| 摘要: | 本篇文章採用Maheu and McCurdy（2004）所提出GARJI模型和GARCH模型估算不蘭特原油期貨、S＆P500指數現貨與美國30年期公債期貨 之風險值。由於GARJI模型可反應市場對於非預期的新訊息所造成的衝擊且具有較好的樣本外波動預測能力，因此本文利用GARJI模型捕捉此 不連續的狀態，並將此報酬不尋常表現的情形納入計算風險值的過程中，同時將偏態係數納入百分位數的修正。由實證結果可知，在通過回溯 測試的前提下，GARJI的穿透率和RMSE均較GARCH模型低，因此其風險管理的績效較GARCH模型優異，而在壓力測試上也有佳的表現。
This study employs GARJI (Maheu and McCurdy, 2004) and GARCH models to calculate value-at-Risk (VaR) of Brent oil futures, S&P500 index, and 30-year US Treasury Bond futures. GARJI model not only captures occasional large changes in price which is induced by the impact of unexpected news arrivals, but also has better forecasting ability of out-of-sample volatilities. Therefore, we adopt GARJI model to take these advantages and modify percentile by conditional skewness coefficient to the computation of VaR. The empirical results indicate that GARJI model has better risk management performance than GARCH model as viewpoints of failure rate and RMSE, and it also performs better than GARCH model in Stress-Testing. |
