Skip to main content

The Globe and Mail

Toss-up or slam dunk? Polls struggle to predict election outcome

U.S. President Barack Obama points to the crowd during an election campaign rally at McArthur High School in Hollywood, Florida November 4, 2012.


With national support for Democrats and Republicans so evenly split, political junkies of all stripes are captivated by the torrent of predictions spit out, hour by hour, by a long list of pollsters. But, despite the daily deluge of data, clarity remains elusive.

Over the weekend, the well-known poll aggregator Nate Silver of The New York Times said that the likelihood of president Barack Obama being re-elected had increased to more than 85 per cent. Meanwhile, other polling companies, such as the firm owned by FOX News contributor Scott Rasmussen, say the race is too close to call. A Sunday poll by The Wall Street Journal and NBC News found that Mr. Obama maintained a slim 48 per cent to 47 per cent lead over Mitt Romney. This after Republican strategist Karl Rove wrote, only four days earlier in The Journal, that a survey of national polls showed Mr. Romney had a slight but persistent edge.

How can such a supposedly scientific process produce so many different results?

Story continues below advertisement

Although partisan pundits are fond of quickly playing the bias card against polls that don't go their way, disparity in polling almost always comes down to method, said Nik Nanos, founder of Canada's Nanos Research.

"Many of the pollsters in the United States are unfairly criticized, and people tend to project onto them," he said. "It's not really partisanship at play ... it's a bit unfair."

For instance, unlike many other recent polls, Mr. Rasmussen's firm, Rasmussen Reports, is showing a near tie, if not a slight edge, to Mr. Romney in the all-important battleground state of Ohio. Democratic supporters like to point to Mr. Rasmussen's history of appearing at Republican events, or what analysts say is a historical trend of Republican biases in his results. But the truth, Mr. Nanos said, probably lies in the nitty gritty of how Mr. Rasmussen produces his numbers – the fine print on his company's website that doesn't translate well into headlines.

Rasmussen Reports relies on Interactive Voice Response, what is often referred to as robo-calling – the least expensive, most efficient method of reaching a lot of households very quickly. But the method has its drawbacks, not the least of which is that its very difficult for a computer to verify whether they are on the line with John Smith, 45-year-old homeowner, or Mr. Smith's seven-year-old, troublemaking son, Timothy. Also, Rasmussen Reports does not call cellphones, which prevents it from reaching a significant – not to mention younger – percentage of the population that don't use land lines.

And then there are the poll aggregators, such as Mr. Silver, who rank and weigh national pollsters according to their historical track records, before plunking recent results from those pollsters into a complex algorithm in order to come out with an average. It's an effective way of covering the country, but they're beholden to the companies supplying the raw numbers. By using averages, they can be consistently accurate, but it's more difficult to nail the precise vote tally because, as Mr. Nanos says, "one rogue poll" can skew their prediction.

But Mr. Silver came close to perfection in the 2008 presidential election, when he accurately predicted results in 49 of 50 states, an achievement that has made him a household name and, in some circles, a defending champion who needs to be unseated. Most recently, a conservative blogger attacked him for being "effeminate" and for having a "soft-sounding voice."

Report an error Editorial code of conduct Licensing Options
As of December 20, 2017, we have temporarily removed commenting from our articles. We hope to have this resolved by the end of January 2018. Thank you for your patience. If you are looking to give feedback on our new site, please send it along to If you want to write a letter to the editor, please forward to