Quotes of the Day

Long before Australians knew the date of the federal election, they were betting on its outcome. Centrebet, one of the country's biggest bookmakers, had taken $300,000 in wagers by Aug. 27; another $30,000 came in last weekend. Expat Justin Wolfers, a professor of business at the University of Pennsylvania's Wharton School, has a special interest in where the money is going. That's not just because he's backed Labor - currently at around $2.30, which he says reflects a 38% probability of victory. Wolfers believes that by the end of the campaign betting prices will reveal, more accurately than opinion polls, which party is going to win.

That's no mere fancy. In a study of the 2001 election, Wolfers and economist Andrew Leigh, of the Australian National University, found that while pollsters differed on the outcome and the winner's vote share, punters' money was solidly on the government. And in marginal contests, the favorite won in 43 seats out of 47. Polls are certainly useful, Wolfers says. But "the media would serve the public better by reporting on betting rather than polls."

Early last century, before opinion polls were invented, the U.S. press tracked the ups and downs of a huge political-betting market centered on Wall Street. According to University of North Carolina economist Paul Rhode, bettors were remarkably prescient: in 12 presidential elections between 1896 and 1940, the underdog won only once. Today Americans can "wager" on elections online through the Iowa Electronic Markets, a small-scale futures exchange that lets people buy contracts on candidates based on their estimated chance of victory. (At press time, George W. Bush led John Kerry 54% to 46%.) Since its 1988 launch, the I.E.M. has outpredicted election-eve polls by an average of 0.5 percentage points.

Why should markets be better forecasters than polls? Economist Friedrich von Hayek argued that markets efficiently gather information held by widely dispersed people, summarizing it in the form of prices. In speculative markets, traders and bettors are rewarded only if they correctly predict future prices. That, Wolfers says, motivates them to get the best information they can, ignoring trivia and trends. "The market seems pretty highly attuned to news that affects the election outcome," Wolfers says. "That's not always the same as what the papers report. The market is unlikely to react to things like party conferences and policy speeches. But it will react to a change of tactics or a new campaign manager, which may not influence opinion polls."

Election punters and pollsters are both trying to figure out how the nation will vote. But when punters make that call, says Leigh, "they're putting their money where their mouth is." Pollsters must rely on what voters tell them. That can lead to error, says Leigh, because "you're sampling only a fraction of the population" and "people might say anything just to get rid of the interviewer." Overall poll results - percentages of the total vote - may not correspond to numbers of seats won. And in marginal seats, on which many elections turn, the margin of error can exceed the margin of victory. In '01, the betting markets' predictive power in these seats "really surprised us," says Leigh. "But the people who were betting on those seats knew them very, very well."

Markets aren't infallible, of course: Australian punters failed to predict the upset wins of Paul Keating in 1993 and Steve Bracks in 1999. And I.E.M. traders wrongly predicted in 2000 that Bush would win the popular vote as well as the presidency. In early July, Centrebet manager Gerard Daffy had several calls from people wanting to bet on the date of the election. He didn't open a book "for obvious reasons," he says. If he had, punters would have had another miss: "The date they were looking for was Sept. 18."

• Elizabeth Feizkhah
• Long before Australians knew the date of the federal election, they were betting on its outcome