Spillovers in space: does geography matter?

We simultaneously assess the contributions to productivity of three sources of research anddevelopment spillovers: geographic, technology and product - market proximity. To do this,we construct a new measure of geographic proximity that is based on the distribution of afirm's inventor locations rather than its headquarters, and we report both parametric and semiparametric estimates of our geographic-distance functions. We find that: i) Geographicspace matters even after conditioning on horizontal and technological spillovers; ii)Technological proximity matters; iii) Product-market proximity is less important; iv)Locations of researchers are more important than headquarters but both have explanatory power; and v) Geographic markets are very local.