A Structural Bayesian Approach To Estimating Inter-Organizational Network Formation

Abstract

Research on inter-organizational networks has extensively examined the determinants of network formation to understand the performance di fferentials among firms. Recent empirical studies that examine this question have used exponential random graphs models (ERGM) to account for the endogenous nature of networks, wherein prior ties influence future relationships established by organizations. In this study, we develop a structural model of network formation that generates exponential random graphs as strategic equilibrium. The advantage of structural modeling is that each parameter has an economic interpretation in addition to the statistical meaning, allowing us to understand fi rms' incentives to create, maintain and delete links. We estimate our model using a Bayesian approach that improves on existing methods by minimizing problems of degeneracy and convergence. An application to the co-investment network of venture capital fi rms in the medical device industry suggests that controlling for the endogenous network structure in equilibrium is crucial to provide a good fi t of real-world network data.