Game-Playing Agents: Unobservable Contracts as Precommitments

The players in most economically important games are agents, not principals. This raises the possibility of the principal's setting a strategic compensation scheme. The central question addressed here is whether unobservable agency contracts can serve as precommitments. I argue that, in terms of Nash equilibrium outcomes, the answer is no when it is common knowledge that there exists a contract that "solves" the standard agency problems and that the principal and agent have the same preferences over income and effort. However, I also show that when these conditions are not satisfied (as they typically will not be), provisions of the agency contract enacted solely to deal with incentive and risk sharing problems of the agency relationship may have the secondary effect of credibly precommitting the agent in the game he plays with other agents. I also briefly consider the effects of contract renegotiation.