eInformation: a clinical study of investor discussion and sentiment

Financial Management (Financial Management Association), Autumn, 2005 by Sanjiv Das, Asis Martinez-Jerez, Peter Tufano

These observations provide motivations for voluntary postings on stock chat message boards. The theoretical justification for posting might be related to models of information disclosure. Suppose each investor receives a noisy signal about future stock price, e.g., their opinion as to the importance of a new product announcement. By sharing their signals with others, they can verify the information before trading, or can share the signal with others after trading, with the hope that their interpretation will lead to the desired movement in share prices. On a more mundane level, stock chat boards can be locations where disgruntled shareholders, customers, employees, and former employees can share their experiences with others.

To the extent that the online community serves as a social group or debating society, its economic impact is probably secondary. However, to the extent that it serves as a vehicle for testing ideas and analyses, it frames some interesting questions that could form the basis for subsequent research. We could ask, what are the relative returns from communal compared to individual analysis? Glenn believed that his analysis would be improved by testing it with others, but this is an untested assertion. More narrowly, can discussion, even among well-meaning investors, have the impact of producing even more severe biases, like the hardening of Glenn's investment bias? Ex post, Glenn reached this self-critical conclusion, but it could be a more general phenomenon.

III. The Substance of the Discussions

Exploiting our clinical research design, we analyze the content of the postings. We look at what subjects are discussed, whether discussions stay on point, and whether the discussions reveal meaningful information.

We conduct two analyses. In the first, we examine the actual news released by firms, and look before and after the news release to understand any foreshadowing of the news prior to the release and the subsequent digestion of the news after the release. From this analysis we find that the discussion boards seem to play an important role in rapidly disseminating news, sometimes "breaking stories" before they are covered widely. In the second analysis, we examine a set of rumors on the boards and track them through time. From this analysis, we conclude that discussion boards are rumor mills for many unsubstantiated claims and a poor source of inside information.

In the first experiment, we selected 16 seemingly-newsworthy press releases by the four companies, based on our reading of the releases and inspection of abnormal returns around the announcements. We performed event studies on the 16 announcements and found that ten of the events have abnormal 1-day or 3-day returns over 5%, however given the high level of volatility in general, these returns are statistically significant in only four cases. For each of these companies we trace how the "news" is communicated to investors through traditional media as well as through postings (Table III lists the events). We also try to understand how the press or message boards provided advance information of the event, and how the companies responded to the event.