股匯市為兩大投資主流，國內超過六百萬人投資台股，而新台幣匯率不但影響海外投資損益，也攸關進出口廠商競爭力，如何精準預測台股以及新台幣匯率的波動性，重要性不言可諭。 學界及實務界經常使用GARCH模型預測波動性，本文同時採用Chou(2005)CARR模型，針對台股指數以及新台幣匯率，實證模型的預測能力。結果顯示，CARR模型對台股指數的預測能力比GARCH模型佳；而GARCH模型對新台幣匯率的預測能力比CARR模型佳。本文以迴歸分析，仍獲得相同驗證；不同的財務金融資料，適合不同的模型。 此外，財務金融資料經常具有「肥尾」特性，本文實證搭配Weibull分配、常態分配、以及肥尾分配的不同假設。結果顯示，CARR模型，以Weibull分配對台股的預測能力較佳；而GARCH模型，以肥尾分配對新台幣匯率的預測能力較佳；不同的假設分配，預測結果不盡相同。 More than six million people invest in Taiwan stock market .When investment extends to overseas , the currency exchange rate will affect return. The currency exchange rate also affects the importer and exporter’s cost. So how to capture the characteristics of volatility and predict it correctly will affect investors and policymakers ,if do well , it can lower risk and have better performance . The academic and the real market usually use GARCH models to predict the volatility. However, Chou(2005) proposed the CARR model and applied in S&P 500 index. We use both GARCH and CARR models to test Taiex and Twd exchange rate. We find that for Taiex CARR model is better than GARCH ,but for Twd exchange rate GARCH still better than CARR . Different data uses different models. Because financial data usually have heavy-tailed characteristic , we also find that Weibull distribution is better for CARR applying in Taiex, and heavy-tailed distribution is better for GARCH applying in Twd exchange rate .Different distributions also fit in different financial data.
