Adaptive Correspondence Experiments

Correspondence experiments probe for discrimination by manipulating employer perceptions of applicant characteristics. We consider the gains from dynamically adapting number and quality fictitious applications each receives to their prior callback decisions. Calibrating behavior experimental data Nunley et al. (2015), we find that it is possible cut required detect a fixed discriminators roughly in half relative benchmark design with per job. These are achieved primarily abandoning jobs very low probabilities those call back Black applicants.