We provide a parsimonious model of leadership in social dilemma situations and test it with a meta-analysis of experimental studies. We focus on studies with treatments that allow for sequential contributions to a public good [as in Güth et al. (J Public Econ 91:1023–1042, 2007)]. The group members observe the contribution of a leader before contributing themselves. We compare the results with simultaneous contribution treatments from the same studies. Our results confirm that the establishment of a leader indeed leads to persistently higher and more coordinated contributions. As predicted, the aggregate effect remains stable over time and increases in group size even though leaders and followers have more divergent contribution patterns in larger groups. We also find empirical support for an explanation of the observed ‘leader’s curse’.