Student effort, race gaps, and affirmative action in college admissions: theory and empirics

In this dissertation, I develop a framework to investigate the implications of Affirmative Action in college admissions on both study effort choice and college placement outcomes for high school students. I model the college admissions process as a Bayesian game where heterogeneous students compete for seats at colleges and universities of varying prestige. There is an allocation mechanism which maps each student's achieved test score into a seat at some college. A colorblind mechanism ignores race, while Affirmative Action mechanisms may give preferential treatment to minorities in a variety of ways. The particular form of the mechanism determines how students' study effort is linked with their payoff, playing a key roll in shaping behavior.

I use the model to evaluate the ability of a given college admission policy to promote academic achievement and to minimize racial academic gaps--namely, the achievement gap and the college enrollment gap. On the basis of these criteria, I derive a qualitative comparison of three canonical classes of college admissions policies: color-blind admissions, quotas, and admission preferences.

I also perform an empirical policy analysis of Affirmative Action (AA) in US college admissions, using data from 1996 on American colleges, freshman admissions, and entrance test scores to measure actual AA practices in the American college market. Minority college applicants in the United States effectively benefit from a 9% inflation of their SAT scores, as well as a small fixed bonus of approximately 34 SAT points. I also estimate distributions over student heterogeneity and perform a series of counterfactual policy experiments.

This procedure shows that AA practices in the US significantly improve college placement outcomes for minorities, at the cost of discouraging achievement by the most and least talented students. The analysis also indicates ways in which AA could be re-designed in order to better achieve its objectives. As it turns out, a quota system produces a substantial improvement relative to either the current system or a color-blind system. However, quotas are illegal in the US and cannot be implemented as such. Nevertheless, I propose a variation on the AA policy already in place that is outcome-equivalent to a quota.