This paper provides a new empirical perspective for analysing the role of social networks for an economic geography approach on regional economic growth by constructing large-scale networks from employee-employee co-occurrences in plants in the entire Swedish economy 1990-2008. We calculate the probability of employee-employee ties at plant level based on homophily-biased random network assumptions and trace the most probable relations of every employee over the full period. Then, we look at the inter-plant ties for the 1995-2008 period because the network is already well developed after five years of edge construction. We argue that these personal acquaintances are important for local learning opportunities and consequently for regional growth. Indeed, the estimated panel Vector Autoregressive models provide the first systematic evidence for a central claim in economic geography: social network density has positive effect on regional productivity growth. The results are robust against removing the old and therefore weak ties from the network. Interestingly, the positive effect of density on growth was found in a segment of the co-worker network as well, in which plants have never been linked by labour mobility previously.
social network homophily probability of ties labour mobility regional productivity growth panel vector autoregression