By Ira Basen

The most nervous people in Canada Tuesday will undoubtedly be the 1,601 men and women running for office. The second most nervous group will likely be Canada’s pollsters.

Over the past five weeks they’ve been paid hundreds of thousands of dollars by media outlets to try to capture the pulse of the nation. Hardly a day goes by without some new poll telling us how people are planning to mark their ballots.

But what if they’re wrong? What if the final results bear little resemblance to what the latest polls predicted? Wouldn’t that be embarrassing? Well, not really.

The snapshot

Pollsters are hard to embarrass. If you want to understand why, you simply need to think about the word “snapshot.” It’s a word pollsters love.

Our polls are simply “snapshots in time,” they tell us. They have no predictive value. A poll taken on October 10 will tell you what the electorate is thinking on that day. It will not tell you what it will be thinking on October 14.

But here’s thing, we have no way of knowing whether the poll on October 10 is an accurate reflection of voter opinion. The only time we can know for sure what voters are thinking is when actual ballots are counted on election day.

But of course pollsters don’t poll on election day. In this election, most polling companies hung up their phones on Sunday. If the numbers two days later don’t conform to the latest poll numbers, it’s not the pollsters’ fault.

Their polls were snapshots in time. Voters changed their minds at the last minute. Imagine that!

The most recent ‘oops’

The last time this scenario unfolded was the federal election on June 28, 2004. Based on the most recent polling data, The National Post proclaimed on June 22, “Harper Widens Seat Lead.” The Globe and Mail on June 25 proclaimed “Dead Heat.”

But when the votes were counted three days later, the Liberals had won 36 seats more than the Conservatives and had beat them in the popular vote by 7 per cent, well outside the margins of error. Oops!

But anyone looking for a mea culpa from Canada’s pollsters would have been disappointed.

“The possibility that all polls were wrong is highly unlikely,” declared Frank Graves of Ekos Research. Graves and other pollsters said there had been a significant shift in voter preference in the last weekend of the campaign, after the pollsters had left the field.

These same pollsters are saying it could happen again this year. People will be sitting around the Thanksgiving dinner table talking politics and many minds might be changed while the gravy boat is being passed.

So be prepared for the possibility the polls will be “wrong” again. Except, of course, they won’t really be wrong, because they were just snapshots. All in all, it’s a pretty sweet deal for the polling industry.

Clouds on the horizon

But that doesn’t mean that all is well in the polling industry. There are several things that will add to their nervousness on election day.

Such as cellphones.

The backbone of the polling industry remains the phone poll in which a random sample of people is selected from phone numbers listed in telephone directories.

But these days, between 10-15 per cent of the voting-age population have no land lines. And cellphone numbers are not recorded in directories.

That means this rapidly growing segment of the population, comprised mostly of young people, are out of the reach of pollsters.

Does that compromise the accuracy of a poll? It is a huge debate within the industry.

‘Sorry, nobody’s home’

When your caller ID shows a polling company is calling your home, do you pick up the phone? If you do find yourself on the line with someone conducting a poll, do you answer all the questions you are asked?

If you’ve answered yes to both those questions, you are part of a small and shrinking minority.

The measure that pollsters use is called the “response rate.” It consists of two parts: the “contact rate,” which is the percentage of sampled households that the pollster is able to reach during the course of the survey, and the “cooperation rate,” which is the percentage of people that complete the survey without hanging up.

In both cases, the numbers are low and getting lower.

It is not unusual for an overnight poll’s overall response rate to be in the vicinity of 10 per cent.

That means that for pollsters to find their desired random sample of 1,000-1,500 Canadians, they might have to make as many as 10,000 phone calls.

Are the opinions expressed by the small percentage of people who agree to be interviewed similar to the vast majority who can’t or won’t participate?

Pollsters say there is no indication that low response rates translate into less accurate results, but they are clearly worried by the declining number of people who will talk to them.

Of course, the poll-consuming public generally has no way of knowing what the response rates are. Pollsters tell us a poll’s margin of error, but they rarely reveal the response rate, perhaps because they are too embarrassed to do so.

A few years ago, pollster Angus Reid called response rates “the big dirty secret of the industry.”

Seat projections

And finally, beware of pollsters bearing seat projections. That’s when a polling company attempts to take its national polling numbers and make predictions on how those numbers will translate into actual seats won.

Pollsters routinely incorporate certain assumptions into their polls, such as how the undecided voters will be apportioned or what questions should be asked — leadership, most important issue? — before inquiring how a respondent intends to vote.

But when it comes to projecting seats, there is more art than science in these predictions and the numbers frequently turn out to be very wrong.

In 2004, the seat projections made by the pollsters were even more off base than the popular vote projections.

After the election, Frank Graves of Ekos Research declared that seat projections were “a mug’s game” and suggested his company might not be doing them anymore. This year, Ekos is one of the few polling firms still projecting.