研究主要探討在全球金融海嘯期間,震盪異常劇烈的俄羅斯股市波動度的特性,標的指數選取具市場代表性的俄羅斯RTS指數。本研究藉由ARCH、GARCH與EGARCH等波動度模型,在市場報酬率普遍存在高狹峰與厚尾現象,何種模型對於俄羅斯股市波動性具有相對較佳的預測能力,同時也透過常態分配、t分配和GED分配等三種殘差分配,找出相對較為符合其波動度的分配。另一方面,本研究引入5日滾動報酬率(5 days rolling return, RR5D)做為真實波動性的代理變數。 而在探討預測績效方面,本研究使用均方誤差(RMSE)、平均絕對誤差(MAE)與泰爾不等係數(Theil Inequality Coefficient, THEIL)三種不同的損失函數來檢定模型預測績效。結論顯示,考量RMSE為損失函數時,較佳的配適模型為GARCH-GED分配、EGARCH-GED分配與GARCH-常態分配,然而當損失函數為MAE時,EGARCH-GED則有最佳的預測能力,故整體而言,EGARCH-GED分配為較佳的配適模型。此外,考量泰爾不等係數時,EGARCH-t分配模型的預測準確度最佳,GARCH-t分配模型次之,顯示殘差分配為 t 分配時較為允當,雖異於損失函數RMSE、MAE的結果,但同樣證明殘差分配的設定較不對稱模型的設定更為關鍵。 整體而言,本研究結果顯示,不對稱模型及殘差分配設定,對於預測具有不對稱特性的俄羅斯股市波動度確實有較佳的效果,不過殘差分配的設定的重要性要較不對稱模型的設定更加凸顯。
This study selects the appropriate model to match volatility of Russia stock market from ARCH, GARCH and EGARCH models and find the appropriate distribution assumption from normal, t and GED distribution. In the meantime, I use “5 days rolling return” to be the proxy of true volatility. This study uses three kinds of loss functions, including RMSE, MAE and THEIL. The empirical result indicates that the GARCH-GED model, EGARCH-GED model and GARCH- ND model have superior forecasting ability of volatility for Russia stock market with RMSE loss functions. However, the GARCH-GED model has the best performance when using MAE as the loss function. As for THEIL, the EGARCH-t and GARCH-t are top 2 models, the former better than the latter. On the basis of the empirical result, there are high performance to forecaste volatility of Russia stock market which is asymmetric when asymmetric models and correct distribution assumption be used but distribution assumption seems more crucial than alternative asymmetric for volatility forecasting.