Optimal Forecasting Models and Trading Strategies of the Futures-Time Series Models with Standard Lévy Process

• Main Authors: Kao, Ching-Wen, 高靜雯
• Other Authors: Chia-Chien Chang
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/41810551773533359047

碩士 === 國立高雄應用科技大學 === 金融資訊研究所 === 99 === We address forecasting and trading strategy by (1) considering a ARMA-GARCH model with standard Lévy process innovations: the Jump Diffusion (JD), Generalized Hyperbolic (GH), Hyperbolic, Normal Inverse Gaussian (NIG), Variance Gamma (VG), GH Skewed T, Student’s t, Classical Tempered Stable (CTS), CGMY as well as Hansen's Skewed T, (2) estimating the model with a sample including 10 years of daily data (including in sample and out of sample), principally focused on the innovation’s skewness, leptokurtosis, fat tails as well as the time varying volatility, and (3) testing empirically the performance of these models for the Dow Jones Industrial average (DJIA) Index Futures during December 16, 1999 to August 17, 2010 and Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) Futures during December 16, 1999 to September 13, 2010 with moving windows. Moreover, we do the futures price forecasting and building a trading rule to make positive profits. The empirical results show that using the ARMA-GARCH model with standard Lévy process innovations for forecasting and trading strategy generates significantly higher profits and greater directional accuracy than those based on normal innovations. To robust the ARMA-GARCH models with various distributions, by moving windows method, we consider 5 days moving window and 30 days moving window, and make the similar results. Therefore, Lévy-ARMA-GARCH models seem to be better in both the forecasting ability and strategy performance than Normal-ARMA-GARCH model.