(Optimal) spatial aggregation in the determinants of industrial location

Empirical studies on the determinants of industrial location typically use variables measured at the available administrative level (municipalities, counties etc.). However, this amounts to assuming that the effects that these determinants may have on the location process do not extend beyond the geographical limits of the selected site. We address the validity of this assumption by comparing results from standard count data models with those obtained by calculating the geographical scope of the spatially varying explanatory variables using a wide range of distances and alternative spatial autocorrelation measures. Our results reject the usual practice of using administrative records as covariates without making some kind of spatial correction.