Growing Cooperation

Cooperation is one of the most important determinants of the success of human groups. However, groups do not always succeed in developing norms of cooperation, especially with anonymous interactions. For instance, cooperation is seldom observed in indefinitely repeated prisoners dilemma experiments with stranger matching. This is the case even when the time horizon, the size of the community and the payoff from cooperation make cooperation theoretically sustainable. In this paper, we test whether initiating small groups that gradually expand foster cooperation in an indefinitely repeated stranger-matching prisoner dilemma game. Starting small is a common and intuitive way to build large groups. It may provide suitable conditions to develop virtuous norms of cooperation. However, whether such a mechanism is effective has not been tested in a clean lab experiment. Our experiment comprises 3 parts. In each part, participants play an indefinitely repeated prisoners dilemma. In the baseline, matching groups start out large, with 8 members. The matching groups remain fixed for the 3 parts of the experiment, and participants interact with strangers for the length of the experiment within their marching group of 8. In the treatment, groups start out small and then grow larger: In the first part, participants are matched in pairs and play the indefinitely repeated prisoners dilemma in partners' matching. In the second part, each pair is merged with another randomly selected pair, and participants play the indefinitely repeated prisoners dilemma in stranger matching within the resulting group of 4. In the last part, each group of 4 is merged with another randomly selected group of 4, and participants play the indefinitely repeated prisoners' dilemma in stranger matching within the resulting group of 8. We find that cooperation is significantly higher in the treatment. We also find that, in the treatment, cooperation decreases as the size of the community increases. This is mostly due to a substantial decrease in cooperation between the first and the second part, i.e. when matching switches from partner to stranger. Turning to our main hypothesis, we find that cooperation in the third part of the experiment is significantly higher in the treatment, where growth was experienced, than in the baseline. This suggests that letting groups grow from small to big provides suitable conditions to develop a norm of cooperation.