Pollster Wars: EKOS Has Been Proven Right

On Nanos, Big Swings, And Vindication

This site has long had a policy against overreacting to Nanos, and it’s something that I maintain is true. The four week roll is bullshit, Nanos is prone to wildly overplaying his hand in his commentary, and the fact that his regionals are both behind a paywall and significantly less useful than everybody else’s by virtue of his dumb and bad decision to lump Alberta in with Saskatchewan and Manitoba in his Prairies tab makes his polling deeply frustrating. The classic example was when the Liberals recovered after signing the health transfer deal a couple years ago, news that was out for about half of one week’s roll. And yet, that’s what the article gave credit for the Liberal recovery, with I’m pretty sure a snazzy quote from Nik. It’s all just frustrating, even when it’s good for the Liberals.

That said, Nanos’ release and some tweets from David Coletto do give us enough to say to make something pretty clear - Frank Graves was right, and a lot of the shots taken at him were wrong. I have been sceptical of the size of the movement, and I think anybody who genuinely thinks we should blindly believe any specific pollster is lying. But when pollsters release their work even when it’s implausible - as I described it - it’s eminently worthy of respect. And it’s even more worthy when it’s vindicated.

Whether or not Mark Carney wins the next election or it’s a messy Hung Parliament or whatever else, the broad strokes of the political realities are now much closer to where Graves had them than where others used to. Beyond him and Mainstreet, Ipsos just showed a 13 point swing left, Nanos is down from a 27% lead to a 8% lead in four weeks, Abacus was at 26% pre peak and coming with a poll this week that we know is going to be much closer, Pallas had a six point race last week, and Leger’s Quebec poll shows a huge swing to the Liberals from their own national poll.

David Coletto tweeted on the weekend that nearly 80% of past Liberals (aka, 2021 Liberal voters) were planning on voting LPC in his preliminary sampling (which I believe was English Canada only), and ~20% of past New Democrats were going to vote Liberal. Those two numbers would be, for the curious, 29% of the vote, and it’s probably 30% if there’s any amount of leakage from 2021 O’Toole moderate Tories. (There’s always some amount of leakage.) That would be a 8% increase in the LPC vote in two weeks.

More interestingly, at the end of last week Coletto tweeted that by Abacus’ “Voter Pools” question - % of the electorate that would consider voting for a party - the Liberal pool was up 12% and the CPC pool was down 8%. Those are numbers that suggest that something very real is happening. Abacus is also one of two polls conducted in the last four weeks currently showing the worst results for the LPC, which means once Coletto releases the final poll, one of the two worst results for the LPC will cycle out of the polling aggregates, or at least be significantly downweighted. If the final results look anything like what Coletto has hinted at, then it’s highly likely my model - which currently has the Conservatives with 179 seats - would tip to the Conservatives being in hung Parliament territory.

(For the 12 people who were confused why Nanos’ big swing left didn’t move the Liberal vote share up - and in fact cost them 1 seat - the reason is that I finally went in and accounted for all manner of candidate effects - removing personal votes from retiring members, giving boosts to new incumbents from 2021 who have accrued this incumbency, and adding where non-incumbent parties are running other elected officials, like Joel Harden in Ottawa Centre. I plainly should have done that before I launched it but I wanted to get the model “done” so I could track the polls as soon as possible.)

Plainly, the debate about whether it’s a 13% lead (Ipsos) or a statistical tie (EKOS) or something in between (Pallas, Nanos, probably Abacus) is fun, but it’s academic in so many ways. This is what a normal polling environment looks like, and one that we’ve seen a lot before. If it’s really a big swing on and an 6-8 point race - which is kind of what it feels like is being congealed to - then it’s not shocking to see some people seeing bigger swings and some lagging. Expecting unanimous readings is asking to be lied to.

But what is worth acknowledging is that whether it’s really a six point race or a 2 point race or flirting with double digits, this used to be a 20%+ race even before Chrystia Freeland resigned and the trend line wasn’t getting any better. We used to argue about whether a couple point swing in a couple polls constituted movement. Now Conservatives are stuck taking solace in the idea it might be a lead that two months ago they would have laughed at as nonsensically low.

Not to be reductive, but what actually matters is that Graves saw a real and credible swing and had the testicular fortitude to stand by it in the face of scepticism, cruelty, and flat out lies about his record as a pollster. He has been vindicated that something big and substantial is happening, and that we are now in a truly competitive election environment. In many ways, this week can and should be the week the debate about whether the Liberals could win the 2025 election dies a final death.

Whether they will, nobody knows. But the idea of a reelected Liberal government has gone from absurdist fantasy to a genuine prospect. And Frank Graves was the first to see it. May we tip our caps.