本篇研究同時考慮解決資產報酬變異隨時間變化的特性與當標的資料分配不符合原始假設的問題。為配合資產報酬變異隨時間變化的特性,通常利用等權移動平均法(SWMA)、一般向量自我迴歸共變異矩陣(GARCH)與指數加權移動平均法(EWMA)等模型來對估計值加以處理。然而,這些方法都是根據樣本的變異數或報酬的共變異數估算的估計值。當原始的標的資料分配不符合假設,則會有估計不具效率的問題。本研究最主要目的在估計這些厚尾高峽峰型態資料之頑強最適避險比率,來驗證台灣股價指數期貨與摩根台灣股價指數期貨並進行動態的避險策略。本研究同時使用條件下等權移動平均法與指數加權移動平均法來進行頑強最適避險比率的估算,在將之與非條件下的個別方法估算出來的估算值進行比較。可以得知頑強最適避險比率估算的避險投資組合變異數比個別方法相對較小,更甚而言之,頑強的最適避險比率所估算的變異數比為避險時所估算的變異數減少相當多的程度,且可在動態避險時大量的減少交易成本。 This article considers solving that conditional distribution of most financial asset return tends to vary over time and the distribution of underlying asset does not conform with ordinary assumption simultaneously. In order to fit time-varying volatility in returns of asset, estimators dealt with simple weighted moving average (SWMA), GARCH, or EWMA models are usually applied. However, these methods are estimated according to sample variance and covariance estimators of returns. When the distribution of underlying asset does not match with the ordinary assumption, the estimators are not in general efficient. The primary purpose in the article is to verify the dynamic hedging strategies in Taiwan stock index futures and MSCI futures by estimating the robust estimation of optimal hedge ratio (OHR) when the data is leptokurtic and fat-tail. This article uses conditional SWMA and EWMA at the same time to estimate the robust estimation of OHR, and compares with the results in unconditional SWMA and EWMA. The variance of hedged portfolio is computed in the robust OHR are less than that in the unconditional way. In addition, the variance of hedged portfolio is computed in the robust OHR are much less than before, thus reducing the transaction costs which produces in dynamic hedging strategies.