Evaluating Skilled Experts: Optimal Scoring Rules for Surgeons
We consider settings in which skilled experts have private, heterogeneous types. Contracts that evaluate experts based on outcomes are used to differentiate between types. However, experts can take unobservable actions to manipulate their outcomes, which may harm consumers. For example, surgeons may privately engage in harmful selection behavior to avoid risky patients and hence improve observed performance. In this paper we solve for optimal evaluation contracts that maximize consumer welfare. We find that an optimal contract takes the form of a scoring rule, typically characterized by four regions: (1) high score sensitivity to outcomes, (2) low score sensitivity to outcomes, (3) tenure, and (4) firing or license revocation. When improvement is possible, an optimal contract for the low quality expert is a fixed-length mentorship program. In terms of methods, we draw upon continuous-time techniques, as introduced in Sannikov (2007b). Since our problem involves both adverse selection and moral hazard, this paper features novel applications of continuous-time methods in contract design.