CAJ | 학술논문

把GED分布引入SV模型,构建SV-GED模型,然后与极值理论相结合拟合标准残差的尾部分布,进而建立一种新的金融风险度量模型——基于EVT-SV-GED的动态VaR模型,并通过沪深300指数和恒生指数的每日收盘价进行实证分析,研究显示：EVT-SV-GED模型能有效刻画中国股票市场的波动性特征,并且能够既有效又合理的度量金融市场风险.
파GED분포인입SV모형,구건SV-GED모형,연후여겁치이론상결합의합표준잔차적미부분포,진이건립일충신적금융풍험도량모형——기우EVT-SV-GED적동태VaR모형,병통과호심300지수화항생지수적매일수반개진행실증분석,연구현시：EVT-SV-GED모형능유효각화중국고표시장적파동성특정,병차능구기유효우합리적도량금융시장풍험.
With the development of financial liberalization,financial globalization,and asset securitization,the economic tie among various countries continues to deepen and new financial instruments continue to emerge.Financial market risk management is becoming more important because each market is exposed to more pressure and challenges.VaR （Value at Risk） is the maximum loss of a financial asset or an investment portfolio that may arise under normal market fluctuations.The VaR theory asserts that VaR method can help measure the risk value of financial assets.Research on how to improve VaR model‘s forecasting accuracy is mainly reflected in how to accurately characterize the distribution of financial assets＇＂fat tail＂ feature.However,foreign empirical studies have pointed out that the measure of VaR appears to be fragile when the GARCH model is used for financial time series,such as ＂Peak fat tail＂ and ＂Leverage effect＂.The SV model is another heteroskedasticity model relative to GARCH because it draws the random process into the variance expression.Theoretical studies have shown that SV models have greater advantage than the GARCH model under financial time series,and the SV class model is more consistent in depicting the volatility and financial market characteristics.The SV type model that is used to depict the financial return on assets is more consistent with the actual situation; however,its description on extreme financial events （mainly shows tail data exception） appears to be powerless.Therefore,stress testing should be conducted as for a supplement of risk measurement.The stress test is based on historical or potential market volatility data to assess the impact of market price change on asset value change.Extreme Value Theory （EVT） is often used to do stress tests and the method can have better measurement in the risk of loss under extreme cases.In China,the risk measures of the extreme value theory for financial assets are mostly concentrated in the normal distribution of capital gains and the GARCH models.The SV model that has better ability in depicting financial times series is more limited to simple normality VaR estimation method,and lack of consideration of the tail characteristics in the return on assets.Based on the stochastic volatility model and extreme value theory,this study re-constructs a new risk measurement model by introducing the fat tail distribution.Particularly,grip GED distribution draws into the SV model,and builds the SV-GED model by combining the extreme value theory with the standard fitting residual tail distribution.This method helps establish a new financial risk measurement model.By analyzing daily closing price of the CSI 300 Index and the Heng Sheng index,the result shows that EVT-SV-GED model can effectively depict the characteristics of the Chinese stock market volatility and can be both effective and reasonable measuring financial market risks.Overall,the model can effectively depict the volatility characteristics of stock market in China,and has a strong predictive ability for the extreme risk in the market.In addition,the posteriori test also shows that the calculation of VaR forecasts is both effective and reasonable.The new method is able to accurately and conveniently reflect the financial market risks,and good in dealing with fat tail phenomenon which is conducive to deep-seated and comprehensive financial risk management.This exploratory study for extreme VaR method still needs further improvement by combining financial market characteristics with EVT theory in the calculation of VaR.