A Statistical Framework and Optimal Policy for Affirmative Action in Higher Education Admissions

Abstract

In many countries, individuals are admitted to university based entirely on their scores on standardized tests or admission exams. Candidates from traditionally disadvantaged subgroups typically attain lower scores on these exams than the rest of the population and as a result are denied educational opportunities unless they are given preferential admissions based on some affirmative action scheme. We present a statistical characterization of the disadvantage suffered by candidates from underrepresented subgroups in the context of score-based admissions in higher education; we characterize disadvantage in terms of difference in the score distributions between subgroups. Based on this characterization, we propose an affirmative action scheme that combines the best features of preferential boosts and quotas. It ensures proportional representation of disadvantaged subgroups without having to legislate pre-fixed quotas. It also addresses multiple sources of disadvantage such as race, gender and poverty as well as the degree of each disadvantage. Furthermore, we show that our scheme is optimal in that no policy can do more to offset the disadvantage.