“…In general, the analysis of stochastic games is different from that of repeated games, because the action today influences the distribution of the future states, which in turn influences the stage‐game payoffs in the future. To avoid this complication, various papers in the literature (e.g., Dutta (), Fudenberg and Yamamoto (), Hörner et al ()) consider a model that satisfies the payoff invariance condition, in the sense that when players are patient, the feasible and individually rational payoff set is invariant to the initial state. In such a model, even if someone deviates today and influences the distribution of the state tomorrow, it does not change the feasible payoff set in the continuation game from tomorrow, so continuation payoff can be chosen in a flexible way, just as in the standard repeated game.…”