A folk theorem for stochastic games with private almost-perfect monitoring (Q2016219)

A stochastic game on infinite horizon is considered where each player gets a noisy information about the others' moves. Under suitable technical hypotheses, which include in particular the conditions that (1) the initial condition will be `forgotten' in a precise sense, (2) the information about the others' moves is sufficiently informative, and (3) the discount factor is sufficiently close to 1, it is shown that every individually rational feasible discounted payoff is attainable, thereby establishing a version of the `folk theorem'.