Beginning with the work of Becker , economists have been concerned with the relationship between competitive market forces and the persistence of labor market discrimination. (1) A major insight of Becker and the literature that followed is that competition could tend to eliminate discriminatory pay differentials. If competitive forces do eliminate discrimination, then empirical findings of persistent discrimination are either a result of faulty testing or market imperfections.
Economic analysis of the link between competition and discrimination depends on the origins of prejudice. Becker identified at least three possible sources: employers, co-workers, and consumers. (2) In each of these instances, the group against which prejudice is directed (e.g., women or minorities, which will be termed "minorities" in this paper) may be able to escape discrimination. First, minorities can work for nondiscriminatory employers (if there are any). Moreover, Stiglitz  shows that with constant returns to scale, such employers will take over markets from the discriminators, eliminating any pay gap based on prejudice. (3) Second, if there is co-worker prejudice, then minorities can work in segregated firms which, under constant returns to scale, may have a cost advantage over integrated or all-majority firms. Again, discrimination will be driven away. (4) Third, minorities can theoretically escape customer discrimination by specializing in the production of goods for which there is no contact between producers and consumers, a case discussed by Cain .
While competitive forces may in some cases erode discriminatory wage differentials, the rate of decline of race and sex wage differentials has evidently not been rapid enough in the eyes of policymakers. U.S. anti-discrimination policy has been guided by two major vehicles: Title VII of the 1964 Civil Rights Act and the Office of Federal Contract Compliance Programs (OFCCP). (5) The Equal Employment Opportunity Commission (EEOC) was created to enforce Title VII and is empowered to aid plaintiffs or bring suit itself against employers. The Commission monitors the employment practices of federal contractors. Not only are contractors barred from discrimination against protected groups, these firms are also required to take "affirmative action" not to discriminate.
The purpose of this paper is twofold. First, it demonstrates that in a general equilibrium framework, the case of customer discrimination is theoretically distinct from employer or co-worker prejudice. In particular, while constant returns to scale and the existence of nondiscriminatory firms can eliminate the latter two forms of discrimination, customer discrimination may persist even under such conditions. The persistence of discriminatory wage differentials is shown to depend on consumer demand, technology, and the relative size of the minority labor force. (6)
Second, the general equilibrium framework is then used to evaluate the impact of government anti-discrimination policy where customers are the source of discrimination. It is shown that if the government requires equal pay and proportional representation for minorities, then the resulting allocation of resources will, under particular assumptions, be the same as if customers did not discriminate. I then contrast these theoretical results with the predicted general equilibrium impact of affirmative action when there is employer or co-worker prejudice. Among previous general equilibrium models of affirmative action, Welch  assumed no discrimination, while Freeman  did not specify the source of discrimination. It is shown here that the source of prejudice markedly affects one's conclusions about affirmative action. Further, even in some cases where segregation has eliminated discriminatory wage differentials, the introduction of affirmative action has allocative effects. Finally, evidence from studies evaluating the impact of affirmative action is used to distinguish among the various equilibrium models. …