Cournot or Walras? Agent Based Learning, and Long Run Results in Oligopoly Games

Recent literature shows that learning in oligopoly games might in the long run result in the Cournot or in the Walras equilibrium. Which outcome is achieved seems to depend on the underlying learning dynamics. This paper analyzes the forces behind the learning mechanisms determining the long run outcome. Apart from the fact that there is a difference between social and individual learning, the key parameter is shown to be the degree of rationality of the learning agents: Learning the Cournot strategy requires the agents to acquire a large amount of information and to use sophisticated computational techniques, while the Walras strategy can be shown to be a particular ‘low rationality result‘.