Can Universities Maintain Diversity without Directly Considering Race in Admissions?
In 2015, the Supreme Court decided to rehear the Fisher challenge. At the heart of the case now under consideration are the vaguely defined terms “workable” and “race-neutral.” Despite a lot of ambiguity about what the Court thinks those terms entail, universities have been tasked with the challenge of establishing alternative admissions methods that are both race-neutral in the eyes of the courts and workable to ensure a racially diverse student body. Is that possible? I conclude that currently fashionable alternatives to affirmative action – such as the use of non-racial characteristics of applicants as predictive proxies for race – have substantial limitations. Often, the alternatives are unworkable or raise important new legal or social concerns.
A Close Look at an Alternative
One frequently proposed alternative to affirmative action is to predict a student’s minority status using other known characteristics. I call this method “proxy-based affirmative action.” To date, there has been insufficient empirical evidence about the capacity of such a system to achieve levels of diversity similar to those achieved by traditional affirmative action approaches that take applicants’ race explicitly into consideration. To learn more about the workability of the proxy-based approach, I did simulation models of admissions decisions at the University of Texas, using data of all kind on applicants to the university.
My models looked at what the admitted class looks like when traditional affirmative action criteria are used in admissions, versus what it looks like if indirect proxy measures of student characteristics are used instead of race itself. I address such questions as:
- Could a student’s race be accurately predicted by various possible proxies?
- Could the same level of diversity be achieved for the admitted class if various proxies were used instead of race?
- What effect would such a proxy based admissions system have on the academic qualifications of admitted students?
- Would this selection method be “workable” in a practical sense, politically sustainable, and “race-neutral” from a legal perspective (given uncertainty about the definitions of these terms)?
Limitations in Practice
My findings show that it is possible to achieve desired levels of racial diversity using a proxy-based system instead of traditional affirmative action. But there are significant limitations, ranging from the administrative cost of collecting the necessary information to the reduced academic performance of the student body admitted under a proxy system.
- Using 195 characteristics from the Educational Longitudinal Study, I was able to correctly predict the race and ethnicity of student applicants 82% of the time. However, many of the key characteristics are not collected on admissions applications and would be difficult to collect without infringing on the applicants’ privacy or inviting fraudulent answers. For example, several of the most predictive non-racial characteristics referred to the race of applicants’ best friends. Obviously, such questions are awkward to ask, and it would be easy for applicants to manipulate their answers if they knew they would affect admissions decisions. With the application data the University of Texas already collects, I was only able to correctly predict applicants’ underrepresented minority status 54% of the time.
- Diverse incoming classes can only be admitted using proxies for race, but only inefficiently. To admit the same proportion of minority students, the university must place 3.5 times as much weight on applicants’ indirectly predicted minority status as it would have placed on actual minority status! This strategy is not cost-free, because it means placing less weight on other student characteristics – including grade point averages and academic test scores.
- My simulations show that the student body’s overall academic performance would be lower under proxy-based affirmative action. Because the proxy system must place great weight on indirect correlates of race, academic performance ends up mattering less for admissions. As a result, the student body’s cumulative grade point average and likelihood of graduating would decline. For every 10,000 enrollees, I found that University of Texas could expect to graduate 75 fewer students if it employed the new proxy system.
- Another feasibility issue is that certain indirect predictors of applicants’ race are unlikely to be legally “race-neutral” or socially feasible for a state university. After all, the best proxy – the race of best friends – is itself not especially “race neutral.” In addition, university admissions officials would almost certainly face wide criticism from applicants and their families if they asked sensitive proxy questions or deemphasized academic qualifications.
In short, my research highlights that using indirect proxies for race – instead of just weighing race as one factor in college admissions – could very well undercut academic performance and student privacy. Proxy alternatives to affirmative action may exist, but these methods should be empirically explored before they are imposed, to make sure they can sustain student body diversity without side effects much more harmful than problems the new approaches are meant to correct.