VaR Estimation of Bond Portfolios Using Principal Components Analysis

• Main Authors: Wen-Bin Huang, 黃文彬
• Other Authors: Mei-Ying Liu
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/89527332499181886359

碩士 === 東吳大學 === 企業管理學系 === 92 === The measurement of interest rate risk plays a central role in the risk management of bonds. As a result of many studies, a number of methodologies have been developed that quantify the risk exposure of bond positions. The non-linearity of the price-yield relationship makes the VaR estimation process more complex in the case of bonds. In this paper, we employ the duration method to construct the term structure of interest rate in the Taiwan government bond market and deal with the principal components analysis of interest rate movements. We then look for the principal components that affect the yield movements to serve as a foundation for measuring bond position risk. Furthermore, we use the principal components model proposed by Golub & Tilman (1997), who use the multi-variables statistical method, to estimate the VaR for the bond portfolio. In addition, we also adopt traditional approaches to VaR estimation, including the variance-covariance approach, the historical simulation method and the Monte Carlo simulation method, for comparison purposes, while also evaluating the merits and demerits of the various models. The results demonstrate that the yield curve for the Taiwan government bond market currently has a large hump. The first principal component can explain 98.434% of the yield movements, and the first three principal components can explain almost all of the yield variability in the Taiwan government bond market. It is found that the Golub & Tilman model is certainly incapable of effectively capturing the risk exposure in relation to bond positions. The Monte Carlo simulation method is found to perform more accurately and efficiently in the VaR estimation than the other approaches.