Return Models with Regime Switching and Applications on VaR computation

• Main Authors: Chien Shien Lee, 李建賢
• Other Authors: Shyang-Hua Wu
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/86370155152197559946

碩士 === 輔仁大學 === 應用統計學研究所 === 96 === The thesis mainly searches for a suitable model to measure VaR, for investors and financial institution, to estimate the risk of portfolio. With this new model, investors can maximize the profits under control of portfolio risk. The sample data we used are collected from Taiwan TEJ database. It includes financial index and Electronic index in Taiwan stock market from 2006/1/6 to 2008/5/28. Among AR-GARCH, MAR-GARCH, and GRS-GARCH models, this study evaluates which one can have better prediction ability to estimate the VaR of portfolio. Under 95% confidence level, RS-GARCH (1,1)-T can provide reasonable VaR for financial index return; Markku and Pentti MAR(2;0,0)-GARCH(1,1;1,1) and RS-GARCH (1,1)-T can provide reasonable VaR for Electronic index return in our empirical analysis. In the empirical evidence, comparing to AR-GARCH and MAR-GARCH models, GRS-GARCH model indeed has better prediction ability in estimating the VaR of portfolio in stock returns. Besides, GRS-GARCH model can moderately diminish VaR to reduce financial institutions’ capital costs, which are not too small to hide the high risk behind, and it’s believed to be a great instrument in evaluating portfolio risk.