How Is This Polling Possible?
Polls tell us that the current election is exactly--to a T, state-by-state, exactly--where the 2020 election ended. I find that mathematically implausible.
Something’s been bugging me lately about the state of the Presidential election, and it escalated to something I had to write about when I read this in Nate Cohn’s New York Times column Wednesday:
There’s no sign of the political chaos of the last few months. Instead, the results look typical: Nationwide, Kamala Harris leads Donald J. Trump by three percentage points, 49 percent to 46 percent. Across the battleground states, the race is a dead heat. In every state and nationwide, the polling average is within 1.5 points of the result of the 2020 presidential election. [Emphasis added; several emojis should be too]
Here’s my question: how in the hell can the 2024 race, coming out of Labor Day weekend, look exactly the same as the conclusion of the 2020 election?
Now, I’m not a political scientist—I am a journalist, which means that when I have a question like this I might typically ask a political scientist. But, first of all, they’re all in Philadelphia at the American Political Science Association annual meeting (including the other two authors of this newsletter). And secondly, Jonathan keeps positing—I think correctly—that this election cycle is too anomalous for political scientists to know what past rules can apply, if any.
To repeat: all seven “battleground” states tracked by the Times —Arizona, Georgia, Michigan, Nevada, North Carolina, Pennsylvania, and Wisconsin—have polling average margins between Harris and Trump nearly identical to the election result gap between Biden and Trump four years ago. The FiveThirtyEight site agrees on all seven, and on all four additional states for which it also shows a polling average: Florida (current Trump +4.6, 2020 Trump +3.3), Minnesota (current Harris +7.6, 2020 Biden +7.1), New Hampshire (current Harris +7.0, 2020 Biden +7.2), and Texas (current Trump +6.1, 2020 Trump +5.7).
That’s freakish. Isn’t it? It’s as if I checked the baseball standings one day and found every team within a game-and-a-half of where they finished last season relative to one another. That doesn’t happen.
I know we think of contemporary America as locked into rock-solid partisan polarization, with everybody planted unwavering in red or blue soil—or perhaps more accurately, in the pro- or anti-Trump camp. So of course the margins never change. But the polling decline of Joe Biden, and reversal upon his departure, belies that theory to some degree. Doesn’t it?
An enormity of change has transpired since November 3, 2020. Economic “fundamentals,” which I’ve always been told drive so much of voting, swerved from pandemic disaster through recovery with high inflation and interest rates, to our current complexity. The country was at war then, and is not now. Inward migration surged and then stabilized; violent crime spiked and then plummeted. Fifty years of abortion law vanished in a nation-rattling incineration.
The returning candidate has been through an entire soap opera cast’s worth of drama since that election, including a revolt against the transfer of power, trials for fraud and sexual assault, ceaseless court appearances, and a near-death experience. His opponent emerged as the nominee after the assumed candidate floundered and withdrew in unprecedented fashion. Two new running-mates were put forward. A celebrity-level third party candidate ran, and then threw his support to one of the major party tickets.
I mean, there’s a reason Jonathan warns against applying metrics that have worked in other cycles—this one has been full of never-happeneds and what-the-hells. A million and one things, each of which might have nudged the contest a little in one direction or another—on top of the usual turnover of one cohort dying, and a different one becoming eligible to vote. Yet here we are apparently at the very same place we were then. Did none of these things have any effect at all? Did they all nudge just enough to cancel each other out and bring everything back to square one?
And even if that’s so: something should have shifted within the states. The partisan composition of states isn’t static.
From 2012 to 2020, the overall vote margin barely changed, from Obama +3.9 percent to Biden +4.5 percent. But the margin in Arizona swung nearly 10 points toward the Democrats; Iowa moved from a six-point D win to an eight-point R victory; Ohio moved 11 points; Texas 10; Wisconsin 7; Michigan 6; Georgia 8. Hillary Clinton’s national margin of +2.1 percent was similar to those elections before and after, but her state-by-state margins varied from both.
Yet this year we are to believe that there are no such shifts. None. As best as we can assess them, every state stands upon precisely the same polling perch a full four-year cycle further along.
Maybe this is just a temporary, bizarrely coincidental alignment of the data—perhaps new polling will quickly reconfigure the numbers, and my question will be moot. Not that I know if it’s meaningful even if we really are trapped in some sort of Groundhog Day election scenario.
Perhaps it’s just this simple: Trump and anti-Trump are almost immovable quantities, Biden’s frailty was a rare exception, and once that anomaly was removed the equilibrium re-emerged.
But, then why do today’s polling averages match the election results of 2020, rather than that year’s polling averages? Which, on both a national and state-by-state basis, very much did not match the election results anywhere near as well as the current polling margins do four years later and with a different candidate and all those other changes?
Have today’s polls corrected for something that the 2020 polls had wrong? Or added in a new bias, that happened to shift the numbers to exactly where the last election ended? Have all the margins of error by chance aligned to create this mirage, this simulacrum of the 2020 election results? Might they show me winning lottery numbers next week?
Clearly, I am out of my depth; as I said, I am not a political scientist. If any out there can offer me an explanation, I would welcome the input.
This is a great column David -- and I'll be interested in Julia and Jonathan's response post-APSA. I think Jonathan is right that this is a special cycle and we should be wary. IMHO, this is not like baseball because teams change more than voters in a year but it does seem bizarre for it to be so precise. Polarization has led to more "team" voting and that would explain some consistency as "red" or "blue" voters cannot imagine crossing party lines. But that doesn't explain those in the middle.
Well, we know from Washington primary results that the election should land somewhere (in terms of D+ margin) between 2020 and 2018. Special election results (https://docs.google.com/spreadsheets/d/1ajyphWQru9TgDDiBe8kvEmApBEXND2wl9MVaxi1dndk/edit?gid=995309150#gid=995309150) are telling us the same thing. Polling is landing there, too, if you account for margin of error.
Given what we're seeing (special elections + Washington primary + polling), I expect we'll see the 2020 presidential election map plus North Carolina in the D column.