What's the deal with Biden's poor polling with young voters?

There's no shortage of explanations for why young voters are unhappy with Biden — without having to unskew the polls.

I’ve been covering polling on American elections for 15 years. Throughout that period, I’ve been fairly consistent in my attitude toward polling. My view is basically: take the average, calculate the uncertainty — which is often larger than people assume — but don’t try to outthink the polls. The polls will sometimes be wrong, but it won’t necessarily be in the direction that you’re hoping for or expecting. Overall this approach has treated me pretty well — the track record of the models I’ve built is for the most part pretty good. In fact, the biggest mistake I’ve made in those 15 years is when I didn’t believe the polls, finding lots of fancy-seeming excuses to dismiss polls showing Donald Trump ahead of the 2016 Republican primary field — which they correctly predicted he’d win.

I have a particular aversion to what’s sometimes called “crosstab diving” — digging into the internals of a poll until you find something you don’t like. One reason for this is that it provides for nearly unlimited opportunities to cherry-pick evidence. A poll of 1000 voters might have a sample of something like 50 or 60 Hispanic women, for example, which comes with a margin of error of around 13 points. It’s very common to hear complaints like “THIS POLL HAS BIDEN DOWN 42-51 WITH HISPANIC WOMEN? NO WAY. IT’S GARBAGE.” These complaints nearly always reflect the crosstab-diver’s political orientation — you don’t see Democrats complaining when Joe Biden polls implausibly well with some subgroup — and almost never make any effort to account for the margins of error associated with small subgroups or the problem of multiple comparisons. If a poll lists 15 or 20 demographic breakouts, it’s a statistical near-inevitability that you can find something to make the poll look bad.

DALL*E’s impression of a dangerous activity: crosstab diving.

It’s much better if instead of looking at individual polls, you compile polls together to get a decent sample size and evaluate whether a demographic trend is consistently reproduced across different surveys. One such trend in recent polls has been the relatively poor performance of President Biden with young voters. Here, from Nate Cohn, is a summary of Biden’s numbers against Trump among young voters — which may be defined as age 18 to 29, 18 to 34 or something else depending on the poll — in as many recent national polls as he could find:

There’s a lot of variation from poll to poll — although that’s what you’d expect given the margins of error involved, plus other factors like whether third-party candidates are included. But overall these are very tepid numbers for a Democrat. Biden leads Trump 46-42 among young voters on average, with large numbers of young voters saying they’re undecided, will sit out the election, or — if given the option by the poll — will vote third-party. (In fact, when RFK Jr. is explicitly included as an option, he actually leads the youth vote in some surveys.)

This has touched off a lot of debate online about whether such numbers are feasible, with crosstab-divers citing partisan nonresponse bias or offering other explanations for why they don’t buy the numbers. Again, my experience is that denying what you see in polls usually doesn’t end well and that these critiques tend to be about 80 percent motivated reasoning and 20 percent actual science.

Instead, my attitude is more like this: I’d keep in mind that the polls are a snapshot of current public opinion. I’d take the over on Biden eventually winning young voters by more than 4 points a year from now. But there are lots of reasons to think these numbers describe the current state of the electorate fairly well and this is a real potential problem for him.

Before we get into those reasons, it’s worth putting this data in historical context. There’s a perception that such a large swing in the youth vote would be unprecedented, but that’s flatly untrue: young voters often shift quite substantially from election to election. Here is the historical data from election-night exit polls.

I’ve included the numbers from Cohn’s compilation of recent polls for reference, though it isn’t quite an apples-to-apples comparison; exit polls bracket voters aged 18-29 together, while some current polls use a higher age cutoff like 34. Even so, a 20-point swing against Biden — from Biden +24 in 2020 to Biden +4 in 2024 — wouldn’t be unprecedented. There was a 20-point swing toward Reagan in the youth vote between 1980 and 1984, a 17-point swing away from Democrats in the youth vote between 1996 and 2000, and a 25-point swing toward Democrats between 2004 and 2008. (Barack Obama was much more appealing to young voters than John Kerry.) Keeping in mind that exit polls are themselves highly imperfect instruments — people sometimes treat demographic estimates from past elections as “hard” data when it absolutely isn’t — swings of this magnitude are not uncommon.

That may be partly because groupings like this are based on voters’ age and not generations or cohorts based on birth year — and age isn’t a constant. Fully a third¹ of voters in the age 18-29 bracket in the 2020 election (everyone aged 26 or older) will have aged out of it by 2024, as will two-thirds of the age 18-to-29 voters from the 2016 election and all of them from 2012. So if you’re inclined to think something like “gee, did all those young voters who backed the Obama-Biden ticket in 2012 really turn on Biden now?”, stop doing that. Those voters are now in the 30-to-41 age bracket instead.

There are lots of obvious reasons why Biden could be struggling with the youth vote

We’re not lacking for hypotheses for why these changes could be real. In fact, there are several obvious candidates:

1. Biden is extremely old for a presidential candidate and this is a big problem for young voters.

In the recent New York Times/Siena College poll, a generic unnamed Democrat led Trump by 21 points voters aged 18-29, right in line with how under-30s voted in 2012 and 2016. Biden himself led by just 3 points, however:

The gap between “generic Democrat” and Biden tracks extremely closely with age. This cuts against the notion that Biden’s poor performance with young voters in polls reflects some sort of technical issue with how young voters are being polled. Surveys are finding plenty of young voters who are willing to vote for a Democrat — but far fewer who are excited to vote for Biden. That may simply be a consequence of Biden just having turned 81 years old and having plenty of moments where he’s shown his age.

2. Young voters sharply break with the rest of Americans on Israel-Palestine and some other issues.

There are very sharp, age-related divides in voters’ sympathies on the Middle East. In a recent Quinnipiac poll, Americans voters overall sympathize with Israelis over Palestinians by a margin of 54-24. But among voters under 35, the numbers are nearly the reverse of that, with young voters more sympathetic to Palestinians by a 52-29 margin

Note how sharp the divide is. Voters aged 35-49 are quite pro-Israel while voters aged 18-34 are quite pro-Palestine. The numbers remind me somewhat of these ones on LGBTQ identity.

Again, we see a sharply nonlinear trend. The number of LGBTQ identifiers in Generation X is still fairly low. But then it zooms up very rapidly in Millennials and in Generation Z — mostly in the form of how comfortable people are identifying as bisexual.

You can also find stark age-related divides on issues such as free speech. And you saw them in the 2016 Democratic primary. Bernie Sanders won voters aged under 30 by an 84-14 (!) margin in the 2016 Iowa caucus, even though Hillary Clinton actually (just barely) won the state.

I’m increasingly coming around to the idea that there’s been some sort of step change that cuts roughly midway through the Millennial generation — so say, people born in 1990 or later — that affects people’s social and political attitudes across a wide range of issues. Perhaps it’s related to social media usage and media consumption habits; that’s beyond the scope of today’s newsletter. But whatever the cause, it’s becoming more common to see very stark age-related divides in data on social and political attitudes and that makes me more inclined to take them at face value when I come across them.

3. The Obama generation has aged out.

I’ve already said this, but none of the voters who were in the age 18-29 bracket the last time Obama was on the ballot in 2012 will be in that group next year. Today’s young voters may not have particularly clear memories of Obama or the somewhat kitschy, nostalgic treatment of Biden during his years as Vice President.

Instead, a voter in their late 20s will now have faced three consecutive elections with candidates they weren’t very enthusiastic about. Sanders, not Hillary Clinton or Biden, was the choice of young liberals in 2016 in 2020. This year, Democrats haven’t really had a contested nomination at all. “Hold your nose and vote for this old person that you didn’t want to be nominated” is a tough message to sell three times in a row. To be clear, I think the sales pitch will work with some young voters, particularly around issues like abortion and gay and trans rights. But Democrats aren’t making it easy on themselves by consistently nominating presidential candidates with little youth appeal.

4. Young voters have less experience with inflation.

I have a longer feature on inflation coming soon, so I’ll keep this quick. But young voters grew up in an environment of very little inflation and near-zero interest rates. Now, they’ve woken up to prices that are still nearly 20 percent higher than they were three years ago:

If you’re an older American, you’ll have had some experience with this — certainly if you’re old enough to remember the late 1970s and early 1980s. And if you’re in your 40s like me, you’ll at least recall the early 1990s and the moderately high inflation leading up to the financial crisis. But if you’re 25 you won’t have any of that experience. You’ll be used to prices that barely changed at all, then suddenly they shot up a lot. In the forthcoming article, I’ll make the case that its price changes relative to expectations that matter for consumer sentiment, and those expectations will be different for younger consumers.

---

Last question: should Biden do more to cater — or if you prefer, to pander — to young voters? Actually, I think that’s not so clear. A more pro-Palestine stance probably would help Biden with young voters, for instance — but it’s unpopular with the rest of the electorate and might hurt Biden overall. More generally, voters who want to vote Democrat but don’t like Biden are coming from all over the place — some think Biden is too conservative, but others think he’s too progressive. Democrats are also increasingly having issues in polls with young Black, Hispanic and Asian men who didn’t attend college, and they won’t necessarily have the same concerns as campus progressives.

Overall, young voters have been saddled with a lot in recent years: a financial crisis, a pandemic that severely disrupted their education, and a warming planet that it will be their job to clean up. Boomers have loaded a lot of figurative and literal debt onto younger generations, in other words. And yet, the candidates most likely to be nominated for president this year are 81 and 77 years old. Trump and the current, right-wing-dominated Republican Party don’t have a lot of natural appeal to young voters, so in that sense they’re still Democrats’ voters to lose. But we should take the polling at face value when it suggests Democrats are tempting fate.

1. Actually more than a third considering that adults in their late 20s are more likely to vote than those in their late teens or early 20s.