"No Man Is an Island": An Empirical Study on Team Formation and Performance

Many organizations rely on decentralized arrangements where employees choose their projects and teams. Most of the empirical literature on working collaborations instead focuses on teams that are exogenously formed. I develop a structural entry model with heterogeneous strategic interactions where agents decide whether to join a project. The decision depends on who else may potentially join the project, the project quality, as well as other individual and project characteristics. In turn, this decision affects the probability of project completion. I estimate the model using a novel dataset from an important scientific collaboration. I find that agents' decisions to select into projects highly depend on the pool of teammates and the size of the team whereas projects' quality is of lesser importance. Heterogeneity in agents' characteristics explains this selection, which needs to be accounted for to obtain unbiased estimates of teams' performance. With a counterfactual experiment, I show that moving from a decentralized to a centralized arrangement leads to fewer completed projects.