Good morning. Polls might not tell us much at all about how Americans will actually vote next week – more on that below, along with the hunt for a child-care spot in Canada and the Conservatives’ latest fundraising haul. But first:
Today’s headlines
- U.S. Senator Elizabeth Warren tells prosecutors that TD Bank executives should have been punished more severely
- Alberta tables legislation regulating pronouns in school and gender in sports
- Israel strikes Beirut’s southern suburbs and Gaza after a U.S. ceasefire push
U.S. Election
Calling the shots
Depending on which U.S. poll you take a look at, Donald Trump is down by one point nationally (Reuters/Ipsos), up by two points (AtlasIntel) or dead even with Kamala Harris (New York Times/Siena). The candidates are tied in Pennsylvania (CNN/SRSS), definitely tied in Pennsylvania (FiveThirtyEight) or Harris trails by two (Quinnipiac). She seems to lead in Michigan, he might have Georgia on lock, but pretty much all those results fall easily within the margin of error. So even though the race appears to be a nail-biter, FiveThirtyEight cautions that Trump and Harris are both one polling error away from an electoral blowout.
And the polls have screwed up before: As The Globe’s international correspondent Nathan VanderKlippe points out in his new report, there’s been a tendency to underestimate Trump’s support. In 2016, roughly 90 per cent of polls – and one unforgettable New York Times election needle – predicted incorrectly that he would lose. In 2020, the polls underestimated Trump’s national performance so badly that they produced the worst presidential polling error of the past 40 years. Of course, the popular vote only counts for so much, and these two very tight races underscore the importance of the Electoral College – which is why I’ve spent a ludicrous amount of time playing around with The Globe’s swing state tool.
Pollsters need to ensure that the respondents in their samples reflect the overall electorate, both demographically – age, sex, race, etc. – and politically. The general consensus is that in 2016, they failed to adjust for an overrepresentation of white college graduates, who tended to support Hillary Clinton. In 2020, they failed instead to correct for too few Trump supporters (who’d been told the polls were “fake” and didn’t much respond to surveys) and too many Democrats (who were likelier to be stuck at home during the pandemic, happy to chat with a pollster).
This time around, several pollsters told VanderKlippe, their samples will better match the demographic and political makeup of the country. But it can be a real headache to successfully reach voters of any stripe. Cellphones offer the benefit of caller ID and the bother of constant spam calls, so people often screen any number they don’t recognize. (I’m disinclined to talk on the phone even when I do know the number.) The vast majority of polling is now conducted online, but that brings its own problems: It’s hard to get a truly random sample from the internet, and results are usually misleading, because people – especially young people – just lie. In one 2022 poll from the Pew Research Center, 12 per cent of respondents under 30 claimed they were licensed to operate a nuclear submarine. The correct percentage is closer to zero.
And so pollsters must once again tweak the knobs on their data to tamp down any bias, which also means making educated guesses about turnout, electorate composition, new voters and late-deciding ones. This week, Vanderbilt University political scientist Josh Clinton took one national survey from mid-October and subjected its results to a variety of adjustments, weighted to the demographics, party identification and voter enthusiasm of three previous elections. Each adjustment produced a wildly different outcome, ranging from a near-even split to a nine-point leg-up for Kamala Harris.
But as we’ve all heard over and over (and over), these are unprecedented times, and it’s hard to know if past patterns can predict future behaviour. Despite the very best efforts of some very smart folks, correcting for the lessons of 2020 means you’re “fighting the last battle,” as Marjorie Connelly, a senior fellow at The Associated Press-NORC Center for Public Affairs Research, told VanderKlippe. How on Earth, then, are we meant to read the 2024 polls?
I asked VanderKlippe for his best advice: “In all likelihood, new reasons for error will become clear after the votes are counted,” he told me. “Which means it seems entirely likely that current polls will once again miss the mark. In an election that already appears to be a toss-up, the only rational reading is that we should not be surprised by just about any outcome.” In that case, instead of frantically stress-refreshing the latest battleground polls, maybe we should all just go outside and touch grass.
The Shot
‘Finding a child-care spot in Parry Sound is like winning the lottery.’
Child-care fees may have dropped across the country, but available spaces are still enormously hard to come by, especially in smaller communities. Read more about the parents stuck in Canada’s daycare deserts here.
The Wrap
What else we’re following
At home: As the clock ticks down to a federal election, the Conservatives are still raking in millions more than the Liberals or NDP, in what the party expects will be another record-setting fundraising year.
Abroad: Since the war in Sudan began last year, as many as 150,000 people have died and more than 11 million have been forced from their homes – but the UN isn’t prepared to send in a peacekeeping force yet.
More cuts: The Bank of Canada hasn’t ruled out chopping interest rates by a half-point again in December.
More chills: Fall’s latest crop of mystery books will keep you up all night. (Start with the new Louise Penny.)
More Taylor Swift: With Taylor about to touch down in Toronto and Vancouver, The Globe wants to see your very best Eras concert outfits.