Economic Theory
November 2019
I explore how a principal dynamically chooses among multiple agents to utilize for production without full commitment. The principal chooses at most one agent to utilize in every period affecting the states of the agents. A utilized agent changes its state because it is utilized, but the nonutilized agents do not remain at rest: they also change their state. The analysis requires a novel methodological approach: the agency problem that the principal faces with each agent is shown to be an appropriately designed restless bandit, thus creating a multiarmed restless bandit. The principal’s optimal contract is characterized by an index rule for the restless bandit problem.