Unintended triadic closure in social networks: The strategic formation of research collaborations between French inventors

Most of the empirical and theoretical literature aimed at understanding the behavioral patterns that lead to the formation of social networks argue that such networks are clustered because agents like social closure, since it facilitates cooperation enforcement, for instance, or increases match quality. We argue that, in certain circumstances, network clustering may arise for other reasons, even though agents may actually not like redundancy in connections. We propose a theoretical model of the formation of new research collaboration that we estimate on the longitudinal evolution of the French co-invention network. We show that if this type of social network is closed it is because it correlates with exogenous metrics affecting the costs of direct link formation, not because agents prefer to close triangles per se. This result is obtained thanks to the richness of our dataset, allowing us to control for dyadic fixed-effects and various costs of network formation (geographical distance, technological specialization, and institution boundaries and attributes) omitted in previous studies.