如何有效率的分配有限的資金在各種不同的投資工具上對投資人而言是一個重要的課題，尤其是處於動盪不安的經濟情況下。最近在金融實證的文獻中指出，股價報酬率間呈現複雜的尾端相關性，我們可以使用 copula來描述這些複雜的相關性結構。此類方法所產生的隨機變數不同於傳統的分配，可以將邊際分配與相關性結構作任意的組合，使得在聯合分配的配適上有較大的靈活度。而此論文使用Pair-copula constructions則提供一個彈性方法來建構高維度copula。在Markowitz 的Mean-Variance架構之下，我們比較報酬率在傳統多維度常態分配假設與pair-copula constructions建立的模型之下的投資組合表現。若報酬率間的相關性結構偏離傳統假設，則pair-copula constructions的確具有較佳的投資組合表現。How to allocate his/her wealth among di erent investment tools more e ciently forindividual investor is an important issue especially in the volatile economic situation.The stock index returns exhibit complex patterns of tail dependence which can becaptured by copula models. We apply pair-copula constructions for reducing theload of estimation. Under Markowitz's mean-variance framework, we construct twoportfolios based on two di erent return models: the multivariate normal distributionand the C-vine pair-copula decomposed model. By examining four Taiwan stockindices from 2002 to 2011, we nd that C-vine provides a better performance.
