Risk and Temptation: A Meta-Study on Social Dilemma Games

This paper reports the results of a meta-study of 89 prisoner's dilemma experiments comprising more than 3000 participants across 6 countries. We organize existing evidence and explain seemingly contradictory results in the existing literature by focusing on two dimensions of the dilemma: "risk'' (the percentage loss of unilaterally cooperating against a defector) and "temptation'' (the percentage gain of unilaterally defecting against a cooperator). We find in particular, that (i) risk best explains the variation in cooperation rates across random matching ("stranger'') and one-shot treatments, while (ii) temptation best explains the variation in repeated ("partner'') interactions, (iii) consistently with reputational models temptation is more effectively contained the longer the time horizon of the game, (iv) there is more cooperation in partner than stranger if and only if risk is high and temptation low and (v) women are more cooperative than the average man if risk is low and less cooperative if risk is high, but there are no gender differences on average. These results can be useful for discriminating between competing theories of why people cooperate. For policy making, it is important to understand which dimension of the dilemma (risk or temptation) is best targeted via interventions to yield improved cooperation.