VANCOUVER - Among the biggest losers in the B.C. election campaign are the pollsters who for months have been predicting an NDP majority.
"I think people are going to re-examine the truthfulness of polls,"' Premier Christy Clark said shortly after learning her party would form the next B.C. government.
"If there is any lesson in this, it's that pollsters and pundits and commentators do not choose the government. It's the people of British Columbia that choose the government.''
Ipsos Reid polled samples of British Columbians on their voting intentions as far back as February and on each occasion found the NDP had at least a six-point lead over the Liberals.
>> ELECTION RESULTS MAP
>> MORE ELECTION STORIES
On May 13, just one day before B.C. elected a fourth consecutive Liberal government, Ipsos published polling results that put the NDP eight points ahead of the Liberals. On April 14, before the campaign began, their results suggested a 19-point spread between the two major parties.
Meanwhile, Angus Reid Public Opinion Polling pegged the NDP lead at 20 points on March 21 and nine per cent on the eve of the election.
Forum Research published a poll that showed a narrower margin than others, suggesting the NDP would garner 43 per cent of the popular vote while the Liberals would get 41 per cent - a two-point spread that does not exceed the poll's margin of error.
Innovative Research Group Inc.’s managing director Greg Lyle said that 58 per cent of respondents to an Angus Reid online poll released the day before the election said they believed it was time for a change and for another provincial party to be elected.
Only 28 per cent said they would prefer to re-elect the Liberals, the same number former Ontario Premier Dalton McGuinty was polling at before his most recent re-election Lyle said.
He added that the polls failed to ask “agree or disagree: I’m angry with the Liberals, but I still think they’re the best choice to run the government.”
“I think that answer really is the story of tonight,” Lyle said late Tuesday. “It’s not that they forgave the Liberals for everything, it’s that, despite their anger at the Liberals, they still felt they were the better party to trust with the reins of power than the NDP.
“In the end they just could not bring themselves to vote NDP, they were too scared, they were too nervous.”
Like past B.C. Social Credit supporters, these “10-second Liberals” cursed the government, but ended up supporting them for those crucial moments in the voting booth, Lyle said.
Lyle, who directed Gordon Campbell’s failed bid to become premier in 1996, said after miscalculating recent election wins in Alberta and Quebec, polling firms must re-evaluate an online political polling system that may be inherently flawed.
“This is actually the third one where the polls have been out (of whack),” he said. Some of the last online polls before Tuesday’s election were almost 10 points off the actual level of support the parties received from voters.
That’s because these online poll respondents are more likely to be younger people who enjoy sharing their points of view, which skews support in favour of the NDP Lyle said. He added that other flaws inherent in the online polls are that 20 per cent of voters can’t be found online and close to 10 per cent of B.C.’s population has a lower level of English that prevents them from participating.
“I believe a big part of the problem was relying on online polls without using something like telephone weighting on something like party (affiliations),” Lyle said. “The only way you can be sure you have a complete sample is with telephone.”
A Forum telephone poll conducted later in the campaign [see it here] found a razor-thin gap between the Liberals and NDP, which correlated with some private polling his firm completed Lyle said.
Polling also tended to dominate the discourse of media coverage during the campaign, he added.
“Right off the bat it set expectations, there was basically an expectation that the Liberals would lose and the NDP would win and that other parties weren’t real players,” Lyle said. “A lot of discussion about one or two point differences, and not really the depth of discussion there might have been, for example, about health care or education.”
Their failure to predict voting behaviour in the days and weeks leading up to the election is reminiscent of the 2012 Alberta provincial election, where polling data suggested Danielle Smith and the Wild Rose party would dethrone the long-ruling Progressive Conservatives.
Right up until election night, polls suggested the first change in government in 41 years but were proven wrong when Alison Redford won 61 ridings to Danielle Smith's 17.
But Hamish Telford, a political scientist at the University of the Fraser Valley, said the situation in Alberta was different.
"They were quite clear in Alberta that what happened there was a last-minute shift. The polls had been fairly accurate up until the night of the election, and over the last weekend, and their election was on a Monday, there was a shift,'' Telford said.
"There was no reason to believe that a similar shift was going to happen here,'' he said.
"The Wild Rose had run a flawed campaign with some problematic candidates tarnishing the image of the party and I think some voters had reason to become nervous about electing this untested party, whereas here, the NDP, I don't think they ran a strong campaign but they didn't make many mistakes. I don't think they made any big errors to cause voters to reconsider.''
Instead, said Telford, pollsters in B.C. may have failed to properly account for undecided voters.
"The last polls I saw over the weekend said it was about 20 per cent of the electorate was undecided. It would appear that that undecided broke heavily for the Liberals,'' he said.
"Pollsters know there are a certain number of people who are undecided. That's something they are going to have to pay more attention to when we look at polls,'' Telford said, adding that some people he called ``leaners'' may have changed their minds.
Pollsters must account for voters who are leaning towards voting a certain way, but have not decided. Telford suggested pollsters do not have a good way of accounting for leaners.
Telford also wondered if there may have been a language barrier between pollsters and voters that do not speak English, or speak very little English.
"There are a lot of voters in and around Metro Vancouver whose first language is not English. Were the pollsters adequately tapping into those communities or were they getting a high degree of non-responses from those communities?'' asked Telford.
"Pollsters are going to have come clean and give us an explanation.''
With a file from The Canadian Press