A Forecast Comparison of Volatility Models in Taiwan Stock Market

• Main Authors: Hsieh tin-cheng, 謝廷政
• Other Authors: Chin-Hsien Lo
• Format: Others
• Language: zh-TW
• Published: 100
• Online Access: http://ndltd.ncl.edu.tw/handle/68649912044114696677

碩士 === 國立高雄應用科技大學 === 金融資訊研究所 === 99 === Abundant GARCH models are proposed in the econometric literatures, but which one is the most appropriate to describe the data has not been consistently concluded. This study use the SPA test of Hansen(2005)and Step-SPA of Hsu,Hsu and Kuan(2008) to compare the forecasting ability of alternative GARCH models in Taiwan stock market. To resolve the forecasting performance distortions aroused from microstructure noise, the realized volatility and realized range-based volatility are used to proxy the real latent volatility, the MSE and QLIKE are used as the loss functions. The empirical results indicate that the forecasting performances of most GARCH models are similar. APARCH and EGARCH are the most superior among the various models.