Can a sentiment survey of a bunch of individual investors actually predict stock price performance? Crowd Technologies—a neophyte online investing community whose Web site Piqqem enables people to vote on the direction they believe individual stocks will follow—has begun to collect data that shows a correlation between changes in sentiment over short periods of time and the propensity for earnings to outperform or underperform market forecasts.
Piqqem is based on the concept of the wisdom of crowds, the belief that a sufficiently big and diverse group of people will over time be more accurate in their estimates and predictions than even the most talented individuals. Jett Winter, Piqqem's chief executive, has a background in software and has headed multiple startups in Silicon Valley over the past 25 years. He also runs an investment bank, Winter Advisors, and is interested in the trading opportunities that will come out of crowdsourcing stock sentiment.
During the second quarter, Winter chose eight stocks that had reached a critical mass of activity and tracked changes in sentiment in each over the course of the quarter until the day before each company's earnings release. Feelings about all but McDonald's (MCD) predicted same-direction earnings surprises. But what he found most interesting were two stocks—Amazon (AMZN) and Microsoft (MSFT)—whose sentiment scores dropped as the earnings dates neared, even as the market continued to push their prices higher in anticipation of better-than-expected results. Both companies missed analysts' forecasts, and their stock prices fell after reporting earnings, showing that the crowd's intuition had been more accurate than the market's.
Wall Street analyst ratings are addedWinter plans to expand from 8 to 20 or 25 stocks for his third-quarter analysis. After another quarter's worth of data, he expects to be able to start simulating trades. "We would buy a stock early on and sell it later [after the stock had appreciated with sentiment], or short it, and see how crowd sentiment is becoming a trading tool."
On Piqqem, users can vote by stock direction or can get more specific by choosing a price they believe a stock will reach by a given option-expiration date. The latter data gets factored into the site's sentiment readings—which are chosen as single or double up, neutral, or single or double down—and measure how strongly a user feels about a stock. Besides user votes, Piqqem pulls in Wall Street analysts' ratings from the Yahoo Finance web site as they are adjusted.
In his book The Wisdom of Crowds, James Surowiecki identifies four conditions that characterize wise crowds: diversity of opinion, independence (individuals aren't being influenced by others' opinions), decentralization (users share specialized and local knowledge), and aggregation (a mechanism for turning individual judgments into a collective decision). The errors each person makes in arriving at an answer will cancel one another out, leaving only the information component of each person's guess, Surowiecki writes in the book's opening chapter.
Because independence is a priority, Piqqem users aren't privy to how other people voted or the identities of the most popular stocks until after they have voted. Once participants have voted, however, they can revote as many times as they like, with only their most recent opinion counted in a stock's sentiment score.
twentysomething techies on two coasts"The reason we allow revoting is because a vote has a life," Winter explains. "We believe people gain additional knowledge and insight and can [change their mind]." Each vote stays active for 90 days, and voters are alerted via e-mail when their vote is about to expire so they can renew it.
Piqqem's active voters tend to be concentrated on the East and West Coasts, especially in the San Francisco Bay and New York areas. The top professions among users are engineering and finance, while the dominant age range is 18 to 29. Winter plans to roll out a revenue model based either on subscriptions, advertising, and/or syndication to brokerages and financial-content sites within the next three to four months.
Winter is eager to make the data as statistically relevant as possible, which he defines as a critical mass of at least 50 votes per stock over time. To make it faster and easier, Piqqem offers lists of similar stocks whenever you vote on a specific stock. The aim is to keep people coming back and staying engaged.
With about 2,000 active voters, Piqqem has to grow considerably to catch up with PredictWallStreet—which pulls in tens of thousands of predictions every day through partnerships with TD Ameritrade (AMTD), the Broadridge Investor Network, and Zecco—and Motley Fool CAPS, which currently has 150,000 people rating stocks, including about 180 financial institutions.
CAPS, which launched in October 2006, now rates 5,500 stocks that must have a market cap of at least $100 million, trade no lower than $1.50 per share, and have at least 20% of their opinions come from All-Stars—users with superior stock-picking track records. Weighting votes according to users' track records for accurate picks and returns in excess of the Standard & Poor's index of 500 stocks "enables us to calculate a community score for each stock and we can do a relative comparison that puts them into quintiles from a one-star [rating] to five stars," says John Keeling, general manager of Motley Fool CAPS. The weighting method corrects for what he calls a "positive selection bias" among relatively inexperienced and overly optimistic investors, and is important given that "we ask the community to become an analyst and rate as many stocks as it can," he says.
Piqqem's Winter: superstars crashInvesting site Covestor has migrated to a wealth management model through which users can, for a fee, replicate the trades of certain investors or professional asset managers whose allocation strategies they agree with—or who have proven track records of picking winners. Winter at Piqqem says he's not interested in identifying superstars for people to follow, because they are apt to crash and burn after a long hot streak.
It's not clear yet if investors should be making real trades based on Piqqem's data. Winter sees actionable ideas emerging as the site's managers zero in on specific data points to track, such as earnings surprises. As the data set increases, Piqqem plans to hire quantitative analysts to look for correlations around the sentiment readings, which may include whether sentiment is a leading or lagging indicator and if it's more relevant to the performances of large-caps than smaller companies.
Zack Miller, a one-time hedge fund analyst and former head of business development for investing site SeekingAlpha, is including a section on Piqqem in a book he's writing about the new rules for investing. "If it can get critical mass with the way they've built it, I think there will be some interesting analytical tools" that emerge eventually, he says.
Miller believes that rather than transform how people invest, Piqqem's sentiment readings will be one of many inputs investors consider when deciding which stocks to add to or remove from their portfolios. "It's something I would pay to get access to, to add to that investment calculus," he says. And just as CAPS incorporates CNBC stock guru Jim Cramer's stock picks into its data set, Miller says he would, too. "I may not agree with Jim Cramer, but it's important to me to know what he's saying about a particular [stock]."
is the market too big to be wise?For now, the demographic breakout that Piqqem provides for stock sentiment is fairly basic—limited to gender, age, and professional categories. But Miller sees potential for more specific demographic analysis, such as what engineers or other professionals living in northern California think of tech stocks such as Apple, which might be more predictive of earnings strength. However, generating data of that caliber will require a much larger and more diverse data set than Piqqem currently has, he adds.
Winter's response to skeptics who say that the wisdom of the crowd is already reflected in a stock price is to respond that the stock market reflects the disproportionate influence of big-block traders and analysts and is more a clearing house for information than a predictive market. Other critics argue that votes should be weighted according to how much money users have actually invested in a given stock because people are less likely to try to game the system when they have more to lose if they're wrong.
Dr. Craig Kaplan, chief executive of PredictWallStreet, believes there may an advantage in simply asking people what they think instead of putting more faith in the actions they take. He's found that people tend to be more willing to tell you what they think about a stock than to trade it. "We're able to get a read on what people are feeling before it crosses their threshold where they're ready to act," he says. "If you wait until somebody has already acted, how much edge do you have on the market? The market is a pretty efficient mechanism, but it can only process information put into it through trades."
It's too early to tell if Piqqem will remain beholden to the collective wisdom of the crowd or—like rivals—start to place more credence in the opinions of its superstars. Perhaps this crowd is smart enough to know that sentiment data should not be the only tool in buying and selling stocks.