| Economic globalization and financial integration strengthen the connection among the global financial market and the relationship among them are becoming increasingly close and complex. To to adapt changes in the financial markets, the international financial institutions have constantly introduced new financial products to strengthen their competitive ability. However, measures of financial supervision and financial system is far less than the speed of product innovation, the financial supervision institution can not implement effective supervision measures on financial institutions,which has contributed to financial institution speculation, exacerbating the volatility in financial markets, even buring the seed for the financial crisis. The United States subprime crisis that broke in2007, has quickly spread to the world, causing a global stock market turbulence and economic panic, then leading to the Greek sovereign debt crisis. These extreme market risk events make financial institutions and financial supervision institutions aware of the necessity and urgency of the financial risk supervivision around the world.Only if the financial insititutions properly take effective measures on the possibility and measurement of the market risks.we could implement effective risk management of financial market. Therefore, risk measurement has become an important research topic for contemporary managers and scholars.For risk measurement and investment portfolio strategy selection, to make sure of asset distribution is one of the key factors.The traditional assumption in general use normal distribution to describe the distribution of assets and asset portfolio. However,when coming to different types of assets, or even similar assets owning different marginal distribution, the joint distribution of a number of assets based on the establishment of the traditional multivariate normal distribution is inconsistent with reality.This paper selects the Copula function to connect the marginal distribution,which makes up for the deficiencies of the traditional risk measure methods, Copula function can connect a variety of marginal distributions and flexibly construct the multivariate distribution.The paper makes the comparison of Gumbel Copula,Frank Copula, Clayton Copula and t-Copula and choose t-Copula as connection function of Shanghai and Hong Kong, Shanghai and Shenzhen. The article points out that the correlation coefficient between Shanghai and Hong Kong is smaller and weaker,while the correlation coefficient between Shanghai and Hong Kong is larger and stronger. When one market fluctuates, another marke will quickly show the relevant fluctuation. The investment portfolio of Shanghai and Hong Kong can more effectively reduce the risk of investment.Although Coupla is widely used in risk modeling,most of the application of Copula for portfolio modeling did not consider non-symmetry of marginal distribution. This paper starts from the characteristics of Chinese stock market index returns analysis and chooses Shanghai Composite Index,Shenzhen Component Index and the Hang Sheng stock index as the research object.In order to better describe skewness and volatility clustering characteristics of return tail of the stock market index,the paper selects EGARCH-t model to describe marginal distribution characteristic, then applys extreme value theory to the standard residual distribution of tail,portrays the Chinese stock market extreme risk and obtains the semi-parametric marginal distribution EGARCH-EVT model. Thirdly, use t-Copula to connect the Shanghai and Shenzhen, Shanghai and Hong Kong respectively, obtaining the EGARCH-EVT-tCopula model, and establishing the multivariate joint distribution of returns on portfolio risk Shanghai and Shenzhen, Shanghai and Hong Kong respectively, measuring two portfolio value. Discusses the risk value under different confidence level. Compare the risk value based on the historical simulation method and based on the model of EGARCH-EVT-tCopula value which use Monte Carlo to simulate portfolio yield future scenario. Then use the backtesting to examine the accuracy of risk value. Finally draw the conclusion. Not only can EGARCH-EVT-tCopula model accurately describe the peakness,thick tail characteristic,heteroscedasticity and asymmetry of the marginal distributions, but also can more accurately measure portfolio risk. |
