The past two decades have seen the development of a large and vital literature on repeated games with discounting. Research in this area has been stimulated both by a succession of theoretical breakthroughs and by a rich variety of economic applications that have emerged, notably in industrial organization, macroeconomics and international trade. Until recently, the models under study shared the feature of "perfect monitoring": in each period, a player in the repeated game knows precisely what actions were taken by other players in all previous periods. Researchers investigating applied problems in different fields became increasingly dissatisfied with the realism of the perfect monitoring assumption and developed models with special kinds of imperfect monitoring. The contribution of this project consists of developing more general and powerful models of imperfect monitoring and confronting the methodological problems that occur in these more realistic models. Traditional agency theory assumes that the principal has no more information about the agent's actions than the enforcement authorities have. This project explores various agency supergames that modify this standard informational structure. In one such model the principal, but not the enforcement agency, can observe the agent's actions. Results obtained by assuming that the principal can only write one-period contracts are extended to allow the principal at any time to freely write long-term contracts, which offer the agent a salary in every future period depending on the publicly observed outcomes. The contracts can be renegotiated at the end of every period. This work provides the basis for extending agency theory to a dynamic setting. Such a theory requires the analyst to solve problems of loss of recursive structure and the negotiation of disequilibrium agreements.
