If you believe most of the public opinion polls so far in the federal election campaign, the Conservatives have this one pretty much in the bag. Oh wait, maybe not.
Canadians have been inundated with daily polls tracking the party leaders and their popularity since the start of the federal election campaign more than two weeks ago.
The numbers have been all over the place.
One day the Libs are up six points in Ontario, the next they’re down six. One day the Libs are projected to win 28 seats in Atlantic Canada, the next day it’s eight.
Why the wild fluctuations? That depends on who you ask.
The consensus seems to be that many Canadians haven’t been paying attention. They’ve been too busy enjoying their last weeks of summer and getting the kids ready for school. They say that those who are paying attention mostly vote Conservative. The all-important leaders’ debates this week will play a critical role in giving Canadians a more accurate appraisal of voting intentions.
But there have been two constants in all the polling. Number one is that the Conservatives seem to be slightly ahead. Number two is that, for most Canadians, Liberal leader Justin Trudeau remains the preferred choice for PM.
That would seem to be counterintuitive given that most polls have been telling us the Liberal leader has the worst favourability ratings of any of the main party leaders. It also doesn’t jibe with surveys telling us that most Canadians (some 61 per cent) believe the Libs will win. Those numbers include 40 per cent of New Democrat voters, 53 per cent of Bloc and 45 per cent of Green Party voters. That’s twice as many as those who believe that the Conservatives will win.
But polling is not an exact science. The bald numbers are only meant to be a snapshot of a moment in time. Alone, they don’t tell us much, but strung together over time they start to sketch out a picture. It’s when so-called expert (and not-so-expert) media columnists begin to interpret polls that bias enters the equation.
For all the fluctuations in the daily numbers over the first two-plus weeks of the campaign – and select news reports of an unstoppable Conservative “freight train” coming down the track – in terms of public support for the parties we’re about where we were in 2019.
The Cons’ support nationally as of Monday is at about 34 per cent (two points higher than they were at this time in 2019) and the Libs around 31 per cent (one point lower than where they were in 2019). The NDP are slightly higher than they were in 2019 at anywhere between 18 and 20 per cent.
You wouldn’t necessarily know that from some of the breathless reporting using the overall percentages to project how many seats each of the parties will win. Those projections tell us the Libs may be in line to lose as many as 30 ridings. It’s a mug’s game based on statistical probability.
The regional breakdowns of voter intentions, for example, paint a different picture.
Those poll numbers suggest a race much like 2019, only with the Conservatives looking right now like they’re poised to win more seats than they did last time in BC, Ontario and Atlantic Canada. The Libs, meanwhile, look like they may be able to make up most of those projected losses in Quebec if their support there continues to rise.
Notice the use of the words “may” and “suggest”? You won’t find them in reporting on polls these days. Too often, media outlets report on poll results as if they are a fait accompli, leaving the false impression among readers that any given poll at any given moment in time will reflect the eventual outcome.
Context is important in evaluating poll numbers. And right now most Canadians are not yet paying serious attention to the election, according to a recent Leger poll. Another 12 per cent or so are undecided. As high as 40 per cent in Leger’s poll in Quebec suggest they’re open to changing their mind, all of which puts a different blush on the trajectory of poll numbers in the current campaign.
Vote concentration and how support in different regions of the country may affect the outcome of the election is another variable you don’t often read about in media reports evaluating poll numbers.
The Conservatives, for example, actually won the popular vote in 2019 by slightly more than one percentage point over the Libs (34 to 33), but the Libs won 157 seats to the Conservatives’ 121. That’s because the Liberal vote is more spread out, whereas the bulk of Conservative support is mostly in the Prairie provinces.
The pattern was similar in 2015. The Conservatives garnered 32 per cent of the vote but won only 99 seats. The Libs won 184 seats with 39 per cent of the vote.
Not all polls are built the same. They use different methods to gauge public opinion. Some pollsters gather data from random people online (note to readers: ignore these). Some gather data from panels they pay to offer their opinions on various subjects. Depending on the polling company, these panels tend to be made up of folks over 55 and tend to skew Conservative.
Other polls collect data from automated responses by phone, which makes achieving the “representative sample” that reflects the general population trickier since older voters tend to use land lines and younger voters tend to use cellphones.
Then there is the all-important sample size. Generally speaking, the more people that are surveyed, the lower the margin of error. Sample sizes of 1,000 people or more are considered to be accurate two or three per cent 19 out of 20 times. Depending on the number of people surveyed, each carries a different margin of error. But some surveys use three-day rolling averages of sample sizes of 400 people or less to arrive at projections on voter intention.
One thing that all polls do have in common is that they are corrected (or weighted) to reflect the demographics of the general population. This is to ensure accuracy. But the exercise relies on statistical analysis that may or may not actually reflect what’s happening on the ground.
In polling parlance, what are known as “lagging indicators” seem to be pointing in the Libs’ favour.
Critics of public opinion surveys say they can be used to suppress voter turnout. Others argue that they’re only useful to media and political parties interested in framing a certain narrative. Seems a number of the polls of the last two weeks, for example, were designed to question whether an election is needed at all. So for whom exactly do the polls toll? The short answer for this election is that much remains in play.
