FIEGARCH模型之風險值計算──以新台幣兌美元匯率為例

Abstract

本研究在資產報酬率之波動具有緩長記憶(long memory)的部分整合指數廣義自迴歸條件異質變異數（fractionally integrated exponential generalized autoregressive conditional heteroskedasticity；簡稱FIEGARCH）模型下，配合極端值理論(extreme value theory)計算風險值(value at risk)。本研究以新台幣兌美元匯率為例，實證結果發現其日報酬率之波動具有緩長記憶現象，且FIEGARCH模型配合極端值理論之風險值計算法的表現較為穩定。

In this paper, we use the FIEGARCH model combined with extreme value theory to compute Value at Risk (VaR) measure for daily asset returns. Our empirical example in using TWD/USD exchange rate returns shows that the data reveal long memory phenomenon and the FIEGARCH model combined with extreme value theory provides a good representation in VaR estimation framework.