Positive feedback in skill aggregation across Chinese cities

ABSTRACT

This paper examines whether human capital levels diverged across Chinese cities between 1990 and 2010. By employing a series of instruments for the initial human capital share, consistent positive feedback between current human capital levels on the subsequent growth of human capital in the 2000s is found. The findings are also robust to alternative instrumental variables. Further evidence suggests that a city's industry structure does not explain the positive feedbacks, but rather urban amenities appear to be the primary causal factor. To mitigate large provincial inequality in development, China should mitigate growing differentials in human capital by improving local urban amenities.

KEYWORDS:

• human capital
• agglomeration economies
• skill aggregation
• city growth
• regional inequality

JEL:

• R11
• R12

ACKNOWLEDGEMENTS

The authors thank the editor and Sarah Hands.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. The Cultural Revolution was a movement launched by Chairman Mao Zedong between 1966 and 1976. Its goal was to preserve ‘true’ communist ideology in the country by purging remnants of capitalist and traditional elements from society. During this period, all schools and universities were closed. Many people, especially educated teachers, scholars and students, were persecuted and often sent to the countryside. The 10-year movement virtually destroyed the educational system and significantly disrupted the economy. The Third Front movement refers to the massive industrial development led by Chinese government in remote rural areas in the south-western and north-western provinces (mostly Gansu, Shaanxi, Sichuan, Ningxia, Guizhou, Yunnan, Chongqing, Henan, Hubei, Hunan and Qinghai) between 1964 and 1980. This relocation of industry was driven by national defence considerations. Large-scale investments were mainly in national defence-related industries such as technology and basic industries, including manufacturing, mining, metal and electricity. This planned-economy policy substantially influenced the economic geography of China in the pre-market reform era.

2. The concept of ‘skill agglomeration’ alludes to a concentration of skilled workers in a limited space (i.e., density), while the concept of ‘skill aggregation’ is a much weaker notion of geographical concentration. The difference between the two terms can be seen with this example. A geographically dispersed distribution of skilled workers in a relatively large region is not a ‘skill agglomeration’, but it is still a ‘skill aggregation’ because of the large share of college graduates (but lower density). Considering that Chinese prefecture cities can be geographically large and ‘skill aggregation’ incorporates a wider set of spatial possibilities, we adopt this term to refer to the spatial distribution of skilled workers across Chinese prefectural cities.

3. The Jinshi was the highest examination for public office positions during the imperial era. It was extremely difficult and prestigious to pass it. The total number of Jinshi holders was about 14,116 during the 276-year Ming dynasty. We matched the birthplace of each Jinshi degree holder based on their bibliography to the modern administration unit to calculate the number of Jinshi holders of each city during the Ming era, the last imperial dynasty of the Han Chinese.

4. This instrument is only used when the dependent variable in the IV regressions is the change in the college graduate share between 2000 and 2010.

5. The intuition for the Lewbel (2012) instrument is that omitted variables that produce any endogeneity also produce heteroscedasticity of the residuals. This heteroscedasticity is used in the construction of the instrument.

6. Consistent boundaries are used in the calculation for each city. The Integrated Public Use Microdata Series (IPUMS) is used to generate variables to reflect consistent city boundaries between 1982 and 2000 that can differ from the actual 2000 administrative boundaries.

7. Some control variables are unavailable for 1990.

8. Amenities variables are measured between 1971 and 2000.

9. The only available control variables for 1990 are location of cities, city population and manufacturing employment share.

10. Alternatively, when using the number of universities and full-time faculty per capita (as opposed to the raw count) is adopted as IVs, the results are similar (see Table A6 in Appendix A in the supplemental data online).

11. Specifically, the sample only includes the more underdeveloped western provinces of Gansu, Shaanxi, Sichuan, Ningxia, Guizhou, Yunnan, Chongqing, Henan, Hubei, Hunan and Qinghai.

12. This analysis is impossible for the 2000s because of the confidentiality of the individual level data in the 2010 and 2005 Censuses.

13. If a large number of the ensuing college graduates remained in their college's city (which are typically large cities with more colleges), the higher education expansion would have contributed to the positive feedback mechanism, undercutting to some degree the hypotheses about base mechanisms. Thus, the finding that even cities without colleges had the same positive feedback mechanism suggests that this enrolment policy was not a key driver of the base results, though it could have other effects. To be sure, to the extent that there was a common national upward shift in the 2000–10 change in the graduate share, that would be reflected in a larger constant term in the regression and not in the 2000 college graduate share coefficient and would not affect the base results.

14. The hukou system has been significantly relaxed for high-skilled workers. Yet, to the extent that it still was binding in various cities, another way that urban amenities played a role is that high-skilled non-hukou residents would consider their accessibility to urban services in their migration decisions.

15. The healthcare and educational proxies are measured by the number of per capita, number of hospital beds per capita, number of middle-school teachers per capita and number of primary-school teachers per capita.