Incentivizing Exploration and Stopping
Description: We study a continuous-time principal-agent model where an agent controls whether and when to take on a risky project, and a principal would like to acquire more information about the project's value. The principal commits to dynamically designing experiments subject to the agent's limited commitment and flexibility to gather information. Our main result states that there is an optimal solution where the principal acquires information via a Poisson process, and the stopping rule is to halt after the first confirmatory signal. While the principal wishes to balance the experiment's frequency and precision of breakthroughs, the agent's lack of commitment may distort this trade-off.
More Dependence and Correlation
With Hector Chade
Description: We study a principal multi-agent model with moral hazard to investigate a suitable notion of dependence over agents' stochastic outputs. We use this notion to analyze the following comparative statics question: how does an increase in dependence affect the principal's expected profit? We provide two broad cases where we show that the principal prefers more dependent outputs. These questions are theoretically challenging because more dependent outputs need not imply more informative ones. Our results also shed light on how organizations can optimally sort agents in teams through the dependence structure.
How Clinicians' Decisions Affect Recipient Life-Years from Transplantation (LYFT)
With Tomas Larroucau, Ellen Green, E Glenn Dutcher, Jesse D Schold, and Darren Stewart
Description: Within the allocation of deceased donor kidneys, clinicians play a key role in making acceptance decisions on behalf of recipients in the waitlist. However, as the literature documents, it is unclear why there is substantial clinician-level variation in decisions, what channels drive it, and how it impacts recipients' survival outcomes. Using administrative data, this paper studies these questions to evaluate the effect of clinicians' decisions on recipient life-years from transplantation (LYFT). We exploit the exogenous variation of on-call data to identify clinician-specific unobservables which induce selection on both their acceptance decisions and each recipient's survival outcomes. We build a structural model of clinician's acceptance decisions which reflect arbitrary correlation structure between recipient, donor, clinician, and match-specific unobservables. Our goal is to estimate the model and conduct counterfactual exercises to assess how several types of clinicians affect assignment outcomes.