VaR最適計算方法之選擇

Abstract

VaR (Value at Risk)為現今風險管理的重要工具。自從巴賽爾協定開放各家金融機構可自由選擇VaR的計算模型後，如何在眾多計算方法中選出適當的方法便成為了各機構關心的課題。 本文提出兩個新的VaR計算方法。其中一個方法是每日先由數個文獻中常見的VaR計算方法中挑選出能通過Pérignon and Smith (2008)所提出之回溯測試（Backtesting）的計算方法，再以這些方法之VaR平均值作為VaR的估計值。另一個方法則是每日選擇回溯測試p-value最大值的方法所計算出之VaR值作為VaR的估計值。在過程中所使用到的VaR計算方法包含：GARCH、GARCH-T、EGARCH、EGARCH-T、GJR、EWMA (Exponentially Weighted Moving Average)與HS (Historical Simulation)法。 本文以臺灣指數資料來驗證新VaR計算方法之效率性。實證結果發現，與GARCH、GARCH-T、EGARCH、EGARCH-T、GJR、EWMA、HS及McAleer et al. (2009b) 文中的積極策略和保守策略相較，本文提出的兩個方法表現良好。

VaR (Value at Risk) has evolved as a standard risk measure. As financial institutions are allowed to choose their internal VaR model, the selection of the most appropriate method has become an important issue for financial institutions. In this thesis, two new VaR calculating methods are proposed. One involves first using a backtesting method, introduced by Pérignon and Smith (2008), to choose several VaR models from those commonly used by financial institutions. Then the average of the VaRs calculated by those chosen methods is the estimated VaR. The other selects VaR calculated by the common method that has the highest p-value as the estimated VaR. In the process, the common VaR models used are: GARCH, GARCH-T, EGARCH, EGARCH-T, GJR, EWMA (Exponentially Weighted Moving Average) and HS (Historical Simulation). Taiwan index data are used to testify the efficiency of the new VaR calculating methods. Empirical result shows that, compared to GARCH-T, EGARCH, EGARCH-T, GJR, EWMA, HS, as well as the aggressive and conservative strategies (McAleer et al. (2009b)), new VaR calculating methods perform relatively well.