关键词: 四阶段DEA, 能源效率, 非期望产出, 区域差异

Abstract: In the past 10 years, China's economy has developed rapidly. The rapid development of industry and real economy has prompted China's energy consumption to rise rapidly. While promoting energy consumption, extensive development mode makes hidden danger of economic growth and social development. The energy issue is not only a Chinese problem, but also a global issue which is closely watched by all the other countries in the world. Under the new normal, increasing the efficiency of energy consumption is the inevitable trend of economic development. The random disturbance and the external environment are an unavoidable to study the energy efficiency.Traditional DEA model can't eliminate the influence of non-expected outputon efficiency. In this paper, through the construction of four stage Bootstrap-DEA-Malmquist model, the provincial panel data in 1990~2014 is used to adjust energy input index by input slack, eliminate disturbance of external environment and random factors on energy efficiency and total energy productivity, and the Bootstrap random sampling method is used to further reduce the impact caused due to sample differences. From static and dynamic perspective, China's energy efficiency and total factor productivity in 1990~2014 are measured.It is found that the external environment and random disturbance estimation have significant impacts on energy efficiency. By eliminating the influence of external environment, technical efficiency change and the change of total factor productivity decrease in varying degrees. In addition, the energy efficiency drops when the environmental factors are considered. However, the energy efficiency value increases again after using bootstrap method to eliminate the influence of random factors. The main contribution of this paper is to separate environmental factors and stochastic factors from the calculation of energy efficiency and total energy productivity, and obtain more accurate values of energy efficiency and total energy factor productivity. In addition, the input-output data of energy consumption from 1990 to 2013 are systematically combed, especially for pollutant emission indicators, such as sulfur dioxide emissions, nitrogen oxides emissions, smoke and dust emissions. Besides, there were a lot of indicators data missing before 2001, and it needs to be integrated. A large number of calendar statistical yearbooks are referenced to complete the missing data, and the panel data collected in this paper is more comprehensive and complete than the previous references.

Key words: four-stage dea, energy efficiency, non-expected output, regional difference