The efficient market hypothesis assumes that stock prices are solely dependent on the rational processing of information and, thus, cannot be predicted. In his talk Marten Risius shows that social influences on social media platforms affect information dissemination and, ultimately, stock returns.
Thus, we study (a) whether buy- and sell-signals of Twitter messages have a predictive power of stock prices, and (b) how the influence of users and messages relate to this effect. We analyze 5 million stock-related Twitter messages issued about the S&P 100 companies, and identify buy- and sell-signals by applying a supervised text mining algorithm. Based on stock market data on a minute basis, we focus on short-term effects of the messages, and measure how they change within 15 to 60 minutes. The results confirm that financial predictive analytics needs to focus on negative messages especially from users with a large follower base across all time-frames. Analyzing the spread if negative messages shows that the share original messages are only relevant within the first 15 minutes, within 30 minutes only the retweet numbers matter in determining stock prices.
Dr. Marten Risius: Clemson University, South Carolina, The University of Queensland
Weizenbaum Institute, Hardenbergstraße 32, 10623 Berlin, Room A103 – A105
Thursday, 20. June 2019
02:00 - 03:00 PM
Back to previous page