In social dilemmas, leading a team by making heroic efforts may prove costly, especially if the followers are not adequately motivated to make similar sacrifices. Attempting to understand what motivates these seemingly selfless individuals to lead, we report the results of a two-stage public good experiment with endogenous timing. Even though it turns out to be costly on average, a large proportion of our subjects volunteer to lead. Our findings suggest that a fraction of these leaders are socially concerned, while others expect to distill some personal gain, possibly of non-pecuniary nature. The composition of the team also matters, as publicizing certain attributes of a subject's teammates has an impact on her decision to lead. Lastly, though voluntary leaders improve efficiency in their team, they are not necessarily more influential than randomly imposed leaders.