Paul Saint-Pierre Plamondon favori dans la course à la direction du PQ, mais rien n'est joué

 Quelques semaines après la course à la direction du Parti Conservateur, je porte un regard rapide sur la course à la direction du PQ. Et comme pour le PCC, j'utilise un mélange de sondages et de données du financement pour arriver à mes projections.

Voici un petit résumé rapide. Analyse plus détaillée après.

- PSPP est favori mais pas de beaucoup. Il est compétitif au 1er tour mais surtout, il semble pouvoir bénéficier d'être le 2e choix des partisans des autres électeurs.

- Étant donné le peu de données et la nature imprévisible des courses à la direction, on ne peut actuellement exclure une victoire d'aucun des 4 candidats!

- Le financement est plutôt défavorable à Gaudreault alors que le sondage le plaçait en tête. PSPP fait bien dans les deux.

- En utilisant le sondage et les données du financement, j'obtiens les projections (moyennes) suivante pour le 1er, 2e et 3e tour:


Note: oui cela ne somme pas à 100% car le financement avait une catégorie "autre" qui representait 1%. Et il y a aussi des arrondis. Au final ce n'est vraiment qu'un détail.

Certains aiment prétendre que l'on ne peut absolument pas prédire les courses à la direction. Ce n'est pas vrai. Un mélange sondages+données du financement avait bien fait pour le PCC en 2017 (bien que j'avais Bernier comme favori), pour le PQ en 2016 (j'avais prédit une victoire de Lisée mais moins facile que sa victoire réelle) ainsi qu'il y a deux semaines pour le PCC (j'avais au final prédit une victoire d'O'Toole). Oui ces courses ont davantage d'incertitude qu'une élection "normale" mais nous ne sommes pas non plus dans le noir total.

La première source est le sondage Segma paru dans le Journal de Montréal. Il a été fait auprès des membres du PQ. Une distinction importante par rapport aux deux précédents sondages faits par Léger qui avaient sondés les partisans du PQ ou le grand public. Cette dernière option favorisant largement Guy Nantel de part sa célébrité.

Segma n'est pas la firme avec forcément la meilleure réputation (ils ne sondent pas si fréquemment) mais c'est mieux que rien. Dans leur sondage, Sylvain Gaudreault était en tête avec 30% des voix avant répartition des 24% d'indécis. Suivaient PSPP avec 20%, Guy Nantel à 15% et finalement Frédéric Bastien à 10%. Le sondage montrait aussi que les jeunes électeurs préféraient PSPP alors que Gaudreault bénéficiait du soutien des électeurs plus agés.

Outre les marges d'erreurs (je parle ici des marges d'erreur effectives, pas théoriques) qui sont toujours assez larges pour une course à la direction, l'autre problème est le 24% d'indécis. C'est une chiffre assez élevé. Lors d'une élection traditionelle, les sondeurs répartissent en général ces indécis proportionnellement (ce qui revient à faire la mauvaise hypothèse que les indécis vont soit voter comme les électeurs décidés ou ne pas voter du tout). Personnellement, j'accorde en général davantage d'indécis aux grands partis et surtout au gouvernement sortant. Mais une telle hypothèse est difficile à justifier ici.

Au final, j'ai décidé d'accorder la moitié des indécis de manière proportionnelle et l'autre moitié de manière uniforme. Oui c'est une hypothèse et non je peux difficilement la défendre au-delà de vouloir introduire une bonne dose d'incertitude dans le modèle. Au final, il n'y a pas vraiment d'hypothèse claire à appliquer ici, c'est plutôt subjectif.

L'autre source de données est le financement. Lors de la dernière course au PQ, c'est ainsi que j'avais JF Lisée devant alors que les sondages avaient tous Cloutier comme favori. C'est aussi la raison pour laquelle j'avais finalement prédit une victoire d'Erin O'Toole. J'utilise les données mises à jour quotidiennement par le DGEQ. Je ne regarde pas les montants contribués mais le nombre de donateurs. Un vote est un vote, peu importe que vous donniez $5 ou $100. Je m'assure de ne compter que les donateurs uniques (certaines personnes donnent plusieurs fois au même candidat). Si une personne donne à deux candidats, je compte le vote deux fois pour l'instant. Une meilleure façon serait peut-être d'attribuer le soutien au dernier candidat soutenu financièrement mais c'est davantage de boulot et on ne parle pas d'un nombre énorme de donateurs dans cette situation.

Voici (dans un tableau pas beau du tout, désolé) ces données présentées. Comme vous pouvez le voir, le financement est bien moins favorable à Gaudreault que le sondage Segma. Personellement, je fais davantage confiance au financement mais je ne peux pas ignorer le sondage.

PSPP est dans le top 2 peu importe la méthode utilisée. Ainsi je crois que l'incertitude est davantage de savoir qui il affronterait au 3e tour. Mes projections actuelles ont Gaudreault mais Nantel a ses chances. Bastien fait aussi plutôt bien en se basant sur le financement.

J'utilise aussi les données pour estimer les 2e choix. Le sondage Segma ne contenait aucune information sur ce sujet malheureusement. Je regarde les donateurs qui ont contribué à plusieurs candidats et je peux ainsi créé une matrice des combinaisons. À ce petit jeu-là, le plus grand nombre de double dons est entre PSPP et Bastien. La 2e combinaison la plus répandue est entre PSPP et Nantel. La 3e est entre PSPP et Gaudreault. Vous voyez une tendance ici: PSPP est le 2e choix des partisans des 3 autres candidats. C'est probablement l'élément le plus important de ce billet aujourd'hui et pourquoi je suis relativement confiant en déclarant PSPP favori.

Voyez la matrice ci-dessous, en pourcentages.



Un problème ici est que cette matrice ne me donne pas le nombre d'électeurs qui n'ont pas de 2e choix. Je pourrais partir du principe que les donateurs qui n'ont contribué qu'à un seul candidat n'ont pas de 2e choix, mais ce serait largement surestimer ce nombre. Si j'avais un sondage avec les données sur cette question, cela me fournirait une base. Au final, ce n'est pas ultra important, je ne fais que surestimer le nombre de votes qui sont redistribués. En réalité, un candidat qui tire de l'arrière après le 1er tour aurait davantage de difficultés à remonter si beaucoup d'électeurs n'ont pas de 2e choix. Mais comme mes projections du premier tour ont PSPP et Gaudreault très proches l'un de l'autre, cela ne change rien à mes projections finales. Gardez aussi en tête que cette matrice est estimée avec un nombre plutôt faible de donateurs.

Quoi d'autre à ajouter? La tendance semble bonne pour PSPP, comme le témoigne l'évolution des donateurs telle que présentée dans ce tweet.

Cette personne semble mettre à jour les données assez régulièrement, je vous suggère de le suivre. Remarquez aussi que PSPP est en tête peu importe que l'on regarde le nombre ou les montants collectés.

Au final, la course au PQ est suffisamment serrée pour être considérée comme une course à 4. En effet, si l'on applique des marges d'erreurs de 5-10% (donc environ le double d'une élection générale où nous avons plein de données fiables), on voit que Bastien pourrait être même être en tête après le 1er tour. Est-ce probable? Non, mais ce n'est pas à exclure totalement non plus. Cela étant dit, il semble bien plus probable que PSPP gagnera à la fin. Il devrait terminer dans le top 2 le 1er tour et pourrait largement bénéficier du report des votes.

Je ferai une analyse plus appronfondie dans quelques semaines. J'espère que nous aurons un autre sondage d'ici là.

Erin O'Toole wins the CPC leadership. Peter MacKay misses an empty net.

 

After hours and hours (omg so many hours) of waiting for the Conservatives to find a way to open and count ballots, Erin O'Toole won the leadership race and is now the new leader of the Tories. This is a surprise given that polls had Peter MacKay as favourite.

My last blog post had MacKay slightly ahead but my final projections, posted on Twitter, had O'Toole winning. He ultimately won much more easily than anticipated in the last round. My method of mixing polls and fundraising data proved to be effective once again.



Why did my projections moved from slightly for MacKay to slightly for O'Toole? Because I incorporated the fundraising data from the third quarter (data recently released) and those were terrible for MacKay (contrary to what some CBC articles claimed). He was a distant third in number of unique donors. If anything, the fundraising was showing Lewis on the rise and MacKay collapsing.

The Mainstreet poll had the following numbers (for the share of points): 40.5% for MacKay, 32.1% for O'Toole, 16.2% for Lewis and 11.2% for Sloan. The actual results were, respectively, 33.5%, 31.6%, 20.5% and 14.5%. So mixing fundraising and polls did a much better job. It put MacKay lower and Lewis significantly higher. Sloan over performed both.

The polls were especially off in Quebec where Mainstreet had MacKay ahead, 56% to 30% for O'Toole. MacKay ultimately lost that province in the first round, 34% to 45%! Similarly to how polls were showing Bernier with a big lead in Quebec in 2017, the race proved to be very different.

MacKay had to run the scoreboard in Atlantic (he mostly did) and in Quebec to take a big lead on first votes. We knew he was weak on second votes, especially from Sloan and Lewis. His 34% in Quebec were definitely not enough and it was clear that O'Toole was going to win after the results of the first round. I always said MacKay needed to be at 40% overall after the first round in order to win. He only finished with 33%.

Leslyn Lewis almost created the surprise. She actually finished first in votes after the second round thanks to a very strong transfer of the Sloan votes to her. Her vote was quite inefficient, mostly because she was very low in Quebec (which makes sense, she doesn't speak French at all). But make no mistake, Lewis did phenomenally well and likely has a bright future in the Conservative party.

For MacKay, this is a really humiliating defeat. He was the big name, the supposed favourite. He was supposedly so ahead that many big names declined entering this race. He was meant to win. But he did a really crappy campaign (which did improve towards the end). He didn't offer much and he alienated a good share of this party. The fact he got so few of the Sloan and Lewis votes is indicative of that. Of course we wouldn't expect social conservatives to vote for MacKay, but they did vote for O'Toole who is, politically speaking, quite moderate and close to MacKay. O'Toole just ran a much better campaign.

Will O'Toole be able to defeat Trudeau? Well that's a big question. Personally, I feel that O'Toole is a giant upgrade over Scheer (MacKay would also have been). If Scheer managed to win the popular vote and come within a few seats in the GTA to win a plurality, I don't see how O'Toole doesn't at least achieved that. I also believe O'Toole to be particularly smart politically. He knows the CPC needs a better data game  for instance to compete with the Liberals. His leadership campaign showed he knows how to win. The road to a majority won't be easy however. Can he win in Quebec? He crushed it in the leadership but a general election versus Trudeau and the Bloc is a different story. So we'll see. I'll say this however, O'Toole has presented a more substantive plan than MacKay. So I expect the Tories to at least offer something bolder than what Scheer ran on in 2019.

Thanks for reading me and see you soon.

The Conservative Leadership 2020: MacKay and O'Toole favourite but Lewis has a chance

Three years after Andrew Scheer dramatically defeated Maxime Bernier on the 13th and last round (after being behind for the first 12th!), we are less than a week away from having a new CPC leader.

While there are four candidates, only Peter MacKay, Erin O'Toole and, to a lesser extent, Leslyn Lewis have a shot. Derek Sloan is pretty much guaranteed to finish fourth and be eliminated after the first round.

Projecting a leadership race is always tricky. Polls are rare and usually less accurate (hard to poll members only). The CPC also uses a weird electoral system where each riding is worth 100 points, no matter how many people vote in the riding. That means that votes in some Eastern riding, in particularly in Quebec, can be worth a lot more points than votes in Conservative members rich ridings of Alberta.

Before going into the numbers in-depth, here's an executive summary for the busy people that you are:

- MacKay is favored to finish first after the first ballot but should fall quite short of the 50% mark. The real issue for him is getting the second votes of Sloan and Lewis or O'Toole. Polls and fundraising data indicate that MacKay is by far not the main second choice of any other candidate. As a rule of thumb, MacKay must be around 40% after the first round to have a chance.

- O'Toole has a real shot at winning. As a matter of fact, it's not really possible to make a call between him and MacKay. The data we have doesn't paint a clear favourite and is quite noisy. I feel MacKay has a small edge, mostly because I expect his vote to be more efficient.

- Maybe the real surprise is that Leslyn Lewis has an actual chance. Don't get me wrong here, it would be a pretty big surprise if she won. But it wouldn't be a monumental one. As a matter of fact, fundraising data indicates a close 3-way race. She can also count on a massive transfer of votes from Sloan's 2nd choices.

Note: most of the tables in this post don't look nice, visually. Sorry but I didn't have time to make it look pretty.

The data

As mentioned previously, we don't have the luxury of dozen of polls like for a federal election. The calculations in this post are entirely based on two sources: 1) the most recent Mainstreet poll among CPC members and 2) The fundraising data from Elections Canada for the first and second quarter. Using fundraising data has proven to work in the past (CPC leadership in 2017 although I had Bernier as favourite because, well, I didn't factor how effective the dairy farmers would be) and especially the PQ race in 2016 where polls kept showing Cloutier ahead but the fundraising has Lisée leading.

The Mainstreet polls had been quite good in 2017 (although, as with my projections, they also had Bernier as winning at the end with a roughly 75-80% chances). They did less well for the NDP leadership the same year.

At the end of the day, we definitely do not have perfect data, far from it. But we are also not navigating in the complete dark or guessing. Moreover, the polling and fundraising data actually agree more often than not.


The (weird) electoral system

The same that our general elections are decided by a stupid and archaic electoral system, the Tories elect their leader with a weird one where each riding is worth 100 points, no matter how many members vote in it. Some ridings in Quebec in particular will only have a few dozen votes while some in Alberta will have a few thousands. This creates an extra layer of uncertainty as it's now possible to have more votes but fewer points.

In 2017, Andrew Scheer won 50.95% of the points but he actually got 53% of the votes. Bernier's vote was more efficient, mostly because he was winning the more urban ridings where there were fewer members. The difference was +2% for Bernier and -2% for Scheer. It's not huge but it was almsot enough for Bernier to win with fewer votes.

This year, the logical bet is that MacKay's vote will be more efficient. Mainstreet seems to agree as their detailed results (shared with me) show MacKay going from 37.5% of the votes to 40.5% of the points. So a boost slightly higher than what Bernier got. This seems reasonable to me but it's important to realize that this is mostly an assumption (although I guess Mainstreet has done some calculations based on who the respondents were from). Ultimately, I gave MacKay a 2% increase from votes to points and O'Toole and Lewis have a 1% drop. This is based on Mainstreet and the 2017 deviations.


The first round

Here below you have the predictions for the first round (% of points) from Mainstreet poll alone and from my projections that incorporate the fundraising data.


The fundraising data is much more favourable to Lewis. In pure raw number of unique donors, she's actually ahead in the second quarter (I do an average of the first and second quarter with a weight of 75% on the latter). But her donors are not evenly distributed across Canada, with below-average numbers in Atlantic or Quebec, thus making her votes to points conversion weaker. MacKay, on the other hand, could benefit greatly from crushing it in the Atlantic and in Quebec. He's getting a whopping 46% of his 1st-round points from those two regions while O'Toole only collects 29% of his points there and Lewis is at only 18%. This right here might well be the key to the victory of Peter MacKay on Sunday.

You have the percentages of votes for each province below, based on the poll and fundraising.


So no matter which method you use, you get the same qualitative answer: MacKay first, O'Toole second, Lewis third and Sloan fourth.

The fundraising numbers are exceptional for Lewis. She isn't an elected MP (never has actually) and I don't think anyone would have predicted her to be doing so well. She's a social conservative (well it's complicated, let's move on) and she might please some of the more militant base of the party. I therefore wouldn't be surprised if she ultimately performs worse than what the fundraising numbers indicate as the base might be more likely to donate and contribute.


The 2nd choices

Since it appears likely that nobody will be above 50% after the first round, the second votes will enter into play. Derek Sloan should be the first one to go. Where will his roughly 10% of votes go to? The answer is quite clear: a majority will go to Leslyn Lewis.

How do I know that? Well the Mainstreet polls indicates as much. Also, using the fundraising data, I identified people who gave to more than one candidates. Doing so showed me which pairs of candidates was more likely. By far the most common overlap is Sloan-Lewis. The table below show you how similar the results are whether we use the poll or fundraising data (note: for the fundraising data, I couldn't just assume that whoever only donated to one candidate had no second choice. I thus use the % from the polls and I scaled everything to match; Still, this alone isn't enough to explain why the two sources provide essentially the same results). For my projections, I simply did an average of the two.



What do we see? Well Sloan's votes will mostly go to Lewis. I don't think there is any doubt on that one. Lewis' votes are more evenly distributed but MacKay is clearly the 3rd option. Finally between O'Toole and MacKay, it appears that the MacKay to O'Toole is more common than the opposite. It will likely not matter as O'Toole and MacKay will face each other in the final round. Still, if Lewis were to sneak in and finish top 2 after the 2nd round, we need to realize that O'Toole might actually push her ahead of MacKay.

More generally, the various Mainstreet polls have shown that MacKay has decreased over time in the number of second votes he's getting. In the poll from May, MacKay was clearly the second choice of O'Toole supporters. That doesn't seem to be the case anymore.

All of this should be the number one concern of Peter MacKay: his capability of getting second votes from Sloan and Lewis is really limited. Like Bernier 3 years ago, MacKay might well finish 1st after the first round but lose at the end. If anything, MacKay is significantly weaker on 2nd choices from Sloan and Lewis than Bernier was overall.

We also need to deal with the 3rd choices of Sloan's supporters. We can assume that once they transferred (to Lewis mostly), that they act like supporters of that person. That's not perfect obviously. Mainstreet does show us the attrition from second to third with only about half of the supporters who have a second choice expressing a third. So that's a ratio we can use.

Anyway, given all this info and numbers, we get the following projections for round 2 and 3. It is assuming the same electoral advantage (the votes to points conversion) assumed earlier. If you hate this assumption, then take 2% away from MacKay and add 1% to Lewis and O'Toole (or 2% to O'Toole in the last round).


Without the electoral advantage, the last round ends as an almost perfect 50-50 where MacKay wins with... 50.01% of the points! And is naturally behind in votes! So it's really impossible to determine who of MacKay and O'Toole would win at the end.


Can Lewis win?

The answer is yes but it'd require the Mainstreet polls to be off. I haven't done formal simulations but essentially, she needs to beat her projections by about 5-7% in the first round. She does NOT need to be top 2 after the first round, she just needs to be close enough to O'Toole (or MacKay... but let's assume here that MacKay will be first originally) so that her votes+60% of Sloan votes are enough to finish top 2 after the second round. Since Sloan will be around 10-11%, Lewis can create the surprise as long as she's within 5-6 points of O'Toole.

Based on the polls, she is much further behind. Based on the fundraising, she is pretty much exactly where she needs to be. So, just for fun, I only used the provincial averages based on the fundraising and I got the following second round:

Sloan: 0%
Lewis: 31%
O'Toole: 32%
MacKay: 38%

So she's still falling short but it's super close!

Given how noisy and uncertain the data we have is, we really can't exclude a surprise Lewis win. Leadership races are full of uncertainties, much more than regular elections. So projections being off by 5-7 points is really not that far fetched. Especially with the weird electoral system the CPC use.


Other indicators

Using Google Trends over the last 90 days, it's surprisingly close. You'd imagine MacKay being much more ahead (note: Google only seems to know who MacKay is and we therefore can't do a comparison using a topic search, which is usually better. So I picked the search term option.



Purely anecdotally, but my Twitter feed has had a ton of people voting for MacKay. But Twitter in general, and my feed in particular, is not representative. If it were, Bernie Sanders would be cruising towards the presidency.

Finally, other polls done among CPC partisans (and not necessarily voting members) have had MacKay ahead, quite significantly. MacKay has the name recognition that the three others don't.


Conclusion

It's... too close to call. If you account for the uncertainty of any poll, add the uncertainty that it's a leadership race and they use a weird electoral system, it really isn't possible to make a definitive call. It's likely that it'll be between MacKay and O'Toole but Lewis can create a surprise.

MacKay was supposed to easily win this race, he was the big name that many others were afraid of. The fact that I can't even make the call that he's favourite is pretty crazy. It shows how lackluster his campaign has been (although it has improved from the catastrophic start). It also shows a pretty divided Conservative party.

Looking ahead, MacKay does have the profile to win where the CPC needs to win (Atlantic, Eastern suburbs), at least on paper. Mainstreet showed that among the 'pragmatic conservatives' (the ones who just want to defeat Trudeau), he's well ahead. But O'Toole could also do well and I believe both should be significant upgrade to Scheer. As for Lewis, well it'S hard to say. Her run is impressive and she doesn't fit the typical picture of a Tory (she's a woman, she's Black). I have no idea what would happen if she won.

Post-Mortem canada 2019: Trudeau pulls a Trump, NDP vote not coming out

After what will likely not be remembered as a great election, Justin Trudeau managed to win a surprisingly strong minority. This happened with a mix of polling failure - namely the NDP vote clearly not coming out - along with an incredible vote efficiency of the Liberals. The number of ridings won by a few hundred votes is astonishing. Multiple ones in the GTA and pretty much every single close ridings in Quebec against the Bloc ultimately went red.

So like Trump, Trudeau won while losing the popular vote. With only 33% of the vote, this is one of the lowest score for a governing party in a long time. Coupled with a weaker turnout - around 65% - this means the Liberals lost a little over 1 mio votes this time around. Trudeau can pretend he won a clear mandate but I think many would disagree. With that said, the NDP will clearly not want to go back in election any time soon - they don't have money - so I suspect they'll pass whatever the Liberals want to pass.

The Conservatives won the popular vote and actually got more votes than in 2015, but that wasn't enough to beat Trudeau. The GTA is one of the reason, although it only explains a few seats - usually close races. No the real difference between the projections and the outcome is the fact the NDP vote didn't get out and was incredibly inefficient. Maybe it's strategic voting, but if that happened, it happened at the very end really. Because even the polls done on Sunday were still good for the NDP.

The weaker NDP allowed the Liberals to win quite a lot more ridings, in Atlantic Canada, the 416, etc. Coupled with the close races in Quebec where the Liberals systematically beat the Bloc, that explains why the Liberals are over 150 seats.

It's pretty shocking how a somewhat small polling mistake, coupled with incredible vote efficiency,  can make a big difference. I'm not sure what you can do. Some races could have been called better by using more strongly the Mainstreet riding polls, but they were still severely overestimating the Liberals. And looking at them, many were wrong, especially in Quebec.

Anyway, here are some numbers. The model made the correct call in 293 ridings, that's an accuracy of 87%. That's quite okay for projections models. I mean you sometimes get lucky and go over 90% but anything above 85% is perfectly decent. Using the actual vote percentages, I would have predicted 128 CPC, 139 LPC, 28 NDP, 3 Green and 38 Bloc. With the correct vote percentages, 90% of the ridings are called correctly (so 35 mistakes). 10 mistakes in Quebec, same with the polling average or the actual percentages.

This is an accuracy rate similar (actually better) than during the last Quebec election. The difference? Mistakes cancelled out in Quebec (mostly). Not tonight.

So this isn't a systematic failure of the model like in 2015 where, even with the correct percentages, I still had 60 mistakes. No the model worked this year and I didn't miss a late surge in the polls or anything (see below).

So the polls caused a gap of about 10 seats for the Liberals (or NDP). The rest is mostly the vote efficiency of the Liberals in Quebec (and a few ridings in the GTA). The fact the Liberals are just ahead of the Bloc in votes (34.2% vs 32.5%) and yet won the most seat is surprising. You'd expect the Bloc to win more seats thanks to the francophone vote (or the concentration of the Liberal vote among the English community). So clearly the Liberals got their votes out where it mattered. They won Gaspésie, Sherbrooke -this mistake annoys me as it was super close and that would have been a nice correct prediction to get- Chateauguay or Quebec with, in each case, only a few hundreds votes lead. There as well it looks similar to Trump with his numerous wins by close margins.

I count at least 10 ridings currently going Liberals by a margin of fewer than a 1000 seats. Imagine if those ridings had flipped differently. The Liberals could easily be at 147 seats, which is quite in line with what the model would project based on the percentages.

Finally, let's talk polls. They did okay. They almost all underestimated the Conservative but somehow that didn't have a big impact - mostly because they underestimated them in Saskatchewan and Alberta. The overestimation of the NDP is systematic with only two polls having the NDP within the margins of error. And even there, one of them is only because the sample size was small, thus larger MoE. The NDP's actual result is below every single poll. So clearly something didn't work for them.



Leger did spectacularly well while Ekos was, to the surprise of no one, the least accurate. As mentioned above, my average was fine. Again I didn't miss anything or introduce any bias. It's just that all the mistakes pretty much all went in one direction. Another way to put it: the Liberals took advantage of the NDP under performing and also won pretty much every single close race they were involved in.

Alright, that's it for me. As I said before, I don't think I'll make projections for another election. It's getting boring and repetitive. It's also very derivative, no matter how some might pretend otherwise. I'll leave the blog open and might write once in a while. And who knows, I might change my mind. See you!


Projections finales: une minorité et le retour du Bloc!

Note: if you want the long analysis of the projections, this post in English is for you. Si vous voulez une analyse détaillées des projections, cliquez ici mais c'est en anglais.


Le nom de ce blogue, du moins en anglais, est enfin approprié pour mes projections finales: c'est vraiment trop serré pour déterminer qui des Libéraux et Conservateurs terminera devant. Une chose semble cependant acquise: il s'agira d'une minorité.

Et la raison principale est la renaissance du Bloc Québécois. Qui l'aurait crû en 2011 alors que ce parti ne faisait élire que 4 députés et voyait la vague orange rafler tout le Québec? En 2015 le parti avait réussit à augmenter son nombre de sièges, mais c'était surtout en raison de la chute du NPD. Après tout, le Bloc avait baissé son pourcentage de vote, la 4e baisse en 4 élections.

Mais voilà, 4 ans de gouvernement Trudeau, un NPD qui revient à son niveau "normal" dans la Belle Province et un chef conservateur qui ne peut pas vraiment parler français -et quand il le fait, c'est pour devoir s'expliquer sur sa position sur l'avortement- tous cela fait en sorte que le Bloc va redevenir le premier parti au Québec. Du moins en termes de sièges.

Comment en sommes-nous arrivés ici? Regardons le graphique des intentions de vote:



On voit la montée continuelle du Bloc depuis le début de la campagne, alors que ce parti était sous les 20%. Il y a ensuite eu le débat face à face sur TVA que le chef Bloquiste, Yves-François Blanchet, a remporté. La hausse s'est ainsi accentuée. Il a aussi bien fait au 2e débat en français, ce qui n'a pas arrêté la progression. On voit une petite hausse NPD en fin de campagne, aux dépends des Verts.

En fin de campagne, le Bloc a continué sa progression en particulier en raison du déclin des Conservateurs. Les sièges PCC étaient en majorité plutôt sûrs mais ce n'est plus le cas. Si le Bloc devait battre les sondages, il ferait mal tant au PLC (Gaspésie, Abitibi, etc) qu'aux Conservateurs.

La tendance de dernière minute est aussi excellente pour le Bloc avec plusieurs sondeurs voyant le Bloc à son plus haut vers la fin. Mainstreet a le Bloc devant en termes de votes, tout comme Nanos -ce sondeur avait le PLC largemement en avance il y a juste quelques jours de cela et dans les faits, le Bloc n'est passé devant qu'aujourd'hui! Ainsi il se peut que le Bloc fasse encore mieux que les sondages et on peut facilement imaginer ce parti au-dessus des 40%. tant que son vote sortira bien sûr. Forum a ce parti carrément à 37%! Par contre Ekos ne le voit qu'à 22% mais reconnaît qu'il s'agît probablement d'une erreur (au moins ils l'avouent).

Le chef Blanchet a réussit à positionner son parti pour attirer tant les électeurs du PQ mais aussi ceux de la CAQ. La loi 21 aide beaucoup. Pour mes projections ici, j'ai aussi regardé une transposition des résultats de l'élection provinciale de 2018 sur la carte fédérale. Cela m'a convaincu qu'une vague Bloc, assez similaire à la vague CAQ, va arriver. La seule différence est que les Libéraux fédéraux ne montrent pas signe d'écroulement comme le PLQ.

Regardons la carte pour voir les possibles gains.


Si le Bloc devait finalement obtenir davantage de votes que les Libéraux -ce que la moyenne des sondages ne montrent pas, mais c'est possible- le 2e comté en Abitibi tomberait rapidement dans le bleu clair.

Ensuite regardons l'est de la province.


La Gaspésie pourrait aussi facilement virer de bord. Dans les faits, il se peut que mes projections soient fausses ici. Montmagny—L'Islet—Kamouraska—Rivière-du-Loup passe des Conservateurs au Bloc. Juste au nord, Beauport-Côte de Beaupré pourrait en faire de même (les données pour le PCC dans la région de Québec sont très mauvais en fin de campagne d'ailleurs).

Dans la ville de Québec, le Bloc fait deux gains, y compris Louis-Hébert, un comté qui devait être sûr pour le PLC. Il faut dire que les données des sondages dans cette région montrent un Bloc plus élevé que prévu.

Au sud la Beauce devrait faire élire le premier et unique député du PPC, Maxime Bernier. La baisse des Conservateurs vers la fin devrait aider.

Au Saguenay et Lac-Saint-Jean, ce qui était un temps une source de sièges pour les Conservateurs est maintenant au Bloc, sauf Chicoutimi et encore là, il faut espérer pour les Conservateurs que l'effet candidat local qui avait permis une énorme victoire dans la partielle soit encore là.

Allons maintenant à Montréal et dans le centre du Québec.


Le 450 va voir une vague Bloc. La Rive-Nord avait commencé en 2015, la Rive-Sud suit cette année. Ce sont des comtés NPD et Libéraux qui changent de couleur.

Sur l'île, Rosemont devrait rester NPD, l'est va retourner au Bloc. Laurier-Sainte-Marie pourrait être l'une des surprises NPD mais les sondages donnent le Bloc gagnant.

Dans le centre du Québec, les Conservateurs ont une chance dans Trois-Rivières grâce à un bon candidat local. Un sondage Mainstreet avait le candidat PCC devant mais la vague Bloc en fin de campagne fait en sorte que je prédise une victoire Bloc. Juste à côté, dans Berthier-Markinongé, la surprenante Ruth Ellen Brosseau -celle qui était allée à Las Vegas en vacances durant la campagne en 2011 et avait été réélue à ls surprise générale en 2015- va probablement perdre. Le sondage local lui donnait des chances mais c'était avant que le Bloc ne passe les 30%.

Pas sur la carte mais j'ai le NPD gagnant dans Sherbrooke. Un comté qui était allé à QS l'année passée, un comté jeune. Serré mais je suis confiant dans ma prédiction.

Donc voilà, le Bloc est de retour et avec lui les gouvernments minoritaires. À moins bien sûr que les sondages aient tort.

Bonne soirée électorale! Suivez-moi sur Twitter @2closetocall pour avoir des analyses et commentaires durant la soirée.

Final projections for Canada 2019: a minority

Note: a French article about Quebec only is available here.

After almost 10 years of writing on this blog, I finally have final projections where, wait for it, things are too close to call! I'm dead serious, there is a limit to the accuracy of polls and especially projections models and I can't make a call either way.

It hasn't happened before. There were close-ish elections, there were elections where I ultimately was wrong (BC 2013), but I never have not made a call at the end. Well, this is it. The 2019 federal elections is literally too close to call. All I can say with high confidence is that it should be a minority (or plurality, whatever terms works for you, depending on how pedantic and annoying you want to be. Heck, go with "hung parliament" if you really want to show people you have a stick up your ass).

Okay so this post will go as follow. First I'll present all the numbers and the map, for the busy people who don't have no time to read. Then I'll provide an analysis of the polls and finally of the seats. In this section, I'll illustrate as well as I can why I can't really call which party will finish first. Notice that who finishes first might not matter ultimately as Trudeau, as incumbent PM, will still be PM and can try to form a government first. Politically however, finishing first might matter.


1. Projections and map








2. The polls

There were two parts to this campaign, both in Quebec and the ROC. And no, it isn't a pre- and post-blackface. In Canada, the situation was incredibly stable -one might say boring- until the the last two weeks when the NDP started an impressive rise. It coincided with the English leaders' debate that Jagmeet Singh won. See for yourself:




It might seem unimpressive to see the NDP around 18% but given where they started from, and the fact they are down a lot in Quebec, this is actually quite a feat. 4 weeks ago I wasn't even sure the NDP would get 12 MPs and be officially recognized as a party. Today? They will likely make gains in the rest of Canada -Quebec will be brutal for them, no matter what.

Let's put this way: the NDP went from fighting for its life and existence against the Green to being in a legitimate position to hurt the Liberals and Conservatives in Ontario and BC especially. Alternatively, the Green Party, who went into this campaign with big hope and momentum, will have to settle for 4th and 3 seats in the most likely scenario. A result that is, in itself, not bad, but disastrous compared to expectations. Plus the popularity of Elizabeth May has decreased during the campaign. My guess is that they'll want to find a new leader.

The NDP appears to be doing particularly well with the 18-34 voters. The ones who voted a lot more last time around and likely helped Trudeau win a large majority. That could create some surprises in some urban environments (I have the NDP winning Sherbrooke for this reason for instance).

The fact that both the Conservatives and Liberals have been so close all campaign long and they have been dropping together for the last two weeks or so is fairly unique. It really shows how Canadians aren't excited with the political offering. Trudeau has definitely lost a lot of his charm in 2 years -back then, nobody thought he wouldn't win a second majority- mostly because of some policies and the scandals (I'd personally add his attitude). Andrew Scheer has clearly failed to convince Canadians. Running a boring, non-ambitious campaign where most of the proposals are simply going back to what Harper was doing clearly didn't work. The fact that he is himself a social conservative and wasn't very convincing at answering questions about abortion for instance didn't help either. And let's not forget all the stupid "scandals" such as whether he had a license to sell insurance! Small scandals in themselves but it distracts from the message when you keep having to defend yourself. I'm of the opinion that if the CPC has had a better leader, they'd have a shot at a majority tomorrow. Also, Scheer's French is just not good enough and his performances during the French debates were disastrous. The Tories had a good thing to try in Quebec this election, they could have made gains. Instead, they'll be lucky/happy if they escape with 9 seats tomorrow.

The story of this election, beyond the rise of the NDP at the end, was the resurgence of the Bloc Quebecois. This party has a legitimate shot at winning the most seats in Quebec tomorrow. In any case, they'll win a ton more than last time -and even more compared to the 4 seats of 2011. Yves-François Blanchet won the first debate (TVA face à face) while also doing well in the second French debate. He managed to position his party to appeal to both the PQ voters but also to the CAQ ones. By defending bill 21 everywhere, this allowed him to attract many French Quebecers. It's almost like bill 21 replaced independence, which is even better for the Bloc as support for bill 21 is much wider than for another referendum.



The rise of the Bloc means the path to a majority for Trudeau is really limited -I'd argue almost impossible but that might be too strong. Not that the Liberals are collapsing in Quebec -the rise of the Bloc is mostly driven by NDP voters coming back- but they can't expect to make the gains necessary to compensate their guaranteed losses in other provinces.

Finally, let's look at Ontario and BC. The former has been very stable while the latter saw the NDP catching up fast.




Notice the Lib-Cons gap in Ontario, very stable in average. Some polling firms (Mainstreet, Nanos, etc) have a much bigger gap and that has been the case all election long. They might be right but a site like mine has to use the average.

Finally BC with the NDP surge and the Green collapse.


First past the post will allow the Tories to really benefit from vote splitting in BC.

2.1 Riding polls

Mainstreet has been publishing a ton of riding polls, behind a paywall however. Those have, generally, been very favourable to the Liberals and really bad for the NDP. I know the ridings polled aren't necessarily representative of the province but we have enough that averaging them should at least give us a rough estimate. Doing so -and no matter if I used all the polls or just the most recent ones- shows a situation where the Liberals are much higher in Quebec and Ontario. I'm talking of the LPC being above 35% in Quebec and easily over 42% in Ontario.

If those polls are right, then a Liberal majority is indeed possible. But those polls are definitely at odds with the polling average. In Ontario, it seems to be a Mainstreet house effect with this firm having a bigger Lib-Cons gap than the average, consistently. Mainstreet also has had the NDP lower in general -most IVR polls did actually- but the effect is amplified in the riding polls.

Look, those same ridings polls were pointing to a larger Doug Ford victory and it happened. They were showing a larger CAQ victory and it happened. To a lesser extent, it did the same in Alberta. So betting on those polls isn't a bad idea. I'm not however. Mostly because they really put the NDP too low. It just doesn't seem to make sense, especially not with the NDP surge at the end. I did use these polls to make adjustments but I have not used them to influence my overall average. I was planning on doing so but I think it's too risky to put all my eggs into Mainstreet's basket. Still, it's important to be aware that there is data supporting a Liberal majority or very strong plurality.

Side note: when using those polls to adjust my own projections, I first adjusted the Mainstreet numbers for the systematic bias they had. For instance in Quebec they had the NDP around 8 points lower than expected, so if one riding poll had the NDP at 12% and I had them at 20%, then no adjustment was necessary.


2.2 Late trends?

In 2015 polls done during the weekend clearly showed a late surge for the Liberal. The same surge that ultimately gave them a large majority. I missed it 4 years ago so I was careful to look for one. But there doesn't seem to be one this time around. Nanos and Mainstreet both saw the Conservatives rising a little bit during the weekend while Ekos and Forum saw the opposite.

Some polls show the Liberals increasing their lead in Ontario in the last few days while others haven't (namely Nanos). There does seem to be a positive trend for the Bloc in Quebec. So I wouldn't be too surprised if the Bloc won over 37 seats.

At the end of the day, polls might still be wrong but I truly believe I didn't miss a late surge or anything.


3. Seats and projections

Ultimately this is what matters. There are many close races, but there always are to be fair. I didn't program my simulations this time around, so I don't have well calculated probabilities. Still, I can give you an idea of the uncertainty that exists.

3.1 How do the projections work?

My model uses the results of the 2015 election and the swing observed in the polls. It's based on the uniform swing -if the Liberals are up 2 points, they are up 2 points everywhere- with many adjustments. I take into account of the region (some are more volatile than others), whether the long-term incumbent retired or there is a star candidate as well as some other modifications based on riding polls and others. Whenever possible, I also adjusted my projections based on the within-province numbers -Leger for instance gives us the estimates for Montreal, Quebec and the rest of the province.

Also of importance, I average the polls slightly differently. Pollsters usually allocate undecided proportionally. I don't. I allocate more to the CPC and LPC as they are the big parties, while Green would get less (and PPC and others get none). In Quebec, LPC is assumed to get a bigger share, then Bloc, CPC and NDP, in this order. Given that we don't have that many undecided in most polls, it doesn't make a big difference. But I thought I'd mention it. And if you think I'm just assuming stuff, you are right. But allocating them proportionally is also an assumption. And one that is less defendable and empirically worse.


3.2 Majority possible?

Talking of a Liberal majority here, a Conservative one is borderline impossible. First of all, let's do some simple math. In the Atlantic, the Liberals will surely lose at least 5 seats and possibly 15, so let's go with the projections and pick 11. That means the Liberals are now only 3 seats above a majority (184 seats in 2015 - 11=173; A true majority is 170).

Then there is Quebec. Best case scenario for them is to lose only 5 seats. The good news is that I don't think they can collapse like the Quebec Liberals last year during the provincial election. So total losses are at 16 and no more majority.

Let's skip Ontario if you don't mind, we'll get back to it.

In the Prairies, the Liberals are significantly down. They won a couple of surprising seats last time around and I can't see them holding onto those seats. They might save a few in Winnipeg and Regina but they'll otherwise have to fight the angry Conservatives or the rising NDP. So that's a guaranteed 3 losses. That puts out total to 18 at a minimum.

Alberta could be swept. If it isn't, it'll likely be because of the NDP, not the Liberals. So that's 4 losses there, almost for sure. Total is 22.

In BC, the Tories haven't increased much but could benefit from vote splitting and make major gains. The Liberals created all a surprise last time by winning 17 seats and they'll be lucky if they limit the losses to 4 or 5. So the total is now 26.

That means they are, at best, at 184-26=158 seats in the best case scenario.

Now back to Ontario. They won 80 out of 121 last time around. They are polling BELOW their level of 2015 even in the most favourable polls to them there. So the Lib-Cons gap will be, at best, the same -thus meaning NDP and Green and PPC are higher. Also, they already won pretty much everything they could in the GTA, so even if they do increase there, the number of gains are seriously limited. They would then need to go and win seats in the Southwest of Ontario. Some riding polls from Mainstreet have shown this to be possible, including in Windsor -where the NDP should be comfortably ahead. Still, the path to a majority is really, really narrow. It requires both a polling mistake and improved vote efficiency.

Some might say that I was saying the same 4 years ago and they'd be right. My model didn't perform particularly well that night. To be fair, most projections had a bad night but mine was especially affected. Part of the reason is because I didn't see the Liberal surge at the end and actually assumed they wouldn't get a higher share of undecided. That was my mistake and I'm not making it this year. As for vote efficiency, well it's already now built into the model through the 2015 results. So it'd require this incredible vote efficiency -the Liberals won way more seats than projected even with the correct vote percentages 4 years ago- to increase. I'm very skeptical. And the sub-Ontario numbers I've seen (like from the GTA) don't indicate this to be happening.

Look, at the end of the day the Liberals are down and significantly so. They are down in every single province, including Ontario. In some (like the Atlantic provinces) they are down by close to 20 points. I just fail to understand the projections of some people -some pollsters for instance- who think the Liberals are within reach of a majority. I know our electoral system is shit but you don't win a majority with 32% of the vote!

So I don't think a majority can happen, not without major polling failures. Speaking of which, now is a good time to remind people that Canadian polls are indeed not perfect. Even by averaging them, you can still be way off. As a matter of fact, the table below shows you the empirical margins of error of Canadian polls. Read this Post to get more details.



So while we haven't had a massive polling failure at the federal level, we did have many at the provincial ones. Also, the last 18 months have seen multiple elections where the polls underestimated the right-wing option (Alberta, Ontario, Quebec, PEI, BC referendum). Will we see this again tomorrow? Not impossible, especially since advance voting seem to favour the Conservatives.

So where are the seats in play? Well you can look at the list yourself but I want to mention some races. In Quebec, there is the uncertainty regarding the Bloc. This party is right at the door of possibly winning 40 seats and more. While most of the Montreal suburbs are already projected to go light blue, there are a many battlegrounds in rural Quebec. In the east, the Bloc could wipe everything from Gaspésie to Quebec city, taking many seats from the Liberals and Tories. The second Abitibi could also go Bloc as well as a few here and there. On the other hand, don't underestimate the resistance that some local candidates might put. In Berthier-Markinongé, the NDP incumbent was polled ahead -although it was before the full rise of the Bloc- while the Conservatives might have a few strong candidates who can win on their own such as in Trois-Rivières.

I'll say this, if there is a surprise tomorrow compared to my projections, I wouldn't be shocked if it were the Bloc winning way more seats than projected.

It doesn't mean I haven't done my job here, I even used the transposed results of the 2018 Quebec election onto the federal map to see which seats could flip. It gave me essentially the same: Bloc between 30 and 40 depending on turnout and what not. If Leger is right and the Bloc is so much ahead among francophones and rural Quebec, then 40 seats is possible. However, polls in average show the Bloc actually doing better than expected around Quebec city, not rural Quebec. This could hurt the Conservatives but would not bring as many seats.

The second big source of uncertainty is the GTA. The 905 specifically as Toronto proper is likely remaining red -although we could see a few surprises from the NDP and its young vote. This is where will be decided who will finish first. There are easily 15 seats in play here with a margin of 5-10%. Turnout will be crucial. Data from Campaign Research or Léger show a GTA that is as expected given the drop of the two main parties in this province.

Beyond the GTA, we obviously have races here and there, but nothing that should be correlated with each other -beyond a general polling failure of course but we can't predict those. The GTA is where it could make the difference from 131-131 to 140-120. Politically, those two outcomes are very different but based on the info available to us right now, they are quite close.

Here are a few races I want to mention:

- In Quebec, I think the NDP could keep win more seats than projected. I have them ahead in Sherbrooke, a riding that went QS last time around. A very young riding. In Laurier-Sainte-Marie, the Mainstreet polls were horrible for the NDP but I believe they could create a surprise

- The Green will likely win 3 and nothing more. If there's going to be some really surprising Green seats -i.e: outside of Vancouver island- keep an eye on Fredericton and Thunder-Bay Superior North. Still long shot, especially the latter

- The Maritimes could have many surprises. Polls have been all over the place and it's always tricky with 4 provinces. Also PEI has riding that are very small and changes can happen more easily. I'd expect some surprises there.

- In BC, Jody Wilson-Raybould is favorite to be reelected, which could be seen as a symbol at the end of the night of the rejection of Justin Trudeau after what could well be a tough night.


Alright, this is it. I hope you enjoyed the coverage during this federal election. I was less active than usual. That will very likely be my last election. I'm getting a little bit tired of doing projections. This is ultimately a very repetitive and derivative job. So thanks for reading me!

Next to last update

Alright, busy weekend for me. So let,s go straight to the numbers. The late trend is favourable to the Liberals who have increased their lead in Ontario. At this point, the Conservatives will need the polls to be wrong to win the most seats. In any case, a minority is ultra likely.

Final projections tomorrow evening.


Projections




Map (percentages are, in order, CPC, LPC, NDP, Green and Bloc)




Riding by riding