A top professional horseplayer once told me that compared to handicapping horse races, elections were a cakewalk. "You're better off listening to a gambler than a pollster," he said, "Polls can be inaccurate. Gamblers are accurate or they lose."
There's considerable evidence that he was on the money. In 2004, the Gallup poll failed to forecast the winner of the popular vote, just as it had in 2000. Meanwhile, halfway through Election Day 2004, when a CNN poll showed Kerry taking the lead, the British site Betfair had Bush with a 91 percent chance to win.
"Prediction markets," according to Inkling Markets* CEO Adam Siegel, "are a reflection of all the information available and ask what people think will happen versus what they want to happen. In this way they can be more objective and efficient than polls."
This echoes the horseplayer's sentiments: "Gamblers have more experience with cheaters. They take voter fraud into their metrics. Polls don't. Nor do polls take into account intangibles like how each state's secretary of state factors in or systems within a state designed to eliminate voters."
Betfair also had all 50 states right in 2004. As did Irish site Intrade.
Such markets' successes are nothing new. In presidential races beginning in 1896, the New York Times, Sun, and World provided daily betting quotes; The newspapers' sources were betting firms that had reporters present at candidates' speeches to record audience reaction and gather additional intel. In the fifteen elections between 1884 and 1940, the betting firms missed just once, Wilson's 1916 upset of Hughes -- and they might have had a perfect record had they had stayed open long enough to take into account late-breaking news from the West.
The advent of polls ended an era. "Prior to Gallup's introduction in 1936, newspapers had little else to report about the election horse races other than the betting markets," says Koleman Strumpf, a University of Kansas economist who tracks betting trends. "When scientific polls came along, newspapers had something to report other than markets they were oftentimes uncomfortable with." Similarly uncomfortable states gradually eliminated election betting.
Internet wagering (as gambling on elections remains illegal within the United States, the bookmakers are overseas) initiated a debate about the relative merits of prediction markets. The debate has grown in intensity with advances in polling methodology. In the 2008 general election, while Intrade went 48 for 50 on states, poll-based FiveThirtyEight.com correctly picked 49.
Columbia's Robert Erikson, one of America's leading political scientists, says, "While it is always interesting to consult election markets, they generally are a less reliable guide than simply checking the latest polls." He offers five elections' worth of substantiating data in a Public Opinion Quarterly piece he wrote with Temple's Christopher Wlezien.
In the subsequent Public Opinion Quarterly, however, Wharton economist David Rothschild crunched the numbers to a different conclusion. As he told me, "Prediction markets include all of the polling information plus information that is not publicly available, so there is no reason that the prediction markets won't be as accurate or more accurate."
Currently, as it happens, the gamblers and pollsters are on the same proverbial page. "But where the polls show many races as being very tight," says Intrade analyst Carl Wolfenden, "our markets are showing more certainty regarding the outcomes." He pointed to California, where polls show Boxer leading Fiorina 46 percent to 43.5 percent; Intrade has Boxer up 76 percent to 24 percent.
"This difference is largely due to the different question being asked," Wolfenden explains. "The poll number tells us that 46% of those sampled plan to vote for Boxer, while the Intrade market tells us there is a 76% probability that Boxer will win."
Intrade shows an 88.2 percent chance the GOP will takeover the House, with a 71.9 percent chance the Republicans will take 227 or more seats. The betting shows a 58.5 percent probability of 232 seats and a 37 percent probability of 237 seats -- a 60-seat swing.
Also of note, the probability of the GOP winning 51 seats to take control of the Senate is a low 15.1 percent. For gamblers' takes on specific races, search here and see realtime charts below.
width="255" height="145" frameborder="false" scrolling="no" marginheight=" 0 " marginwidth="0"> width="255" height="145" frameborder="false" scrolling="no" marginheight=" 0 " marginwidth="0">
As Strumpf says, "On Election Night I'll look at the movement on the betting sites to see what's going on. I watch CNN too, out of the corner of an eye, but it's not necessary."
*Beginning today and continuing through Election Day, Inkling is running a free prediction market focusing on big questions such as which party will take Congress and the Senate, and on Governor races that polls are saying are too close to call." To participate, visit elections2010.inklingmarkets.com. Leaders will be posted and if you finish on top, you'll have an opportunity to share your methodology. Or it might pay -- come the next election -- to keep it to yourself.
Follow Keith Thomson on Twitter: www.twitter.com/kqthomson