Investor herding behavior in social media sentiment

We investigate the mechanisms of investors' herding behavior using machine learning and textural data analysis from social media and the impact of sentiment in forming the herding behavior. We find that the abnormal information creation activity (AICA) for the retail investor is positive and statistically significant with the herding behavior, while informed investors with access to valuable information are negative with relation to the AICA. The herding behavior in firms traded by the retail investor is strongly related to the sentiment in social media at the cross-sectional level and has been more effective after COVID-19.