An Extreme Value Model Based Value at Risk Estimation for Equity Portfolio in Taiwan

• Main Authors: Yen-hsin Chen, 陳炎信
• Other Authors: Yang-cheng Lu
• Format: Others
• Language: zh-TW
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/29371310424697468620

碩士 === 銘傳大學 === 金融研究所 === 87 === In using VaR systems to estimate risk degree, it should be considered about the variation of risk assets occurred by abnormal extreme events. And the RiskMetrics model proposed by JP Morgan and general econometrics model can''t capture the behaviors of this kind of fat-tailed distribution effectively, so the VaR estimations should be adjusted upward. In this thesis, we compare RiskMetrics model、historical simulation model and extreme value model proposed by Danielsson and de Vries in estimating VaR for industry portfolios of public-offering stocks in Taiwan. The results are as following, 1. In this study, we found the price limit of Taiwan Stock Exchange affects the stock price. And this kind of situation make the VaR estimation of extreme events which are rare occurrence meaningless. 2. Form empirical analysis, the estimated ability of extreme value model is superior than RiskMetrics models and historical simulation model in simulation data or empirical data. 3. According Bank for International Settlement''s requests of adequate capital, financial institutions use extreme value model to be VaR system that could reduce the requirement of adequate capital and make it more stable.