Incentives for experimenting agents

type="main"> We examine a repeated interaction between an agent who undertakes experiments and a principal who provides the requisite funding. A dynamic agency cost arises—the more lucrative the agent's stream of rents following a failure, the more costly are current incentives, giving the principal a motivation to reduce the project's continuation value. We characterize the set of recursive Markov equilibria. Efficient equilibria front-load the agent's effort, inducing maximum experimentation over an initial period, until switching to the worst possible continuation equilibrium. The initial phase concentrates effort near the beginning, when most valuable, whereas the switch attenuates the dynamic agency cost.