在台灣公債市場免疫策略的實證研究㆗， 傳統的免疫模型( 例如M-square、M-absolute、M-vector 等模型)，皆可以得到不錯之免疫績效。Nawalkha et al. (2003) 改良了前述傳統免疫模型，利用與現㈮流量到期㈰函數㈲關的債券報酬函數之泰勒展開式，並提出了Generalized M-vector 模型。然而此㆒Generalized M-vector 模型，是否能在台灣公債市場達成較好的免疫績效，仍不得而知。本研究首先採用指數基礎樣條模型為基礎，估計台灣公債市場的利率期限結構，並用以建構債券投㈾組合，以檢驗Generalized M-vector 模型在台灣公債市場免疫策略之投㈾績效。實證結果發現：(㆒)公債投㈾組合㆗不存在債券賣空的限制㆘，Generalized M-vector 模型的免疫績效是㊝於M-vector 模型；(㆓)若公債投㈾組合㆗存在債券賣空的情況㆘，則GeneralizedM-vector 模型的免疫績效，並未㆒致性的㊝於M-vector 模型。

In Taiwan Government bond market, the empirical studies of bond immunization have shown that the traditional duration vector models, including the M-square, M-absolute and M-vector model, can offer good immunization performance. Recently, Nawalkha et al. (2003) corrected the aforementioned traditional duration vector models and derived a Generalized M-vector model,which is based on the Taylor series expansion of the bond return with respect to specific functions of the cash flow maturities. However, there is still no empirical result for applying this Generalized M-vector model in Taiwan Government bond market. This paper uses the exponential B-spline model to estimate the term structure of interest rates in Taiwan Government bond market and investigates the
hedging performance of government bond portfolio constructed by the Generalized M-vector model approach. The empirical results indicate that (1) the Generalized M-vector model will provide better hedging performance than M-vector model when short selling in cash bond market is not allowed, (2) if short selling in cash bond market is allowed, we conclude that the Generalized M-vector model is not necessarily better than M-vector model in hedging the interest rate risk of bond portfolio.