To determine if spillovers from university research are localized or extend hundreds of miles.
BACKGROUND
Anecdotal evidence of links between US university research and growth of high technology industries such as biotechnology, computer software, and semiconductors reinforces the view that proximity to research universities is important to innovation and economic growth. Belenzon and Schankerman (2013) use the choice sampling method to document state-level localization effects for knowledge spillovers for patents assigned to universities. Mowery and Ziedonis (2007) investigate patent-to-patent citations from four universities to firms at the metropolitan level disaggregated by technology. They find that distance is negative and significant and distance squared is positive and significant.
Our contribution is providing a more precise estimate of the geographic distance over which university knowledge affects others using a spatial interaction model.
DATA/METHODS
In a spatial interaction model the knowledge flow (patent citations) between the university (origin) and metropolitan area (destination) depends on the mass at each location measured by the stock of patents and the separation between them. Separation is quantified by physical distance, a state border effect, and technology compatibility. Distance is measured by indicator variables for distance bands between the university and metropolitan area. Spatial effects are incorporated to capture origin and destination dependence.
We use the NBER patent database. Citing patents are between 1990 and 1999 and the cited patents (backward citations) are between 1963 and 1999. Data on 92 universities (origin) and 143 metropolitan areas (destination) yields 13,156 pairs. The mean citation flow is 3.13 and 57 percent of the pairs had a zero citation flow. The maximum citation flow is 807 between the Massachusetts Institute of Technology (MIT) and firms located in Boston. Because of pairs with zero citation flow, the model is estimated using Poisson and negative binomial regression.
RESULTS
Preliminary findings show that the distance indicator variables have an above average number of citations made to patents granted within 200 miles. Beyond 200 miles the impact drops to 10 percent or less, and becomes negative at 1000 miles. For firms in the same city as the university, citations are 1196 percent above average. Technology compatibility has a positive impact on knowledge flow.
POLICY IMPLICATIONS
The result that academic knowledge spillovers are highly localized and decay rapidly as distance increases supports managerial decisions on the location of R&D near universities and state policy that encourage university research.