Overview

We study a dynamic principal–agent relationship in which an agent must exert costly effort to learn a privately observed binary state before taking an action. The principal wants to match the action with the state, while the agent is biased toward one action, generating both a moral hazard (effort choice) and an adverse selection (action choice) problem. The principal disciplines the agent through verification (at a cost), reduced workload and termination. We show reduced workload is always a valuable instrument, even when the cost of verification is small and the loss from shirking is large. By promising a reduced workload in the future, the principal can lower verification costs across multiple periods. For high biases, verification and reduced workload are insufficient instruments, and the principal must rely on firing along the equilibrium path. The threat of future firing complements verification and saves verification costs over time.