Minding the Racial Wage Gap

For a long time, economists doubted that racial bias of white employers was a significant factor in explaining black workers’ lower wages relative to whites. New research revisits this issue and finds evidence as to why prejudice deserves more weight in the debate.

Do you think there should be laws against marriages between blacks and whites? If your party nominated a black candidate for president, would you vote for him if he were qualified for the job? Responses to these and other similar questions can give researchers a valuable look into the racial sentiments of individuals and to ultimately study how these feelings affect social and economic outcomes. In a recent paper entitled “Prejudice and Wages: An Empirical Assessment of Becker’s The Economics of Discrimination,” University of Chicago professors Kerwin Kofi Charles of the Harris School of Public Policy and Jonathan Guryan of the University of Chicago Booth School of Business use this type of survey data to analyze a rather controversial issue: whether racial preferences affect the pay of black relative to white workers and how those preferences are carried out.

That prejudice matters for determining wages may be seen by looking at the survey responses among whites in certain areas and the corresponding black-white wage gap in those areas. Charles and Guryan found, for instance, that blacks have much lower wages compared to whites in places where a higher fraction of whites report opposition to interracial marriage. The same is true in places where whites would not vote for a black president. These results suggest a strong link between prejudice and wages, but they don’t explain how exactly racial preferences determine the wage gap.

Nobel laureate and University of Chicago professor Gary Becker’s influential 1957 work on employer discrimination laid the foundation for understanding the relationship between racial bias and wages of blacks relative to whites. One of his model’s surprising insights is that the racial aversion of the most prejudiced white employer does not affect the difference in pay between black and white workers. This might seem counterintuitive if one expects that an area that is relatively more prejudiced, that is, shows a higher average level of prejudice, should have a larger wage gap.

Although discrimination has been an intensely studied topic in economics since the publication of Becker’s research, no one had actually tested his model’s predictions. Critics suggested that a wage difference based on prejudice cannot survive in the long run in a competitive labor market, casting doubt on whether racial preferences play a role at all alongside other factors that could determine the wage gap such as differences in skills. Charles and Guryan are the first to test the key implications of Becker’s model. Not only did the authors find overwhelming support for Becker’s predictions, but they also found that about one-quarter of the overall black-white wage gap is due to racial prejudice.

Meet the Marginal Discriminator

Becker defines prejudice as a distaste or aversion for cross racial contact. His model says that if employers are white and white employers have a choice between black and white workers who are perfect substitutes, then prejudice would make an employer behave as if hiring a black worker is more expensive than it really is. This is because a prejudiced employer would have to be compensated for interacting with blacks. Put differently, a prejudiced employer would hire black workers only if blacks receive a sufficiently lower wage than whites in order to make up for his discomfort. This suggests that the difference in wage between black and white workers will depend on just how strong an employer’s racial bias is.

However, employers will surely have varying degrees of prejudice. Some may hate black workers, but others may merely dislike them. Whose racial preferences will actually matter for determining the wage gap?

Imagine that all employers were lined up according to their level of prejudice, with those on the right more prejudiced than those on the left. Those at the right end of the distribution– the employers who intensely dislike blacks–may not hire black workers at all because they consider the cost of interacting with them to be too high. Therefore, their racial bias should not matter for determining what black workers will be paid. Those at the left tail of the distribution, on the other hand, will more likely hire blacks because it is less costly for these relatively less prejudiced employers to do so. Ultimately, the wage gap will be determined by the most prejudiced employer of all employers who end up hiring a black worker.

This fascinating result implies that it is the racial preference of the “marginal discriminator” and not the average prejudice of all employers that matters for determining the wage gap. For instance, even if the most prejudiced employer in market A is 100 times more averse to blacks than the most prejudiced employer in market B, the black-white wage gap need not be bigger in A than in B. It all depends on the racial preference of that last employer who gave a black person a job. If for some reason that employer’s bias against black workers worsens in the future, then the wage gap will widen as well. Furthermore, if the number of black workers in a market increases relative to the number of white workers, then blacks will have to be sorted to ever more prejudiced employers along the distribution. The marginal employer–the one who determines the wage–will naturally be someone who is more averse to interacting with blacks. As a result, holding the level of prejudice constant, the difference in pay between blacks and whites will be bigger as the relative supply of black workers in a market increases. The wage gap could be zero even if there are many prejudiced employers in the market, as long as the supply of black workers is such that all of them can be hired by unprejudiced employers.

Criticisms and a Second Look

Becker’s predictions have been sharply criticized by many over the years, beginning with Becker himself, but more famously by another Nobel laureate, Kenneth Arrow of Stanford University. Arrow memorably remarked that Becker’s employer discrimination model “predicts the absence of the phenomenon it was designed to explain.” He argued that prejudiced employers in a competitive market will be forced to shut down in the long run because they sacrifice profits by discriminating. If there is no reason to believe that the productive abilities of black workers are inferior to white workers, then those employers whose decisions are not clouded by their dislike for blacks will stand to earn a greater profit because whites are more expensive to hire than blacks. Investors observe this difference in profit and will direct capital toward the most profitable companies. Over time, competition should ensure that prejudiced employers will be driven out of the market and the racial wage gap will disappear.

Such a criticism–and its force in the literature–may be one reason why it took some 50 years before Becker’s insights could be tested using real data. But recent work has shown that prejudice can lead to persistent wage gaps if there is some form of imperfect information or imperfect competition. Moreover, even under perfect competition, Charles and Guryan have shown in an earlier paper that racial wage gaps need not vanish in the long run.

They argued that it is quite possible for a prejudiced employer to stay open for business–even if he has to sacrifice profits–in order to enjoy isolation from a group he dislikes. After all, an employer who is contemplating shutting down knows that he will eventually have to get a job, and that job might entail interacting with blacks in his future workplace. “Implicit in the criticisms of Becker’s model is this unrealistic idea that prejudice is not portable across roles in the labor market,” explains Charles. “That if I’m prejudiced as an employer, I somehow miraculously become unprejudiced if I shut down and become a worker.”

Charles and Guryan recognized that there are enough good explanations out there for why racial wage gaps can survive in the long run that would justify testing Becker’s predictions. To check whether Becker’s insights hold up using real data, Charles and Guryan look at the prejudice of white individuals as estimated from survey responses and relate that to the black-white wage gap across states. They empirically analyzed the importance of the average and marginal levels of prejudice on the wage gap, and conducted a horse race between different points–the median, 10th, and 90th percentiles–on the prejudice distribution. The marginal level of prejudice is estimated by taking the pth percentile of the distribution, where p is the fraction of the state workforce that is black.

Just as Becker predicted, the study found that the wage gap is largely determined by the tastes of the marginal discriminator and not by the average prejudice of whites. The relative supply of black workers matters too; the wage gap widens with larger fractions of black workers in a state’s workforce. Moreover, the horse race revealed that the racial preferences of the relatively less prejudiced whites at the lower end of the distribution (the 10th percentile) have a large and negative effect on the wage gap, while that of the most prejudiced whites (the median and 90th percentiles) have no effect at all. The study’s results also can be used to compute the costs borne by black workers due to racial bias. If an 18-year-old black male, for example, were to choose between two states in which to work, Charles and Guryan estimate that by living in Florida rather than Massachusetts–the 75th and 25th percentiles of the marginal prejudice distribution, respectively—the value of his lifetime earnings would fall by about $34,000. The penalty is much bigger if he opts for Mississippi rather than Wisconsin, states that are at the 90th and 10th percentiles of the distribution, respectively. His earnings would drop by about $115,000.

Room for Prejudice

While the study clearly found an important role for racial prejudice in generating differences in wages between black and white workers, it also suggests that three-quarters of this gap is due to something else. For instance, firms may observe differences in skills between blacks and whites such that they sensibly decide to pay a lower wage to black workers. Another possibility is something called “statistical discrimination.” Employers don’t necessarily dislike blacks, but because they cannot properly assess the skills of black workers due to lack of information, they fall back on stereotypes and generalizations that lead them to discriminate against blacks in terms of wages. If the black-white wage gap were due either to skill differences or statistical discrimination, then a policy designed to equalize skill or make the assessment of the underlying skill easy and perfect for firms may have a chance at narrowing the gap. But if racial bias is as important as the study suggests, then it might be more difficult or at least require a different strategy to close that gap. This logic extends to prejudice toward other groups as well. “We say that even if test scores converged, or even if women do as well as their peers, there’s a possibility that the reason I observe a wage difference has nothing to do with skills at all,” Charles says.

The study by Charles and Guryan enriches the scholarly understanding of racial wage gaps by encouraging researchers to reexamine the role of racial preferences, including the extent to which it can influence the prevailing arguments. “It raises the interesting possibility that differences between racial groups are not exclusively due to skill and information differences, but might, in addition, be due to this thing called prejudice,” says Charles. “There has been a move in much of recent literature to discount that last possibility. Our results, at the very least, retard that discounting."

“Prejudice and Wages: An Empirical Assessment of Becker’s The Economics of Discrimination.” Kerwin Kofi Charles and Jonathan Guryan. Forthcoming in the Journal of Political Economy.