波動性在財務上扮演著關鍵的角色，若能適當的描述波動性模型，將有助於投資組合配置的最適化，進而能有效的控管風險。GARCH模型在波動性的預測上已被廣泛的應用，而且也能在實證上得到良好的成效。然而Chou (2005)將GARCH 模型結合變幅在波動性預測上的優勢進一步提出條件變幅自我相關(Conditional Auto-Regressive Range，CARR)模型，並且在S&P500 股價指數波動性預測實證上獲得優於GARCH模型的結論。本文中將介紹CARR 模型及其性質，並以日圓與星幣為研究對象，分別進行CARR 模型和Skew-t GARCH 模型在樣本內及樣本外波動性的預測能力比較；結果與Chou (2005)認為CARR模型對股價指數波動有較佳的預測結果相異。故認為CARR模型在預測波動性上並非具有完全的優勢，不同商品應適用不同的模型進行預測，以期得到最適的預測波動性，提升投資決策的效率。 In finance, volatility plays a key role in several sub-fields. Whether the construct of portfolio is optimal or not, partly depends on the control of volatility. GARCH family models have been used in the forecast of volatilities, and have performed well in many empirical studies. Recently, Chou (2005) proposed the CARR (Conditional Auto-Regressive Range) model. The main concept of the CARR model is to use a simple dynamic structure for range to characterize the volatility process. In Chou (2005), comparing the CARR model and traditional GARCH model, the former is better in the volatility forecasting based on the data of the S&P 500 index. We use both CARR and GARCH models to test JPY and SGD exchange rate. But we find that different data uses different models. In order to obtain the most accurate projection of volatility and improve the decision-making efficiency, it’s better to apply specific volatility forecast models to different products.