The geographic reach of knowledge spillovers: A functional regression approach with precise geo-referenced data

This paper applies functional regression to precise geo-coded register data to measure productivity spillovers from high-skilled workers. We use a smoothing splines estimator to model the spatial distribution of high-skilled workers as continuous curves. Our rich panel data allows us to address spatial sorting of workers and the entanglement of spillover and supply effects with an extensive set of time-varying fixed effects. Our estimates reveal that spillovers from highskilled workers attenuate monotonously with distance. Effects disappear after approximately 20 kilometers. Furthermore, our findings illustrate the benefits of applying functional regression to modern (spatial) economic data.