Group Membership and Communication in Modified Dictator Games

This paper presents a laboratory experiment to measure the effect of group membership on individual behavior in modified dictator games. The results suggest that this effect is influenced by the degree of group membership saliency. A within-subject design is employed: in stage 1, each subject decides individually; in stage 2, the subjects are divided into groups of three and one person is selected at random from each group to make the decision (the 'hierarchical decision rule'). In stage 3, additional pre-play communication in the group is allowed before the decision and, in stage 4, the decisions are again made on an individual basis. Interestingly, the dictators behave more selfishly when group members are not allowed to communicate. However, if groups are allowed to communicate, decisions do not differ from individual choices. Chat content shows that groups are concerned with reaching a consensus, even though talk is 'cheap' and only one group member will make the binding decision.