Policy Commitments in Relational Contracts

How should an organization choose policies to strengthen its relationships with employees and partners? We explore how biased policies arise in relational contracts using a flexible dynamic game between a principal and several agents with unrestricted vertical transfers and symmetric information. If relationships are publicly observed, then optimal policies are never biased - they are always chosen to maximize total continuation surplus. In contrast, if relationships are bilateral - each agent observes only his own output and pay - then the principal may systematically choose backward-looking, history-dependent policies to credibly reward an agent who performed well in the past. We first show that biased policies are prevalent in a broad class of settings. Then we argue that biases can manifest in interesting ways in a variety of simple examples. For instance, hiring may lag demand following a recession, investments may be delayed and awarded inefficiently, and allocation decisions may "stick with" an inefficient worker.