Dynamic Contracting with Moral Hazard Under Incomplete Information

I study a continuous time principal-agent model in which an unknown parameter and the agent's hidden effort affect the distribution of observable outcomes. The principal and the agent learn about the parameter by observing past outcomes. The agent's current effort has an implicit long-term effect through the belief dynamics and a deviation in effort creates a persistent disparity between the principal's and the agent's beliefs. This disparity affects the rate of learning as well as how the two evaluate the expected distribution of future outcomes which in turn affects their evaluation of future payoffs. Placing minimal restrictions on how effort and the parameter interact, I derive necessary and sufficient conditions for incentive compatible contracts. In addition to the agent's promised utility, the covariance between the on-path posterior beliefs and the agent's total payoff serves as a second state variable capturing the marginal long-run effects of effort.