The Ratchet E¤ect: A Learning Perspective

We examine the ratchet e¤ect arising from moral hazard and symmetric learning about a new technology. A principal and agent are symmetrically uncertain about the di¢ culty of a new job. If the agent shirks, this increases his future payo¤s since the principal will, on average, believe that the job is harder than it actually is. So shirking must be deterred by high-powered incentives. These induce the agent to over-work, since he can quit if the principal believes that the job is easy. With continuous e¤ort choices, no interior e¤ort is implementable. We show that this problem is endemic in learning models where the agent has discrete as well as continuous choices. Rent-sharing, where the principal pays more than the outside option, can restore implementability, since the agent loses rents if he over-works. The ratchet e¤ect is strongest when the marginal productivity of e¤ort is relatively independent of the type of the job, and is mitigated either if job type and e¤ort are complements, or if they are substitutes.