Santa Clara University

Guided by Chatter

Social Media Can Help Predict Stock Fluctuations

Investors are always looking for an edge, and good information at the right time can give them a considerable advantage. Recognizing this, academic researchers have been looking at the connection between public information in the mainstream and social media and shifts in stock prices, or the ability to predict those shifts.

David Zimbra, who became an assistant professor of operations management and information systems (OMIS) in the fall of 2011, pushed that research to a more detailed level and found evidence that a careful and systematic analysis of online forum content can be used to predict stock prices with considerable success. His paper, “A Stakeholder Approach to Stock Prediction Using Finance Social Media,” co-authored by Hsinchun Chen, his graduate advisor, has been published in IEEE Intelligence Systems.

“Before this study most researchers in the field had been looking at social media forums as if the forum participants were all from one group with the same background and relationship with the firm,” Zimbra says. “My thought was that open social media spaces have a very diverse group of participants and draw a lot of different constituencies, or stakeholders, into the discussion. If you could break down the conversations by the groups they came from — for example, investors, activists, and employees — it would provide a more refined analysis of the forum discussion.”

By following the content of a Yahoo Finance chat room for a year Zimbra saw an enhanced ability to predict stock price movements.

It took Zimbra several years to develop an automated method of evaluating the comments, using artificial-intelligence methods. For the case study presented in the paper, he chose the Yahoo Finance Wal-Mart forum as the information source, although more recently he has applied the analysis to other forums and firms with similar success.

David Zimbra Associate Professor of OMIS

Zimbra used social network analysis techniques to identify members of stakeholder groups, and sentiment-analysis techniques to measure opinions conveyed in forum discussions. He then developed statistical models using historical information to determine the relationship between forum discussions and stock behavior over time. The model predictions were then utilized to perform daily trading of the Wal-Mart stock over a year based on observed forum discussions.

Using the approach of a predictive technician, who would respond quickly to capture profits before the market fully absorbs new information into stock prices, he began with a hypothetical investment of $10,000 in Wal-Mart stock. Every trading day for a year, his program tracked the comments on the forum, broke them down by stakeholder group, and at the end of the day, based on the comments, made a decision to buy, sell, or hold the stock.

At the end of the year Zimbra found that if he had bought the stock and held it for a year, he would have realized a one percent gain. Trading stock based on forum comments as a group would have yielded a return of 16 percent. Trading based on the forum comments, broken down by stakeholder group, would have brought a stunning 44 percent return.

“The impressive performance of the stakeholder-level model represented a statistically significant improvement in the accuracy of predictions,” Zimbra says. “To the degree that different constituencies were reflected in the analysis, there was a more accurate assessment of the forum discussions.”

Given the rapid growth of social media, there is more research to be done in the area, but Zimbra feels that the results of this experiment indicate a potential for much broader use of the techniques he developed.

“The implications of this research go beyond stock prediction,” he says. “It can help companies assess how people feel about their products and provide a wealth of information that has been drawn from focus groups up to this point. There are many ways in which businesses can maximize the value of social media.”


Want to download your own PDF. copy of the Winter 2012 Issue of Minds@Work?
Get it Here:

Winter 2012 Contents

Watch My Eyes
Savannah Wei Shi used state-of-the-art eye-tracking technology to determine what works best on an Internet page layout.

Holding Up Production
Ram Bala and his colleagues have developed a model for helping companies calculate the production capacity they’ll need when a patent on a major product expires.

Good Mood, Good Decision
John Ifcher’s controlled experiment demonstrated a correlation between being in a good mood and making good financial decisions. The shelf of papers behind him shows the amount of raw data generated in a similar experiment.

Foiling the Data Snoopers
Haibing Lu did pioneering research on how to define “data cubes” so as to both share and protect information effectively.

Guided By Chatter
David Zimbra found that following the content of a Yahoo Finance chat room for a year led to an enhanced ability to predict stock price movements.

Minds@Work Home

Printer-friendly format