本研究以新台幣兌美元匯率為主要研究對象估算其風險值,並應用Bollerslev (1986) 提出的GARCH、Glosten,Jaganathan,and Runkle (1993) 提出GJR-GARCH、及Tse (1998) 提出的FIAPARCH模型來衡量新台幣兌美元匯率之報酬波動,並以GARCH (1,1) 為基礎,使用GJR-GARCH及FIAPARCH為擴展模型。為了更精確計算風險值故採用常態分配及t分配來做模型估計及風險值之計算,並進行回溯測試 (back testing),比較各模型間估計績效。實證結果顯示ARMA (2,2) -FIAPARCH (1,d,1) -N為最佳估計風險值之模型。
This paper focus on the VaR of the exchange rate of NTD/USD. In addition to value the return volatility of the NTD/USD exchange rate by applying the Bollerslev’s (1986) GARCH model, Glosten, Jaganathan and Runkle’s (1993) GJR-GARCH model, and Tse’s (1998) FIAPARCH model, an extended GJR-GARCH and FIAPARCH model based on the GARCH (1,1) is used. Normal distribution and t-disruption are used to achieve higher accuracy of the estimation of VaR. The efficiency among different models are tested by a back testing. The result shows that the ARMA (2,2) -FIAPARCH (1,d,1) -N model is the best model to estimate the VaR.