We spoke with Dr. Daniel Westlake, a USask faculty member in political science, about what drives polling shifts and how quickly polls can change — he said Canadians have switched focus from anger at an incumbent government to concern with tariffs and American hostility.
“In the lead-up, when the Conservatives were polling well, the dominant issue was things like cost-of-living, and things where voters are going to assign a lot of blame to the incumbent government, regardless of what role the government is playing in leading to those outcomes,” Westlake explained.
“Now, that’s not necessarily the issue on the forefront of everybody’s mind. You have a different issue, the question of tariffs and the relationship with the United States. I’m not sure it’s necessarily an issue the Liberals have an advantage on, but it’s not nearly as harmful as the question of the cost of living and the issues that were dominating the public discourse back when the Conservatives had their large lead.”
One of the questions on the public mind now is how accurate those polls could be, raising the question of how pollsters collect their information. How biased are the results?
Westlake noted that the entire goal of gathering survey data is to get what’s called a ‘representative sample’: That is, rather than gathering just the opinions of office workers on lunch in downtown Victoria, pollsters want responses from a mix of rural and urban, wealthy and not-so-wealthy, from tradespersons to social workers, from BC to the Prairies to Newfoundland and Labrador.
“Polling firms all have different methods. The general thing they’re trying to do is get a sense of where the population is by getting as close as possible to a random sample,” Dr. Westlake said. “Historically, you used to be able to just dial random phones numbers ... the move away from landlines and the real drop-off in the number of people who respond to requests for random phone surveys, has made that more and more difficult.
“And so, different firms have different ways of getting around this.”
Whether it’s establishing a regular pool of respondents that can then be randomly sampled from, or using various means of contact to solicit responses, polling firms like 338Canada or Angus Reid stake their reputations on accuracy.
Some of the biases they attempt to compensate for include Selection Bias, when the sample captures only a certain segment of the population and not a fully representative slice (only farmers in rural Saskatchewan, for example); Nonresponse Bias, when certain groups in the survey are less likely to respond than others (people who don’t feel strongly about politics vs those who hold powerful convictions); and Confirmation Bias, when researchers unconsciously skew their results based on their own expectations.
It’s all a balancing act, and everyone knows there is no crystal ball.
“Polls are only ever a single image, or an attempt, to capture where public opinion is at the time,” Westlake explained. “We know public opinion is sensitive to events.
“There are two quite dramatic things that happened between when the Conservative Party had their 20-point lead and where we are today: Justin Trudeau resigned as Prime Minister ... and then you have the swearing-in of Donald Trump, and the tariffs that have been imposed, as well as the threats to make Canada the 51st state. And that has created an opportunity for the Liberal Party.”
It raises the question, he added, about how solid such a quick shift in public opinion can be.
“A lot of people just changed their mind. They said they were going to vote Conservative, now they said they’re going to vote Liberal. Is that a commitment that people have made, or is that very soft? As people get to know Carney, as the election happens further, are the Liberals going to be able to hold onto that?”
Westlake doubted any further shifts would put the ball fully back into the Conservative court, but the two leading parties could be more even by April 28.
If the federal NDP finishes where they are currently polling, on the other hand, it will be the worst result for that party since 2000.