In November, 2008, Colorado and Nebraska voted on amendments that sought to end race-based affirmative action at public universities in those states. In anticipation of the vote, the University of Colorado at Boulder (CU) explored statistical approaches to support class-based (i.e., socioeconomic) affirmative action. This dissertation introduces CU's method of identifying socioeconomically disadvantaged and overachieving applicants in undergraduate admissions. In addition, sensitivity analyses were conducted to gauge the impact of technical decisions that were made when these measures were devised. Two experiments were carried out to determine whether or not implementing this approach would change the racial and socioeconomic diversity of accepted classes. Finally, historical student records were examined to explore the likelihood of college success for the beneficiaries of CU's class-based approach.

The sensitivity analyses identify particularly consequential issues that architects of class-based systems may face, including modeling application to college, defining target populations, and addressing missing data. The experiments suggest class-based affirmative action can potentially increase acceptance rates for low-SES and minority applicants, particularly if it is used alongside race-conscious admissions. Analyses of historical data do not rule out the possibility of college success for the beneficiaries of class-conscious admissions, but they do argue for the provision of robust academic support to marginally qualified, low-SES students when they matriculate. This dissertation is intended to serve as a resource for postsecondary institutions considering class-based admissions policies. If race-based approaches are overturned, universities like CU could struggle to develop race-blind metrics to identify applicants who have faced adversity. This research examines one method of quantifying the barriers these students encounter.