| Repeated games with private information and communication can be analyzed by integrating mechanism design techniques with dynamic programming techniques. However, this integration poses some novel problems and issues. This dissertation begins by showing that, under standard assumptions, the folk theorem can be extended to many infinitely repeated games with statistically interdependent private information and interdependent utility functions. But these standard assumptions---borrowed from static mechanism design---imply a strong institutional environment underlying the game. The second chapter explores optimal equilibria in environments lacking institutional support, and shows that optimal equilibria are generally inefficient. Techniques for calculating and characterizing optimal equilibria are provided. One application of the results is to provide a new explanation for why cartels may engage in price wars. The third chapter focuses on a repeated trade game, under a range of institutional settings. In the absence of institutions, repeated game equilibria can be no more efficient than the optimal one-shot equilibrium. But there are several types of weak institutions that can support efficiency or near efficiency in the repeated game. For instance, when the traders have access to a joint, bounded savings account, they can achieve approximate efficiency even in the absence of other institutions. |
