Social Data Predictive Power Comparison Across Information Channels and User Groups: Evidence from the Bitcoin Market

Abstract

In  the  context  of  Bitcoin,  we  examine  the  relationship  between  Bitcoin  price movement  and  social  data  sentiment.  Baseline  findings  reveal  that  social  media provides value-relevant information in both short-term and long-term predictions. By  comparing the  predictive  power  across  different  information  channels  and different user groups, we found that (1) while speculative information predicts both long-term and short-term returns effectively, fundamental-related information only predicts long-term returns, and that (2) prediction accuracy is higher for less active users than for active users on social media, especially in long-term prediction.