We conduct a randomized controlled trial to study how the self-selection of individuals into teams changes the impact of a team’s gender composition on gender preferences, team performance, and individual performance. We randomly divide a sample of high-performing high school students into two groups: we assign students in one group to teams of varying gender composition, and we allow the students in the other group to form teams freely. We find that the latter choose more male-predominant teams than the former, and if self-selected into gender-biased teams, students prefer even more gender-biased teams ex-post. We also find that female-predominant teams underperform other types of teams when we form teams exogenously, but these differences disappear when students form teams endogenously.