ABSTRACT
In a research program that began in 2014, my colleagues and I studied a method of achieving diversity goals in admissions using an operations research technique called constrained optimization (CO). Here, I provide details about an analysis conducted using applicant data from a graduate program in the physical sciences. The CO approach was used to select a hypothetical class from the applicant data, and the resulting class was compared to the class that had actually been admitted and to a hypothetical class admitted solely on the basis of GRE scores. The CO class, which contained the same number of students as the actual admitted class, had a larger percentage of women (40.6% versus 30.8%), a larger percentage of students from under-represented ethnic groups (15% versus 7.5%), and slightly better average GRE scores. The CO class was much more diverse than the class selected solely on the basis of GRE scores and had average GRE scores that were almost as good. The analyses serve as a proof of concept—an illustration that mathematical techniques can help in the selection of diverse and high-performing classes.
