Font Size: a A A

Empirical Analysis Of The Stock Market Risk Based On The Mixed Strong Copula Theory

Posted on:2014-07-02Degree:MasterType:Thesis
Country:ChinaCandidate:H H LuFull Text:PDF
GTID:2269330401459047Subject:Probability theory and mathematical statistics
Abstract/Summary:PDF Full Text Request
With the accelerated pace of financial globalization and financial innovation, the riskanalysis of financial markets has developed rapidly, and the copula function becomes a newfinancial analysis tool. It can capture the non-linear、asymmetric、tail dependence among thevariables, providing investors with competitive advantages and substantial income, so it hasbeen widely used at home and abroad. Simple copula function is only suitable for the specialcase of the financial markets, which has a difference with the complex financial markets. Themixed copula function can not only describes the upper tail dependence, lower taildependence and symmetric dependence among the stock markets, but also describes theasymmetric model that the upper tail dependence is coexist with the lower. The model basedon the mixed Copula is similar to the real financial markets. The linear copula mixed model iscommon, this paper we’ll propose a new copula mixed model-geometric weighted averagemixed model.The focus of this article is to verify the effectiveness of the mixed model. We selectfinance index and real estate index as our empirical analysis’s data from Jan4,2007to Dec31,2012and the total number of observation is1460.And we’ll study from the followingaspects:Firstly, the determination of marginal distribution. Marginal distribution is not the focusof study in this article, so we commonly selected GARCH (1,1)-t model. Considering that theheteroskedasticity model need verify the ARCH effect at first, we select the graphics testmethod-the correlation diagram of the residual square.Secondly, the choose of the copula function and the determination of the parameters. Themajor difficulty in the article is how to choose the suitable copula function for a geometricweighted average mixed model. We used the least squares estimation method and gibbssampling, which solve the value of parameters effectively, considering there is moreparameters.Thirdly, the validation of the model. In this paper it is not easily to get VaR by joint distribution function, so we can use Monte Carlo simulation to fit VaR of several models.Finally, we carried out the posteriori test. By analyzing the results of posteriori tests, wefound that the mixed copula model has the less number of the failure days, compared with thesingle copula model. As a new mixed model, it improves the accuracy of risk prediction. Thatis to say, the new model can truly describe the risk of the portfolio, which has a certaineconomic value.
Keywords/Search Tags:geometric weighted average mixed copula, least squares estimation, Gibbs Sampling, GARCH
PDF Full Text Request
Related items