綜合國內外有關匯率報酬風險值的文獻研究，大部分都是以單一模型來討論匯率報酬其尾部指數的估計及風險值，但甚少考量報酬序列存在波動叢聚與自我相關特性。本文採用1993年1月2日至2007年9月30日六個臺灣主要貿易國家匯率日資料，在考慮報酬序列具有GARCH效果下，搭配極值理論來估計報酬序列尾部指數及其風險值。實證結果顯示，使用極值理論估計尾部指數與風險值時，若能將原始序列加入GARCH效果，則風險值估計將更為準確。 Majority of the previous studies about VaR of Foreign Exchange Rate were using simple model to estimate the tail index and VaR of exchange rate return with very little consideration of volatility clustering and the autoregressive errors of financial asset returns. This research analyzed the Foreign Exchange Rate of 6 main trading nations of Taiwan from January 2, 1993 to September 30, 2007 to evaluate VaR. We estimate the exchange rate return under GARCH effect and evaluate the tail index of exchange rate return through EVT method. The empirical result shows that the unconditional tail index estimate under GARCH effect is smaller than that ignores GARCH effect, in other words, the tail distribution of financial asset returns with GARCH effect is more accurate than that ignores GARCH effect.
