Population agglomeration and the effectiveness of enterprise subsidies: a Chinese analysis

ABSTRACT

This paper investigates the impact of population agglomeration on the relation between government subsidies and firm innovation. It constructs a parsimonious model to show that the positive effect of government subsidies on innovation by small-cap enterprises should be pronounced only in regions with higher population densities. Moreover, using the ‘mass entrepreneurship and innovation’ policy implemented by the Chinese government in 2015, and the resulting boost in government subsidies to small-cap enterprises as a natural experiment, the paper confirms the theoretical prediction and demonstrates that population density strengthens the positive relationship between government subsidies and firm innovation in small-cap enterprises. When the population density of a region is below a certain threshold (1100 people/km²), the positive connection between government subsidies and technological innovation disappears.

KEYWORDS:

• technological innovation
• government subsidies
• population agglomeration

JEL:

• H2
• O3
• R1

ACKNOWLEDGEMENTS

The authors are thankful for the insightful suggestions made by the associate editor, Professor Derudder, and three anonymous reviewers. The authors are listed as joint first authors in alphabetical order and contributed equally to this study.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. The Baidu index did not include the phrase ‘大众创业万众创新’ (mass entrepreneurship and innovation) into its searching vocabularies until almost 2015, which is consistent with our conjecture. Results for ‘创业创新’ (entrepreneurship and innovation) are also provided for comparison.

2. Here we use $(M_i+S_i)$ instead of $S_i$ to represent the total market power after receiving government subsidies. Simply using $S_i$ would make the second term at the right side of equation (1) equal to 0 in the case that $S_i=0$, which is obviously in conflict with the reality that enterprises can also have some market power even without any subsidy.

3. If $H_j^{h-l}<K_i-M_i$, we have $\mathrm{\partial }{R}_i/\mathrm{\partial }{S}_i<0$ from equation (4), implying that the crowding effects of subsidies are dominant in regions with low population densities. We consider the effects on R&D inputs in the fourth section.

4. Our model also indicates a potential final equilibrium that innovative firms gather in denser regions and non-innovative firms in less dense regions, which is in line with the reality (take the aforementioned example, the siting of Huawei Co., Ltd and Foxconn Technology Group). The reasons are as follows. First, agglomeration effect results from economies of scale and network externality; however, there are also many factors that prevent the clustering of economic activity, for example, problems of crowding and administrative regulation (like the ‘Hukou’ system in China), the trade-off between economies and diseconomies allows cities to grow, but keeps them from becoming too large. Second, firms with large capital and enough market power may have a less urgent requirement to enlarge their markets share through innovation, especially those highly mechanized firms that obtain profits mainly from capital instead of innovations, moving to highly densely populated regions is unnecessary for them. In contrast, firms willing to innovate more could exploit the advantages of high innovation efficiency in regions with high population density.

5. For details, see http://live.china.org.cn/2015/03/14/premier-li-keqiang-meets-the-press/.

6. About US$9.40 billion, by the average exchange rate in the last quarter of 2015.

7. About US$15,660, by the average exchange rate in the last quarter of 2015.

8. The ‘mass entrepreneurship and innovation’ policy mainly increased government subsidies for small firms in China. For details, see http://old.moe.gov.cn//publicfiles/business/htmlfiles/moe/s8203/201409/175089.html.

9. We use the total assets to size firms due to the consideration that firm innovation belongs to the field of corporate finance, in which most literature use total assets as the measure of firm size. Dang, Li, and Yang (2018) investigate 100 empirical papers that involve measures of firm size from top finance, accounting and economics journals.

10. The authors thank the anonymous referee for this suggestion.

11. For details, see http://caifuhao.eastmoney.com/news/20190406074056405687390; and http://www.mnw.cn/news/cj/1784924.html.

12. The detailed results are reported in Figure A1 in Appendix A in the supplemental data online.

13. The detailed results are reported in Table A1 in Appendix A in the supplemental data online.

14. Chinese patent law requires applications for invention patents have outstanding substantive characteristics and remarkable progress compared with existing technology. The requirement for utility model patents is that they have substantive characteristics and progress relating to the shape or structure of products which are fit for practical uses. Design patents refer to new designs relating to the appearance that creates aesthetic feelings and are suitable for industrial applications. Obviously, invention patents are the most original and design patents the least (He, Tong, Zhang, & He, 2018). Owing to the differences among the three types of patents, we use two variables for the patents counted separately and a variable for the sum of the three types as explanatory variables to observe the effect of government subsidies on patent applications. In a certain sense, Invention and Utility measure high- and low-quality innovations, respectively.

15. We control for two-digit industry fixed effects according to the guidelines for the classification of listed companies revised by the China Securities Regulatory Commission in 2012. Moreover, we also take all explanatory variables in lags to ensure the robustness of results. The authors thank the anonymous referee for this suggestion.

16. See http://cn.gtadata.com/. Specifically, data for R&D expenditure, government subsidy and patent application are from the sub-database of R&D and Innovation (https://www-gtarsc-com.web.bisu.edu.cn/SingleTable/DataBaseInfo?nodeid=34503); financial information is from the sub-database of Financial Reports.

17. The authors thank the anonymous referee for reminding us to report Table 3.

18. The whole sample includes 20,617 observations; thus, it about 412 of them have to be winsorized. Specially, in Table 3, more than 25% of the patent intensity variables and R&D intensity take the value of 0, thus only 206 observations are affected with this cleaning process for these four variables, Invention, Utility, Tpatent and RDS.

19. Before the empirical tests, we verify the parallel trend assumption to ensure the validity of the treatment group and the DID method in our sample. The parallel trend test shows that the treatment and control groups have parallel trends before the policy shock, and that the treatment group experiences a significant increase in government subsidies in 2016, which indicates that our construction of the treatment group satisfies the parallel trend assumption of the DID method. The detailed results are reported in Table B1 in Appendix B in the supplemental data online.

20. Specifically, we perform several robustness tests: (1) using the sample of Chinese industrial enterprises; (2) alternatively taking the firms with a subsidy surge after encountering the policy shock in 2015 as treated firm; (3) using the subsample of private enterprises; (4) replacing the measure of population agglomeration with employment density; (5) standardizing government subsidies with a firm’s operation revenue instead of total assets; and (6) using the bottom 10% of firms by size as the treatment samples. The results from these tests are all consistent with the threshold argument of population agglomeration on the relation between government subsidies and firm innovation, as reported in Tables 5 and 6. For detailed results, see Appendix D in the supplemental data online.

21. When we include density as a continuous variable in the regression, the coefficient of the quadruple interaction, GS*Post*Treat*Density, is not significant. With the threshold argument, density > 1100 people/km² does not show an even greater return for higher levels of density. Thus, our theoretical and empirical results both support a binary effect instead of a continuous effect of population density.

22. For detailed results, see Table C1 in Appendix C in the supplemental data online.

Additional information

Funding

Kebin Deng acknowledges the financial support from the Fundamental Research Funds of the Central Universities [grant number XYZD201905]. Zhong Ding acknowledges the financial support from the National Social Science Foundation of China [grant number 12CJY047] and the Guangdong Higher Education Training Program for Outstanding Young Teachers [grant number YQ2015064]. Mingli Xu gratefully acknowledges the financial support from the Humanities and Social Science Foundation of the Ministry of Education of China [grant number 19YJC790159], the Guangdong Higher Education Project for Young Innovative Talents [grant number 2018WQNCX015] and the South China Normal University Research Foundation for Young Teachers [grant number 18SK17].