Evidence of discrimination

This paper develops a decision-theoretic model of evidence production in the context of a discrimination trial. Producing evidence is assumed to be costly, and the cost can vary depending on what type of defendant behavior (and plaintiff characteristics) the evidence bears upon. The goal of the trial is to uncover a possible behavioral bias in the defendant (intent to discriminate). I then ask how a social planner would structure the production of evidence in a trial in order to best achieve this objective, taking into account the cost of evidence production. I show that it is sometimes efficient to sequence the production of different kinds of evidence (burden shifting) or even to allow a decision based on limited evidence (for example, disparate impact alone, as a proxy for intent to discriminate). A key variable is the availability of evidence concerning the productivity of the plaintiff.