Social dilemmas such as greenhouse gas emission reduction are often characterized by heterogeneity in benefits from solving the dilemma. How should leadership of group members be organized in such a setting? We implement a laboratory public goods experiment with heterogeneous marginal per capita returns from the public good and leading by example that is either implemented exogenously or by self-selection. Our results suggest that both ways of implementing leadership only have small effects on contributions to the public good. Self-selected leaders—in particular self-selected low-benefit leaders—tend to set better examples than imposed leaders, but they are also exploited more strongly by followers. Leaders seem to need additional instruments to be more effective when benefits are heterogeneous.