Designing Affirmative Action Policies under Uncertainty

We study university admissions under a centralized system that uses grades and standardized test scores to match applicants programs. In the context of this system, we explore affirmative action policies seek narrow gap between admission rates different socio-demographic groups while still accepting students with high scores. Since there is uncertainty about score distribution who will apply each program, it unclear what policy would have desired effect on groups. address challenge by using predictive model trained historical data help optimize parameters such policies. find learned does significantly better than relying ideal for last year. At same time, also large pool yields similar results as our approach data. Due more complex nature approach, conclude simpler should be preferred if enough available (e.g., long-standing, traditional programs), but not newer programs other cases in which strategy can prove helpful.