We analyze equilibrium outcomes under costly screening of candidates from distinct categories. Candidates choose how much effort to invest in skill acquisition, and the screener decides how carefully to screen each candidate. Typically, there are multiple equilibria, and ex ante identical individuals in different categories can receive different equilibrium treatment. The imposition of a quota for an initially disadvantaged group can result in increased screening intensity and skill investment by destabilizing the least active equilibrium. The endogeneity of screening enables powerful comparative statics, which we use to unify different strands in the literature on discrimination.