本研究延續Moschini and Myers（2002）二元GARCH模型，文中假設資產報酬的共變異矩陣具有自我迴歸異質化的性質，放寬BEKK模型的參數設定並將星期效果納入共變異矩陣考量，以檢定台股指數期貨和摩根台股指數期貨的最適避險比率是否是一固定避險比率。本文蒐集1998年9月1日至2005年12月30日台灣加權股價指數與台股期貨指數，以及摩根台股指數與摩根台股指數期貨資料，資料期間橫跨2921日，期間包含1867交易日。本研究檢定拒絕「現貨市場與期貨市場報酬的條件共變異數和期貨市場報酬的條件變異數的比值為常數」的虛無假說，且拒絕「最適避險比率會出現週期性變化」的虛無假說。研究結果可知資料期間投資人的「最適避險比率」會隨時間變易，且星期效果變數無法完全解釋此一現象。若比較求算的最小風險避險比率在二期指市場的避險績效，則發現在未考慮星期效果的原始模型中所求算的最適避險比率可以降低台灣加權股價指數和摩根台股指數投資組合的風險。另外，在模型加入星期一到星期五效果後，則發現由所求算的最適避險比率可以降低台灣加權股價指數投資組合和摩根台股指數投資組合的約80％的風險。 Allowing for a more flexible BEKK form of time-varying volatility and with the day-of-the-week effect embedded in the variance-covariance matrix, the study follows a bivariate GARCH parameterization from Moschini and Myers (2002) to test the hypotheses that the optimal futures hedge ratios of MSCI Taiwan Index futures and TAIFEX Stock Index futures are constant over time. The time period covered is from September 1, 1998 through December 30, 2005, including 1867 daily observations over a span of 2921 calendar days. The empirical results show that the null hypothesis of a constant hedge ratio is statistically significantly rejected and the time-varying optimal hedge ratios cannot be explained solely by the day-of-the-week effect. It is also found that over 80% of the variance of the unhedged portfolios returns can be reduced by the hedging strategies suggested in the study for both MSCI Taiwan Index futures and TAIFEX Stock Index futures.
