Repeated Congestion Games with Bounded Rationality


International Journal of Game Theory

August 2012, vol. 41, n°3, pp.651-669

Departments: Economics & Decision Sciences, GREGHEC (CNRS)

Keywords: Folk theorem, Braess’s paradox, Network routing games, Private monitoring, Public monitoring, Anonymous games, Strategic complexity, Contagion strategy, Calendar strategy

We consider a repeated congestion game with imperfect monitoring. At each stage, each player chooses to use some facilities and pays a cost that increases with the congestion. Two versions of the model are examined: a public monitoring setting where agents observe the cost of each available facility, and a private monitoring one where players observe only the cost of the facilities they use. A partial folk theorem holds: a Pareto-optimal outcome may result from selfish behavior and be sustained by a belief-free equilibrium of the repeated game. We prove this result assuming that players use strategies of bounded complexity and we estimate the strategic complexity needed to achieve efficiency. It is shown that, under some conditions on the number of players and the structure of the game, this complexity is very small even under private monitoring. The case of network routing games is examined in detail.