Collaborate to Compete: An Empirical Matching Game Under Incomplete Information in Rank-Order Tournaments

This paper studies the collaboration of talents in rank-order tournaments. We use a structural matching model with unobserved transfers among participants to capture the differentiated incentives of participants that spur collaboration, with a specific focus on incorporating incomplete information and competition in the matching game. We estimate our model using data from a leading data science competition platform and recover the heterogeneous preferences and abilities of participants that determine whether and with whom they form teams. Overall, teamwork enhances performance and competition fosters collaboration, whereas incomplete information about potential coworkers’ ability hinders collaboration. Using the estimation results, we conduct counterfactuals to investigate how the information on potential collaborators’ ability and competitive pressure affect collaboration and performance outcome. Our results suggest that the platform could further improve collaboration and yield better outcomes by providing more informative signals of ability and further concentrating the allocation of rewards to top performers.