Affirmative Action with Multidimensional Identities

Abstract

Affirmative action policies are widely employed in college admissions, hiring, and other decisions to reduce underrepresentation of disadvantaged groups. When identities are multidimensional, the basic unit of analysis is the intersectional group (i.e., identity vector). Affirmative action policies are, however, predominantly nonintersectional, being based on the identity dimensions, not the intersectional identities. We demonstrate that any nonintersectional policy can almost never achieve a representative outcome. In fact, nonintersectional policies can increase the underrepresentation of underrepresented groups in a manner undetected by standard (nonintersectional) measures. Examples based on race and gender in the United States reveal significant (hidden) losses from nonintersectional policies. Accounting for interactions between identity dimensions, we show how to construct intersectional policies that achieve proportional representation.

Keywords: Affirmative action, education, inequality, underrepresentation, identity, intersectionality

JEL Classification: J7, I24, D02