Abstract:
Markov perfection has become the usual solution concept to determine the non-cooperative equilibrium in a dynamic game. However, Markov perfection is a stronger solution concept than subgame perfection: Markov perfection rules out any cooperation in a repeated prisoners' dilemma game because the history of previous cooperation does neither change the future action space nor the possible payoffs in this setting. This paper demonstrates that a dynamic modelling approach may sustain cooperation by Markov perfect strategies in situations which are usually modelled as repeated prisoners' dilemma games. The idea is that past defection from cooperation changes a compliance state variable which enters the utility function. The corresponding dynamic games are discussed for the trigger strategy and for a strategy which is weakly renegotiation-proof. Finally, the paper shows that dynamic game modelling improves the chances for strong renegotiation-proofness in the corresponding repeated game.