Monitoring Decreases Discrimination

There is broad but ambiguous evidence on ethnic discrimination. In particular, there is a disparity between lab and field. In order to assess a possible influence of monitoring (e.g. in an experimental setup) on discrimination we implement a recruitment setup on an online crowdsourcing marketplace with US residents. Our three treatments vary the context of a recruitment situation. Participants are either (1) simply asked to perform a pre-selection in a recruitment process, or (2) additionally made aware they are participating in an experiment, or (3) additionally made aware their decisions will be evaluated ex post. We use a causal forest to uncover heterogeneity in ethnic discrimination. On average, participants favor resumes with names typically associated with Black Americans by a small amount. A small minority of participants strongly favor resumes with names typically associated with White Americans in the control treatment, but this effect disappears if participants are told that they are participating in an experiment or that their decisions are being monitored. Our study thus demonstrates that socially acceptable discrimination pertains while undesirable discrimination is mitigated by monitoring. This extends to experimental setups, explaining part of the disparity between lab and field.