以邏輯式平滑移轉迴歸模型研究遠期溢酬異常和動能效應

Abstract

未拋補利率平價說(Uncovered Interest Parity， UIP)一直是國際金融和總體經濟學裡，對匯率模型的一個重要假設，此模型隱含兩國的利差應該等於兩國的幣值變動，而且利率高的貨幣會貶值，2008年金融危機後，美國開始貨幣寬鬆政策維持低利率的環境，許多利差交易者改為借入流動性更好的美元，用來購入其他高息貨幣來進行利差交易，Froot and Thaler(1990)整理過去的75篇研究中發現，實證結果大部分都顯示有遠期溢酬的異常現象(Forward premium anomaly)現象，不論是研究實證結果或是市場上的利差交易活動，我們都能發現UIP理論不成立的現象，但如今學者對UIP不成立的現象仍未有一致的說法。 過去探討遠期溢酬異常現象的研究大多利用披索問題(Peso problem)或隨時間變動的風險貼水(Time-varing risk premia)來證明，後來有許多學者開始用非線性模型說明匯率有regime-specific的特性導致UIP不成立。本文選擇採用LSTR非線性模型來區分利差交易報酬的不同情境，探討利差交易的動能效應並討論UIP不成立的情形，研究結果發現在低波動時，外匯動能效應顯著，此時UIP不成立。此外，和一樣研究匯率市場動能效應交易策略的文獻相較下，本研究還擁有能夠找出動能效應顯著的時間點之好處。

Uncovered interest rate (UIP) is one of the key assumptions in exchange model used in both international finance and macroeconomics. It implies that the interest rate differential should be equal to the expected exchange rate. After 2008 financial crisis, Federal Reserve started the quantitative easing policy to lower interest rate. Because of the low interest, many arbitragers borrow the more liquid US dollar instead of Japanese yen to invest in higher-yield target currencies. Besides, Froot and Thaler (1990) find the forward premium anomaly has been consistently found for most published studies. To sum up, we can see the breakdown of UIP not only from empirical studies but also from the carry trade activities but there isn’t currently a consensus on how to explain the failure of UIP. The traditional explanations based on the presence of peso problems or time-varing risk premia. Afterwards, some scholars start to use nonlinear model to find out the relationship between interest rate differential and expected exchange rate movement tends to be regime-specific. This study relies on a linear framework to differentiate carry trade return into two regimes and examines the role of carry trade and momentum strategies in the failure of UIP. The result generally concludes that UIP does not hold in the higher exchange rate volatility regimes where shows the significant momentum effect. Besides, this study can also find out the time of the momentum effect compared to other FX momentum papers.