Improved Research On VaR Measuring Method Of The Financial Market Risk

VaR(Value-at-Risk)is an advanced instrument of metrics,indentification and management for the financial market risk. It is also an inner model of management recommended by Basel Committee's. Many researchers have done vast works for the improvement of precision of VaR in the future market distribution, volatility estimation and potfolio pricing. Financial time series data are not normal distribution.They not only show thick tail but also asymmetry, that is ,they have skewness and kurtosis. For the fat tail student-t distribution and General Error Distribution (GED) have been put forward. But all these are symmetric distribution and do not conside the deflexion.For solving the problem, an asymmetric innovation coeffient is intruduced into ARCH models.In practice the skewness and kurtosis are not independence.The skewed student-t distribution (SKST) which not noly reflects the skewness but also the kurtosis is introduced into APARCH model. The traditional student-t distribution is only one special case of SKST. The empirical research shows that STST is better than student-t distribution and GED both in volatility goodness-of-fit and VaR estimation.Since the ARCH model is proposed by Engle. It has been developed surgent by econometricians. But the coefficients of traditional ARCH models are constant. It neglects the facts that there are occasional, discrete jump from one state into another in financial markets.When such a switch occurs, the distribution of the data seems to change dramatically. The popular ARCH models fail to capture the feature. SWARCH (regime-switching-ARCH), as an alternative model, is used and the unobserved state of system is assumed to be govered by a first-oder Markov process. The empirical study shows that this model can measure VaR exactly and characterize probabilistic inference about whether and when the switching occurred based on the observed behavior of the series.CAPM(Capital Assert Pricing Model)is still the foundation stone of finance.But it has faced great challenge for its assumptions of normal distribution, constantβcoefficient and without considing the profile of sknewness and kurtosis both market index and the portfolio. CAPM based on Markov process is proposed. Theβcoefficient and the condition variance are time-varying. The trational CAPM is only one special case of this model.With the basement, the SSRM model which is used to calculate the VaR of portfolio is proposed. The model allows the market index and portfolio are two indepence Markov processes. The empirical research shows this model has more superiority than the model of combining the traditional CAPM with GARCH model in estimating VaR of portfolio.