This Ontario election is turning out to be a real snooze fest. There is almost no movement in the polls. Whenever we think something is happening (OLP rising, etc), it gets cancelled out by the next 2-3 polls.

The result? Well, projections are super stable. The most noteworthy change of this update is the fact the Green are now projected to win a second seat in Parry Sound -- Muskoka. The lack of a Liberal candidate, the loss of a long term incumbent for the PC and the strong Green candidate are all adding up to the perfect condition for a stunning upset.

With one week to go, I don't really think there is any remaining uncertainty. If the progressive vote was to coalesce around one option, it'd have done so already. Sure they are a fair number of close races in Toronto and its suburbs (mostly PC vs OLP races) but even if the Liberals were to overperform there, Ford would likely still get his majority. Plus polls show that older voters, as well as those who are sure to vote, favour the PC more. That means the Conservatives could be underestimated.

So anyway, here are the projections

Polling average; Seat projections with 95% confidence intervals; Chances of winning the most seats

The map

The riding by riding projections

Ontario Proj May 25 by bryanbreguet on Scribd

Quick update to the projections after many new polls (Leger, Abacus, Mainstreet, Research Co.). We see a tightening of the race, although the PC is still virtually guaranteed to win the most seats. However, the chances of a majority are now down to 71%. Still good obviously but we are now getting to the point where a slight overperformance of the OLP in the GTA (or an underperformance of the PC) could cause a minority, in which case I suspect Del Duca would govern with the help of Horwath.

Polling average; Seats projections with 95% confidence intervals; Chances of winning the most seats

The North, which was once very promising for the PC has fully reverted back to the NDP. Toronto back to the Liberals and the NDP could well lose 2-3 of the DT seats very soon. The PC majority/minority will be decided in the GTA/905. As many as 12 close races where the PC is currently winning by less than 5%, most of them against the Liberals.

So, how could we end up with a minority? Well the OLP vote could just become very efficient and win those roughly 10 seats. The other solution would be for the OLP to 'brute force' it by simply getting closer to the PC province-wide. My estimates show that the PC majority is safe (i.e: over 50% chances of happening) as long as the PC keeps a lead of about 5.25 points over the OLP). It used to be that my estimates where that the PC could do it with a lead as small a 4-4.5. So we are definitely getting close to a too close to call territory. See below.

We'll see of the debate allowed Del Duca the pish he needs to bring Ford back below 62 seats. David Coletto from Abacus has hinted on Twitter that his early data was showing the opposite: Ford did well. So let's wait and see.

Finally, thanks to the lack of a Liberal candidate in the riding, the Green have a shot at winning Parry Sound - Muskoka. Still a long shot but not impossible.

Ontario Proj May 18 by bryanbreguet on Scribd

Here are my first projections for the Ontario election. Don't expect updates as regularly as during the federal election though. The reason being fewer polls being published. Also, Ontario has a tendency to have big fluctuations between polls (especially based on whether they are online or by phone). I therefore think there is a lot of noise in the data and updating too regularly amplifies this noise.

Anyway, here are the numbers, the map and the detailed projections. As usual, feel free to use the model and make your own projections.

Polling average, seat projections with 95% confidence intervals, chances of winning the most seats


Play with the map here (it's a little bit slow, give it a few seconds after zooming in or out). You can see that the extended GTA has most of the close races.

Riding by riding projections at the bottom of this post.

There is not much to say right now. Ford's PC has a huge lead and is almost guaranteed to win the most seats. The uncertainty is really only about getting a majority. See below for a deeper analysis on that front. The PC wins most rural seats except in the North where the NDP is down (quite significantly according to Mainstreet) but is resisting. There has been a trend going on in the North for years now where the NDP is seeing its support eroding and the Conservatives are rising. Will it be enough to flip many ridings? Right now, no, but it could certainly happen and that would be devastating for the NDP.

This is especially the case since the NDP isn't making up for the drop in the North by increasing in urban environments. It's the Liberals, Ontario's natural party for many, that is back ahead in Toronto. After a disastrous election in 2018, the Liberals have a new leader who has the charisma of a traffic cone but has been running an okay campaign so far (the few bad vetting of candidates notwithstanding). The Liberals being the main challenger to the PC has more to do with the NDP and Horwath being quite useless than the Liberals being that good though.

Will the progressive votes coalesce around one option, à la 2015 at the federal level? Maybe but the conditions are a lot less optimal. Del Duca isn't Trudeau, Ford isn't disliked like Harper (he was before Covid though). But there is certainly a chance where the PC falls short of 62 seats. It's currently unlikely but based on my analysis (see graph), we are only a few percentage points from such a situation. Indeed, if the PC's lead over the OLP falls to 4 points, then it's 50-50 for a majority. In that case, it would come down to the close races in the GTA.

Detailed projections (yes I know the Liberals are running only 122 candidates, I'll make the adjustments later).

Ontario Proj May 13 by bryanbreguet on Scribd

 Please participate if you watched the debate. Please answer both questions. If you don't see the polls, disable your ad blocker.

Polls PollMaker

 Please give your opinion below.

If you only want to see the results, click on the link and you'll have the option.

Poll Creator

While I offered a first look a few weeks ago, I hadn't made any projections until now. The reason being that we simply didn't have good enough data to even try. But this changed yesterday with the release by Elections Canada of the fundraising numbers from the first quarter.

Fundraising is by far the best indicator of where a leadership race stands. It isn't perfect and surprises can still happen but it beats polls and other measures. And despite what some like to pretend, it is absolutely possible to project leadership races -- it's just that there is more uncertainty. I have been quite successful with the last two CPC races (although I did have Bernier winning in 2017, mostly because Scheer ended up winning more points than expected in Quebec) as well as some of the PQ races.

So I took the data and did the following. I calculated the percentages per province (or regions for the Atlantic and the Territories) of the total donations (amounts) and the number of donors. The latter is usually a better predictor (it's one member, one vote ultimately, which means that it doesn't matter if a member gave $100 or $1,000, it's still only one vote at the end. And yes I'm aware of the electoral system but that's another story, see below). I then did a weighted average of the two with a 75% weight attributed to the number of donors and 25% to the amounts. While not simply use the number of donors only? Because having big donors, like Charest, can be a sign of a strong organization and it can help. So I want to use both. Ultimately I'd like to also mix in some polls but we haven't had good enough ones so far. Once I had the percentages per province/region, I assigned the number of points (so if you are at 50% in Quebec, you get 3900 points).

Here are the results for the first round:

Poilievre is easily ahead. Which isn't surprising, whether you use the fundraising data, endorsement, the polls or Google Trends, Poilievre is always ahead and close or above of a first-round win. If the election was tomorrow, there is little doubt Poilievre would win. His position also appears (at least to me) more solid than Bernier's or MacKay's. What I mean by that is that Poilievre is liked (loved) by many in the CPC. He's not leading because nobody better is running. He's leading because that's the leader most want, period. The one province where he likely needs to work harder (and organize better) is Quebec. If he could do this, he'd win in the first round easily.

Yes we should include some margins of error. I'll do so by looking at the accuracy of the projections with the results in 2017 and 2020. But I didn't have time today to do so (I was in an economic articulation meeting all day).

Charest is doing quite well if we only look at the overall amounts collected ($488k to 545k for Poilievre). But Charest achieves this only thanks to a lot of large donations. While Poilievre averages $163 per donation and has 3,336 of those, Charest is at $855 in average but only has 571 donations (note: those are donations, not single donors. There as well I'll need to clean the data in the next few days). Moreover, Charest is really only competitive thanks to Quebec where he represents 60% of the donors and 90% of the amounts collected! Outside of Quebec, Charest is more often than not a distant third, including in Ontario where he only has 7% of the donors. It gets worse as we move West.

Lewis is exactly where we expected her to be. As the only SoCons candidate, she is pretty much guaranteed 15-25% of the votes. Unfortunately for her, there isn't a Sloan candidate whose voters could then help her climb higher this time around. I suspect Baber's voters might prefer Poilievre or Lewis over the other two but it's still wouldn't be enough for Lewis to win. However, it is absolutely possible that Lewis would make it to the last round versus Poilievre. In which case I really wonder what would happen with the second-choices of Charest's voters. I suspect many wouldn't have ranked Poilievre and even less Lewis, so many votes would likely be dropped.

Aitchison is doing what we would expect (he had a hard time raising the 300k necessary to qualify) while Baber is doing better than I thought, but I guess he has a certain strong following and fundraising could overestimates him.

I want to talk about Brown and the really weird campaign he's running. We know, from how he won the Ontario PC leadership, that he's supposedly good at signing up members, especially from ethnic and religious minorities. But so far, we have no evidence that he's successful. At some point, if he's really signing up a ton of members, we should see some evidence! Polling numbers, fundraising, etc. So far, Brown is a complete failure. Also, it's worth pointing out that his campaign is super weird, to the point that it feels phony at times. His Twitter accounts has over 60k followers but his tweets usually get 20-50 likes. He has not given a single interview to traditional media. And the one interview we have is full of red flags. He shows clear contempt for the current CPC base and is openly talking of a takeover. Many of his oppositions, including on Israel, are completely incompatible with the CPC. His average donation is close to $1,600, by far the highest among the six candidates. This doesn't really scream grassroot support among ethnic and religious minorities to me. I honestly don't know what he's doing. All I know is that so far, based on every metric we have, he's below 5%. I originally thought he'd be the main challenger to Poilievre but I have fully changed my mind on this. I also maintain what I said: if Brown were somehow successful, the caucus wouldn't accept him. They'd likely find a way to get rid of him.

Finally, yes the CPC uses a weird electoral system where each riding is worth 100 points (unless that riding doesn't even have 100 votes, in which case the total number of points is equal to the number of votes; This is a situation that could mean a few ridings in Quebec). So projections like these ones aren't able to capture the possible impact of where the votes are. We know, for instance, that the Bernier vote was more efficient than Scheer's as Bernier was popular with urban voters (it was in 2017), so winning ridings with fewer members. We saw the same phenomenon with MacKay. So Charest might be slightly higher than what you see here but he's still closer to Lewis and fighting for second than of first place.

Alright, that's all for this very early and rough projections.

Ce dimanche, les électeurs Français se rendent aux urnes pour le premier tour de l'élection présidentielle. Il y a 12 candidats en lice, allant de l'extrême droite à l'extreme gauche. Les deux candidats qui arriveront en tête seront qualifiés pour le 2e tour, deux semaines plus tard. Il n'y a pas actuellement de grand suspense quant à ce top 2, le président sortant Emmanuel Macron et Marine Le Pen, chef du Rassemblement National (anciennement le Front National) devraient normalement terminer en tête. Il y a une petite chance que le chef de la France Insoumise (très à gauche), Jean-Luc Mélenchon, crée la surprise mais les probabilités sont relativement faibles. On est loin de l'incertitude de 2017 où 4 candidats avaient de vraies chances de se qualifier.

Par contre, la grande différence concerne le 2e tour. Alors que 2017 avait vu Macron facilement remporter l'élection avec plus de 66% des voix, les sondages montrent que Marine Le Pen se rapproche du président sortant, elle est parfois aussi élevée que 48.5%!

Dans ce court billet, je présente mes projections actuelles pour le premier tour ainsi que le modèle utilisé. Ensuite, je m'essaie à projeter le 2e tour, une pratique plus incertaine car il y aura deux semaines de campagne et un "Front Républicain" (où tout le monde fait barrage à l'extrême droite) pourrait apparaître. Ce barrage est moins probable qu'en 2017 (ou 2002) en particulier car Le Pen a réussit à adoucir son image. Elle a aussi fait une campagne beaucoup plus centrée sur le pouvoir d'achat au lieu de parler de la sécurité et de l'immigration (Zémmour l'a remplacé sur ce créneau et lui a permis d'apparaître moins extrême). On voit aussi que Le Pen peut maintenant espérer aller chercher un nombre significatif de votes chez Pécresse et même Mélenchon. Cela peut sembler un peu bizarre que la droite dure partage des électeurs avec la gauche dure, mais les deux sont populistes et séduisent les jeunes. Aussi, Le Pen n'est pas vraiment de droite économiquement.

Le 1er tour

Mes projections, basées sur 50,000 simulations:

L'évolution des sondages:

Source: Wikipédia

L'avance de Macron est relativement significative et il devrait terminer en tête malgré une tendance à la baisse depuis quelques semaines (l'effet Ukraine s'estompant). Il faut savoir que les sondages Français sont particulièrement précis, bien davantage que les sondages Canadiens. La moyenne des sondages n'est en général pas éloignée des résultats par davantage que 1-2 points. La plus grande erreur observée était en 2007 avec une sur-estimation de Jean-Marie Le Pen de 3.5 points (probablement une surcorrection des sondages qui avaient sous-estimé Le Pen en 2002 par 3.5 points, ce qui avait permi à ce dernier de terminer 2e juste devant Jospin qui avait été, lui, surestimé de 1.8 points). En 2017, les sondages avaient eu tous les candidats dans une marge de 1 point, ce qui est extraordinaire. Tout cela pour dire qu'une avance de 3-4 points dans les sondages devraient normalement se réaliser.

Le modèle inclut de l'incertitude de plusieurs manières. Tout d'abord, il y a naturellement la marge d'erreur des sondages. En utlisant les sondages de 2002 à 2017, j'ai calculé une erreur moyenne de 1.3 points de pourcentage pour les candidats principaux (top 5). Cela correspondait à une marge d'erreur effective de 3.4% 95% du temps. Je fais ainsi mes simulations avec une taille d'échantillon de 545 afin d'avoir les bonnes marges d'erreur.

Il y a aussi la possibilité que les électeurs changent d'opinion. Pratiquement tous les sondages incluent une question sur la sûreté du vote. À ce jeu-là, Macron et Le Pen font en général le mieux avec plus de 80% de certitude, alors que Jadot ou Pécresse traînent derrière. Je pars du principe que la moitié des indécis votera finalement pour leur candidat favori mais l'autre moitié est redistribuée selon les 2e choix. J'ajoute un peu d'incertitude en faisant une distribution du nombre d'indécis. Par exemple, si (disons) 12 % des électeurs de Mélenchon sont incertains, il y a des simulations où c'est en fait 16% et d'autres où c'est seulement 8%. Inclure cette incertitude a un effet important sur mes intervalles de confiance. D'un côté on pourrait argumenter que cela ajoute trop d'incertitude compte tenue de la fiabilité des sondages Français, mais mes intervalles ressemblent assez à ceux publiés dans certains sondages (comme ici, page 34). Au final, je préfère inclure un peu trop d'incertitude que pas assez.

Le 2e tour

Gardez en tête que projeter le 2e tour avant même le 1er comporte beaucoup d'incertitude. Pour mes simulations, j'utilise les résultats simulés du premier tour et je redistribue les votes selon les reports indiqués dans les sondages. On y voit par exemple qu'une large majorité des électeurs de Jadot iront vers Macron alors qu'une majorité des électeurs Zémmour se rapporteront sur Le Pen. Cela devient plus intéressant lorsque l'on regarde Pécresse où ses électeurs semblent favoriser Macron mais une fraction non négligeable pourrait aller vers Le Pen (actuellement j'ai 45% qui iraient chez Macron et 27% chez Le Pen). Quant aux électeurs de Mélenchon, ils se diviseraient 33-27 en faveur de Macron! Pour Le Pen, si elle veut remporter le 2e tour, il lui faudra aller chercher un peu davantage d'électeurs chez Pécresse et Mélenchon. Je suis vraiment intéressé de voir si ce dernier donnera une consigne de vote!

À noter que je ne fais les simulations que dans le cas où Macron affronte Le Pen. S'il devait affronter Mélenchon, il gagnerait beaucoup plus facilement selon les sondages.

Au final, j'ai Macron toujours favori mais c'est bien plus serré qu'en 2017. Nous verrons si Le Pen continue sa bonne campagne ou s'effondre comme elle l'avait fait en 2017 (elle avait eu, en particulier, un très mauvais débat télévisé). À remarquer que les chances de gagner ici sont pas mal les mêmes que les prédictions des marchés de paris.

Je posterai une mise à jour final vendredi ou samedi.