This week: How did the pollsters and pundits get it so wrong in B.C.?
On Tuesday, the B.C. Liberals stunned political watchers and pollsters by pulling out a majority win in the province's election.
When the campaign started polls, showed the Liberals trailing the NDP by 20 points. But while that gap closed significantly throughout the course of the campaign, polls on Friday still put the Liberals behind the NDP by six to nine points.
In the end, the Liberals won 50 seats, with 44.4 per cent of the vote, giving them a majority government.
The NDP came second with 33 seats and 39.5 per cent of the vote. The Green Party won one seat and 8 per cent of the vote and one independent MLA was elected, with 2.4 per cent of the vote. The B.C. Conservatives failed to win any seats, but received 4.8 per cent of the popular vote.
"This is a big head slap for the pollsters that were polling in British Columbia," Nik Nanos said.
Nanos said there were a number of things that came into play and it's clear the polling that was done did not fully capture the "Liberal uptick."
"In our experience in the past, there's a certain segment of voters that are kind of like last-minute shoppers," Nanos explained. (Nanos Research did not do any polling in British Columbia for this election.)
Nanos also said polls done more than a week before election day can be discounted, you can't project an election that far out. But looking at the polls done within days of voting, there's an issue.
The main methods of polling used during the B.C. campaign included online surveys and IVR or robocalling technology, which have a history of being less accurate when compared to more traditional methods such as random telephone surveys, Nanos said.
Nanos Research uses random telephone surveys that include landline and cellphone lines for its election tracking. They are the most expensive tools, but are also more accurate.
Nanos points out that polls do not project seats, but it is fair to compare polls to the popular support.
"The Liberals won by five points. Many of the pollsters predicted that the NDP would win by nine, that's outside of the margin of error for basically almost anyone. And that's not a great showing for online surveying and for the IVR, robocall pollsters," Nanos said.
"If you're polling close to election day, there's really no excuse not to get it within the margin of error for your survey," Nanos said.
Nanos said pollsters can get it right, but it will require an investment.
Recognized as one of Canada's top research experts, Nik Nanos provides numbers-driven counsel to senior executives and major organizations. He leads the analyst team at Nanos, is a fellow of the Marketing Research and Intelligence Association, a research associate professor with SUNY (Buffalo) and a 2013 public policy scholar with the Woodrow Wilson International Center for Scholars in Washington DC.