Experts Playing the Traveler's Dilemma

Abstract

We analyze a one-shot experiment on the traveler’s dilemma [Basu, K. (1994), AER 84(2), 391–395] in which experts, i. e. members of the Game Theory Society, were asked to submit both a (possibly mixed) strategy and their belief concerning the average strategy of their opponents. Only very few entrants expect and play the unique Nash equilibrium, while we observe a fifth playing the cooperative solution of the game, which means playing a strictly dominated strategy. We show that analyzing the game as one of incomplete information, with an irrational type choosing the ‘cooperative strategy’, experimental data are in line with the prediction of Bayesian Nash equilibrium behavior. In particular strategies observed are in the possible support of equilibrium strategies. Since the Nash equilibrium strategy of the original game is not in the support of the Bayesian Nash equilibria of the incomplete information game but we do observe it being played, we add a second irrational type always playing that strategy. This does not substantially change the results.