Rational Behaviour in Extensive-Form Games

Abstract

It is argued that subgame perfect equilibrium behavior is not an inevitable consequence of common belief of rationality among players in noncooperative extensive-form games. The argument centers around a particular extensive-form game with perfect information. In the context of this game, it is argued that if rationality is not common belief throughout play of the game, then it can be perfectly rational to play contrary to subgame perfection, even if rationality is common belief at the beginning of the game. It is then shown that it is impossible for rationality to be common belief throughout every play of the game.