An Estimable Model of Production Interactions in Endogenous Networks

We propose a model where forward-looking agents first decide to form links with each other and, then, engage in a production activity jointly with their linked peers. Exogenous linking opportunities facilitate the creation of network connections and the return to productive effort varies with the personal attributes of the connected agents. We apply our model to a purposely built dataset of college students containing information on the endogeneous networks of study partners linked with administrative records on the students' characteristics and academic performance. Identification relies upon the random assignment of students to classrooms, which generates exogenous opportunities for socialisation. Using the estimated structural parameters, we investigate the implications of two counterfactual experiments, one where students are streamed into classes by ability and one with single-sex classes.