Does intraday high-frequency investor sentiment help forecast stock returns? Evidence from the MIDAS models

Purpose This paper aims to enhance the predictability of stock returns. Existing studies have used investor sentiment to forecast stock returns. However, it is unclear whether high-frequency intraday investor sentiment can enhance the forecasting performance of low-frequency stock returns. Design/methodology/approach Thus, we employ the MIDAS model and the high-frequency intraday sentiment extracted from the Internet stock forum to forecast Chinese A-shares returns at daily frequency. Findings The results illustrate that high-frequency sentiment data are better than daily sentiment data in predicting daily stock returns, and the sentiment in non-trading hours has been proved superior to those in trading hours. Originality/value First, our study adds to the growing literature on investor sentiment. We are the first to construct a proxy for high-frequency investor sentiment using intraday postings collected from Chinese Internet stock forum. Second, we confirm that sentiment in non-trading hours has a stronger predictive ability than those in trading hours. Third, we also contribute to the performance comparison of MIDAS-class models. The good performance of U-MIDAS is confirmed in our empirical applications.