U.S. President Barack Obama speaks at an election campaign rally in Concord, N.H., Nov. 4, 2012.JASON REED/Reuters
For months, pollsters correctly predicted that the U.S. presidential race would be close and by the morning of election day they were nearly unanimous in handing Barack Obama his victory.
But polls can be deceiving. Just ask the people of Alberta who were told in the week before they marked their ballots last April that the Wildrose party had a six– to 10-point lead – and then watched as the Progressive Conservatives won a majority.
Still, experts say, the recent survey trends in the United States were so clear and consistent in Mr. Obama's favour that it would have been nothing short of astonishing if Mitt Romney, the Republican, had managed to prove them wrong. So even if the final decision came late in the evening, it was far from a surprise. The pollsters got it right.
Ronald Rapoport, a professor of American politics who studies public opinion at the College of William and Mary in Virginia, said even though the divide between the two candidates in every poll was small and most were within the margin of error, the combination of multiple poll results was far more conclusive.
"The findings of consistency in those [surveys] certainly would lead you to expect a very likely win for Obama," he said. There were routes to a Romney victory, said Dr. Rapoport, but the Republicans needed to have things break so perfectly that the chances of a GOP win were severely limited.
A tracking model of national and statewide polling created for the Huffington Post by Simon Jackman of Stanford University suggested Tuesday that Mr. Obama had a 91 per cent chance of defeating Mr. Romney in the bid for his country's top job.
Nate Silver, the political numbers cruncher at the New York Times who accurately forecast the results in 49 of 50 states in 2008, aggregated a range of polling results to come up with the same odds.
And eight out of 12 major surveys conducted since Thursday of last week gave the edge to the incumbent president with two suggesting the candidates were tied. Only Rasmussen Reports and Gallup put Mr. Romney ahead – and not by very much.
Even at the state level – a better indicator of strength in the U.S. system than the overall measure of popular support – Mr. Obama was predicted to win key battlegrounds.
Richard Johnston, who holds the Canada research chair in public opinion at the University of British Columbia, says different polling firms have different methods, but people who aggregate polls like Mr. Silver and Mr. Jackman are able to weed out technological biases.
Since polling became a big part of elections, he said, the averages of polls conducted in the final week of a campaign have been very close to correct – and the closer to election day, the better the results. The polling industry would have had to be off by more than two points for Mr. Obama to lose, said Mr. Johnston.
"We haven't seen the industry be that far off in an election as close as this one over the last 50 years."