Polling Isn’t Broken

It’s just telling us we don’t know anything.

It was a bad week for Joe Biden; it was a good week for Joe Biden.

First the bad. The Economist and 538—the two most respected polling models—released their 2024 presidential projections this week. Both of them were well stocked with nightmare fuel. 538’s model, which considers fundamentals and polling, had Biden/Trump in a dead heat, 50/49. The Economist, which weighs polls more heavily, had Trump with a 2 in 3 chance of winning.

To balance that out, and just generally to confuse things, though, there were actual election results this week, in the form of an Ohio special election (OH6). Trump carried the district in 2020 by 29 points; the Republican incumbent, Bill Johnson, who just resigned, won reelection in 2022 by 35 points. The Democratic challenger, Michael Kripchak, was an unknown waiter who only spent a measly 25,000 to challenge well funded state senator Michael Rulli. Everyone expected a blowout.

Instead, Kripchak came within 9, a stunning 20 point overperformance expected by literally no one. More, he mainly overperformed in rural areas. That complicates the argument that this was a case of a low turnout election being swept by more engaged Democratic suburban voters.

Polls indicate Biden is in big trouble. Special elections suggest the GOP is going to get annihilated—if Ds actually overperformed by 20 points—or even 10 points—in Ohio in November, it would mean the biggest Democratic presidential win in probably 50 years. Both of these things can’t be true.

¯\_(ツ)_/¯

So, what should we make of this? There is a clear answer to that. The answer, as political scientist Jonathan Bernstein pointed out earlier this week, is that no one knows.

The data we have for this election just does not make a lot of sense. The economy is currently humming along with stunning job and wage growth, flat inflation, and low unemployment. But economic sentiment remains horrible; more than half the country thinks we’re in a recession. Biden’s approval reflects that dismal outlook; he’s currently at a wretched 38% approval, 18.5 points underwater.

That seems bad for November. But also, again, half the country thinks we’re in a recession and the president’s party just overperformed in a special election by 20 points. The president’s party usually gets crushed in specials. When you’re in a recession, the president’s party usually gets destroyed in specials. Instead, rural voters in Ohio said, “we’re in a terrible recession, I’m going to vote for the president’s party!”

Again, and I can’t emphasize this enough, nobody knows what the fuck is going on. If someone tells you they know what the fuck is going on? They are deluded. Or lying. Maybe both. Because (again) they do not know what the fuck is going on.

And to be clear, this also applies to people who tell you the polls are broken. Polls in 2022 were in fact very—almost bizarrely—accurate. It’s too early now for polls to be very predictive, but we don’t have strong evidence that they’re broken or not working. Maybe we will come November. Maybe we’ll know then that the specials were more predictive (which isn’t usually the case.) But right now? ¯\_(ツ)_/¯

Polling Isn’t Useless

Again, a lot of people insist that this all means that polling is useless and that data analysts are just conning us.

But I don’t think that’s accurate. Most reputable data analysts (like 538 and Split Ticket) will be quick to admit up front that polling this far out isn’t going to predict the election, and that polling always has a good deal of uncertainty. The problem isn’t that polls are broken. The problem is that people expect polls to predict the distant future with certainty, which they just are not ever going to be able to do.

They can provide some useful information though. In this case, polls are telling us that Donald Trump has a real chance to win. That’s important. In 2016, Democrats were convinced Trump could not possibly win—which I think reduced Democratic turnout, and may well have contributed to Trump’s victory. In 2020, strong Democratic polling led many to believe the election would be a blow out—again, the overconfidence didn’t serve Democrats well. So it’s not a bad thing that we’re all much more aware, now, that Trump has a lot of support, and could well be the next president.

Polls also are helpful when they are confusing. We know economic sentiment is out of sync with economic fundamentals; we know that Biden’s approval makes no sense in light of Democratic overperformance in specials. Without polls, we might think we have a good sense of what’s going on. But thanks to polls, we know we don’t really. The presidential election, and the House, and the Senate, are all a black box right now. We know we don’t know. That’s not reassuring, obviously. But I think it’s better than overconfidence. The best we can say right now is that we can win, but it’s going to be a fight.