Rewarding success and failure: Moral hazard and adverse selection in strategic experimentation

A principal hires an agent to learn about the cost of a project (experimentation) and then to execute it (production). The agent is privately informed about the probability that the cost is low, with the high-type agent more optimistic than the low type. Learning is modeled as a series of experiments, where the agent privately chooses his effort in each experiment. Due to the joint presence of adverse selection and moral hazard, even the low type can misrepresent his type, shirk during experimentation and earn a rent. In the optimal contract, the principal uses the length and outcome of experimentation as well as the timing of payments to induce effort and screen the agent. The principal may find it optimal to reward the low type after failure in experimentation. We provide sufficient conditions for both success and failure to be rewarded. We show that over-experimentation can be optimal as it reduces the asymmetric information after a series of failed experiments. We also consider whether it may be optimal to separate experimentation and production between two different agents. When the same agent works on both tasks, the principal can leverage the adverse selection rent to mitigate moral hazard. Therefore, integrating experimentation and production is optimal when adverse selection is severe.