Polling Roundup: Carney Knocking On Majority’s Door?

On Leger, EKOS, And The Polls

There are two very important things to keep in mind as the first Carney-as-LPC Leader polls start rolling out this week.

1. Even if the polls are correct, being correct about the state of play in mid-March is not a prediction of the state of play in late April when the election will be.

2. We are blessed to have lots of good pollsters in Canada - do not overreact to one data point no matter how good or bad.

We’ve seen some of both come into play so far - Tuesday morning’s Innovative Research Group poll was proof Carney is leaking air despite barely moving from the week before on vote intention, Monday’s Pollara data was proof he’s strong. We need to narrativize this, so we overreact to single data points. It’s dumb and bad.

This week will see a rush of data beyond what we’ve already gotten - both Mainstreet and Abacus are confirmed for releases this week, plus Pallas is probably due, having been a month since its last release. But with the release of Leger and EKOS, it’s probably time to take stock of where we are.

I’ll start with EKOS, not because it’s the more important data point but because I’m hopeful the Leger crosstabs will be out by the time I’m done this section. (Editor’s note: they weren’t.) EKOS and Frank Graves has the CPC down 18%, a Liberal landslide of the kind that last happened against the united right in … ‘68? (Hilariously, also a minority parliament swap of leaders.)

It’s almost assuredly wrong in magnitude, but honestly I think everything at EKOS probably agrees on that to some extent. What is more interesting is whether it’s directionally correct or not, and it probably is. Prompting for Carney, as opposed to New Leader or whatever generic EKOS was doing, is worth some points, and they already had a Liberal lead. It’s not shocking the lead is in double digits based on those two factors. It’s just shocking it’s 18% and not 11%.

I will say EKOS does itself no favours by the seemingly random nature of EKOS’ releases. I do not believe the conspiracies but the lack of predictability and the inconsistent nature of the release (screenshots of the data viz on Twitter versus a full release, how many days worth of respondents to include) makes it easy to create them. If EKOS were to standardize those questions - say, biweekly releases of a 5 day roll on the same day every two weeks - it would go a long way to erasing (false) claims we only see EKOS when the data is good for the left.

On the other hand, you have Leger, which shows a tied race nationally and has shot my model left. It is a pure fact of statistical reality that a poll that is already good for the Liberals moving left will move the model less than a bad poll for them moving to a good poll. The inclusion of Leger has the Liberals nearly tipped over into majority in my model with 168 seats, which is why I lead this piece with point #1.

(This is a wonky point but today’s Leger led to an outbreak of my absolute favourite discourse, Why Do Certain Models Have So Many Liberals In Alberta? The answer isn’t incompetence or even Liberal tendencies, it’s math. The Liberals got 15% in Alberta last time and Leger has them at 27%, a 12 point increase. It’s also ~180% of their previous vote share. Models that use proportionate swing overstates the swing of a party off of a low base.

Take Calgary Centre, where the Liberals got 30% of the vote and lost by 20%. Straight swing on that poll would have them lose by ~2% on that Leger poll, as the Liberals get to 42%, the Cons fall to 44%, and the Cons hold narrowly. In a proportional model, the Liberals break 50% of the vote and win. It’s a methods choice, and one that I stridently disagree with exactly for situations like this. It’s a method that works fine for big parties but unduly punishes and benefits smaller parties, or major parties in places where they’re irrelevant like Liberals in Alberta.)

Mark Carney, if the election were tomorrow and the polls were accurate, would win the most seats, and it would be lineball for majority government. The election, of course, is not tomorrow. It hasn’t even been called yet. The polling doesn’t account for the fact that Carney is a new candidate who may stumble. It doesn’t account for Doug Ford saying Carney “understands finance like no other person” and getting breakfast together this morning like best buds. It doesn’t account for any of the 64000 things Donald Trump will say about Carney between now and polling day.

If Abacus swings left this week, then the model will probably tip into majority. They’re the worst poll for the Liberals still in the model, and so if they replace their last with something more favourable (even if a Poilievre lead), it’ll move the model. Mainstreet CEO Quito Maggi said the Liberals are up 10 in Ontario right now in a continuation of his Twitter war with Nick Kouvalis - if that’s confirmed by later sampling then that would also be enough. Until pollsters with good Liberal numbers start swinging back, it’s not enough to move the model back. (The funny thing is, Nanos could well swing right nationally soon and be better for the Liberals. Their vote is way too good to be true in the Prairies, which is how the national vote is essentially tied, but they’re down 10 in Ontario. A big Prairie correction but Ontario swinging left could see the Liberals gain in seats and losing vote share nationally.)

The Conservatives are in a lot of trouble. They can reverse it, but they’re acting like chickens with their heads cut off. Carney is a credible challenger they can’t get their heads around, and it shows. Things could fall apart, obviously - but there’s every reason to also think that Poilievre’s machine, a machine that was never as good as the hype and always propped up by Liberal ineptitude, is falling apart and can’t be put back together in time.