What do those election polls really tell us?

0
35
What do those election polls really tell us?


With the federal election a month away, a national poll taken last week had both the Liberals and Conservatives with 33 per cent support — a tie.

A different poll taken last week had the Conservatives ahead with 36 per cent, and the Liberals lagging with 33.

A third, taken days earlier, had the Liberals in front with 34 per cent and the Conservatives behind at 33.

What does it all mean? Not much, say political scientists and pollsters.

Horse race polls — which measure popular support for major parties — are like sprinkles on a birthday cake: they are everywhere and they are eye-grabbing, but they don’t offer much sustenance.

“Even though it’s the thing media and consumers generally are the most interested in, this far out from an election, the horse race is actually the least interesting and probably instructive data that we can collect,” says David Coletto, a professor at Carleton University and CEO of Abacus Data.

At this point in the 2015 federal election campaign, one poll indicated the Conservatives had enough popular support to win a majority and give Stephen Harper his fourth term as prime minister. The NDP had been in first place for months, and Justin Trudeau was a distant third when respondents answered who would make the best prime minister.

That poll, and the others mentioned above, were all well-designed surveys of their type, and conducted by reputable polling firms. The problems arise from how we interpret them, say experts.

YOU MIGHT BE INTERESTED IN…

This “is nothing against the polling firm — it’s just the way it is,” says Daniel Rubenson, a professor of political science at Ryerson University and co-principal investigator of the Canadian Election Study, a data collection project.

“It’s a mug’s game to put too much stock in individual polls at one time that aren’t very big samples.”

For starters, most polls have a margin of error. The poll that indicated a Conservative-Liberal tie, for example, had a margin of error of plus or minus three percentage points, 19 times out of 20.

What this means is that if the pollsters were to go back out and redo the survey with an entirely new set of random respondents 19 times, they are confident the results would be no more than three percentage points higher or lower than the results they actually got. One time out of 20, they could accidentally poll an extremely high number of rabid NDP stalwarts, or an extremely high number of people who just can’t stand the sight of Trudeau, and get a skewed result that doesn’t map out onto reality.

The pollsters hope the survey they did isn’t the skewed one full of hair-haters, but there is a small chance it could be — about 5 per cent. So it’s important not to put too much stock in a single poll, says Rubenson.

YOU MIGHT BE INTERESTED IN…

But even more important is to remember that if you add or subtract three percentage points from either the Conservatives or the Liberals, the race is no longer a tie: the gap between them could be as large as six points. That is true for the other two polls mentioned too. The leading party and the losing party could be much closer together, or even reversed.

Reporters, this one included, don’t typically like to start stories with sentences like “No conclusions could be drawn from the latest national poll,” however.

Media tend to “overemphasize very small, statistically insignificant differences and changes,” says Rubenson, so it’s more helpful to look at many polls than at one.

Rubenson also points out that our political system doesn’t elect parties or leaders based on a nationwide popular vote. Governments are formed based on how many seats they win in parliament, which is much harder to calculate from a survey that samples a few dozen or a few hundred people in each province.

“In terms of seat projections, those are much more complicated models of the world. Polls are ingredient in those models, but they’re not the only thing.”

David Coletto says there are some relevant insights that can be derived from polling this far out from election day. They just tend not to generate headlines.

Questions about whether respondents believe the country is headed in the right direction, or the issues they think are the top priority, can all help gauge the country’s mood.

On the other hand, surveys can also show that many Canadians have not even tuned into the election yet — as Coletto’s firm’s own polls indicate, he says. That means there is lots of room for parties to attract voters.

“Almost half of Canadians haven’t even really tuned into the election yet, and yet they’re still answering how they might vote,” says Coletto.

“That choice that many people are picking at this stage is really their starting point. It’s where they typically would vote, it’s where they voted the last few elections provincially or federally, but it may not be where they end up.”

Get The Lead newsletter

Start getting your whip-smart guide to Canada’s 2019 federal election in your inbox.

Sign Up Now

And another big chunk of voters may change their minds. Coletto cites the big NDP slide in the 2015 race.

“Campaigns matter,” says Coletto.

It’s why we have elections, and not just polls.

Kate Allen

Kate Allen is a Toronto-based reporter covering science and technology. Follow her on Twitter: @katecallen





Source link