Optimal Contracts for Experimentation

This paper studies long-term contracts for experimentation in a principal-agent setting with adverse selection about the agent’s ability (pre-contractual hidden information), dynamic moral hazard, and private learning about project quality. We show that profit maximization by the principal generally leads to under-experimentation by an agent of low ability, even though there would be no distortion in the absence of either adverse selection or moral hazard. The structure of optimal contracts is shaped by a variety of considerations including dynamic agency costs and the possibility of post-contractual hidden information about project quality. We derive two explicit menus of contracts that can be used to implement the second best solution: “bonus contracts” and “clawback contracts”. Both feature history-contingent dynamic streams of transfers.