Introduction

The Folk Theorem suggests the possibility of cooperation through a self-enforcing agreement. The phrase ‘Folk Theorem for repeated games’ refers strictly to a theorem stating that a certain large class of single-shot game payoffs can be obtained as Nash equilibrium payoffs if the game is repeated infinitely many times and the players do not discount their payoffs (Aumann, 1960). It refers more broadly to similar results, supported by subgame perfect equilibria (SPEs), for games that are finitely repeated and infinitely repeated games in which players may discount payoffs. Perhaps the most striking implication of these results is the possibility of apparently cooperative outcomes in non-cooperative games when these games are knowingly played many times by a fixed collection of players.

I have been aware of, and impressed by, the work of Louis Phlips for about a quarter of a century. His work on competition policy is meticulous and distinguished. The topic of my contribution bears on competition policy, because the Folk Theorem literature spells out means by which firms can attain outcomes that appear collusive without, necessarily, engaging in overt collusion – or, indeed, even discussing together what to do.

Infinite horizon oligopoly is a well known example of a repeated game. The Cournot model is the most widely taught example and it is typical in that the Cournot equilibrium (i.e. the Nash equilibrium of the single-shot Cournot market) generally results in a payoff profile that lies inside the payoff possibility frontier.