Stationary almost Markov perfect equilibria in discounted stochastic games (Q2806812)

Recent counterexamples by \textit{Y. Levy} [Econometrica 81, No. 5, 1973--2007 (2013; Zbl 1354.91016)] show that a nonzero discounted Markov game may not have a stationary Markov perfect equilibrium.NEWLINENEWLINEIn the paper under review, the authors consider the lightly weaker concept of stationary almost Markov perfect equilibrium (SAMPE): i.e. a Nash equilibrium that depends only on the current and the previous states of the game. They prove the existence of such SAMPE for discounted stochastic games having NEWLINE{\parindent5mmNEWLINE\begin{itemize} \item- a Borel state space \(S\), \item - for each player \(i\), a family of time dependent action sets \(A_i(s), s\in S\) subsets of some compact metric space, \item - reward functions, which are Carathéodory functions, \item - a transition probability \(g\), convex combination of a a finite number of state dependent probabilities, which are all dominated by the same probability measure. NEWLINENEWLINE\end{itemize}}NEWLINENEWLINEAfter proving this result, a large section of the paper is devoted to examples of dynamic Cournot games which admit SAMPE.