Industry Choices and Social Interactions of Entrepreneurs: Identification by Residential Addresses

This paper shows that industry choices of entrepreneurs are determined by their social networks. The separation of residential and business addresses helps us establish the causality, because we can safely argue that residential addresses determine social networks but do not directly affect industry choices. In a large cross-section of London neighborhoods (average area: one squared-mile), we show that new generation of entrepreneurs are more likely to enter industries overrepresented by their residential neighbors. We further show that industry composition of a neighborhood is more persistent when social interactions are more intensive and of higher quality, as proxied by higher ethnic homogeneity, more sociable housing structures, or higher entrepreneurial population density. The effect is also stronger in industries that require more informational interactions, as proxied by greater geographic agglomeration of entrepreneurs. The median home- business distances in our sample is nearly six kilometers, thus the persistence of a neighborhood’s industry specialization is unlikely to be driven by unobservable common product market conditions. We also control for industry specialization at borough level (each borough contains around 20 neighborhoods), to further remove the effects of unobservable factors. Finally, we also use various sub-groups of entrepreneurs to test for a series of alternative hypotheses, and we do not find support for them. We do not find failure rates to be significantly different for entrepreneurs who follow their neighbors’ popular choices. Overall, the results suggest that entry of new entrepreneurs tend to reinforce agglomeration, while exits do not reverse it. These evidences (weakly) lend support to the existence of agglomeration economies.