匯率預測-縱橫平滑轉換自我迴歸模型之應用

Exchange Rate Forecasting-An Application of Panel Smooth Transition Auto-Regression Model

林雅瓶(Ya-Ping Lin)

指導教授 ： 吳博欽

中原大學/商學院/國際經營與貿易研究所/碩士(2013年)

https://doi.org/10.6840/cycu201300294

摘要

在經濟自由化與國際化原則的運作下，國際金融衍然成為國家經濟發展重要的一環，而連繫各國通貨往來之重要媒介即為匯率。匯率的變動是一個極為複雜的問題，在面臨國際金融市場的不確定時，外匯暴險常吞噬跨國企業鉅大盈餘。因此，當企業積極進行全球佈局的同時，正確預測匯率之重要性不言可喻。

文獻上用以預測匯率的因素不少，其中以利率、經濟成長率或通貨膨脹率等因素最常被採用。考量經濟決策者在搜集匯率預測所需資料的成本，以及對這些影響匯率預測的變數之預測可能產生偏誤，故本文採用落後期匯率以作為預測當期匯率的解釋變數。有鑑於利率是影響匯率的重要因素，對發揮經濟槓桿亦有重要的意義，亦即透過利率可以清楚說明貨幣政策的變化，故本研究選取利率為Panel STAR模型中的轉換變數。此外，匯率的走勢可能存在非線性型態，故必須採用非線性的模型預測匯率動態路程。因此，本文以González與Teräsvirta(2005)所提出的縱横平滑轉換自我迴歸模型(Panel Smooth Transition Autoregression，PSTAR)預測匯率的非線性走勢。

實證上，以G20的匯率為對象，在排除資料不齊全之巴西、阿根廷和沙烏地阿拉伯三個國家，以及德國、法國與義大利因共同使用歐元後，將各國貨幣表示為以美元為計價單位，共得十三種匯率。研究期間為2002年1月至2012年8月的月資料，共計1664個觀察值。

實證結果發現，匯率是以非線性的方式移動且存在至少一個門檻值，即匯率變動的迴歸方程式應以縱橫平滑轉換自我迴歸模型來描述更貼切。不同落後期的匯率估計係數皆顯著異於零，表示落後期匯率對當期匯率有顯著的影響，亦即匯率變動具有遞延性與持續性，且受利率的非線性影響：利率愈高，匯率變動率的持續性愈高；利率愈低，匯率變動率的持續性愈低。

關鍵字

持續性 ； 門檻值 ； 匯率預測 ； 縱横平滑轉換自我迴歸模型

並列摘要

With the operating of economic liberalization and internationalization, international finance has played an important role in a country’s economic development. Meanwhile, the linkage between a pair of currencies is exchange rate. However, the change in exchange rate is an extremely complex problem. In face of the uncertainty in international financial markets, foreign exchange exposure often erodes huge earnings of multinational enterprises. Therefore, carry out the global layout, the question regarding how to correctly forecast becomes an important task for the enterprises to actively expand their territory.

In literature, some factors are used to forecast exchange rates, including the interest rates, economic growth rates and inflation rates. Based on the considerations of searching costs of the above variable that influencing exchange rates, and the probable biased forecasting of the variables, this research uses the lagged exchange rates as the regressors to forecast current exchange rates. In addition, exchange rates may display a nonlinear time path, therefore, this research employs the panel smooth transition autoregressive model (PSTAR) proposed by González and Teräsvirta (2005) to trace out the dynamic path. In view of interest rate is an important factor that influences the change in exchange rate and can describe the change of monetary policy, the paper uses interest rate as the transition variable in the PSTAR model.

Empirically, we choose the G20 countries as sample objects. After the exclusion of incomplete information in Brazil, Argentina and Saudi Arabia and the Euro countries: Germany, France and Italy, there are 14 countries chosen. There 14 countries form 13 exchange rates used the USD as denominated. Sample period spans from January 2002 to August 2012; therefore, a total of 1664 observations.

Empirical results show that the exchange rates display a nonlinear dynamic path and there are at least one threshold value. Thus, the PSTAR model is more proper used to describe the change in exchange rate. In addition, lagged exchange rates have significant effects on current exchange rates, implying that the persistence of exchange rate is obvious.

Finally, the changes in exchange rates are influenced by interest rate in different regimes. The higher (lower) the interest rates are, the larger (smaller) the persistence of exchange rates would be.

並列關鍵字

Persistence ； Threshold Value ； Exchange Rate Forecasting ； Panel Smooth Transition Autoregressive Model

參考文獻

吳博欽、申志偉、潘聖潔(2009)，匯率的非線性調整、套利與經濟價值可預測性，人文及社會科學集刊， Vol.21 , No.1 p.101－142. ( TSSCI )

黃永昇(2008)，「現金流量風險值之估計—Linear、STAR與STARX模型之比較分析」，中原大學國際經營與貿易學系研究所碩士論文。

潘聖潔、黃永昇、吳博欽(2011)，現金流量風險值之估計，管理與系統，Vol.18， No.1 p.35－70. ( TSSCI )

Bulter,Alison ,1990.A methodological approach to chaos: are economists missing the point?

Federal Reserve Bank of ST.Louis:Review,Mar./April, 36-48