When dispersed teams are more successful: Theory and evidence from software

Who collaborates with whom when physical co-location is no longer a constraint? How does the geographic composition of teams influence project success? We develop a model of global team formation and collaboration in which individuals have heterogeneous and partially observable skills, collaboration incurs geographic frictions, and project success depends on the best idea developed in the team. The model yields five testable predictions: (I) collaboration is more likely among geographically proximate individuals; (II) only highly skilled individuals form long-distance collaborations due to selection effects; (III) geographically diverse teams produce higher-quality outcomes; (IV) the positive impact of diversity on success is stronger for more complex projects; and (V) there is a non-linear relationship between team size and project quality. We test these predictions using a highly granular dataset on open-source software development, where project success is measured through downstream usage. The open-source setting offers a unique empirical advantage: its transparency and absence of physical constraints allow us to isolate the role of team composition and collaboration frictions. Our findings confirm the model’s predictions and suggest that the benefits of dispersed collaboration extend beyond software development to other knowledge-intensive activities.