September 28th: Liberals still slightly ahead. Diverging polls in the GTA

Quick update to the projections. Almost nothing changes as the polling average hasn't really moved either. Mainstreet has had the Liberals dropping by almost 2 points in two days but other pollsters, such as Abacus or Research Co see the party of Trudeau actually ahead.

My feeling is that nothing will move until the debates at the earliest. And even there, I wouldn't be surprised if this election was almost entirely decided by the turnout.

Anyway, here are the usual things you want to see. In order: an infographic with the polling average and the projections, by province. Then the map (and heat maps for the main parties). Finally the riding by riding projections. Below I do a quick analysis of the GTA and how pollsters don't seem to agree.

The map. Ignore the first two numbers when you hover over a riding. The next numbers are the percentages for the CPC, LPC, NDP, Green and Bloc, in order.

Finally the riding by riding projections:

   Proj 28.09.2019 by bryanbreguet on Scribd

The GTA and polls that disagree

The GTA is by far the most important region of the country, as far as this election is concerned. While Quebec could decide whether Trudeau gets a majority or not, depending on the Bloc, the GTA is where Andrew Scheer needs to make gains if he wants to become Prime Minister.

Let's look at the current map:

As you can see, Toronto and the close suburbs are like the village in Asterix and Obelix resisting to the Romans. It's blue all around. But the red seats do add up to a very significant number.

Compared this to 2015 (map from Wikipedia):

We can clearly see that while the core of Toronto (what is usually referred to as the 416) hasn't changed much, the suburbs have. The problem for the CPC is that not enough red seats have switched to blue. Why? Well because the polls in Ontario show that this is pretty much the only province where the Conservatives are not up significantly compared to 4 years ago. As a matter of fact, the current polling average (34.3%) shows the Tories down! Fortunately for them the Liberals are down even more (from 45% to 38%). Still, this is nowhere close to being enough to win most of the GTA.

Unless, of course, the provincial polls are underestimating the level of support for Scheer in the the GTA specifically. For instance we could imagine the CPC down in Toronto proper but up in the suburbs. Or down a little bit in rural ridings (won't matter for them) and up in the GTA. Doug Ford won big in the GTA last year after all, so it's not like this region can't vote Conservative.

We don't get many sub-provincial polls, mostly because the sample sizes would be too small. There was one from Campaign Research showing the Tories ahead in regions of the GTA (except Toronto proper). But this pollster has had diverging numbers from the average this election. The same poll had the CPC ahead by 4 points province wide, something very different from most pollsters (many polls actually have the Liberals ahead by 7-10 points in Ontario).

Then we have Nanos, very reputable firm. They did a 905-specific poll (the 905 is the area code of the GTA... mostly... not everyone is defining those regions the same way unfortunately). The results? Conservatives and Liberals tied in the GTA at 40%! To put this in perspective, my projections currently have the Liberals at 41% and the Conservatives at 38%. So the Nanos poll is in line with what my model would predict based on provincial polls.

Finally, Mainstreet published a GTA poll yesterday. In it, the Liberals are at 45% versus 31% only for the Tories (the numbers are for the GTA minus Toronto itself). But Mainstreet has had the Liberals with a bigger lead in Ontario overall in pretty much all of their polls.

Which leads us to the riding polls from Mainstreet as well. We've got quite a lot in the GTA and with the exception of Whitby where my projections have the CPC slightly ahead while Mainstreet has the opposite (it's really close in both cases), every single one has had the same projected winner as my projections. But what about the actual numbers? In Oshawa, the poll had the NDP much, much lower than expected and the Liberals higher. But in Durham for instance, projections and polls were almost identical. Same story in Aurora-Oak Ridge. Waterloo's poll had the Liberals almost at 50%, so more than my projections and maybe a sign of the rise of the Liberals in really urban ridings. All in all however, none of the riding polls were super surprising and indicative of a very different situation.

If I only use those riding polls (in Ontario, not only in the GTA), I would get that the Liberals are higher than expected based on provincial polls and the NDP lower. However, Mainstreet's polls also have the LPC with a bigger lead and the NDP lower than other polls in average. So it appears to be a Mainstreet effect rather than a riding polls vs provincial ones.

To conclude this look at the GTA, let's see what a couple of percentage points could do. For instance, if I boost the Tories by 2 points and lower the LPC by 2, the seat count changes from 44-65-12 to 50-59-12. So a 4-points swing was enough to switch 6 ridings. In other words: a simple polling error in this region and that's the difference between the Conservatives winning the most seats or not.

For Andrew Scheer, he knows he needs those GTA seats. His entire campaign is pretty dedicated to them. But right now it doesn't seem to work. Maybe because of the unpopularity of Doug Ford in the province. In the GTA, he's doing better than Harper in 2015, but not remotely close to 2011 (see below). To be fair, he likely doesn't have to in order to finish first.