the one hand, an increase in the productivity of agricultural labor has raised the output of the unit agricultural labor and the income of rural residents and weakened the driving forces of the transfer of agricultural labor. On the other hand, an increase in the output of agricultural products has enabled the agricultural labor to be released from industrial sectors and prompted its transfer. The total mechanical power of agricultural labor (the ratio of the total agricultural machinery power to the population employed in agriculture) is selected to measure the productivity of agricultural labor, as analyzed in the subsequent sections.
(2) Important variables
Other important variables selected in this section include the degree of openness, the proportion of state-owned enterprises, the level of public education expenditure and the scale and efficiency of financial development.
(i) Degree of openness
With reference to the practices in the previous literature, the degree of openness is measured by the ratio of the FDI and the total volume of exports–imports to GDP. The degree of openness may affect the transfer of agricultural labor through multiple channels. On the one hand, the higher the degree of openness of a region is, the more conducive to the introduction of advanced technology, the absorption of advanced management experience and the improvement of total factor productivity, thus further affecting the transfer of agricultural labor; on the other hand, China is now still at the peak of the transfer of labor, and the higher degree of openness leads to more active economic activities and a greater deepening of capital, thus promoting the transfer of labor from the agricultural sector to the non-agricultural sector. In addition, the degree of openness may also affect the transfer of agricultural labor from the perspective of factor allocation. Bentolila and Saint-Paul point out that any factor affecting the degree of imperfect market competition may affect factor allocation.21 As for the specific situation of the Chinese market, the FDI and the total volume of imports–exports can be used to measure the degree of competition in the product market. The strengthening of market competition will reduce the cost of the transfer of agricultural labor.
(ii) Proportion of state-owned enterprises
The restriction of the system of household registration makes it difficult for the agricultural labor to become the staff of state-owned enterprises. A large proportion of state-owned enterprises indicates high monopoly power of state-owned enterprises, which tends to reduce the transfer of agricultural labor. In this chapter, the ratio of the total output of stateowned and state-owned holding industrial enterprises above the designated scale to the total output of all industrial enterprises above the scale has been adopted as a measurement of the proportion of state-owned enterprises.
(iii) The level of public education expenditure
The per capita public education expenditure is used to measure the level of public education expenditure. In fact, public education expenditure is a resource allocation. Generally speaking, the urban residents have a higher level of education than rural residents, thus leading to a higher marginal output of investment in rural residents by public education expenditure. In case of an even urban–rural distribution of public education expenditure, the increase in the level of per capita education expenditure is conducive to promoting the transfer of agricultural labor. However, the unevenly distributed public education expenditure between urban and rural areas may hinder the transfer of agricultural labor.
(iv) Scale and efficiency of financial development
This chapter introduces the ratio of the total amount of loans to the GDP as an indicator of the scale of financial development and the ratio of the total amount of loans to total deposits as a proxy variable for financial efficiency. The two indicators may affect the urban–rural income gap and the transfer of agricultural labor to varying degrees in the specific environment of China’s economic development. In terms of the distribution of financial resources, China’s financial system shows a clear tendency toward urbanization, which is inclined to the state sector in credit allocation. Such an unbalanced development may hinder the transfer of agricultural labor.22 According to the analysis of Zhang Qi et al. and Ye Zhiqiang et al., financial development has significantly expanded the urban–rural income gap.23 They also note that the improvement in financial efficiency with the development of financial scale may alleviate the tendency of urbanization and state-owned enterprises, and financial development may bring about the narrowing of the urban–rural income gap. Yao Yaojun’s analysis shows that the efficiency of financial development is negatively correlated with the income gap between the urban and rural areas, despite the positive correlation between the scale of financial development and the urban–rural income gap.24
(3) Other factors
In addition to the foregoing variables, the variable of the rate of urban unemployment and the variable of return on capital are also taken into account because they may affect the rate of the transfer of labor by affecting the demand for labor in the urban sector. In the regression analysis, the former is measured by the changes in the rate of urban unemployment and the latter is measured by the ratio of the total profit of industrial enterprises to the net value of fixed assets of industrial enterprises. In the benchmark regression, the rate of urban unemployment is measured by the rate of registered urban unemployment. Because of statistical problems, the indicator of the rate of registered urban unemployment cannot reflect the unemployment rate of China’s urban sectors well. The more ideal measurement indicator is the rate of surveyed urban unemployment. However, the National Bureau of Statistics does not fully disclose the data regarding the rate of surveyed urban unemployment, and only microdata from the urban household surveys in some provinces are made available. Through the calculation of the microdata from the survey on urban households, it is possible to obtain the estimated data regarding the surveyed urban unemployment rate in the nine provinces from 1992 to 2009. Therefore, the surveyed urban unemployment rate is used to facilitate the regression analysis (the results show no significant difference, as reported in Exhibit B of Appendix B).
To eliminate the possible impacts caused by price changes, the data for the nominal variable have been adjusted based on the CPI of various provinces and regions in 2000, such as per capita public education expenditure. In addition, the impact of inflation is controlled. The statistics of the above variables are reported in Table 2.1.
Table 2.1. Statistics of Regression Variables of the Driving Factors of the Transfer of Agricultural Labor.
Source: China Statistical Yearbook, statistical yearbooks of various provinces and regions, traffic yearbooks of various provinces and regions, official website of the provincial department of transportation, Compilation of Statistical Data of 60 Years in New China, Compilation of Agricultural Statistics Data of 60 Years in New China and CEIC Database; the rate of surveyed urban unemployment has been estimated using the microdata from urban household surveys. The sample interval of each variable is from 1992 to 2010, and the sample interval of the communications infrastructure is from 1998 to 2010. Due to the missing data of some observations in individual provinces, municipalities and autonomous regions, the number of observations of each variable was not completely equal.
2. The results of the empirical analysis
With the use of the panel data of China’s provinces and regions, the transfer of agricultural labor has been used as an explained variable to analyze the determinants of the transfer of agricultural labor. By reference to the practices in the previous literature, both SAR and SEM models are used for analysis in this section to overcome the influence of a potential spatial correlation and to carry out a maximum likelihood