本文以VaR之變異數-共變異數法為理論基礎,期望建構更為精確之風險衡量模型。為達此目標,本文分三階段進行:(1)考量「局勢變遷」因子,在預期未來局勢趨於穩定時,透過灰預測修正資料,來捕捉可能之實際損失;(2)處理「資產權重」因子,引入熱力學之熵測度來處理投資組合之資產權重;及(3)針對「波動性」因子,採用實證力較強之指數加權移動平均法(EWMA),並建構直交門檻一般化自我迴歸條件異質變異數(Orthogonal TGARCH)模型來強化捕捉波動性之能力。最後,本文透過美國那斯達克生技類股之九支股票來對上述三階段進行實證研究,並驗證本研究所提出之投資組合風險衡量模型的修正變異數-共變異數法是否在波動性之解釋能力以及模型精確度檢定方面較未修正變異數-共變異數法更為精確。
Taking the value at risk (VaR) variance covariance approach as a theoretical base, this study sets out to develop a more accurate risk-measuring model. The study is divided into three stages to achieve this objective. In the first stage, the factor of ”environmental change” is considered. When we expect the environment to become stable in the future, a gray forecasting model is used to modify the data and to estimate actual losses. In the second stage, the factor of ”asset weight” is processed. This study adopts the concept of concept in entropy in thermodynamics to calculate the asset weight of a share portfolio. In the last stage, the factor of ”volatility” is processed and precipitated. This study uses a strongly practical model-the exponential weighted moving average (EWMA). At the same time, the orthogonal TGARCH is incorporated into the proposed risk-measurement model to enhance the ability to allow for volatility. Finally, this study puts these three stages mentioned above into practice in relation to nine Nasdaq biotech stocks in the U.S.A., and verifies whether the modified variance-covariance approach in this proposed risk measuring model for investment portfolios is more accurate or not in extreme scenarios.