Incentives for Experimenting Agents

We examine a repeated interaction between an agent, who undertakes experiments, and a principal who provides the requisite funding for these experiments. The agent's actions are hidden, and the principal, who makes the offers, cannot commit to future actions. We identify the unique Markovian equilibrium (whose structure depends on the parameters) and characterize the set of all equilibrium payoffs, uncovering a collection of non-Markovian equilibria that can Pareto dominate and reverse the qualitative properties of the Markovian equilibrium. The prospect of lucrative continuation payoffs makes it more expensive for the principal to incentivize the agent, giving rise to a dynamic agency cost. As a result, constrained efficient equilibrium outcomes call for nonstationary outcomes that front-load the agent's effort and that either attenuate or terminate the relationship inefficiently early.