If the riding polls are right, the NDP is heading for disaster

Mainstreet has been releasing riding polls pretty much on a daily basis, only for subscribers. Looking at them, the NDP is even lower than what we'd predict based on national polls. I'm talking of an absolute disaster here. In this blog post, I won't publish all the numbers but I'll try to explain how bad those polls are for Jagmeet Singh and his party.

But first, quick update to the projections. Not a lot of changes nationwide but the Liberals took a bigger lead in Quebec and Ontario, resulting in them gaining some seats. The results are partially influenced by the latest poll from Ekos. A poll that has, to say it nicely, very weird results. The Bloc is at 14% in Quebec, tied with the Conservatives, 3 points behind the Green and 1 above the NDP. It's obviously ridiculous but I can't really exclude it from the average. Still, let be clear here, I don't believe this poll is correct. If those were my final projections, I'd have to think long and hard about including it.

Map. Ignore the first two numbers if you hover over a riding. The percentages are the numbers for the CPC, LPC, NDP, Green and Bloc.

Riding by riding:

   Proj 01.10.2019 by bryanbreguet on Scribd

The riding polls and the NDP

So far we got 16 polls in September in Quebec and 13 in Ontario, plus a bunch elsewhere. I'll focus mostly on the two big provinces since we have enough of those riding polls to average them.

If I use those polls to calculate and average the swings compared to 2015, I find the following.

In Quebec:

CPC: +5.5%
LPC: +4.4%
NDP: -21.1%!!!
Green: +3.8%
Bloc: +1.5%

Yes that's right, minus 21% for the NDP. That means this party would be at 4.25% province-wide!! This is insane. This is a party that got 25% in 2015 and over 40% in 2011. 4.25% is even lower than what Jack Layton was getting in 2008 for instance.

In Rimouski, one of their best ridings in 2015 and what we thought were one of the only hopes in 2019, the NDP incumbent is polling at 15%, down from 43%. On the island, in Hochelaga, the NDP is at 12%. That's barely ahead of the Conservatives. In La Pointes-de-l'île, the NDP was polled at 4.6%.

Provincial polls have the NDP, in average, down by 15%. So those riding polls are worse by a full 6 points. The projections still have this party winning Rosemont but we should get a poll there very soon and we'll see if the NDP has any hope of keeping even one seat in Quebec.

So, who is benefiting in those riding polls? The Liberals. Using the swing above (4.4%), that would put this party around 40%, about 4 points higher than the provincial polls. This could be a crucial difference. Using only the riding polls, we get the Liberals at 40% and the Bloc at 21%. That would likely be enough for Trudeau to win over 50 seats in Quebec alone.

In Ontario:

CPC: +0.3%
LPC: -2.4%
NDP: -8.4%
Green: +6.4%

That would put the CPC at 35% provincially and the Liberals at 42.4% while the NDP would be below 10%, at 8.2%. Compared that to the provincial polls that have, respectively, 33.4%, 39.5% and 13.6%. As in Quebec, we see the Liberals a few points higher and the NDP a few points lower.

In terms of seats, that would change the projections from 38-74-9 to 36-84-1! That's right, one seat in Ontario for the NDP (Windsor West, a riding won by the NDP in 2015 with over 51% of the votes). And actually Mainstreet had a poll in this riding on august 21st that had the Liberals at 38% and the NDP at 36%. So the NDP could well win zero seat in Ontario as well.

Just think about it, if the riding polls are right, the NDP will likely win zero seat in Quebec and in Ontario. How freaking crazy is that? They would be at 9 seats Canada-wide and that'a assuming they keep their ridings in BC. The only three polls there (Victoria, Vancouver Granville and Kamloops) have shown the NDP down by over 20 points in Victoria and Kamloops (Vancouver Granville is harder to use since there is the effect of the independent running for reelection, Jody Wilson-Raybould). So it's well possible that the NDP would also lose way more than expected in BC.

Elsewhere, the news aren't better. In the riding of Winipeg Elmwood-Transcona, won by the NDP by 0.14% in 2015, the riding poll put this party third at only 20%. In Edmonton Strathcona, won by NDP in 2015 as well, the poll had the NDP down by over 20 points.

We are literally talking of the possibility that the NDP not win enough seats to be recognized as an official party! They need 12 seats for that.

So now the big question obviously is: are those riding polls right? Because they are definitely painting a very different outcome for the NDP than the provincial polls.

My answer is: I don't know! I wish I did. Here's what I have to say. In Ontario and Quebec last year as well as in Alberta earlier this year, the riding polls proved to be much more accurate. They had the CAQ higher and the PLQ really low (which happened). In Ontario, they predicted a much easier victory of Ford than other polls. In Alberta they had the UCP over 60%, which they got.

But there is a major difference this time around: in all those elections, the traditional Mainstreet polls (not the riding ones) were mostly in line with other polls. In Quebec they didn't have the Liberals as low as their riding polls for instance. In Ontario Mainstreet was very close to the results, but so did most of the IVR polls (IVR polls did have the PC-NDP gap much larger and they ended up being right). In Alberta, Mainstreet polls actually had one of the highest score for the NDP and a clear overestimation from the actual results.

What I'm saying here is that in previous elections where the riding polls from Mainstreet were showing something different, those polls were also different from the other polls of the firm.

This time however, this is different. If we use the Mainstreet polls (last 6 days) in Quebec and Ontario and compare them to the polling average, we get the following:

In Quebec, the LPC is higher by 1 point and the NDP lower by 1. Not a big difference but it means the 4 points difference from the riding polls can be partially explained by a Mainstreet effect in Quebec.

This is even clearer in Ontario. Mainstreet polls have the Liberals way ahead of the Tories, much more than other polls in average. Specifically, they have the Liberals higher by almost 4 points and the NDP lower by almost 3. They have the LPC ahead of the CPC by about 12 points, much more than the 4-5 points of the polling average.

In this case, if I take the average swing from the riding polls and correct it for the "Mainstreet effect", I actually get that the Conservatives are at 37% and the Liberals at 38.5% only. In other words: the differences in the riding poll can be completely explained by a Mainstreet effect.

So we have a big decision here. We can trust those riding polls. But that means trusting Mainstreet more than other pollsters. This is a big gamble. Mainstreet is a very good firm with an excellent track record (outside of course of Calgary). But am I ready to bet the entire projections on this firm? Not really.

So I'm currently using the riding polls to adjust some ridings when there are big differences. I am not adjusting my averages with them. That was my intention at the beginning, but I'm now reluctant to do so because it really seems Mainstreet polls are different from others, both the traditional and riding polls. But for the NDP' they better hope Mainstreet is dead wrong, that's for sure. The Conservatives too since those riding polls would almost surely mean a Liberals majority.