"When Do the Crowds Lose Their Wisdom?" by Prof. Susanna Ho
- 2:30 p.m. — 4:00 p.m.
Professor Susanna Ho
ANU College of Business and Economics
The Australian National University
Information systems (IS) researchers have recently leveraged the wisdom-of-the-crowd effect on social networks to make predictions. In this paper, we argue that wise crowds do not always function in the event of social contagion. Two dimensions quantify social contagion effects: intensity and extensity. We take the approach of predictive analytics and identify three network characteristics—argument diversity and sentiment diversity (to model intensity of contagion) and tree size (to model extensity of contagion)—to predict when the wisdom of the crowd disappears. The context of our examination is the prediction of stock returns of initial public offerings (IPOs); the loss of crowd wisdom is manifested as an abrupt change in stock returns. Our social network data come from StockTwits. We extract 18 million tweets for 859 IPOs launched between 2008 and 2016, corresponding to 177,053 firm-week observations. Based on the results of logistic regression and multinomial logistic models, we find that a network comprised of large trees and posts with similar arguments (or sentiment) leads to a loss of crowd wisdom, and that these factors increase the predictability of an abrupt change in stock price (e.g. crashes and jackpots) in the subsequent week. Further, we assess predictability levels and measure the predictive power of the models with out-of-sample predictions to confirm the practicality of our developed models. Theoretical and practical contributions will be discussed.