A Remark on Color-blind Affirmative Action

Faced with legal challenges to explicitly race-contingent admissions policies, elite educational institutions have turned to criteria that meet diversity goals without being formally contingent on applicant identity. We establish that under weak conditions that apply generically, such color-blind affirmative action policies must be nonmonotone, in the sense that within each social group, some students with lower scores are admitted while others with higher scores are denied. In addition, we argue that blind rules can generate greater disparities in mean scores across groups conditional on acceptance than would arise if explicitly race-contingent policies were permitted. Elite colleges and universities in the United States have recently faced a number of legal challenges that restrict their use of explicitly race-contingent admissions policies.1 Because these institutions continue to seek broad representation from different social groups (and to view campus diversity as an essential ingredient in the provision of a first-rate education), they face strong incentives to adjust their admissions criteria in order to attain diversity goals through less direct means. There is considerable evidence that this Debraj Ray, Department of Economics, New York University ([email protected]) and Rajiv Sethi, Department of Economics, Barnard College, Columbia University ([email protected]). We thank (without implicating) Dennis Epple, Erik Eyster, Glenn Loury, and Tim Van Zandt for comments on an earlier version. Received October 21, 2009; Accepted November 23, 2009. Among the most visible of these are the 2003 Supreme Court ruling in Gratz v. Bollinger , which struck down the undergraduate admissions policy at the University of Michigan, and California’s Proposition 209, which in 1996 prohibited the use of race, sex, or ethinicity as an admissions criterion in public education; see Chan and Eyster (2009) for other examples. C © 2010 Wiley Periodicals, Inc. Journal of Public Economic Theory, 12 (3), 2010, pp. 399–406.