Faculty Researcher Richard Zeckhauser, Frank P. Ramsey Professor of Political Economy; Christopher Avery, Roy E. Larsen Professor of Public Policy and Management;Harvard Kennedy School Paper Title The CAPS Prediction System and Stock Market Returns Coauthor Judith Chevalier, School of Management, Yale University
“The theory of efficient markets is the centerpiece of modern financial research,” assert Christopher Avery, Judith Chevalier, and Richard Zeckhauser. The theory holds that the prices of stocks and other assets automatically incorporate all available information and rapidly adjust to incorporate new information. This makes it virtually impossible for an individual — amateur or professional — to consistently outperform the stock market (in terms of predicting future returns) since the prices for a given stock already reflect any information that he or she might rely on. One might do better through sheer luck for a short while, but not over the long haul.
“This theory is extremely well established,” Zeckhauser notes. “It’s what they teach at Harvard Business School.” That’s why he and his coauthors are intrigued by a stock price prediction system called CAPS that seems to beat the market on a regular basis. “If this system really works, it would mean the efficient market hypothesis cannot be true,” Zeckhauser adds.
So what is this CAPS that threatens to topple a linchpin of contemporary finance theory? For starters, it’s a web-based system run by The Motley Fool, a financial services company. (CAPS is not an acronym, although many possible longhand versions — including “Crazy Amateurs Pick Stocks”— have been suggested.) Tens of thousands of individuals regularly feed information to this system in the form of recommendations for and against particular stocks, with more weight given to the predictions of players with proven track records. All these recommendations are tallied (using a proprietary algorithm), resulting in daily rankings of companies whose stocks are accorded anywhere from one (lowest) to five (highest) stars.
Avery, Chevalier, and Zeckhauser analyzed a year’s worth of data (from November 2006 through October 2007) provided by The Motley Fool, encompassing more than 1.2 million stock picks made by more than 60,000 individuals. Following those recommendations, by buying all five-star-rated stocks and selling all one-star-rated stocks, would have yielded a whopping 18 percent return on investment over the course of the year, the researchers found — a level of performance that’s completely at odds with standard theory.
How could something like this work? CAPS, Zeckhauser explains, is an example of “collaborative filtering in which large numbers of people, who are linked by the Internet, have scraps of information that can be aggregated together in ways that might be important.”
“There have been lots of studies of entries posted on Internet stock message boards, but none of them found strong evidence that the multitude of messages posted on those websites actually contain new information,” Avery notes.
The greatest mystery about CAPS — apart from the fact that it seems to work — is figuring out what information these 60,000 or so people might possibly have. At present, Zeckhauser admits, “We have no idea, nor do the folks at The Motley Fool, as to the basis on which people are making their recommendations.” The apparent success of this approach suggests that markets aren’t necessarily incorporating all relevant information. “It appears that lots of information is not being aggregated, and this is one way of doing it,” he adds.
The team plans to extend its analysis to find out, for instance, whether the system works better or worse as more people participate. The researchers will pay particular attention to how CAPS fared during the financial meltdown of late 2008 and early 2009, looking specifically at recommendations for (or against) stocks that tanked during this tumultuous period. The potential exists for a system like CAPS to serve as a kind of “early warning system,” alerting us to financial trouble — within individual companies or an industry as a whole — that might otherwise go unnoticed. CAPS or something like it might also provide valuable information to regulatory agencies and corporate overseers.
“We’ll just have to see what the data turns up,” says Zeckhauser. “I try not to predict things like that.” — Steve Nadis