Forget about the opinion polls; follow the money

Gambling on presidential elections is hardly new. As a recent paper by Paul Rhode and Koleman Strumpf, both economists at the University of North Carolina, Chapel Hill, shows betting on the presidential elections dates back to George Washington's time. In the early 1900s, newspapers published daily quotes of the odds. Betting volumes peaked in 1916, with punters betting more than twice the amount of money spent on the campaigns that year. In the elections they studied, the odds nearly always picked the winner.

Cyberspace betting on political events has flourished since the last American presidential election, in 2000. Websites such as www.betfair.com, based in Britain, and www.tradesports.com, based in Ireland, let gamblers back and lay odds on their favourite candidates for a host of American election contests with as much money as they want. An older exchange (and the only legal one based in America, where gambling over the wires is forbidden), the Iowa Electronic Markets (IEM), sponsored by the University of Iowa, has been letting selected experts bet modest amounts for years.

Relying on such markets makes sense in many ways. An opinion poll gauges sentiment on the day it is taken; the betting exchanges are trying to predict a future event, so are often less prone to over-react to the news on any given day. People's responses to polls depend a lot on how the questions are phrased, how the sample of voters was taken and whether they will actually vote. Betting exchanges, more vitally, test how far people are prepared to risk their hard-earned money.

The markets have not ignored the polls. Mr Bush's odds of victory fell dramatically after his first debate with Mr Kerry. But Leighton Vaughan Williams, director of the Betting Research Unit at Nottingham Trent University in Britain, points out that Mr Bush's biggest single-day slide, which took place on the day of the second debate, may have had an economic reason: his number slid hours after disappointing jobs figures were announced and before the debate started. Punters, in other words, are perhaps putting more weight on the economy than on politicians' words.

In the period after the second debate, Mr Bush barely managed to scrape back to a 60% chance of victory, and has fallen from there in recent days, pushed down in part by poor consumer-confidence figures. On October 28th, as The Economist went to press, Mr Bush's odds of winning had slumped to 56%.

Online exchanges, to be sure, are not perfect. Mr Vaughan Williams notes that IEM's odds were far off the mark in the 2000 election, predicting that Mr Bush would win a larger fraction of the popular vote than he did. Moreover, IEM failed to predict the Republican victory in the Senate in the mid-term elections in 2002. He thinks that this is perhaps because Iowa's markets keep bets to a small size, which puts less pressure on punters to get their predictions right. But don't bet on it.