Localization, knowledge spillover, and R&D investment : evidence of Chinese cities

We develop a simple Cournot type, two-stage competition model that reveals that firms tend to reduce their R&D investment more in denser locations than in lesser ones with the presence of knowledge spillover. This implies that local agglomeration increases the negative impact of knowledge spillover on firms' R&D efforts. We then use high-tech, firm-level data from China to test our theoretical predictions. We develop a technological similarity index as a proxy for knowledge spillover and use total employment in the same two-digit Chinese Standard Industry Classification System industry as a proxy for localization agglomeration. Our Tobit model yields estimated results that are consistent with our predictions. That is, R&D effort is negatively correlated with knowledge spillover, and the magnitude of the negative relationship increases with localization agglomeration. These results are robust by different subsamples and different estimators. We also find evidence suggesting that cost-saving, rather than expropriation-avoidance, is the primary reason for firms' R&D investment reduction with knowledge spillover.