Learning in Relational Contracts

We study relational contracts in a firm-worker relationship with mutual uncertainty about match quality. The worker's actions are publicly observed and generate output, and hence information about the match quality. Despite the absence of private information, uncertainty about match quality creates a holdup problem causing inefficiencies in relational contracts even in the frequent action limit. We characterize the optimal relational contracts and show that they involve actions that yield both a lower payoff and less information than another action. Yet, such actions may be a modest way to learn about the match quality when more efficient actions are not credible