摘要
過去探討利率商品市場關聯性的文獻中，學者多以債券市場為研究的重心，由於對利率交換之成交利率資料取得困難，國內幾乎對台幣利率交換之利率走勢及關聯性研究付之闕如；本文即以選取台灣2年、5年及10年公債對應同年期台幣利率交換之利率來探討兩市場間的互動關係，以2003年1月10日至2006年5月12日的日時間數列為樣本分析期間，實證結果顯示：台灣公債2年、5年及10年期利率與利率交換2年、5年及10年期利率均為I(1)的時間數列。台灣公債5年及10年期利率分別與利率交換5年及10年期利率具有長期共整合趨勢，且存在長期穩定的均衡關係。以向量誤差修正模型分析得知，本研究的實證支持長期間兩個變數間存在密切的均衡關係。另外利用Granger因果檢定，檢定各變數兩兩間的因果關係，發現在5%的顯著水準下前一期的利率交換2年及5年期利率會分別影響下一期的台灣公債2年及5年期利率，另外前一期的台灣公債10年期利率會影響下一期的利率交換10年期利率；此檢定結果顯示整條殖利率曲線，兩個變數間都有單一方向可影響對方。此外就衝擊反應而言公債殖利率與交換利率對本身的衝擊較其他變數影響來的大，特別是在第一期，爾後逐漸縮小，但基本上各變數之間皆為正向反映。
並列摘要
Among articles published in the past on the relationship between different interest rate products, scholars focused on bonds only, hardly on IRS. This is because the Taiwan Dollar IRS is traded on OTC market. As such，past data on IRS are difficult to obtain. This thesis aims to study the relationship between the government bond and IRS yirlds. I pick Taiwan 2, 5, and 10 year government bond and IRS daily yields from January 10, 2003 to May 12, 2006 as study samples. The empirical result shows that the yields between Taiwan 2, 5, and 10 year government bond and IRS are both I(1) time series. The yield relationship between Taiwan 5 and 10 year government bond and IRS demonstrate long term cointegration. and exhibit long term stable equilibrium . From Vector Error Correction model analysis, this study also show that close equilibrium relationshipexists between these two variables. By way of Granger Cause Test, I test the causality relationship between these variables. I find that at 95% confidence level,t(0) IRS2Year and IRS5Year will affect t(1) BOND2Year and BOND5Year. In addition, t(0) BOND10Year will affect IRS10Year. At 90% confidence level, t(0) BOND2Year will affect t(1) IRS2Year. This test also shows that , through the entire yield curve, one variable shows the same directional influence on the other variable. Moreover, the yields between two- year government bond and IRS exhibit a bi-direction at 90% confidence level. Besides, in Impulse-response terms, the government bond yields and IRS are most strongly affected by themselves (i.e. Bond t(1) is most dependent on Bond t(0)；Similarly IRS t(1) is most dependent on IRS T(0)). This impact dwindles with time. Generally speaking ,these variable exhibit positive correlation amongst themselves。