The information content of analysts’ textual reports and stock returns: Evidence from China

Abstract

We apply a Long Short-Term Memory (LSTM) deep-learning approach, which can capture the order of words, to extract textual tones from analyst reports written in Chinese. The market reaction is significantly and positively correlated with the textual tone of the analyst reports. Furthermore, we find that the market reaction is stronger to negative tones. Additionally, investors are more responsive to the textual tone from star analysts and the analysts from larger brokerage houses. We also investigate how the margin trading and Shanghai-Hongkong (or Shenzhen-Hongkong) stock connect impact the informativeness of textual opinion.