Minority Penalty in Holistic College Admissions

Abstract

We test for minority bias in holistic assessments in Danish college admissions exploiting a novel data source on the universe of rankings of applicants by programs. We find that programs consistently rank minorities lower than majority applicants with similar high school GPA. To investigate whether this is evidence of bias we perform outcome-based tests at the margin of program admission and find strong evidence of preferential treatment of minorities. The marginal minority admission underperforms the marginal majority admission by about 2 percentage points, suggesting either a taste-based preference for minority applicants or systematic overestimation of the likelihood of completion for minority applicants. While on average the bias is small and negative, it varies substantially across programs. Our data on the full rankings of applicants based on holistic assessments also allow us to estimate infra-marginal bias. Conditional on program assessment, minority applicants underperform majority applicants within a large majority of programs and by about 4 percentage points on average. The infra-marginal bias varies considerably less across programs. While marginal and infra-marginal bias are on average similar and strongly correlated, we find strong attenuation of large marginal biases away from the margin. Joint paper with Edwin Leuven.