Antweiler and Frank (2004) study the frequency and tone of stock message-board
posts on Yahoo! Finance and Raging Bull about 45 large US stocks in the year 2000.
Their main finding is that message-board posting frequency positively predicts stock
return volatility and trading volume, even controlling for the frequency of WSJ stories.
The authors use an algorithm called Naı
¨
ve Bayes to classify posts as bullish, neutral, or
bearish based on the pattern of word occurrences. They report only weak relationships
between posting tone and market activity.
7
Tetlock et al. (2008) analyze the tone of firm-specific newspaper stories. In contrast to
studies of selected columns about the market, the authors analyze a comprehensive sam-
ple of WSJ and DJNS news ...