Seat projection suggests a narrow Liberal victory

A new projection prepared for Maclean’s by Innovative Research suggests the Liberals will lose seats but keep power. The biggest gains may be made by the NDP.
Trudeau makes a campaign stop at the University of Windsor in Windsor, Ont., on Sept. 17, 2021 (Sean Kilpatrick/CP)
Liberal leader Justin Trudeau makes a campaign stop at the University of Windsor in Windsor, Ont., on Friday, Sept. 17, 2021. THE CANADIAN PRESS/Sean Kilpatrick

Justin Trudeau’s Liberal Party of Canada appears poised to keep power after Monday’s election, albeit with a reduced share of seats in the House of Commons, according to a seat projection prepared for Maclean’s by the Innovative Research Group Polling Firm.

The projection, based on party support in clusters of geographically or demographically similar seats, suggests the Liberals are poised to win about 148 seats, down from 157 in the 2019 election. Erin O’Toole’s Conservatives would win approximately 112 seats, also down from 121 in 2019. The Bloc Québécois under Yves-François Blanchet would win 35 seats, up three from their 2019 score. And the largest gains would be made by the NDP, which is on track to win about 42 seats, up from 24 in 2019 and just shy of the 44 seats Tom Mulcair’s NDP won in 2015.

“If you think about the Liberals fighting a three-front war, they have a narrow win over the CPC, a stalemate versus the Bloc and a big loss versus the NDP,” Greg Lyle, the president and founder of Innovative Research Group, said.

“The re-alignment we saw in the spring has faded. The CPC have rebounded in their core seats. A key factor in this is likely the strong backlash among CPC voters against the government’s handling of Afghanistan as well as backlash against the election call.”

To produce this analysis for Maclean’s, Lyle combined the samples from four national surveys throughout the campaign, with the first survey in August and the last from Sept. 10-12,  to produce a total sample of 6,009 voters. The result isn’t a snapshot of national voter sentiment at a precise moment in time. But a sample this big allows greater precision in identifying trends at a smaller geographic and demographic scale. Using this technique near the end of the 2015 campaign, Innovative Research was able to spot a stronger Liberal performance than had been predicted in most seat projections up to that point.

Lyle divided his large national sample into 12 clusters of seats with comparable voter dynamics, based on which parties have been most competitive in those clusters across the last three federal elections, and also incorporating regional dynamics in some cases.

Many of these clusters are showing new dynamics that didn’t figure in previous recent elections. In the 44 CPC Strong (Prairie) seats the Conservatives swept in 2011 and held against Trudeau in 2015 and 2019, the Conservatives vote is the lowest it’s been in a decade at 46 per cent, down from 71 per cent in the 2019 election, and NDP support the highest at 24 per cent, double the party’s 2019 vote in these ridings. Lyle anticipates the Conservatives losing a few of these prairie seats to the NDP.

But the Conservatives have reasserted their lead in the 37 CPC Strong (Non-Prairie) ridings they’ve held for a decade, ridings like Durham and Sarnia—Lambton in Ontario and North Okanagan—Shuswap in British Columbia.

In the 30 CPC-LPC Swing (Ontario) ridings that went mostly Conservative in 2011 and mostly Liberal since, Innovative found both Liberal and Conservative support lower than in 2019—at 35 per cent. down from 43 per cent in these ridings in 2019, for the Liberals and 30 per cent, down from 37 per cent, for the Conservatives. What’s changed? NDP support has doubled from 12 per cent to 25 per cent in these ridings. That’s likely to produce a number of three-way races.

Similarly, in the 31 CPC-LPC Swing (Non-Ontario) ridings, NDP support is up 9 points since 2019, so the three big parties are in for a bunch of three-way contests.

Quebec is proving to be a tough nut for the Conservatives to crack. They’re up in Montreal ridings where they’ve never had a realistic chance, but down in their relative stronghold around Quebec City, while both the Bloc and the Liberals are up in those seats.

No poll is a prediction, and it’s been an eventful week since the final wave of responses in Innovative’s campaign-long survey was gathered last weekend. Voter turnout is only one of several variables that could affect the final result. But in interpreting these results, it’s worth noting that in last weekend’s survey, Innovative found a larger Liberal lead than some other pollsters have found: 33 per cent to 28 per cent for the Conservatives, 22 per cent for the NDP, 7 per cent nationally for the Bloc, 4 per cent for the Greens and 5 per cent for “other” parties, chiefly the People’s Party. So Innovative’s model showing 148 or so Liberal seats is predicated on a comfortable Liberal lead in the overall popular vote. If that Liberal vote doesn’t come out, seat tallies could change.


This report combines the results of four online surveys conducted in August to September 2021. In total, the dataset contains a representative sample of n=6,009 Canadians, 18 years or older. The online samples of these surveys were conducted through INNOVATIVE’s Canada 20/20 national research panel with additional respondents were provided from Dynata and Lucid, leading providers of online samples.

The dates and sample sizes for each survey were: Dynata & Canada 20/20 August 2021: Conducted from Aug. 26 to 30, 2021 with an unweighted sample size of 1,433 (weighted to 1,200)

Lucid August 2021: Conducted from Aug. 26 to 30, 2021 with an unweighted sample size of 1,373 (weighted to 1,200)

Dynata & Canada 20/20 September 2021: Conducted from Sept. 10 to 12, 2021 with an unweighted sample size of 1,831 (weighted to 1,200)

Lucid September 2021: Conducted from Sept. 10 to 12, 2021 with an unweighted sample size of 1,372 (weighted to 1,200)

The combined sample is weighted to n=4,800 by age, gender and provincial sub-regions using the latest Statistics Canada Census data. Results are weighted to ensure that the overall sample’s composition reflects that of the actual population to provide results that are intended to approximate a probability sample. Results for the latter three surveys were weighted to match the reported past vote from the first survey to avoid any biases across panels.

Respondents were grouped together into their federal electoral districts based on their postal code. A weighted total of 97 respondents could not be grouped into a federal electoral district because they did not provide a postal code or their postal code matched multiple districts.

INNOVATIVE provides each panellist with a unique URL via an email invitation so that only invited panel members are able to complete the survey, and panel members can only complete a particular survey once. Sub-regional quotas are set within regions to ensure there is a representative sample of respondents from across the entire region.

This is a representative sample. However, since the online survey was not a random probability-based sample, a margin of error cannot be calculated. Statements about margins of sampling error or population estimates do not apply to most online panels