University patenting and knowledge spillover in Japan: panel-data analysis with citation data

Abstract

University-based patents, by their nature, tend to embody scientific knowledge, which can contribute to technological innovation. University patents, therefore, can be an important catalyst between science and technology. This article examines the contribution of university patents to technological innovation. Using patent citation data, we investigate how the knowledge embodied in university patents influences technological innovation. We conclude that university-based patents convert scientific knowledge into generic knowledge and contribute to technological innovation.

Keywords:

• R&D
• innovation
• knowledge spillover
• patent citations

JEL Classification:

• O32
• O33
• O53

Notes

¹ Science linkage is developed by National Institute of Science and Technology Policy (NISTEP) based on the data by CHI Research Inc. The index is designed to indicate the interaction between scientific and technological research. It is defined as the ratio of the number of citations to scientific papers to the number of patents.

² The index indicates the range of technological fields influenced by a patent. If a patent is cited by other patents in a wide range of technological fields, then the generality index is high.

³ When the main inventors are affiliated with universities, we consider the patents as university-based.

⁴ As the focus of Maruseth and Verspagen (2002) was the inter-regional knowledge spillover, they examined citations between two regions, whereas our analysis focuses on knowledge spillover in general. Therefore, we consider the overall number of citations.

⁵ Some previous studies, including Jaffe and Trajtenberg (1998), assume that the number of citations is a nonlinear function of time. This is primarily because these studies focus on estimating the distribution of the time lag associated with citations. These studies consider two distinct impacts of the time lag: diffusion and obsolescence. The nonlinearity of the model is assumed to address these two impacts.

However, unlike these previous studies, we focus on the propensity, in which each patent is cited by the others. Since our main interest is in the relative importance of university patents, we do not consider such distinct impacts. The variable YEAR is only included to control for the joint effect of the time lag in each patent.

⁶ This is because the use of cumulative stock of patents may impose a serious endogeneity problem.

⁷ R&D expenditures are obtained from Science and Technology Indicators by OECD.

⁸ Pedroni panel cointegration tests have rejected the null hypothesis of no cointegration (Table A1 in the Appendix). Since we are interested in the long-run relationship, we will not discuss the short-run dynamics of the model. The results of the error correction model are presented in Table A2 in the Appendix.

⁹ We have also estimated the model with Autoregressive (AR) term, using the lagged values of the dependent variable and the regressors as instrumental variables. We have observed that the inclusion of AR term does not considerably alter our findings.