The Geographic Distribution of Human Capital in China

In this paper, we basically made efforts to answer 3 questions: （1）What is the geographic distribution of human capital in China? (2) How the regional differences of human capital change over time? (3) What are the main determinants of geographic distribution of human capital? To answer the first question, we choose the education attainment approach and build an indicator system which includes both a human capital stock indicator and human capital structure indicators. Average years of schooling is the human capital stock indicator, and percentage of people with higher education, high school education, middle school education, primary school education (according to the final education level) and illiteracy rate are the 5 human capital structure indicators. We collect data of 31 regions from 1997 to 2008, and run the cluster analysis. We find that Beijing, Tianjin, Shanghai are most developed in human capital, and Tibet is ranked the last. Furthermore, northern regions and some central regions are better developed in human capital; western regions are generally poor in human capital. However, not all economically developed regions have an abundant educated population, like Zhenjiang and Fujian; and not all poor regions lack human capital, like Shanxi. To answer the second question, we compute the correlated variation of each indicator in the observation period. Through comparison among indicators, we find that percentage of people with higher education varies most in 31 regions, then the illiteracy rate, then the percentage of people with high school education. The correlated variations of the rest indicators are relatively small. Meanwhile, the regional difference of illiteracy rate tends to increase during the 12 years, and so does the regional difference of percentage of people with primary school education. And these two indicators represent the situation of basic education. Regional differences of the rest decline. Since it is unreasonable to have a high regional difference in illiteracy rate, we compute the results by excluding Tibet again, and find that this change does not affect other indicators but do decrease the correlated variation of illiteracy rate. But the increasing trend does not change. To answer the third question, we do two things. First, we run two panel data regressions. The dependent variable of first one is average years of schooling, and of the second one is percentage of the population with higher education. The model we use is the fixed effect model including both entity and time fixed effects. Nine regressors are added in and two of them are control variables. We find that income, the number of health personnel per 10,000 inhabitants and the number of street lights per cities are positively related with the two dependent variables. The last two are variables reflecting nonpecuniary benefits offered by the region. Cost has a negative impact on average years of schooling, but does not affect percentage of people with higher education. Government’s education expenditure per person does increase average years of schooling in a statistical perspective, but government’s expenditure on higher education per student does not affect the percentage of people with higher education. The differences in the results of the two regressions may suggest different behavior patterns among people with different education backgrounds. The unemployment rate is not statistically significant in either of the two regressions. Two possible reasons could explain this: first the data we used is not the unemployment rate of educated people, but the registered unemployment rate of urban residents, because the first one is not available. Secondly, lack of full information may also lead to this result. The other thing we do to answer question (3) is constructing an internal migration model to explain the formation of human capital in labor force. In this model, people possess different levels of latent ability, and have to make two decisions: migration for education and migration for job. Furthermore, people assign different weights to nonpecuniary benefits. As a result, education cost, living cost is negatively related with human capital stock in that region, on the contrary, starting wage offered by employers and nonpecuniary benefits are positively related with human capital stock in that region. A brief discussion of preferences is also made