Dynamic Measurement Of Industry Portfolio Risk Based On Vine Copula

With the development of economic globalization and financial liberalization,the opening degree of domestic financial market increase in recent years.The domestic market is more vulnerable to the international financial risk.At the same time,with the development of financial innovation,the types of financial products increase,as has the volatility of the market.In the practice of financial investment,investors often choose multiple assets for portfolio investment rather than one type of asset.As for the choice of portfolio assets,investors often prefer to allocate assets in multiple industries at the same time,to disperse industry risks and improve investment performance effectively.Under the background of the increasing volatility of the domestic financial market,how to effectively manage the industry portfolio risk is not only an important task faced by the financial sector,but also a focus of academic attention.The efficient measurement of financial risk is the core of financial risk management.The Value-at-Risk(VaR)method is regarded as a standard method of risk measurement,but it cannot satisfy the subadditivity.At the same time,it does not consider the tail risk in financial assets,thus cannot depict portfolio risk accurately.On that basis,some scholars proposed the Expected Shortfall(ES)model to meet the requirements of coherent risk measurement.Accurately describing the dependence structure of industry assets is the basis for portfolio risk measurement.The traditional linear models cannot accurately depict the nonlinear state of financial markets.The binary Copula models face the dimensional disaster problem and multivariate Copula lacks flexibility,while the vine copula method can solve the problems better.Existing studies often use C-vine and D-vine models to measure industry portfolio risk,but few scholars use the structure of R-vine which is relatively flexible to depict it.To discuss the effectiveness of risk measurement,backtesting method should be used.Existing studies often conduct backtesting under the equal weight of assets,and then set new investment weight for robustness test.However,equal weight is only a special case of portfolio investment,and the robustness test has great subjectivity for the weight setting.Based on the analysis above,this paper selects the financial and real estate,energy,industry,raw materials,major consumption,optional consumption,utilities,and medical and health index in the CSI 300 to establish an industry portfolio.Firstly,we use ARMA(1,1)-GARCH(1,1)-t model to depict the stylized facts of financial series and obtain the standardized residuals.Secondly,combined with the extreme value theory(EVT),the six types of vine copula models(C-vine,D-vine,R-vine,R-vine all Frank,R-vine all Gumbel,R-vine all Clayton)are constructed to describe the dependence relation between industry assets.Finally,the ES is rolling measured using the Monte Carlo method and backtesting is carried out on the equal weights and the mean-CVaR optimization weights respectively,to compare the accuracy difference of different risk models.The specific conclusions are as follows:Firstly,according to the descriptive statistical results,it is found that the industry return series show significant stylized facts,such as leptokurtosis,fat-tail,non-normality,volatility clustering.This paper uses ARMA(1,1)-GARCH(1,1)-t-EVT model to describe these stylized facts and construct marginal distributions.The K-S and BDS tests show that the marginal distributions satisfy i.i.d and uniformly distributed on(0,1).So,it is appropriate to use the model to establish marginal distribution,thus vine copula models can be further constructed.Secondly,the construction results of R-vine model show that the non-conditional dependence between domestic industries is strong and establishing multiple industry portfolios by introducing conditional markets can reduce the risk level.Specifically,when we construct portfolio in five or fewer industries,there is still relatively strong conditional correlation in some portfolios.While portfolios in six or more industries can better disperse the risk degree.Thirdly,based on the backtesting results,vine copula models can effectively predict the expected shortfall of industry portfolio.In the different vine copula models,R-vine can more flexibly describe the dependence structure between assets and achieve a better measurement result.Based on the empirical results,the following three suggestions are proposed:(1)There are many stylized facts in the financial markets.If these facts are not effectively captured,the Copula model cannot effectively depict the actual dependence relation between assets.Besides,it is more important to construct the marginal distribution accurately for the vine copula model.Therefore,investors should fully consider the stylized facts of financial markets.With the financial volatility increasing,investors should pay more attention to the financial extreme risks.The extreme value theory can model extreme distribution of financial asset series,and then accurately depict the extreme dependence relation between financial assets.(2)Although there are some differences between industries,the unconditional dependence relations between markets remains strong,due to the correlation between industries as well as the influence of domestic political and economic environment.When the stock market fluctuates violently,it is more likely that two industries fluctuate in the same direction.So,investors should avoid portfolio investment in only two industries.In addition,although the introduction of conditional market can reduce the industry portfolio risk to some extent,when we construct portfolio in 5 or below 5 markets,the conditional correlation in some portfolios remain strong due to the lack of conditional market and so on.Investors should avoid building portfolios in industries where conditional correlation are strong.Finally,when the amount of conditional markets reaches four,the industry risk can be fully dispersed and the conditional correlation is relatively small.When the market fluctuates violently,investors can consider this kind of portfolio to reduce investment risk.(3)The traditional linear dependence models cannot accurately depict the nonlinear dependence relation between multiple assets,i.e.,the nonlinear binary Copula and multivariate Copula models.But the vine copula method can be more effective to describe the dependence between multiple assets.In different vine copula models,the R-vine model can be constructed based on the actual state of the markets and is more adaptable to the complex dependence between financial assets.