Pew on non-voters | Weekly roundup for June 29, 2025

Last week in political data: Pew says non-voters leaned Trump in 2024; Polls on bombing Iran; Everyone missed the NYC mayor primary.

Dear readers,

According to the Strength In Numbers new reader survey, the most requested new feature of this Substack (at 69% ) is a weekly roundup of new political data and analysis published by other sources around the web. I am but your humble servant, so this Sunday, I’m kicking off a weekly roundup post with all the notable data published by the press, pollsters, and other analysts in the last 7 days.¹

Here’s the way this is going to work: Every week, I’m going to pick one big headline from the new polls and political data published in the prior week to write a short commentary on. Then, the rest of this article will be an annotated list of recommended reading elsewhere (what people used to call a “listicle”).

I figure that since I’m reading all this stuff anyway for my job, I might as well provide some additional value for readers by compiling material for you so you don’t have to sort through all the noise like I do.

With that introductory note out of the way, here’s what’s new this week!

---

Pew says non-voters leaned Trump in 2024

The Pew Research Center published its analysis of the 2024 election this week. Pew’s analysis is special because the organization combines a massive set of interviews from election polling with post-election “voter file” data from election officials that records who actually turned out to vote in November. This lets Pew separate out changes in election results since 2020 that are due to changing minds, versus changing turnout. They also benefit from appending their survey data to the voter file in that they can look at the percentage of the electorate that is white or Black, young or old, Democrat or Republican, high-income or low-income, etc — variables that states don’t gather on voters.

The hard work behind all this data-wrangling lets Pew find out a lot of things you can’t know with polls alone. For example, Pew says that 15% of people who voted for Joe Biden in 2020 didn’t vote in 2024, and that Donald Trump won 2020 non-voters in 2024 14% to 12% (the vast majority of 2020 non-voters didn’t vote last year). In 2024, Trump won 85% of his 2020 voters.

You may infer from this that the people who didn’t vote in 2024 were Democratic-leaning, since more 2020 Democrats sat out than 2020 Republicans. But Pew finds that many of those people must have changed their preferences since 2020: According to their polling, while Trump won by 1.5 percentage points among voters in 2024, among Americans who did not vote in 2024, he led Harris in pre-election surveys by 4 points, 44 to 40. That’s an advantage that reverses Democrats’ longtime edge with non‑voters. This is a trend that has been moving against Democrats since 2012.

Pew’s finding makes the top spot in this newsletter for three reasons:

The first is that this is the highest-quality and most comprehensive public analysis that assesses vote preference among validated non-voters in 2024. Several private firms have released their own estimates, but don’t publicly explain how they arrived at them. Others, like the polling team at the New York Times, have done some estimation but not released a full report.

The second reason this is notable is for what it says about turnout. After the 2024 election, lots of Democrats claimed that if turnout had been higher, Harris would have won. But Pew shows that higher turnout might have widened Trump’s margin.

On this subject, the implicit argument is that if only Democrats had turned out their voters who stayed home, they would have won. But that’s unfalsifiable. Figuring out how to do it is the billion-dollar question. My theory is that what convinces people to move from Republican to Democrat is also likely a strong force on getting non-voters to turn out at all. A popular campaign is a persuasive campaign. In 2024, Trump managed to convince both voters and non-voters to move to the right, and to convince more of his prior non-voters to turn out. Those trends probably didn’t happen in isolation.

In a hypothetical world where Democrats who stayed home in 2024 turn out, there are probably other knock-on effects for persuasion. That world looks very different from ours. So if Democrats can find a message for 2028 that persuades non-voters their way, that’s probably also increasing their vote share with independents and moderates. Turnout and vote share, moving in the same direction. If they find the right message.

Finally, I think Pew’s report is valuable because the findings show just how volatile the electorate is. It’s not just that Trump made big gains with certain demographics, but that roughly one out of every four voters behaved differently in 2024 than they did in 2020. In an age of partisanship and calcification, this is a good reminder that things do matter to people and behavior isn’t predetermined. Campaigns matter!

---

What you missed at Strength In Numbers

SIN published three articles last week:

• Part 1 of the June Q&A, on Tuesday. I answer questions like: Are right-wing pollsters biasing Trump approval averages? Should America switch to ranked-choice voting? And what's up with the Democratic Party brand?

• Part 2 of the Q&A, on Wednesday: Who supports bombing Iran now? How much do people know about politics? How should analysts update their priors?

• And on Friday, a Chart of the Week on Iran: After Trump's Iran attack, MAGA voters are not isolationist anymore.
  

---

Even more numbers!

Democratic primary for NYC mayor

• Pollsters broadly missed Zohran Mamdani’s 8-point win in the Democratic primary for mayor of New York City on Tuesday, missing his margin by an average of 20 points. A discussion of that error is here. The pollster that did best in the contest, Public Policy Polling, explains why they think they were ahead of the curve.

• It wasn’t just the pollsters that did poorly in NYC. Prediction markets dramatically missed Mamdani’s win, giving him just a 6% chance of win a few days before the contest. this appears to be because they were underestimating error of ranked-choice primary polls. Funnily enough, this is despite the many historical arguments from the people who like these markets that they are supposedly accurately pricing in information about the accuracy of polls and non-polling information. Hopefully, people will take the hint soon and stop thinking of these things as oracular.

  • In the future, you can use this heuristic: Cuomo was up 12 in the average of polls a week before Election Day. With an average mayoral primary polling error of 10 points, that would have put his odds of winning at 90%. Assuming more error, 80%.

  • The late endorsement by Brad Lander also rendered older polling borderline useless. Why didn’t markets price that in?

• Here are some maps and scatter plots on the NYC mayor race from fellow Substacker and election forecaster Jesse Richardson. He suggests that Mamdani increased youth participation and succeeded in getting middle-class voters to show up for him. Other analyses have suggested the same thing.

• Plus, even more maps (these from Michael Lange) on how Mamdani won.

  Here’s more proof of the Mamdani effect on turnout.

U.S. bombing of Iran

• YouGov finds a huge increase in GOP support for the war. The cool thing about this is that they were interviewing voters for this week’s poll while the U.S. military strike took place, so they could split their sample in two (pre- vs post-attack) and look at opinions.

• Charlie Cook explains why you shouldn’t expect a large increase in support for the bombing among independents and Democrats (this is not like 2001).

Trump approval

• Echelon Insights finds broad disapproval of Trump on the issues, dragging down his net approval

• Trump’s pollster warns Republicans about growing backlash to Medicaid cuts in the GOP budget bill

Other data points

• The NYT’s Nate Cohn comments on the Pew data.

• Sabato’s Crystal Ball compares the Pew report to other 2024 post-mortems. The conclusions are directionally similar but differ largely in magnitude.

• Tom Wood finds that Americans have become less protectionist since 2024, as Trump has sparked an increase in free trade attitudes from his tariff plan:

  

• My former colleague at FiveThirtyEight Galen Druke explains how he thinks Democrats can win in 2028: By allowing more ideologies into the tent. My spin on this: Democrats can do this by attracting low-engagement voters with malleable ideologies by running an outsider, anti-establishment, pro-working-class candidate. This is what most Democrats have been calling for since 2020, and it’s arguably what helped Barack Obama win in 2008 and 2012.

---

Update from the data portal:

• Trump’s approval rating is down:

  

• I’ve added more issues to our issue approval tracker:

  

• And aggregate economic statistics have started showing the impact of Trump’s policies:

  

The charts on the data portal update every day. Feel free to use and cite them liberally!

---

And that’s it for this week!

If you have any feedback on the format or content of the new weekly roundup post, email me at mail@gelliottmorris.com. My job is to write independent political analysis and provide value to readers, so any thoughts on how to do those things better are very valuable to me. I read every email I receive.

Elliott

1. Yes, I really do read the new survey! For example, 80% of people said they wanted monthly Q&A posts, which I started this month. And 45% of people say they want a podcast, which I’m not inclined to do (I don’t have enough time if only half of readers want it), but I'm open to hearing ideas on this.