雖然Markowitz 模型在實務上是可用的，但是它仍存在許多缺點，尤其是在估計最適權重時，常會因為參數過度敏感的問題而導致嚴重的估計誤差產生。為了解決這個估計誤差的問題，我們使用Black-Litterman model，藉由它將市場隱含報酬率與投資人的觀點做聯結去修正預期報酬率，此外為了能更準確刻畫投資人之觀點誤差我們也將標準的Black-Litterman model 模型結合不同之GARCH 模型去估計隨時間變化的共變異數矩陣。最後，我們利用台灣股票市場中的五檔產業指數提供了一些實證分析。;Asset allocation using Markowitz model has many disadvantages, particularly because the optimal weight is sensitive to the estimation error of the model. To overcome the problem of estimation error, we follow Black-Litterman model, where the initial expected returns are linked to market implied return and subjective views of investor for each asset to adjust the expect return. To adjust the heteroscedasticity of the volatility, we further combine the standard Black-Litterman model with several GARCH-typed models to estimate time-varying covariance matrix. Finally, we conduct an empirical analysis using five industry indexes in Taiwan stock market.
