透過您的圖書館登入
IP:3.149.251.154
  • 學位論文

期貨商品最適預測模型之研究-以台灣市場為例

The Research of Futures Optimal Forecasting Models-An Example of Taiwan Futures Market

指導教授 : 洪茂蔚
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


本研究使用GARCH家族裡的四個模型:ARCH(1)、GARCH(1,1)、TGARCH(1,1)和EGARCH(1,1)來捕捉國內期貨市場的四個熱門商品:臺股期貨(TX)、電子期貨(TE)、金融期貨(TF)和小型臺指期貨(MTX)自2005年12月底到2008年4月共28個交易月的波動性走勢。且將樣本分成前24個交易月的樣本內資料和後4個交易月的樣本外資料。樣本內資料經模型配適後,利用AIC與SBC兩個模型配置準則評估模型配適度的好壞,之後再利用樣本外資料進行模型預測力的評估,評估方法為預測力指標RMSE、MAE及MAPE。 在模型配適度方面,臺股期貨、電子期貨和小型臺指期貨都顯示以TGARCH(1,1)模型的配置度最好,唯獨金融期貨以GARCH(1,1)模型的配適度較佳。在模型預測力方面,實證結果顯示臺股期貨、電子期貨和小型臺指期貨都支持以TGARCH(1,1)模型的預測能力較好,金融期貨顯示以EGARCH(1,1)模型的預測能力較好。 實證結果推論,國內金融市場的確存在槓桿效果,使得考慮到槓桿效果和變數較多TGARCH(1,1)模型配適度較佳,而沒考慮到槓桿效果的對稱性模型ARCH和GARCH的表現較差,符合國內外文獻實證的結果。推論也許是因為全球整個環境是一個大的經濟體,使得各國家之間的經濟狀況和金融商品的變動都有連帶關係,所以本研究會出現和國外實證一致的結果。經本研究結果推薦投資者在投資國內期貨商品時,可利用TGARCH模型來判斷波動性的走勢和預測,以之做為投資的參考。

並列摘要


This thesis use four models in the GARCH family, ARCH(1), GARCH(1,1), TGARCH(1,1), and EGARCH(1,1), to capture four popular financial products including TX, TE, TF, and MTX in domestic future markets in Taiwan from late December, 2005, to April, 2008. It revealed volatility trend in 28 reading months. I classified all samples into 24 pre-reading-months, in-sample data and four post-reading-months, out-of-sample data. fitting in-sample data in models, use AIC and SBC of the Model Selection Criterion to evaluate the goodness of fit of model, and then use out-of-sample data to evaluate model forecast ability through forecast ability index RMSE, MAE, and MAPE. In the goodness of fit of model, TX, TE, and MTX all indicate that TGARCH(1,1) model fits in the best; only TF indicates GARCH(1,1) model has better fitness. In model forecast ability, TX, TE, and MTX all indicate that TGARCH(1,1) model has better forecast ability; only TF indicates EGARCH(1,1) has better forecast ability. Since the leverage effect does exist in domestic financial markets, we implied from the empirical study that TGARCH(1,1) model considers more variables and the leverage effect, TGARCH model has better model fitness than parallel model, ARCH model and GARCH model which did not consider the leverage effect. This phenomenon accords with the results from international and domestic documents. We supposed that since the global financial market is a big economy, the economic situations and financial product fluctuations of each country are related to each other. We would suggest investors to use TGARCH model to evaluate volatility trend and forecast when invest domestic future products.

並列關鍵字

VaR TX TE TF MTX GARCH model

參考文獻


7.蔡宗和(2005),「厚尾GARCH模型在台灣金融資產之應用」,淡江大學財務金融研究所碩士論文。
6.林楚雄,高子荃,邱瓊儀(2004),「結合GARCH模型與極值理論的風險值模型」, 《管理學報》2005年,Vol.22,No.1,133-154
4.鄭天德(2002),「ARMA-TGARCH模型之建立」,國立交通大學經營管理研究所碩士論文。
1.陳裴紋(1995),「台灣股票市場報酬率與波動性預測之研究-ARCH-family模型之運用」,台灣大學財務金融學研究所碩士論文。
1.Bollerslev, T.R. “Generalized Autoregressive Conditional Heteroskedasticity,” Journal of Econometrics, Vol.31, pp.307-327.

延伸閱讀