Après le Léger de samedi, voici le premier Mainstreet. Cette firme devrait publier un sondage quotidien, fait par téléphone (Léger est en ligne).


Mainstreet est une firme qui a souvent les partis de droite (PCC au fédéral par exemple) plus élevés. Elle a aussi les jeunes qui sont davantage à droite de manière quasi systématique. C'est vraiment une conséquence de leur méthodologie. En général leurs sondages ne sont pas mauvais mais gardons cela en tête. Léger avait le PCQ à seulement 14% par exemple.


Si j'ajoute ce sondage à ma moyenne, cela ne change pas grand chose. La CAQ est tellement dominante quasiment partout que changer quelques points de pourcentage n'a pas une grande influence. Regardez par exemple les intervales de confiance à 95%, la CAQ est entre 84 et 97. Ce n'est pas particulièrement large pour un parti en tête. Cela nous montre simplement que la CAQ n'est pas en danger mais, en même temps, n'a plus beaucoup de sièges qu'elle pourrait remporter. En gros, soit elle gagne facilement soit elle perd de beaucoup, si on simplifie.




La carte des projections

Page au complète.


Projections par circonscription


Soyons honnêtes, l'histoire de cette élection pourrait fort bien être la percée du Parti Conservateur d'Éric Duhaime. La CAQ étant maintenant bien pl;us centriste qu'en 2014, il y a une ouverture à droite. La fatigue de la pandémie aide aussi (mesures Covid, etc). Le PCQ n'est pas actuellement projeté gagnant dans aucun comté (et cela magré un bonus au chef Duhaime dans Chauveau) mais le parti est compétitif dans 4-6 comtés. Les nouveaux partis ont en général besoin d'un petit coup du pouce via un candidat local fort pour remporter des sièges (pensez QS). Est-ce le PCQ a de tels candidats? À part Duhaime, on verra. Il nous faudra surement des sondages par circonscription pour vraiment mesurer.


Bon voilà, demain je publierai une petite analyse/scénario où je montrerai à quel point il serait difficile pour la CAQ de ne pas gagner une majorité.

Petit poste rapide avec les projections basées sur le dernier sondage Léger publié ce matin. Pas grand chose à dire pour l'instant tant la domination de la CAQ est totale.


Les projections:

CAQ entre 87 et 99 sièges, 100% de chances de gagner une majorité. La seule informaiton intéressante est que la CAQ gagne actuellement quasiment tous les sièges possibles (94). Donc pas de grande possibilité de croissance, les quelques comtés qu'elle ne gagne pas semblent hors de portée.


PLQ entre 16 et 25. Ce parti n'existe quasiment pas dans le Québec franco. Si Anglade devait perde le vote anglo, ce parti se retrouverait sous les 10 sièges.


QS entre 7 et 11. Cette formation conserve essentiellement ses acquis de 2018 mais pas de réelle possibilité de gains.


Le PQ entre 1 et 4, en gros compétitif dans l'est de la province et c'est tout.


Le PCQ entre 0 et 1. Il lui faudra concentrer son vote bien davantage dans la région de Québec et espérer des effects locaux avec ses candidats. Duhaime a de vraies chances dans son comté.


La carte:


Les projections détaillées sont ici.

We now have enough data to make a (quasi) final call on this race. Quite honestly, we could have made this call back in February. It may well end up being one of the most boring leadership races ever -- Poilievre was the favorite from the start, took an early lead and kept it the whole way.


The TL;DR version is: Poilievre will be the next leader. I truly believe there is 0% uncertainty here. The only unknowns are whether he'll win on the first ballot (almost guaranteed) and whether he'll be closer to 50 or 70%. Here are the projections with the 95% confidence intervals. Poilievre win 100% of the time but in some simulations (0.2% of the time), he needs a 2nd or 3rd ballot.


The 56-74 range is based on 10k simulations with sample size 100 (yes very small but that's to make sure I included enough uncertainty in the model; I based this estimate off the performance of the projections in 2020). Interestingly, 56-74 is very similar to the range of the 4 metrics I used. Poilievre is at 72% based on Google Trends but only at 56% in the Mainstreet poll (once I redistributed the undecided and the Brown's voters).


Projections by province (it's percentage of points).


As you can see, Poilievre leads everywhere, including in Quebec. In this province, the only metric that would put Charest on top is the fundraising but only looking at the amounts collected, not the number of donors. Charest just got many large donations there. But the fundraising by riding would put Poilievre above 60%. Charest has a 12% chances of finishing first in Quebec. West of Ontario, Charest is likely to finish 3rd or worse.

I personally think Ontario is the most fascinating province with a extended 3-way race for second. Charest is favored here with all the measures but was actually fourth (!) using the fundraising by riding!

One more stat: Charest has a 94% chance of finishing ahead of Lewis overall.


As I've said, there are very few simulations where Poilievre fails to clear the 50% threshold in the first round. Is he at risk then? After all, that's what happened to Bernier and MacKay! The answer is not at all as he's actually at the center of this field and would therefore get a lot of second votes. See below my estimates based on the Mainstreet poll and the people who donated to more than one candidate:




As you can see, Poilievre has a lot of reserves among Lewis and Baber. He's also not doing that badly with Brown, Charest and Aitchison. I think it's fascinating how 16% of Charest supporters have Poilievre as second choice (and more if you include the 3rd choices of those who would go to Lewis) but only 6% of Poilievre supporters have Charest as second. It's similar to 2020 where MacKay's supporters had O'Toole as second choice but the opposite wasn't true.

Just to illustrate how dominant Poilievre is, in the eventuality of a last round with him vs Charest, Poilievre would win 79 to 21! Anybody saying Poilievre can only win if he wins outright in the first round is just being delusional. This isn't Bernier or MacKay.


Finally, here's the map per riding. I simply used my fundraising map and did a uniform swing to match the % above. I also redistributed the voters of Patrick Brown. Poilievre wins 308 ridings, Charest 18 (14 in Quebec), Aitchison and Baber each win their own. Lewis doesn't win a single one.

You can see the full map here.


Methodology and commentary


Leadership races are harder to project than regular elections. One obvious reason being the fact that only members can vote. It is therefore almost impossible to poll. However, we get a better metric: the fundraising data. We also have the data from Google Trends that has proven fairly accurate in the past (more on this later). So here are the four data sources I used:


1) The fundraising by riding. I created this map recently where I aggregated the unique donors by ridings. This is likely the best metric we have although it likely underestimates some candidates in some ridings (where there are very few donors). It's especially useful to measure the efficiency of the vote.

2) The fundraising at the provincial level. That's the metric I used last time. Yes it's technically the same data as 1) but it might provide for a less distorted version than the riding one. As I said, many ridings only had a few donors and it could amplify the number of points received by a candidate. Moreover, I used an average of the number of donors and the amounts collected. The former should be more accurate (and has proven to be in the past) but the latter can be a proxy for organization and who is supporting a candidate (Charest, for instance, gets a lot of high donations. While it's still one person, one vote, it could be the sign he has the support of older, richer people who are more likely to return their ballots). Amounts received really help Charest as he barely got more donors than Baber and fewer than Lewis but his average donation amount is much higher. This is particularly the case in Quebec. So see 2) as a slight boost to Charest's chances.

3) Google Trends, partially adjusted. In 2020, we would have got the following projections: O'Toole at 30%, MacKay at 27%, Lewis at 26% and Sloan at 17%. They actually received, respectively, 32, 34, 21 and 14%. So MacKay was underestimated, O'Toole slightly underestimated, Sloan and especially Lewis were overestimated. Still not a bad projection. This year, I used the data from Google Trends and I assumed that Poilievre was the candidate overestimated (a phenomenon that seems common to more right-wing politicians), Charest was underestimated and the rest was mostly okay (Lewis was overestimated in 2020 but she's doing a lot worse this year, so I didn't adjust her down). Even after adjustments, this is the metric that is projecting the biggest margin of victory for Poilievre. This is also the only metric that could be updated between now and September 10th. I haven't decided if I'll bother updating my projections.

4) That one Mainstreet poll done among CPC members. It was published all the way back in June and I have confirmed with them that they had the 2020 membership list, not the new one with 675k members. So it's not the most useful poll. It's likely a good estimate of the voting intentions among pre-existing members (that are also more likely to return their ballots).

I ultimately gave 25% weight to each metric (so the fundraising, combined, was at 50%). In 2020, I only used the fundraising and the Mainstreet polls. I had the following projections: MacKay at 37%, O'Toole at 30%, Lewis at 21% and Sloan at 12%. So it was pretty close with MacKay performing worse mostly because he failed in Quebec (he was projected to win it with 44% of the points but only got 34%). I did not introduce any of my own biases into the data analysis. If anything, whenever I made adjustments, they were justified and in favour of Charest.


There are people out there who argue leadership races are unpredictable. I strongly disagree. If we didn't have the fundraising data, maybe. But we do and it has proven to work remarkably well on multiple occasions. Also, more importantly, literally every metric we can use this year points at the same outcome: a first ballot victory. Anybody pretending this race is close, tightening or whatever is either spinning or terrible at data analysis (or both).


A few words about each candidate:

Poilievre. I don't have much to add. His campaign has been quite impressive, if maybe repetitive by the end. He has clearly decided that his path to power was by first uniting the right (read: getting the PPC back) and less by trying to convince so-called Blue Liberals swing voters. I have been clear that I believe there is a lot of truth there. I don't buy that there are a many Liberal voters who are unhappy with Trudeau and would jump to the CPC if that party had a 'fiscally conservative, socially progressive' leader. 2019 and especially 2021 have shown that the 32% or so of people who vote for the Liberals are very solid in their choice. Is it possible to get them? Maybe in the perfect conditions. Polls have usually shown that some LPC voters could switch if Charest was leader but not Poilievre (alternatively, the PPC would be in double digits territory with Charest as CPC leader). As I've said, I have my doubt that such a shift would survive a 35 days campaign. But anyway, we are years away from the next election (likely). As far as this leadership is concerned, Poilievre is the clear favorite. Despite what you might have read on Twitter or Reddit, he doesn't actually divide the party. All the polls have shown that very few CPC members or voters dislike him. If there is one divisive candidate, it's actually Charest (we got many polls showing that a large share of voters wouldn't support the party if he were leader). Looking at the second choices also show that: Poilievre actually has a lot of reserve of votes in Lewis and Baber. He's actually not even that far behind Charest among the Brown and Aitchison voters. There really only are a tiny minority that dislike him among conservatives. But this minority is very loud.

He's the only politician on the right to have a social media game. Many dumbasses have called him the Trump of Canada but I personally think that the only true comparison is that he's the Trudeau of the right. Young, incredibly popular among his base and hated by the other side. He signed up more members than the total the party had in 2020! His announcement video was watched over 2 million times. It's really something quite unique by Canadian standards.

Will he need to pivot after the leadership? Well... that's not his thing. I do think there will be a change in tone (somewhat) and focus. But don't expect Poilievre to all of a sudden not want to defund the CBC or not support the convoy. We'll see if he can assemble a coalition large enough to win in 2025. I personally think that his strategy should guarantee 140 seats. Beyond that? Well he'll need to do better in the GTA and/or Quebec. He'll need to do better with women (especially soccer moms) and people over 55+. This is where a small change of tone could do wonders. He has potential but it's too early to tell if he'll be successful or not in my opinion.


Charest. His campaign started very slowly and I strongly suspect he got dragged into this by bad advice from Boomers and the Quebec caucus. These people probably told him he'd be the savior of the party if he ran. Have you seen this video announcement? That was pathetic. His campaign is also plagued by incredibly incompetent people around him. When his co-chair is saying nonsensical stuff such as "freedom is a dog whistle" and she spends more time selling her book than her candidate, you know there is something wrong. I don't think Charest ever had a path to win but a better campaign could have made things more interesting. As it stands, Charest will be fortunate to finish above 20%. And outside of Quebec, he might well be fighting more with Baber than Poilievre.

I don't personally have any problem with Charest or his policies (as a matter of fact, I quite like both) but he and his team clearly made mistakes. They attacked Poilievre way too hard for supporting the convoy (a majority of CPC voters and especially members supported it). He put out, in my opinion, too many very detailed policies. A leadership race isn't the time for this, he'd have been better off just focusing on 2-3 big ideas.

He and his team also spent weeks gaslighting us about "it's not about the numbers, it's about the points". The fundraising analysis by riding showed that Charest will indeed likely get more points than votes (thanks to Quebec) but it's a difference of about 4%. The idea that 80k members (rough estimates from me) could be so efficiently distributed that they'd beat the 311k of Poilievre is ridiculous. Poilievre's vote is actually not very inefficient. Sure he 'wastes' some in Alberta in ridings with thousands of members but he's still winning everywhere. As a matter of fact, the Charest vote is only efficient at the aggregate level. What I mean by this is that the % of votes is actually higher than the % of points in most provinces, including Quebec (Charest's voters are too concentrated in Montreal and Estrie). His vote only looks efficient Canada-wide because a higher share of his voters come from Quebec. In any case, efficiency won't save Charest from a likely very humiliating defeat to end his career. But hey, the pundits and the media will salute the 'last principled conservatives' or something like that, so it's not all negative for him.


Lewis. She came so close to winning last time (she really only failed because she didn't speak French). It's quite different this time around. I think there are mostly two reasons. The first one is that while she can still count on the SoCons members, they represent a smaller share of the total. They have represented about 15-20% in 2017 and 2020 but that was with roughly 275k members, not the current 675k. On top of this, many of her voters in 2020 were libertarians, anti system, etc. Most of those voters have moved to Poilievre or Baber. She's also still terrible in debates and in French. All in all, she has no chance this year. She could, however, finish second. The fundraising data actually showed that her odds were quite good. Lewis is weird to me. One day she puts out a nice, compassionate email. The next day she spreads conspiracy theories. Not sure how Poilievre will take advantage of her in his shadow cabinet.


Brown. He's disqualified but let's talk about him a little bit. His campaign claimed to have signed 150k members but everybody with a brain knows it's BS. All the metrics we have put him below 10%, sometimes below 5%. And before one of you argue that his strategy (pandering to every racial and religious group out there to sign up new members) wouldn't register in such metrics, you are wrong. When he won the Ontario PC leadership in 2015, he actually won on Google Trends. He actually won the fundraising. So there is no reason to believe he was on track for more than 10% this year. I really wish he hadn't been disqualified. I wish the CPC will actually reveal the number of ballots cast for him (they said such ballots will directly go to their second choices).


Baber. Watch out for him in Ontario and BC. The fundraising data put him ahead of Charest in those two provinces. He knows he can't win this overall but his GOTV strategy has been pretty good and focused on "no vote splitting with ranked ballot", thus trying to get some Poilievre supporters to rank him first. Beyond this, I don't have much to say. His themes are quite similar to Poilievre's and so are his voters. As a matter of fact, there is clearly a right side to the CPC where everybody likes Poilievre, Baber and Lewis. The second choices and multiple donors showed that.


Aitchison. Is fighting with Charest over the centrist voters. I  personally think he ran an interesting campaign but leadership races are popularity contests and his style isn't very suited for that. He'll have increased his profile within the CPC and I guess that's all he wanted?

Note: I updated this post and map on August 6th. The preliminary version only included data from Q2, I have now added the fundraising from Q1 as well. As usual, if you want more frequent updates, you should follow me on Twitter.

The fundraising is by far the best metric when it comes to projecting leadership races. It beats the hell out of polling of voters (because members and voters of a party aren't the same). Usually, however, I used the data at the province level. But this year I decided to improve on it. Since the data about the donations from Elections Canada include the postal codes, I can map them at the riding level!

This is exactly what I did here. I also made sure not to count a unique donor more than once. Some people donate more than once and to more than one candidate. Still, they'll only have one vote. If someone donated to more than one candidate, I split their vote. For instance, someone who donated to both Poilievre and Lewis (the most common occurrence of double donations), I gave 0.5 vote to each in that riding. This is why the map shows some vote totals with decimals.

If you want the full screen version, click here. Remember that Tableau maps can be slow, so zoom in and wait 1-2 seconds. The name says Q2 but it does include Q1, I just couldn't rename the project.



So what are the results? Well Poilievre wins easily. He gets 70% of the votes (Canada wide) and 67% of the points. He also wins 320 ridings (note: technically there were two ties with Charest, in Beloeil-Chambly and Mount-Royal Given the regions, I gave the latter to Charest and the former to Poilievre.

Charest wins 14 ridings, 12 of which are in Quebec. The only CPC-held riding he wins is Richmond-Arthabaska, most likely thanks to Alain Reyes. Charest's vote is indeed efficient (as he likes to repeat it everywhere) as he goes 8% of the votes to 12% of the points. It's nice but ultimately in line with what I had been saying: the differences between the percentage of votes and points hasn't been that big during these leadership races. The biggest gap was 4 points for Lewis in 2020. So Charest is doing the same here. Nowhere near enough to win this thing. The fact he also only wins two ridings in the GTA (hilariously, University-Rosedale and Toronto St. Paul) really should put to bed the idea he's the candidate who can win this region. MacKay ran a bad campaign but he at least dominated in the GTA!

Look, I think Charest is a good guy, a principled conservative who would be a decent leader and PM. But his time has passed. His leadership campaign has alternated from the ugly (the first 2 weeks) to the good (his policies, performances during the debates). I suspect the issue might be his team who is really not advising well (most likely because they are a bunch of Boomers who don't understand the modern CPC). Ultimately, he's not even winning Quebec.

Lewis is the only one not winning a riding. People have been wondering how the candidate who almost won in 2020 (she won the popular vote!) could fall so much. I believe there is a fairly simple explanation. First of all, the SoCons members represented 15-20% of the membership when there were about 275k members. But now that there are 675k, her share has decreased. Also, many of her 2020 voters were libertarians, anti-system, etc. They all left for Poilievre or Baber. Still, she could well end up with more votes than Charest and finishing second in points isn't impossible (she has 10% of the votes and 9% of the points).

I included Brown since he was disqualified after Q2. His numbers don't look good. 2.5% of the voters, 3% of the points. His strategy clearly failed. He pretends to have signed up 150k members but I don't believe that. His numbers have been terrible in all metrics, from polls to Google Trends. Some will argue his strategy is different (basically signing up enough new members, mostly from religious and ethic minorities to overtake the other members). Sure. But if he were successful, we should still see some signs. When he won the Ontario PC leadership in 2015, he actually dominated on Google Trends and in the fundraising. But hey, he's winning two ridings in Brampton, the city he's the mayor of. I'm just glad his numbers are so low because I don't have to sweat about how to re-allocate them now that he's disqualified.

Baber might be the surprise of this race. a guy that was mostly unknown outside of Ontario 6 months ago is getting over 7% of the votes and points. He's actually ahead of Charest in Ontario (by a lot!) and BC. He's also winning his former riding. His message is often quite close to Poilievre's, so I do believe it's relatively impressive how he managed to be so high while fighting the Poilievre juggernaut. I wonder if he's jump into federal politics after this.

Finally, Scott Aitchison has been running a nice campaign with some great policies. But he isn't very successful. The moderate/Red Tories crowd is with Charest. He is winning his Ontario riding, so there's that. Scott didn't run to win this leadership but to increase his profile. I feel he was ultimately successful but I sure hope for him that he gets more than 1% at the end.

I'll use other sources when I make my final projections. Still, it's probably hard to beat a dataset of over 50k donors. Sure many ridings have low numbers of observations, but this is likely the most accurate map you can create right now. If I had to guess, I'd say the fundraising might be overestimating Poilievre a bit and underestimating Charest. But this map is just another example of a metric putting Poilievre well over 50% in the first round. At this point, anyone arguing that Poilievre will need multiple rounds isn't looking at the data correctly.