Industrial Agglomeration and Spatial Persistence: Entry, Growth, and Exit of Software Publishers

We use geocoded administrative data from Texas on all business establishments to estimate the effects of localization economies on the spatial persistence of industrial employment for the software industry. We decompose this persistence into components arising from entry rates, firm growth, and exit rates. Unlike previous research that has used geographies based on county and MSA divisions, this analysis takes place at a very high level of spatial resolution in which the industrial composition is identified within areas as small as one mile in radius. The choice of the software industry allows us to isolate the effects arising from human capital spillovers and the effects arising from the labor pool channel from other sources of agglomeration economies. Moreover, the decomposition of the employment persistence in entry, growth and exit, and the high level of spatial resolution allow us to distinguish between these two effects and has a number of other advantages. The results suggest that a location, defined as a 1-mile radium circle, with an initial concentration of software industry employment, retains a disproportionate number of software industry employees 6 years later. Software industry employment in the surrounding area has a small and often insignificant effect, i.e., any agglomeration effects dissipate rapidly over space. The results are not driven by higher growth rates of software establishments in high concentration locations or by differences in the survival probabilities. Rather, they are fully accounted for by two factors: (i) the retention of jobs lost by an establishment in a location by other establishments in that same location and (ii) an increased propensity of software establishments to enter in or near locations with prior software establishment presence. The entry effect diminishes sharply beyond one mile. These findings are mostly consistent with labor channel effects, including the possibility of spin-offs locating near existing firms, but disembodied human capital spillovers might also be present to some extent.