Which sentiment index is more informative to forecast stock market volatility? Evidence from China
Chao Liang,Linchun Tang,Yan Li,Yu Wei
DOI: https://doi.org/10.1016/j.irfa.2020.101552
IF: 8.235
2020-10-01
International Review of Financial Analysis
Abstract:<p>In this paper, we investigate the predictive ability of three sentiment indices constructed by social media, newspaper, and Internet media news to forecast the realized volatility (RV) of SSEC from in- and out-of-sample perspectives. Our research is based on the heterogeneous autoregressive (HAR) framework. There are several notable findings. First, the in-sample estimation results suggest that the daily social media and Internet media news sentiment indices have significant impact for stock market volatility, while the sentiment index built by traditional newspaper have no impact. Second, the one-day-ahead out-of-sample forecasting results indicate that the two sentiment indices constructed by social media and Internet media news can considerably improve forecast accuracy. In addition, the model incorporating the positive and negative social media sentiment indices exhibits more superior forecasting performance. Third, we find only the sentiment index built by Internet media news can improve the mid- and long-run volatility predictive accuracy. Fourth, the empirical results based on alternative prediction periods and alternative volatility estimator confirm our conclusions are robust. Finally, we examine the predictability of the monthly sentiment indices and find that the two sentiment indices of social media and Internet media news contain more informative to forecast the monthly RV of SSEC, CSI800, and SZCI, however invalid for CSI300.</p>
business, finance