Are stock message boards worth reading? If so, what clues point to useful information? In their November 2005 paper entitled “eInformation: A Clinical Study of Investor Discussion and Sentiment”, Sanjiv Das, Francisco Martinez-Jerez and Peter Tufano examine relationships among on-line stock message board discussions and related news and stock prices. They further employ content analysis software to measure the intensity and dispersion (level of disagreement) of message board sentiment. They focus on 170,000 messages from four stock message boards (Yahoo!, The Motley Fool, Silicon Investor and Raging Bull) for four stocks chosen to represent extremes of stock information flow (Amazon, Delta Air Lines, General Magic and Geoworks) during July 1998 through January 1999. Integrating the message board content with contemporaneous news items, stock price movements and one interview of a frequent message board poster, they conclude that:
- Stock message boards offer: rapid dissemination of new information before the media widely covers it; some non-public information before companies announce it; substantial on-point exchanges (for about eight hours after news releases); and, many unsubstantiated rumors (that tend to die quickly), but little or no inside information.
- Message board discussion volume generally mirrors stock trading volume and relates positively to level of disagreement.
- There are a few frequent posters among a much larger number of occasional posters. (See chart below.)
- About 40-50% of messages convey a buy or sell sentiment, with the aggregate leaning modestly toward buy. This low level of net positive sentiment reflects a typically high level of disagreement. (See table below.)
- There is a close relationship between message board sentiment and news sentiment.
- The sentiment of message board posters does not predict stock returns. Sentiment rather tracks stock price, suggesting that aggregate message board thinking tends to be trend-following rather than contrarian.
- The interviewed frequent poster is a well-informed college graduate (engineering) who uses message boards to test his ideas regarding what professional stock analysts may be missing and how news stories may be misleading. This individual speculates that he became more stubborn about a stock as a result of taking a public (though anonymous) position on it, thereby degrading his investment performance.
The following chart, taken from the paper, shows the number of messages posted by screen name over the entire seven-month period. For example, one screen name posted over 5000 messages, and 8899 screen names posted only one message. It shows that a relatively few posters dominate the exchange.
The following table, extracted from the paper, shows the classification of all messages by stock for the entire seven-month period. “Opinion” is the percentage of messages that are either Buy or Sell. “Sentiment” is the number of messages that are Buy minus the number that are Sell. “Sentiment %” is Sentiment divided by the total number of messages. “Disagreement” is [Sentiment/(Number of Buy + Number of Sell) -1]. These results indicate that stock message boards tend to be neither mostly “pumping” nor mostly “bashing.”
In summary, stock message boards accelerate information dissemination and maybe analysis, but they do not reliably reveal new information (lots of opinions and little or no private information). Message board sentiment does not predict stock returns.
