Based On The Financial Market Risk VaR Measure Research Of The Semi-parametric Copula

With the rapid development of financial markets, the diversification of financial products and market instability factors caused the reduction of assets risk controllability, thus the study of value at risk is very necessary.The key problem of asset portfolio risk is a single asset marginal distribution.Although the marginal distribution of financial assets models, using semi-parameter method to model is not the first time.However,the modeling method of marginal distribution usually adopt fixed model to model, and doesn’t have accurate modeling analysis of marginal distribution.Thus this method has many limitations.This paper first introduces the basic theory of the VaR and Copula. Second, the marginal distribution of yield rate of financial assets models by choosing the proper threshold.The value distribution of financial time series is divided into the upper end and lower end and middle part. We put forward a kind of hybrid parameter and non-parametric method about the distribution of financial assets.This method is called semi-parametric modeling method.A new modeling method disseminate only selecting the limitations of fixed type fat-tailed distribution in previous papers.Finally, in the empirical part, this paper selected parameter model which has good fitting degree of marginal distribution (the upper end uses t distribution, middle part uses the Laplace distribution and lower end uses the t distribution), the non-parametric model and a semi-parametric model (that is, the end uses GDP distribution, the middle part uses kernel estimation, the tail uses Laplace distribution) calculates the VaR of real estate, CITS and portfolio.The results showed that the three models are all through the inspection of the failure rate. Semi-parametric model is superior to the parametric model and non-parametric model is superior to the parametric model. Three models of the failure in the number of days, the failure of the portfolio days is less than the failure of a single asset number, portfolio can reduce risk.The contrast of using different methods to calculate VaR shows that regardless of the distribution type can cause significant difference of candidate model, further proves the necessity and effectiveness of the model.The corresponding candidate Copula connection of different models also exists significant difference, but the optimal results are from the candidate Copula.The option of Copula connect type remains to be further improved.During the process of building marginal distribution in the thick tail of the distribution of financial assets,it adopts four kinds of common distribution.People can explore other distribution whether they can also study on the financial data of the tail.