Affirmative Action in a Labor-Market Signaling Game
Larry T. McRae, Appalachian State University
Objective: This paper applies theoretical reasoning to the impacts of affirmative action policies in academic admissions and subsequently in firms’ hiring decisions. Its major purpose is to explore the implications of such policies for employer expectations and behavior.
Background: Although affirmative action policies, of one sort or another, have been in place for more than thirty years, there remain confusion about what such policies entail and controversy about what they accomplish. The numerous empirical studies of job-market discrimination are often cited in support of continuing affirmative action, but these studies show a puzzling lack of effect from existing affirmative action policies, with the narrowing of male/female and black/white wage gaps attributable mostly to changes in the characteristics of the female and minority labor forces. Theoretical studies of affirmative action are relatively rare and generally focused on efficiency arguments. This paper seeks to explore the information content of affirmative action policies, in particular to ask how such policies affect the signal sent by educational attainments to employers. The theory draws on the literature begun by A. M. Spence and advanced by Andrew Weiss and others.
Methods: Firms and prospective employees may be thought of as playing a sequential game; the game is initiated by the individual sending a signal through his level of education. In response to the signal, the firm makes a wage offer. In a signaling context, the positive effect of affirmative action may be conceived as lowering the cost to a member of a minority for sending an education signal. But such policies may also change the probabilities of different levels of productivity which a firm assigns to education. In the game context, we explore how this effect may change a firm’s willingness to hire minority workers and, incorporating Bayesian reasoning, its willingness to promote if forced to hire.
Results: The value of education lies at least partly in its value as a signal of productivity. If affirmative action causes the value of the signal to be degraded, the value of education is reduced to the individual member of a minority group, whether or not his acquisition of the signal is primarily due to affirmative action. Moreover, in those jobs in which signaling is important, direct measures of productivity are difficult. If affirmative action in education degrades the signal while affirmative action policies force firms to hire minority workers, the Bayesian priors established by affirmative action will affect the firm’s perception of the employees’ actual performance, so that subsequent promotion is less likely. We find that while affirmative action may benefit low-productivity minority workers, the spillover effects of changed subjective probabilities may hamper the success of high-productivity minority workers. Majority workers may actually benefit from affirmative action: while the risk of displacement at initial hiring is increased, they may gain relative preference in promotion.
Discussion: Examining the signaling aspect of affirmative action leads to the conclusion that affirmative action in education may reduce the value of education as a signal to all members of the groups favored by affirmative action. This may provide a partial explanation of the apparently greater advance of women than of blacks in recent decades, since white females at least are not perceived as benefiting to the same extent as black males from affirmative action in college admissions. The results also allow examination of the claim made by conservatives that affirmative action actually harms the economic prospects of minority groups. In the context of a signaling game, we can examine this claim and the assumptions necessary for it to be true. Although this paper advances no new statistical results, its theoretical results may help to explain some existing empirical results and policy claims.