The Research Of The International Stock Markets' Complex Networks Based On Tail Risk

Complex network theory mainly emulates complexity systems through the form of networks,the principle is to abstract the objects in the network into nodes,a certain relationship between the objects is used to be the edge of the network.From this,the statistical characteristics of the whole or part of the network are analyzed to explore the nature of the complex system.The international stock market is a complex system.Each country's stock index acts as a node in the network.The risk impact between them is used as the edge to form a complex network.The characteristics of the international stock network are discovered through research on the network topology structure.The law of evolution,and then analyze the status and role of countries in the stock market.The focus of financial risk control is tail risk.The tail correlation can study the extent to which stock markets in other countries are affected by extreme situations such as a sharp rise or a sharp fall.In general,the sequence of returns has the nature of "spikes and thick tails".The traditional correlation coefficient that follows the normal distribution as an indicator of risk measurement has not been able to meet this kind of characteristic,and there is a big deviation.This article takes the international stock market as the research object and incorporates the tail risk into the international stock market associated network.Firstly,based on the deficiencies of the correlation coefficient in the process of the original complex network construction,this paper considers introducing the Co VaR model into the complex network and including the tail risk.Secondly,unlike the traditional CoVaR method,the CoVaR model coefficient is used as an indicator to measure the marginal spillover effect of risk,and the original binary CoVaR model is extended to a multivariate CoVaR model.For the estimation of the CoVaR model,quantile regression is used because different levels of quantile can be used to obtain risk associations under different conditions,so the market state can be divided rationally;at the same time,considering the threshold method to construct the network with subjective defects,adaptive LASSO regularization for variable selection is added to the multivariate CoVaR model,which can avoid the subjective problem of complex network threshold selection.Finally,in the empirical analysis part,we construct three different directional weighted networks with quantile levels,analyzing network statistics from a microscopic and macroscopic perspective,respectively;constructing the basic static network and the dynamic network to study the structure and nature of evolution over the course of time,and focusing on the changes in the international influence of Chinese stock market.