Sunday, August 1, 2004

Disclosing Information on the Internet: Is It Noise or Is It News?

STANFORD GRADUATE SCHOOL OF BUSINESS—It doesn't take a rocket scientist to figure out that the Internet has revolutionized the way people communicate with one another. Indeed, the Internet has enabled individuals to broadcast messages to a potentially vast audience at very little cost. On the flip side, this cacophony of messages has led to a noisy environment, making the medium a less than ideal vehicle for meaningful communication.

Yet, despite the volume of disclosures and the potential noise in the system, the often anonymous disclosure of information on the Web is finding a receptive audience, and useful information is getting through. Two finance professors at the Stanford Graduate School of Business, Anat R. Admati, the Joseph McDonald Professor of Finance and Economics, and Paul Pfleiderer, the William F. Sharpe Professor of Financial Economics, set out to investigate how this process works.

The seeds of their recent research germinated in the notion that the key to culling information from the Web is sorting out the useful nuggets of news from the noise. Further reading about the effects of Internet disclosure on various companies' stock prices fueled their interest. In particular, their attention was drawn to the case of the 14-year-old New Jersey boy who got busted by the Securities and Exchange Commission for profiting by dabbling in stocks he had researched and touting those companies on the Internet. Pursued on the basis of manipulation laws, the boy's so-called crime is hard to distinguish from an analyst's role. Using data from Amazon.com book review ratings as a model for Internet disclosure, Admati and Pfleiderer's recent research has implications for any field affected by public disclosure.

The Web opens a whole new world where disclosure by individuals is extremely easy, says Admati. "In the past, you would have to buy an ad in the newspaper if you wanted to let people know your views on a book or some company's stock" she says. For example, at Amazon.com, one needs only to check off a rating number and possibly add a few sentences to register an opinion on a book or product and become an instant reviewer.

In their study, Admati and Pfleiderer develop an economic model to address what happens when consumers, for whatever reason, want to share information with fellow consumers and they have a simple avenue to transmit their opinions. Will information be transmitted in the most efficient way when forces are allowed to work unhindered? And will results be distorted if it costs something to send the message, or if some senders' opinions are not relevant to others' experience?

While Admati calls the model a simple one, it nevertheless lays the foundation for further research leading to increased understanding of the effects of financial and political disclosure in the new webbed world. "What we did so far was scratch the surface," she says. "We first wanted to understand the basic issues."

They found surprisingly rich results. They assume that there is a finite number of "messages" that the sender can use to communicate his or her information and that the sender is attempting to bring the receivers' opinions as close to his own as possible. If both the senders and the receivers of information understand the true likelihood that the sender's information is relevant, then the result is often that information is communicated in the most efficient way. Surprisingly, however, the most efficient communication might involve some messages being used much more frequently than others. "There is a preponderance of extreme messages—a lot of five-star ratings, for example—with some available messages, like two-star ratings, essentially unused," says Admati." Our results are generally consistent with our observations that the customer rating levels on the Internet are not evenly distributed."

A well-documented psychological phenomenon is that people are often overconfident concerning the usefulness and accuracy of their own knowledge and information. "If the sender in our model is overconfident," says Admati, "in other words, he believes that his information is definitely useful while, in effect, there is a chance it is not, then the sender will exaggerate in his disclosure, choosing more extreme messages." Nevertheless, overconfident senders may actually contribute to more informative communication. For example, if there is a small cost associated with broadcasting opinions, an overconfident sender is often more likely to be willing to incur the cost than a rational sender who realizes his information might not be useful to receivers.

In analyzing this particular interaction that has not been investigated before, Admati and Pfleiderer are preparing the groundwork for the next step of their research: developing models to apply to message boards and chat rooms related to the financial market, and perhaps examining the added dynamics of a sequence of reviews. "The Internet has dramatically changed the way investors and other economic actors can communicate. Understanding that process is critical to understanding how the financial market might behave in the Internet age," Admati concludes.