Household Registration System Reform and Firm Innovation: Evidence from China’s Scaled Industrial Firms

ABSTRACT

China has relaxed its household registration system (HRS), making it easier for rural-to-urban migrants to settle in cities. By improving the stability of high-educated workers in firms, the HRS reform should enhance these firms’ innovation. We investigate how the HRS reform influenced scaled industrial enterprises’ innovation by exploiting the staggered HRS reform in different cities from 1998 to 2007. We find that the HRS reform significantly increased these firms’ innovation output, measured by patent counts. We further find that this positive effect mainly existed among knowledge-intensive firms.

KEYWORDS:

• Household registration system reform
• firm innovation
• patent
• China

JEL:

• O10
• O30
• L26
• P20

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Data come from the Statistical Bulletin of National Science and Technology Expenditures in 2018.

2. Data come from the website of the SIPO.

3. The cases include total output less than or equal to zero, export value less than zero, fixed assets less than zero, and fixed assets greater than total assets.

4. We also use design patent counts as the dependent variable and find that the coefficient on the reform dummy is insignificant. It is not surprising as design patents are not innovative. Additional to the number of invention patents, Chen et al. (2020) uses two other innovation measures: one is the sum of the number of utility patents and design patents and the other is the sum of the number of invention patents, utility patents, and design patents. We thus further use these two innovation measures as well as the sum of invention patents and utility patents to estimate the effect of the HRS reform on firm innovation. As shown in columns (1) to (3) of Appendix Table A2, the coefficient on the reform dummy remains positive and significant at the 1% level no matter which measure is used. Additionally, as a robustness check, we also use citation-weighted number of invention patents as the dependent variable and reach similar results (not reported).

5. Our housing price data come from Fang et al. (2016). The data cover 123 cities from 2003 to 2013. This data set has been widely used in the literature (e.g., Chen and Wen 2017; Glaeser et al. 2017; Wei, Zhang, and Liu 2017). To calculate a city’s house price in 2002, we simply use its housing price growth rate in 2003 as its growth rate in 2002.

6. Matched firms in control and treatment cities can bring about a more similar set of characteristics, making them more comparable and reducing systematic confounding effects. Following Chen et al. (2020), we select three firms using the following criteria for each firm in the treatment city. First, the firms in the control group are in the same industry as the firms in the treatment group; second, the distance from the firms in the treatment group is within 100 miles; and third, the firms are the closest to the firms in the treatment group. Based on the constructed matched samples, we redo the baseline regression and report the results in Appendix Table A4. It shows that the coefficient of the reform dummy is significantly positive at the 5% level, indicating that our results still hold using the matched samples

7. We also repeat these regressions in the same period using the sample of listed firms. We find the coefficient on the reform dummy insignificant. It is also true when the dynamic effect is examined. These findings suggest that the short-term effect was ignorable among listed firms during this period.

8. Our study ends in 2010 due to a lack of firm characteristic variables in the ASIE database after year 2010. Specifically, the dependent variables, invention patents and utility model patents are from 1999 to 2010, and we lag the independent variables by one year. One city, Taiyuan had the HRS reform in 2007; five cities in 2008, including Anshan, Dalian, Kunming, Shenyang, and Zhuhai; two in 2009, Guangzhou and Qiqihar. Since the data in Kunming show an abnormal pattern after 2007, we exclude the observations in Kunming from 2008 to 2010.

9. See Chen et al. (2020) for more evidence in this direction.

10. We do so also because only in 2004, the ASIE data disclosed relevant information about the composition of employees’ education level, categorized by junior high or below, senior high, associate, college, and above.