Some have criticized government antidiscrimination programs for causing efficiency losses both by devoting public resources to monitoring and enforcement and by distorting personnel decisions. This paper examines the efficiency consequences of such programs when discrimination is caused by a market failure and when regulators are imperfectly informed about individual firms' personnel policies.This paper outlines a simple model of “statistical discrimination” and shows that an equal opportunity rule, which constrains employers from offering different wage schedules to different groups, will improve the efficiency of this economy. Employers will attempt to evade such regulation, and the paper describes an evasion strategy in which employers search for worker “qualifications” that can act as proxies for race or sex. The paper discusses problems involved in monitoring and enforcing compliance, and considers two possible responses by the regulatory agency. The first is analogous to “disparate impact” judgments in which the type of information that employers can use in personnel decisions is regulated. The second is an affirmative action‐type policy that monitors the outcomes, rather than the process, of hiring and compensation decisions. Affirmative action policies have some advantages since they permit employers to use information more efficiently than do policies that attempt to regulate personnel decisions directly. However, affirmative action policies are likely to lead to charges of preferent