The role of proximity to universities for corporate patenting: provincial evidence from China

Abstract

This paper investigates whether proximity to universities matters for corporate patenting in Chinese provinces. The investigation is based on estimating regional knowledge production functions using a Chinese provincial data set for the years from 2000 to 2008. Geographic proximity of companies to universities is taken as a key element to measure firms’ accessibility to university research. In addition, quality-adjusted accessibility measures are considered in extended models to take into account quality difference in university research. The results suggest the existence of spatial academic effects on corporate patenting activities in China as found in the previous literature for Western economies. In China, however, these effects are especially strong for realising technologically less demanding non-invention corporate patents than for invention corporate patents. Moreover, companies’ geographic proximity to universities dominates over university research quality difference for determining the relevance of universities as knowledge sources for companies. Extended models are estimated for robustness checks which ascertain the main results.

Appendices

Appendix A

The first concept applied to construct the quality-adjusted ACCE variable is as follows:

$$\begin{aligned} \bar{{d}}_{it}^\mathrm{a1}&= \left( {-1/\gamma } \right)\log \left[ {\sum _{j=1}^J {\left( {\text{ NO}_{jt}^\mathrm{uni} \exp \left( {-\delta QR_{rt} } \right)} \right)\exp \left( {-\gamma \text{ DIS}_{ij} } \right)} } \right]\end{aligned}$$

(4a)

$$\begin{aligned} \bar{{d}}_{rt}^{a1}&= \sum _{i\in r} {\bar{{d}}_{it}^\mathrm{a1} } \left( {{\text{ NO}_{it}^\mathrm{ind} }/{\sum _{i\in r} {\text{ NO}_{it}^\mathrm{ind} } }} \right)\end{aligned}$$

(4b)

$$\begin{aligned} \text{ ACCE}_{rt}^\mathrm{a1}&= -\bar{{d}}_{rt}^\mathrm{a1} \end{aligned}$$

(4c)

Compared to Eq. (3a) the exponential term ‘$\exp (-\delta QR_{rt})$’ is the only one newly added term in Eq. (4a). This calculated quality-adjusted average distance ($\bar{{d}}_{it}^\mathrm{a1}$) replaces the original average distance in the second and third step. The variable $QR_{rt}$ refers to the quality ranking of universities by province from zero to 29 with decreasing number of invention patent applications filed by universities in each province under the assumption that universities from the same provinces are of the same quality. The effect of quality differentials between university research on corporate patenting is reflected in the quality decay parameter ($\delta$). A positive value of $\delta$ means that only universities with the best quality will be counted fully as relevant universities for companies, while universities with lower quality will be counted in Eq. (4a) as if fewer universities existed. We assume $\delta$ equal to $0.01\,\text{ rank}^{-1}$ as our base value, meaning that universities with a quality of one level lower than the best ones are considered as if there were only 99 % of the existing universities relevant to companies instead of the full population of universities, assuming the same geographic distance from companies to the best universities and to the universities with a one-level lower quality. To check robustness, we consider $\delta$ equal to $0.005\,\text{ rank}^{-1 }$ and $0.05\,\text{ rank}^{-1 }$ as well.

The second concept differs from the first concept only in the first step to construct the new variable:

$$\begin{aligned} \bar{{d}}_{it}^\mathrm{a2} =\left( {-1/\gamma } \right)\log \left[ {\sum _{j=1}^J {QL_{rt} \left( {{\text{ NO}_{jt}^\mathrm{uni} }/{\sum _{j\in r} {\text{ NO}_{jt}^\mathrm{uni} } }} \right)\exp \left( {-\gamma \text{ DIS}_{ij} } \right)} } \right] \end{aligned}$$

(5)

The new element $QL_{rt}$ refers to the number of invention patent applications filed by universities in province $r$ at the time $t$. Given the same number of universities existing in city $j1$ and city $j2$, both located at the same distance from a company in city $i$, universities in city $j1$ provide more academic knowledge for that company than universities in city $j2$ if the number of invention patent applications allocated to the universities in city $j1$ is higher than that in city $j2$. As above, we assume that universities from the same province are of the same quality, indicating that the number of invention patent applications allocated to universities in each city is determined by the city’s share of universities in the same province in addition to the total number of provincial academic invention patent applications. The quality-adjusted average distance obtained, $\bar{{d}}_{it}^\mathrm{a2}$, replaces the corresponding $\bar{{d}}_{it}^\mathrm{a1}$ in Eqs. (4b) and (4c), thereby deriving the quality-adjusted accessibility measure ($\text{ ACCE}_{rt}^\mathrm{a2} )$.

Appendix B

See Figs. 1, 2, 3 and Table 6.

Fig. 1

Total corporate patent applications (invention vs. non-invention) and industrial R&D expenditure. Note Sums of the corresponding province-level statistics of 30 provinces in China are presented here. Original data source NBSC-CNSYST (various years). Own presentation

Fig. 2

Total patents (invention, utility model and external design) filed by universities. Note Sums of the corresponding province-level statistics of 30 provinces in China are presented here. Data source SIPO-SARP (various years). Own presentation

Fig. 3

Total academic journal publications and universities’ R&D expenditure. Note Sums of the corresponding province-level statistics of 30 provinces in China are presented here. The articles here refer to the Chinese scientific papers taken by major foreign referencing system such as SCI (Science Citation Index), EI (Engineering Index) and ISTP (Index to Scientific & Technical Proceedings). Data source NBSC-CNSYST (various years). Own presentation

Table 6 Key descriptive statistics of variables considered in regression models