Optimal Contract to Reward Private Experimentation∗

This paper studies how a principal motivates an uninformed agent to learn about, and reveal, his quality through private experiments. The principal commits to a reward scheme and she aims to assign the rewards to correspond as closely as possible to the quality of the agent. To get a high reward, the agent experiments privately and discloses the results selectively. I show that the optimal reward scheme features an increasing step function: the early steps encourage a potential good type agent to continue experiments after early successes; the later steps are designed to deter a bad type agent from over-experimentation after a failure, and the scheme becomes flat when enough successes are reported. If the agent’s incentives to deviate from the intended path of experimentation are weak, a one-step function is optimal: the agent receives a bonus if he reports enough successes; otherwise, he only gets a non-negative compensation. I characterise the conditions where the principal achieves the same efficiency level relative to a public information environment.