An Analysis Of The Relationship Between Regional Energy Consumption Intensity And Economic Growth Based On Cross-Section Dependence

As the panel data is influenced by various factors, such as with the increasing strengthen of the trend of globalization, there is a remarkable relationship between different countries, which make cross-section dependence in the real economic study. On the other hand, cross-section dependence make the analysis of the panel data, such as the analysis of unit root tests and cointegration, becoming more complex. However, if you ignore the exist of cross-sections dependence, which will not only make models less accuracy, and even get wrong conclusions. Therefore, applying cross-section dependence tests and unit root tests or cointegration tests to the panel data has great theoretical and practical significance.The first generation panel unit root and cointegration tests are assume that the cross-section members are independent,such as the unit root and cointegration tests proposed by Levin, Lin and Chu(1997), Im Pesaran and Shin (1997). However, O’Connell(1998) stated that in unit root tests, it is inappropriate to ignore cross-section dependence, which can have distortionary impacts on the panel inference by studying the theory of PPP. So the academia proposed and developed the second generation panel unit root and cointegration tests. But the domestic study of cross-section dependence tests is still immature and there is a great potential to research it.In recent years, with the rapid development of technology of compute simulation, as well as Bayesian methods can make full use of prior information and sample information which enhance the credibility of models. Bayesian methods have been applied widely in the world. This paper, based on foreign literature about the Bayesian approaches, will introduce the Bayesian panel data inference method to test exist of cross-section dependence.Before the introduction of Bayesian panel data inference method, this paper will propose the panel clustering method based on the K-means algorithm, which can identify and classified the panel data.and then apply this method to classify the energy consumption patterns of our country. With the establishment of low-carbon economic policy and the transfomation of economic growth, the study of the relationship of energy consumption and economic growth has become the central issue. On one hand, classifying the energy consumption of our country has remarkable practical significance on setting regional policies according to local conditions. On the other hand, testing the dependence of regional energy consumption and analyzing the long-run equilibrium relationship between the intensity of energy consumption and economic growth have important theoretical significance.The main jobs and contributions can be summarized as follows:First, using a large sample of panel data including energy consumption and real GDP per capital and energy-intensity of29provinces of China, considering the differences among different parts of China, the paper established indicators of the energy consumption, combined with regional industrial structure of our country. Secondly, proposing the panel clustering method based on the K-means algorithm and applying it to divide our country into four different energy consumption modes. Thirdly, Bayesian panel data Inference method was introduced to test the spatial correlation of the four energy consumption regions and the empirical result suggested that there is remarkable spatial correlation of regional energy consumption. Finally, adopting PANIC method proposed by Bai and Ng(2004) to analyze the long-run equilibrium relationship between the intensity of the energy consumption and economic growth and the empirical result suggested that energy-intensity and real GDP per capital still has a long-run equilibrium relationship in our country even under the case of cross-section dependence.