A Structural Model of Firm Collaborations with Unobserved Heterogeneity

Abstract

We develop and estimate a structural model of strategic network formation to study the determinants of firms collaborations for patenting new technology in the medical device industry. Our aim is to bridge the strategy literature on interorganizational networks and the economic literature on structural estimation of network models. In our model, firms have payoffs that depend on linking costs and benefits, as well as externalities from common partners and popular partners. Firms are characterized by observed and unobserved characteristics, that affect both their opportunity and their willingness to form links. The equilibrium networks are sparse and match the aggregate clustering levels observed in the data. We use the network of patent collaborations among medical device firms, to estimate the structural parameters using a Bayesian approach. Our results show that firms tend to partner domestically and collaborate with companies in similar markets, perhaps due to technological complementarities or regulation effects. Unobserved heterogeneity matters: we find that firms' payoffs vary by type. Finally we show that the estimated model including unobserved heterogeneity provides a better fit of crucial features of the data.