A Model of Racial Profiling in Higher Education

Competitive public and private institutions of higher education in the U.S. take race into consideration in admissions and decisions about financial aid when able to do so. In public universities in states that have proscribed use of race, substitute policies, intended to promote minority attendance, have been enacted. Universities seek out the highest achievers among those from under-represented races. This paper develops a theoretical model, with computational counter-part, that explains university admissions policies. In the model, race provides a low-cost signal to admissions officers about likely hardships faced by potential matriculants. In a recent case challenging the University of Michigan's affirmative-action practices, the Supreme Court circumscribed use of race by mandating a holistic approach to admissions. Implications of this decision are analyzed. In the model, a university maximizes an academic quality index that increases with the predicted academic potential of the student body and with the educational inputs the university provides. Student potential is imperfectly observed. A set of readily observable student measures, such as SAT score and high-school GPA, provide information about potential. However, better prediction requires observation of "hardship" experienced by applicants while growing up. Hardship lowers performance in grade- and high-school and masks simple measures of academic potential. Hardship will tend to increase with poverty and membership in minority populations, but can only be accurately assessed at considerable cost. An unregulated university selects, based on relatively low-cost observables, the set of applicants, if any, for whom hardship is accurately assessed. For the latter students, admission and financial aid is directly linked to the readily observed variables that predict potential (e.g., SAT score) and to the hardship measure. For other students, the university employs racial profiling, with admission and financial aid directly linked to race. The model predicts colleges' equilibrium tuition and financial aid policies, expenditures on educational inputs, and the allocation of students of varying characteristics across educational institutions. The recent Supreme Court decision is interpreted as requiring assessment of hardship for any admitted students, thus as disallowing racial profiling. Because performance of assessments is costly, the set of potential student types that are considered is affected. The effects on the characteristics of the student body, tuition and financial aid policies, and provision of inputs are examined. Specific predictions about these effects are made by employing the counter-part computational model that is calibrated to U.S. data