Paul St-Pierre Plamondon and the Parti Québécois secured victory in yesterday’s provincial election, but came up five seats short of a majority government.
And this despite the complete and utter collapse of the Coalition Avenir Québec, a party that only a few years ago seemed nearly unbeatable. Instead, the PQ’s path to a majority were blocked by the three other major parties on the ballot — as well as the PQ’s own self-imposed, referendum-induced ceiling.
Les chiffres : PQ minoritaire, la CAQ éclate - Pour ce nouvel épisode du balado Les chiffres, Philippe et moi décortiquent les résultats de l’élection québécoise, qui s’est conclue par une victoire du Parti québécois — mais avec seulement 28 % des voix et un gouvernement minoritaire.
The PQ won 59 seats, five short of the 64 needed for a majority government in the 127-seat National Assembly, with 28% of the vote. This election marks the first time in the history of Canadian federal and provincial elections that no party was able to capture at least 30% of the vote. The PQ’s tally is the most seats it has won since the 1998 election, though its share of the vote is still lower than any of the party’s results between the 1973 and 2012 elections. A win’s a win, of course, but by historical standards this one does not rank with the Parti Québécois’s past triumphs in 1976, 1981, 1994, 1998 or, arguably, 2012.
The pledge to hold a referendum in the party’s first mandate (after Donald Trump leaves the White House) undoubtedly played a role in limiting the PQ’s potential. It was stuck at or below 30% for the entire campaign, putting it at the same level of support as support for Quebec independence. By corralling most of the sovereignist vote behind the PQ banner, the party was able to win in a divided electorate. But it probably also prevented it from winning bigger.
The Liberals won 40 seats and will form the official opposition after capturing 24.2% of the vote. While a marked improvement over 2022, when the Liberals won only 21 seats and took 14.4% of the vote, the party’s vote share nevertheless ranks as the second-worst in the party’s history. But in an election with five viable parties, everyone is graded on a curve.
The Conservatives made a splash with 20.7% of the vote and 19 seats. Those 19 MNAs are the first to be elected under a Conservative banner in a Quebec provincial election since 1935. Québec Solidaire, meanwhile, secured nine seats and 12.4% of the vote, down from the 2018 and 2022 elections but higher than where it was previously — an indication that QS is here to stay.
That brings us to the Coalition Avenir Québec, which took 13.4% of the vote but failed to win a single seat. It’s a catastrophic result that was predicted, but is nevertheless gobsmacking to see in reality. The last party to go from majority government to being shutout was the New Brunswick Progressive Conservatives in 1987. But even in that election the PCs still managed nearly 29% of the vote. It’s hard to see how the party survives this.
Or why it should. With the current make-up of the National Assembly, the question has to be asked: if the CAQ didn’t exist, would anyone invent it? The party was launched as a third way between sovereignty and federalism with a right-of-centre policy offering that stood it apart from the social democratic PQ and the centrist Liberals. It replaced the Action Démocratique du Québec, which had previously filled this roll. Now, it seems likely the Conservatives will take up the torch — leader Éric Duhaime has pitched himself as searching for more autonomy for Quebec without independence, and he is clearly to the right of the political spectrum. It’s not obvious what electorate remains to the CAQ going forward, as federalists have the Liberals as a viable option, small-c conservatives have the big-c Conservatives now, and the nationalists who were soft on sovereignty and tired of the PQ have, for all intents and purposes, gone back to the PQ.
Throughout the campaign, I discussed how there were many races being contested between the PQ and the CAQ, and that the survival of the CAQ would be one of the factors that could block the path forward for the PQ. That was true when the CAQ was at 20% in the polls. It was even true when they were at 17%, maybe even 15%. But the drop to 13% on election day simply removed the CAQ from the equation entirely. In only five ridings — FIVE! — did the CAQ finish within 10 percentage points of the winner. In only two ridings did the CAQ come within four points (Trois-Rivières, where leader Christine Fréchette was on the ballot, and Jonquière). They simply weren’t competitive anywhere else.
But the other parties took up the task of blocking the Parti Québécois’s majority, helped along by the PQ’s softening support in the closing days of the campaign. The Liberals and Québec Solidaire blocked the PQ in Montreal, the Liberals blocked them in the Montérégie region and the Conservatives in central Quebec. And all three parties blocked the PQ in Quebec City, where the party failed to win a single seat — and even lost one of its few holdings at dissolution.
A lot of races were settled by narrow margins. Only 3,036 votes in six ridings separated the PQ from finishing with 65 seats. The Liberals eked out narrow wins against the PQ in four of those (Vachon, Montarville and La Prairie in the Montérégie and Jean-Talon in Quebec City) and the Conservatives in the two others (Champlain in the Mauricie and Louis-Hébert in Quebec City).
Without the Liberals and Conservatives winning those seats, the PQ would have secured a narrow majority.
But that is not to say that the PQ was robbed of a majority at the margins. They also got to 59 seats thanks to their own close shaves. If 3,036 votes stood between the PQ’s result and 65 seats, only 3,244 votes ensured they won 59 seats instead of 54. The party won by fewer than 1,000 votes over the Conservatives in Drummond—Bois-Francs, over the Liberals in Papineau and Saint-François and over the CAQ in Trois-Rivières and Jonquière. The night could’ve gone very well for the PQ. But it could have also been quite disappointing.
The Parti Québécois experienced significant gains across the province, but they needed to make up more ground than they did. The party was up 16.5 points in the suburbs of Montreal and in the regions, but the Liberals were also up 11 points in the Montreal suburbs and in the regions, cutting into the PQ’s margins. The Conservatives were up 16.6 points in the Quebec City region, dwarfing the PQ’s 9.5-point gain in and around the provincial capital.
The CAQ’s wipeout is attributable to its rather uniform loss throughout Quebec, down 29 points in the Montreal suburbs, 31 points in and around Quebec City and 32 points in the regions. But the PQ needed to take more of those points to secure its majority — the Conservatives were the bigger beneficiaries around Quebec City and the Liberals took their fair share throughout the province as well.
The shifts for Québec Solidaire are worth noting. The party was up nearly three points on the island of Montreal, finishing second with 25%. But it was down everywhere else. Though the party actually only lost seats in Montreal, holding its own in Quebec City and Sherbrooke, the party’s change in vote share actually made its support base more disproportionately concentrated in Montreal than it was before.
Also notable is the Conservatives’ lack of progress on the island of Montreal, up only 0.2 points. The polls suggested that Duhaime’s party was poised to make inroads among anglophones, but that does not seem to have materialized. In the six majority-anglophone ridings on the island, the Conservatives averaged just 13% support to the Liberals’ 64%. Polls had the Conservatives around 20% among non-francophones, but these numbers do not reflect that kind of support.
How the polls did in Quebec
It was actually a rare (marginal) flub for the polls, which on the whole did very well in the Quebec election. There were joint winners of the pollster sweepstakes, with Léger (in its polling for Le Journal de Montréal/Québec and TVA Nouvelles) and Pallas Data taking the title with a total 4.8-point error, equating to less than one point per party.
Forum Research had a similarly good result, but it should be mentioned that Forum has not polled in Quebec for years and only dropped its poll late on Sunday night. The poll also had a seat projection that awarded 19 seats to the CAQ despite the party being at 13%. There are no points for style on that, and the fact that Forum did not poll the campaign and instead put out a late-night report that had no broader purpose than for Forum to have a result to point to post-election — virtually no one in the public would’ve seen the poll, since it went unreported because of its late publication — makes me loathe to give Forum any laurels. I firmly believe in the public benefit of polling in that it helps inform voters in their decision-making and understanding of a campaign, and this poll did not do that on Sunday night, and Forum certainly never attempted to do that over the last six weeks.
Research Co. also had a decent performance, though was a little too high on the Parti Québécois. Liaison Strategies also performed well enough, though was too high on the CAQ. Mainstreet Research significantly over-estimated the CAQ and under-estimated QS and was the worst performer of the pollsters who published data in the last week of the campaign.
On the whole, however, pollsters did a good job, hinting at a weakening in PQ support in the final days and a rising tide for the Conservatives and Liberals as the CAQ collapsed. The regional numbers were also good, showing the rise of the Conservatives in Quebec City and the regions and the growing strength of the Liberals in Greater Montreal.
Kudos to the pollsters, particularly Léger and Pallas Data, who were in the field throughout the campaign, who put out good quality work that helped illuminate media coverage and inform voters about the real state of the race from start to finish.
How the model did
This election was the first test of The Writ’s revamped vote and seat projection model, and overall I am pleased with the results. Some lessons have also been learned, which I intend to apply to the ongoing B.C. election campaign, as well as to the federal model (and future provincial models).
The vote projection was a little too high for the PQ and CAQ and too low for QS and the Liberals, though every party was projected to within two percentage points.
The PQ was favoured in 65 seats, six more than the party actually won. But the result of 59 seats was well-within the low and high ranges and was quite close to the Avg. Projection, which I think proved its value by signalling that the PQ was more likely to under-perform than over-perform. The PQ was given a 58% chance of winning a minority and a 36% change of a majority. Considering how the count unfolded, that was a very reasonable projection.
The Liberals, Conservatives and QS were off by only a few seats, with the result falling within the confidence intervals.
The CAQ was favoured in zero seats, which was on the money. But it had 9.8 seats in the average projection, clearly too high considering the party was not within spitting distance of more than five seats. The high range of 39 was also ludicrously too high.
This is one of the lessons I’m taking from the results — that the level of uncertainty in the model is too great at the margins. I believe I know the reason, as the model’s ranges for a party’s performance in a seat is based on past errors, but not on the party’s projected support in any given seat. This means that a party’s +/- range in a riding was the same whether it was at 40% or 20%, which gave the CAQ a chance in too many ridings where the leading party was projected to have a relatively low level of support. It’s reasonable to assume the CAQ could’ve beat its projections by 10 points in a riding where it had 45% support, but unreasonable to expect a 10-point over-performance in a riding where it was down to 15%.
Correcting for this would’ve lowered the CAQ’s average seat projection and increased it for the other parties, which would’ve likely brought the average projection closer to the mark. It would not have changed the number of seats each party was favoured to win, however.
And on that score the model did respectably well. The wrong winner was projected in only 13 seats, meaning the model had the right winner in 114 of 127 ridings, or in 89.8%. If we include the tosses as “correct” calls, then the model only had seven true misses, for a 94.5% accuracy rating. Close enough to the 19 times out of 20 for me!
The chart below shows the calls made and whether they were right. In the misses, the projected margin is shown for the party that was wrongly called as the winner. As you can see below, in five of the 13 misses the projected margin was three points or less, while in nine it was seven points or less.
The seven true misses are interesting cases. Three of them are in Quebec City, where the PQ unexpectedly lost one of their own seats (Jean-Talon) and failed to flip Louis-Hébert and Taschereau. The latter might have been more predictable due to the presence of a QS incumbent, but the Liberals’ upset win in Jean-Talon came as surprise (it was the model’s biggest miss). The PQ’s inability to win these three seats can be attributed to its broader under-performance in Quebec City.
The Liberals were projected as the favourites in Abitibi-Est due to the presence of floor-crosser Pierre Dufour. But, clearly, he was not as much of a boon to the Liberals as the model expected (which, considering he was a caquiste before, might have been predictable). Ungava being projected Liberal is another special case considering the large Indigenous population in the riding that generally turns out in lower numbers, but usually votes Liberal when it does turnout.
The PQ under-performed on the island of Montreal (by about five points compared to the polls), which seems to be behind the model’s miss in Rosemont. Montarville going Liberal instead of PQ is another clean miss, like Jean-Talon, and perhaps a suggestion that the Liberals have re-connected with parts of a francophone base that abandoned it in 2022.
Another lesson I’m taking from the performance is that perhaps I should simplify my classification of projection calls. Six different categories is clearly too much, though they broadly performed as expected:
The model correctly predicted as many Lean, Likely and Safe ridings as it was designed to predict, but was less confident on Toss and Close calls than it needed to be. Going forward, I might try to consolidate these into three or four categories to simplify things a little.
So, a little work to be done to adjust and refine. But, overall, I’m pleased with the performance of the model and I hope it helped you understand the dynamics of this election campaign. There will be more models and, unlike in the case of Quebec and B.C., I hope to have the next ones up and running long before the next writs are dropped in Alberta, Manitoba and Prince Edward Island.
And, as a bonus, the Quebec model will start anew once the adjustments are made and the 2026 results are incorporated. Odds are the model will be put to another test in Quebec sooner rather than later. Thanks for following The Writ over the last six weeks. There’s plenty more to come!
And Albertans praying the UCP will follow the CAQ!!!