Pollsters are celebrating the uncanny accuracy of some of their predictions ahead of this week's federal election, which represented a much-needed credibility boost for an industry that is still recovering from high-profile miscalls in recent provincial campaigns.
Nanos Research, which did nightly tracking for The Globe and Mail and CTV News, was remarkably on target throughout the campaign, recording the nuanced shifts as the Liberals moved from third to pull ahead of the NDP and Conservatives.
"There's right and almost right and perfect," said Nik Nanos, who thinks his company's results couldn't have been much closer to the mark.
The final Nanos poll pegged the Liberal vote at 39.1 per cent; Elections Canada recorded it as 39.5 per cent. The Conservatives were polled at 30.5 per cent and the final result was 31.9 per cent. The Nanos poll had the NDP at 19.7 per cent, which is exactly what the NDP got according to Elections Canada results.
"I was kind of excited we pegged the NDP support exactly to one-tenth of a percentage point. That doesn't happen very often," Mr. Nanos said.
"Some people might say that's luck, but the reality is you can't have that level of precision on a regular basis and say it's luck."
Mr. Nanos said the polling was accurate for several reasons. First, Nanos Research did nightly tracking, instead of jumping in for occasional snapshots. And secondly, the company polled on Sunday, just 24 hours before Canadians voted. That polling, he said, caught voters who made up their minds at the last moment.
Mr. Nanos said another reason for the accuracy was that Nanos Research relied entirely on random telephone calling to voters, not online surveys or other new technologies that are faster and cheaper but less accurate. "Doing random telephone surveys with live agents … is the most expensive but the most mature methodology out there," Mr. Nanos said.
But the failures, such as the miscalls in recent provincial elections in B.C. and Alberta, are what people remember, University of Calgary political scientist Paul Fairie says. "I think polling generally does very well," Mr. Fairie said Tuesday.
"There are many more successes [than failures by pollsters], but why would you remember something like that?" he asked. "The problem is it's a lot easier to remember the failures. Everyone remembers B.C. in 2013 and Alberta in 2012 but for most of the other elections, polling was within a couple of percentage points of the final results, which is the best you can hope for."