Polls, Biases, Shy Tory, And Historical Defaults

On The Tsunami Of Bad Polling Takes

How much should we care about vote share?

Let’s be charitable, and walk through this math, shall we? Let’s give the Bloc 30% in Quebec (6.9% nationally, which we’ll round up to 7%), the micro parties and independents a combined 1%, and assume that between them the Greens and PPC get 5%, which is probably fair. The Bloc aren’t polling 7% nationally, and the Greens and PPC getting 5% would be a good result for two parties being squeezed, especially with what we know happens to minor party vote shares as election day gets closer. That’s 13% of the vote combined, down from 15.82% in 2021 for those three categories of votes.

Then you’ve got the NDP, which the Toronto Star’s polling average has at 7.3%. If you call it 8, to be generous, then you’ve got 21% of the vote going to parties outside the majors. That number in 2021? 33.64%. So, you’ve got 79% that can go to the two major parties, and that’s with generous assumptions about the Bloc and NDP.

If the NDP get 7%, which seems likelier to me, and the Bloc get 25% in Quebec and not 30%, then the math gets even easier, but let’s be real about this: the loser of this election - unless either the Liberals or Conservatives race out to a 10% lead - is going to have a popular vote share that most winning parties don’t achieve. Even a comfortable plurality - say, 4% - would pencil out to a 37.5% vote share for the losers, even with generous minor party assumptions. If it gets closer, you’re looking at a loser’s vote share that rivals what Chretien got in 1997 in a majority government winning election.

Why does this matter? Well, one, you’re starting to see a lot of people talking about vote share as if it’s actually predictive - if Party X gets Y share, they’ll win because they’ve won the last Z times they’ve won Y share. It’s a facile and bad bit of electoral math, a version of the 2020 Georgia Runoff stat that in the history of runoffs, Republicans had done better on runoff night than at the General Election in all 8. Since those stats were quoted, the GOP are 0/3. It’s bad numerology, not actual analysis.

The other thing that’s worth talking about is the idea there’s some “Shy Tory” effect in Canada, which misunderstands the history and who it is likely to help in Canada. There is a decent chance the Liberals are being overestimated in the polls, slightly, in Alberta and SaskyToba. The Sasky NDP just underran their polls pretty significantly in 2024 and the UCP popular vote margin was bigger than the polls in 2023 and 2019, as well as the Federal Alberta polling in both 2019 and 2021. There’s a decent chance that this is inflating the Liberals slightly. But this is mostly irrelevant, because there’s so few Liberal-CPC battles in those three provinces. Respectfully, whether the Liberals lose Saskatchewan by 20% or 50% will not flip a single seat, but that’s a difference in the popular vote. It’s almost like I wrote a whole column about it this week.

But that’s a different thing from some form of “Shy Tory” effect, coined for the UK Conservative shock victory in 1992 and oft-referenced around Trump. Here’s the problem; what is often called “Shy Tory” is actually just the historical tendency for undecided voters to break for the party of that country’s historical dominance. The British Tories are the dominant party of Britain, and Labour governments are almost definitionally aberrations. The same is true in Australia, where the only interesting national polling miss in recent times aided the right, but where Labor governments are rare.

However, move outside of these two countries, and you see a different pattern - the French far-right almost always underrun their polling, as the last two Presidentials and the snap Parliamentary elections show. In France, at least in the 5th Republic, French politics is oriented around stopping the extremes - originally the communists, now the far right. The centre almost always holds.

Where does Canada fit into this? It’s simple - the Liberals are, for all our flaws, the default option. Conservatives have to work harder to win elections than Liberals do, because Liberal government is seen as the default in this country. Liberals don’t have to win elections so much as not lose them, while Conservatives have to win them. It’s probably unfair, but it’s the same way with British Tories and the Australian Coalition - it’s reality. And a lot of what looks like Shy Toryism in the UK is just that reality - tie goes to the Tories.

None of this is to say the Liberals will win, or sweep, but a lot of the bad assumptions about the polls are nonsensical. This is much more likely to be something like UK 1983, where a crisis saved a government against an official opposition leader who was more fulsomely and stridently ideological than his predecessor. (To be clear, Poilievre isn’t close to being a Michael Foot level disaster, as much as it’s a fun joke to make.)

The problem with international comparisons on polling is that the intricacies of the local politics are just reduced to left/right overperformances, ignoring incumbency, history, and context. We love a good trend, but it’s irrelevant.

The point I made a lot in the run up to 2024 is that if America were any other country I’d bet on the Democrats beating their polls, and probably sizably. There is no international comparison for the polls consistently being off the same direction three times for a national election. Misses in the UK split (2015/2017), in Australia the miss didn’t repeat in 2022, and in Canada to the extent we’ve had misses they mostly cancel out over time. But, I made the point that America is not every other country. Well, Canada’s not either.

We are one of the few countries on earth where our default party of government is left of centre. It’s a truth that most democracies default to the right - certainly true in Australia, since the introduction of MMP in New Zealand, definitely in the UK, Germany, and famously Japan (though whether Japan really counts as a democracy is, you know, complicated). Canada is abnormal in this regard, and you can’t just lazily assume Shy Tory applies because you saw some meme about a polling miss. It’s not how it works.

The polls in Canada have been remarkably accurate. We have a large collection of pollsters with different methods and samples and approaches that are all getting the same basic story, which is a Liberal government and a not even particularly close battle for it. Everything else is a distraction from this basic fact, now fucking relax and enjoy the shitshow that is the fact the Conservatives are already at each other’s throats in that Global story.