天然氣價格風險值衡量

Abstract

本文使用歷史模擬法、GARCH、EGARCH、GJR-GARCH與CARR模型針對紐約商業交易所(New York Mercantile Exchange,NYMEX)天然氣商品(Natural Gas,NG)為研究對象，在誤差項分別假設為常態分配與t分配下計算其風險值，本文嘗試比較不同窗口長度和顯著水準下之模型特性；在模型評估方面，本文利用回朔測試，計算失敗次數、失敗比率，並以Christoffersen (1998)發展之Likelihood Ratio test 分別檢定未受條件限制下與受條件限制下之檢定統計量，衡量各模型績效。經過實証分析後發現，誤差項設定以常態分配較能準確評估天然氣期貨資產特性，失敗比率上五種模型都有不錯的表現，隨著顯著水準的降低，模型將趨於穩定，在LR test中以CARR與EGARCH模型表現較佳。

This paper use historical simulation approach, GARCH model, EGARCH model, GJR-GARCH model and CARR model, under different error term distribution hypothesis to estimate the Value at Risk of NYMEX Natural Gas price. We use different number of days and significant level to test our model and observe their performance. To value our model, we use back test, calculate failure frequency, failure ratio, and use Christoffersen’s Likelihood Ratio Test to test unconditional and conditional test statistic. According to our analysis, error term should be suppose to normal distribution, that can fit the NYMEX Natural Gas property. The model have good performance in failure ratio, as the significant level goes down, the model will be more stable, CARR and EGARCH model have better performance in LR Test.