Equilibrium analysis of dynamic models of imperfect competition

Motivated by recent developments in applied dynamic analysis, this paper presents new sufficient conditions for the existence of a Markov perfect equilibrium in dynamic stochastic games. The main results imply the existence of a Markov perfect equilibrium provided the sets of actions are compact, the set of states is countable, the period payoff functions are upper semi-continuous in action profiles and lower semi-continuous in actions taken by rival firms, and the transition function depends continuously on actions. Moreover, if for each firm a static best-reply set is convex, the equilibrium can be taken in pure strategies. We present and discuss sufficient conditions for the convexity of the best replies. In particular, we introduce new sufficient conditions that ensure the dynamic programming problem each firm faces has a convex solution set, and deduce the existence of a Markov perfect equilibrium for this class of games. Our results expand and unify the available modeling alternatives and apply to several models of interest in industrial organization, including models of industry dynamics.