Information-based discrimination

Abstract

The distribution of a single college’s incoming freshman cohort by high school of origin differs from the population distribution of high school seniors (Bernhard, 2013), a fact typically attributed to taste-based or statistical discrimination. This paper models a third explanation which I call information-based discrimination: information asymmetries between a college and high schools incentivize the formation of a relationship between the college and one ex ante randomly selected high school, resulting in de facto discriminatory admissions even in the absence of taste-based and statistical discrimination. I construct an admissions game between one college and $N$ identical in expectation high schools. The game features a perfect Bayesian equilibrium in which the ex ante randomly chosen high school provides truthful information to the college in exchange for future favorable admission treatment. When agents are sufficiently patient, this relationship between the college and the initially randomly chosen high school is sustained in equilibrium. This equilibrium increases overall admissions efficiency relative to but does not Pareto improve on a dominant strategy implementable alternative: the welfare gains accrue entirely to the college and high school in the relationship while harming students from all other high schools.