Statistical analysis of economic development issues in China.

摘要

In the first thesis I apply the logic of the compound distribution to examine the effects of certain factors on the number of deaths in Chinese coal mining industry, such as price, type of ownership and labor quality. I apply the logic of compound distribution which is widely used in the actuarial models. Such models combine two distributions: loss frequencies and loss severities. In my case, loss frequencies are frequencies of accidents, which follow the Poisson distribution, and loss severities are the number of deaths in each accident, which follow the Logarithmic series distribution. Then the compound distribution is the negative binomial distribution, which will fit the number of deaths over a fixed interval of time.;The second thesis presents an empirical analysis of labor productivity in China's coal mining. The overall motivation for this work is the explanation of observed changes in labor productivity from 1980 through 2004. I employ the productivity indices defined by Stoker (2005)---fixed effect index, scale effect index and time effect index---which are in line with the panel data concepts. In this period the price effect, a main component of the time effects, accounted for a substantial portion of the change in labor productivity, which was in line with the process of price and wage reform that began since 1979.;In the third thesis, I apply semi-parametric smooth coefficient model (SSCM) to analyze the determinants of economic growth and regional disparities. The analysis applies the combination of SSCM and augmented Solow model, using data from 1991 to 2004 on the provincial level. The SSCM allows geographic factor and time factor to work as determinants of the marginal effects of policy variables, instead of explanatory variables as dummies. I find that the marginal effects of the policy variables vary significantly across regions and over years. It suggests that in comparing the traditional constant coefficient models, the SSCM has more capabilities to capture the reality.