Research On Financial Complex Network Based On Cointegration

With the deepening of China’s financial reform, the improving of national’s level of knowledge of financial and business capital operation ability, the Chinese stock market plays an increasingly important role of providing access to raise funds for enterprises, providing investment channels for investors, and its relation with the real economy are increasingly closely. In this context, to study the association between stock and stock, the influence of a stock or group of stocks to others, will undoubtedly have forward-looking implications to risk management and securities investment. At present, more and more scholars are getting into the study of complex networks in the financial sector, especially the study of stock market applications. However, most scholars focus on the study of the structural characteristics of the complex network, but ignore the in-depth study of relating it with the actual situation. In this context, the research of applying complex networks in financial area is written in this paper.Instead of the correlation coefficients used by most scholars, cointegration coefficients are used in this paper to construct directed and weighted cointegration networks in China’s stock market. In this paper, the network is obtained starting from the matrix of cointegration coefficients calculated by the first step of Engle–Granger cointegration test between all pairs of 167 stocks. Then the diagonal elements and the cointegration coefficients whose corresponding p-values which calculated by the second step of Granger cointegration test are bigger than 0.05 were removed. The cointegration coefficient matrix is the adjacency matrix of a complex network of financial.In this paper, the basic tools of complex networks to filter information, such as threshold cointegration network and cointegration planar graph which is improved based on planar maximum flat filter. Threshold cointegration network node degree distribution is analyzed to verify their distribution are power-law distributed. Then cointegration planar graph network is analyzed with the Page Rank algorithm while stocks’ influence rankings were established as well as the stock industry’s influence rankings. The conclusion is as follows:Among 167 stocks of CSI 300 component stocks, the power industry’s stocks have greater influence. The most influential stock, not necessarily be the first position in the industry, but it must have a larger total assets, a higher ability of earning profit. Electricity industry’s influence on other stocks, is mainly from the closely relationship of power industry and national economic development.