Title: Twits versus tweets: Does adding social media wisdom Trump admitting ignorance when forecasting the CBOE VIX?
Authors: Steven Lehrer - Queens University (Canada)
Tian Xie - Shanghai University of Finance and Economics (China) [presenting]
Xinyu Zhang - Academy of Mathematics and Systems Science, Chinese Academy of Sciences (China)
Abstract: A rapidly growing literature has documented improvements in forecasting financial return volatility measurement via use of variants of the heterogeneous autoregression (HAR) model. At the same time, there is an increasing number of products made from social media that are suggested to improve forecast accuracy. We first develop a model averaging heterogeneous autoregression (MAHAR) model that can account for model uncertainty. Second, we use a deep learning algorithm on a 10\% random sample of Twitter messages at the hourly level to construct a sentiment measure that is being marketed by the Wall Street Journal. Our empirical results suggest that jointly incorporating model averaging techniques and sentiment measures from social media can significantly improve the forecasting accuracy of financial return volatility.