We finally got multiple polls from firms with online samples (Angus-Reid, Léger, Abacus and Research Co.). While Léger was quite favourable to the Tories (lead by 4 nationwide, ahead in Ontario, Bloc strong in Quebec), Abacus, Ekos and Research Co all had good Ontario numbers for the Liberals.

The projections have shifted quite a lot but it's a little bit misleading. The Liberals are projected at 155 seats, yet only 14.5% of a majority. That seems weird but it's because they are currently ahead in many close races. In Quebec only, the Liberals are projected to win 8 seats by less than 5%. If they drop a little bit (or if their vote isn't as crazy efficient as in 2019), those 8 seats could flip really quickly.

Maybe the best way to illustrate this is to see that while the Liberals are projected with 155 seats, the average of the simulations puts them at only 145. So the race hasn't shifted that much with the polls today. Something to keep in mind as there is still considerable uncertainty at this point. Maybe presenting the average of the simulations (like what 338Canada does) would make more sense than doing the polling average and looking at the outcome with those numbers specifically. Oh well.

Anyway, here are the projections:

Map

Full page here.

Detailed projections

Proj Canada 31 August 2021 by bryanbreguet on Scribd

If you are looking for the most recent projections, they are here.

Le Québec bouge souvent beaucoup lors des élections. Que ce soit la hausse de Stephen Harper en 2006 (lorsqu'il avait fait passé son parti d'essentiellement zéro à 25%), la vague orange en 2011, la vague Trudeau en 2015 ou la forte hausse du Bloc en 2019, il est assez évident que les Québécois ont l'habitude de changer leurs votes.

C'est pourquoi il est un peu bizarre de constater que le Québec est la seule province qui n'a à peu près pas bougé depuis le début de cette campagne. Alors que l'Ontario est passée d'une avance Libérale de 10 points à une égalité statistiques ou la CB qui a vu les Conservateurs prendre une avance importante après avoir été derrière tout l'été, le Québec est essentiellement constant.

Il faut dire que le Québec n'a pas tendance à bouger au tout début, ça prend du temps. Il est ainsi raisonnable de penser que les Québécois commenceront à prêter attention à cette campagne cette semaine, surtout avec le débat à TVA ce jeudi, le fameux Face à face. Un format de débat qui a séduit par le passé et qui attire une grande audience.

Le but de ce billet est de démontrer que les Québécois sont assez "fluides" entre les principales options. Voyez par vous-mêmes les 2e choix en 2019 ci-dessous, estimés via l'Étude Canadienne Électorale.

Les électeurs Bloc en particulier semblent très partagés dans leur 2e choix. Si Erin O'Toole veut aller chercher des électeurs supplémentaires, il semble qu'il devrait regarder auprès du Bloc mais aussi chez les Libéraux.

Peut-être plus important pour les Conservateurs, voici les 2e choix mais hors des grandes villes et banlieues:

On voit que près d'un quart des électeurs Libéraux avaient le PCC comme 2e choix. Chez le Bloc, les Conservateurs sont maintenant le principal 2e choix (de peu, il est vrai).

Imaginons le scénario parfait pour O'Toole qui irait chercher ses électeurs. Son parti serait à 33% hors des centres urbains. Ce serait suffisant pour être en tête dans cette partie du Québec. Ça ne lui donnerait pas des sièges à Montréal mais il remporterait largement plus que 10 sièges au total.

Je ne dis pas ici que ce scénario est probable, il ne l'est pas. Mais les Québécois sont connus pour être volatiles et leurs 2e choix montrent (en partie) pourquoi. Une mini vague Conservatrice n'est de loin pas impossible au Québec. Il a positionné son parti beaucoup plus proche d'une CAQ fédérale.

De manière plus générale, le problème pour le Bloc est que ses électeurs peuvent s'en aller de tous les côtés. Cela explique possiblement pourquoi ce parti est en baisse dans les sondages par rapport à 2019. Blanchet a positionné son parti plsu au centre que Duceppe et sa formation est ainsi une sorte de grande coalition avecd es électeurs de gauche et de droite. Ça lui donne un grand potentiel mais aussi un plancher assez bas si les choses vont mal.

Au final, je crois que le Québec va commencer à prêter attention à cette campagne cette semaine. Et je serais surpris si nous ne verrions pas du mouvement dans les deux prochaines semaines. La tendance hors Québec indiquerait qu'une montée PCC est probable mais le Québec aime bien faire les choses à part. Pourrions-nous voir une poussée NPD, à la Québec Solidaire en 2018? Ou alors les Québécois vont décider qu'ils aiment Trudeau et ils vont lui donner sa majorité? À suivre.

Reminder: you can use the model with your own number here. And follow me on Twitter if you want more regular updates.

Short post where I just want to provide the most up to date projections. It's almost a 50-50 race at this point. If we get a full polls next week (likely on Tuesday) from Léeger, Abacus or Ipsos also showing the Tories ahead (like it has been the case with Mainstreet, Ekos and Nanos), then I expect the projections to flip and have the Conservatives ahead.

Map

Remember that it's a live map, so it might not match the content of this blog if you read it after August 29th. Full version here.

Detailed projections

Proj Canada 29 August 2021 by bryanbreguet on Scribd

Here below are the most up to date projections based on the polls published last week. As you can see, the Liberals are holding up maybe better than what you could have thought after reading some stuff on Twitter. But it could all change pretty quickly if today's polls from Ekos and Mainstreet are right. Both are showing the Conservatives well ahead nationwide (by like 5-6 points) and, if they are right, the Tories could well win more seats (still short of a majority although that possibility is now a lot more real than previously).

Anyway, here are the projections. Below them, I provide a short analysis of how close the Liberals are to losing a lot more seats (hint: it involves the GTA and Quebec. Yes, shocking!).

The map:

Full page version here.

And the detailed projections with chances of winning.

Proj Canada 27 August 2021 by bryanbreguet on Scribd

If I only used today's polls (Ekos and Mainstreet as well as Nanos who is less bullish on the Tories than the other two), I would actually get 138 CPC, 147 Liberals, 29 NDP,1 Green and 20 Bloc. The chances of winning the most seats would be, essentially, 40% for the Tories and 60% for the Liberals. The Conservatives would even have a 9% chance of a majority! Yes, what was completely unthinkable 2 weeks ago is now at 1 in 10 chances. It's pretty crazy. I think we are one debate win away from genuinely discussing an O'Toole government.

Looking at the relationship between popular vote (Canada wide) and the number of seats, we see that Trudeau at least isn't close to a complete collapse. But we also see that the 170 magical number for a majority is quite elusive unless the Liberals can get to around 36%.

The one reason the Liberals are still holding up is mostly because they are still ahead in Quebec and Ontario. I'll leave the former aside as it's mostly against the Bloc and it isn't as crucial as the latter (it should be noted, however, that the Liberals are at risk of dropping by a good 10 seats if they were to drop a tiny bit in the popular vote there). The graph below shows how the Canada-wide number of seats is heavily a function of the popular vote in Ontario. The seat heavy GTA can obviously tilt the scale on its own and we are starting to see a lot of races and flips there. Earl Washburn from Ekos, who is usually very good at seat projections, have noticed this as well in their data.

Please note that the relationship here does also depend on the across provinces correlations. In other words, the model assumes that if the Liberals are doing worse in Ontario, they'll likely do worse in other provinces. I thought I'd mention it because no, a 20 points swing in Ontario alone (from +10 to -10) can't really cause the Liberals to drop by 70 seats. I should have kept the numbers in other provinces constant but that was extra work. The point was more to give a general rule of thumb of when Trudeau would be in real trouble than a true causal effect.

If the Liberals can't regain a commanding lead in Ontario (like the +10 they had 2 weeks ago), a majority is quite unrealistic. And, as you can see, should the Tories take the lead (so on the left-hand side of the vertical axis), then the Liberals would actually be in danger of not winning the most seats. A 'simple' 4 points lead for the Conservative in Ontario would basically put the two parties tie. The Trudeau would absolutely need the Bloc to remain in the 20s or he'd in serious risk of finishing second. And yes, I'm aware that O'Toole finishing first doesn't guarantee him power.

The next few days will be crucial. The Liberals need to stop the bleeding. And fast. They are actually fortunate Quebec doesn't seem to have been paying attention so far, I suspect things could get much worse once Quebec does.

A lot of polls today (Abacus, Léger and Ipsos on top of the usual 3 trackers from Mainstreet, Ekos and Nanos). The overall trend is very clear and favours the Conservatives of Erin O'Toole, in particular in Ontario and BC. The NDP doing quite well also, especially in the West where they are slowly replacing the Liberals as the main progressive alternative.

In Quebec, it's still mostly stable but I doubt it'll remain like this for the next two weeks (remember, the French 'debate' (called face to face) on TVA is on September 2nd and it attracts a huge audience, probably more than the official leaders' debate.

So let's take a look at the projections:

The map (note: it's interactive and up to date, so if you're reading this article days from now, the map won't match the rest of the projections)

Full page here. Hold 'Shift' and use the left-click to move around.

In Quebec, voting intentions haven't changed much. The Bloc is still down compared to 2019 and the Liberals are making a few gains. However, there are quite a lot of races the Liberals are currently winning by a margin of less than 5% (8 to be exact). Any drop in Quebec and/or increase of the Bloc would make those seats flip. And if it happens, Trudeau would have almost no way to a majority based on the current polling numbers. Please note that I am not saying Trudeau can't come back during the next four weeks.

Let's also remember the Liberals had an incredibly efficient vote in Quebec in 2019. So it's well possible that some of those close races have already flipped.

In Ontario, the Tories have been surging in the last few days. Not enough to be ahead in the polling average (although a few polls do have them ahead, including Ekos and Mainstreet). Compared to 2019 when the Liberals won the province 42% to 33%, the current situation (36% to 35%) is basically turning half of the GTA into a race.

If the Conservatives can indeed win Ontario, then we can start genuinely discussing the possibility of Erin O'Toole having the most seats at the House of Commons. Would that be enough to become Prime Minister? Maybe with the Bloc but it's too early to tell.

In the Prairies and Alberta, the only interesting ridings are the urban ones. With the Tories being down generally there (ironically it's making their vote more efficient as they waste less of it in the rural ridings in Alberta for instance), both the Liberals and NDP can expect to make a few gains. This is also a region where the PPC and Maverick could do well and I might have to adjust my projections because of that. More generally, if the PPC voters ultimately go back to the CPC, at least in urban ridings in the West, that could create an ideal situation for the CPC where it could still win almost every single seat (the vote splitting in rural ridings won't be an issue).

See Manitoba here

And Calgary and Edmonton here:

Yes Trudeau would win seats in Alberta again after being shut out of it in 2019.

Finally, BC has seen the biggest increase for the CPC. To be fair, it might be a regression to the mean after some weird polls in July placing the Tories as low as 20%. The Lower Mainland is full of 3-way races and the Liberals will either need o get back up or will require their incredible vote efficiency again.

Alright, so all in all, we now have a situation where the Conservatives winning the most seats is almost as likely as the Liberals winning a majority. This is a very different situation from just one week ago. We'll see if O'Toole can continue his good campaign and if Trudeau can wake up. As for Singh, the Angus-Reid poll clearly showed that people like him and liked his campaign. If he can establish himself as the progressive alternative in the West, that could be huge for his party.

Detailed projections:

Proj Canada 24 August 2021 by bryanbreguet on Scribd

Remember that you can use the model yourself here, with your own numbers. You can also follow me on Twitter for more regular updates.

This election has so far seen BC going back to 'normal' with the Tories back in first place (after polling terribly low for a while last month) and we are starting to see movement in Ontario. Quebec hasn't really tuned in as voting intentions have, so far, been stable. La Belle Province is actually saving Trudeau by providing him with the few gains he'd make, mostly thanks to a drop of the Bloc.

So here are the projections:

The Liberals are currently projecting to win a significant number of close races. Should the polls slightly overestimate this party and Trudeau could well win fewer seats. The average of the 5,000 simulations has the Liberals at 161 seats. The actual probabilities of a majorities are now only 38%, stable over the last few days but a sharp drop compared to 2 weeks ago. At 11% chances for a Tory plurality, progressive voters likely don't need to worry about vote splitting (yet?).

Remember that outside of Quebec, the LPC is not really making gains and is even losing seats in Ontario. All the potential gains in the West have evaporated in 1 week and Trudeau must be hoping for Quebeckers to keep not paying attention. If the Bloc were to climb back above 30%, a majority would become quasi impossible. Keep an eye on Ontario in the next few days, this could become really interesting. It's also pretty logical that O'Toole would be doing better in Ontario as 2019 was marked with the Ford effect.

Map:

Full page map here. Hold 'shift' and use the left-click of your mouse to move the map. Performance from Tableau isn't the best, so allow 1-2 seconds if you zoom in or out.

Detailed projections:

Proj Canada 23 August 2021 by bryanbreguet on Scribd

After almost a week of campaigning, polls have moved. The Liberals aren't as high as they were before and their 'guaranteed' majority is anything but. The Conservatives have climbed back in the West, especially in BC, while slightly closing the gap in Ontario. Frank Graves from Ekos even said that his new numbers, that will be published today, have the CPC ahead in Ontario!

In Quebec, we are seeing signs that we could have a four-way race with all 4 parties between 15 and 30%. And Atlantic Canada has seen a massive drop of the Liberals.

With that being said, the Liberals are still polling, in average, 9-10 points ahead of the Tories in Ontario and by the same margin in Quebec over the Bloc. The model therefore still sees them in majority territory as opposed to other models. Please keep in mind that as opposed to 338, the topline numbers are not averages. This is why my seat numbers don't have decimals. Yes I run simulations to get ranges and probabilities, but the topline numbers are the results of the average polling numbers only. This explains why I have the Liberals at 174 but the chances of a majority are only 43%. What it means is that if the polls are spot on, Trudeau likely wins a majority. But if the polls are a little bit wrong or if the Liberals are just slightly less efficient, they win a minority. To give a specific example, my 'official' projections have the Liberals at 174 but the average of the 5,000 simulations only gives them 165 seats. It's just two different approaches and there is no right or wrong. But I wanted to point it out to explain the numbers.

I think we are seeing that people are paying attention (Google Trends shows that as well) and things can definitely move fast. O'Toole is definitely performing better than expected while the Liberal campaign has not been at its best. As for Singh, he is, as expected, quite good on the campaign trail. We have known this since 2019.

So it is possible that things will change quickly, we'll see. But as it stands, with the polling average, I still think the Liberals are clearly ahead and either winning a majority or close to it.

Map

Full page version here. Hold 'shift' and use the mouse left-click to move around. Give it 1-2 sec after zooming in or out, performance isn't the best.

Detailed projections

Proj Canada 21 August 2021 by bryanbreguet on Scribd

Ce matin, le plus récent sondage quotidien (fait sur trois jours) de Mainstreet a un résultat assez époustouflant: le Parti Conservateur d'Erin O'Toole est 2e au Québec! Je ne peux pas révéler tous les chiffres car ils sont techniquement derrière un mur payant. Mais disons que le PCC fait aussi bien qu'en 2006 alors que le Bloc chute de beaucoup par rapport à 2019.

Je ne me prononcerai pas sur la vraisemblance de ce sondage. Avec une taille d'échantillon de 351, les marges d'erreur sont naturellement élevées. Aussi, soyons très clair ici, les autres sondeurs n'ont pas le PCC aussi élevé. Cependant, les sondages ont montré une certaine affinité entre les électeurs Bloc et Conservateurs. Ils sont plus agés, plus masculins et plus ruraux. On a souvent vu, comme dans le sondage Angus-Reid d'hier, que les électeurs du Bloc ont souvent le PCC en 2e choix. Donc l'idée que ce parti se retrouve entre 20 et 30% n'est pas complètement irréaliste.

Au lieu de débattre de savoir si les Conservateurs sont vraiment si élevés, essayons plutôt d'analyser qui risquerait de perdre le plus de sièges si O'Toole devait réussir à amener son parti vers les 29-30%. Le graphique ci-dessous vous montre les résultats des 5,000 simulations.

On remarque que le PCC a une zone assez plate où son nombre de sièges ne change quasiment pas entre 15 et 25%. C'est assez logique avec notre mode de scrutin mais l'effet est renforcé pour cette formation au Québec en raison de la concentration de son vote. En gros, le PCC est assuré de gagner 10 sièges mais il lui est difficile d'en gagner plus que 15.

Normalement la courbe commence à être plus pentue vers les 25% mais pour le PCC, on voit que c'est vraiment plutôt vers les 30%. En d'autres mots: une poussée conservatrice au Québec n'apporterait probablement pas davantage que 20 sièges à Erin O'Toole. Pas à moins d'avoir une vraie vague.

Remarquez qu'O'Toole serait ravis de remporter 15-20 sièges!

Et qui perdrait alors? Le graphique suivant nous donne une réponse très claire:

Les Libéraux ne sont pas vraiment affectés par une montée du PCC entre 20 et 30%. Au contraire, le parti de Trudeau pourrait en fait en profiter car une montée des Conservateurs ferait baisser le Bloc et ouvrirait la porte à des victoires Libérales dans le 450. C'est vraiment remarquable de voir le PLC gagner 40 sièges peu importe que le PCC soit à 20 ou 30%.

Le Bloc souffrirait beaucoup. Et c'est assez logique, une poussée conservatrice ferait juste étendre la base PCC actuelle. Regardez la carte ci-dessous qui vous montre les projections basées sur les chiffres du Mainstreet:

Le PCC remporte 15 comtés. Dans la ville de Québec, il gagne Beauport--Limoilou, il gagnt aussi Jonquière et le Lac-Saint-Jean. Aussi trois-Rivières. Les Conservateurs sont dans la courses en Abitibi et Drummond et même dans Sainte-Hyancinthe! Si le PCC devrait gagner encore 5%, il remporterait ces comtés et étendraient la partie bleu foncée. Tous ces comtés (ou presque) sont au Bloc actuellement, pas aux Libéraux. Ces derniers ne commencent à perdre des sièges (Abitibi, les deux à Québec, Compton--Stanstead) que si le PCC s'approchent des 30% au niveau provincial.

En conclusion, une poussée conservatrice ne fournirait pas nécessairement un nombre important de nouveaux sièges à Erin O'Toole. Les gains se feraient cependant aux dépens du Bloc. Justin Trudeau doit même espérer une petite montée qui diviserait le vote PCC-Bloc.

The Leaders' debates commission changed the rules to be invited to the debate this year compared to 2019. A party can qualify 3 ways. Either it elected one MP (under the actual color of the party -- no crossing the floor switcheroo), or the party got at least 4% of the vote last time, or the party is polling at 4% or higher 5 days after the beginning of the election.

For Bernier and the PPC, the last criterion is the only way to be invited. I looked at the detailed rules about which polls would be considered and how they'd average them. Based on everything I read, I don't believe Bernier will qualify. It'll likely be close but not that close. and that might be a good thing because there could easily have been a controversy (more on that later).

The rule is that only polls from reputable firms (it's not the exact wording but that's the spirit) will be included. Those polls have to have been conducted and published between 9 days before the election started and 5 days after. That means polls done and published between August 6th and 20th. Those polls also need to explicitly include the PPC as an option (can't simply be 'other') and need to have at least 1,000 respondents (that's where it could have become tricky). The commission will take a simple average (no weights or anything). Pretty straightforward, right? Well sure, except I'm not sure how the commission will deal with trackers like Mainstreet or Ekos where they do a 3-days rolling average (that means they published one fully new poll based on a fully new sample every 3 days). Do they include one of those polls every 3 days? To qualify, a party has to be at 4% in average, unrounded.

So, which polls are included and where does the PPC stand? Well here are the polls published during that period as of 11:30pm Pacific time on the 18th.

[Update]

I have been informed by the commission that I missed a detail: only the latest poll from a firm will be included. So that really solves the whole 3-days tracker thingy. I don't want to change the entire post now, but here is the updated calculations, including the latest number from Mainstreet. As you can see, the PPC is even further away from qualifying right now and it would need to increase a lot in the Ekos/Mainstreet that will be published on Friday. And even there, the Ekos would need to have a 1,000 respondents, which hasn't been the case until now.

[Back to original post]

Please notice that I'm including all the daily trackers from Mainstreet and Ekos. This is wrong since those polls share some of the same samples. But as I've said, I'm not sure which days should be selected (like do I take the Mainstreet from Monday and Thursday or Tuesday and Friday?). Also, Wikipedia didn't list the PPC score of some of these polls (like Angus-Reid) because the info had to be found in a detailed pdf, so I added it.

The PPC is not at 4% but not far behind at 3.71. But here's the kicker: some of the best polls for the PPC might not count: the Ekos trackers. Why? Well they are just below 1,000 sample size!!

Imagine if, by the 20th, the difference between the PPC being at 4 or not is... including the Ekos trackers (one every 3 days) or not! I mean, I know rules are rules but wouldn't it raise some questions? Why were the Ekos polls just below 1,000? And given that methodology (IVR versus online) seems to be the main factor explaining a high or low PPC, should we really care that much about sample size? Shouldn't we care more about doing an average with an equal number of methodologies? Also, just to add some spiciness to is, the commission noted that the Conservatives complained about Ekos being biased against them. So is Ekos purposely using samples below 1,000 as to make sure they aren't the ones qualifying the PPC? Yes, conspiracy theory time! Follow me on Twitter to know what I think of 9/11! (It's a joke, please keep reading)

Okay, so let's try to forecast where we'll be by Friday, the 20th. Mainstreet and Ekos will have posted 2 more updates. Let's also assume no other firm will publish. It seems most of the major ones (Abacus, Ipsos, etc) have published recently and I can't imagine they'll want to pay for another poll by Friday. As for Nanos, their tracker is supposed to start this weekend (again, why? Is it to be sure not to be included in this average? Yes, more conspiracy theories!).

I'm going to assume the commission will want to use the most recent trackers (so the numbers published on Friday). Then, since it's a 3-day rolling average, they would have another poll from each firm, the ones published yesterday (field dates of 14 to 16 while the numbers of Friday will have a field dates of 17 to 19). That means that we need to remove the Mainstreet/Ekos polls with the last dates of 15th and 17th in the table. Then we obviously can't do the full calculations as we don't know which score the PPC will get on the Ekos/Mainstreet published on Friday. Let's assume it'll be the same numbers (despite what appears to be a downwards trend). So I'm adding two polls with the PPC at 4.2% and 5%. That gives us an average of 3.53%.

Now what happens if I exclude the Ekos numbers, as per the rules? The PPC average drops to 3.36%.

So, right now, the fact that Ekos has sample sizes below 1,000 is NOT causing any problem. To qualify, the PPC needs two more polls with the party at 9% (assuming no other polls are published). I can't see it happening.

Bottom line: Bernier doesn't look like he'll go to the debate. But he'll likely miss the mark by very little. And the PPC would be even closer to the 4% mark if we only used polls conducted after the election started. It's mostly food for thoughts at this point though. The cold hard truth for Bernier is that he'll likely not be invited this year. Fair is fair and the bar wasn't set particularly high at 4%. Still, it does seem that online polls (Leger, Ipsos) will have cost Bernier his spot. If we have had more IVR polls, I believe Bernier would have qualified (also, the fact that Mainstreet didn't include an explicit option for the PPC in its poll on the 11th could also prove costly!). For a polling nerd like myself (and you, let's face it, you wouldn't be on this blog otherwise), this is super interesting stuff.

I doubt the commission will change its rules but there is a case to be made that Bernier and his party are actually at 4%. But, does Bernier really want to go? He'll say yes but I kind of believe this is better for his brand not to be invited. He can play the victim and show how anti-system he is.

Jagmeet Singh is the most popular leader at the federal level. Many polls have shown this. His party, the NDP, is also up a good 3-4 points compared to the 2019 results and seems to have some momentum. Quite a contrast with the beginning of the last federal election when the NDP was fighting the Green around 10-12%.

It is therefore far from impossible to observe a mini orange wave. I say mini because this post isn't going to look at scenarios where the New Democrats were to increase as much as they did in 2011. I just want to see what happens if the NDP reaches 22, 24 or 26%. Would it affect the Conservatives or the Liberals more? All the graphs should be interpret in context of the current projections of this site.

First thing first, let's see if the NDP is becoming better at turning votes into seats.

At 19%, the NDP is projected to win around 30 seats. As you can see with the vertical points at 19 however, the NDP could win over 40 seats if efficient (lucky?) or as few as 20. It obviously depends on many things including where the NDP support is. But over 5,000 simulations, things even out. remember however that reality only happens once.

We can see that the average is picking up at around 22-24%. So the NDP is currently just below what we call 'la zone payante' in French. I guess we could say the 'touchdown zone' in English? Our electoral system usually creates such a zone at around 25% where parties start actually winning seats. The relationship is clearly not linear.

Who loses when the NDP increases? Well both but the Liberals are more affected.

It might be super obvious but the line is definitely steeper for the Liberals than for the Tories. The latter are losing 0.9 seats for every 1 percentage point increase of the NDP while the former are losing 1.9 seats (if I approximate with a linear effect).

Maybe a better visualization is to see what happens to the CPC and LPC seats when their leads (in percentages) over the NDP shrink. It's a more direct approach (in the graph above, it was possible for the NDP to be up by 1% but for the Liberals to also be up compared to 2019 for instance, thus mitigating the impact of the NDP's rise).

The Liberals are currently polling at 35% while the NDP is at 19%, so a lead of 16 points. For every point decrease in this gap, the Liberals are losing 3.7 seats. The Tories are only losing 2.7 seats whenever the gap shrinks. Again, I understand some of you will argue using the national numbers is meaningless and I should focus on the leads in Ontario and BC. That is fair but the simulations have correlations across provinces. So if the Liberals are dropping and the NDP gaining nationwide, it'll be the case in Ontario and BC as well.

And now maybe the weirdest graph: the number of seats for the CPC and LPC as a function of the number of seats for the NDP:

It is weird because the Conservatives have a very slightly positive relationship! That's right, the more seats the NDP wins in my simulations, the more seats the CPC wins as well. It took me a while to look deeper and understand what was happening.

The first part of the answer is what we have seen so far: the NDP has more races with the Liberals than the Tories. The second part is there are some seats that the Conservatives start winning when the NDP rises and the LPC is down. But there aren't as many seats the Liberals win if the NDP is taking votes from the Conservatives. So yes, vote splitting.

It's also a question of looking at correlations between only two variables when it's really a multivariate relationship. If you instead do a regression of the number of seats for the Conservatives over the number of seats for the NDP and Liberals, you get a negative relationship with both of these variables. So don't start thinking that the Tories necessarily love it when the NDP is gaining seats and rising in the polls. They like it to some extent if it means the NDP winning seats in Toronto, but they are a lot less happy when the NDP starts taking seats from the Tories in the West, including in BC.

So, overall, what have we learned? First of all, the NDP needs to likely be around 22% nationally in order to win 40 seats, one of their objectives. They can do it with a lower percentage of votes but higher efficiency of course. Second of all, a mini orange wave would likely hurt the Liberals significantly more than the Tories. It obviously depends where the NDP is rising (is it DT Toronto or BC) but, in average, it is pretty clear the NDP can hurt the Liberals more.

The first few polls of this campaign have been less positive for the Liberals than the ones in recent weeks. Trudeau is still hugely favorite to win the most seats but a majority is now less certain. At 59%, the odds are still in Trudeau's favour but it's significantly less than the 75+ we had just before the start of this campaign.

The Conservatives are doing much better in BC but are still trailing in Ontario. They are back above 100 seats but their chances of even a simple plurality are low (although increasing).

The two good news for the Liberals are Quebec where most polls still give them a big lead over the Bloc (note: except Leger) and what appears to be substantial gains among the 55+, a demographic that votes more than others.

Anyway, here are the latest projections:

Below the seat projections are the 95% confidence intervals

Detailed projections

Proj Canada 17 August 2021 by bryanbreguet on Scribd

Map

Full page interactive version. Hold 'shift' to move the map around with your mouse cursor.

Alright, it's finally officially on. Early federal election starting today with e-day on September 20th.

Let's take a look at where we stand right now. If you want quicker updates, follow me on Twitter.

Projections

Voting intentions, seat projections with 95% confidence intervals, probabilities

Map

Full screen interactive version. Performance isn't the best (it depends), so give the map 1-2 seconds if you zoom in or out.

Short analysis

Trudeau has enjoyed a polling bump ever since Covid started, even though the gap has now been cut in half compared to what it was in April 2020. Still, given that the Liberals weren't that far from a majority in 2019 (with 33% of the vote, thanks FPTP!), it is natural that any bump would bring them closer or above the 169 seats threshold.

One caveat is that the Liberals won 157 seats in 2019 thanks, partially, to an incredible vote efficiency (i.e: winning close races; concentrating their support in the ridings they can win). That creates a somewhat weird situation where the Liberals could be up by a few percentage points and that wouldn't necessary translate into many additional seats. This is why the Liberals really need big leads over the Bloc (in Quebec) and the CPC (in Ontario) if they want a majority. As I showed a few days ago, the Liberals are comfortably in majority territory as long as they keep both leads to 10 points. But should one lead shrinks to 'only' 5 points and the odds of a majority are at 50-50 at best.

This is why I currently have the chances of a majority at 70%. It is there, no doubt about it, but the gamble that Trudeau took this morning might be more risky than it appears. The lead over the Bloc, in particular, doesn't feel as solid as the lead over the CPC in Ontario. I personally would expect Yves-François Blanchet to run another great campaign and bring his party to at least 30%. The most recent Leger poll showed that while people (including Québécois) are usually okay with Trudeau and his government, few would want a majority. This is something the Bloc can easily use to its advantage.

While Ontario is usually the center of attention for any federal election, it might be different this time around. This seems to be the most stable province compared to 2019 and Trudeau is polling actually lower than two years ago. It might be because the 'Ford effect' is now mostly gone, thus compensating the Covid bump? Don't get me wrong, the Tories are not doing better there, it's just that the gap with the Liberals is essentially the same.

Speaking of which, the Conservatives start this election with almost no chances of finishing first (not even a plurality). The party is bleeding on its right (many polls have shown that the CPC has lost more voters to the PPC/Maverick than to the Liberals) all while not making any gains in the center. O'Toole has clearly tried to re-center his party and enlarge his coalition, but it hasn't worked yet. So the CPC has this worst case scenario where the base is unhappy and the moderates aren't joining. The bleeding on the right might end up not costing too many seats (the CPC will just win the rural ridings in Saskatchewan and Alberta with smaller majorities) but it creates a situation where O'Toole has to fight on both fronts. Given the polling numbers, I wouldn't be surprised if the Tories decide to run a boring, catered-for-the-base campaign with the aim of saving the furniture (i.e: winning 100 seats) all while hoping the NDP and Bloc are preventing a majority. This plan might work as long as they don't drop too much in the West, in particular in BC where some polls are straight up disastrous for the CPC.

The NDP is the other party starting this election in good shape. Very different compared to the  beginning of the 2019 campaign where the NDP was polling at the same level as the Green party, around 10-12%. Singh ultimately won that battle (partially by winning the English debate) and enjoyed good momentum at the end of the campaign. It didn't translate into as many seats as they had hope but it created a strong foundation to built onto. Singh also benefits from being, by far, the most liked leader federally. If he plays his cards well, we could see a mini orange wave with the NDP gaining many seats in Ontario and BC. This could be helped by the fact the CPC isn't a 'threat' as far as forming the government is concerned, thus negating any call for 'strategic voting' the Liberals will throw around.

The Green party is collapsing in front of our eyes with a party actively sabotaging the efforts of its new elected leader. This is spectacular to witness but sad democratically. They already lost one MP (Fredericton) who decided she wanted to keep her seat and run for the Liberals (shameful but whatever) and they could well end up with only one elected MP, Elizabeth May, the former leader. I can't see this party getting its shit together during the next 4 weeks and I think the only real question is who benefits from the collapse, the Liberals or the NDP? Polls don't necessarily agree on that front.

Finally, the Bloc is polling significantly lower than in 2019 but things could change quickly. Blanchet is still liked and the Bloc as well. If the election becomes a question about giving Trudeau a majority, the Bloc would be ideally positioned. Blanchet also has the benefit of being able to campaign in Quebec every day while Trudeau will need to spend a lot of time elsewhere. My guess is the Bloc has the objective to regain some or the urban ridings it should have won in 2019 (Hochelaga, ridings on the south shore of Montreal). The Bloc could be under threat of the Conservatives if they were to increase in Quebec (both parties do share part of the electorate) but O'Toole's French is too weak in my opinion for him to make major inroads in La Belle Province. Also, as mentioned previously, the CPC will likely play defense for most of the election. That means spending more time in Ontario and BC than in Quebec.

Alright, that's all for now. Happy election!

Detailed projections

Proj Canada 15 August 2021 by bryanbreguet on Scribd

I have included the polls from Angus-Reid, Abacus and Innovative.

The Liberals are clearly in majority territory and there is almost zero uncertainty regarding who would finish 1st and second. Find the detailed projections below the map.

Map:

If you want an interactive version.

Projections:

Proj Canada 13 August 2021 by bryanbreguet on Scribd

Possible outcomes: