透過您的圖書館登入
IP:3.131.110.169
  • 學位論文

探討利率變化與匯率波動之研究 -以PID 控制與自我迴歸模型比較分析

The Relationship between Interest Rate and Exchange Rate – Comparing PID Control and Autoregressive model

指導教授 : 胡為善

摘要


全球金融市場中有許多投資工具及不同種類的投資標的。而其中各國央行採行的貨幣政策常會牽動一國的資金流向,使得投資人必須評估其匯率風險,以便採行相關的資產配置。因而對投資者與企業經營者而言,投資組合及其匯率風險為投資上不可或缺的考慮因素,因此本研究將未拋補利率平價理論與工程上應用之PID控制系統結合,俾探討使用此模型來預測匯率是否會較AR模型及GARCH模型能有較佳的預測效果。 本研究樣本資料包括美國、歐盟及日本的匯率、十年期公債殖利率,採用TEJ月資料(1995年12月至2016年12月共253筆),與近一年的TEJ日資料(2015年12月31日至2016年12月7日共243筆),進一步運用AR系列與PID模型進行匯率預測,最後透過樣本外預測之誤差分析均方根誤差(RMSE) 與平均誤差絕對值(MAE),來評估各種匯率預測模型之預測能力。 本研究實證結果彙總如下: 1.在美元指數的匯率預測中,以AR模型與GARCH模型在RMSE與MAE的表現,勝於PID模型所作的預測。尤以GARCH模型在月資料的RMSE表現更好。 2.歐元的匯率預測中則以GARCH模型之預測月資料表現最佳,但AR模型在日資料之預測略則優於PID模型。 3.日圓的匯率預測不論在月、日資料上,GARCH模型之表現均為最佳、AR模型次之,PID模型的表現最差。此外,本研究亦發現,針對不同整體之預測結果,以日資料較月資料,更為精確。

並列摘要


There are various investment instruments and investment targets in the global financial markets, in which the monetary policies of the Central Banks of various countries should affect the capital inflows and outflows. So the investors and corporate managers must evaluate the exchange rate risk and adopt the related configurations. The emphasis of the portfolio and its exchange rate risk will be very important for investment. This study combines the interest rate theory with engineering application of the PID control system to forecast the exchange rate from December 31, 2015 to Dec, 7, 2016 and to compare the prediction result with the AR model and GARCH model.  The sample data includes the TEJ monthly data of the exchange rates of the United States, European Union and Japan, as well as the ten-year bond yields (from December 1995 to December 2016) and the daily data (December 31, 2015 to December 7, 2016). The RMSE and the MAE are used to analyze the forecasting power of various exchange rate models through the out-of-sample prediction errors. The empirical results are summarized below: 1.Regarding the US dollar index exchange rate forecasting, this study finds that the RMSE and MAE of the GARCH model outperform those of the PID model and AR model. 2.The GARCH model is also the best one on forecasting Euro. The AR model is better than the PID control model in daily data forecasting. 3.The GARCH model still performs the best in forecasting Japanese Yen, and the AR model is the second best one and the PID model performs the worst. Additionally, this investigation finds that the forecasting result of daily date is more accurate than that of the monthly data.

參考文獻


林政弘 (2011). 結合整體經驗模態分解, 基因演算法與極速學習機於財務時間序列預測之研究. 臺北科技大學商業自動化與管理研究所學位論文, 1-74.
李建慧 與 張珮芬 (2014). 匯率報酬預測績效模型之比較-委託單流量的探討. 商業現代化學刊, 7(4), 257-274.
鄭淑芬, 李壽昇, 黃家偉 與 郭東義 (2011). 線性非時變系統之即時參數估測. 工程科技與教育學刊, 8(2), 330-341.
方文碩 與 張倉耀 (2002). 風險貼水與外匯市場效率性. 管理評論, 21(2), 27-51.
謝宜倪 (2003). 國外利率衝擊, 國外資產累積與匯率動態調整: 代表性個人模型. 經濟論文叢刊, 31(1), 27-50.

延伸閱讀