A Robust Test of Prejudice for Discrimination Experiments

A large experimental literature is devoted to studying discrimination. An important question for policymakers and firms is what drives the discrimination uncovered by those experiments. However, motivations are hard to determine when decision makers pay selective attention to information because their learning is private. We overcome this challenge by deriving conditions on average outcomes that reveal decision makers are prejudiced no matter what they learn about individuals in each demographic group before making their decisions. This provides a test of prejudice that is general, simple, and robust and that can potentially be used to identify prejudice in a wide range of important settings, such as hiring, consumer lending, and housing access. We demonstrate our test of prejudice using two influential labor market experiments.