Unskilled Migration and Vertical Disintegration: Theory and Evidence from China

Abstract

This paper studies how rural-urban migration in China affects the vertical disintegration of firms. We construct a theoretical model and find that the inflow of rural migrants increases the output of firms in urban areas. This leads to a higher degree of vertical disintegration of firms due to the decreasing returns to scale. We also find that the wage of skilled workers relative to that of unskilled workers in urban areas is increased with rural migrants as the vast majority of rural migrants are unskilled. Our empirical findings are consistent with our model predictions using China’s population census and manufacturing firm-level data in 2005.

Keywords:

• rural-urban migration
• vertical disintegration
• skill premium

JEL CODES:

• J61
• L23
• O15
• R23

Acknowledgements

We are grateful to the Editor, and two anonymous referees for very useful comments that significantly improved the paper. We thank Guobin Hong from the Key Laboratory of Econometrics in the Ministry of Education (Xiamen University) for the technical assistance. Jianan Li acknowledges the financial support of the National Natural Science Foundation of China (grant number 72173102).

Disclosure statement

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

Notes

1 See United Nations report (2015).

2 According to China’s population census, about 83.5% of rural-urban migrant workers are under 12 years of schooling, compared to 38.5% of local urban workers.

3 Jones (2013) assumes that all intermediate inputs are produced by final goods and that the fraction of final goods used to produce intermediate inputs is fixed.

4 As mentioned by Imbert et al. (2022), the key identification assumption underlying the shift-share design we use is that shifts (vthat is agricultural income shocks) are numerous and as good as random (Borusyak et al., 2022), which is supported by our empirical tests in Section 4.2.

5 According to Jones (2013), intermediate goods are only one other form of capital.

6 Jones (2013), to focus on how intermediate goods lead to greater multipliers, assumes that each firm appropriates a fixed fraction of output to produce intermediate goods.

7 The amount of labour used to transit final goods into intermediate goods is normalised to 1.

8 This implies that the production function at time t can be written as $Y_t={\left\{{[\varphi H_t^{\frac{\sigma -1}{\sigma }}+(1-\varphi )L_t^{\frac{\sigma -1}{\sigma }}]}^{\frac{\sigma }{\sigma -1}}\right\}}^{\beta }{[{({\gamma }_{t-1}{Y}_{t-1})}^{\theta }+I_{B,t}]}^{1-\beta }.$

9 The labour is assumed to be immobile across borders. Thus, the price of labour is determined endogenously.

10 This approximation method has been used in many literature, for example, Bramoullé and Saint-Paul (2010).

11 2005 China 1% Population Census (2005 Mini Census) is widely used in the literature (for example Imbert et al., 2022; Meng & Zhang, 2010; Tombe & Zhu, 2019). Additional data for robustness tests are obtained from the 2000 China 0.095% Population Census and 2005 China Custom data.

12 Regarding the distribution of education, 83.5% of rural migrant workers stayed in school for 12 years or less. In contrast, 61.5% of urban workers have more than 12 years of schooling.

13 Acemoglu et al. (2010) find that the technology intensity of producer industries is positively correlated with the likelihood of vertical disintegration.

14 Here, according to the similarity of cultural and economic growth across cities, we divide China into three regions: Eastern China including Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and Hainan; Middle China including Shanxi, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, and Hunan; Western China including Sichuan, Chongqing, Guizhou, Yunnan, Tibet, Gansu, Qinghai, Ningxia, Xinjiang, Shannxi, Guangxi, and Inner Mongolia.

15 Following Imbert et al. (2022), the 21 commodities/crops that we use for potential agricultural output construction include banana, cassava, coffee, cotton, fodder crops (barley), groundnut, maize, millet, other cereals (oats), potato, pulses (lentil), rapeseed, rice, sorghum, soybean, sugar beet, sugar cane, sunflower, vegetables (cabbage), tea and wheat. These account for 80%–90% of the total agricultural output over the period.

16 We use the Agricultural Producer Prices from the FAO as the commodity prices.

17 In this regression, t denotes the year fixed effects, and c denotes the crop fixed effects.

18 We define the neighboring cities of city j as all cities except for j.

19 This supports the theoretical finding that the inflow of rural-urban migrants increases the skilled labour supply in urban areas H (see section 1 of the Webappendix, which shows that the urban skilled labour supply H increases with the rural-urban migration).