Cooperation When Strategic Risk is Removed

Abstract

Laboratory experiments on the infinitely repeated prisoner's dilemma, and other similar cooperation games, find little cooperation when the discount factor is near the theoretical cutoff discount factor for which cooperation can be supported in equilibrium. The explanation is that the non-cooperative equilibrium is risk dominant; it is the best response to most beliefs about what the other player will choose. I study a new game where cooperation is risk dominant at the cutoff to understand whether cooperation can be empirically observed to the fullest extent that theory says is possible. The main finding is that there is still not pervasive cooperation, just 41% and 35% in period 0 of the game for the last 10 matches in two respective treatments. Cooperation is a best response to almost all beliefs in the latter treatment indicating that eliminating strategic risk is not sufficient to obtain cooperation.