On time. The Léger firm predicted the results of the federal election almost perfectly. Even taking into account a very tight margin of error.
On September 17, three days before the elections, the Quebec firm’s probe gave 32% of the voting intentions to the Liberal Party and 33% to the Conservative Party. A week earlier, the two parties battled tied at 32% in Léger’s tables, and on September 6 at 33% each this time.
As of Tuesday afternoon’s count, the PLC was filling up with 32.24% of the vote, and the CCP with 33.96%. The latest forecasts for the Bloc (7% predicted, against 7.8% in the ballot box) and for the People’s Party of Canada (6%, versus 5%) also fit. The NPD score was slightly overestimated (at 19%, for a final result of 17.7%). We repeat: pile hair, or just like.
“We are very happy,” said Jean-Marc Léger. We work so hard. It is also ironic that during this election campaign, no other Quebec house published a poll. Nobody noticed it. We were the only ones to publish and it was our firm, a Quebec firm, which was the most precise on a Canadian scale, just as we were very precise in Quebec, Ontario and British Columbia. It’s the key to success. “
No false modesty, then, in the aftermath of the intractable test of reality? “Certainly not,” replied Mr. Léger. A campaign is a very delicate moment, under pressure. People in political parties hate us all. “
Even those in the lead? “Ah yes, ah yes, assures the pollster. It is never to their liking. If you knew the number of calls I have for complaints. But when everyone hates us, it’s a good sign. It’s a bit like for journalists, what. “
Qualify the failures
The rest of the pollster class did not do badly at all. And polls, we’ve had a dime a dozen. The Wikipedia page devoted to the subject listed nearly 140 produced between August 15 and September 19 by a dozen firms actually coming from all but one of the Rest of Canada (Nanos, Ekos, Mainstreet, Abacus, Angus Reid, etc.).
“It’s a good day for pollsters,” says Claire Durand, full professor in the Department of Sociology at the University of Montreal, an essential specialist in this very particular technique for reading the bowels of citizens. She generalizes the very positive appreciation and finally passes the ointment over the whole group, or almost.
Like M. Léger, Mme Durand also corrects the impression that pollsters are often wrong when trying to follow the electorate.
The professor has meticulously studied every federal election since 2000, and only one (in 2004) recorded marked differences between the polls and the ballot box.
There were hiccups, yes, as in 2018 provincial election in Quebec, when the historic and catastrophic collapse of the Liberal Party and the tremendous growth of the CAQ escaped the barometers. There had been similar cases of wacky projections in Alberta in 2012 and in British Columbia in 2013, there, with a 17 point gap between the virtual and the real.
“In 2018, according to the analyzes that we carried out, my colleague André Blais and I, by administering a post-election survey, it seems that people changed their minds during the last weekend and even on the day of the vote,” says the Professor Durand. So there was a shift from the Liberal Party to the CAQ. Opinion changed at the last minute. “
The sampling challenge
That said, a low turnout is puzzling forecasters. We will probably be close to 62% or 63% this time, which remains within the low average of the last decades.
“For a pollster, then you have to work hard to identify the right people,” said Mr. Léger. He himself says he performs well thanks to his unique reference panel, a pool of 400,000 Canadian adults, or one in 100 citizens, which allows him to build a reliable sample questioned online.
“I invested millions in this achievement,” said President Léger. It is the largest panel in the country. It belongs to us and it is constantly renewed. We control the sample. “
Above all, in total, to reduce the chances of royal wanderings, Professor Durand recommends consulting as many surveys as possible to finally establish trends, averages. Survey aggregators (Wikipedia pages, 338Canada, Qc125, etc.) collect and compare data.
Here again, Mr. Léger, of the Léger firm, can only agree. “A survey is a photo of the event. A photo is not enough. You have to watch the film of the event, several photos over several weeks. “