Why Trading on News Stories Could Actually Be Profitable

An interesting thing typically happens to a company's stock after it announces an earnings surprise, either positive or negative: over the next couple of months the stock tends to move in the same direction -- up for beating, and down for missing, analysts' expectations -- even though no new substantive information has come out.

The phenomenon, dubbed the Post-Earnings Announcement Drift, is a spur in the side of efficient market proponents since, theoretically, a stock price is supposed to convey all available information about a company at any given time. But the theorists don't quite have an explanation for the drift. Even the father of the Efficient Market Hypothesis, Eugene Fama, has voiced support for it.

And trading on the drift can be fruitful strategy with one recent study finding that a portfolio which took long-positions in good-news stocks earned abnormal annual returns of 7.5 percent.

The drift was first documented in the late 1960's, and an extensive body of research has tried to explain its existence. Much of this has been focused on the hard numbers included in an earnings release, but only recently have researchers turned to another valuable source of information: the not-easily quantifiable news generated around an earnings announcement, namely things like news stories, conference calls, and interviews.

In a new study, Joseph Engelberg, a PHD candidate at Northwestern University's Kellogg School of Management, looks at Dow Jones Newswire stories surrounding earnings announcement between 1999 and 2005 to see if they can explain some of the post earnings drift.

To find out what kind of information is conveyed in these new stories, Engelberg set up an algorithm that matched negative words as classified by the Harvard Psychological Dictionary with words in the headline and lead paragraph of earnings stories. The idea here being that a story with a lot of negative words is communicating bad news about a company's prospects. (There is too much noise associated with positive words that make them harder to investigate.)

Engelberg found that the occurrence of negative terms did seem to do a good job of predicting the post earnings drift. A portfolio of stocks consisting of long positions on stocks with no negative words and short positions on stocks with the highest level of negative words earned one percent per month abnormal returns.

The effect was more pronounced for the technology sector and companies with high research and development budgets. This makes sense in that the information these types of companies have to offer is harder to convey in the numbers of an earnings announcement.

Why might it take longer for markets to process the non-numerical information? Soft information is more nuanced and takes more time consume, and likely costs more to process, argues Engelberg.

The kinds of soft information hardest to digest are references to positive fundamentals and to a company's outlook. Analysts, for their part, don't do a great job of incorporating this type of information into their earnings forecasts, and "this suggests a possible channel by which this information fails to get into prices," writes Engelberg.

The findings have two interesting implications. First, prediction markets have recently been criticized for not reflecting information quickly enough. This from NYT's David Leonhardt:

After the drug maker Schering-Plough reported strong earnings on Tuesday, for example, its stock price jumped. But the stock is unlikely to continue soaring in coming weeks. The market has already adjusted to the news.

On Intrade, such reactions often happen in slow motion -- and eventually turn into overreactions. Mr. Obama's stock rose for days after he won Iowa, then fell during the two weeks after he lost New Hampshire and rose again in the 10 days after he won South Carolina. The impact of each contest took surprisingly long to sink in.

But Engelberg's findings suggest an alternate explanation: Elections might throw off a lot of hard-to-digest information that takes time to process and be reflected in Intrade prices.

Second, Reuters and others have been offering products which allow for trading on news stories. Engelberg's method for identifying the negative/positive-ness of a story is admittedly crude, but enterprising hedge funds may have -- or more likely, already have -- found more sophisticated methods that show that trading on the backs of journalists is a profitable strategy.