Theoretical Economics 17 (2022), 1403–1450

We consider a moral hazard problem in which a principal provides incentives to a team of agents to work on a risky project. The project consists of two milestones of unknown feasibility. While working unsuccessfully, the agents’ private beliefs regarding the feasibility of the project decline. This learning requires the principal to provide rents to prevent the agents from procrastinating and free-riding on others’ discoveries. To reduce these rents the principal stops the project inefficiently early and gives identical agents asymmetric experimentation assignments. The principal prefers to reward agents with better future contract terms or task assignments rather than monetary bonuses.

Keywords: Principal-agent, moral hazard, experimentation, exponential bandit, contests

JEL classification: D82, D83, D86