Startup Location, Local Spillovers and Neighborhood Sorting
Abstract How critical is spatial concentration for the success of startup firms? This paper uses data on the universe of firms in large Canadian cities to study this question at the level of city blocks and their surrounding neighborhoods. To account for sorting within blocks, I use a newly developed clustering algorithm to construct neighborhoods relevant for each industry within which sorting across blocks is conditionally random. To account for sorting across neighborhoods, I develop a model of neighborhood selection, where entrepreneurs choose neighborhoods based on expected startup outcomes and preferences for location. Results show that spillovers of block average same-industry employment and revenue are hyper-local and mostly fade away after 75 meters. These spillovers have economically significant effects on startups' end-of-year revenue, and survival rates. For a sense of magnitude, going from the 10th to the 90th industry-specific percentile of incumbents' average revenue increases the median startup revenue by 8.2%. These effects are heterogeneous across industries, with employment-intensive industries benefiting relatively more from larger while knowledge-intensive industries benefiting relatively more from better incumbents firms.