How accurate are polls in Canada? Less than you think!

If you want more updates, follow me on Twitter. I'll post new projections later today, I want to wait for the new Nanos numbers.

Every poll published in Canada is supposed to come with margins of error. You know, the plus or minus 3%, 19 times out of 20 thingy. Over the last decade, the rise of online polls has created a new debate among many since those polls don't use a random sample - they instead have panels of hundred of thousands of people who accepted to be on this list - and therefore don't have the classic margins of error.

That's a debate for another day however. What I want to do in this post is take a look at the actual, empirical accuracy of polls. Because, you see, the 3% 19 times of 20 is purely theoretical (and as mentioned, doesn't even apply to online polls). It only represents the random variation occurring because of the random sampling. It doesn't account for other factors such as turnout, biased sampling, people changing their mind, selection bias, etc.

Think of it this way: if really the only uncertainty was due to the random sampling, then polling aggregators like me would have almost perfect accuracy. Indeed, while one individual poll with 1000 observation has margins of error of roughly 3%, a polling average, composed of 5-7 polls, has a theoretical margin of error much, much smaller than that (less than 1% for sure). Yet, empirically speaking, I (or any other aggregators really) have not been that close.

Here are the metrics we'll use. The first one is the average absolute error. This one is the simplest. If a party was polled, in average, at 30% but ultimately got 32% of the vote, the absolute error is 2%. Absolute means you take the absolute value, so it doesn't matter if the polls were under- or overestimating the party. You do this for every party during an election, take the average and voilà.

The second metric is the MSE, or Mean Square Error. This is the average of the deviations squared. This is a very commonly used measure of the precision of an estimator in statistics.

Finally, using the MSE, we can estimate actual, empirical margins of error by simply multiplying the MSE by 1.96. This measure has a nice interpretation and can be directly compared to the theoretical one provided by the polls.

I have looked at the following elections over the last 12 years: the 2008, 2011 and 2015 federal elections, the 2008, 2012, 2014 and 2018 Quebec elections, the 2012, 2015 and 2019 Alberta elections, BC in 2013 and 2017 as well as Ontario in 2014 and 2018. In each case, I took an average of the polls published during the last week, limiting each polling firm to one poll only. No weight or nothing, just a simple average. Some might argue that I should give a bigger weight to polls closer to election day. Fair enough but empirically it really doesn't make a big difference, as we showed in a research paper with David Coletto. I also only used the numbers for the major parties included in the polls. So the number varies depending on the election (5 at the federal level for instance).

So here are the results:

In average polls have been roughly 2 points off. That's not bad you'll say but remember that this is the polling AVERAGE. As mentioned above, this average should be much closer to the actual result if the only uncertainty was really due to random sampling. Being 2 points off for an individual poll? That's great. For the average composed of usually 5-10 polls? Not impressive. And this alone shows that theoretical margins of error published in the media are pretty useless. This is why, personally, I don't care about the phone vs online debate - at the end of the day, what matters is the empirical accuracy. But feel free to have a strong opinion on this and go on Reddit or Twitter to express your deep knowledge of stats.

Maybe the most shocking stat is the corresponding, effective, empirical margin of error: 5.8%! And if you think I made a mistake or that Canadian polls are uniquely bad, you'd be wrong. This same margin of error in the US is close to 7%! In other words, we pretty much never ever have an election where we can be absolutely certain as to who will win based on the polls. Take the current federal election, it means the range at 95% confidence level for the top two parties are 29% to 40% roughly. Yes that might seem absurd but this what the accuracy has been in the last 10 years.

Side note: France presidential elections seem more accurate than our elections, in average, with an effective margin of error below 4%. And that includes the 2002 election where pollsters have the wrong top 2.

Don't believe me? You'll likely mention a list of elections where the polls were super close. Fair enough but let's me retort with many cases of giant mistakes. Alberta 2012, polls had the Wildrose ahead by 7 points, they lost by 10 to the PC! In BC 2013, the BC Liberals got 44% of the vote while the polls predicted around 36%.

More recently the CAQ in Quebec won over the Liberals with a margin of almost 13 points. What were the polls saying? CAQ ahead by roughly 4!

Even elections where polls got the correct winner can have weak accuracy. The recent Alberta election had polls putting the UCP at 48%, 10 points ahead of the NDP. The actual results? UCP got close to 55% while the NDP was only at 33%. Even elections where people think the polls did well aren't that great. The 2018 Ontario election had the PC at 39%, a little bit less than 4 points ahead of the NDP. At the end, Doug Ford won with 40.5% versus 33.6%. Not as bad as the other examples but still far from perfect accuracy. and a good example of how having an average absolute error of 2 points can lead to outcomes very different from the polls and projections.

Here below you have those stats for every election I used:

We can clearly see the big misses of Alberta 2012 and BC 2013, as well as Quebec 2018 or Alberta 2019 (again, the last two are less remembered because the polls at least had the correct winner).

The good news for us? Federal elections have been more accurate than the average so far. The bad news? The polls over the last year or so have been very bad. Ontario 2018 is the best of the bunch and the polls still missed the margin of victory by 4 points!

So keep that in mind when you're looking at polls, even polling averages like on my site. Also, keep an open mind if you see a poll that 'clearly' looks like an outlier. You never know, this one poll might actually be right.

Oh, finally, when I looked at the performance of online versus phone polls, I found no significant difference between the two.

Why is Ontario not turning blue?

If you want me updates and numbers, follow me on Twitter here. If you want to make your own projections, use the simulator here.

We got a few new polls yesterday from Nanos and Mainstreet (expected since they have daily trackers), but Innovative and Angus-Reid too. I'll update the projections during the day.

We should mention how Nanos and Mainstreet see very different trends. While Nanos has seen the CPC increasing steadily over the last week, Mainstreet instead is showing the Liberals gaining ground. Nanos uses live callers while Mainstreet uses the now commonly used automatic phone calls technology, also known as IVR. Hard to say why one method would produce some opposite results though. So I guess we'll have to wait and see.

What is clearer and observed pretty much in every poll is the fact the Conservatives are up everywhere but in Ontario (and Alberta technically, but it doesn't matter and the sample sizes are very small for that province). See the table below showing the swing for the CPC compared to 2015:

So how can we explain this? The likely culprit is of course Doug Ford. His unpopularity in Ontario might have tarnished the 'Conservative' brand. The Liberals have been trying to associate Scheer to Ford for months now. The polling numbers for the Tories were better earlier this year (without being extraordinary), before Ford started being hated so much.

Ontario has always been a province showing resistance to the Tories. Harper spent years campaigning there before finally becoming the main party (in 2008) and getting enough seats for a majority (in 2011 thanks to finally winning big in the GTA). So we know Ontario is never the easiest province for the CPC. But to literally see this party down compared to 2015 is straight up weird. The Mainstreet riding polls in this province are placing the CPC higher than the national polls, but the same applies to the LPC, so it cancels out really.

Just for illustration, my current projections (non published) have the Liberals winning 164 seats Canada wide, with 70 coming from Ontario (versus 140 and 42 respectively for the Tories). If the CPC was 4 points above its 2015 results (leaving the LPC where it is, so we are assuming the CPC would take votes somewhere else), the projections would become 156 for the Conservatives and 58 seats in Ontario. It'd be majority territory if the votes were to come from the LPC.

I already wrote in the past week how Scheer's path to victory was narrowed. He needs some help from the Bloc in Quebec to prevent the Liberals from winning 50+ seats there. But he also absolutely needs to win Ontario. And right now it's not working. Of course, all the numbers used for this analysis were before the brown/blackface incident.

Projections update September 17th: Liberals possibly in majority territory

With new polls from Ipsos, Mainstreet and Nanos yesterday, the projections have moved a little bit. Thanks to strong numbers in Quebec and Ontario, the Liberals of Justin Trudeau are right at the door of another majority.

Before going further, here are the projections:

Untitled infographic
Infogram

The riding by riding projections:

   Proj Canada 18.09.2019 by bryanbreguet on Scribd

I understand that it might seem weird given some of the recent trends regarding the nationwide numbers. Nanos for instance has shown the Tories moving from almost a 3 points deficit to a 1 point lead in 3 days. Mainstreet has also seen the CPC rising.

But Ipsos, that had the CPC ahead, now has them tied with the Grits. More importantly however, the numbers in Quebec and Ontario are just too good for the Liberals. and they might actually be undesestimated there!

As I was discussing the other day (in French), the collapse of the NDP in Quebec is opening wide open a door to many gains for the Liberals. Trudeau and his candidates don't even need to gain votes or anything, they'll likely recover a majority of the 16 NDP seats from 2015. The Conservatives are doing fairly well in Quebec (slightly above 20%) and would likely keep their seats and make some gains. But their support is simply way too concentrated to really prevent the Liberals from winning 50+ seats. Could it change? Sure, I mean everything is possible. But Scheer's French is definitely inferior to Harper's. I don't really see any reason why the Conservatives would suddenly go above 25% in that province.

The Bloc could. Almost dead in 2011, this party is now on the rise. At 20-22% in the polls, it's still too low to really compete with the LPC in most of the province. However, we can easily imagine the Bloc gaining back some of the NDP votes. Should the nationalist party climb back to 25% and above, this could cost the Liberals multiple seats (in the east of Montreal island, the suburbs and elsewhere). The Bloc's message is very well aligned with the CAQ's one and this party remains extremely popular in Quebec.

I should mention that Mainstreet is providing riding polls and so far, in Quebec, they show the NDP even lower than the national polls. It's possible the party of Jagmeet Singh is below 5%! The Liberals, on the other hand, are higher in those polls by a few points. I haven't made any adjustment yet but it'll come soon. So if you're somehow hoping for the national polls to be wrong and overestimate the Liberals, well the riding polls are bad news for you.

Then there is Ontario. This is the province Andrew Scheer needs to win. Without major breakthrough in Quebec, he can't afford to be behind in Ontario. Sure he's in very good positions in the Prairies, Alberta and especially BC to make significant gains, but he'll fall short of the magic number of 169 without Ontario. As a matter of fact, he would fall of even creating any suspense and finish behind the Liberals.

Polls have been fairly clear over the last two weeks: Trudeau and the Liberals are ahead in Canada's most populous province. Yes they might be down compared to 2015 (and it's actually not sure), but they are still ahead. And by enough to maybe win a majority. The seats they'd lose would be compensated by the gains in Quebec.

There as well the riding polls are showing that the Liberals might actually be even higher than what the national polls are indicating. And the NDP is below any expectations. I'm talking of the NDP collapsing below 10% in Ontario. Yes that seems absurd but those are the current numbers.

The one thing surprising to me is that all those numbers don't match well with a very detailed Innovative survey. In this one, we can clearly see that the Liberals are actually down among some major groups of voters, including what they refer to as the "core left". So you wouldn't really imagine the Liberals being able to compensate some losses by having some NDP voters joining them. More importantly, you have the "Canadian dream strugglers" where the Tories are now ahead. Those are exactly the people targeted by the Conservative campaign. So it's a little bit surprising to see the Liberals doing so well in Ontario where non-horse race numbers would point to a less favourable scenario (that includes the numbers regarding a desire for change).

In any case, I think it's fairly obvious the campaign is just starting. The Liberals will need to play defense for one long month. In Quebec the Bloc will attack them non-stop on bill 21. A Leger poll published recently clearly indicated that a vast majority of Quebecois support this bill. Although, it should be said, there might be just enough of a sizable minority for the Liberals to win enough votes.

In Ontario, I'd expect the CPC's message to resonate well in the GTA. Google Trends is showing that people are starting to learn more about Andrew Scheer. And while I don't think his campaign was spectacular during the first week, he also seems to be delivering his message to the people he wants.

 

Is the People's Party hurting the Conservatives?

So it's official (and surprising), the leader of the People's Party of Canada, Maxime Bernier, has been invited to the official leaders' debate. A surprising decision given that the debate commission had initially told Bernier he didn't make the cut. I'm of the opinion that this was the right thing to do as the PPC is running candidates pretty much everywhere and, while it might not elect more than 1 MP (or zero actually), the polls put this party around 2-5%, a score that isn't negligible. It seems to me that should indeed be enough to be invited.

Andrew Scheer's reaction was, to say the least, not positive. See below.

Statement from Andrew Scheer’s Press Secretary on the Bernier debate decision: pic.twitter.com/YvPeFBT3DG

— Katie Simpson (@CBCKatie) September 16, 2019

Given that Bernier came super close to winning the leadership of the CPC (he would have if it weren't for the effort of the dairy farmers, let's be honest), and given the general policies of the PPC, the conventional wisdom is that the PPC is mostly hurting the Tories.

I'd tend to mostly agree with this assumption. A poll from Abacus supported this thesis, although it was done right after the creation of the PPC. My own analysis of the data (for instance I looked at the correlation between polls as well as looking at the before/after of polls from the same firms when they decided to include the PPC), confirmed this. Other polls (for instance polls showing how the 2015 votes is distributed currently) also pointed towards the PPC mostly taking votes from the CPC, but not only. I think it's really hard to give an exact estimate. I'd say anything between 40 to 60%. I'd say we can confidently say a majority (or plurality) of the PPC votes are coming from the CPC, but not all.

In my model, I assumed that the PPC was taking 40% of its votes from the CPC, 15% from the Liberals, 15% from the NDP and 30% from new voters. Some might be surprised that I'd even assumed some NDP voters defecting to the PPC but all the numbers analysis I ran showed that there was indeed a link. Remember, chances are that if you are reading this blog, you are a political nerd who knows the full left-right spectrum. For you, it's not logical to jump from NDP to the PPC. But many voters aren't that logical.

Before looking at the current impact of the PPC on the CPC's chances of winning, I need to mention a somewhat counter-intuitive effect of my model. For the sake of illustration, let's round the Conservatives' results at 32% in 2015. If the PPC is polling at 4% and we assume 50% of those votes are coming from the Tories, then they should be at 30% (32% minus half of 4%) if nothing else had changed. So when the polls are currently placing the party of Andrew Scheer at 35%, it means they are up by 3%+2%, not 3%. In other words, they made up for the 2% lost to the PPC by gaining 2% elsewhere. Depending on where this 2% is coming from, it could actually make the Conservative's vote more efficient (imagine for instance the CPC losing the 2% to the PPC in rural ridings but gaining it back in the suburbs). Key word: could.

So what it means is that keeping the CPC constant at 35%, the higher you input the PPC in the model, the bigger the positive swing for the CPC becomes! Again I understand it might seem counter-intuitive but I actually think it's very logical if you think about. Of course the key element here is whether the CPC is indeed really at 35%, but that's another question.

So, is the CPC currently hurt by the PPC? My latest projections have the CPC at 35.2% nationwide and 140 seats, compared to 34.1% and 165 seats for the Liberals (the Liberal vote is just more efficient, thanks to Quebec. Read this piece in French from yesterday if you want to know why the Liberals could make big gains there). Specifically, here what it looks like:

The PPC is at 3.2% and zero seat (Bernier is in a close race in his riding of Beauce). So let's assume that 40% of the PPC votes would vote for Scheer if the PPC didn't exist. 15% would go to the LPC, 15% to the NDP and 30% would simply not vote. Let's assume it's the same in every province (well except in Quebec where we will assume 40%, 10%, 10% and 20% to the Bloc).

Doing so would give the following projections:

So yes the Tories would indeed be higher (they gained 40% of 3.2% but also a little bit more since 30% of the PPC wouldn't vote, thus making the other votes be worth slightly more). But realistically, it's only changing a few seats.

With that said, a few seats could well be the difference between a majority or minority, or a Conservative or Liberal government. So it's too early to categorically state that the PPC isn't affecting this election. But it does seem to be fairly minor right now. Of course, this exercise is full of assumptions. Moreover, with Bernier at the debates, it's possible he'll actually take more votes away from the Tories.

If we instead assume that 60% of the PPC votes is from the CPC and 40% wouldn't vote (so no exchange between PPC, LPC and NDP or Bloc), we get the following:

So depending on the assumptions, it seems the existence of the PPC is costing the CPC between 1 and 4 seats. Alternatively, it means the gap with the Liberals is decreased from 25 seats to 17 in the most aggressive scenario here. So it's definitely not negligible in this case.

In conclusion, I think it's fair to say the PPC could indeed hurt the Conservatives the most. But it's also not currently making the difference between Scheer becoming Prime Minister or not.

Projections update, Monday September 16th

Quick update to the projections. I added the new daily trackers from Mainstreet and Nanos. Plus a few tweaks here and there (like the riding of Victoria). Nothing major. They are behind paywalls but I can add them to my averages. The Tories are now at 35% but... they still trail in Ontario and actually lose seats in Ontario and actually lose seats in this update.

Remember, this update is done early morning (like 1am Vancouver time). So no, it doesn't include any poll published in the AM on the East coast. Mostly because I'm still sleeping at that point!

I don't include the riding polls yet simply because we don't have enough. However, if I did, the Liberals would be even higher in Quebec while the NDP would be lower. But again, that's based on like 4-5 polls.

Have a nice day!

Remember that you can use the simulator here if you don't trust the polls (but trust me, at least when it's free, to convert percentages to seats).

And the riding by riding projections:

   Proj Canada 16.09.2019 by bryanbreguet on Scribd

élection fédérale 2019 Qui profitera de la baisse du NPD au Québec?

Premier billet en fran?ais de cette campagne. Comme à mon habitude, si le sujet concerne le Québec, j'écris dans la langue de Molière.

Le Québec joue toujours un r?le important lors d’une élection fédérale, ne serait-ce qu’en raison du nombre important de députés dans la province (78 sur 338). Mais cela est encore plus vrai cette année. Le Québec pourrait en effet être la clé d’une réélection pour Justin Trudeau.

Il y a une certaine ironie à voir une province qui a voté majoritairement pour la CAQ et la la?cité il y a 1 an de cela -et qui soutient majoritairement la loi 21- qui pourrait fort bien être la raison principale pour laquelle Trudeau conserva son poste. Mais la politique est parfois bizarre. Aussi, le mode de scrutin étant ce qu’il est, un parti à 35% peut gagner très gros, surtout si l’opposition est divisée.

Il est encore t?t dans cette campagne pour vraiment aller dans les détails. Les projections actuelles sont disponibles ici. Mais une chose semble acquise : le NPD est en forte chute par rapport à 2015. Tellement en fait que le parti de Jagmeet Singh se retrouve parfois derrière les Verts d’Elizabeth May et à des niveaux d’appuis d’avant 2008. En gros le NPD au Québec n’existe quasiment plus.

Les Néo-democrates avaient déjà subit de fortes pertes en 2015, 4 ans après une vague orange qui avait permis à cette formation de devenir l’opposition officielle. Ils avaient quand même remporté 16 sièges. La question devant nous est ainsi claire : à qui iront ces sièges? Regardons cela en détails.

Sur cette liste, Outremont est déjà retournée en mains Libérales après une partielle en raison de la démission de Thomas Mulcair.

Comme vous pouvez le voir, une bonne partie de ces victoires avaient été acquises par des marges très faibles. Avec le NPD sous les 10% dans les sondages au Québec -donc une baisse de plus de 15 points- il n’y a aucune chance que ces comtés serrées restent NDP.

L’autre fait marquant est que le PLC avait terminé 2^e dans 13 comtés sur 16. Pour le parti de Trudeau, l’effondrement du NPD au Québec fournit une banquet de possiblement 13 sièges supplémentaires (pour rappel le PLC avait remporté 40 sièges en 2015).

Les Conservateurs ne sont pas bien placés pour aucun de ces sièges. Cela fait en sorte que pour Andrew Scheer, s’il veut avoir une chance de devenir Premier Ministre, il doit compter sur le Bloc pour empêcher le PLC de récolter plus de 50 sièges au Québec. Car si un tel scénario devait arriver, cela rendrait la tache du chef Conservateur très, très difficile. Il lui faudrait gagner l’Ontario par une bonne marge. Or, les sondages actuels ne montrent de loin pas cela.

Est-ce le Bloc peut profiter de la baisse du NPD pour reprendre du poile de la bête au Québec? Après tout, avant 2011, le Bloc était quasiment assuré de remporter le plus de sièges dans la province. Pendant longtemps les électeurs Bloquistes partageaient beaucoup avec le NPD, sauf bien s?r la question de la souveraineté. Je me souviens que la vague orange de 2011 n’avait pas été une grosse surprise pour moi. Au contraire, je m’attendais à une percée du NPD à un moment ou un autre.

Tout ?a pour dire qu’il n’est pas illogique de penser que ces électeurs NPD retourneront au Bloc. Le Bloc avait en fait baissé en termes de pourcentages de votes en 2015, cela veut ainsi dire que les votes NPD de 2011 perdus étaient allés au PLC. Est-ce que les électeurs restants sont plut?t du type à avoir le Bloc comme 2e choix? Pas impossible. Un tel transfert d’électeurs permettrait au Bloc de non seulement gagner une bonne partie de ces 16 comtés, mais aussi d’aller chercher des sièges au PLC. Utilisons le simulateur pour illustrer cela.

Si le Bloc pouvait grimper vers les 24-25% tout en ayant le PLC sous les 35% (donc sous son résultat de 2015), cela ouvre la porte à de nombreuses victoires dans le 450 et en régions. Tout à coup le Bloc passe de ses 10 députés en 2015 à 20 et plus. Un tel scénario forcerait presque assurément une minorité à Ottawa. Je ne vois en effet pas comment le PLC pourrait retenir sa majorité étant donné les pertes probables en Ontario et ailleurs.

Le Bloc a aussi une carte majeure à jouer lors de cette élection : c’est le seul parti qui soutient le Québec dans son désir d’imposer davantage de la?cité. Si le chef Conservateur a fait savoir plut?t clairement qu’il ne voudrait pas aller en court pour défaire la loi 21 (bien qu'il ne soutienne pas la loi), le chef Libéral a été bien moins clair. à une époque où la souveraineté pure ne semble pas vraiment d’actualité pour le Bloc, se concentrer sur la défense de la loi 21 pourrait payer. Après tout, une forte majorité de Québécois soutiennent cette loi.

Nous verrons bien ce que nous réserve la campagne, mais je ne serais pas surpris de voir le Bloc gentiment remonter vers les 25% et devenir bien plus compétitif. Après les 4 sièges de 2011, ce serait toute une renaissance pour le Bloc!

Quant au NPD, il pourra peut-être sauver quelques sièges en concentrant ses efforts. Le chef Singh fait un bon début de campagne et son fran?ais est meilleur que prévu. Mais il resterait que ce parti ne conservera vraiseblablement que très peu de ses députés dans la province à moins d'un `norme retournement de situation.

First projections for the 2019 federal election: A close race between Liberals and Tories

Alright, time to start making projections for the now on-going 2019 federal election. I apologize for starting a little bit later than expected. But you can expect regular updates from now on.

Also, I'm sure most of my regular readers will be interested in the new simulator. I have added the PPC (more details below). So have fun.

Alright, let's cut to the chase, here are the current projections of this site:

Presentation polls and projections 2019
Infogram

I'll need a little bit more time to get the probabilities going.

You can find the riding by riding projections here below.

   Projections Canada 15 September 2015 - Too Close to Call by bryanbreguet on Scribd

1. What is the situation as the campaign is starting?

Believe it or not, we haven't had that many polls in the last two weeks, and especially not since the campaign officially started. So I don't believe we have a perfect idea of where things stand. Still, we can at least make an educated guess.

Basically, it's two-way race between the Liberals of Trudeau and the Conservatives of Scheer. This much is clear. There is absolutely no way any other party is in the running. This isn't your 2015 campaign with the NDP with a legitimate shot at power. If anything, the party of Jagmeet Singh might be lucky if it keeps its official status at the House of Commons. Yes, it's that bad for the NDP - although polls don't fully agree on the extend of the NDP's collapse, with IVR polls being much harsher than online ones.

The story of the last 12 months has been the rise of the Green party (they also won a second seat during a by-election on Vancouver island). The Green rising is a worldwide phenomenon and has been observed in BC in 2017 for instance. They aren't in a position -yet?- to win many seats but they could hope to finish of the NDP, at least in terms of votes. That is pretty nuts and we'll have to see if the Green vote is really solid as the campaign progresses. The current model doesn't have Green-specific adjustments. If this party stays as high as 10-12%, I'll look into that. I suspect the demographic of the ridings more likely to vote Green can be inferred.

Behind we have the Bloc. Under its new leader -Yves Fran?ois Blanchet- this party appears to have regained some strength. After the collapse of the PQ last fall, it wasn't guaranteed that there would be a competitive Bloc Québécois. Number wise, polls put them slightly above their 2015 results (which was a record low). It's not great but again, they are still there. More importantly, now that the NDP has fully collapsed back to its pre-2011 levels (for many reasons, don't want to debate them), that offers a good opportunity to the Bloc to come back to 25 or even 30%. I personally can't imagine the Bloc becoming the number one party in la Belle Province again, but this party could at least go and challenge the Liberals in the 450 or in the more rural Quebec.

Finally, the new People's Party of Maxime Bernier exists and is alive. Actually it's doing not too bad. Multiple polls even puts this party around 5% which is quite impressive for a new party. I personally find it ridiculous not to include him at the leaders' debate, but that's another discussion.

Let's go back to the big two. The Liberals are clearly down from their 2015 results. But a strong resilience in Ontario and the NDP collapse in Quebec make it such that the LPC would be favourite to win the most seats at this point. Yes they wouldn't wipe the Atlantic like 4 years ago but they could make some significant gains in Quebec. This is where not switching to proportional representation could really pay off for Trudeau. There is a desire for change but it's not as high as what we observed 4 years ago, or in recent elections in BC, Quebec and Ontario. It appears that many people aren't the biggest Trudeau fans anymore but they aren't yet sure if they want to change.

As for the Tories, they seem to have made some gains but mostly west of Ontario. In Quebec they are stable and they still trail the Liberals in Ontario. In this province, any hope of becoming the first party again would rely heavily on improving the vote efficiency. I'm talking for instance of reclaiming some of the suburbs of Toronto.

The main issue for Scheer is that his path to victory requires him to gain in Ontario and hope that another party will prevent the Liberals from winning 55 seats+ in Quebec. I'm saying another party because I possibly can't imagine Scheer, with his fairly weak French, doing better than what Harper ever did in this province. In an ideal scenario for Scheer, the Bloc Quebecois would go back to 25-30 and cost the LPC many seats. Then Scheer would "only" have to do seduce the GTA. This last part is clearly the objective of the Conservative campaigns. The public transit tax credit is proof of that. It's very early in this campaign but it seems fairly obvious the CPC has decided to campaign conservatively (pun intended). No giant promise or anything, just one clear message ("hey middle class in the suburbs, do you want more money?") and voilà. Will it work? Well maybe. If the election were tomorrow, the Tories could well win. But the "Quebec problem" could prove insurmountable. And the soft-change voters don't seem fully convinced by Andrew Scheer. it might help him if his party didn't have a controversy regarding one of its candidates every day...

Also, there is the whole issue of what winning really means. If Scheer wins a minority, does he actually become Prime Minister? I don't know, I guess it would depend. While a Liberal minority would likely require the support of the NDP and the Green (and maybe the Bloc), it also means a Tory minority would need 1 or more of these parties to defeat Trudeau. Because I can't imagine Justin Trudeau just giving up the power even if he wins fewer seats. So Scheer either needs a majority (he's far from it right now) or he needs to get close enough. My guess is it'd make a huge difference if Trudeau would only need the support of the NDP (and Green) or if he'd also need the Bloc.

Anyway, we'll need more polling and data before making a better call.

2. How does the model work?

All projections models -and there are a lot nowadays- use basically the same principle. You take the results in 2015, you add the provincial or regional swing and you get the numbers for 2019. You can tweak this swing, make some adjustments, be pretentious and write that you "account for socio-demographic characteristics" (without explaining how of course) but that's pretty much it.

My model isn't fundamentally different. I do have regional coefficients estimated using past elections. Do they help? I think they can. They usually don't hurt at least. I also have an incumbency effect but this is super weak -and it's always the case in this country. Don't believe anyone telling you about a strong incumbency effect in Canada. I also account for when a long term incumbent retires. Data shows that costs the party around 5 points (sometimes more but that's the average).

I believe I do two things differently however. The first one is how I aggregate the polls. I do NOT distribute the undecided proportionally. Instead I currently allocate 40% of them to the Tories, 40% to the Liberals and 10% each to Green and NDP. Other parties don't get anything (well the Bloc in Quebec does). Why do I do that? Because redistributing the undecided proportionally is a bad assumption. And yes this is an assumption, even if it looks like the natural thing to do. It tends to overestimate the smaller parties. It also assumes that decided and undecided voters will ultimately vote the same, which is dumb. So yes my assumptions of 40-40-10-10 are subjective but I'm clearly letting you know. I believe the undecided will ultimately go for one of the main two parties. The CPC vote is older -so higher turnout- while the Liberals benefit from the usual underestimation of the incumbent. Again, you are perfectly allowed to disagree but remember that using a proportional redistribution is also making a big assumption. My track record on polls aggregation is that it has always helped me compared to simply averaging the polls. Yes it hurt me partially in 2015 because I didn't see the late surge for the Liberals. My bad. But my overall polling average was still better than the CBC one for instance.

I also don't really waste my time trying to weigh each poll differently based on a slightly bigger sample size or whatever. My experience is that giving equal weight (except for obvious exceptions) works perfectly fine. So my rule during the campaign is every poll within two weeks is included as long as I don't have more than 2 polls from the same firms (and the second one is heavily discounted). Polls from the last week get a weight of 1, polls of the week before are at 0.5.

The second thing is that I'll use the riding polls a lot. Not only will I use them to adjust at the riding level -usually if my current projections are really off- but I'll also aggregate them and calculate the swings using these polls only. I'll then do an average of the swing calculated using the riding and provincial polls.

You might not trust the riding polls (especially the ones from Mainstreet which has a mixed reputation -wrongly if you ask me) but collectively, they have proven to be gold. They had the large victory of Ford over the NDP in Ontario. They also pretty much had the perfect results in Quebec, while provincial polls ended up being so wrong we can add Quebec 2018 to Alberta 2012 and BC 2013 in the list of giant polling failures.

We currently only have a few of these riding polls, so I don't have any adjustment yet. But it'll come.

2.5 How did I add the People's Party?

It wasn't simple. Adding a party to a projections model is always tricky. Ultimately, after carefully looking at polls and correlations, I decided to add the PPC the following way: I assume that 40% of the votes of the PPC will be taken from the Conservatives. Then 15% each from the Liberals and NDP (polling correlations did indicate a relationship with these parties, so did an analysis by Abacus). The remaining 30% is taken from new voters or uniformly. This creates a situation where the distribution of the PPC vote resembles the CPC's but it isn't a perfect copy.

Maxime Bernier is assume to keep half of the votes he got as a CPC candidate. This assumption makes it such that the projections match the two Mainstreet polls done in the riding. Personally, I'm fairly convinced he'll win his own riding. 

Alright, that's all for now. Expect daily updates from now on, in English and French. Have a nice campaign!