Skip to main content

B.C. Premier Christy Clark, left, and former NDP leader Adrian Dix react to the 2013 provincial election results – which were considerably different than what pollsters predicted. Ahead of the vote, polls showed the NDP headed for a landslide, but the party instead lost badly to Ms. Clark’s incumbent Liberals.

In 2013, an Angus Reid poll found the NDP ahead by 9 percentage points, but they finished five points behind the BC Liberals , reports Ian Bailey

There was nowhere for pollster Mario Canseco to hide on election night in 2013. As vice-president of Angus Reid Public Opinion, he was on television offering informed commentary as voting results rolled in depicting a significantly different reality than the one Mr. Canseco had confidently predicted.

The problem for Mr. Canseco and other pollsters was that their numbers did not add up. Christy Clark had led the BC Liberals to their fourth straight majority mandate despite expectations, based in part on polls such as Angus Reid's election-eve survey that found the NDP were ahead by nine percentage points. Instead, the NDP finished five points behind the Liberals.

"Muhammad Ali said it best: You learn more from your losses than from your wins," said Mr. Canseco, who added that he did some painful soul searching but came to the conclusion that 2013 was simply unusual.

Related: The future of polling is hard to predict

Related: Polling industry fails on its biggest stage: the U.S. election

In light of the polling debacle of 2013 – and in the context of similar forecasting failures in the 2012 Alberta election of Alison Redford, the Brexit vote last year in Europe and the U.S. election of Donald Trump – the provincial NDP is making some changes in how they will use polling to try and win the election. Meanwhile, the Greens are using polling for the first time. And the BC Liberals appear to be staying the course on measuring voter opinion given, they say, their numbers forecast the result.

A survey of the major polling firms working in the province indicates some will largely avoid publicly disclosed horse-race polling for this election, but will not alter their methods or other aspects of their work, such as private polling.

Pollsters like Mr. Canseco and his former boss, veteran pollster Angus Reid, are talking about sticking to mapping voter views on issues, and seeking a sense of exactly who is going to vote.

Jeff Smith of EKOS Research Associates said his company will not be doing any media polling in British Columbia in the 2017 election, although the BC Green Party said in a statement on Friday afternoon that they have hired EKOS to do their campaign polling. The use of a pollster and a separate unnamed company to do focus-group work is a first for the party, which has one seat in the B.C. legislature. But the Greens see polling as a means of better focusing their campaign resources. In a statement, EKOS subsequently noted that polling for private clients is quite different from media polling.

Mr. Smith said, in a statement, that B.C. is a challenging environment for polling. "Polling in B.C. is a terribly risky business; B.C. voters are highly unpredictable, both federally and provincially."

Mr. Smith wrote that his firm "effectively nailed" NDP support, but "seriously underestimated" Liberal support. Also turnout was only 57 per cent.

"When large segments of the population don't vote and the preferences of non-voters are different from those who do show up to the polls, these types of problems will ensue," he wrote in an e-mail, adding the Liberals got their supporters to the polls while the NDP did not.

Not all companies are holding back. David Valentin, executive vice-president of Ottawa-based Mainstreet Research, said from Montreal that his company did polls for the 2013 election in B.C for private clients and forecast a Liberal win, but did not release the results to the media.

This year, though, the company is working with Postmedia and doing a weekly tracking poll.

Mr. Valentin whimsically refers to B.C. as "the scene of the crime" where people have a distrust of polling deeper than he has seen anywhere else. When Mainstreet posts B.C. polls, he said he gets calls and e-mails from people in B.C., who say it's not possible to trust such polls.

He said his company won't retreat.

"Polling is relevant. Polling does work. There are going to be misfires. We are very lucky we are not associated with those misfires."

Angus Reid says the experience last time has made him extremely cautious about this year's election.

The chairman and founder of the Angus Reid Institute, who has been polling for 40 years, said watching Mr. Canseco on television that night was painful.

"I remember … feeling sorry for him as he did the dance of the seven veils around what was clearly a major miss," Mr. Reid said in an interview.

"I am not meaning to blame anyone because ultimately the buck stops at my desk. Last time, to be frank with you, it was, like, OK. This is just another election campaign that we're covering and it was just kind of handled in a pretty routine way."

Mr. Canseco noted that polling simply did not catch a shift in opinion confirmed on voting day, largely because the data at hand was several days old. He said the methods – online polling tapping the views of a recruited panel of voters, recruited to cover various groups – were fine.

"I don't think it was a methodological issue. It was an issue of people changing their minds," Mr. Canseco said. He subsequently left his job with Angus Reid Public Opinion, though he says it was not because of the 2013 election situation.

But Mr. Reid remains stung and is wary about getting too deep into the 2017 campaign.

Taking the pulse of voters

In their 2015 paper, The Accuracy of Public Polls in Provincial Elections, authors David Coletto and Bryan Breguet map out the varied methods and technologies for conducting polling, including:

Computer-assisted telephone interviewing

A live interviewer conducts a survey over the p hone, assisted by a computer program that prompts questions to the interviewer as the interview progresses. “CATI surveys are still considered by some in the industry to be the best at accurately measuring public opinion because they more closely adhere to probability theory and response rates are generally higher than [interactive voice response] surveys,” the authors say.

Interactive voice response

Respondents receive a call and are prompted by either a prerecorded message or automated voice instructing them to use their keypad to answer questions. Such surveys are said to be inexpensive to field, and can be completed quickly. “The limits are that the surveys have to be much shorter than [computer-assisted telephone interviewing] or internet surveys and the respondent’s identity cannot be confirmed,” says the paper. Also most IVR polls can only call a household once.

Internet surveying

Panels of individuals who have agreed to participate in a survey when recruited. Random panelists complete the surveys through e-mail. “Questions frequently asked about internet surveys surround the representativeness of the panels used to conduct the research and whether the act of completing many surveys has an impact on the opinions and behaviour of panels,” the authors say.

Ian Bailey

He said his team had not decided how to proceed ahead of voting day on May, 9.

"We're approaching this campaign much more cautiously," he said, suggesting his firm may only canvass an assessment of opinion on issues and also trying to understand the views of the electorate.

"There's very little upside and a lot of downside and risk associated, particularly with established companies or organizations like mine, doing anything. I mean, why bother?"

But political parties must.

Like many in the province, the 2013 election results came as a shock to members of the NDP.

In a task-force review of what happened, a party panel concluded the NDP did a large sample size poll in February and March, 2013, but no target riding-specific polling between then and voting day on May. 14.

"In the absence of consistent polling in these targets, the party and the campaign had no specific sense of just how volatile voter intentions were, who was being affected by what arguments in the campaign and whether key elements of the platform and campaign message were resonating with voters in those ridings," said the election review.

The review blamed complacency among NDP members who were lulled into over-confidence on media reports of public polls that had the party way out in front, public polls that were described in the review as "one of the most bedevilling, distracting and, in many ways, demobilizing factors."

This time, Glen Sanford, BC. NDP deputy director, says the party is using polling to get a better sense of what people care about and how to talk to them about those issues.

"We know that we have got to use the tools of polling to ensure that we are communicating effectively about the values that we represent. That is how we intend to use polling in this election," Mr. Sanford said in an interview.

Mr. Sanford declined to get into other details about the NDP polling effort.

Meanwhile, the BC Liberals are carrying on.

The party confirmed that Dimitri Pantazopoulos, the former principal secretary to the Premier, will be polling for the 2017 election. The party made Mike McDonald, the director for the 2013 BC Liberal campaign, available to comment on polling issues.

Mr. McDonald was one of few people interviewed for this story who said he was not surprised at the outcome of the election.

"Because I am a pessimist and my nickname is Eeyore, I wasn't going to believe it until I could see it. I would have been surprised if we lost," he said. "Our numbers would have been completely wrong."

Mr. McDonald, who is acting as an adviser in the current campaign was wary about getting into the details of polling for the current race, but is skeptical about "free media polls" and online panels of voters ready to offer opinion.

"What happened in the last election is these numbers created a narrative that conventional wisdom adopted and then it failed to correct itself when things were actually changing," he said.

The day after the election, Ms. Clark chided the media for relying on polls they had not paid for. "You guys get this stuff for free, and you should take it for what it's worth," she said in one of several exchanges, during a triumphant news conference at her Vancouver office, about how the polls had got it wrong.

"I said this to you guys from the very beginning. The polls do not tell us how people are going to vote because voting day is the only day that they vote. It's like me asking you what you're going to have for dinner a month from now. Well, you know, maybe it's chicken. Maybe it's steak. I don't know. You don't know either."


MORE FROM THE GLOBE AND MAIL: