Leading-by-example and third-party punishment: Experimental evidence

Abstract

In this paper we explore how decisions in a sequential three-person game are influenced by either dynamics of roles or group composition. In the game a dictator decides how much of his endowment to transfer to the recipient. The supervisor can punish the dictator and/or transfer to the recipient after learning about the dictator’s decision. We find that transfers by a dictator are highest and stable when the group of three is fixed, no matter whether roles change or not. There is limited support of a leading-by-example effect, that is, only when both role and group composition of subjects are fixed, supervisors give more the more dictators gave, and dictators transfer more in the next period the more supervisors gave in a period. Punishment partially has a disciplining effect on dictators. Finally, we observe that subjects’ actual actions are consistent with their beliefs and expectations.