Pollsters showed they can still call a winner

Hey there, time traveller!
This article was published 20/10/2015 (3682 days ago), so information in it may no longer be current.

Getting an accurate, up-to-the-minute read of public opinion in the federal election was like trying to grab a sip of water from a churning waterfall. As difficult as it was, however, national pollsters performed fairly well in terms of showing Canadians they were about to elect a Liberal government.

The unofficial election results show slightly fewer than four in 10 Canadians (39.5 per cent) voted Liberal, with 31.9 per cent voting Conservative, 19.7 per cent voting for the New Democrats, 4.7 per cent voting for the Bloc Quebecois and 3.4 per cent voting for the Green Party.

Six national polls were conducted in the last week of the election campaign, with Liberal support ranging from a low of 35 per cent (Angus Reid) to a high of 40 per cent (Forum Research). All of the pollsters were within two points of the Conservative results, while the two polls that only included results collected on the last day of the campaign (by Nanos and Forum) captured the NDP around 20 per cent. Still, every other pollster pegged NDP support as being no higher than 22 per cent.

So overall, it was a good night for Canada’s polling firms. Our industry has spent a number of recent election nights nervously explaining why the polls didn’t hit the mark. Although some will engage in the parlour game of arguing over whose polls came closest to “predicting” the actual outcome, the more important thing to take away from the election in terms of public opinion is that polling painted a very clear picture of what was happening in the country.

Too often, people confuse polling with prediction. What polling does in an election campaign is draw out representative samples of public opinion and show how it changes over time. That is becoming increasingly harder to do as response rates drop for traditional telephone surveys, but pollsters are constantly refining their methods to make sure they obtain an accurate sample from that raging waterfall of public opinion.

In an election with such a strong “change” element in it, capturing public sentiment is a very fluid and dynamic process. Polls captured a shift toward the Liberals in late September and continued to show them on a steady ascent as the days counted down to election night. Here in Manitoba, a poll conducted in late September by Probe Research showed the Liberals and Conservatives tied with 39 per cent each.

The final Manitoba results show the Liberals captured 44.6 per cent of the vote compared with 37.3 per cent for the Conservatives.

Some might look at that and say “your poll was wrong,” but that is an unfair and unfounded criticism. The Probe Research survey sampled public opinion nearly four weeks before election day. The important takeaway is it showed a trend — Liberal support had surged, with the party going from a distant third place in the 2011 election to a tie.

A similar thing showing a sudden wave of public opinion moving toward one party or one candidate happened in last year’s Winnipeg mayoral election. A Probe Research survey for the Free Press conducted three weeks before election day showed Judy Wasylycia-Leis was in the lead, but that Brian Bowman had risen from third to second. Another survey conducted for two other media outlets two weeks later — a week before election day — showed Bowman and Wasylycia-Leis in a statistical dead heat. The result, of course, is that Bowman trounced Wasylycia-Leis.

Bowman’s momentum curve was on an upward trajectory and the polls taken captured voter sentiment as it was at those different points. Looking at the national polls in this federal election, we can conclude the same thing happened. The polls that showed the lowest levels of Liberal support and the highest levels of NDP support included results from three to six days before election day. There is no way of knowing this for sure, but had they been taken a couple of days later, they likely would have shown higher numbers for the Liberals.

There were many complaints about the proliferation of “horse race” polls in this federal election, but many pollsters also did a great job of showing the deeper dynamics of the campaign. One of the most important measures — and one that explains the Liberal breakthrough better than anything — is results showing when leaning and undecided voters were going to make up their minds on who to vote for.

The final Angus Reid Institute survey of the campaign showed that 45 per cent of these “soft” voters were not going to make up their minds on who to vote for until election day. Clearly, these voters broke massively to the Liberals when they entered the polling station. This level of volatility is what makes election polling difficult, as results can be skewed if these soft voters all break one way as opposed to splitting their support proportionally between the parties.

Winnipeg Free Press | Newsletter

Paul Samyn | Editor's Note

Wednesdays

A weekly dispatch from the head of the Free Press newsroom.

Sign up for Paul Samyn | Editor's Note

Sign Up

I agree to the Terms and Conditions, Cookie and Privacy Policies, and CASL agreement.

It is also difficult if it appears that on-the-fence voters are about to give a long-serving party the boot, but then they have second thoughts when it comes time to vote. This explains in large part some of the polling industry’s recent misses in Alberta in 2011 and B.C. in 2012.

In this case, however, the polls and election results showed clearly that Canadians were in a mood for change, and they fully intended to follow through with it this time.

Curtis Brown is the vice-president of Probe Research Inc., a Winnipeg-based market research company. His views are his own.

curtis@probe-research.com

Twitter: @curtisatprobe