Evaluation and Empirical Study for Value at Risk Based on the DVEC-GARCH Model

• Main Authors: Jhan-Jhih Cao, 曹展誌
• Other Authors: Cheng-Der Fuh
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/78a252

碩士 === 國立中央大學 === 統計研究所 === 97 === In recent years, risk management has become an important topic, and value at risk (VaR) is a powerful tool for assessing market risk. This research adopts four methods, including the delta-noraml approximation, the delta-gamma approximation, the Monte Carlo simulation method and the historical simulation method to obtain the measures of accuracy and conservatism. Together with parametric and nonparametric bootstrap resampling methods, we further make valid improvements in estimation of VaR. An empirical study applying historical stock closing prices of twelve Taiwan companies over the past two years is provided to compare the performance of each approach. In the process of estimating VaR, two methods, equally weighted moving average and exponentially weighted moving-average (EWMA), are applied to estimate the volatility, which is necessary. According to simulation studies with data generated based on the DVEC-GARCH model, there does not appear to be signi?cant difference in outcomes between both. However, empirical studies show that estimating volatility via the EWMA method achieves higher accuracy and applicability. In general, it is improved to employ the bootstrap resmalping method, and in view of the mean relative bias (MRB), we have a more conservative conclusion with a nonparametric bootstrap method. As a result, using the delta-normal approximation method to estimate VaR along with nonparametric bootstrapping is recommendable in this research.