- 學位論文
運用多元GARCH模型於銀行投組避險之實證研究
The Empirical Study on Portfolio Hedge of Banks by Multivariate GARCH Models
摘要
這篇論文主要在探討台灣的銀行產業在利率及匯率風險下的投資組合動態避險績效。因此，本文研究台灣資產最大的銀行─台灣銀行。並搭配Bauwens, Laurent and Rombouts (2006)提出關於多變量GARCH模型的三種分類，我們各取一個模型於各分類中，分別為BEKK-GARCH、DCC-GARCH、GO-GARCH，進一步探討其避險績效。 實證結果顯示DCC-GARCH模型相較於另外兩個模型，可以更精確地描述殘差的異質變異效果。不論在樣本內或樣本外，DCC-GARCH模型相較於另外兩個模型皆有較佳的夏普比率和預期效用。最後，我們在樣本外期間使用Diebold and Mariano (DM)單邊檢定，在10%顯著水準下，DCC-GARCH模型有較佳的避險績效。
並列摘要
The dynamic portfolio hedging effectiveness of Taiwan banks with exposure to both interest rate and foreign exchange risks is examined in this paper. Therefore, the paper investigates Bank of Taiwan, the bank which has the largest asset in Taiwan. Because there are three classes of multivariate GARCH models (see Bauwens, Laurent and Rombouts, 2006), this paper chooses one model for each class. Those models are BEKK-GARCH, DCC-GARCH and GO-GARCH respectively, and this paper discusses the hedging performance of those models in advance. Empirical results show that compared with the other two models, DCC-GARCH can explain the ARCH effect of the residuals better. In both in-sample and out-of-sample, DCC-GARCH has the highest Sharpe ratio and expected utility among the three models. In the end, this paper measures the Diebold and Mariano (DM) test statistics for out-of-sample, DCC-GARCH has better hedging performance than the others under the 10% significance level of the one-sided test.