台幣對美元匯率預測之探索性分析

A Heuristic Investigation of NTD/USD Exchange Rate Forecasting

林源馨(Yuan-Hsin Lin)

指導教授 ： 楊奕農

中原大學/商學院/國際貿易研究所/碩士(2008年)

https://doi.org/10.6840/CYCU.2008.00405

摘要

傳統上以經濟理論進行匯率預測是以長期預測為主，因其使用的總體經濟資料是以季、年為變數，且其樣本外預測比較績效甚至不如隨機漫步模型 (Meese and Rogoff, 1983)。有鑑於匯率操作實務上短期預測實際之需求，本研究改用具有日資料的黃金期貨價格、紐約輕原油期貨價格、台股期貨等之近月連續報價及即期匯率等作為自變數，以2002年1月至2006年12月的日資料為樣本，利用多變數動態迴歸模型為估計方法，進行自變數落後1至30期之大規模排列組合的估計，再以係數顯著性、配適度方法進行篩選，找出候選模型，最後與隨機漫步模型進行樣本外預測力績效比較。實證結果顯示，本研究找出之最佳模型在以誤差均方根 (RMSE)、平均誤差百分比值 (MAPE)、平均誤差絕對值 (MAE) 等樣本外預測力績效比較中皆勝過了隨機漫步模型。研究結果顯示，所選出的較佳模型中，匯率除了和匯率落後期有關之外，並與黃金期貨價格落後期呈現負相關的現象且和原油期貨價格、台股期貨報價無關。

關鍵字

隨機漫步模型 ； 匯率預測 ； 多變數動態迴歸

並列摘要

In traditional economic theory, exchange-rate forecasting models focus on long-term prediction, and use macroeconomic data as the source of information. Additionally, compared to the random-walk model, these models generally do not perform well in terms of its performance on out-sample prediction (Meese and Rogoff, 1983). In light of the above issue and the need of short-term prediction in practice, we alternatively develop a dynamic multivariate regression model of the spot NTD-USD exchange rates, based on the large combinations of one to thirty lag periods. Our model also includes three exogenous variables based on the following consecutive daily price data in recent months: (1) gold futures price, (2) light sweet crude oil price (NYMEX), (3) Taiwan index futures prices. The observation period spans from January 2002 to December 2006. We examine numerous models that use different combinations of exogenous variables, and select the best model according to the significance of its parameters and goodness-of-fit. Finally we compare the selected model with the random-walk model according to their performance on out-sample prediction. The empirical result shows that our model outperforms the random-walk model in terms of Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE). The selected best model also indicates a tractable relationship between the NT-USD exchange rate and the variables except light sweet crude oil price and Taiwan index futures prices. This model also shows that the NT-USD exchange rate correlates to its lag exchange rates, and has a negative correlation with the lagged gold price.

並列關鍵字

dynamic multivariate regression ； RW model ； exchange-rate forecasting

參考文獻

鄭美幸與詹志明 (2002)，「灰色理論與時間序列模型在匯率預測績效上之比較」，台灣金融財務季刊，第3輯第2期，頁95-103。

Boothe, P. and J. G.. Mackinnon (1986),“A Specification test for models Estimated By GLS,” The Reviews of Economics and Statistics, 68, 711-714.

Enders, W. (2004), Applied Econometric Time Series, New York : John Wiley and Sons, Inc.

Goodhart, C.A.E., S. G.. Hall, S.G.B. Henry, and B. Pesaran (1993),“News Effects in a High-Frequency Model of The Sterling-Dollar Exchange Rate,” Journal of Applied Econometrics, 8, 1-13.

Kilian, L. and M. P. Taylor (2002),“Why is it so difficult to beat the random walk forecast of exchange rates?,” Journal of International Economics, 60, 85-107.