Research On Financial Market Dependence Modeling And Risk Measures

With rapid development of the world economic integration and financial innovation, dependence relationship between different financial markets becomes more and more complex. How to describe dependence structure of financial market accurately is the most important issue of financial risk management. In order to improve the scientificity and accuracy of financial decision-making, in-depth research on risk management and dependence between different financial markets has important theoretical and practical significance.Vine Copula model based on conditional Copula function and Vine graphical modeling tool can be used to connect the edge distribution function of multivariate financial variables into a joint distribution function. It provides a powerful analysis tool for financial market. Vine Copula model can not only measure the degree of dependency, but also determine the dependence structure between financial variables. It is flexible and effective to describe the characterization of asymmetric, nonlinear and tail dependence between financial markets.This paper modeled the financial market dependence and risk measurement based on relevant theory at first, and then systematically summarized and analyzed the shortage of the research literature. Relevant questions were studied as follow:Firstly, this paper introduced the concept of dependence and its measure index, which could be further divided into whole and local dependence measure index. Then, on the basis of Copula theory, several common Copula functions and their advantages were discussed.Secondly, the GARCH-EVT-Vine Copula model was constructed by GARCH model and extreme value theory (EVT), and the main steps and processes were given in this paper. The establishment of model was composed of two parts:In the first part, semi parametric method based on AR-GARCH method and EVT theory was used to establish the marginal distribution model. In the second part, the Vine Copula model was used to model the dependence of the marginal distribution. This method established parametric model for tails of the empirical distributions using extreme value theory, and had obvious advantages on description of return series’ peak and thick tail characteristics, and therefore, it is helpful to establish a more accurate dependence structure model.Based on GARCH-EVT-Vine Copula model, empirical research on the cross-market dependence was carried out. In order to study the non-conditional dependence and conditional dependence among different financial markets, energy, stock index, gold and dollar market were selected as research object. Three different vine structures (R-vine, C-vine and D-vine) were used respectively to build the dependence structure of selected markets. Results of this research showed that the conditional cross-markets dependence were weak. Modeling accuracy of R-Vine copula was better than C-vine and D-vine Copula. However, it was more complex and thus time consuming than the other two Copula models.Risk measure of high dimension financial assets portfolio was studied based on Copula GARCH-EVT-Vine model.16 stock indexes from major global economies were selected as the research object, using three types of vine copula to model high-dimensional assets dependency structure. Based on combining Monte Carlo simulation method and rolling time window,300 trading days of portfolio value at risk (VaR) were forecasted, and the VaR failure tests were conducted. The research results showed that:those three Copula Vine models had all passed failure tests, in which the R-vine Copula model had better fitting effect and VaR prediction accuracy. The results also confirmed that this model was effective to characterize dependence of complex and high dimension between financial markets. At the same time, Monte Carlo simulation method and out of sample estimation based on rolling time window were used to model time-varying dependence simulation, so as to provide an effective method for quantitative risk management.At the end, based on conclusions of this study, the problems to be further studied were discussed.