Accuracy Improvments of Value-at-Risk Estimation by Using Different Probability Distributional Assumptions - Some Empirical Evidences from Weighted Stock Indexes

• Main Authors: Da-Wei Chien, 簡大為
• Other Authors: Wei-Chung Miao
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/6ew6gf

碩士 === 國立臺灣科技大學 === 財務金融研究所 === 106 === This study proposes to use three common volatility estimation methods ( SMA, EWMA and GARCH ) ,in combination with different assumptions of probability distributions for financial asset return to estimate VaR (Value-at-Risk). We choose eight stock market index data including Taiex, Nikkei225, S&P500, Dow Jones index, DAX, CAC, FTSE and Heng Seng as research targets. The violation rate of VaR is compared in each stock market index from 2000~2017, and what kind of combination can increase the accuracy of violation rate in different stock market periods is discussed in this study. Under high confidence level, using EWMA conbined with Laplace distribution and EWMA conbined with normal-Laplace distribution can outperform the accuracy of violation rate in most of stock market periods. We demonstrate that using leptokurtic probability distribution assumption for financial asset return can alleviate tail risk espeacially when α is high confidence level.