Surprise! Pollsters got it right
Liberal leader and prime minister-designate Justin Trudeau takes a selfie while greeting people at a subway station in Montreal, Quebec, October 20, 2015. (REUTERS/Chris Wattie)
Since we give them grief when they blow election calls, it’s only fair to note Canada’s major pollsters -- at least the ones I follow -- did a pretty good job of predicting the Oct. 19 election.
I’m talking about predicting the popular vote only.
Trying to predict seat counts from national polls is, in my view, a fool’s game, because the sample sizes aren’t large enough and the more you drill down to the ridings, the more unreliable the data become.
Those who attempted seat counts generally underestimated the efficiency of the Liberal vote and thus the size of Justin Trudeau’s victory -- 184 seats out of 338, well above the minimum 170 seats needed to form a majority government.
Canada’s pollsters are concerned about their credibility.
This in the wake of many blowing the call in the 2013 B.C. election, when they predicted an NDP victory in a race the Liberals won, and in the 2012 Alberta election, when they wrongly predicted the Wildrose Party would defeat the Progressive Conservatives.
Recently, some pollsters formed the Canadian Association for Public Opinion Research to establish a code of ethics, although not all pollsters are on side and often snipe at one another publicly for blowing election calls.
A poll early in this campaign commissioned and distributed to the media by the NDP -- not by any of the pollsters I follow -- didn’t help.
It showed Trudeau 11 points behind the NDP candidate in his riding of Papineau.
Another pollster (Mainstreet) using a larger sample size a few days later found Trudeau five points ahead and Trudeau eventually won with 51.8% of the vote, almost double that of the second-place NDP.
On the plus side, predictions on the popular vote were largely accurate.
According to the Elections Canada website on Wednesday, the Liberal party received 39.5% of votes cast, Conservatives 31.9%, NDP 19.7%, Greens 3.4%, Bloc 4.7%.
As for the pollsters:
Nanos' final numbers were Liberals 39.1%, Conservatives 30.5%, NDP 19.7%, Greens 4.6%, Bloc 5.5%.
Forum’s were Liberals 40%, Conservatives 30%, NDP 20%, Greens 3%, Bloc 6%,
Ipsos called it Liberals 38%, Conservatives 31%, NDP 22%, Greens 4%, Bloc 4%.
Mainstreet’s final public numbers were Liberals 38%, Conservatives 33%, NDP 21%, Greens 5% Bloc 4%.
(Mainstreet says its final numbers submitted Oct. 18 to polling aggregator site threehundredeight.com were Liberals 39%, Conservatives 31.9%, NDP 20.9%.)
EKOS had it Liberals 35.8%, Conservatives 31.9%, NDP 20.4%, Greens 5.5%, Bloc 4.9%.
(Pollster Frank Graves in an election post-mortem piece acknowledged EKOS’ numbers were slightly outside his poll’s margin of error in underestimating Liberal support.)
While the pollsters’ results varied somewhat depending on when they did their final polling, all had the Liberals in first and caught the surge in the final days of the campaign to a greater or lesser degree.
Finally, in fairness to a couple of pollsters, Nik Nanos, whose nightly polling was done for CTV and who was regularly accused by some Conservative supporters of being a Liberal hack, was very close to the actual results for the three major parties.
Meanwhile, if, as some Conservatives supporters complain, Frank Graves of EKOS is a Liberal hack, why did his poll underestimate the Liberal vote? Far more likely he was accurately reporting what his polls were telling him.