Labor Market Discrimination: A Statistical Approach

Abstract

Gary Becker's classic treatment of discrimina tion was based upon the notion of racial taste or preference (3). Recent works by Kenneth Arrow, however, have attempted to interpret employer discrimination in an alternative framework as a rational reaction to uncertainty in labor markets (1, 2). Employing a somewhat restrictive set of assumptions, Arrow demonstrates that in a world of imperfect information wage differentials may exist between blacks and whites even in the absence of racial taste on the part of employers. Using the same set of assumptions, Arrow then goes on to consider certain general equilibrium aspects of his model. The purpose of the present paper is two fold: first, to show that Arrow's major partial equilibrium results may be derived in a much more general setting; second, to point out the important fact that each dollar spent by the government on lowering the cost of information to employers would serve to increase equilibrium wage offers by some amount exceeding one dollar, thus yield ing a net social gain. Consider an employer faced with several job applicants, some of whom are white, the others black. If the true productivity of each worker could be determined without cost and if the employer had no racial preference, then he would simply hire all those applicants judged to be suf ficiently qualified to perform on the job, without regard to race. Suppose, however, that the employer possesses no knowledge about the various appli cants other than their racial identities. In order to determine the true productivity of each applicant the employer must undertake a costly processing procedure, which might involve testing, inter viewing, or perhaps even a trial period of employ ment. Let us suppose further that the employer can withhold the results of this processing procedure from his competitors, so that the fact that a par ticular worker has been deemed qualified by one employer does not lower the cost to any other employer of making a similar determination. Following Arrow, assume that the employer assigns a probability pw (pb) to the event that a randomly chosen white (black) worker will be qualified to perform on the job. Consider the consequences of making Arrow's restrictive as sumption that all workers who are above a mini mum acceptable level of qualification are in fact equally qualified, having a marginal product of M P. Now if a white worker is processed and turns out to be qualified, the gain to the firm from hiring him is equal to MP ? Ww, where Ww is the wage commanded by white workers. But he will only be qualified, hence only hired, with a probability of pw. The expected gain to the firm thus equals (MP ? Ww)pw. In a long-run equilibrium the cost of the investment will just equal the expected gain :