“What is the Impact of Foreign Trade on Race Based Discrimination in the Workplace?” Michael Shields May 2013 Abstract This research tests the impact of foreign trade on race based discrimination in compensation and employment in the workplace. Predicated on Gary Becker’s 1957 theory that competition will eradicate taste-based employer discrimination in the long run, as discriminatory firms paying effective wage premiums for preferred workers are driven from the market, this paper tests whether foreign trade serves as an effective approximation of competition. Because the opportunity cost of discrimination for firms depends on the slack in the labor market, increased trade may decrease costs to firms by creating labor surpluses in import competing industries. Thus the likely impact of trade on discrimination is ambiguous. Utilizing data on forty-one distinct manufacturing industries in the US, I find that for US manufacturing firms over the period from 2002 through 2006, increased trade was in fact associated with a reduction in median wage differentials between African Americans and whites. I wish to thank Dr. Francesco Renna for his patience and insight in sharing ideas with me over the course of this project. Also, many thanks to Dr. Steven Myers for helping me to develop the technical skills needed to complete this project and offering guidance along the way. Thanks to Dr. Kate Sheppard for introducing me to the formal tools used in evaluating discrimination. Thanks to Sam Hinkle for reading my work and offering ideas, and to Jeremy Hunter for offering troubleshooting advice. Finally, thanks to Dr. Steve Harp, whose support has helped me to navigate this project alongside many other academic demands, and whose guidance has been invaluable to me over the course of my academic career. 2 Contents Introduction……………………………………………………………………………………………………..……….3 Literature Review…………………………………………………………………………………………..………….8 Data…………………………………………………………………………………………………………….………….13 Theoretical Model……………………………………………………………………………………………...…….15 Empirical Model……………………………………………………………………………………………………...16 Estimation Results……………………………………………………………………………………….………….18 Conclusion and Recommendations for Further Research………………………………….………20 Bibliography……………………………………………………………………………………………………………21 Appendices Appendix A: Data……………………………………………………………………………….…………22 Appendix B: Estimation Results…………………………………………………………………….23 3 Introduction Economists disagree about the likely impact of trade on discrimination in wages and hiring against workers. Two theories from divergent subfields of economics can offer insights into the impact of trade on discrimination. The first is Becker’s (1957) seminal work. According to Becker, competition will tend to reduce workplace discrimination. To explain how, Becker evokes the notion of a “taste for discrimination” which leads some employers (as well as some consumers and employees) to behave “as if [they] were willing to pay something either directly or in the form of a reduced income, to be associated with some persons instead of others”(p.14). Becker calls this assumed cost the “Discrimination Coefficient. Instead of paying a wage, “W,” and hiring workers at the market rate without regard to race, discriminating employers incur a disutility “d” from associating with minorities, so that they perceive the cost of hiring minority workers to be not just “W,” but (W+d), and are willing to pay that amount to a white worker in order not to transact with African Americans against whom they harbor discriminatory preferences. In the case that they do hire a black worker, the wage they pay him or her will be lower than that of a comparable white worker, by a magnitude of d. Because firms with lower Discrimination Coefficients incur lower costs, Becker predicts that discrimination will decrease as more discriminatory firms are either driven from the market or forced to change their discriminatory hiring practices in order to remain competitive. 4 In assessing the likely impact of trade on discrimination, some economists have used trade as an approximation of competition. Thus with increased entry by foreign firms into their industry, firms are predicted to decrease discriminatory hiring and compensation practices. Yet Becker’s assessment may be extrapolated to account for a divergent outcome; to the extent that domestic firms are displaced by foreign competition, trade may in fact reduce the opportunity cost of discrimination for remaining firms, by swelling the labor supply in that industry with displaced workers and driving down the market wage. An employer faced with import competition and thus induced to cut labor costs by lowering wages may still find an abundant supply of preferred workers at a lower wage. Thus, with increased trade, minority workers may enjoy reduced discrimination, or may face still greater discrimination than before, depending on the specific nature of the industry in which they work and the skill set of the worker. The Hechscher-Ohlin-Samuelson comparative advantage trade theory sheds light on how trade will impact discriminatory practices in specific industries. The HO-S theory predicts that with increased trade firms will increasingly specialize in the production of goods and services in which their region has a comparative advantage. As they specialize, they will substitute away from the production factors utilized in the types of production which are diminishing, and toward those factors needed in production which is on the rise. This means that, for example, demand for low-skilled workers in countries specializing in skills intensive production will decrease, while demand for high-skilled workers will grow. While income differentials resulting from changes in demand for one’s skill set are legitimate, 5 market-based differentials, not discrimination, trade induced decreases in demand for one’s skill set also enable firms to discriminate more by decreasing the cost they incur to do so. Recall that in keeping with Becker’s theory, it is not competition per se which constrains firms in their ability to discriminate; it is opportunity cost. What the H-O-S theory indicates about industry specific costs is that firms engaged in import competing production, while facing increased incentives to cut costs, may also draw from larger labor supply pools at a lower equilibrium wage. Thus, their ability to discriminate may in fact expand. Those firms engaged in exporting on the other hand, will draw from a diminishing labor supply pool as more workers possessing the skill sets they need are hired in their expanding industry, prompting equilibrium wages to rise. This study assumes that US firms tend to specialize in skills intensive production, and that with increased trade, demand for high-skill workers will increase, while demand for low-skill workers will decrease. The implications for minority workers are clearly contingent on whether their skill sets match up with those in growing demand. Not only can such workers reap higher compensation based on skill itself but the fact that they are in tight labor markets constrains even prejudiced employers’ ability to discriminate against them. The opposite is true for workers whose skill sets are in decreasing demand in their region (low-skill workers, in the US case). For those workers, not only will their skill type reap diminished compensation, but slack in the labor market for their skills will mean that, barring legal repercussions, employers can discriminate against them with impunity. Important to note is that this study also assumes labor 6 market stratification, at least in the short run, as workers face time and cost constraints on their ability to transform their skill sets to enable them to compete in markets for job types in higher demand. Because African American workers comprise a disproportionate share of the low-skill workforce in the US, increased trade may exacerbate both the wage gap between themselves and their nonblack peers, and the residual wage gap, a proxy for discrimination, which they face. To the extent that “pre-market discrimination” in access to quality schooling exists, African American workers may face distinct challenges to overcoming their segmentation into low-skill sectors. This analysis assumes that African American workers comprise a disproportionately large share of low-skill workers. To the extent that blacks are clustered in low-skill jobs, there are market-based reasons predicated on human capital differentials. While these differentials have nothing to do with discrimination, how these individuals got to the labor market with skill sets inadequate to secure lucrative jobs to begin with may have a great deal to do with so-called pre-market discrimination. This matter, while important to keep in mind, is not the central focus of this study. In this study, I will test whether the residual wage gap between the races increases or decreases as a result of trade. In short, does increased trade result in diminished discrimination against blacks in the workforce, or does trade exacerbate such discrimination? 7 While the literature represents divergent approaches to assessing the likely impact of trade on women workers, the idea of a trade induced reduction in discrimination begs the obvious question of how racial minorities will be impacted. While Black and Brainerd can point to a trend of rising female relative wages, Altonji and Blank (1999) point out that a similar trend in evidence for African Americans in the 1960’s has come to a halt since then, and African American wages have stagnated below those of whites for decades. Based on the Becker Discrimination Model, I will test whether African American workers in trade impacted industries experienced lower rates of employment as a result of increased import penetration into the markets in which they worked. Literature Review Black and Brainerd (2004) devise a formal test to determine whether increased imports have driven the reduction in the gender wage gap. Their work is predicated on the Becker theoretical model, which argues that competition eliminates discrimination over time, since firms engaging in discrimination pay an effective premium to indulge their presumed “taste” for discrimination. According to Becker’s theory, increased competition should trim profit margins and constrain firms’ ability to pay this premium.1 They hypothesize that the wage differential in a It bears note that, while firms must pay an effective premium in order to indulge discriminatory tastes, this does not necessarily imply that the wages they pay, even to “preferred” workers are at or above market rates. While a competitive firm pays its workers a salary equivalent to the marginal revenue product it realizes from the last worker, a monopsonistic employer, possessing market power both as a seller in the goods market and as a buyer in the labor market, limits production to artificially inflate prices and hires fewer than an equilibrium number of workers. In turn it pays an “exploitive” wage and captures a share of the surplus that would, under competitive conditions, accrue to workers (and similarly captures consumer surplus in the goods market). Thus, the “wage premium” may simply mean that the firm is less exploitive of preferred workers than of minority 1 8 given industry will decrease as foreign firms enter the consumer market for that firm’s goods. To test this, they use personal income and demographic information from the CPS March demographic supplement and the 1980 and 1990 Censuses, combined with figures for 188 manufacturing industries from NBER Trade (1998) and Manufacturing (1996) Databases. They develop a difference in differences model in which firms are first classified into competitive and concentrated industries, then the gender wage gap for each is compared on the basis of whether firms experienced a trade shock.2 This methodology is employed in order to filter out the wage impact of non-discrimination related factors such as increased education for women, assumed to be the same across all classifications of firms. Postulating that concentrated firms are insulated from domestic competition, they predict that foreign entry will have a greater impact on wage differentials for concentrated firms. This is supported by the findings. For “concentrated” industries, Black and Brainerd found that a ten percent increase in import share led to a decline in the residual gender wage gap of about 6.6percent. These results are consistent with the Becker theory that competition will reduce discrimination over time. Kongar (2006) expresses concerns that intensified job competition among workers as a result of increased trade may actually enhance firms’ ability to discriminate. Utilizing a similar methodology to Black and Brainerd, she too finds a workers, not that preferred workers realize a premium in absolute terms. Based on Becker’s prediction that competition eliminates discrimination, it emerges that any firm engaged in discrimination must be shielded from competition, and thus free to engage in exploitive behavior of this nature. 2 Black and Brainerd define a “concentrated” industry as one with a four firm market share of at least 40%. 9 reduction in the wage gap in trade impacted industries. Yet, cautious to rule out discrimination on this basis, she notes another important dimension of the trade impact on work outcomes; that of relative employment both overall and in high and low skill jobs. Citing evidence of extensive job losses for low-skill workers, as the HO-S model predicts, Kongar notes that in US manufacturing industries, increased import competition has had a disparate impact on women workers. That is, while a human capital based H-O-S model would predict job losses for low-skill workers, Kongar finds that among low-skill workers, women have suffered the greatest job losses. She claims that women’s overall wages rose in these industries due to these job losses among the low-skilled, which disproportionately affected women, leaving the average female wage to rise due to the higher skill set of those women who remained on the job. Kongar utilizes a Becker theoretical model, and a difference in differences empirical model similar to that employed by Black and Brainerd, but incorporating a variable for technological investment. The key difference between her work and that of Black and Brainerd is that she regresses not just female earnings, but also share of employment. Kongar finds that, for import competing manufacturing firms, increased trade is in fact associated with a rise in relative female earnings, but this is offset by nearly as great a loss in female share of overall employment. Based on these differential findings, Kongar regards the trade impact on discrimination to be inconclusive. Utilizing Columbian firm data, Ederington et. al (2009) investigate key assumptions of the Becker model, and incorporate H-O-S theory to make predictions about future trends. Because Becker’s prediction of what drives decreases in 10 discrimination is predicated on opportunity cost, slack in labor markets is an important component. Based on the H-O-S model, demand for workers is higher in industries in which their countries specialize. Thus, workers in export industries should be in relatively high demand, and firms in these markets might be more constrained (face higher costs) in their ability to discriminate. Ederington et. al test to find whether women comprise a disproportionate share of the workforce in export markets, and find that in fact they do. This is important in that, based on the H-O-S model, it indicates that increased trade should decrease the residual wage gap between genders, since women are heavily represented in export competing industries which should see more demand with trade, thus increasing the opportunity cost of discrimination. Next, they test the Becker hypothesis that discriminating firms will be driven from the market as profit margins narrow, and profit maximizing firms obtain lower cost labor inputs by hiring a larger share of female workers. While they find evidence that firm exit did occur as a result of trade liberalization, they key question is whether those firms which exited comprised a disproportionate number of male workers, indicating that they were paying a wage premium to indulge their “taste for discrimination.” The findings do not support this hypothesis. In light of these results, Ederington et. al. pose an alternative mechanism whereby the Becker model may work. Instead of simply maintaining labor ratios that favored more costly male workers, firms might decide at the margin to forego their discriminatory preferences and hire a greater share of women workers. While this idea provides a plausible explanation for the study’s failure to establish a relationship between firm exit and male employment 11 concentration, empirical testing has not been done, and thus this possible explanation remains speculative in nature. The data come from a plant-level dataset produced by the Colombian Manufacturing census by DANE (National Statistical Institute) for the years 1977 through 1991, which comprised a sample of 6,035 plants in 1984, of which 3,760 remained in 1991. Increased trade competition is measured as industry specific tariff reductions. Berik et. al (2004) utilize an empirical model based on Black and Brainerd’s to test the trade impact on discrimination in Taiwan and South Korea. Taiwanese figures come from the DGBAS and International Economic Databank, while Korean data comes from the South Korean Census, Labor Ministry, and International Economic Databank. Their results indicate a reduction in female relative earnings. Both Taiwan and South Korea witnessed employment losses in low skill sectors as a result of import competition from lower wage countries. These troubling findings indicate that, in some cases, trade may actually increase workplace discrimination. Berik et. al call upon the H-O-S comparative advantage model to show why. Taiwan and South Korea, as net importers of manufactured goods, saw job losses in this sector with increased trade. With slack in the labor markets, the opportunity cost firms faced to discriminate actually fell, and with it fell women’s relative earnings. Menon and Rogers (2008) also argue that discrimination may be positively linked to increased foreign competition. In the case of India, they argue that women comprise a significant share of low skill workers who become increasingly vulnerable to discrimination with increased trade, since India specializes in high 12 skill exports amongst most of its trading partners, and import competition for low skill goods creates slack in the labor markets in which most women work. The most important contribution Menon and Rogers make to the literature is that they utilize a decomposition to isolate the residual wage gap from the aggregate wage gap. Because discrimination is traditionally proxied as the residual, that portion which is unexplained by gender specific differentials in skills, experience, or other human capital based determinants, they use an Oaxaca decomposition. They use data from the National Sample Survey Organization and construct their own concentration ratios.3 With increased trade, Menon and Rogers find that the residual share of the wage gap grew from 56.5percent in 1983 of the wage gap to 77.7percent by 2004. This unexplained wage gap rose 2.8percent per year, indicating that, for firms in India, increased exposure to foreign trade appears to have increased discrimination. Data Data is drawn from four sources. Race, income, and employment data are from the March CPS, utilizing years 2002 and 2006 as a comparison. These years are chosen to coincide with trade data figures through 2005 and have the advantage of being isolated from the recession that took place at the end of the decade. These figures update the data utilized in the Black and Brainerd and Kongar studies, which end in 1994. The shorter time-span is more conducive to the difference-indifferences methodology in that it limits the likelihood of endogeneity. Income data 3 Menon and Rogers’ concentration index is distinct from the Hirfindahl application utilized by Black and Brainerd in that it is a continuous rate, rather than a dummy variable “Concentrated,” predicated on whether the four firm share of an industry comprises at least 40%. 13 is annual income, deflated by the Consumer Price Index base year 1999. Mean and median wages by race are taken for each industry in both 2002 and 2006. Employment is approximated by multiplying weeks employed by usual hours worked. The trade variable is compiled by summing the value of production by industry from the Manufacturing Industry Productivity Database with the value of imports by industry from the NBER Trade Database, then dividing the foreign value figure by this aggregate figure to get a relative share. Figures are taken for 2002 and 2005, with the proportional difference used for the variable. Concentration data is from the Census of Manufacturers. Figures are taken at the four-digit NAICS code level for the beginning period 2002, then a dummy variable is created based on the Hirfindahl four firm 40 percent market share threshold. Because competition is expected to decrease discrimination, the Concentration variable is predicted to have a positive coefficient, indicating that concentrated firms have greater wage differentials than non-concentrated firms. ConcTrade is an interaction term indicating the impact of increased trade on an industry, given that that industry was concentrated as of 2002. Because nonconcentrated industries are assumed to face competition already, the impact of increased trade on them is anticipated to be negligible. Likewise, these firms are predicted to be less discriminatory. For concentrated industries on the other hand, increased trade may reduce profit margins and provide an incentive for firms to give up costly discrimination practices or be forced from the market by 14 nondiscriminating competitors. Thus, ConcenTrade is the primary variable of interest. Theoretical Model Based on the ambiguous nature of the trade impact on gender discrimination in the workplace, and the differential trends between women workers, who are closing the wage gap they face, and African Americans, who are not, I will test the impact of foreign trade on racial discrimination in compensation and employment. The theoretical model is predicated on Becker’s 1957 discrimination theory. In principle, competition, which is here approximated by trade, should decrease discrimination against workers as discriminating firms are driven out of their product markets or induced to change their discriminatory behavior in order to save on costs. The hypothesis maintains that in the case of trade, discrimination may not always be reduced because slackness in labor markets may shelter firms from market forces which should otherwise induce them to minimize factor costs in order to remain competitive. To serve as a guide for which industries are likely to see reduced discrimination, and which may see more, the H-O-S trade theory is incorporated. Based on that theory, I posit that employees who have skills in high demand in their region face less discrimination than those who have skill sets in decreasing demand. Note that this is distinct from market based skills selection per se. While it is true that workers whose skills are in high demand will realize higher wages, it is also likely that minority workers in those fields will face fewer instances 15 of discrimination, since firms face higher costs to discriminate when labor markets are tight. Empirical Model The model will be a difference of differences methodology similar to that used in the Black and Brainerd and Kongar studies. This methodology is useful in that, by making the assumption that factors such as education and workplace attachment are consistent across concentrated and non-concentrated industries, it isolates the portion of the wage gap, or employment gap respectively, attributable to discrimination. This method for isolating the residual is useful in that, by invokig the assumption that such factors are the same for both groups it effectively filters out those market based differentials which are unobservable.4 While this makes the model a desirable method, it bears note that this methodology is most commonly used in controlled tests, and it invokes an assumption of trade exogenaity which, if untrue, will limit the effectiveness of the procedure. An attempt is made at limiting trade endogenaity by focusing on a narrow range of time. This is particularly useful for filtering taste based differentials in commitment to ones career, since measuring utility is notoriously problematic. 4 16 Wage Regression: t(ln(wage)iw – ln(wage)ib) = + ttradei + conceni + (concen*ttradei) Where t ln(wage) is the change in logged wage of persons by industry, according to their race, from 2002 to 2006. Note that subscript “i” denotes industry, not individual. To create this figure, nominal wages as reported in the dataset were deflated using the Consumer Price Index to base year 1999. Individuals were sorted by year, then race, and finally NAICS code industry. From these group totals I took a mean, to be regressed in Model 1, and a median, to be regressed in Model 2. trade is the proportional increase or decrease in foreign share of the industry in which they worked. Concen is a dummy variable indicating whether the industry was concentrated as of the beginning period, 2002, based on the Hirfindahl 4 firm 40percent market share criterion. (concen*ttrade)I is an interaction term measuring the impact of trade change contingent on whether the firm was concentrated. Note that subscript “i” denotes industry, not individual. Employment Regression: t(ln(employ)iw – ln(employ)ib) = + ttradei + conceni + (concen*ttrade)i Where all right side variables are as above and the dependant variable t(ln(employ)iw – ln(employ)ib) is the difference in change of employment share from 2002 to 2006. 17 Results In the first model, “Change in Mean Wage Gap,” I regress the change in logged mean wages on Trade, Concentrated, and ConcenTrade. This model fails to find statistical significance for any variables, set aside ConcenTrade, the primary input of interest. With a very low R-Squared, this model proves not very useful in predicting employer behavior in terms of discrimination. This may be due to outliers in mean earnings, particularly for industries having few employees. Model 2, “Change in Median Wage Gap” regresses median earnings by industry on foreign trade share, industry concentration, and the interaction term for these. By taking the median, I attempt to filter outliers which may bias results. In this model, both industry concentration and increased trade in concentrated industries prove highly significant, at the one percent level. Both signs are appropriate and consistent with the expectations discussed in the model. Concentrated industries are associated with a 38 percent larger wage gap than nonconcentrated industries. Furthermore, the impact of trade on discriminatory wage practices emerges as clearly contingent on whether the industry was concentrated in 2002. While the general trade impact is insignificant, the impact of trade on concentrated firms in particular is significant. Concentrated firms exposed to foreign entry decreased their race wage differential by 215 percent as more than did firms not impacted by trade or not concentrated. That is, the wage gap fell by a magnitude of roughly double for such firms. 18 Model 3, “Change in Employment Hours Differential” indicates that increased trade causes a reduction in the employment differential between whites and African Americans. This model suggests that proportionally more African Americans are hired as trade increases. Incidentally, because neither the estimator for concentration, nor the interaction between concentration and trade is significant, the results do not provide conclusive evidence on the impact of trade on the relative employment of African Americans in firms which have been insulated from competition, the focus of the study. Conclusion and Recommendations for Further Research This study advances our knowledge of discrimination by drawing links between gender based discrimination and discrimination based on race. Based on my findings, increased trade in consumer goods markets may in fact incentivize firms which have been insulated from competition to reduce discriminatory practices going forward. This fact may hold promise for racial minorities working in those industries. This study is constrained by a lack of available data on industry concentration ratios for firms outside of the manufacturing sector. Also, the use of a difference in differences test invokes assumptions of exogenous trade shock which may prove a poor fit in a dynamic global economy. While this methodology has the advantage of employing a filter to cancel unmeasurable group characteristics which could impact incomes for reasons other than discrimination, such as employee attachment to the labor force, it does so only on the assumption that such 19 preferences do not differ on the basis of trade between members of the same demographic. For this reason, future research utilizing a decomposition style model could prove useful in corroborating these findings. 20 Bibliography Altonji, Joseph G., and Rebecca M. Blank. "Race and gender in the labor market." Handbook of labor economics 3 (1999): 3143-3259. Bartelsman Eric J. and Wayne Gray, “The NBER Manufacturing Productivity Database,” Technical Working Paper 205. National Bureau of Economic Research,1050 Massachusetts Avenue Cambridge, MA 02138, October 1996. http://www.nber.org/nberprod/t0205.pdf Becker, Gary. The Economics of Discrimination. The University of Chicago Press, Chicago, 1971. Berik, G., Rodgers, Y. v. d. M. and Zveglich, J. E. (2004), International Trade and Gender Wage Discrimination: Evidence from East Asia. Review of Development Economics, 8: 237–254. Black, S. "et Brainerd E.(2004).“Importing Equality? The Impact of Globalization on Gender Discrimination”." NBER Working Paper 9110. Ederington, Josh, Jenny Minier, and Kenneth Troske. "Where the girls are: trade and labor market segregation in Colombia." (2009). Kongar, Ebru. "Importing equality or exporting jobs? Competition and gender wage and employment differentials in US manufacturing." Competition and Gender Wage and Employment Differentials in US Manufacturing (January 2006) (2006). Menon, Nidhiya, and Yana van der Meulen Rodgers. "International trade and the gender wage gap: New evidence from India’s manufacturing sector." World Development 37.5 (2009): 965-981. Schott, Peter K. “U.S. Manufacturing Exports and Imports by SIC or NAICS Category and Partner Country, 1972 to 2005,” Yale School of Management and NBER, 2010. http://faculty.som.yale.edu/peterschott/files/research/data/sic_naics_trade_20100 504.pdf 21 Appendix A Data Table I 22 Appendix B Estimation Results Table II 23