TCS Daily

'Ask the Crowd'

By James K. Glassman - June 3, 2004 12:00 AM

At this point, presidential polls are volatile and meaningless. In January, the Gallup survey had John Kerry leading George Bush, 55 percent to 43 percent; in March, Bush was ahead by precisely the same margin. Gallup's latest survey, on May 25, has Kerry leading, 49 percent to 47 percent.

There's another way to predict the winner, and, while it's not foolproof, it's proven far better that polling. Don't ask individual Americans for whom they'll vote if the election were today. Instead, ask traders on a specialized exchange -- similar to a stock market -- for whom America will vote on Nov. 2.

One such exchange is called the Iowa Electronic Markets, founded in 1988 by the University of Iowa College of Business. Anyone can trade by betting money (up to $500) to predict the outcomes of events like governor's races and Federal Reserve policy. Results have been uncannily accurate.

Academic papers have found that the IEM, on average, has whipped the polls soundly. For presidential elections, the IEM's margin of error on the brink of the vote was just 1.5 percent, compared with 2.1 percent for Gallup. Three-fourths of the time, the IEM has been more accurate than the average pollster.

In one of the Iowa contracts, traders predict the share of the vote that Bush will receive against Kerry. Winners receive Bush's final vote percentage on Nov. 2 in cents, so this market provides a good gauge of who traders think will win, and by how much. The latest quotation shows Bush leading Kerry, 51 percent to 49 percent.

While polls show what people think now, the Iowa markets show what traders, with their own money at stake, think will eventually happen. "Though ever-changing," the IEM forecasts, writes James Surowiecki in "The Wisdom of Crowds," his superb new book, "are considerably less volatile [than polls] and tend to change dramatically only in response to new information. That makes them more reliable as forecasts."

For example, in contrast to polling, IEM trading has put Bush ahead of Kerry consistently since February. Bush has declined slowly from a peak of 53 percent since mid-April and lately has leveled off. (A contract set to debut in June will pay $1 to bettors who predict the winner of a head-to-head Bush-Kerry contest; watch this one closely.)

The IEM is what economists call a "prediction market" -- where participants bet on political and economic events. Such markets are flourishing on the Internet. One,, currently provides wagering on whether Don Rumsfeld will resign by June 30 (current chances, just 7 percent, down from a high of 32 percent in early May), who will host the Olympics in 2012 (the betting is on Paris) and who will be Kerry's running mate (John Edwards leads, with his stock sharply higher in the last month)., a "pseudo-market," without real money at stake, offers contracts on whether Osama bin Laden will be caught this year (31 percent chance, according to current market prices) and whether Sharon will outlast Arafat (35 percent). A similar market, has betting on gas prices hitting $3 a gallon (22 percent chance right now) and on suicide bombers hitting the U.S. this year (20 percent).

In 2000, the director of the quantum computing program of the National Science Foundation encouraged the Pentagon's think tank, DARPA, to study the usefulness of prediction markets in assessing geopolitical risk. In July 2003, press reports surfaced about a government-sponsored market that could allow betting on events like bioterrorism attacks, military casualties and assassinations. The uproar in Congress, headed by Luddites like Sen. Byron Dorgan (D-ND), led to abandoning the project.

That was a shame because, as Surowiecki shows clearly, well-constructed markets are better predictors than experts. "We should," he writes, "ask the crowd," whose collective judgment tends to be better than the judgment of its smartest participant. In a classic example, the average guess, made by 100 participants, on the number of jelly beans in a jar is typically better than the best individual guess.

In a new paper, Justin Wolfers and Eric Zitzewitz of Stanford see a bright future for prediction markets. Already, they are being used by businesses to predict movie receipts and computer sales. Among the clients of NewsFutures are Eli Lilly, the giant drug company, and Dentsu, Japan's top ad agency.

So, as the election approaches, the smart strategy is to tune out the polls and check out the markets.


TCS Daily Archives