Meritocracy in the Face of Group Inequality *

Abstract

Many institutions use identifiable markers such as race or sex to set standards for candidate selection. The desirability of these systems is often questioned. This paper attempts to inform these debates using a model in which groups have unequal access to resources that affect their performance and performance has to be inferred prior to selection. We show that performance-maximizing allocations are not necessarily group-blind or monotonic in observable traits. This is true even when individuals can underreport scores or underinvest in training, and when commitment to selection policies is possible. By characterizing the complex nature of efficient allocations, our results provide critical benchmarks for broader debates on social equality.

Keywords: merit, affirmative action, group inequality