Affirmative Action in a Two Stage Statistical Discrimination Model

We investigate the effect of affirmative action policies in a statistical discrimination model with two stages. A group of workers face discrimination in both hiring and promotion decisions of the firm, and the regulator is free to intervene in both stages. We show that there are affirmative action policies which can eliminate negative employer beliefs by increasing educational investment in the discriminated group. Further, affirmative action imposes a disincentive effect on educational investment only for policies which require the firm to promote workers who are clearly unqualified. Our results show that while it is always beneficial to intervene in the hiring stage, the welfare effect of promotion stage intervention is ambiguous. We subsequently characterize the optimal affirmative action plan and show that the optimal policy may require intervention in both stages. Lastly, we consider the feasibility of introducing affirmative action policies, in terms of ensuring compliance by the firm. We show that arbitrary policies may indeed violate the participation constraint of the firm. Nevertheless, the optimal policy is always feasible, since it allows the firm positive profit.