Do self-formed teams perform better than other team structures? Using unique data from Virgo, a Nobel-prize-winning scientific organization with self-formed teams, first, I uncover new evidence on team formation and performance. Then, I develop a structural model to i) estimate which teams perform better controlling for self-formation and ii) evaluate the performance of counterfactual team structures. Regarding i), estimation results show that small teams perform better than large teams. Regarding ii), counterfactual results show that randomly formed teams perform worse than the observed self-formed teams, and teams with a more diverse membership perform better.
