以平滑轉換迴歸檢定風險利率平價

The Application of the STR Model: An Empirical Study of the UIP

柴蕙質(Hui-Chih Tsai)

指導教授 ： 楊奕農

中原大學/商學院/國際貿易研究所/碩士(2004年)

https://doi.org/10.6840/cycu200400514

摘要

由於過去已發表文獻之UIP實證結果，大多無法支持UIP中利率差與匯率變動率一對一關係之推論，即以利率差對匯率變動率作迴歸之係數β大多不等於1，甚至小於零 (Froot and Thaler, 1990)，而文獻中提到β估計值小於零之原因，大致上可歸納為隨時間變動之風險貼水因素，以及貨幣管理機構對匯率或利率採取之干預政策，由於政策干預之目的通常在於緩和或穩定匯率或利率之波動，因此很可能使匯率決定模型之參數發生變動，過去已有許多文獻採用參數具可漸近式變動特性之平滑轉換迴歸 (smooth transition regression, STR) 檢定匯率決定理論，例如購買力平價 (purchasing power parity, PPP)、貨幣模型 (monetary model) (Taylor and Peel, 2000; Michael et al., 1997)，並指出匯率決定模型之參數，以較平滑之非線性方式收斂至其採用之匯率決定理論長期均值，而且發現以STR模型估計之模型配適度較原線性模型佳，因此UIP之實證模型是否有相似之現象，是在檢定UIP時值得考量之方向。
鑑於過去UIP之實證研究文獻，尚未有人應用STR以考量UIP實證模型參數發生平滑式變動之可能，因此本文即以STR模型估計及檢定UIP，此外，本文也以一般以OLS及ARCH-M模型估計及檢定UIP，並就其結果加以比較。其中以OLS及ARCH-M模型估計出之係數β皆無法支持UIP之推論，且模型之解釋力很低。而LM type test之檢定結果指出UIP模型之參數具有平滑式轉變之現象，各國資料以ESTR模型估計之結果顯示UIP所估計出之係數β僅在英、法兩國之估計結果中顯著，其中英國資料所估計出之β由0.029 逐漸減少為-0.054，之後又漸增至0.029，而法國之β值則由-0.019慢慢增加為0後再遞減為-0.019，從係數檢定結果來看，不同階段之β值皆仍無法支持UIP，而大部分國家UIP關係式所估計出之參數變動到第二階段之時間恰與廣場協議舉行之日期相近，由於廣場協議屬於國際性之外匯市場聯合干預政策，因此很可能是造成參數發生變動的因素之一，另外是否有其它影響UIP參數變動之因素，或是受限於本文採用之STR模型針對2個以下之結構轉變點進行檢定及估計，因此模型之參數或許還有更多次變動之可能，未來可再進一步驗證，此外，亦可將其它可能使參數發生變動之因素納入估計之考量。

關鍵字

風險利率平價 ； 匯率 ； 平滑轉換迴歸

並列摘要

The uncovered interest parity (UIP) has been rejected in most of empirical studies. According to the literature. Empirical failure of UIP results from two main possible reasons: time-varying risk premium and monetary policy (Froot and Thaler, 1990). Since monetary policymakers might tend to resist rapid changes in exchange rates. This behavior could have smooth effects on relationship between interest rate and exchange rate (McCallum, 1994). Furthermore, it may leads to smooth transition in parameters of the UIP model.
The smooth transition regression (STR) model has been applied broadly in exchange rate determination empirical studies, such as the purchasing power parity (PPP) and monetary models (Taylor and Peel, 2000; Michael et al., 1997). These studies indicate that the STR model provides a more realistic representation while a single structural break model does not. And it may also be viewed as a generalization of special case of some discrete structural break model. Most of these studies documented that the STR model outperforms various linear models. This paper proposes a framework for the empirical analysis of the UIP that allows for smooth transaction process in the UIP relationship by the STR model. The estimation by the STR is also compared with results estimated from the OLS and the ARCH-M model.
The results show that the fitness of STR model is better than the OLS model and the ARCH-M model. Parameter constancy is strongly rejected according to the result of the LM type tests. Transition function changed from unity to zero and reverted to unity around 1985. The timing of shifts in parameters seems to coincide with the official interventions following the Plaza agreement. However the coefficient β estimated by OLS, ARCH-M and STR model cannot support the UIP.

並列關鍵字

exchange rate ； uncovered interest parity ； smooth transition regression

參考文獻

Chow, G. C., (1960) "Testing for Equality between Sets of Coefficients in Two Linear Regressino Relationship over Time." Econometrica, 29, 591-605.

Dijk, D. V., Terasvirta, T., and Franses, P. H. (2000) "Smooth Transition Autoregressive Models – A survey of Recent Developments" Econometric Institute Research Report EI2000-23/A.

Engle, R. F., Lilien, D. M., and Robins, R. P. (1987) "Estimating time varying risk premia in the term structure: The ARCH-M model." Econometrica, 55, 391-407.

Eitrheim, O. and Terasvirta, T., (1996) "Testing the Adequacy of Smooth Transition Autoregressive Models. "Journal of Econometrics, 74, 59-75.

Fama, E. F., (1984) "Forward and Spot Exchange Rates." Journal of Monetary Economics, 14, 319-338.