Skip to main content

If you're a political journalist in this country, you inevitably get asked with some frequency why your profession keeps giving so much attention to public-opinion polls.

Beyond complaints that pollsters have frequently been wrong of late, this question is often accompanied by dreamy speculation about the utopian elections we would have if the media were willing to ignore the horse race, and only focus on loftier matters.

It's a nice thought. But the reality is that it's not going to happen, and that's not altogether a bad thing.

The more realistic aim is to be smarter about how polls are conducted and interpreted and put in context, so they ultimately do a better job of informing readers rather than risking misleading them. And that may require a degree of experimentation, like the project The Globe and Mail is launching this week.

Start, though, with why polls and their prominence are here to stay for the foreseeable future. The fact is, they wouldn't get published or broadcast if there wasn't almost insatiable demand for them. Just look at the traffic for a new poll story published on this newspaper's website or any other, relative to many reports or analyses about things less quantitative. Or just think about how quickly the horse race comes up in any real-world conversation about an election.

We may be overly conditioned to follow elections that way, making it impossible to imagine one in which we really had no clue who would win the most seats until after the votes were counted.

But considering that we live in a first-past-the-post system with at least three competitive parties, it is also reasonable for voters to want some sense of who has a realistic chance of winning.

In the United States, once the primaries are done, voters really only have two choices; unless they have an overwhelming desire to be on the winning side, the polls shouldn't influence their choice. Here, they may be torn between two options and want to prevent a third, in which case the state of play matters. Or they may want to vote for a party but think they'd be throwing away their vote, and polls can give them licence to vote for that party after all – which is, in some measure, what happened with the NDP in the past federal election.

Those of us covering campaigns are acutely aware, too, that parties constantly conduct their own opinion research, and that it informs their tactics throughout a campaign – what they talk about, where they talk about it, which ridings they pour resources into and which they write off. It helps to have some idea of what they're seeing. Unless we take their word for what their polls are showing, which is an enormously risky proposition, we have to rely on public-domain research for that.

The problem is that, in recent elections, much of the public-domain research hasn't been very good. It is less reliable than it used to be, which is problematic for those who might cast votes based on it, and it is vastly less sophisticated than what the parties are doing.

Public-domain pollsters have struggled with the decline of home-phone usage. New methods, ranging from online panels to interactive voice response (IVR), have yielded decidedly mixed results. Voters with English as a second language, who now make up large chunks of the electorate in many of the most important battlegrounds, are often harder to reach and engage than others. Low voter turnout, by historical standards, has complicated matters further.

The parties have faced some of the same challenges, and their own polling has been imperfect. But rather than conducting nationwide or regional surveys like the ones that appear publicly, which can be distorted by high or low support in ridings that are foregone conclusions, they focus only on competitive seats and use large enough sample sizes to really gauge them. Blending their polling efforts with voter data they have collected over the years and new information their door-knocking volunteers, call centres and digital teams pick up, they are able to get a much better sense of what day-to-day shifts in support really mean.

Unless media outlets are willing to spend millions of dollars on research, and can somehow come up with thousands of ground troops across the country, they're not going to be able to match that. So it's necessary to be thoughtful and creative, resisting the urge to over-hype individual polls that offer a sexy headline, and finding ways to paint a fuller picture.

In some cases, that can mean trying to get beyond the horse race, as The Globe did during last year's Ontario election – when it worked with the polling firm Innovative Research Group to explore how different types of voters reacted to different things they were seeing – and will try to do again this federal one.

It also means trying to find ways to cover the horse race better, as with the predictive model The Globe is launching this week. Hopefully, in aggregating and weighing the available public-domain polls with some eye toward historical patterns, the model's emphasis on probabilities (as opposed to pin-point predictions) will give voters a realistic sense of the campaign's ever-changing dynamics.

Other outlets, no doubt, will make their own efforts to understand the modern electorate. It's a fair bet they will all be imperfect, but hopefully they'll be better than what we had in the past election, and next election they'll be better still. Not trying is not an option.