本研究係以台灣及美國10年期政府公債的殖利率資料進行分析，透過二市場自2005年4月至2008年11月間公債殖利率收盤之日資料，並以2007年4月初引爆第一起次貸風暴事件之美國第二大次级抵押資產貸款機構，新世紀金融公司正式提出破產為轉折點，探討在次貸風暴效應發生前、後，該段期間內台灣公債殖利率與美國公債殖利率之短期動態關係與長期均衡關係之變化。在研究方法上採用Kapetanios et al.（2003）之KSS單根檢定法以檢定資料之非線性定態關係，並以Enders and Granger（1998）門檻自我迴歸模型（TAR）及動差門檻自我迴歸模型（M-TAR）進行門檻共整合檢定，再進一步利用Enders and Granger（1998）及Enders and Siklos（2001）門檻誤差修正模型（TECM），來捕捉變數間的長短期非對稱因果關係。 實證結果發現：在非線性KSS與線性NP, PP及KPSS單根檢定法檢定台債殖利率與美債殖利率之時間序列皆為I（1）數列；而在門檻共整合檢定部份，次貸風暴前以AIC法則選出調整機制最佳的模型為M-TAR模型，次貸風暴後之最佳模型則為TAR模型。經由門檻共整合檢定結果可發現，次貸風暴前、後，美債殖利率與台債殖利率皆呈現共整合關係；在門檻誤差修正模型的實證結果顯示，次貸風暴前、後，台債殖利率皆明顯受美債殖利率影響，且次貸風暴後受影響的期數增加；在Granger因果關係檢定中發現，次貸風暴前二地債券殖利率之走勢皆較為平緩時，美債對台債具有單向領先關係；而次貸風暴後二地債券殖利率皆呈現較大幅度波動時，美債與台債之間則互有領先落後關係。 This study examines the relationship of government bond yield between Taiwan and US using daily closing data over the period Apr. 2005 to Nov. 2008.We employed the threshold error-correction model (TECM) to investigate the asymmetric causal relationship between Taiwan and US bond market before and after the subprime mortgage crisis. The KSS nonlinear stationary test and NP, PP, KPSS conventional unit-root tests suggested that both data series are integrated of order one, i.e. I(1) series. The AIC rules suggested that the most preferable model for our adjustment mechanism is the M-TAR model before the subprime crisis and TAR model after the subprime crisis.Empirical results found the existence of threshold co-integration between Taiwan and US bond market before and after the subprime crisis.Our empirical results also showed that existence an unidirectional causality relationship running from US bond market to Taiwan bond market before the subprime crisis, and a bidirectional feedback causality relationship between US and Taiwan bond market after the subprime crisis.
