Amid Canadian political-poll mayhem, meet the man worth listening to

Article content

After a long stretch of dominance in the polls, the Conservative Party of Canada now finds itself the potential underdog. Mark Carney, Canada’s newly appointed (but not yet elected) prime minister is kiting his international credentials, shaking hands with the king and a Europe keen on stability.

This dramatic shift in voter sentiment sparks lively banter among political pollsters. Nik Nanos predicts the upcoming federal election in Canada between the Liberals and the Conservatives will be akin to a “knife fight in a telephone booth.” Others speculate on the duration of leadership honeymoons; they don’t last forever but sometimes they last long enough.

Article content

Riveting stuff for politicos, but watching this contest play out, poll by poll, will be agonizing for some. And sometimes, pollsters just get it wrong, making one cynical about the whole notion of polling. A mid-March Ekos poll — predicting the federal Liberals leading in the province of Alberta — rattled my cage.

“Going from one poll to another to another to another is like driving on a street without shock suspensions; it’s very noisy,” says Philippe Fournier, the creator of 338Canada, an independent, apolitical poll aggregator and analyst known for sticking to the math and letting the numbers speak without spin.

“I wrote a model that incorporates all this (polling) data, adds it up with demographic data,” he explains, “and I try to project odds of winning in each riding, and with that, odds of winning an election.” 338Canada’s stats are exceptional: Philippe’s model has covered 18 general elections in Canada, nailing the winner in 89.3 per cent of the electoral districts projected.

A physics and astrophysics professor at Cegep de Saint-Laurent in Montreal — studying black holes and planets and galaxies, by day — Philippe spends the rest of his time aggregating polls from several pollsters and analyzing that data using a statistical model he designed. Polls are weighted by sample size, field date and a firm’s past accuracy.

Article content

Article content

“The reason why Ipsos and Leger and Abacus and the other good pollsters in this country are good, is because they’re consistently good. Everybody can be right once, so I look at the long-term track record,” says Philippe, who also co-hosts the excellent The Numbers podcast with Eric Grenier.

Yet polls may not be as independent as the public might think; polls can be funded by or tethered to a political ideology, the sample size can be too small, the questions inherently biased. I ask Philippe about the mid-March Ekos poll, and I’m heartened when he assures me, “If you go on my website, you will see that this poll is not listed.” It’s rare for 388Canada to exclude a poll in the calculation, he says, shaking his head, “but the Liberals are not ahead in Alberta.

“All parties — provincial, federal — all parties will sometimes leak their internal polls to me,” he acknowledges. “You won’t be surprised to know that they never leak polls that are bad for them. They only leak polls that are favourable to them,” he chuckles. What does 338Canada do with these polls? “I will take them into consideration with a low weighting if the polling is done by a professional,” he says, “But I will be very careful not to take it at face value.”

Article content

And, he continues, all pollsters are private companies and often, they conduct political polls at a net loss. “This is like advertising for them,” he says. “When they self-fund a poll, it’s to make sure it has maximum media exposure .… so then they can turn to their private clients — airlines, orange juice or PR companies — to say, look, we were right in the election, look how precise it can be for your business.”

I take his point, but can’t quite expunge a slightly jaded feeling, especially after observing the U.S. election last November.

“American polls are not as good as Canadian,” he replies, without hesitation. In Canada, there’s a Darwinism at play that puts bad pollsters out of business. “If you constantly put out numbers that are off the mark, you will not get contracts … In the U.S., bad pollsters, some bad pollsters, can stay afloat. In Canada, it doesn’t happen. That’s why we only have a dozen pollsters and generally, they do pretty good … aside from some misses in Saskatchewan. Saskatchewan is a tough place to poll.”

Philippe pushes back — hard — against the suggestion that polls sway voters toward the purported winner. “Good polls did not help Hillary Clinton,” he asserts. “Information influences you,” he continues, “and polls, when they are done correctly by professionals, is objective information you get in a campaign. Everything else is spin.

Article content

“When I hear that polls influence, well … information influences. Advertising influences you. Campaigns try to influence you. Everything in a campaign — the events, the advertising — everything is designed to influence you.

“A poll, when it’s done correctly, will give you the score of the game,” he says. If you’re watching a game, and at the end of the second period the announcer reports, “Oh, you’re not doing well, the Habs are trailing 3-1 against Boston,” Philippe chuckles, “you don’t shout, ‘Don’t say the score! You influenced the game!’” The announcer — like a pollster — is just giving you the score of the game at that moment in time.

Polling techniques have evolved over time but Philippe isn’t worried AI will subsume the polling business anytime soon; polls still require timely, on-the-ground data inputs from real people. And while he acknowledges pollsters’ challenges in reaching younger people, that’s not anything new: “It was true in the ‘80s,” he says. “It was true in the ‘90s, when everybody had a landline and nobody had a cell phone.” What’s increasingly difficult, Philippe reports, is reaching voters who are anti-system: people who believe in online conspiracy theories aren’t likely to respond to a pollster.

Article content

Philippe’s a scientist and he sticks to the data. “I look at it coldly,” he says. Observing the intensity of the man on my video-call screen — in his black shirt and black-framed glasses, animating key points with a firm clasp of his fingers — I don’t doubt him for a moment. And if the numbers are cooked, he assures me, he’ll call out a pollster: “I will expose them; sunlight is the best disinfectant.” It’s a point he makes repeatedly in our conversation.

338Canada may well temper some of the bumps in the election ahead. Still best to buckle up.

Our website is the place for the latest breaking news, exclusive scoops, longreads and provocative commentary. Please bookmark nationalpost.com and sign up for our daily newsletter, Posted, here.

Article content