The Unpollable Election

There’s no point in reading this newsletter next year.

I shouldn’t say that, as it’s bad for business. Let me rephrase: You should definitely read this newsletter next year and should happily pay for the privilege of doing so. Maybe even send Steve and Jonah a note with your subscription order saying, “Thanks for being you.”

What you shouldn’t do is take anything you see in this column too seriously, at least as it pertains to forecasting the election.

If Donald Trump ends up annihilating Ron DeSantis and Nikki Haley in the early states, as I expect, the general election campaign will begin in earnest by the end of January. Over the nine months that follow, we’ll spend many hours together parsing polling. There will be endless statistical sturm und drang and much earnest navel-gazing by yours truly about what the latest 3-point shift among demographic XYZ might mean.

And it’ll all be very silly. The hard truth is that, in this cycle of all cycles, no one knows anything.

Well, fine: We know that Joe Biden is in grave danger of losing. That’s something.

But whether he’s ahead or behind, how deep a hole he’s in, whether his base might come home to him, whether Trump is scaring off voters tempted to give him a second chance—these will likely remain mysteries to a considerable degree all the way to Election Day, whatever the polling might say. I’ll frankly be surprised if we don’t have an historically enormous share of exasperated “double-haters” making up their minds in the final weeks of the campaign.

Elections weren’t like that when I was a kid.

In the 1980s and 1990s, it was easy to tell who’d win a presidential election many months in advance. All you had to do was ask, “How’s the economy?” A bad economy meant the incumbent was out, a good economy meant the incumbent would cruise. 

In 2024, people can’t even agree on whether the economy is good or bad. If they do ultimately end up agreeing, there will still be multiple major “X factors” that could plausibly swing the election either way at a moment’s notice. And if that isn’t confounding enough, there are reasons to think presidential polling right now isn’t capturing the true state of the race—although whether it’s underselling Biden more than it is Trump, or vice versa, is unclear.

No one knows anything.

On Tuesday the New York Times published a national poll that had Trump ahead of Biden among registered voters … and Biden ahead of Trump among likely voters.

That wasn’t the crazy part. The president led among likely voters by only 2 points, which would be enough to win the popular vote again but probably not enough to win the presidency because of the structural advantage the GOP enjoys in the Electoral College. Whichever Times numbers you use, Trump likely ends up being reelected.

Here’s the crazy part:

In this particular poll, the split isn’t just between midterm and non-midterm voters. It’s between people who voted in the 2020 general election and those who didn’t. Mr. Biden leads by six points among voters who participated in the 2020 election, while Mr. Trump holds an overwhelming 22-point lead among those who did not vote in 2020. In our estimation, needless to say, 2020 nonvoters are less likely to vote in 2024, and that’s why we show Mr. Biden ahead among likely voters.

We should be careful about taking people at their word when they’re asked whether they voted in 2020. According to the Times’ Nate Cohn, 51 percent who said they voted that year claim to have voted for Biden versus 41 percent who say they voted for Trump. That 10-point margin is more than twice the actual margin of Biden’s victory.

But if it’s true that the president is leading comfortably with 2020 voters, whereas Trump is leading overwhelmingly with nonvoters, you can see why this election will be so difficult to poll. Everything will depend on somehow measuring sentiment and projecting turnout accurately among a cohort that couldn’t get motivated to get off the couch the last time these two candidates faced each other.

You might argue that those nonvoters will show up for Trump this time because, after all, the country is in a foul mood and he’s been good at persuading low-propensity voters to show up for him in the past. He famously overperformed his national polling in both of his runs for president, didn’t he? Benjy Sarlin of Semafor is right, though, that after seven long years of his bizarre demagogic shtick, he may have finally reached the point of diminishing returns. By definition, he failed to persuade 2020 nonvoters to turn out for him last time. Perhaps he’s mobilized every “forgotten man” capable of being mobilized by Trumpism. 

You might also argue the opposite: that 2020 nonvoters won’t break as sharply for Trump as expected because that group contains millions of young Americans who only became eligible to vote in the last three years. Young voters favor Democrats traditionally, with Biden having won the 18-29 group by 24 points in the last cycle. He’ll be sitting pretty if that trend holds in November given that a huge bloc of left-leaning Zoomers is poised to join the electorate, replacing another huge bloc of right-leaning senior citizens who have since passed away. Some pollsters believe it could make all the difference.

Here we run into another polling problem, though. The most recent presidential surveys taken have younger voters breaking toward, er, Trump.

Yes, really. In the Times poll, young registered voters split 49-43 for the Republican, a shift of 16 net points since July. (I wonder why.) A Fox News survey published on Monday found Trump leading 41-28 within the same group. Morning Consult had a comparatively good result for Biden, finding him still ahead among voters in battleground states born after 1997, but in that case his lead was just 6 points. If he goes from a 24-point advantage among young voters in 2020 to a 6-point win in 2024, he’s finished.

I don’t expect Trump will win the Generation Z vote—“poisoning the blood” isn’t a message tailor-made for a racially diverse left-ish bloc like twentysomethings. But when a demographic is in political flux to this extent, only a fool would feel confident forecasting where they’ll settle. Sarlin made an arresting point about their political loyalties:

They’re not “anchored” to Joe Biden. Granted, apart from drawing occasional analogies between Trump and Hitler, Democratic organizers and ad teams haven’t gone to work on them yet. At some point soon, God willing, the war in Gaza will end and Hamas will lie in ruins, lowering the temperature among young voters who are sympathetic to the Palestinians. But high consumer prices and unaffordable housing will linger; so might the grudge that progressives bear Biden for having supported Israel after the conflict is over.

Projecting the result of the election requires projecting precisely where, amid all of those hurricane-force political winds, young voters will finally land. Good luck to all of us on that.

But wait. We haven’t gotten to the really tricky part yet.

How confident should we be that Americans are telling pollsters their true feelings about Donald Trump?

That question has bedeviled political junkies since his victory in 2016. The polling that year missed so badly in some swing states that the pros concluded there must be some hidden explanation. The most common theory was “social desirability bias,” a fancy term for respondents refusing to reveal their true preferences to pollsters for fear of being judged negatively. Simply put, some so-called “shy Trumpers” may have felt embarrassed telling the nice people at Gallup that they really did intend to vote for a dissolute loudmouthed playboy sociopath from Page Six. Result: The polls underestimated Trump’s support.

Pollsters adjusted their methodologies afterward, resolving to do better in 2020 by taking care that voters without a college degree, the heart of Trump’s base, were properly sampled. Their reward for their hard work was underestimating his support nationally again, this time by a wider margin than they had in 2016. The theory offered for this latest round of malpractice was “non-response bias,” which describes the impact on polling when members of one party are consistently more reluctant to speak to pollsters than members of the other party are. Because Trump had spent years teaching his fans to distrust and even disdain polls, and because the more highly educated left was more likely to be working from home during the pandemic than the frontline workers of the lower-educated right were, presidential surveys just couldn’t get enough Republicans on the phone to measure accurately how many of them might show up on Election Day.

There will be further tweaks to the methodology in this year’s polling to try to account for that, but let’s be real: Measuring public opinion about Trump has only gotten harder since 2020.

It could be that there are fewer “shy Trumpers” now than there were before. “Shy Trumper” is almost a contradiction in terms, in fact, after eight years of relentless MAGA propaganda aimed at the American right. When a voter supports him today, they’re only too willing to let you know. They’ve never had more reason to be embarrassed about their preference, yet they have never seemed less so. As for the non-response problem, it could be that Trump’s steady lead over Biden in national surveys of late will cure or ease it. The more often pollsters show their guy ahead, the less reason Trump diehards will have to suspect that they’re biased against him and to refuse to cooperate as a result.

It could also be, though, that there are more “shy Trumpers” now than there were before. It’s easy to see why: Since Election Day 2020, Trump attempted a coup, got impeached a second time, spent three years whining about a massive vote-rigging conspiracy he still can’t prove, got slapped with four separate criminal indictments totaling 91 charges, and has resorted to rhetoric on the stump that is, shall we say, Trumpier than ever. If it embarrassed a swing voter in 2016 to tell a pollster that they regarded him as the lesser of two evils on the ballot, imagine that voter’s quiet mortification at still feeling that way in 2023.

We can make this trickier, though. Precisely because the broader right has grown so cultish on Trump’s behalf, it’s possible that some quietly mortified Republicans who don’t intend to support him again have become what we might call “shy anti-Trumpers.” Consider the reaction when NBC News chatted recently with some right-leaning voters in Iowa:

“If you go against Trump, like—you’re over,” said Kyle Clare, 20, a member of the University of Iowa’s College Republicans.

“I don’t talk about Donald Trump a lot because I’m afraid of the backlash,” said Jody Sears, 66, a registered Republican from Grimes, Iowa.

“If you would say something negative about Trump, we had one person that would just go bang for your throat,” said Barbra Spencer, 83, a former Trump voter describing her experience living in senior apartments in Spillville, Iowa.

“One Iowan said they plan on saying they caucused for Trump when asked by members of their community but will actually caucus for Vivek Ramaswamy,” the report continued. How many people in Red America answering a pollster’s call will state their presidential preference as Trump because they fear even their spouse knowing that it’s actually Biden? How many will fear irrationally that their pro-Biden sympathies might somehow be publicized by the pollster himself and that it’ll eventually get back to the neighbors?

We can play this game with the incumbent as well. The fact that multiple national polls now show Trump leading Biden among younger voters suggests that that trend is real. But how “real” is it? When twentysomethings tell pollsters they prefer the Republican, do they earnestly want to see Trump reelected, or are they simply trying to throw a scare into Biden to get him to put more pressure on Israel? “I don’t want to vote for someone who is not aligned with my own personal values, as Biden has shown he is not when it comes to Gaza,” one such voter in California told the Times. “Do I vote for Biden or do I not vote at all? That’s really difficult, because if I don’t vote for Biden, I open up the possibility that Trump will win, and I really do not want that.”

How many “shy Biden” voters are out there? How are pollsters supposed to adjust for all of these secret preferences in their topline numbers?

Hold on, though. Thanks to the many “X factors” this cycle, projecting the election will be even more challenging than all of the above might suggest.

As I said earlier, when I was a spoiled child of the late Cold War era, elections were humdrum referendums. The out-party would nominate someone, voters would consider whether they were happy with the economy, and the incumbent would be reelected or not based on the answer to that question. Those elections were easy to poll.

Weirdly, and against all odds a year ago, in this cycle the economy might end up being one of the more stable major variables pollsters have to contend with. Consider the less stable ones.

In this cycle, the incumbent is the oldest president in American history, very much looks it, and is reportedly so intent on trying not to look it that he’s at risk of causing a health event that would destroy his campaign. It’s quite possible that the U.S. economy will have a very nice 2024 and Joe Biden will lose badly anyway because voters no longer deem him fit for office.

In this cycle, for the first time in more than a century, we have a contest between two candidates with presidential experience. The usual dynamic in which the election is treated as a referendum on the incumbent’s term probably can’t hold despite the best efforts of Republicans. It’ll be a “choice election” between not just two parties but two de facto incumbents, and both are broadly unpopular. There’s no precedent in polling for that.

In this cycle, for the first time in more than 20 years, we have at least one third-party candidate who looks strong enough to affect the outcome. Robert F. Kennedy Jr. has a gold-plated name, a high media profile, and a degree of populist appeal on the fringes of both parties. He might soon be joined in the race by Joe Manchin, one of the most formidable centrists in modern America and a man with considerable governing experience. Voters disenchanted with the two major-party nominees may flit between Trump, Biden, RFK, and/or Manchin—and back again—for months, confounding polling further.

In this cycle, for the first time in, er, ever, we have a candidate who stands a solid chance of being convicted of a felony before Election Day. That candidate happens to be the favorite to win the election at the moment. If he’s found guilty and his polling tanks, some of his wayward supporters will talk themselves into supporting him in the end after all. Others will not. It’s anyone’s guess what the proportions will be, how stable they’ll remain during the campaign, and whether members of either group will even be willing to state their honest intentions to pollsters. Who wants to confess to hoping to be governed by a criminal?

In this cycle, also for the first time ever, it’ll be a major issue in the campaign that one of the nominees appears to harbor authoritarian ambitions as president. Some voters will support him for that reason but dare not admit it; others may doubt the scaremongering but grow anxious about his intentions toward Election Day; others, unhappy with Biden and resolved to vote for an independent, might find themselves spooked at the last moment by the thought of what Trump might do and end up reluctantly supporting the president after all, especially if Trump says something hair-raising on the trail down the stretch. Pollsters will need to catch those shifting sentiments somehow.

In this cycle, we have a number of major crises that could easily tip the balance of the election against the president yet are largely out of his control. Ukraine’s defense might collapse as aid from the United States dries up; Israel could escalate further in Gaza, breaking the loyalty of progressives to Biden irreparably; inflation could pick up again or the job market could slow down. There could be another major surge at the border—which is in the White House’s control to some degree, of course. But only some.

Ultimately this will be an election between two candidates whom no one to the left of Matt Gaetz truly wants to vote for. That’s not unlike what we had in 2016, another race that proved famously hard to predict and one in which major developments that appeared at first blush to be fatal to a candidate’s chances turned out … not to be.

And so it’s trivially easy to imagine a meaningful number of voters shifting their preferences—or nonvoters shifting their intentions on whether to turn out or not—throughout the campaign based on whom the lesser of two evils at any given moment appears to be. Picture a huge crisis at the border circa September; “we cannot do four more years of this,” disgusted undecideds will say of Biden. Now picture Trump being convicted on, say, 10 charges a week later. “We cannot do four more years of this,” those voters will cry.

We cannot do four more years of this, whatever “this” happens to be, but somehow we’re going to. Pity America’s poor pollsters, charged with trying to reliably poll an election like that.