Empirical Study On High-dimensional Portfolio With Vine Copula

With the development of science and technology,financial theory has been developing towards the direction of complex and meticulous depiction in recent ten years.In the process of depicting financial thory meticulously,the measurement of the correlation between different assets is very essential.In the aspects of asset pricing,combination allocation and risk management,it is very helpful for investors and managers to accurately measure the related linkage between assets in the financial market.Nowadays,in an economic structure system,different industries form an interrelated complex system.Facing more and more complex financial environment,more stringent requirements for risk control in various industries are put forward.In recent years,the related research on individual assets has been relatively mature,but the theoretical research and empirical analysis of the measurement of multiple asset portfolios is still in the development period.Therefore,this paper selects ten categories of China's first class industry index as the main research object,covering the related assets data of ten industries,such as energy and materials,and mainly studies the application of rattan Copula method in the high dimensional portfolio.The main research contents are summarized as follows:1.Through cointegration test and Granger causality test analysis,it is proved that the short memory between the industry index in China is not the same in the short term,but there is a long-term equilibrium relationship between all the asset returns.2.This paper establishes different edge distribution models.In detail,the nonparametric kernel density estimation method,parameter ARMA GARCH partial t method and the method of combining kernel density estimation and extreme value model are used in the marginal distribution research,and the empirical cumulative distribution function and QQ diagram are used to verify the fitting effect,and the JB test and KS test are also used to meet the Copula modeling requirements.3.On the joint distribution of multiple assets,based on the traditional Copula method,the article uses different vine structure Copula to achieve high-dimensional asset construction.Besides considering two special rattan structures,C-Vine and D-Vine,a more general and flexible R-Vine structure is also considered.Under the different edge distribution model,the Copula model of different vine structure is constructed,and the information criterion AIC,BIC and maximum likelihood value are compared vertically and vertically,and the optimal model Kernel-R-Vine-Copula is finally obtained.Using the Monte Carlo simulation method with better operational performance,we measure the risk of the portfolio of ten industry stocks,calculate the VaR,and verify the validity of the model by using the back test.The research methods in this paper are richer in edge distribution or in rattan structure selection.The final Copula model under R vine structure can describe the whole correlation and the tail correlation among the high-dimensional industry indexes flexibly and accurately.