由於預測-資產組合之未來的涉險值(Value at Risk)時,必須投入個別資產之報酬的變異數,及彼此之間的共變異數,因此本文分析多變量變異數預測模型對涉險值預測之精確度的影響。針對-外匯資產組合,本文比較的模型包括食物上最常用的指數權移動平均法(EWMA)、單變量(univariate)GARCH模型、與多變量(multivariate)GARCH模型。實驗結果顯示,較具效率性的多變量變異數模型能夠有效提高涉險值預測模式之精確度,利於投資者或市場參與者更有效的控管風險性資產。
The purpose of this paper is to study the impact of a more efficient covariance-variance estimate on the forecast of value-at-risk (VaR). One-day VaR forecasts are calculated based on the variance-covariance forecasts from the exponential weighted moving average method, currently the most commonly used method, the univariate GARCH-type models, and the BEKK multivariate GARCH models. Using a portfolio composed of three foreign exchange rates, the Japanese yen/US$, British pound/US$, and Deutschemark/US$, this paper evaluates the forecasting performance of VaR obtained from the above approaches. The empirical results show that the VaR calculated from the BEKK multivariate GARCH model outperforms other estimates from competing variance-covariance models.