Ontario election polling practices questioned
The plethora of public opinion polls done during the Ontario election campaign proved reasonably predictive of the actual results, but questions remain about the polling industry and how its numbers were used.
John Wright, vice-president of Ipsos Reid, railed Friday at the conduct of some polling firms, and at the media for failing to be more critical.
Some pollsters, he said without naming names, used another company's respondent pool.
He also said one media outlet published a poll in the morning saying the Tories would win, and another in the afternoon predicting a Liberal win.
"I have never witnessed anything as blatant and ridiculous as we saw in this [election]," said Wright, who has been in the business for 23 years.
"What it's going to lead to, I fear, is the call for the ban of polls during election campaigns."
Wright's post-election complaint echoed a letter he and Ipsos CEO Darrell Bricker put out last month that, among other things, criticized "marginal pollsters" and savaged the media for failing to hold them to account.
On Friday, Wright insisted the issue went to the very heart of democracy.
"There is going to be editorial creep where we are going to see polls in election campaigns match the editorial positioning of those outlets," he said. "This is [already] going on."
Pre-campaign, polls captured what appeared to be a huge lead in support for the Progressive Conservatives.
After the campaign began and people began to focus on the contest, polls showed a much tighter race. By voting day Thursday, the polls predicted a Liberal victory. In fact, the Liberals did win, but fell just shy of a majority.
"We predict that the Ontario Liberal Party will win a majority, but a final tightening of the race means that this is not a certainty," is how Ekos put it at the 11th hour of the campaign.
"Ironically, we have seen a modest tightening of the race just as others have found a major breakout for the Liberals."
Derek Leebosh, a vice-president with Environics Research, said the closeness of the race tended to focus more attention on the polls and their margins of error.
Still, when averaged out, he said, they were all "pretty close" to the mark.
"After all that huffing and puffing, their final poll ended up being off by more than anybody else's," Leebosh said of Ipsos.
Still, he said, Wright was right to highlight the potential pitfalls of polling, especially if surveys aren't done properly.
Wright responded by saying his final survey was accurate within its margin of error, and that if his firm had been in the field two days longer, the poll would have been bang on.
An 'assault on democracy'
Still, he called the post-election chatter about poll accuracy "irrelevant."
"Everybody can dance around on the heads of pins claiming that they were [accurate] on the last day, but what we witnessed during this campaign was unforgivable," Wright said. "Because you put a poll out, it's given credence and it's covered. This is the beginning of an assault on democracy."
What made little difference in the accuracy of the predictions was whether a pollster called to ask how you would vote, a machine "robo-called" you, or you responded to an online poll.
"There was no one mode that seemed to be more consistently more accurate," Leebosh said.
Jon Pammett, a political science professor at Carleton University, said he didn't think polls have much impact on how people vote.
There's little evidence of a "bandwagon effect" in which people see polls and get on board with that trend of opinion.
"I really don't put very much stock on the fact that polls really lead opinion," Pammett said.