To value firms and price stocks, financial economists and accountants typically rely on “hard information,” looking at a firm’s fundamentals through the lens of balance sheets, income statements, and related quantitative information.
While undeniably useful in assessing firm value, the quantitative focus may have limitations, says Professor Paul Tetlock. “Firms like to portray themselves in a favorable light. And analysts may have conflicts of interest and other human imperfections,” Tetlock says. “If so, qualitative data that isn’t represented in quantitative analysis may also be informative about firm value.”
News stories presented a natural testing ground for Tetlock to explore this theory. Does so-called “soft information,” as conveyed by the language used in news stories, uncover relevant market information that standard tools of financial analysis do not? Could soft information help shed light on firm value, or predict events such as a firm’s positive earnings announcement?
Tetlock worked with Maytal Saar-Tsechansky of the University of Texas at Austin and Sofus Macskassy of Fetch Technologies to use textual analysis to convert this soft information into hard information. The researchers first examined all Wall Street Journal and Dow Jones Newswires stories for references to all firms in the S&P 500 index between 1980 and 2004. Then, they matched the references back to corresponding firm codes in databases containing stock price, accounting, and analyst forecast data.
When the researchers combined these data with their simple counts of positive and negative words in each story, they found that the market tended to respond more to negative words. The researchers hypothesized that soft information could be indicative of a firm’s future profitability. To test this, they examined firms with many negative words written about them in the month prior to their quarterly earnings announcements: these firms’ profits fell short of stock analysts’ forecasts and were lower than their profits from the same quarter in the prior year.
“In theory, analysts should have already picked up on the news and acted on it,” Tetlock says. “The analyst forecasts came at points in time when the analysts had an opportunity to respond to all the negative words in the news. Despite that, analyst forecasts were too high relative to a statistically optimal forecast that took the news into account.”
So the market reacted to negative news stories, but did it react too much, too little, or just right? Were investors taking the information available in these stories into account when pricing a firm? “We found that they do, to a degree,” Tetlock says. “But the one-day market reaction to a story seems to be insufficient, in the sense that stock prices of firms continue to fall on the next day and even a little bit thereafter.” This underreaction lasts only a week or so, suggesting that the market corrects itself fairly quickly.
To try to better understand whether the market really was underreacting to the value-relevant information in these news stories, the researchers then compared news stories that specifically mentioned firm earnings to stories that did not mention earnings at all. Perhaps not surprisingly, the researchers found that earnings stories are better predictors of firm earnings and elicit stronger market responses, implying that such stories are more relevant for valuation.
Yet the researchers also discovered that the degree of market underreaction to earnings stories is even greater than the degree of underreaction to other stories. “It is as if the market can tell the difference between value-relevant and value-irrelevant news, but it doesn’t fully appreciate just how important value-relevant news is,” Tetlock says.
Nor does the market fully appreciate just how irrelevant value-irrelevant news is, according to follow-up research conducted by Tetlock. “News about current events often unfolds over a long timeframe,” he explains. “For somebody who hasn’t been following the firm constantly, it can be challenging to sort out what’s old and what’s new. So stories often refer to old events to give context for readers to interpret a given news event.” Essential though it may be, older news is decidedly less relevant — or should be — for the market.
To assess whether investors might confuse old news with new news, Tetlock conducted further research that measured the degree of textual overlap between a news story appearing on a given day and the prior ten news stories for that same firm, expecting that a story with more textual overlap with older stories would also contain more old information. Tetlock found that the market responded less strongly to stories containing a lot of textual overlap with previous stories, implying that the market did recognize that such “stale stories,” as Tetlock called them, were less value-relevant. Notably, despite the lower degree of reaction, the market still overreacted to stale news: firm returns were high on a day when positive stale news appeared, but tended to be low in the following week, suggesting that the market was correcting for its initial overreaction.
An investor could use a relatively simple portfolio trading strategy to capitalize on either the market overreaction to stale news or the underreaction to relevant news. To bet on the underreaction, one could buy shares of firms with the fewest negative words in news stories and sell shares of firms with the most negative words on the day the news comes out. In a simulation, Tetlock, Macskassy, and Saar-Tsechansky waited to buy at the close of the trading day when negative news appeared, resulting in 20 percent annualized returns. But trading costs are the catch: the strategy calls for daily trading, which could lead to extremely high costs. Investors who are able to accurately interpret the tone of the news and keep the cost of executing trades down could net considerable profits from this strategy. Less ambitiously, investors could use the tone of news to help them time their already-planned purchases and sales of stocks, potentially saving a bundle of money.
Although these specific strategies are very short-term, Tetlock notes the possibility that the same principles apply to long-term market reactions to firm news: the market may not fully distinguish value-relevant and -irrelevant news. “A discerning investor has a greater potential to profit from interpreting the news by really digging into what’s behind the negative news and its implications for firm value,” Tetlock notes. “In that sense, the paper’s message is quite consistent with the ideas of Warren Buffett.
“In our complex modern economy, it’s tough for investors to distinguish value-relevant and -irrelevant news. They qualitatively get it right but quantitatively don’t fully appreciate the distinction,” Tetlock says. “As long as others aren’t appropriately responding to the news, investors can find value in analyzing it more carefully.”
Paul Tetlock is the Roger F. Murray Associate Professor of Finance in the Finance and Economics Division at Columbia Business School.
Professor Tetlock's research interests include behavioral finance, asset pricing, and prediction markets. One area of his research examines how firms' stock market prices respond to the content of news stories. His 2007 Journal of Finance study on the impact of negative words, such as "flaw" and "ruin", won the Smith-Breeden Prize for the best article in asset pricing. His research...
Read the Research
"All the News That's Fit to Reprint: Do Investors React to Stale Information?"