最適避險比率期貨避險策略的研究相當可觀，同時也存在著多種不同衡量避險績效的方程式，然而每一個避險策略皆有其優缺點。本文著重在長天期期貨避險之檢討與改善。首先，傳統的最小變異數方法可以求出最小變異數的避險比率，但是隨著避險期間拉長，此法將面臨樣本減損的問題。另外，儘管資料分析已成為金融研究的核心活動之一，但是傳統的分析方法大多受限於資料形式須為穩態或者線性。我們提議以一個新發展出來的適應性資料分析法解決這些問題，此法稱為經驗模態分解法。我們將在文章中簡短的解釋這個方法，並且呈現出其如何應用在長天期最適避險比率的設計上。我們運用此法推演出的最適避險比率和經由傳統最小變異數方法求出的最適避險比率，分別建構投資組合進而衡量其避險績效的好壞。最後，我們將分析樣本內和樣本外，避險期間對避險績效的影響效果。 There are considerable amount of studies on futures hedge with optimal hedge ratio and many different functions to evaluate the performance. However, each hedge strategy has its own strength and weaknesses. This thesis will focus specifically on the issue of optimal futures hedge ratio for longer horizon. First, conventional minimum variance (MV) method can get the minimum variance hedge ratio, but it still suffers from the sample reduction problem when the hedge horizon extends. Second, as data analysis has been one of the core activities in financial researches, most existing methods are confined to stationary or linear data. To solve these problems, we propose a newly developed adaptive data analysis method, empirical mode decomposition (EMD) method for hedging with futures in longer horizon. We will briefly explain the method and demonstrate applications on our derivation of an optimal hedge ratio. We use the hedge ratio to form a hedged portfolio then compare the hedge performance with the MV method. Finally, we discuss the effect of hedge horizon and hedge performance for both the within sample and out-of-sample periods.
