The next federal election may be decided by who votes, rather than for which party Canadians cast a ballot. But it's not entirely clear which party stands to benefit most from the disparity between the voting and non-voting populations, according to two recent polls.
The most recent comes from Angus Reid, which surveyed 1,514 Canadians online between Feb. 12 and 16. The poll found that 33 per cent of all eligible decided voters supported the Liberals, while the Conservatives had the support of 28 per cent and the New Democrats finished third with 26 per cent. The regional levels of support among eligible voters polled by Angus Reid would likely deliver 123 seats to the Liberals, 117 to the Conservatives, and 95 to the New Democrats, using the seat projection model of ThreeHundredEight.com.
But among likely voters, the results change significantly. The Conservatives place first with 32 per cent, followed by the Liberals at 31 per cent support and the New Democrats at 26 per cent. Based on the regional support levels of likely voters, the Conservatives would probably win around 142 seats, with the Liberals taking 104 and the New Democrats winning 88. Nine seats swing to the Conservatives in British Columbia (primarily from the NDP) and 11 in Ontario (from the Liberals) due to the difference between the general and voting populations.
This is the kind of result one would expect, considering the Conservatives have done better at the ballot box than they have in opinion polls in recent elections. This is in large part due to the Conservatives' traditional advantage among older voters, who tend to vote in larger numbers than younger Canadians.
However, a recent poll by Ipsos Reid turns that logic on its head. In their online survey of 1,001 Canadians, conducted between Jan. 31 and Feb. 4 for CTV News, the voting intentions of eligible voters were remarkably similar to Angus Reid's findings: 33 per cent for the Liberals, 29 per cent for the Conservatives, and 27 per cent for the New Democrats. Based on Ipsos Reid's regional numbers, the Conservatives would eke out a slim plurality with 121 seats to 118 for the Liberals and 62 for the New Democrats (this is largely due to the Tory lead in Ontario in this poll).
But among the most likely voters, the situation actually improves for the Liberals. Their support is boosted to 38 per cent, while the Conservatives and New Democrats fall to 28 and 25 per cent support, respectively. Support among likely voters was not broken down regionally, so it has to be estimated, but based on the voting intentions of likely voters as defined by Ipsos Reid, the Liberals would win 146 seats to 112 for the Tories and 66 for the New Democrats.
Why the discrepancy between the two results? Both polls had almost identical numbers among all eligible voters, but Angus Reid found the Conservatives to hold an advantage among likely voters while Ipsos Reid gave the Liberals a significant edge.
The reason lies in how likely voters were identified by the two firms, and the difference highlights one of the unknowns of the 2015 federal election.
An inability to determine who was likely to vote was the main factor behind the polling miss in British Columbia's 2013 provincial election, and this is why more polling firms are taking new approaches to this issue. In this case, Angus Reid is estimating who is likely to vote based on, according to their methodological report, "known variations in voter turnout – specifically across age groups" and "respondents' own identified past voting patterns and habits."
In other words, Angus Reid is giving more weight to older voters and to respondents who say they have voted in the past in order to determine who is likely to vote in the future Ipsos Reid's approach is more straightforward. It simply asks respondents "if an election were held tomorrow, which of the following best describes how committed you are to actually go out and vote?" and calculates the voting intentions of those who responded "nothing short of an unforeseen emergency could stop me from getting to the voting booth and casting my vote." By including "I would do my best to vote, but sometimes things get in the way" as a potential response but not counting them as likely voters, the well-intentioned but uncommitted (who can make up a large proportion of a sample) are excluded.
Whether one approach will perform better than the other is a question that will be answered only after the next election. Models using assumptions based on past behaviour can perform well when the profile of the voting population does not change markedly from one election to the next. If that is what happens in 2015, the Conservatives may be the beneficiaries. But with all three parties running in such a tight race, even marginal changes in the profile of the average voter may make all the difference.
Éric Grenier writes about politics and polls at ThreeHundredEight.com.