本研究之精神在於提供另一種判斷Kalman filter是否為匯率預測之優良模型的方法。因為過去的文獻在探討預測模型是否配適得好,都是經有比較各模型的RMSE、ME,而本文利用Kalman filter做匯率的預測在結合匯率與股票指數之因果關係,其預測績效所強調的是「方向預測的準確度」,而非「預測誤差的大小」。因為當企業在利用衍生商品避險時,準確的預測匯率未來的走向,才能達到較佳的避險效果。而本研究分別探討的範圍包含法國、德國、台灣、日本以及新加坡等五國家的貨幣兌美元的匯率,並且依照資本市場開放程度的不同來劃分研究期間以進行比較。最後,本研究也將Kalman filter與Random walk以及GARCH兩模型作兩種預測績效上的比較。
This research offers another way to decide whether the Kalman filter is a good model to forecast exchange rate or not. When It discussed whether the model was appropriate for the models past literature, they were comparing RMSE with MAE in each model. This research uses Kalman filter to forecast exchange rate that combines the causality of exchange rate with stock index. The forecasting performance is emphasized the degree of accuracy of direction, not the size of errors. If a firm hedges with derivatives, it should use the proper model which can make the forecasting accurately, and it can achieve a better hedge effect. This study confers on the exchange rate to U.S dollars of five countries—France, Germany, Taiwan, Japan and Singapore. And we divide our sample into different periods depended on the degree of opening of capital market. Finally, this research compares Kalman filter with Random walk and GARCH model.