Coalition-Stable Equilibria in Repeated Games

It is well-known that subgame-perfect Nash equilibrium does not eliminate incentives for joint-deviations or renegotiations. This paper presents a systematic framework for studying non-cooperative games with group incentives, and offers a notion of equilibrium that refines the Nash theory in a natural way and answers to most questions raised in the renegotiation-proof and coalition-proof literature. Intuitively, I require that an equilibrium should not prescribe in any subgame a course of action that some coalition of players would jointly wish to deviate, given the restriction that every deviation must itself be self-enforcing and hence invulnerable to further self-enforcing deviations. The main result of this paper is that much of the strategic complexity introduced by joint-deviations and renegotiations is redundant, and in infinitely-repeated games with discounting every equilibrium outcome can be supported by a stationary set of optimal penal codes as in Abreu (1988). In addition, I prove existence of equilibrium both in stage games and in repeated games, and provide an iterative procedure for computing the unique equilibrium-payoff set