Endogenous learning and the persistence of employer biases in the labor market

I present a statistical discrimination model of the labor market in which persistent negative employer biases about the productivity of a group of workers arise through hiring and learning about the group. Bayesian profit-maximizing employers endogenously develop biased beliefs based on their hiring experiences which lead to asymmetric learning about the group's productivity across employers. Optimal hiring follows a cutoff rule in posterior beliefs and market-clearing wages below which employers stop hiring from the group, preserving negative biases and leading to a negatively-skewed aggregate distribution of beliefs. Long-run discrimination in the form of a wage below the group's expected productivity can arise even with market competition, without productivity differentials across worker groups or prior employer biases, and regardless of worker signaling or investment decisions. The model generates predictions analogous to the Becker taste-based model, in a statistical framework with beliefs replacing preferences, rationalizing apparent prejudice as the result of "incorrect" statistical discrimination.