Cunning of the Bulls? Discussion Networks and Crowd Wisdom in Stock Market Predictions

• In Event: Analytic and Computational Approaches in Social Psychology (Co-sponsored by Mathematical Sociology Section)

Abstract

In uncertain decision-making environments, we often rely on peers to provide guiding signals. While such social influence can produce “wiser” crowds, the tendency to self-sort into homogeneous enclaves can also trap members of the crowd in information-poor echo chambers. This paper examines the intelligence of networked predictions using data from Stocktwits.com, a Twitter-like social network in which amateur and professional stock traders follow stocks, post comments, and share predictions. Applying network measures, computational text analysis, and panel data models to approximately 160 million posts spanning nearly a decade, we show that predictions about future price movement were more accurate for stocks discussed by “bridging” crowds of users from different parts of the discussion network, and when the discussions themselves were more diverse in content. While these two mechanisms—network bridging and discursive diversity—are often conceptually linked in previous work, we surprisingly find that their effects were largely independent. This study provides large-scale observational evidence for the wisdom of (network) crowds but also highlights the critical ways this wisdom depends on how those crowds are embedded in network and cultural spaces.

Authors

• Jungsik Choi, Pennsylvania State University

• Daniel DellaPosta, Pennsylvania State University