in the developed countries duringthis period. For example, in the UnitesStates—a country whose tariff policy resem-bles the policy of most other developedeconomies—the average tariff was only 4.8percent in 1982; tariffs declined on averageby 0.6 percentage points to 4.2 percentbetween 1982 and 1992 (Andrew B.

9 Article XVIII of the GATT granted exemption fromtariff concessions to developing countries.

10 For each country, the actual year used to describethe period before and after the reforms is recorded belowthe country name in column 1.

Bernard, J. Bradford Jensen, Peter K.Schott 2005).

In addition to tariff reductions, the unilat-eral trade reforms also reduced NTBs.Unfortunately, as discussed earlier, the infor-mation on exact measures of NTBs is oftennot available, especially for longer periodssurrounding trade liberalization episodes.However, the available data on average NTBcoverage ratios in manufacturing industriesbefore and after the reforms (presented incolumns 3 and 4 of table 2) suggest thatNTBs were high prior to trade reforms andthat liberalization drastically reduced theirlevels. For example, in Colombia the NTBcoverage ratio declined from 72.2 percent in1986 to 1.1 percent in 1992. In Mexico, theshare of manufacturing production subjectto import licenses dropped from 92 percentin 1985 to 23.2 percent in 1988. In India, theshare of manufacturing imports covered bynontariff barriers dropped from 80 percentin 1990 to 17 percent in 1999 (Prachi Mishraand Utsav Kumar 2005). Although we do nothave access to measures of NTBs in othercountries, NTBs were virtually eliminated inChile (Rudiger Dornbusch and SebastianEdwards 1994) and Brazil (Donald A. Hay2001), while Argentina eliminated all importlicenses (Sebastian Galiani and PabloSanguinetti 2003).

Table 1 suggests that, subsequent to uni-lateral trade reforms, several countries alsolowered their trade barriers vis a vis specifictrading partners through regional tradeagreements. The most notable example isMexico’s entry into a free trade agreement



TABLE 2GLOBALIZATION IN SELECTED DEVELOPING COUNTRIES

Trade Liberalization Trade Flows (% GDP) FDI inflows (% GDP)

Average Tariff Average NTB Exports ImportsBefore After Before After 1980 2000 1980 2000 1980 2000

Argentina 45 12 n.a. declined 5.1 10.8 6.5 11.4 .88 4.09Brazil 58.8 14.4 n.a. declined 9.1 10.9 11.3 12.1 .81 5.50Chile 105 10 n.a. declined 22.8 31.8 27.0 30.8 .77 5.21Colombia 50 13 72.2 1.1 16.2 21.9 15.6 20.4 .47 2.92Hong Kong n.a. n.a. n.a. n.a. 89.9 150.0 90.8 145.3 n.a. n.a.India 117 39 82 17 6.1 14.0 9.7 16.6 .04 .51Mexico 23.5 11 92 23.2 10.7 31.4 13.0 33.2 .96 2.31

Notes: The following years are specific dates refered above as before and after trade liberalization: Argentina(1988, 1994), Brazil (1987, 1994), Chile (1974, 1979), Colombia (1984, 1992), India (1991, 2000), Mexico (1985,1988). Information on trade liberalization in selected countries is from Galiani and Sanguinetti (2003), Pavcnik etal. (2004), Dornbusch and Edwards (1994), Goldberg and Pavcnik (2005), Mishra and Kumar (2005), andRevenga (1996). World Development Indicators CD is the source of trade flow and FDI data.

with the United States and Canada in 1994(NAFTA). Argentina and Brazil joinedMercosur in 1991, along with Uruguay andParaguay. These regional trade agreementslikely induced changes in the geographiccomposition of trade in these countries; how-ever, the changes in trade policy implied bythese agreements were substantially smallerthan the declines in trade barriers observedduring the unilateral trade reforms.

Furthermore, several countries (mostnotably Mexico and Hong Kong) experi-enced increases in trade in intermediateinputs associated with global productionsharing. For example, after the capital con-trol liberalization in Mexico in the mid-1980s, many U.S. companies shiftedrelatively low-skill intensive stages of pro-duction to Mexico by setting up foreignassembly plants (maquiladoras). Intermediateinputs were imported to Mexico, assem-bled in maquiladoras, and the final prod-ucts exported to the United States. Theimportance of maquiladoras for theMexico–U.S. trade was growing during the1980s and 1990s so that by 2000 maquilado-ras accounted for 35 percent of Mexico’simports from the United States and for 48

percent of its exports to the United States(Hanson 2004). Similarly, when China liber-alized its markets, many firms in Hong Kongshifted their relatively less-skilled-labor-intensive activities to Chinese borderregions, while specializing in higher-skillintensive activities, such as headquarterservices, at home. As a consequence, theshare of intermediate inputs that wereimported from China in Hong Kong’s totalintermediate inputs rose from less than 10percent in 1976 to almost 50 percent in 1996(Hsieh and Woo 2005). A related develop-ment has been the growing presence of affil-iates of multinational companies indeveloping countries during the 1980s and1990s following their capital market reforms.This is illustrated by the increased impor-tance of FDI inflows in the economies ofdeveloping countries. Table 2 reports FDIinflows as a share of GDP in select countriesand illustrates that, while the share of FDI intotal GDP was below 1 percent in 1980s inthese countries, it grew to about 3 percent in2000 for Colombia and Mexico, to 4 percentin Argentina, and 5 percent in Brazil. InIndia, however, it is still about 0.5 percentof GDP.



Finally, table 1 indicates that many devel-oping countries experienced large currencyfluctuations during the 1980s and 1990s. Insome instances, these exchange rate changesmay have exposed the relevant countries tointernational markets more than the tradereforms. Eric A. Verhoogen (2006), forexample, argues that Mexico’s 1994 peso cri-sis, during which the peso lost half of its orig-inal value, overshadowed the average tariffchanges from NAFTA.

3.2 Inequality

The information on inequality is based onempirical studies that have utilized microsurveys of households or firms from thecountry in question. The relevant sourcesare cited in the notes to the table. Table 1reports several measures of inequality: skillpremium, wage inequality, income inequali-ty, and consumption inequality. Note thatbecause of data constraints, some of thesemeasures, most frequently consumptioninequality, are missing for many countries.

We begin by examining the evolution ofthe narrowest measure of inequality: thewage gap between more and less skilledworkers (the so-called skill premium). Wheninformation on an individual’s education isavailable, we use the returns to completeduniversity degree as a measure of the skillpremium and report evidence based on aMincerian regression; when data on the edu-cational attainment of workers are not avail-able, as is the case with plant surveys, we usethe relative wage of white- to blue-collarworkers (or, alternatively, the relative wageof nonproduction to production workers), tomeasure the skill premium. Several broadpatterns emerge.

When we consider the 1980s and 1990s asa whole, all countries seem to have experi-enced increases in the skill premium. Theskill premium increases were largest inMexico, where the return to university edu-cation (relative to primary education)increased by 68 percent between 1987 and1993 (Michael Ian Cragg and Mario

Epelbaum 1996). In other countries, theskill premium increased too, but by less: forexample, the return to a university degreeincreased by 16 percent (relative to primaryeducation) in Colombia between 1986 and1998 (Orazio Attanasio, Goldberg, andPavcnik 2004), by over 20 percent (relativeto no complete education) in Argentinabetween 1992 and 1998 (LeonardoGasparini 2004), by 13 percent in India (rel-ative to primary education) between 1987and 1999 (Yoko Kijima 2006), and by 10 per-cent among men (relative to no completeeducation) in Brazil (Gasparini 2003). Giventhat relatively large skill premium increaseshave been documented for several countries,it is unlikely that they are all a figment of themeasurement problems discussed in section2, although the exact magnitudes of thechanges may be affected by these problems.

A further pattern evident in table 1 is thatthe skill premium does not steadily increasethroughout the two decades in all countries.Interestingly, the skill premium increasesseem to chronologically coincide with thetrade reforms in several countries. Forexample, the skill premium grew steadilyduring the 1980s and 1990s in Mexico,11

which implemented a large trade reform inthe mid-1980s and was continually exposedto other forms of globalization, such as out-sourcing or FDI, for the next two decades.On the other hand, skill premium increasesin Colombia, Brazil, Argentina, and Indiawere mainly confined to the 1990s; the lattercountries implemented the bulk of theirtrade reforms in the early 1990s. In Chile,where the reforms took place during the1970s, the skill premium increased duringthe 1970s and 1980s, declined in the early

11 Most evidence on Mexico points to a rising skill pre-mium, at least until the mid-1990s. Gasparini (2003) andHanson (2004) document skill premium increases overthe entire decade using nationally representative house-hold survey and population census data, respectively.However, Raymond Robertson (2004) argues that the skillpremium declined (or remained relatively stable) after themid-1990s in urban areas.



1990s (Donald J. Robbins 1996; HaraldBeyer, Patricio Rojas, and Rodrigo Vergara1999), and then increased again between1990 and 2000 (Gasparini 2003). Thesetime-series patterns have led many casualobservers to conclude that globalization wasthe main source of growing inequality inthese countries. As we argue in the next sec-tion, inference based on these before andafter comparisons can be misleading.

Finally, note that changes in the educa-tion-based measure of the skill premiumand the relative wage of white-collar toblue-collar workers tend to move in thesame direction in countries and periods forwhich both measures are available. Forexample, in Mexico the average relativewage of nonproduction workers increasedalmost by a factor of 1.5 between 1987 and1995 (Robertson 2000). This parallelmovement is reassuring for studies thatrely on the white-collar/blue-collar distinc-tion (or nonproduction/production workerdistinction) as a measure of skill.

The observed changes in the skill premiumare generally (but not always) reflected inchanges in the wage inequality (usuallymeasured by the Gini coefficient of logwages, or the 90–10 log-wage differential).As with the skill premium, wage inequalityincreased in Mexico12 in the 1980s and earlyto mid-1990s, in Chile during the 1970s and1980s, and in Colombia, Argentina, and Indiaduring the 1990s. Interestingly, increases inthe skill premium are not mirrored inincreases in wage inequality in Brazil, wherethe Gini coefficient remains remarkably sta-ble during 1980s and 1990s (CarolinaSanchez-Parama and Norbert Schady 2003;Francis Green, Andy Dickerson and JorgeSaba Arbache 2001; Gasparini 2003). Green,Dickerson, and Arbache (2001) attribute thisfinding to the small share of university grad-uates in total population. Unfortunately,

12 Increases in wage inequality in Mexico during the1990s are more pronounced in the first half of the decade,especially in urban areas.

studies that decompose changes in wageinequality into changes in the distribution ofobservable skills (such as education),changes in the prices of observable skills,and changes in unobservables, which arecommon in the literature on the evolution ofinequality in the United States, rarely existfor developing countries. Kijima (2006) pro-vides an example of such decomposition.She formally shows that most of the increasein the postliberalization wage inequality inurban India can be attributed to increases inthe prices for observable skills, and in partic-ular to the return to tertiary education.However, the wage inequality increase of the1980s (when returns to tertiary educationremained relatively stable), was largely dueto changes in the quantity of observed skill.Similarly, Gasparini (2004) finds that wageinequality increases during the 1990s inArgentina can be to a large extent attributedto the rising skill premium, while changes inthe educational composition of the work-force importantly contributed to growingwage inequality in the 1980s (when the skillpremium actually slightly declined).

Income-based measures of inequalityhave been used less widely in the literatureon globalization and inequality. As men-tioned earlier, this is partly due to the lack ofreliable survey data on nonwage sources ofincome (especially in Latin American coun-tries). Surveys that contain such informationare more recent and often less frequentlyconducted than labor market surveys. Thelimited information available in LatinAmerican countries (mainly drawn fromGasparini 2003) suggests that incomeinequality and wage inequality move in thesame direction, although changes in incomeinequality are at times less pronounced thanchanges in wage inequality or the skill pre-mium (for example, in Mexico and Colombiaduring the 1990s). Finally, a consumption-based measure of inequality is to our knowl-edge available over this period only forIndia, which has a nationally representativeconsumer expenditure survey that spans the



1980s and 1990s.13 In urban areas, con-sumption inequality moves in the samedirection as income and wage inequality; it isrelatively stable during the 1980s (a periodprior to major liberalization) but increasesduring the 1990s. Although this pattern can-not be generalized to other countries, it isreassuring that at least in the one case whereboth income and consumption inequalitymeasures are available, they both move inthe same direction.

In summary, the evolution of variousmeasures of inequality suggests that most ofthe developing countries experienced anincrease in inequality during the past twodecades. More importantly, we find no evi-dence that any measure of inequalitydecreased over this entire period when com-pared to earlier periods characterized by lessglobalization. As we note in the introductionto this section, our discussion abstracts fromseveral potentially important countries, mostnotably China. Lee Branstetter and NicholasLardy (2006) provide an excellent detailedreview of the process through which Chinaincreasingly liberalized its trade and foreigndirect investment policies during the 1980sand 1990s, culminating with the country’sentry into the WTO in 2001. During thisperiod, income inequality in China hasincreased (Wei and Wu 2002), so thatChina’s experience is consistent with thepositive correlation between inequality andexposure to globalization noted above.

The survey of the evidence confirmsWood (1999), who noted that inequalityincreased in several middle-income LatinAmerican countries that liberalized theirtrade regimes during the 1980s and 1990s. Itfurther suggests that this positive relation-ship also holds in the cases of India, China,and Hong Kong. As noted previously byWood (1999), the experience of developingcountries that globalized during the 1980s

13 There is however a large debate on whether thesesurvey data allow for over time comparisons during the1990s given the changes in the survey questionnaire. SeeDeaton and Jean Dreze (2002) for an excellent discussion.

and 1990s contrasts with the experience ofseveral Southeast Asian countries (SouthKorea, Taiwan, Singapore) that underwenttrade reforms in the 1960s and 1970s. Thelatter observed a decline in inequality as theyopened up their economies to foreign mar-kets. We discuss the possible explanationsproposed by Wood (1999) for the differ-ences in these correlation patterns betweencountries that globalized in the 1960s and1970s and countries that globalized in the1980s and 1990s in section 5.1.1.Unfortunately, neither detailed data on tar-iffs nor micro surveys are readily availablefor the early globalizers during the periodsof reform to allow us to examine whether thedeclining inequality in these countries wascaused by globalization as opposed to beingjust coincidental.

In general, one needs to be careful draw-ing conclusions regarding the link betweenglobalization and inequality simply based onbefore and after comparisons. Table 1 listsother important reforms that took place dur-ing periods of external liberalization inselected countries. Perhaps the most strikingfeature of these reforms is the fact that not asingle country implemented trade reformsor FDI liberalization in isolation from otherpolicy changes. For example, the most dras-tic trade policy liberalization in Colombia in1990–91 coincides with changes in labormarket regulation that substantiallyincreased the labor market flexibility.Mexico’s 1985 trade reform took placeamidst privatization, labor market reform,and deregulation. These concurrent policyreforms combined with the simultaneouschange of several globalization measuresmake it particularly difficult for the researcherto disentangle the effects of trade liberaliza-tion (or other aspects of globalization) fromthe effects of other policies.

4. Identification of Trade Policy Effects

The previous section documents thatmany developing countries experienced an



increase in inequality as they became moreexposed to various dimensions of globaliza-tion. But establishing a causal link betweenglobalization and inequality by providingcredible empirical evidence poses severalchallenges beyond the measurement issuesdiscussed in section 2. We highlight theseissues for the case of trade policy below,however similar concerns exist with respectto other aspects of globalization

Although there are several channelsthrough which trade policy can affectinequality within a country (we discuss thesechannels in detail in the next section), acommon theme in many of the mechanismsdiscussed in the theoretical literature is thattrade policy affects wage inequality bychanging the relative demand for skilledworkers. The main empirical challenge ishow to isolate the effects of trade from othercontemporaneous changes in the economicenvironment that may have induced shifts inthe relative demand and supply of skilledlabor. Governments in developing countriesoften implement trade reforms concurrentwith other economywide policy changes,ranging from labor market reform to indus-trial delicensing, tax reforms, and privatiza-tion. Table 1 illustrates the prevalence ofreforms that may have had confoundingeffects on wage inequality.

4.1 A General Equilibrium Approach

Any study that attempts to address theambitious question of “what is the overalleffect of trade liberalization on inequality ina country” thus requires strong modelingand identification assumptions. Porto (2006)is an example of such a study. He examinesthe implications of the Argentinean tradereform for the distribution of household wel-fare in a general equilibrium model of trade.In his framework, trade policy influenceshousehold welfare by changing the relativeprices of goods, which in turn affect laborincome and consumption. Because house-holds in different parts of the prereform wel-fare distribution differ in the composition of

their consumption bundles and their educa-tion endowments, they will be differentiallyimpacted by price changes. For example,households in the left tail of the welfare dis-tribution spend a higher share of their budg-et on basic items, such as food, and are lesseducated than richer households. Themodel, combined with predictions about thechanges of traded good prices, estimates ofwage–price elasticities, and estimates of theresponsiveness of the nontraded good pricesto traded good prices, can be used to simu-late the effect of trade policy changes on thedistribution of household welfare (i.e.,household expenditure per capita).

The main advantage of this approach isthat it ultimately yields an answer to theimportant question of how trade reformaffects the welfare distribution within acountry in a general equilibrium setting thatexplicitly accounts for intermediate goodlinkages and nontraded goods. However, thepredictions of the model depend in a crucialway on estimates of parameters that are typ-ically not known: the wage–price elasticities,the elasticity of nontraded good prices withrespect to traded good prices, and thedegree of pass-through from trade policychanges to product prices. These parametersare difficult to estimate consistently withtime-series data on wages and prices in a set-ting when many other policies change con-temporaneously with trade. Moreover, iflabor and capital are assumed to be mobilewithin a country, as is often the case in long-run general equilibrium trade models, thenthe level of industry aggregation needed toempirically implement the general equilibri-um approach is very high, implying thatthere is not enough variation in the data toidentify the relationship between trade poli-cy and the variables of interest (prices,wages, etc.). For example, in theHecksher–Ohlin model, both skilled andunskilled labor are assumed to be perfectlymobile, so that—no matter what the cross-sectional pattern of trade protection or liber-alization is—the wages for skilled and



unskilled wages should be equalized acrossdifferent sectors in an economy. If this weretrue, it would eliminate the prospect ofexploiting any cross-sectional variation intrade barriers and wages in order to identifythe relationship between trade and wagechanges.

4.2 Differential Exposure Approach

An alternative approach to identifying theeffects of trade liberalization on the wage orincome distribution is taken by severalrecent studies that have focused on cross-sectional variation in changes in trade pro-tection. Such studies examine whether,within a country, industries or regions thatwere more exposed to trade liberalizationexperienced smaller or bigger changes inwage or income inequality than less-exposedindustries/regions (Topalova 2004a; Hansonforthcoming; Goldberg and Pavcnik 2005;Wei and Wu 2002). The empirical frame-work in this line of work usually exploitshousehold survey data that include informa-tion on individuals’ industry of occupation,wage, region of residence, and variousdemographic characteristics, such as age,education, etc., to construct measures ofaverage wages by industry (after controllingfor relevant worker, industry, and job charac-teristics) or measures of wage or incomeinequality by region. These measures arethen related to trade policy changes over thespan of a trade liberalization episode to iden-tify the effect of trade barrier reduction oninequality. To the extent that the tariff changesdiffer across industries/regions and are exoge-nous (or can be instrumented for), the differ-ential exposure of various industries/regionsto tariff changes enable the researcher to sep-arate the effects of trade liberalization fromthe effects of concurrent policy changes. Theunderlying premise of this line of work is thatlabor is not perfectly mobile across industriesand/or regions (or at a minimum that certainskills are sector-specific and not easily trans-ferable across industries), so that wages arenot equalized across sectors/regions. This

premise is plausible in the short and medi-um run but questionable in the long run. Atany rate, it is important to note that failure ofthis premise to hold in practice does notinvalidate the approach; it simply impliesthat one would not find any differential tradepolicy effects across industries/regions inthis case, as wages are equalized acrossindustries/regions. However, studies thathave exploited industry or regional variationin developing countries do find effects, sug-gesting that the assumption of constrainedlabor mobility is more appropriate in thecontext of developing countries.

The main advantage of approaches thatexploit differential time changes in tradeprotection across cross-sectional units is thatthey require much weaker identificationassumptions than the general equilibriumapproach described above, so that the causallink between trade and inequality is perhapsmore convincingly established. On the otherhand, such approaches can only identifyindustry- or region-specific deviations fromaggregate trends that could, in principle, inpart be due to trade policy. This limitation isdiscussed in Topalova (2004a), who exploitsthe differential exposure of Indian districtsto trade liberalization to identify the effectsof trade on poverty. Her results indicate thatdistricts that were more exposed to liberal-ization experienced a relative increase inpoverty (or, more accurately, a smallerdecrease in poverty). However, povertydeclines dramatically in India over this peri-od. While her approach can plausibly identi-fy the role of trade in explainingdistrict-specific deviations from this aggre-gate trend, it cannot identify the role oftrade liberalization in explaining the trenditself.

An additional limitation of the aforemen-tioned studies is that their usual focus onnominal rather than effective tariff ratesimplies that they ignore intermediate inputlinkages, so that they are ultimately partialequilibrium in nature. This focus is, however,not inherent in the nature of the identification



approach but rather dictated by data con-straints; effective rates of protection areavailable for a few, isolated years at best and,even then, they tend to be noisy.Fortunately, for the few years for whicheffective rates are available, the correlationbetween nominal and effective rates of pro-tection appears to be positive and high, sothat the findings based on nominal tariffrates are likely to be robust to using effectivetariff rates as a measure of protection. At anyrate, this latter shortcoming could, in prin-ciple, be addressed with better data thatwould allow one to compute effective ratesof protection.

4.3 The Endogeneity of Trade Policy

Another challenge facing the literature ontrade and inequality is that trade policy is theoutcome of a political process and thusendogenous. While there is a large theoreti-cal and empirical literature on the determi-nants of the protection structure acrossindustries, empirical work on trade andwages has only recently focused on theendogeneity of trade protection and liberal-ization. The concerns about the endogeneityof trade policy and political economy of pro-tection apply to all studies. For example, instudies focusing on the overall effects oftrade reform, the political economy of pro-tection might affect the assumptions on theexpected price changes subsequent to thereforms and the consistency of the estimatesof wage and cross-price elasticities.

Similarly, studies that exploit cross-industryor cross-regional changes in the pattern ofprotection have to answer the question, is itvalid to treat such changes as exogenous?Fortunately, the nature of the tariff reformsin several developing countries, such asColombia, Brazil, Mexico, and India, makesthe usual concern about the endogeneity oftrade policy in the context of these countries’trade liberalization potentially less severe.Their governments’ goal in implementingtrade reforms was to lower tariff levels acrossindustries to more uniform levels that were

negotiated with the WTO, rather than caterto special lobby interests; as a result, indus-tries with initially higher level of protectionexperienced greater tariff declines. Thus,trade liberalization did not simply lower tar-iff levels but also changed the structure ofprotection across industries. In fact, studiesdocument that industries with larger pre-reform tariffs experienced larger tariffchanges in Colombia (Goldberg and Pavcnik2005), Brazil (Pavcnik et al. 2004) and India(Topalova 2004a). This pattern suggests thatindustry lobbies may have had less influenceon the magnitude of the tariff changes duringthe reform period.

With these methodological issues in mind,we now examine the existing evidence onvarious channels through which trade policyhas affected inequality.

5. The Relationship between Globalizationand Inequality

Globalization affects individuals throughthree main channels: changes in their laborincome, changes in relative prices andhence consumption, and changes in house-hold production decisions. Consistent withthe income- or wage-based measurement ofinequality, most research to date hasfocused on the first channel. The first fiveparts of this section are therefore devoted tosummarizing the evidence related to theeffects through the labor income channel.Since the increase in the skill premium hasbeen identified as one of the main con-tributing factors to rising wage inequality,we start by reviewing the main explanationsfor the widely documented increase in theskill premium (part 5.1). Next we discussother ways through which globalization mayhave impacted the income distribution:transitional unemployment (part 5.2);changes in industry wages (part 5.3); uncer-tainty (part 5.4); and potential effects onlabor market standards (part 5.5). The sixthpart of this section focuses on the effects ofglobalization on household production andconsumption decisions.



5.1 Explanations for the Increase in theSkill Premium

Whatever explanation for the wideningwage gap between skilled and unskilledworkers in many developing countries oneadopts, there seems to exist wide agreementthat the skill premium increase was drivenby an increase in the demand for skilledworkers. The main evidence on this issuecomes from studies that have documentedthat wages and employment in various skillcategories have moved in the same direction,implying that demand shifts dominated(Robbins 1996; Sanchez-Paramo and Schady2003). The related arguments are similar tothe ones used in the context of the inequali-ty debate in developed countries but appeareven more convincing when applied todeveloping countries, as many of them(Latin American countries in particular) didnot experience the same increase in the sup-ply of educated workers as the U.S. and EastAsian economies (Attanasio and Szekely2000; Sanchez-Paramo and Schady 2003).The causes of the increased demand forskilled workers have however been the sub-ject of intense debate.

5.1.1 Stolper–Samuelson Effects

The most direct link between trade open-ness and changes in the skill premium is pro-vided by the best known general equilibriummodel of International Trade, theHeckscher–Ohlin model. This model hasshaped thinking about the distributionaleffects of trade openness in the lastdecades, even though the theoretical andempirical shortcomings of the model arewidely recognized by now.14 In its simple2 × 2 version, the model predicts that coun-tries that are relatively rich in unskilled

14 On the theoretical side, the model rests on extreme-ly restrictive assumptions such as perfect competition,perfect labor and capital mobility within a country andfixed technology. On the empirical side, there has been nosupport for the predictions of the model, at least not in itsstrict version.

labor will specialize in the production ofgoods that are unskilled-labor intensive.The connection to the income distributionis provided by the model’s companion theo-rem, Stolper–Samuelson, that links changesin product prices to changes in factorreturns. A trade-liberalization-inducedincrease in the price of unskilled-labor-intensive products should, according toStolper–Samuelson, increase the return tothe factor that is intensively in the produc-tion of these products, unskilled labor. Incontrast, the expected decrease in the priceof the skilled-labor intensive imported prod-ucts should lead to a decline in the wage ofskilled labor. Based on this theorem and theempirical evidence suggesting that develop-ing countries are richly endowed withunskilled labor,15 one would expect the dis-tributional changes induced by trade liberal-ization in developing countries to favor theunskilled workers.

The general equilibrium nature of theHeckscher–Ohlin model makes it extremelyhard to bring it to the data. Given that themodel’s predictions refer to economywidefactor returns, one has only one observationper year to work with. In theory, one couldtry to identify Stolper–Samuelson effects byrelating trade-policy-induced relative pricechanges to factor returns over time but, inpractice, this approach is fraught with prob-lems. Price data are often incomplete,while the changing mix of goods producedwith different factor proportions withinstatistically defined product categoriesmakes price comparisons over time less

15 The most influential paper on this issue has been thestudy by Anne O. Krueger et al. (1981) that calculated thefactor content of trade in manufactures for several devel-oping countries in the 1970s and showed that the export-ing sectors were less skill-intensive than theimport-competing sectors. These results have been con-firmed in several other studies (see Wood 1999 for anoverview). The only caveat is that most of these studies donot cover the last two decades (they typically include dataup to the mid-1980s) and do not differentiate betweenhigher-skilled and lower-skilled trading partners for eachdeveloping country.



informative than one would have hoped.Furthermore, prices are determinedendogenously and may change for reasonsunrelated to trade. For these reasons, adirect link between goods and factor pricesas suggested by general equilibrium trademodels has been empirically elusive.

Despite the lack of direct evidence onStolper–Samuelson effects, it seems safe tosay that the widening wage gap betweenskilled and unskilled workers in the last twodecades in many developing countries thatare presumed to be relatively abundant inunskilled labor seems inconsistent with thespirit of the Hecksher–Ohlin theory. In prin-ciple, it is possible to reconcile the evidenceon wage inequality with the theory by con-sidering various extensions of the originalmodel. However, several other patterns doc-umented in developing countries seeminconsistent with Hecksher–Ohlin.

First, a fundamental prediction of factorendowment based trade theories is that theadjustment process to trade reforms wouldinvolve labor reallocations from sectors thatexperience price declines, and hence con-tract, toward sectors that experience relativeprice increases and hence expand. However,most studies of trade liberalization in devel-oping countries find little evidence in sup-port of such reallocation across sectors. Thelack of labor reallocation following tradereform has been documented by Ana L.Revenga (1997), Hanson and Harrison(1999), and Zadia M. Feliciano (2001) forMexico; by Attanasio, Goldberg, and Pavcnik(2004) for Colombia; by Janet Currie andHarrison (1997) for Morocco; by Topalova(2004a) for India; and by Romain Wacziargand Jessica Seddon Wallack (2004) in across-country study of trade liberalizationwhere, however, trade liberalization is cap-tured only through a time dummy. Thesestudies attribute the lack of labor realloca-tion in response to trade reform to eitherrigid labor markets (so that the adjustmentto trade liberalization occurs through rela-tive wage adjustments (Colombia, Mexico,

India)), or to the existence of imperfectproduct markets (so that firms respond bylowering of profit margins (Mexico,Morocco) and not through labor reallocationacross sectors). An alternative line of expla-nation for the lack of labor reallocation issuggested by recent work by Kaivan Munshiand Mark R. Rosenzweig (2005), who alsodocument very low spatial (and social)mobility in India.16 They argue that thesocial insurance provided by subcaste net-works creates a disincentive to migrate orout-marry out of fear of losing the services ofthese networks. Interestingly, the increase ininequality (possibly due to the Indianreforms) is shown to lower mobility (whichwas low to start with) even further, as sub-castes successfully coped with the conse-quences of rising inequality. In contrast,Gene M. Grossman (1986) and Revenga(1992) find greater employment than wagesensitivity to trade shocks for the UnitedStates. These differences in the adjustmentmechanisms are consistent with greater labormobility in the United States compared tothe developing economies.

A second piece of evidence that seemsinconsistent with Stolper–Samuelson effectsis that empirical work on developing coun-tries typically finds that the share of skilledworkers has increased substantially withinmost industries in the last two decades.Within-industry increases in the share ofskilled workers have been reported forArgentina, Brazil, Mexico, Chile, andColombia (Robbins 1996; Sanchez-Paramoand Schady 2003; Attanasio, Goldberg, andPavcnik 2004); Hong Kong (Hsieh and Woo2005); and India (Kijima 2006). The highershare of skilled workers in most industrieshas often been interpreted as evidence in

16 Further evidence on this issue includes Topalova(2004a), who documents little spatial mobility across dis-tricts in India during the 1980s and 1990s, and DanielChiquiar (2004), who finds little mobility of individualacross Mexican regions in five-year intervals surroundingthe Mexican trade reforms in the late 1980s and 1990s(see table 5 in his paper).



favor of skilled-biased technologicalchange.17

Given these patterns, several explanationsfor the increase in the skill premium havebeen suggested in the literature. The firstone is to consider simple extensions of themodel that would reconcile the theory withthe evidence. Specifically, the simpleStolper–Samuelson predictions may beoverturned if one introduces nontradedgoods or additional factors in the model.Suppose, for example, that there are threefactors of production, skilled labor,unskilled labor, and land (or naturalresources, or primary factors). Suppose fur-ther that some developing countries are rel-atively abundant in land (as is the case withmany Latin American countries) and thatland is a complement to skilled labor. Thengreater trade openness will favor land-intensive goods in these countries. If pro-duction of these goods requires a higherratio of skilled to unskilled workers, tradeopenness will benefit skilled workers. A sim-ilar argument can be applied to the role ofnontraded goods. Such arguments rest,however, on the rather implausible assump-tion that land, or natural resources, requirea higher ratio of skilled workers; to ourknowledge, there has been no evidence sup-porting this claim. Moreover, even in thatcase, one would expect labor reallocation tobe the mechanism through which thechanges in the wage distribution are trans-mitted and, as noted above, the evidence onsuch reallocation has been lacking for devel-oping countries.

Another line of explanation for theincrease in the skill premium focuses on thepattern of protection prior to trade liberal-ization in many developing countries and theskill intensity of the sectors that wereimpacted the most by trade reforms.Several studies have noted that, contrary to

17 This interpretation is not uncontroversial. We dis-cuss it in more detail under “Skill-Biased TechnologicalChange” in one of the following subsections.

expectations, it was the unskilled labor-intensive sectors that were protected themost prior to trade reform. This protectionpattern has been reported for Mexico(Hanson and Harrison 1999; Robertson2000, 2004 for pre-NAFTA period), Morocco(Currie and Harrison 1997), and Colombia(Attanasio, Goldberg, and Pavcnik 2004).The same studies document that it was infact the unskilled-labor-intensive sectors thatwere impacted the most by tariff cuts. Giventhis evidence, the increase in the skill premi-um is exactly what Stolper–Samuelsonwould predict: since trade liberalization wasconcentrated in unskilled-labor-intensivesectors, the economywide return tounskilled labor should decrease—at least inthe period immediately following thereforms. In fact, the only study that has toour knowledge exploited price data,Robertson (2004), documents that relativeprices and relative wages in Mexico areclosely related along the lines suggested bythe Stolper–Samuelson theorem (see figures4 and 5). This argument demonstrates theadvantages of exploiting the sectoral varia-tion in tariff changes as opposed to relyingon time variation alone to identify the effectsof trade policy changes. Studies that simplyuse “before–after comparisons” to uncoverthe effects of trade liberalization miss theimportant fact that—unlike in textbooks ofInternational Trade—the comparison is notbetween autarky and free trade, but ratherbetween protection and “less-protection” sothat the pattern of protection across sectorsprior to liberalization is crucial in determiningthe effects of trade reforms.

Still, this argument is not completely satis-factory as it again implies sectoral labor real-location—a prediction that no empiricalstudy of trade liberalization in a developingcountry has found strong support for.Moreover, the initial pattern of protectionthat favored unskilled-labor-intensive sec-tors seems a puzzle by itself. Why did coun-tries abundant in unskilled labor find itdesirable to protect the low-skill-intensive



sectors when the pattern of comparativeadvantage would have suggested otherwise?One possible answer is that the protectionpatterns reflected political economy consid-erations that had little to do with compara-tive advantage. Another is that high tariffs inlow-skill-intensive industries, such as textilesor footwear, were left over from a time inwhich these sectors were capital and high-skill intensive and when the protection ofthese sectors would have been compatiblewith the patterns of comparative advantage.Hisahiro Naito (1999) offers an alternativeexplanation from a public finance perspec-tive: He argues that, contrary to the view oftraditional trade theory, tariffs imposed onunskilled-labor-intensive products canPareto improve welfare in a small openeconomy that uses a redistributive nonlinearincome system; this is because the tariffschange the unskilled/skilled wage ratio,which in turn reduces the incentive problemof income redistribution. Since the unskilledearn more, the tax burden of the skilled canbe reduced; the reduction of the tax burdenhas a first order effect on welfare, while thedistortion introduced by the tariff is onlysecond order. This argument applies even todeveloped countries with flexible income taxsystems but much more so to developingcountries that have fewer alternative meansof redistribution. A trade liberalization thensacrifices this redistribution in the hope ofachieving enough growth to eventuallycompensate the less skilled.

Another possibility is that the recent entryof China and other low-income developingcountries (India, Indonesia, Pakistan, etc.) inthe world markets shifted the existing pat-terns of comparative advantage in middle-income countries. This possibility isexamined in detail in Wood (1999). Woodpostulates that, while in the 1960s and 1970smiddle-income countries had a comparativeadvantage in goods of low-skill intensity, inthe 1980s and 1990s, when low-incomedeveloping countries started exporting to therest of the world, the comparative advantage

of middle-income countries shifted to goodsof intermediate skill intensity. This shiftingpattern of comparative advantage mightexplain why many middle-income countriesfound it necessary to protect their low-skillintensive sectors from imports from low-income countries. It would also explain whygreater openness in these countries wouldnot necessarily benefit low-skill workers, asthe trade barrier reductions in low-skill-intensive sectors (such as textiles) may lead toan increase in the imports from China ratherthan an increase of domestic production andexports. While this argument is a priori plau-sible, there has been no direct empiricalinvestigation of its implications to date. Amore disaggregate analysis of imports andexports of middle-income countries thatwould differentiate between “more skilled,”high-income trading partners and “lessskilled,” low-income trading partners mightshed light into this question in the future.

A final explanation for the apparent ten-sion between the increase in the skill premi-um and theoretical predictions is that tradeaffected the wage distribution through chan-nels other than the ones suggested by simpleHecksher–Ohlin theory, or that there wereother forces at work (some of which mayhave interacted with, or even been inducedby, trade openness). A common theme insubsequent research on alternative channelsthrough which trade affects inequality is tofocus on the mechanisms that lead toincreased relative demand for more edu-cated labor within industries (as opposedto across industries). We consider thesearguments next.

5.1.2 The Role of Intermediate Goods andOutsourcing

Most trade models assume that all tradeoccurs in final goods: this assumption wasalso implicit in the above discussion of theHecksher–Ohlin mechanism. However,recent work by Feenstra and Hanson (1996,1997, 1999, 2003) has challenged thisassumption and emphasized the growing



importance of trade in intermediate goods,the so called “outsourcing” or “global pro-duction sharing.” They argue that the rapidexpansion in “global production sharing”over the past two decades can explain part ofthe observed increase in demand for skilledworkers in both developed and developingcountries.

The basic framework in Feenstra andHanson relies on the premise that produc-tion of final goods can be split into interme-diate stages and that intermediate inputsdiffer in their skill intensities. Consequently,firms find it optimal to “outsource” some ofthe production stages to cost-minimizinglocations abroad. Trade liberalization, cou-pled with a removal of capital controls,opens new opportunities for firms to shiftthe production of some of these intermedi-ate goods from developed to developingcountries. While products shifted to devel-oping countries would be characterized asunskilled-labor-intensive from a developedcountry’s perspective, they appear skilled-labor-intensive when compared with existingdomestic production activities from thedeveloping country’s point of view. As aresult, “outsourcing” increases the averageskill intensity of production in both thedeveloped and developing economies,inducing an increase in the skill premium inboth places.

While descriptive statistics on trade flowssuggest that outsourcing is a potentiallyimportant phenomenon in the developingworld (especially in Southeast Asia), theimpact of outsourcing on wage inequality indeveloping countries has so far been exam-ined only for Mexico and Hong Kong.Feenstra and Hanson (1997) find strongsupport for the “global production sharing”hypothesis for Mexico, where many U.S.firms export intermediate inputs tomaquiladora plants that assemble the inputsinto final products. Similarly, Hsieh and Woo(2005) document a large increase in the rel-ative demand for skilled workers in HongKong after firms reallocated much of the

(relatively unskilled-labor-intensive) produc-tion facilities from Hong Kong to China fol-lowing China’s FDI liberalization in the late1970s. Hong Kong, in turn, specialized inskill-intensive manufacturing activities andoutsourcing-based services such as market-ing and distribution.

The lack of empirical work on other devel-oping countries is partly explained by thepredominant interest in the role of outsourc-ing in developed rather than developingcountries. But it may also reflect the factthere are few developing countries that havereceived as large FDI flows as Mexico orthat have outsourced as big a share of theirproduction as Hong Kong. However, it isunlikely that outsourcing affects wageinequality only in Mexico and Hong Kong,due to the proximity of these two countriesto the United States and China, respectively.Evidence from Southeast Asia (Keith Headand John Ries 2002), Central America(Andres Rodriguez-Clare 2001), andEastern Europe (Dalia Marin 2006) suggeststhat many other developing or transitioneconomies engage in production sharingwith developed economies. In fact, a recentstudy by Andzelika Lorentowicz, Marin, andAlexander Raubold (2005) confirms theimportance of outsourcing as an explanationfor wage inequality in a transition economysetting. They find that outsourcing activitiesof foreign multinationals in Poland are asso-ciated with a large increase in the relativedemand for skilled workers in Poland.Examining the relationship between out-sourcing and wage inequality in other devel-oping countries remains a topic for futureresearch.

5.1.3 Increase in Capital Flows andComplementarity of Capital withSkilled Labor

A basic premise of the Stolper–Samuelsonmechanism is that capital and labor, whileperfectly mobile within a country, areimmobile internationally. This premise isclearly inconsistent with the recent adoption



of outward-oriented policies in developingcountries that has been in many cases associ-ated with substantial increases in interna-tional capital flows. If globalization leads toan increase in capital inflows into developingcountries and if the utilization of capitalrequires the use of a higher share of skilledlabor, then the increase in capital flows willbe associated with higher demand for skilledworkers. This argument is put forward inCragg and Epelbaum (1996) for pre-NAFTAMexico and Jere R. Behrman, NancyBirdsall, and Szekely (2000) for several LatinAmerican countries. Both studies focus onthe role of trade reforms in reducing theprice of capital goods as the mechanism thatgenerates higher demand for both capitalgoods and skilled labor. Similar argumentsare developed in theories of endogenoustechnological change (e.g., Daron Acemoglu2003), since new technology is often embod-ied in capital good imports. Since these the-ories are most frequently used in the contextof (endogenous) skill-biased technologicalchange, we discuss them in more detail inthe next subsection.

5.1.4 Skill-Biased Technological Change

The main alternative explanation for theincreased demand for skilled labor has beenskill-biased technological change. “Skill bias”is inherently hard to measure and, becausemost of the measures commonly employedin the literature are based on endogenousoutcome variables (e.g., the share of skilledworkers in a firm’s wage bill) rather thanexogenous technology shocks, there exists nouncontroversial measure of skill-biased tech-nological change. Nevertheless, the repeat-ed finding of an increase in both the share ofskilled workers and their relative wage with-in fairly narrowly defined industry categoriesin both developed and developing countrieshas been interpreted as evidence for aworldwide skill bias in new technologies.

As with the evidence on Stolper–Samuelsoneffects of trade, it is possible to come upwith alternative explanations for this well

documented empirical phenomenon. How-ever, none of these explanations seemsentirely convincing. Edward E. Leamer(1998), for example, argues that sector biasand not factor bias determines changes inthe wage distribution: skilled-biased tech-nological change that is concentrated inunskilled-intensive sectors benefits un-skilled workers in the general equilibrium,while skilled-biased technological changeconcentrated in skilled-intensive industriesbenefits skilled workers. This argumenthowever requires that product prices do notchange, which is unlikely to be true during aperiod of trade reforms. Moreover, the(admittedly very scant) empirical evidencedoes not support this theory; Attanasio,Goldberg, and Pavcnik (2004) do not findany statistically robust evidence that skill-biased technological change in Colombia wasconcentrated in skilled-intensive industries;if anything, the (statistically insignificant)point estimates of their regressions suggestthat skilled-biased technological change wasconcentrated in low-skill sectors, whichwould have generated a decrease in the skillpremium in the general equilibrium.

The past decade witnessed an intense andlively debate between those who favoredthe trade-openness-based explanations forthe increase in the skill premium and thosewho considered skilled-biased technologicalchange as the primary force behind the doc-umented changes in the wage distributionworldwide. By now it has been recognizedthat the most credible explanations for thedistributional changes witnessed in the pastfew decades would most likely involve inter-actions of trade openness with skilled-biasedtechnological change. Along these lines, sev-eral recent papers have postulated that, eventhough skilled-biased technological changemay have played a greater role in increasingthe skill premium that particular trade policychanges, technological change was itself anendogenous response to more “openness” sothat globalization was indirectly responsiblefor the increase in the skill premium.



One of the earliest studies in this vein isWood (1995), who introduced the term“defensive innovation” to describe firms’response to trade openness. According to hishypothesis, intensified competition fromabroad may induce firms to engage in R&Dor take advantage of existing new technolo-gies that they may have had little incentiveto adopt prior to liberalization. This theoryis developed further in Mathias Thoenigand Thierry Verdier (2003). While thisargument seems more suitable to explain-ing the increase in inequality in the devel-oped world, it may be applicable tomiddle-income developing countries, suchas Colombia or Brazil, that underwent sig-nificant trade reforms in the 1980s and1990s. By that time, low-income developingcountries (e.g., China) had entered theworld markets, and the import competitionmiddle-income countries faced from thenew entrants in their low-skill-intensive sec-tors may have induced faster technologicalchange in these sectors. On the empiricalside, a common implication of these modelsis that in the short- and medium-run, skill-biased technological change should be morepronounced in the sectors that liberalizedmore. Attanasio, Goldberg, and Pavcnik(2004) indeed document that, during1984–98, the increase in demand for skilledworkers in Colombia was largest in thosesectors that experienced the largest tariffcuts. This provides some support for the the-ory that skilled-biased technological changewas itself an endogenous response to tradeliberalization.

A different mechanism through whichtrade liberalization can induce (or acceler-ate) skill-biased technological change is sug-gested by Acemoglu (2003), who develops amodel of endogenous technological changeand argues that, in the case of developingcountries, this technological change maytake the form of increased imports ofmachines, office equipment, and other cap-ital goods that are complementary to skilledlabor. Trade liberalization affects the

demand for skilled workers by reducing theprices of the relevant capital goods and,hence, increasing their imports. From anempirical point of view, this model has twodistinct implications: first, following a tradeliberalization episode in a developing coun-try, total imports for office equipment andadvanced machinery from developed coun-tries should increase; and second, theincrease in the demand for skilled workersshould be more pronounced in sectors thatimport more foreign machinery. This sec-ond implication is investigated for the peri-od surrounding the 1980s Mexican tradeliberalization by Harrison and Hanson(1999), who find that within each Mexicanindustry, firms that import machinery andmaterials are more likely to employ a high-er share of white-collar workers than firmsthat do not import these inputs. Pavcnik(2003), on the other hand, finds that theincreased relative plant demand for white-collar workers by Chilean plants in early1980s cannot be attributed to the use ofimported materials and foreign technicalassistance to these plants once one controlsfor time-invariant plant characteristics.Marc-Andreas Muendler (2004) reportsthat the use of imported intermediate prod-ucts plays only a minor role for productivityimprovements by Brazilian firms followingthe trade reform, while Ana M. Fernandes(forthcoming) notes a positive associationbetween the use of imported intermediateproducts and productivity of domesticplants in Colombia. The evidence on therole of machinery and office equipmentimports in transmitting new technology andcreating demand for skilled workers istherefore mixed.

An alternative mechanism through whichtrade liberalization can affect technologicalchange, and thus indirectly inequality, is sug-gested by Philippe Aghion, Robin Burgess,Stephen Redding, and Fabrizio Zilibotti(2005). In their model, firms’ response totrade liberalization depends on how closethey are to the technology frontier. Firms



that are sufficiently close to the frontier cansurvive or deter entry of (foreign) competi-tors by innovating; firms that are far fromthe frontier may not be able to fight externalentry. Hence, the average effect of trade lib-eralization will depend on the fraction offirms and sectors that are sufficiently closeto the frontier to fight for their survival. Inaddition, Aghion et al. emphasize the role ofdomestic institutions, labor market restric-tions in particular, and their interactionswith technology adoption for the distribu-tional effects of trade policy. In the empiricalpart of their paper, Aghion et al. look at theIndian trade liberalization of 1991 for sup-port of their theory. Consistent with theirtheoretical arguments, productivity andprofits increased by more in industries thatwere close to the Indian productivity fron-tier and in states that had more flexible labormarket institutions. This differential impactof trade liberalization across industries withdifferent proximity to the technology fron-tier and states with different regulatoryregimes had strong inequalizing effects.These conclusions find less support inTopalova (2004b), who documents, usingfirm-level data and detailed information onIndian industry tariffs from India, that tariffdeclines were associated with productivityimprovements in firms with both high andlow productivity prior to the trade reform.

Overall it seems fair to say that, eventhough the premise that trade openness hasinteracted with skill-biased technologicalchange to increase the demand for skilledlabor seems both a priori plausible and the-oretically well founded, the empirical evi-dence on the role of particular mechanismsthrough which this increase occurred, ismixed and inconclusive. Clearly, more evi-dence from other developing countries isneeded before one can draw general conclu-sions. There is also very little empirical worklinking skilled-biased technological changein developing countries to the rise in exports.Given that exchange rate realignments haveaffected exports in many developing coun-

tries in recent years, this is a promising areafor future research.

5.1.5 Compositional Changes WithinIndustries: Exporting and “Quality”Upgrading of Products, Plants, andWorkers

Recent literature has emphasized theimportance of firm heterogeneity in inter-national trade (see James R. Tybout (2003)for a survey). In particular, studies of theeffects of trade reforms on productivity thatexploit plant- or firm-level data typicallyfind major market share reallocationstowards more efficient plants (often withinthe same industry) in the aftermath of liber-alization. This finding seems to contrastwith the documented lack of labor realloca-tion across industries in response to tradeshocks. One possible explanation for theseseemingly conflicting findings is that thedocumented reallocations are in reality sim-ple “revenue-share” reallocations that couldpotentially result from changes in firms’market power, rather than factor realloca-tions. This is due to the fact that the plantlevel surveys that are typically employed tomeasure productivity do not contain data onphysical output or inputs, neither on plant-specific prices, so that the above variablesare measured in value terms while the priceindices that are used to deflate them are sec-tor specific. Another possibility is that dueto the factors discussed earlier on page 59,labor market regulation in particular, laboris in many developing countries less mobilethan capital. Finally, it is also possible thatthere is in fact a lot of labor movementacross firms, often within the same industry,but this movement is not visible at theaggregation level at which the industry ofemployment is reported in household sur-veys. In fact, one recent study by JohnHaltiwanger, Adriana Kugler, MauriceKugler, Alejandro Micco, and CarmenPages (2004) finds substantial labor reallo-cation within sectors in several LatinAmerican countries.



This latter possibility suggests a reorienta-tion of empirical analysis away from coun-tries or industries, toward firms or plants, asthe relevant units of observation, a move-ment that parallels recent developments ininternational trade theory. The focus of tra-ditional trade theory and empirics on sectorsor industries abstracts away from the sub-stantial heterogeneity of products and firmsthat are included in statistically definedaggregates. Products that fall into the sametwo- or three-digit SIC category may be pro-duced with different factor proportions,while individual firms may vary both interms of their efficiency or “quality” and interms of the type of workers they employ.Recent work has made this heterogeneitythe main focus of the analysis by stressingthe importance of compositional changeswithin industries in response to trade liberal-ization, which may induce reallocation ofboth capital and labor toward “better” firms.

The basic idea is that trade opennessinduces a “quality” upgrading of firms,where quality can mean either “firm produc-tivity” or “product quality.” The qualityupgrading in response to trade openness canitself arise either because firms in importcompeting sectors try to avoid competitionfrom cheaper countries by differentiatingthemselves or because trade can shiftresources from nonexporters to exporters(see Marc J. Melitz 2003 for a related argu-ment), and there is ample empirical evi-dence that exporters tend to be more“productive” than nonexporters. Despite thetheoretical appeal and plausibility of thesearguments that emphasize firm and plantheterogeneity within an industry, the empir-ical evidence on how this channel affectsinequality is still scant and mostly indirect.

What is essential for establishing a con-nection between compositional changeswithin an industry and the inequality debateis that “higher quality” firms have a higherdemand for skill, so that quality upgradingleads to an increase in the skill premium.For example, one dimension along which

firms within an industry differ is theirexporting status. If production for exportmarkets is relatively more skill-intensivethan production for developing countries’domestic markets, increased demand forexports will increase the relative demandfor skilled workers within industries andlead to a higher skill premium. Empiricalevidence from the United States suggeststhat exporting is indeed a skill-intensiveactivity (see Bernard and Jensen 1997).Harrison and Hanson (1999) also find thatexporters employ a higher share of white-collar workers than nonexporting plants inMexico.

Production of higher quality products maybe one reason why exporting firms in devel-oping countries may require relatively moreskilled labor than domestic firms. In addi-tion, “product quality” varies significantlyacross exporters from different countries.Schott (2004) provides strong evidence ofcomplete specialization by countries withinproduct categories, with the skill- and capi-tal-abundant countries specializing in theproduction and export of higher unit valueproducts, and unskilled-labor-abundantcountries specializing in the production andexport of low-unit value products. If oneaccepts his premise that unit values withinvery narrowly defined product categoriesreflect differences in product “quality,” thenthe implication of Schott’s findings is thatdeveloped countries specialize in higherquality products while developing countriesspecialize in lower quality products withinthe same product category. While these find-ings do not directly tell us how countriesadjust to trade liberalization, it seems plausi-ble to assume that, as middle-income devel-oping countries become more open to trade,they start upgrading their products like themore developed countries. If higher qualityproducts indeed require a higher share ofskilled workers, then the shift toward higherquality products will benefit skilled workers.Recent findings by Susan Chun Zhu (2005)are consistent with this idea. She finds that



wage inequality has increased by more incountries and industries that (because ofproduct cycles) shifted the within-industrycomposition of exports away from low-skill-intensive exports that were historically asso-ciated with less developed countries towardmore sophisticated products that had beenexported by richer countries in the past.

A somewhat different mechanism involv-ing upgrading is discussed in a recentpaper by Verhoogen (2006). In Verhoogen’sstudy, trade openness leads to an upgrad-ing of the average product quality inexporting plants, which in turn generatesdemand for a better qualified workforce.The upgrading of the workforce can, how-ever, be satisfied by upgrading the existingworkforce in each plant, rather than hiringof new, better qualified workers, so thattrade openness does not necessarily gener-ate labor reallocation across plants. Ratherthan focusing on trade liberalization,Verhoogen exploits a major exchange ratedepreciation episode (the 1994 peso crisisin Mexico) to study the response of firms toincreased openness. The peso depreciationclearly benefited exporters. Instead offocusing on the effects of an increase inaggregate exports on productivity ordemand for skilled workers, Verhoogenconsiders the effects of the exchange ratedepreciation on firms of different produc-tivity. More productive firms produce high-er quality products and export; lowerproductivity firms produce lower qualityproducts and sell in the domestic marketonly. The basic hypothesis is that theincrease in exports was associated with adifferential quality upgrading withinMexican manufacturing as higher-produc-tivity, exporting plants shifted their withinplant product-mix toward higher qualityvarieties in order to appeal to U.S. con-sumers. But this shift toward higher quali-ty products required an upgrading of theworkforce. As a result, the peso deprecia-tion induced quality upgrading benefitedskilled workers.

The increased demand for “skill”18 withinexporting plants could be met either byattracting new, better-educated workers orby increasing the productivity of the existingworkforce. In a subsequent paper, David S.Kaplan and Verhoogen (2005) argue that itwas the second mechanism that was at work:the higher demand for skill in exportingplants translated to higher efficiency wagesin these plants, rather than changes in theproportions of white-collar and blue-collarworkers within each plant. Higher efficiencywages could in turn reflect additional train-ing or effort by the white-collar workersemployed in exporting plants. This findingimplies that there was little labor realloca-tion across plants in the aftermath of thepeso depreciation.19 The higher demand forskill was instead satisfied by increasing thewage premia of the workers alreadyemployed in exporting plants.

The main challenge of this literature is todefine “quality” in an operational way. AsErkan Erdem and Tybout (2004) havepointed out, a separation of “firm produc-tivity” and what we typically mean by“product quality” is not possible given theavailable data sets. Moreover, the term“quality” is itself elusive from an empiricalpoint of view, especially in the context of ahorizontal differentiation model in whichconsumers value products differently.

18 We use the term “skill” here in the most generalsense of the word to include general human capital asreflected in a worker’s educational attainment: specifichuman capital, motivation, and effort. Importantly, thisinterpretation does not match the white-/blue-collarworker dichotomy often used in the literature to differen-tiate between skilled and unskilled workers. The mecha-nism discussed in Verhoogen (2006) demonstrates thelimitations of this latter narrow definition in capturing thetrue quality of the workforce.

19 Kaplan and Verhoogen (2005) exploit matchedemployer–employee data from the Mexican social securi-ty agency, so that they can follow workers and their wagesover time. A potential caveat of their analysis is that thedata do not contain information on worker education sothat one cannot be sure that within-plant changes ofworker wages do not reflect changing returns to a partic-ular worker characteristics, such a education, during thatperiod.



Schott (2004) tries to circumvent this prob-lem by implicitly assuming a vertical differ-entiation model so that higher unit valuescorrespond to higher quality. Verhoogen(2006) uses a set of proxy variables (forexample, a plant’s total sales) or, alterna-tively, a latent variable approach to capture“product quality.” However, from the per-spective of the inequality debate, it doesnot matter what definition of “quality” oneadopts. What matters is the proportion ofskilled and unskilled workers that isrequired to produce goods before and aftera trade liberalization or currency deprecia-tion episode. If the demand for skillincreases within firms, this is going toinduce an increase in the skill premium.Hence, rather than resorting to particularinterpretations of product quality that maybe controversial, empirical work in thisarea could directly examine how within-firm relative demand for skill is affected bytrade liberalization and whether this effectis different for firms with initially low ver-sus high skill-intensity (where “initial”refers to the skill-intensity observed priorto the trade reform or exchange rate depre-ciation episode).

The second challenge facing this litera-ture is that, for the results to be relevantfor the inequality debate, it is important tohave accurate measures of skill.Unfortunately, the information on workerand job characteristics provided in firmlevel data sets is much more limited thanwhat is usually provided in household sur-veys, so that researchers have to resort tothe familiar dichotomy between produc-tion and nonproduction, or white- andblue-collar, workers. In the absence ofmore detailed information, there is littleone can do in the short run. In the longerrun, more information about the character-istics of workers employed by differentfirms (or plants) will be essential for estab-lishing a connection between firm hetero-geneity and changes in the wagedistribution.

5.1.6 Changing Returns to Skill-IntensiveOccupations

In some developing countries, theincrease in the skill premium has beenlinked to the increase in the returns to par-ticular occupations that require a higherlevel of education. Cragg and Epelbaum(1996) find strong support for this hypothe-sis in the case of pre-NAFTA Mexico, forwhich they document a rapid increase in theoccupational premia of professionals andadministrators (including public administra-tors).20 The authors attribute the increase inthese occupational premia to the rapidchanges introduced in the economy byreforms that increased the demand for indi-viduals who could enact these reforms: man-agers and professionals. The link toglobalization is indirect: trade reformsimpacted these changing returns to occupa-tion only to the extent that they were part ofthe general reforms that generated demandfor highly educated individuals. In relatedwork, Kijama (2006) finds that increases inthe returns to tertiary degree were especial-ly pronounced for individuals in managerial,professional, and technical job in urbanIndia subsequent to 1991 reforms. Studieson other countries have however found lesssupport for rapidly changing returns toskill-intensive occupations.21

5.2 Transitional Unemployment

Perhaps the most commonly expressedconcern regarding globalization in develop-ing countries is that trade openness will lead

20 These changing premia to skill-intensive occupa-tions account for a significant fraction of the estimatedskill premium increase: controlling for occupation com-presses the original estimate of the change in the premi-um of postsecondary to secondary education from 67percent to 40 percent. Similarly, the increase in the pre-mium of postsecondary to primary education drops from70 percent to 42 percent once occupation is controlled for.

21 In Colombia for example, Attanasio, Goldberg, andPavcnik (2004) document that occupational returnsremained relatively stable over the 1986–98 period, withthe exception of 1992, for which there was a short-livedspike in the returns to “managers and other professionals.”



to transitional unemployment as the econo-my adjusts to new conditions. To the extentthat this unemployment disproportionatelyaffects the poor, it will have important con-sequences for income inequality.

Despite the prominence of this concern inthe public debate, there is remarkably littletheoretical or empirical work on its rele-vance. On the theory side, unemployment isabsent in the mainstream models of interna-tional trade, which typically assume fullemployment. A notable exception is thework of J. Peter Neary (1978, 1982) thatexplores the consequences of factor speci-ficity in the short run. In Neary’s framework,it is possible that labor markets are at dise-quilibrium in the short run as the economyadjusts to a terms-of-trade shock. Thisframework seems particularly relevant fordeveloping countries that are often charac-terized by severe labor market rigidities (seealso David Tarr and Steven J. Matusz 1999for a discussion).

On the empirical side, the lack of evidenceon the relationship between trade and tran-sitional unemployment is mainly due to theabsence of appropriate data. Aggregate sta-tistics on total unemployment by year seemto suggest that macroeconomic recessionshave a larger impact on unemployment thantariff reductions, but inferences based onmacroeconomic trends can be misleading, asthey do not indicate which industries andwhich population groups are most affected,what the causes of unemployment andchances of reemployment are, and how longthe duration of unemployment spells are.Such information is important for relatingunemployment to measures of well-beingand inequality. The link between trade poli-cy, unemployment, and inequality could bebetter identified by relating detailed indus-try tariff changes to changes in industryunemployment. The difficulty in pursuingsuch an approach stems from the fact thathousehold surveys in developing countriestypically do not report in which industriesthe currently unemployed used to work and

in which industries they seek new employ-ment; even when they do (as is the case inthe Colombian NHS for example), theyreport the industry at a very aggregate level(one-digit ISIC). As a result, it is not possibleto relate industry unemployment to moredisaggregate tariff changes. Furthermore,empirical work in this area needs to dealwith truncation issues, as workers who areemployed in any given survey interval canonly be assumed to be employed up to theend of the particular survey interval, and,similarly, unemployed workers can beassumed to be unemployed only to theextent that they have not found a new jobbefore the end of that survey period.

An attempt to relate trade liberalization totransitional unemployment was undertakenby Attanasio, Goldberg, and Pavcnik (2004)in the context of the Colombian trade liber-alization. The authors examine whether theincrease in the probability of being unem-ployed was greater for workers in the manu-facturing sector (where tariff cuts were thelargest) than for workers with the sameobservable characteristics in nontraded-good sectors (such as wholesale and retailtrade, restaurants, hotels, construction, etc.)in urban Colombia. They find that increasesin the probability of unemployment beforeand after tariff reductions were not larger inmanufacturing than in nontraded sectors.However, this evidence is based on a veryaggregate industry definition, while theinformation on unemployment is not direct-ly linked to changes in trade policy.Moreover, no attempt is made to linkchanges in probability of unemployment toinequality.

5.3 Industry Wages

Among those who are and remainemployed, our discussion so far has focusedexclusively on the impact of trade opennesson changes in the economywide skill premi-um. We now turn our attention to otherways in which globalization may have affect-ed wage and income inequality. The first



one is through changes in industry wagepremia.

Industry wage premia are the part ofworker wages that cannot be explained byobservable worker characteristics, such asgender, age, education, experience, etc., butcan be attributed to workers’ industry affili-ation. While studies have found that indus-try wage premia account for a significantportion of individual wage variation, there isless agreement as to whether these premiareflect compensating differentials, efficien-cy wages, industry rents, or returns toindustry-specific skills.

There are several plausible channelsthrough which trade policy changes mayaffect industry wage premia. In short- andmedium-run models of trade where workerscannot easily move across sectors, tariff cutstranslate into proportional declines in thewage premia of those industries that experi-ence larger than average tariff declines.22

This possibility is particularly important indeveloping countries characterized by labormarket rigidities (James J. Heckman andPages 2000). These rigidities may be irrele-vant in practice because of the existence ofinformal labor markets and the vast non-compliance with labor market regulation.However, the lack of labor reallocationacross sectors in the aftermath of dramatictariff declines in several countries that wediscussed earlier supports the premise ofrigid labor markets. A further channelthrough which trade may affect industrywages is suggested by models of imperfectcompetition and union bargaining. If prof-itable industries share part of their rentswith workers because of union bargainingpower, tariff cuts in these industries maylead to lower wages as the industry rentsstemming from protection disappear.

22 If such industries had lower wage premia in the pre-reform period, then such changes will further increase thewage dispersion, making those who received lower rela-tive wages to start with even worse off. This turns out tobe in fact the case with the trade reforms in Mexico in the1980s and Colombia in the 1990s.

Moreover, industry wage premia may beaffected in cases where unions share inindustry rents through employment securityguarantees rather than wages, and whereemployment security is obtained throughhigher trade protection (Grossman 1984).

Finally, trade policy could affect industrywage premia via industry-level productivitychanges. Several recent empirical studieshave found that trade liberalization was asso-ciated with productivity improvements indeveloping countries.23 If these improve-ments are passed on to workers in the formof higher wages, trade could increase wagepremia in the sectors that experienced high-er productivity gains due to their higherexposure to trade liberalization. A relatedargument is presented in the two studies byVerhoogen (2006) and Kaplan andVerhoogen (2005) we examined earlier,although the (efficiency) wage increases intheir mechanism are not generated by tradeliberalization but rather a peso-crisisinduced increase in exports destined for theU.S. market.

The empirical evidence on the responseof industry wage premia to trade reforms ismixed: no association between tariff reduc-tions and industry wage premia (Feliciano2001 for Mexico; Pavcnik, Andreas Blom,Goldberg, and Schady 2004 for Brazil), pos-itive association (Goldberg and Pavcnik2004 for Colombia), and negative associa-tion (Mishra and Kumar 2005 for urbanIndia).24 Feliciano (2001) reports a positiveassociation between declines in importlicenses and industry wage premia. The

23 See Harrison (1994) for Cote d’Ivoire; PravinKrishna and Devashish Mitra (1998), Aghion, Burgess,Redding, and Zilibotti (2005), and Topalova (2004b) forIndia; Euysung Kim (2000) for Korea; Pavcnik for Chile(2002); Fernandes for Colombia (forthcoming); andMuendler (2004) and Hay (2001) for Brazil.

24 Studies that rely on average firm or industry wagesrather than industry wage premia also report mixedresults: Currie and Harrison (1997) find no associationbetween changes in industry wages and tariffs inMorocco; Revenga (1997) on the other hand reports apositive association for Mexico.



heterogeneity of findings in the above stud-ies is perhaps not surprising given the largenumber of possible channels through whichtrade could affect industry wage premia.Kaplan and Verhoogen (2005) present evi-dence based on panel data that the wageincreases in plants with higher productivityVerhoogen documented in his earlier workare due to an increase in the efficiencywages of the workers employed in theseplants, rather than higher wages of newhires.

These findings have potentially importantimplications for the effects of trade opennesson wage inequality. Interestingly, both inColombia and Mexico, studies that have doc-umented a decline in industry wages inresponse to trade liberalization also find thatthe sectors that experienced the largest tariffcuts had the highest shares of less educatedworkers and the lowest wages to start with(see, for example, Attanasio, Goldberg, andPavcnik 2004). As a result of trade liberaliza-tion, the initially low wages declined evenfurther increasing wage inequality. In thissense, one could argue that less educatedworkers were hit twice: not only did the skillpremium increase during that period but thewages in industries that employed a propor-tionately higher share of unskilled workersdeclined relative to the average wage in theeconomy. Kaplan and Verhoogen’s results goin the same direction: even though wagesincrease in absolute terms in the aftermathof the peso crisis, the wages of white-collarworkers employed in high productivityplants increase by more, thus contributing toan increase in wage inequality.

While these effects go in the direction ofincreasing wage inequality, their magnitudeis estimated to be small, and so it is ques-tionable whether they are the primary forcebehind increases in wage inequality. InColombia, for example, the estimates sug-gest that the average tariff reduction in man-ufacturing sector of 37 percentage pointswould be associated with 4 percent declinein industry wage premium. Industry wage

premia account for about 2 percent ofexplained variation in log hourly wages con-ditional on workers’ observable characteris-tics in this country. Thus, while changes inindustry wages contribute to the increase inwage inequality, it seems unlikely that thechange in industry wage premia is a firstorder effect. One potential explanation forthe relatively small magnitude of industrywage responses and the simultaneous lack oflabor reallocation across sectors is the exis-tence of an informal sector in many develop-ing countries. This sector offers anadditional margin through which firms canadjust to trade shocks. We investigate thisexplanation in section 5.5.

5.4 Uncertainty

A body of research has examined the ideathat globalization not only affects incomelevels but also exposes workers to increasedeconomic uncertainty through less secureemployment and more volatile income.Conceptually, most empirical work in this lit-erature relies on a simple labor demand andsupply framework with a stochastic labordemand (see Kenneth Scheve and MatthewSlaughter 2002 for an in depth review of thisliterature). In this setting, trade reformmight increase wage uncertainty in twoways.

First, trade liberalization can lead togreater price volatility and productivityshocks (as in Dani Rodrik 1997, 1998),which in turn generates greater volatility inwages and employment. Scheve andSlaughter (2002) convincingly argue thatempirical studies do not reach a consensuson whether trade liberalization increasesprice variation. This state of affairs is per-haps not surprising. While trade liberaliza-tion exposes domestic consumers andproducers to the volatility of world prices, atthe same time the exposure to foreign mar-kets mitigates the effects of potentially largedomestic shocks on prices. Theoretical workby John McLaren and Andrew Newman(2002) makes a similar point, suggesting that



the relationship between globalization andrisk is ambiguous.

Second, Rodrik (1997) argues that tradereforms may increase wage uncertainty notonly through potentially greater demandvolatility but also by increasing the (absolutevalue of) the own-price elasticity of labordemand. The reasoning is as follows. For agiven vertical shift of the labor demandcurve (arising from productivity or productdemand shocks), a more elastic labordemand implies greater variation in wagesand employment. In this case, greater open-ness increases the uncertainty faced by indi-viduals only indirectly, not by exposing themto greater demand volatility but by magnify-ing the effects that any given demand shockwill have on their wages and employment.

One way in which trade reform canincrease the elasticity of labor demand is byintensifying product market competition(and thus increasing the elasticity of productdemand from which the labor demand isderived). In fact, James Levinsohn (1993),Harrison (1994), and Currie and Harrison(1997) find empirical support for increasedproduct market competition following tradereforms. They show that domestic firms low-ered their markups following the tradereforms in Turkey, Ivory Coast, andMorocco, respectively. Alternatively, tradeliberalization may make labor demand moreelastic by providing firms with increasedaccess to substitutes for domestic labor suchas imported intermediate products. Our dis-cussion in section 5.1.2 emphasizes the rapidexpansion of trade in intermediate goods.

Unfortunately, the empirical work thatlinks trade reforms to wage uncertainty isscarce, especially in the context of develop-ing countries. Most studies examine the linkbetween trade and wage uncertainty indi-rectly by studying the relationship betweentrade reform and labor demand elasticity.The results of these studies are mixed.Krishna, Mitra, and Sajjid Chinoy (2001)find no evidence that trade liberalizationincreased (the absolute value of) labor

demand elasticity in Turkey. Rana Hasan,Mitra, and K. V. Ramaswamy (forthcoming)find that labor demand becomes more elas-tic following the 1991 Indian trade reformsand that more protected industries havelower labor demand elasticities. Their studyis particularly interesting because it alsoexamines the differential effects of tradereform on labor demand elasticities in sec-tors with differential tariffs located in stateswith different labor market regulation. Thestudy finds that labor demand elasticities aregreater in Indian states with more flexiblelabor laws and that trade reforms increasedlabor demand elasticities by greater degreein states with more labor market flexibility.To our knowledge, Tom Krebs, Krishna, andWilliam Maloney (2005) is the only studythat directly examines the link betweentrade reform and income variability. Usinglongitudinal income data on workers beforeand after the Mexican trade reforms in the1980s and 1990s, they find that tariffdeclines are associated with increasedincome uncertainty.

To the extent that globalization increasesincome uncertainty, risk averse individualsmight be worse off even if trade reform doesnot affect or increases their expectedincomes. That said, the question stillremains open whether and how increaseduncertainty affects inequality. To the extentthat increases in uncertainty and/or riskaversion vary across individuals of differenteducation and/or ages, globalization induceduncertainty could add to greater inequalityacross individuals. Yet, we are not aware ofany study that links liberalization-inducedincreases in uncertainty to inequality.

A different but related point is that a moreuncertain product demand may induce firmsto adapt hiring practices that increase afirms’ flexibility to hire/fire workers inresponse to changing product demand. Forexample, a firm that operates in a more vari-able product market may find it beneficial torely more heavily on informal or temporarylabor to maintain flexibility. This, in turn,



could lead to greater wage variability. We arenot aware of any empirical work that exam-ines how greater product demand uncertain-ty affects firms’ choice of workers andcontract types and, ultimately, inequality.However this issue is partly related to abroader line of work that examines theeffects of globalization on the use of infor-mal labor and compliance with labor marketstandards. We examine this issue next.

5.5 Labor Market Standards

Many globalization opponents haveargued that globalization may have adverseeffects on inequality in the broader sense byinducing noncompliance of firms with labormarket standards and by increasing the pro-portion of workers in the informal sector ofthe economy. The informal sector is general-ly defined as the sector of the economy thatdoes not comply with labor market regula-tions, such as minimum wage or minimumworking age laws, and it is associated in thepublic’s mind with lower pay and worseworking conditions. It accounts for a sizableshare of the labor market in developingcountries: for example, 50 to 60 percent ofthe labor force in urban Colombia isemployed in the informal sector during the1980s and 1990s.

The claim that the informal sector offersworse working conditions is controversial.On one hand, several studies (DouglasMarcouiller, Veronica Ruiz de Castilla, andChristopher Woodruff 1997; Goldberg andPavcnik 2003; Pavcnik et al. 2004) documentthat workers with otherwise comparableobservable characteristics are paid lowerwages in the informal sectors of Peru, ElSalvador, Brazil, and Colombia; moreover,workers employed in the informal sector areconsiderably less likely to receive nonwagebenefits, and in household survey question-naires they express less satisfaction withtheir working environment and job quality.On the other hand, some individuals maychoose to work in the informal sectorbecause they value the greater flexibility in

work arrangements offered by this sector; tothe extent that this is true, the observed dif-ferences in pay between formal and informaljobs may be partly driven by selection ofindividuals based on unobservable tastes orcharacteristics.

The usual argument that trade liberaliza-tion will increase informality is that foreigncompetition forces firms to cut costs, whichthey in part do by employing a higher pro-portion of informal workers. Goldberg andPavcnik (2003) present a model that formal-izes this idea and show that, under certaintheoretical assumption, firms within anindustry may find it optimal to hire relative-ly more informal workers after a permanentdecline in industry tariffs. To the extent thatjobs in the informal sector are associatedwith relatively lower pay and worse workingconditions, the relative expansion of theinformal sector following a trade liberaliza-tion episode could contribute to growinginequality, especially since the informal sec-tor tends to employ a higher proportion ofless-educated workers. Alternatively, it ispossible that the expansion of the informalsector in the aftermath of trade liberalizationreflects the entry of new firms into themarket in response to new opportunitiescreated by the reforms. Such firms are like-ly to start small and informal, especially incountries with rigid labor markets, and shiftinto the formal sector only later if they aresuccessful.

Evidence on the link between informalityand trade reform is scarce due to the lack ofdata on informality and labor market regula-tion compliance of firms. The few studies onthese issues yield mixed results. Currie andHarrison (1997) find that firms in Moroccostarted hiring more temporary workers(who are not entitled to benefits) in theperiod following the trade reform.Goldberg and Pavcnik (2003) focus onColombia and Brazil, two countries thatexperienced expansions of their alreadylarge informal sectors in the years followingthe trade reforms, and examine whether



trade liberalization can explain the docu-mented increase in informality. Interestingly,most of the observed increase in the share ofinformal workers in the total labor forceoccurred through within-industry increases,rather than through shifts in employmentacross industries with different informalityintensity. However, the association betweenwithin-industry tariff changes and probabili-ty of employment in the informal sectorvaries across countries and time and seemsto be related more to the flexibility of thelabor market than to trade policies. In Brazil(a country with a relatively flexible labormarket according to Heckman and Pages(2000)), industry tariff declines were notassociated with changes in the probability ofemployment in the informal sector. InColombia (a country with more rigid labormarket institutions), industry tariff cuts wereassociated with increased probability ofinformal employment, but only in the peri-od prior to the implementation of labormarket reforms that substantially increasedthe flexibility of labor markets.

Related work has examined firms’ compli-ance with minimum wage laws. This aspectof the labor market regulation is particularlyrelevant in the globalization and inequalitydebate, because minimum wages affect pri-marily workers at the bottom of wage distri-bution. Harrison and Jason Scorse (2004a,2004b) study differences in compliance withIndonesian minimum wage legislation acrossexporters, foreign-owned firms, and domes-tic firms using Indonesian surveys of manu-facturers from the 1990s. In their study, theyconsider a firm to be compliant with the leg-islation if the average wage of productionworkers in the plant exceeds the minimumwage. They find that foreign owned plantsare actually more likely to have productionwages above the minimum wage. The use ofaverage wages as a measure of compliancemay also conceal individual instances ofwages below the legislated minimum wage.Unfortunately, data constraints precludethem from examining whether foreign-owned

plants subcontract to domestic establish-ments that may pay below the minimumwage. Relying on information on wages ofindividual workers, Goldberg and Pavcnik(forthcoming) find no association betweenthe likelihood of industry compliance withminimum wage laws and industry tariffreductions in Colombia.

Overall, existing studies provide little evi-dence that trade liberalization or FDI con-tribute to growing inequality by expandingthe size of the informal sector and inducingnoncompliance with minimum wage laws.However, more work is needed in this area.For example, the results in Goldberg andPavcnik (2003) suggest that the relationshipbetween trade reform and informalitydepends on the institutional setting in whichtrade reforms take place. To investigate thispossibility further, one would ideally exploitthe heterogeneity of labor market institu-tions over time and across administrativeareas within a country such as India, wherelabor market regulation varies across states.In addition, a disadvantage of the within-country analysis presented above is that itcannot by its nature shed light on the empir-ical relevance of “race-to-the-bottom” argu-ments; that is, arguments that suggest thatfirms that have the option of relocating theirplants will choose, everything else beingequal, the country with the lowest laborstandards. This induces competition amongcountries for footloose firms and leads tofurther degradation of labor standards.Bernard and Jensen (2003) and Bernard andFredrik Sjoholm (2003) present some evi-dence for the United States and Indonesiathat suggests a higher propensity of multi-nationals to relocate; after accounting for thefact that foreign affiliates are typically largerand more productive than domestic firms,foreign affiliates are more likely to shut downthan domestic firms in the host country.Determining whether these shut-down deci-sions (and the set up of new operations else-where) are driven by labor market regulationremains a topic for future research.



5.6 Household Production andConsumption

Our discussion so far has focused on thelabor market effects of globalization. Twoadditional channels through which globaliza-tion may affect inequality are household pro-duction and consumption. These channelsare particularly relevant in poorer develop-ing countries, yet they have received littleattention in the literature, perhaps becausethe main focus of the globalization relatedresearch to date has been middle-incomedeveloping countries.

Many individuals in poor countries are notemployed in the formal labor market forwages, but instead work in their householdbusiness or family farm and devote a sub-stantial amount of time to production ofgoods/services used for own consumption(Rosenzweig 1988). For example, inVietnam in 1993, about 19 percent of adultsage 20–64 report working for wages, while90 percent of adults report working withintheir own household (Eric V. Edmonds andPavcnik 2006). Similarly, in Indonesia, lessthan 30 percent (45 percent) of rural (urban)men and less than 12 percent (20 percent) ofrural (urban) women worked in wage workbefore the Indonesian crisis (James P. Smithet al. 2002). In India, 46 percent of the laborforce (rural and urban) works for wages(Suresh D. Tendulkar 2003).

The main reason for the limited amount ofempirical work on within household produc-tion and consumption is data constraints.Specifically, many surveys focus only on theformal labor market and thus exclude theself-employed. To the extent that the self-employed are surveyed, measures of profitsor net earnings associated with their busi-nesses are often missing or, to the extent thatthey are available, they tend to be noisy.Moreover, because labor market surveys donot contain information on householdexpenditures or consumption, the implicitvalue of products produced by households fortheir own consumption cannot be captured.

Abstracting from household production andconsumption may be defendable when onestudies the consequences of manufacturingtariff declines on urban households in a mid-dle income country such as Mexico orColombia. However, it is substantially moreproblematic to ignore these channels in pooreconomies such as India, Indonesia, orVietnam, especially in rural areas and incases when trade liberalization affected theagricultural sector.

The only study that has included house-hold production in studying the relationshipbetween trade reforms and inequality is toour knowledge Topalova (2004a), whoderives measures of inequality and povertybased on household expenditure data. Asmentioned earlier, her results suggest thatpoverty declined less in districts that liberal-ized more, but the findings regarding inequal-ity are less clear-cut: the point estimates inmost of her specifications suggest that biggertariff cuts were also associated with biggerincreases in inequality within a district,but these findings are never statisticallysignificant.

Other work has investigated how house-holds allocate their time between formal wagemarkets and within household work, but hasnot explicitly examined the relationshipbetween this allocation decision, globalizationand inequality. The general lesson from thiswork is that adjustment of household produc-tion is an important way through which fami-lies in poor countries respond to economicshocks. Smith et al. (2002) and ElizabethFrankenberg, Smith, and Duncan Thomas(2003) show that Indonesian families copedwith the 1998 crisis by increasing their within-household production. Along the same lines,Edmonds and Pavcnik (2006) find that house-holds allocated time away from householdproduction toward wage work following therice market liberalization in Vietnam.Determining how these adjustments affectinequality remains a topic for future research.

Household consumption is equally impor-tant as a channel through which globalization



may impact inequality. Most internationaltrade models assume that individuals haveidentical and homothetic preferences. Inthese models, trade-policy-induced changesin relative prices of goods change the con-sumption of individuals with differentincomes in proportional terms; as a result,trade does not affect people’s relative posi-tion in the welfare distribution through theconsumption channel. However, a large lit-erature in development economics hasshown that poorer households devote a dis-proportionately large share of their house-hold expenditures to basic items such asfood. To the extent that household con-sumption depends on the relative positionof households in the welfare distribution,globalization-induced price changes mayaffect inequality through consumption.Furthermore, the increased availability andlower prices of traded goods may shiftdemand away from nontradable services(e.g., household services, such as housekeep-ing, cooking, etc.) toward tradable goods(washing machines, dryers, microwaves,etc.), further depressing the earningprospects of the poor.

The consumption channel has beenlargely ignored in empirical work for thereasons discussed in detail in section 2.2 ofthe paper. Porto (2006) is the only studythat explicitly considers how trade policyaffects the welfare distribution throughconsumption.25 As we discussed in section4, he examines the implications of theArgentinean trade reform for the distribu-tion of household welfare using a generalequilibrium framework. Porto’s analysisyields two interesting insights. First, hismodel implies that the structure of theMercosur-induced tariff cuts translated intoincreases in the prices of relatively low-skill-labor-intensive goods such as food and bev-erages. These goods have a larger share in

25 To the extent that consumption responses to tradereform differ across districts in India, this channel is alsocaptured by Topalova (2004a).

the budget of households in the bottom tailof the welfare distribution. Second, hismodel also implies that changes in the pricesof traded goods lowered (through generalequilibrium effects) the prices of nontradedgoods such as health, education, and leisuregoods, which are consumed in greater pro-portion by the rich. Consequently, the con-sumption channel implied an increase ininequality in the case of Argentina’s entryinto Mercosur. Although these findings aresubject to the same caveats discussed earlierin section 4, Porto’s study nicely illustratesthe importance of the consumption channel.Furthermore, the pattern of predicted pricechanges serves as a reminder that it is impos-sible to make general statements about theimpact of trade liberalization on inequality,as the effects depend crucially on thespecifics of the reform in question, in partic-ular the structure of tariff changes acrossindustries.

Interestingly, at the end of the study Portoconcludes that the impact of theArgentinean trade reforms on inequality viathe consumption channel was substantiallysmaller in magnitude than its impactthrough the labor income channel. Portoattributes the difference in the magnitude ofthe two effects to the underlying assumptionof perfect factor mobility and the associatedmagnification theorem that states thatchanges in relative goods prices generatemore than proportional changes in factorprices. Based on Porto’s results, it is tempt-ing to conclude that the usual neglect of theconsumption channel in the globalizationand inequality debate may not be a firstorder concern. However, more work needsto be done to establish whether his findingsgeneralize.

6. Conclusions

The substantial amount of evidence wereviewed in this article suggests a contempo-raneous increase in globalization and



inequality in most developing countries.However, establishing a causal link betweenthese two trends has proven more challeng-ing. Despite the ambiguities involved inidentifying the relationship between open-ness and distributional changes, it seems fairto say that the evidence has provided littlesupport for the conventional wisdom thattrade openness in developing countrieswould favor the less fortunate (at least inrelative terms).

Our survey has identified several channelsthat may explain why the recent experienceof developing countries did not conform tothe “naive” thinking about globalization. Ourunderstanding of the consequences of glob-alization for inequality has improved as theconceptual framework used in empiricalwork expanded to include trade in interme-diate products, international flows of capital,trade-induced skilled biased technologicalchange, short-run factor immobility, and firmheterogeneity.

Overall, there is little support for thepremise that adjustment to changing eco-nomic conditions would occur through laborreallocation from declining to growing sec-tors of the economy, at least at the aggregateindustry level usually considered in tradi-tional international trade models of compar-ative advantage. A common finding ofstudies of the effects of trade reforms indeveloping countries is the lack (or small mag-nitude) of sectoral labor reallocation (althoughit is possible that there is reallocation acrossfirms within sectors that is not visible at therelatively high level of aggregation used inlabor market surveys).26 In some instances,the data also suggest that the wage responseto trade barrier reduction is more pro-nounced than the employment response.

While these findings are subject to manycaveats—the high level of industry aggrega-tion being the perhaps most importantone—the cumulative evidence seems to

26 Recent evidence on constrained spatial mobility indeveloping countries is also in line with these findings.

point to constrained labor mobility as oneplausible explanation for the lack of sectoralreallocation. Indeed, the strict labor marketregulation that many developing countrieshad in place prior to the recent reforms is apotential source of labor market rigidities.The importance of these rigidities is likely todiminish in the long run, especially sincemany developing countries have by now sig-nificantly liberalized their labor markets.Still, from an empirical point of view, the dis-tinction between short- and long-run hasalways been elusive. We have surprisingly lit-tle knowledge as to how long it takes aneconomy to adjust to external shocks, andwhat time frames we should use in practicewhen we consider the short- versus long-runeffects of particular policies.

The lack of sectoral reallocation could alsoreflect that most of the adjustment to tradereform occurs within industries, but at alevel of detail that cannot be detected in thehousehold or firm level surveys usually usedin this line of work. Our survey highlightsseveral globalization-based explanations forthe increased relative demand for more edu-cated workers within industries. In somecases, trade reforms liberalized in additionto goods flows, factor flows, most important-ly capital, that may have generated addition-al demand for skilled workers. In otherinstances, globalization affected not onlytrade in final goods, but also and foremosttrade in intermediate goods that from thedeveloping country perspective were skill-intensive. Even in those cases where liberal-ization was concentrated on final goods, thehighest trade barrier reductions were oftenconcentrated—contrary to conventionalwisdom—on low-skill sectors that had origi-nally enjoyed a higher level of protection.Technological change that favored skilledworkers may have interacted with tradereforms to further depress the demand forlow-skilled workers. Increased exposure tocurrency fluctuations boosted exports fromdeveloping countries in some cases and pro-vided incentives to upgrade the product-mix



of their domestic plants. These composition-al changes may have fostered a qualityupgrading of plants that further contributedto the widening of the wage gap betweenskilled and unskilled.

Overall, it appears that the particularmechanisms through which globalizationaffected inequality are country, time, andcase specific; that the effects of trade liberal-ization need to be examined in conjunctionwith other concurrent policy reforms; andthat implementation details of particularpolicies matter. This conclusion may seemdisappointing, as it offers no simple predic-tions regarding the distributional impact ofglobalization and, hence, no straightforwardrecipe for remedial measures to alleviatepotentially adverse impacts. Yet, it is hardlysurprising given the heterogeneity of coun-tries, reforms, and overall globalizationexperience within the developing world.

Finally, we should emphasize that most ofthe existing evidence refers to narrow meas-ures of inequality, such as the skill premiumor wage inequality. Broader concepts ofinequality that focus on consumption andgeneral well-being have received substantial-ly less attention. The very scant evidence thatexists on these issues however seems to sug-gest that the labor market effects of globaliza-tion dominate its effects on consumptionthrough relative price changes, so perhaps thefocus on wages alone is not as limiting as onewould have thought. Along the same lines, weknow surprisingly little about one of the mostfrequently voiced concerns regarding global-ization: its potential to generate transitionalunemployment that might disproportionatelyaffect less skilled individuals. It would betempting to characterize these open ques-tions as areas of future research, but the truthis that the same factors that have inhibitedresearch on these topics in the past (lack ofappropriate data being the primary one) arelikely to do the same in the future. The mostpressing research priority in this regard isthe collection of additional data and theimprovement of existing collection methods.

As the nature of globalization keepschanging, the channels through which thedistribution of resources within countries isaffected changes too, and so does theresearch agenda investigating the relation-ship between globalization and inequality. Inrecent years, it has become increasinglyapparent that trade is more than the flow ofgoods between countries as traditionallymodeled in international trade theory. Traderepresents exchange between firms that arelocated in different countries. As traditionalcross-border restrictions are disappearing,the focus of the analysis is shifting from thecountry to the firm, as the relevant unit ofobservation. Accordingly, questions, such aswhat type of firms produce what goods andfor which markets, which firms export andwhich ones produce for the domestic mar-ket, what are the characteristics of workersemployed by different types of firms, etc.,are becoming more prominent in the litera-ture. Mechanisms that emphasize composi-tional effects of globalization, qualityupgrading in response to intensified importcompetition from lower-income countries or,alternatively, to higher export demand bymore developed economies, and reallocationof resources across firms or plants within asector, or even across products of differentquality within a firm, seem more relevant todeveloping countries these days. The mainchallenge facing the empirical literature inthis area is that the heterogeneity of firms,plants, products, and workers emphasized inthe theoretical arguments implies the needfor highly disaggregate data. Such data aretypically available for plants and contain fair-ly detailed information on many plant char-acteristics, including occasionally theirproduct lines. However, what is missingfrom such data sets is information on thecharacteristics of the workers employed byeach plant/firm, which is the crucial stepneeded for establishing a connection to dis-tributional questions. Hence, we do notknow for sure whether plants that are moreproductive employ better educated workers;



or whether the production of higher qualityproducts requires a more skilled labor force;or whether changes in the product mix orproduct quality are accompanied by changesin the characteristics and compensation ofthe workforce. It is these kinds of questionsthat future research will need to address inorder to provide insight into how ongoingglobalization will impact inequality indeveloping countries.

REFERENCES

Acemoglu, Daron. 2003. “Patterns of Skill Premia.”Review of Economic Studies, 70(2): 199–230.

Aghion, Philippe, Robin Burgess, Stephen Redding,and Fabrizio Zilibotti. 2005. “Entry Liberalizationand Inequality in Industrial Performance.” Journal ofthe European Economic Association, 3(2–3):291–302.

Anderson, James E., and Eric van Wincoop. 2004.“Trade Costs.” Journal of Economic Literature,42(3): 691–751.

Attanasio, Orazio, Pinelopi Koujianou Goldberg, andNina Pavcnik. 2004. “Trade Reforms and WageInequality in Colombia.” Journal of DevelopmentEconomics, 74(2): 331–66.

Attanasio, Orazio, and Miguel Székely. 2000.“Household Saving in East Asia and Latin America:Inequality Demographics and All That.” In AnnualWorld Bank Conference on Development Economics2000, ed. B. Pleskovic and N. Stern. Washington,D.C.: World Bank.

Banerjee, Abhijit, and Thomas Piketty. 2004. “Are theRich Growing Richer? Evidence from Tax Data.” InData and Dogma: The Great Indian Poverty Debate,ed. A. Deaton and V. Kozel. Houndmills: MacmillanPress.

Banerjee, Abhijit, and Thomas Piketty. 2005. “TopIndian Incomes, 1922–2000.” World Bank EconomicReview, 19(1): 1–20.

Behrman, Jere R., Nancy Birdsall, and Miguel Székely.2000. “Economic Reform and Wage Differentials inLatin America.” IADB Working Paper, no. 435.

Berman, Eli, John Bound, and Stephen Machin. 1997.“Implications of Skill-Biased Technological Change:International Evidence.” NBER Working Papers,no. 6166.

Bernard, Andrew B., and J. Bradford Jensen. 1997.“Exporters, Skill Upgrading, and the Wage Gap.”Journal of International Economics, 42(1–2): 3–31.

Bernard, Andrew B., and J. Bradford Jensen. 2003.“Firm Structure, Multinationals, and ManufacturingPlant Deaths.” Unpublished.

Bernard, Andrew B., J. Bradford Jensen, and Peter K.Schott. 2005. “Trade Costs, Firms, and Productivity.”Unpublished.

Bernard, Andrew B., and Fredrik Sjoholm. 2003.“Foreign Owners and Plant Survival.” NBERWorking Papers, no. 10039.

Beyer, Harald, Patricio Rojas, and Rodrigo Vergara.1999. “Trade Liberalization and Wage Inequality.”

Journal of Development Economics, 59(1): 103–23.Branstetter, Lee, and Nicholas Lardy. 2006. “China’s

Embrace of Globalization.” NBER Working Papers,no. 12373.

Chiquiar, Daniel. 2004. “Globalization, Regional WageDifferentials, and the Stolper–Samuelson Theorem:Evidence from Mexico.” Bank of Mexico WorkingPaper, no. 2004-06.

Cragg, Michael Ian, and Mario Epelbaum. 1996. “WhyHas Wage Dispersion Grown in Mexico? Is It theIncidence of Reforms or the Growing Demand forSkills?” Journal of Development Economics, 51(1):99–116.

Currie, Janet, and Ann E. Harrison. 1997. “Sharing theCosts: The Impact of Trade Reform on Capital andLabor in Morocco.” Journal of Labor Economics,15(3): S44–71.

Deaton, Angus. 1997. The Analysis of HouseholdSurveys: A Microeconometric Approach toDevelopment Policy. Baltimore and London: JohnsHopkins University Press.

Deaton, Angus. 2003. “Measuring Poverty in aGrowing World (or Measuring Growth in a PoorWorld).” NBER Working Papers, no. 9822.

Deaton, Angus and Jean Dreze. 2002. “Poverty andInequality in India: A Reexamination.” Economicand Political Weekly, 3729–48.

Dornbusch, Rudiger, and Sebastian Edwards. 1994.“Exchange Rate Policy and Trade Strategy.” In TheChilean Economy: Policy Lessons and Challenges,ed. B. P. Bosworth, R. Dornbusch, and R. Laban.Washington, D.C.: Brookings Institution, 81–104,112–15.

Edmonds, Eric V., and Nina Pavcnik. 2006. “TradeLiberalization and the Allocation of Labor betweenHouseholds and Markets in a Poor Country.” Journalof International Economics, 69(2): 272–95.

Erdem, Erkan, and James R. Tybout. 2004. “TradePolicy and Industrial Sector Responses in theDeveloping World: Interpreting the Evidence.” InBrookings Trade Forum 2003, ed. S. M. Collins andD. Rodrik. Washington, D.C.: Brookings InstitutionPress, 1–43.

Feenstra, Robert C., and Gordon H. Hanson. 1996.“Foreign Investment, Outsourcing, and RelativeWages.” In The Political Economy of Trade Policy:Papers in Honor of Jagdish Bhagwati, ed. R. C.Feenstra, G. M. Grossman, and D. A. Irwin.Cambridge and London: MIT Press, 89–127.

Feenstra, Robert C., and Gordon H. Hanson. 1997.“Foreign Direct Investment and Relative Wages:Evidence from Mexico’s Maquiladoras.” Journal ofInternational Economics, 42(3–4): 371–93.

Feenstra, Robert C., and Gordon H. Hanson. 1999.“The Impact of Outsourcing and High-TechnologyCapital on Wages: Estimates for the United States,1979–1990.” Quarterly Journal of Economics,114(3): 907–40.

Feenstra, Robert C., and Gordon H. Hanson. 2003.“Global Production Sharing and Rising Inequality: ASurvey of Trade and Wages.” In Handbook ofInternational Trade, vol. 1, ed. E. K. Choi and J.Harrigan. Malden, Mass: Blackwell, 146–85.

Feliciano, Zadia M. 2001. “Workers and Trade



Liberalization: The Impact of Trade Reforms inMexico on Wages and Employment.” Industrial andLabor Relations Review, 55(1): 95–115.

Fernandes, Ana M. Forthcoming. “Trade Policy, TradeVolumes and Plant-Level Productivity in ColombianManufacturing Industries.” Journal of InternationalEconomics.

Ferreira, Francisco H. G., and Julie A. Litchfield. 1999.“Calm after the Storms: Income Distribution andWelfare in Chile, 1987–94.” World Bank EconomicReview, 13(3): 509–38.

Frankenberg, Elizabeth, James P. Smith, and DuncanThomas. 2003. “Economic Shocks, Wealth, andWelfare.” Journal of Human Resources, 38(2):280–321.

Galiani, Sebastian, and Pablo Sanguinetti. 2003. “TheImpact of Trade Liberalization on Wage Inequality:Evidence from Argentina.” Journal of DevelopmentEconomics, 72(2): 497–513.

Gasparini, Leonardo. 2003. “Different Lives: Inequalityin Latin America and the Caribbean.” Unpublished.

Gasparini, Leonardo. 2004. “Argentina’s DistributionalFailure.” Unpublished.

Goldberg, Pinelopi Koujianou, and Nina Pavcnik.2003. “The Response of the Informal Sector to TradeLiberalization.” Journal of Development Economics,72(2): 463–96.

Goldberg, Pinelopi Koujianou, and Nina Pavcnik.2004. “Trade, Inequality, and Poverty: What Do WeKnow? Evidence from Recent Trade LiberalizationEpisodes in Developing Countries.” In BrookingsTrade Forum 2004, ed. S. M. Collins and C. Graham.Washington, D.C.: Brookings Institution Press,223–69.

Goldberg, Pinelopi Koujianou, and Nina Pavcnik.2005. “Trade, Wages, and the Political Economy ofTrade Protection: Evidence from the ColombianTrade Reforms.” Journal of International Economics,66(1): 75–105.

Goldberg, Pinelopi Koujianou, and Nina Pavcnik.Forthcoming. “The Effects of the Colombian TradeLiberalization on Urban Poverty.” In Globalizatonand Poverty, ed. A. Harrison. Chicago: ChicagoUniversity Press; Washington, D.C.: BrookingsInstitution Press.

Green, Francis, Andy Dickerson, and Jorge SabaArbache. 2001. “A Picture of Wage Inequality andthe Allocation of Labor through a Period of TradeLiberalization: The Case of Brazil.” WorldDevelopment, 29(11): 1923–39.

Grossman, Gene M. 1984. “International Competitionand the Unionized Sector.” Canadian Journal ofEconomics, 17(3): 541–56.

Grossman, Gene M. 1986. “Imports as a Cause ofInjury: The Case of the U.S. Steel Industry.” Journalof International Economics, 20(3–4): 201–23.

Haltiwanger, John, Adriana Kugler, Maurice Kugler,Alejandro Micco, and Carmen Pages. 2004. “Effectsof Tariffs and Real Exchange Rates on JobReallocation: Evidence from Latin America.”Journal of Policy Reform, 7(4): 191–208.

Hanson, Gordon H. 2004. “What Has Happened toWages in Mexico since NAFTA? Implications forHemispheric Free Trade.” In Integrating the

Americas: Ftaa and Beyond, ed. A. Estevadeordal etal. Cambridge: Harvard University David RockefellerCenter for Latin American Studies, 505–37.

Hanson, Gordon H. Forthcoming. “Globalization,Labor Income, and Poverty in Mexico.” InGlobalizaton and Poverty, ed. A. Harrison. Chicago:Chicago University Press; Washington, D.C.:Brookings Institution Press.

Hanson, Gordon H., and Ann E. Harrison. 1999.“Trade Liberalization and Wage Inequality inMexico.” Industrial and Labor Relations Review,52(2): 271–88.

Harrison, Ann E. 1994. “Productivity, ImperfectCompetition and Trade Reform: Theory andEvidence.” Journal of International Economics,36(1–2): 53–73.

Harrison, Ann E., and Gordon H. Hanson. 1999. “WhoGains from Trade Reform? Some RemainingPuzzles.” Journal of Development Economics, 59(1):125–54.

Harrison, Ann E., and Jason Scorse. 2004a. “Moving upor Moving Out? Anti-Sweatshop Activists and LaborMarket Outcomes.” NBER Working Papers, no.10492.

Harrison, Ann E., and Jason Scorse. 2004b. “TheImpact of Globalization on Compliance with LaborStandards: A Plant-Level Study.” In Brookings TradeForum 2003, ed. S. M. Collins and D. Rodrik.Washington, D.C.: Brookings Institution Press,45–96.

Hasan, Rana, Devashish Mitra, and K. V. Ramaswamy.Forthcoming. “Trade Reforms, Labor Regulations,and Labor-Demand Elasticities: Empirical Evidencefrom India.” Review of Economics and Statistics.

Hay, Donald A. 2001. “The Post-1990 Brazilian TradeLiberalisation and the Performance of LargeManufacturing Firms: Productivity, Market Shareand Profits.” Economic Journal, 111(473): 620–41.

Head, Keith, and John Ries. 2002. “OffshoreProduction and Skill Upgrading by JapaneseManufacturing Firms.” Journal of InternationalEconomics, 58(1): 81–105.

Heckman, James J., and Carmen Pages. 2000. “TheCost of Job Security Regulation: Evidence fromLatin American Labor Markets.” NBER WorkingPapers, no. 7773.

Hsieh, Chang-Tai, and Keong T. Woo. 2005. “TheImpact of Outsourcing to China on Hong Kong’sLabor Market.” American Economic Review, 95(5):1673–87.

Kaplan, David S., and Eric A. Verhoogen. 2005.“Exporting and Individual Wage Premia: Evidencefrom Mexican Employer–Employee Data.” Unpub-lished.

Keller, Wolfgang, and Stephen R. Yeaple. 2003.“Multinational Enterprises, International Trade, andProductivity Growth: Firm-Level Evidence from theUnited States.” NBER Working Papers, no. 9504.

Kijima, Yoko. 2006. “Why Did Wage InequalityIncrease? Evidence from Urban India 1983–99.”Journal of Development Economics, 81(1): 97–117.

Kim, Euysung. 2000. “Trade Liberalization andProductivity Growth in Korean ManufacturingIndustries: Price Protection, Market Power, and



Scale Efficiency.” Journal of DevelopmentEconomics, 62(1): 55–83.

Krebs, Tom, Pravin Krishna, and William Maloney.2005. “Trade Policy, Income Risk, and Welfare.”Unpublished.

Krishna, Pravin, and Devashish Mitra. 1998. “TradeLiberalization, Market Discipline and ProductivityGrowth: New Evidence from India.” Journal ofDevelopment Economics, 56(2): 447–62.

Krishna, Pravin, Devashish Mitra, and Sajjid Chinoy.2001. “Trade Liberalization and Labor DemandElasticities: Evidence from Turkey.” Journal ofInternational Economics, 55(2): 391–409.

Krueger, Anne O., Hal B. Lary, Terry Monson, andNarongchai Akrasanee, eds. 1981. Trade andEmployment in Developing Countries. Vol. 1.Individual Studies. Chicago: University of ChicagoPress.

Leamer, Edward E. 1998. “In Search ofStolper–Samuelson Effects on U.S. Wages.” InExports, Imports and the American Worker, ed. S.M. Collins. Washington, D.C.: Brookings InstitutionPress, 141–214.

Levinsohn, James. 1993. “Testing the Imports-as-Market-Discipline Hypothesis.” Journal of Inter-national Economics, 35(1–2): 1–22.

Lorentowicz, Andzelika, Dalia Marin, and AlexanderRaubold. 2005. “Is Human Capital Losing fromOutsourcing? Evidence for Austria and Poland.”University of Munich Discussion Paper, no. 2005-22.

Machin, Stephen, Annette Ryan, and John VanReenen. 1996. “Technology and Changes in SkillStructure: Evidence from an International Panel ofIndustries.” IFS Working Papers, no. W96/06.

Marcouiller, Douglas, Veronica Ruiz de Castilla, andChristopher Woodruff. 1997. “Formal Measures ofthe Informal-Sector Wage Gap in Mexico, ElSalvador, and Peru.” Economic Development andCultural Change, 45(2): 367–92.

Marin, Dalia. 2006. “A New International Division ofLabor in Eastern Europe: Outsourcing andOffshoring to Eastern Europe.” Journal of theEuropean Economic Association, 4(2–3): 612–22.

McLaren, John, and Andrew Newman. 2002.“Globalization and Insecurity.” Unpublished.

Melitz, Marc J. 2003. “The Impact of Trade on Intra-Industry Reallocations and Aggregate IndustryProductivity.” Econometrica, 71(6): 1695–1725.

Mishra, Prachi, and Utsav Kumar. 2005. “TradeLiberalization and Wage Inequality: Evidence fromIndia.” IMF Working Papers, no. 05/20.

Mistiaen, Johan, and Martin Ravallion. 2003. “SurveyCompliance and the Distribution of Income.” WorldBank Policy Research Working Paper Series, no.2956.

Muendler, Marc-Andreas. 2004. “Trade, Technology,and Productivity: A Study of BrazilianManufacturers, 1986–1998.” Unpublished.

Munshi, Kaivan, and Mark R. Rosenzweig. 2005.“Why Is Mobility in India so Low? SocialInsurance, Inequality, and Growth.” CID WorkingPaper, no. 121.

Naito, Hisahiro. 1999. “Re-examination of UniformCommodity Taxes under a Non-linear Income Tax

System and Its Implication for ProductionEfficiency.” Journal of Public Economics, 71(2):165–88.

Neary, J. Peter. 1978. “Short-Run Capital Specificityand the Pure Theory of International Trade.”Economic Journal, 88(351): 488–510.

Neary, J. Peter. 1982. “Intersectoral Capital Mobility,Wage Stickiness, and the Case for AdjustmentAssistance.” In Import Competition and Response,ed. J. N. Bhagwati. Chicago: University of ChicagoPress, 39–67.

Nicita, Alessandro. 2004. “Who Benefited from TradeLiberalization in Mexico? Measuring the Effects onHousehold Welfare.” World Bank Policy ResearchWorking Paper Series, no. 3265.

Pavcnik, Nina. 2002. “Trade Liberalization, Exit, andProductivity Improvement: Evidence from ChileanPlants.” Review of Economic Studies, 69(1): 245–76.

Pavcnik, Nina. 2003. “What Explains Skill Upgradingin Less Developed Countries?” Journal ofDevelopment Economics, 71(2): 311–28.

Pavcnik, Nina, Andreas Blom, Pinelopi KoujianouGoldberg, and Norbert Schady. 2004. “TradeLiberalization and Industry Wage Structure:Evidence from Brazil.” World Bank EconomicReview, 18(3): 319–44.

Porto, Guido G. 2006. “Using Survey Data to Assessthe Distributional Effects of Trade Policy.” Journalof International Economics, 70(1): 140–60.

Revenga, Ana L. 1997. “Employment and WageEffects of Trade Liberalization: The Case of MexicanManufacturing.” Journal of Labor Economics, 15(3):S20–43.

Revenga, Ana L. 1992. “Exporting Jobs? The Impact ofImport Competition on Employment and Wages inU.S. Manufacturing.” Quarterly Journal ofEconomics, 107(1): 255–84.

Robbins, Donald J. 1996. “Evidence on Trade andWages in the Developing World.” OECD TechnicalPaper, no. 119.

Robertson, Raymond. 2000. “Trade Liberalisation andWage Inequality: Lessons from the MexicanExperience.” World Economy, 23(6): 827–49.

Robertson, Raymond. 2004. “Relative Prices and WageInequality: Evidence from Mexico.” Journal ofInternational Economics, 64(2): 387–409.

Rodriguez-Clare, Andres. 2001. “Costa Rica’sDevelopment Strategy Based on Human Capital andTechnology: How It Got There, the Impact of Intel,and Lessons for Other Countries.” Journal of HumanDevelopment, 2(2): 311–24.

Rodrik, Dani. 1997. Has Globalization Gone Too Far?Washington, D.C.: Institute for InternationalEconomics.

Rodrik, Dani. 1998. “Why Do More Open EconomiesHave Bigger Governments?” Journal of PoliticalEconomy, 106(5): 997–1032.

Rosenzweig, Mark R. 1988. “Labor Markets in Low-Income Countries.” In Handbook of DevelopmentEconomics, vol. 1, ed. H. Chenery and T. N. Srinivasan.San Diego and Amsterdam: Elsevier Science.

Sanchez-Paramo, Carolina, and Norbert Schady. 2003.“Off and Running? Technology, Trade, and theRising Demand for Skilled Workers in Latin



America.” World Bank Policy Research WorkingPaper Series, no. 3015.

Scheve, Kenneth, and Matthew Slaughter. 2002.“Economic Insecurity and the Globalization ofProduction.” NBER Working Papers, no. 9339.

Schott, Peter K. 2004. “Across-Product versus within-Product Specialization in International Trade.”Quarterly Journal of Economics, 119(2): 647–78.

Smith, James P., Duncan Thomas, ElizabethFrankenberg, Kathleen Beegle, and Graciela Teruel.2002. “Wages, Employment and Economic Shocks:Evidence from Indonesia.” Journal of PopulationEconomics, 15(1): 161–93.

Székely, Miguel, and Marianne Hilgert. 1999. “What’sBehind the Inequality We Measure: An InvestigationUsing Latin-American Data.” IADB Working Paper,no. 409.

Tarr, David, and Steven J. Matusz. 1999. “Adjusting toTrade Policy Reform.” World Bank Policy ResearchWorking Paper Series, no. 2142.

Tendulkar, Suresh D. 2003. “Organised Labour Marketin India Pre and Post Reform.” Unpublished.

Thoenig, Mathias, and Thierry Verdier. 2003. “ATheory of Defensive Skill-Biased Innovation andGlobalization.” American Economic Review, 93(3):709–28.

Topalova, Petia. 2004a. “Factor Immobility andRegional Impacts of Trade Liberalization: Evidenceon Poverty and Inequality from India.” Unpublished.

Topalova, Petia. 2004b. “Trade Liberalization and FirmProductivity: The Case of India.” IMF Working

Papers, no. 04/28.Tybout, James R. 2003. “Plant- and Firm-Level

Evidence on “New” Trade Theories.” In Handbookof International Trade, vol. 1, ed. E. K. Choi and J.Harrigan. Malden, Mass: Blackwell, 388–415.

Verhoogen, Eric A. 2006. “Trade, Quality Upgradingand Wage Inequality in the Mexican ManufacturingSector.” Unpublished.

Wacziarg, Romain, and Jessica Seddon Wallack. 2004.“Trade Liberalization and Intersectoral LaborMovements.” Journal of International Economics,64(2): 411–39.

Wei, Shang-Jin, and Yi Wu. 2002. “Globalization andInequality without Differences in Data Definition,Legal System, and other Institutions.” Unpublished.

Winters, L. Alan, Neil McCulloch, and Andrew McKay.2004. “Trade Liberalization and Poverty: TheEvidence So Far.” Journal of Economic Literature,42(1): 72–115.

Wood, Adrian. 1995. “How Trade Hurt UnskilledWorkers.” Journal of Economic Perspectives, 9(3):57–80.

Wood, Adrian. 1999. “Openness and Wage Inequality inDeveloping Countries: The Latin American Challengeto East Asian Conventional Wisdom.” In MarketIntegration, Regionalism and the Global Economy, ed.R. Baldwin et al. Cambridge; New York andMelbourne: Cambridge University Press, 153–81.

Zhu, Susan Chun. 2005. “Can Product Cycles ExplainSkill Upgrading?” Journal of InternationalEconomics, 66(1): 131–55.

Date post:	04-Jul-2020
Category:	Documents
Upload:	others
View:	0 times
Download:	0 times

Distributional Effects of Globalization in Developing ...noy/362texts/inequality.pdf · 1....

Documents