新台幣兌美元匯率波動性預測及其與遠期匯率之關聯性-預測模型比較及納入成交量之探討

Volatility Forecasting of USD/NTD Exchange Rate and Its Relationship with Forward Exchange Rate: Effects of Forecasting Performance and Trading Volume

本文旨在探討新台幣兌美元匯率波動性之預測與即、遠期匯率間之關聯性，因此本文利用隨機波動模型、GARCH模型、GARCH-M模型、EGARCH模型、TGARCH模型及GJR-GARCH模型，針對新台幣兌美元匯率波動性進行預測，並比較各模型之預測績效，進而找出最適之VEC-TGARCH模型以探討即期與遠期匯率間之關聯性，其實證結果如下：(1)新台幣兌美元之即、遠期匯率皆為非恆定數列且具有相同的I(1)整合級次。(2)利用Johansen共整合檢定法對新台幣兌美元即、遠期匯率進行檢定，結果發現即、遠期外匯市場存在共整合現象。(3)即、遠期外匯市場存在波動叢聚現象與不對稱效果。(4)當納入遠期匯率報酬率與外匯成交量時，會使得波動叢聚效果下降，模型進而配適更佳。(5)就即、遠期外匯市場為彼此互饋的因果關係且遠期外匯市場對新訊息的反應較快速。(6)在單變量模型中，隨機波動模型之預測績效為最佳、TGARCH模型之預測績效次之，接著的排列順序依情境不同而有所改變。然而，再加入遠期匯率與外匯成交量後之多變量模型，各模型之預測能力有提升，就整體績效來看，雙變量模型之預測能力比單變量模型之預測能力為佳。

The purpose of this study is to probe the volatility forecasting of NTD/USD exchange rate and the relationships between the spot and forward exchange markets. In order to forecast the volatility of rate of change for exchange rate, this study applies stochastic volatility model, GARCH model, GARCH-M model, EGARCH model, TGARCH model and GJR-GARCH model to proceed this purpose. Comparing the forecasting performance of each model, we find that the VEC-TGARCH model is better to describe the relationship between the spot and forward exchange markets. The sample period is from January 2,2001 to November 30,2005. Major conclusions of this study are shown as follows. Firstly, the result of the unit root test shows that the NTD/USD spot exchange rate and the NTD/USD forward exchange rate are non-stationary series and have the same integration order I(1).Secondly, by using Johansen co-integration test, the result find that there is a co-integration relationship between the spot and forward exchange markets. Thirdly, there are a volatility clustering phenomenon and an asymmetric effect in spot and forward exchange markets. Fourthly, while taking the return of the forward exchange rate and the trading volume into account, it decreases the volatility clustering effect and fits the model well. It means that the investors should consider the influences of the forward exchange rate and trading volume to make better investment decisions. Fifthly, there are reciprocal cause and effect relationships between spot and forward exchange markets and the reaction of the forward exchange market to the new news is much faster. Sixthly, referring to the single variable model, the performance of the stochastic volatility model is the best, the second comes to TGARCH model. However, when adding the forward exchange rate and trading volume, the forecasting performance become better. As to the overall forecasting performance, the forecasting ability of the bi-variable model is better than that of the single variable model.