ABSTRACT
Labor economists agree on the existence of persistent wage differentials by race and sex in the US labor market, but disagree on the source of these differentials. There are two dominant explanations of these wage differentials that are the basis for most empirical research on this topic. The first is that employers discriminate against women or minorities in some fashion that results in lower wages for them even when they are equally productive. The second is that women or minorities come to the market with productivity shortfalls. Typical empirical studies of these alternative explanations proceed by estimat-
ing regressions of log wages on race or sex variables controlling for variables that are proxies for human capital, which in turn is assumed to determine productivity.1
However, although wage differentials by race and sex usually survive such exercises (although they are smaller), their interpretation is not conclusive, because one can always argue that there are remaining productivity differences between, say, blacks and whites, which are not captured by the included proxies. Improving on this typical exercise, this chapter considers the relationship between starting wages, race, and sex, in a data set in which performance ratings are also available. The performance ratings afford the opportunity to better control for productivity differences than in the typical wage regression with human capital proxies. The chapter first documents that gaps in starting wages by race and sex persist even when we account for measured performance on the job. This difference in starting wages could reflect taste discrimination, as in
Becker’s seminal work (1957) in which minorities or women are paid less because of employers’ distastes for hiring from these groups. However, if employers base starting wages on expected productivity or performance, and average performance is lower for minority workers (as it is in these data), then these estimated differentials could reflect simple statistical discrimination. From a policy perspective, whether taste discrimination or statistical discrimination plays a major role in race and sex differences in wages is significant. If taste discrimination accounts for the unexplained lower wages of women and minorities, then antidiscrimination legislation may be the only appropriate response. On the other hand, if statistical discrimination is important, especially in conjunction with some
other factor that leads to group discrimination, then better means of assessing workers’ productivity-including apprenticeships, skill certification, job testing, etc.—may contribute to the reduction of discrimination at the individual or group level. The goal of differentiating between these two forms of discrimination is the reason this chapter focuses on starting wages. Starting wages can potentially reflect either of the two types of discrimination-taste discrimination and statistical discrimination-in a fairly simple way, and an analysis of starting wages leads to some straightforward tests of these alternative models of discrimination.
