Pollster Wars: Are The Liberals Reviving?

On EKOS, Abacus, And Conflicting Data

Frank Graves has created quite the controversy, with the latest Conservative lead at a measly 4%. It’s movement that is directionally similar to what some others have seen, but at a level that’s yet to be replicated. Mainstreet had a 10% reduction in the CPC lead in their last poll, but that is a 10% swing left from a 29% deficit, and some of that 10% swing might have just been Mainstreet normalizing from a very pro-CPC base.

Abacus, on the other hand, hasn’t seen much. Their weighted, partial sample stood at a “20-ish” point Conservative lead as of Friday midday, which is a 6-ish point swing from their last poll but also not exactly a race in single digits. (Abacus CEO David Coletto will be on the Scrimshaw Show next week to talk about the state of politics, which is awesome.) And with this polling disparity, there’s been a lot of intrigue. So, let’s dig into this, and see what’s real and what’s not.

It’s worth starting with the obvious - the politics of the proprietors of these companies is not particularly relevant to their outputs. Frank Graves didn’t stop being a progressive when he had the Conservatives winning by 25% after the Freeland resignation. Mainstreet, which is currently being called a fake Liberal pollster in Ontario and whose proprietor is an Ontario Liberal member, had the most pro-Conservative polls federally until their last poll, and missed low on progressive parties in both New Brunswick and BC. (In both cases, narrowly and well within margins of error - this is by no means an anti-Mainstreet point.) Ipsos is run by an annoyingly Conservative asshole in Bricker, but they’re generally on the call - and in 2021, it was Graves and Quito that saw the huge Conservative momentum in August that Angus Reid and Ipsos never really did. The idea that the ideology of the people running these companies is hugely relevant is nonsense.

It gets even weirder when it’s people assuming an ideological bent that isn’t in evidence, because it’s utterly insane. I’ve paid attention to Coletto’s work for nearly a decade and I couldn’t tell you who he’s voted for at any level of government ever. He was also the pollster who had some of the Conservatives’ worst numbers in the COVID Parliament, so just on the public evidence it’s bullshit. But it is a hilarious conspiracy theory that he is some Conservative plant or a right winger, given the fact that he’s come on my show twice and given me free, exclusive first looks at Abacus data twice. If he were truly committed to his supposed right wing politics I find it hard to believe he’d give me data for me to use, given my shitlib tendencies. It’s just ludicrous.

I think what happened here is actually pretty simple - phone polls are bouncier in general and more prone to swings in enthusiasm as opposed to genuine change in sentiment. It’s why the panel pollsters in 2021 never saw the big swings the telephone and IRV pollsters did. That sometimes means that panel pollsters are slow to respond to big swings, but there’s also sometimes quite a lot of noise from phone samples. Both approaches have their virtues, and both their vices.

I think EKOS and Mainstreet (which is now running a hybrid text/online survey but definitely isn’t an “online” poll in the same way the panel pollsters are) are picking up a legitimate uptick in pro-Liberal sentiment. This is an understandable uptick because of how singularly unpopular Justin Trudeau was. The idea - as I’ve written about 4000 times since the 2024 Budget was full of broadly popular or at least not unpopular ideas and yet the government’s polling went down - that the government would rebound when Trudeau resigned is logical. Now, it being logical and it being certain aren’t the same thing.

Abacus’ last poll left the field Tuesday the 14th, the day after the Daily Show Carney interview aired. Whatever impact that first impression had did (essentially) nothing to impact the Abacus poll. Graves said he started to see the big swings that week, which means Abacus would have left the field at the worst time. That Coletto is now seeing a real, though by no means huge, swing suggests there’s a real trend occurring.

The other part of this is that Abacus’ Ontario number was probably a bit high last time. Nick Kouvalis, who is probably (read: definitely) polling for the PCPO, released an Ontario-only Federal number that had the CPC up 17 in the province. Coletto had them up 27%. The combination of a slightly high CPC Ontario number and the field dates led to this situation where it seemed like they were truly seeing different things, as opposed to seeing the same trend at very different levels.

The other part of this, however, is enthusiasm. Panel pollsters in the abstract tend to stick to a very rigid structure, meaning that whether you get 25% of your unweighted sample as having a University degree or 55%, you’re not going to get wildly divergent results, because upon weighting it goes back to the predetermined weighting scheme. This is why they tend to be less bouncy. One of EKOS’ releases had 55% of the sample as University educated, and the Liberal rebound is significant amongst the educated. (About 26% of the Canadian electorate has a University degree.) Now, that 26% number is not gospel, because differential turnout is real, but plainly what we’re seeing from EKOS is a turnaround essentially without precedent. It would make a lot of sense if part of why it’s so big is that they’re getting an overly educated sample, and educated voters - who pay more attention to the news, are likelier to know who Mark Carney is and have seen the Daily Show clips, and generally are likelier to have been moved by recent events.

Far from being some conspiracy, it seems rather likely that what we have is the virtue of a broad polling environment. It’s possible that the panel pollsters are missing something, and that actually the movement is closer to the 10% or even higher, as disaffected Liberals and progressives who would have sat out a Trudeau-led contest are energized for a new one. It’s possible that Coletto is correct that it’s a real but more muted swing back. It’s also entirely academic because it is impossible to discern which pollsters’ non-writ polling was more correct. If Carney somehow led the Liberal Party to victory because Pierre Poilievre started doing Roman Salutes every day, that wouldn’t make Abacus wrong. If Carney admits to having been part of the plot to kill Kennedy and he loses by 35%, that doesn’t mean Graves was wrong.

The pollster wars are fun, and I’m obviously going to ask Coletto about all of this Monday. But we have a volatile news environment and a lot of things happening very quickly. Expecting unanimity from the pollsters in this time is a lost battle. The Liberals seem to be recovering. We have no real idea how fast. Let’s relax while it all shakes out.