Study On The VaR Of Portfolio By Copula Be Chosen

Recently, with gradual penetration of the financial globalization, the finance risk management technology has developed violently. The VaR method which has been raised in 1990s, as a result of its good nature, is the most important and widely applied method of risk measurement. The Copula function can put the marginal distributions together to form a joint distribution. The dependence relationship is entirely determined by the Copula function, while statistical descriptions of each variable are entirely determined by the marginal. Copulas have become a popular multivariate modeling tool in many fields where the multivariate dependence is of great interest and the usual multivariate normality is in question. Copula gives a model for the dependence structure that reflects more detailed knowledge of the random variables than Conventional methods. So more and more scholars use Copula function to measure VaR and get good result.So far, the key problem of the Copula function in application is the choice of function form. Although some cultural literature once put forward a homologous suggestion for the choice problem of Copula function, this problem is not solved. According to this, in this paper, on the basis of research method nowadays, a kind of Bayesian Copula selection method is chosen so that the Copula parameters do not have to be estimated and the order of fitting results of different Copulas is got. Combine it with graphic method, research has been done on the Chinese Stock Market so that the best Copula is got. Furthermore, in this paper, as a result of the closing price has the characteristic of"sharp peak, fat tail", we choose the Laplace distribution to be the marginal. Meanwhile use the best two Copulas which are got from the order of the fitting results before to form a new Copula called"Mixed-Copula"as the best Copula to reduce the risk of getting the wrong Copula. Put it together with Laplace distribution to get a multivariate distribution function. At last we get the portfolio at different confidence leves, it turns out that the method owns a good result to mearsure portfolio risk judged by back testing.