Markov equilibria in discounted stochastic games (Q1290840)

This paper first adress the question of the existence of Markov perfect equilibrium in stochastic games which have a state space, which is a complete separable metric space, and action spaces, which are compact metric spaces. In this case if the transition probability is weak-continuous in yesterday's actions there exists a Markov perfect equilibrium strategy for finite horizon games. For infinite horizon games the assumption is strengthened to the requirement that the transition probability is norm continuous in yesterday's actions. The transition probabilities are assumed to be absolutely continuous with respect to some measure \(v_t\) every period. Secondly it is shown that if the stochastic game is stationary (\(v_t= v\) for all \(t\)), and the measure \(v\) is nonatomic, then there exists a semi-Markov equilibrium strategy (i.e. a behavior strategy that depends only on the state and actions of the preceding period and the current period).