Leger GTA And The Zombie Poll-ification Of This Election

On Bad Data Takes

There’s a big-sample Leger out today, which in addition to its usual fare, broke out regional information for Ontario, BC, and Quebec. The Ontario numbers have people getting riled up, as it’s merely a 5% Liberal lead in the GTA, as opposed to a 10% lead province wide. So, is it right, and does it matter?

Whether it’s right is hard to say, but it’s probably directionally correct. It’s nearly impossible to check the swing on the data because I have no idea what Leger is counting as “GTA”, because every pollster has a different definition. This has been a recurring battle in both 2019 and 2021 as well, because how far outside of Toronto counts is unclear. Whether Michael Chong’s Wellington seat counts matters, as does how far east does it go. To use a fun example, does Jamil Jivani’s Bowmanville-Oshawa South count as GTA? Does Peterborough count?Cause if it goes out that far a 5% lead isn’t as big of a swing as just thinking about the “Core GTA” (for lack of a better way of putting it) would suggest.

The reason not to be that terrified is if it’s right, the Liberals are taking some losses but also sweeping Windsor, London, Niagara, Hamilton, Northern Ontario, beating Skippy in Carleton, and almost assuredly still winning government. Remember, with Quebec and BC where they are the Liberals start this race with an extra dozen MP at least, meaning the Tories would need to net out 30+ gains in Atlantic Canada, the Prairies, and Ontario just to get back to parity. So yes, sweeping Halton and splitting Brampton and Mississauga and winning 7 seats in the 416 proper would be a great result, but they’d lose Kenora, Niagara Falls, Carleton, Parry Sound, Ferreri and Bay Of Quinte if they don’t count as GTA, and that in totality isn’t enough.

What's likelier is that the swing in the GTA - both suburbs and 416 proper - is heavily uneven. What's likely is Poilievre does a lot better with ethnic minorities in Brampton, parts of Mississauga, Scarborough, and other assorted parts of the GTA, which means that Bill Blair gets run much closer than usual in his Scarborough seat but Karina Gould skates back in Burlington. That would match with what Mainstreet’s sub-regional polling suggests, which is much higher ethnic realignment. It’s also what we saw in BC last year, as Surrey broke for Rustad while whiter parts of Vancouver proper didn’t to the same degree.

If that were to happen, you could have something resembling a total annihilation of the Conservatives on the table - their vote gets so much less efficient as every Brampton seat goes from Liberal+20 to Liberal +8 but they don’t flip, while the votes they lose are in electorally useful places like Niagara Falls, Peterborough, Bay of Quinte, Jivani’s seat, and even potentially Carleton. The thing about realignments is that they’re messy, and sometimes you can thread the needle perfectly and sometimes you get screwed.

When Hillary Clinton lost, she did best in generations levels for Democrats in a lot of affluent white suburban areas, especially in the South. Those gains were enough to help Biden win Georgia in 2020 when Trump had less working class votes to gain while Biden had more suburbanites to get, but in 2016 Trump’s coalition was hugely efficient. He bled a lot of votes in states he didn’t need (California) or where he had room to bleed (Texas, Georgia, Arizona), while maximizing the gains in the ones he did (Ohio, Florida, Wisconsin, Michigan, Pennsylvania). In a similar way, Liberals might strike that perfect balance here.

They obviously might not, too - if Leger is literally right the Tories will flip Brampton seats and win (relatively whiter) Halton seats too. But we’ve done this before with regional crosstabs, and I don’t think this makes much sense. The idea the Liberals are doing better in non-Eastern Ontario, non-GTA Ontario than they are in the GTA - which is this poll’s inherent claim, given the Liberals win South, North, and Hamilton-Niagara by a combined ~9% - doesn’t pass a basic smell test. I mean, the Liberal 10% lead province wide is definitely being powered by *checks notes* the places that think the Liberals care too much about wokeness and that hate how anti-gun Liberals are, for sure. Yup. Definitely.

Whether it even matters is, again, unclear - if the Liberals are gaining 8-12 seats west of Thunder Bay and another dozen-ish in Quebec, then the Tories can gain 5 seats north of Toronto, 3 Scarboroughs, York Centre and Eglinton-Lawrence, 3 Bramptons, 3 Mississaugas, all 5 Haltons, and hold everything else in Ontario and they’re still not winning a plurality of seats. The basic realities are really, really, truly horrendous for the Conservatives.

This is also only one pollster. Pallas yesterday had a 19% lead in Ontario, Angus Reid has a 14% lead, Pollara has a 12% lead. Yes, Mainstreet is down to a 2.5% lead as of this morning, but from watching their rolling sample I’m pretty sure they have a good Ontario CPC number heading out of their sample tonight. Even if I’m wrong, it’s one pollster out of 13 who have released since the start of the writ.

The other problem, as Mainstreet’s own model shows, the Liberals are still comfortably in majority government if their poll is literally accurate. What people are doing is cherry picking the best crosstabs from the various pollsters, stitching together a zombie poll that points to a close race, and then saying that the trend they’ve created in their own head must continue. If you take Abacus’s BC numbers and Research Co’s Atlantic tab and the Angus Reid Quebec number where the Bloc still gets 30% plus this Mainstreet number with the Leger GTA tab, plus assume the Conservatives lose nothing in SaskyToba and Alberta, and you tilt your head at a 77 degree angle and then fart to the rhythm of Billie Jean while watching a compilation of Jose Mourinho interviews, you can convince yourself that we kind of, sort of, maybe have a close race.

One of the ways to think about upset likelihood, whether in politics or sports, is to think about how many things would have to go wrong before the favourite would have to lose, and how likely it is to happen. Take a theoretical Sens-Leafs Round 1 playoff series - all the Sens really need to win is Linus Ullmark to be meaningfully better than whichever of Stolarz and Woll the Leafs end up going with, which is not exactly a huge hurdle. Compare that to what the Sens would need to beat Florida, which is Ullmark to outduel Bobrovsky, Matty Tkachuk not to play or to be legitimately useless, Ekblad to be rusty when he returns, and Marchand not to mesh with his new teammates. See how one of those is a lot less likely?

The Conservatives have a path, just like the Sens do if Florida wins the Atlantic, but it’s so chock full of conditions that it might as well not materially exist. The number of things that have to go right for the Conservatives - claw back huge ground in BC, hold everything in the Prairies, thread a very difficult needle in Ontario, get a Bloc revival in Quebec, and then make big ground up in Atlantic Canada - is preposterous. There’s just no other way to say it. There will either be a broad based revival in Conservative fortunes across a series of the 12 national pollsters who have released this cycle, or there won’t be. But this idea that the data we have right now suggests anything other than a landslide is crap.

I keep harping on the number of polls we have right now for a reason - no individual poll is worth that much right now because it can’t be. Even if I were to do some fancy weighing by past performance thing, at 12 pollsters you’re looking at an average weight per poll of 8.5% of the average. Even if I gave Leger 15% and Abacus 10% and down weighted Franky, it’d still be basically the same fucking results because none of this is very different.

There’s no maximalism in saying this, as one columnist accused me of yesterday, but right now, on all the evidence, there is no evidence of anything other than complete Conservative annihilation. If your response to that is to say that maybe the Tories have a chance while refusing to put a number on it or in any way actually tie yourself to a falsifiable prediction, that’s your right. Let’s not pretend that it’s actually analysis.

A Conservative recovery is possible, though incredibly unlikely. But if it happens it won’t be visible because you can cobble together something that vaguely looks like something resembling a statistical case for it, it’ll be because you don’t sound like Kramer doing the fucking Magic Loogy bit from Seinfeld explaining how it happened.

If I sound dismissive of this shit, it’s because I am. This campaign is not anything resembling close right now. It might get there, but it’s not right now. Cherry-picking bits of data to make your point and believing data that strains all credulity just because it makes your side look better isn’t analysis. It’s hackery. I would know - I’ve done it plenty in my life.