Some pollsters think their results are exaggerating Democrats’ support, and angry ex-Trumpers are the cause

Welcome to Wednesday’s edition of Washington Secrets. If a friend sent you this and you’d like to sign up, CLICK HERE. Today we go full-on polling nerd to look at one issue that could be overestimating support for Democrats. But don’t get carried away — Republicans are still likely to take a beating. Plus, the bookmakers are coming for Katie Zacharia after her first White House briefing.

In 2024, a rash of cheap online polls skewed forecasts, underestimating Donald Trump’s eventual margin of victory.

This time around, a handful of pollsters have spotted another possible error in a type of survey that finds participants by sending out SMS messages, called text-to-web polling.

They say it is returning a disproportionate number of disenchanted Trump voters who have switched sides.

And it could mean that forecasts predicting a blue tidal wave, such as Nate Silver’s much-quoted model, with Democrats’ chances of winning the Senate at over 73%, are overestimating their chances.

Most pollsters use a range of methods to reach respondents, and then balance things out so that their final sample reflects the voting population for age, race, educational attainment, geography, and a host of other factors.

That isn’t working said Landon Wall, founder of Republican polling firm GrayHouse. He said his live interviews by cellphone were giving different results from the text-to-web portion.

“There is a mode split happening,” Well told Secrets. “And even when we control for demographics, so that the sample from text looks demographically and party registration-wise exactly like the live-to-cell portion, that live-to-cell portion is significantly more right leaning.”

He has rejigged his methods to increase the number of live calls in his samples and reduce the text portion, even though it means bigger bills for his clients.

The phenomenon was first spotted by James Johnson, co-founder of J.L. Partners, another Republican leaning firm. He ran an experiment to quantify what he was seeing with a poll in Iowa.

Online polling showed Democrats ahead by a single point. Live call polling also had Democrats ahead by a single point. Yet the SMS-to-web portion showed Democrats up by 12 points.

“My problem is not that SMS-to-web is getting too many Democrats; it’s that it’s getting too many angry Trump voters, and what I mean by angry Trump voters is people that voted Trump in 2024 who are now saying they will vote Democrat,” Johnson said. “That is harder to control for.”

“There’s no benchmark for that,” he added. “There’s no measurement in the population for how many angry ex-Trump voters there are expected to be.” 

In some ways, pollsters face the same problem they have always faced: How do they take their samples and make sure they are weighted to reflect the overall population?

If you know that a certain percentage of Democrats tend to turn out in the midterm elections, or the proportion of black men aged 30 to 40, or the number of people with college educations, then you can balance your findings to give a final result that reflects the expected voting population.

But what do you do when it involves a new species of voter?

Trump’s extraordinary rise in 2016 and his broad 2024 coalition, combined with his collapsing popularity in 2026 — his approval rating is down to around 33% in poll averages — means pollsters have to account for a chunk of ex-Trumpers who are motivated and angry.

Do they behave like other Democratic voters? Does their disgruntlement make them more likely to turn out? Or less?

Not everyone is as worried. Several other pollsters said they were confident that proper weighting could help control for such sampling discrepancies.

“I found that I do get more MAGA-leaning voters and Trump approvers via live call, but it’s not as neat and clean as they are making it out to be,” said Adam Carlson, founder of Democratic-aligned polling firm Zenith Research. “I think it really depends on how you’re sampling and who you’re reaching.”

Carlson put it down to previously known effects, such as SMS-to-web surveys reaching higher earners and more-educated voters, who skew Democratic.

READ MORE: EXPERTS AREN’T SOLD ON DEMOCRATS’ POLLING ADVANTAGE

President Donald Trump arrives to speak at a campaign rally at the Curtis D. Menard Memorial Sports Center, Tuesday, Oct. 13, 2026, in Wasilla, Alaska. (AP Photo/Alex Brandon)

But he also agreed that he was concerned about whether pollsters were picking up enough Republican voters. One problem was that they were reaching what he called super voters, who turn out in every election.

“What I’m seeing in my data, if you don’t control for that, the people who are more likely to take polls in general are more likely to be the kind of the super voters who turn out every midterm … maybe off-year elections too, and they’re more likely to be voting for Democrats,” Carlson said. They’re more engaged and more fired up.”

That could signal a turnout advantage for Democrats in November, he added. Or it could mean less frequent voters, who will still turn out in November, are being missed if the correct weighting is not applied.

To be clear, no one is saying Republicans are in danger of winning in November. If the polls are off, it will likely be by a point here and there, but that’s enough to distort some of the forecasts about Senate control.

“I think the Nate Silver 75% chance of Democrats winning the Senate is based on a lot of the polls that have the problem that we’ve just been discussing,” said Johnson, “and that puts me closer to 55% than it does 75%.”

You are reading Washington Secrets, a guide to power and politics in D.C. and beyond. It is written by Rob Crilly, who you can reach at [email protected] with your comments, story tips, and suggestions.

How long will Katie Zacharia last?

That didn’t take long. The new press secretary is already under fire for the brevity of her first briefing, and left-wing media has questions about the nature of her legal qualifications.

And now bookmakers are offering odds that Katie Zacharia is out of the job before the end of the year.

William Hill has her at 2/1 to step down before 2027.

“That price implies a one-in-three chance of her stepping down before the turn of the year, following what’s being described as a ‘brutal’ debut as Donald Trump’s press secretary,” said spokesman Lee Phelps.

“Zacharia’s odds for her to leave her current role could well have been shorter if it wasn’t for the U.S. president’s public backing and defense against the criticism reported after her first appearance in front of the cameras,” Phelps added.

Zacharia is traveling with the president at the moment. And we know that he won’t want to lose her so soon after he picked her. But her next briefing will be the subject of intense scrutiny.

In truth, her performance was fine. Nervous? Sure. Brief? Definitely.

She was only announced as press secretary on Friday and to be thrown into the briefing room on Tuesday meant a baptism of fire. 

To be sure, there is a huge difference between being a talking head on conservative television and dealing up close with dozens of journalists every day. Zacharia needs to bring in some of her own people and take time to properly prepare for the next time she takes the podium.

There were moments when her comments landed, particularly when talking about her family. But she needs someone on her team who can pummel her with questions and prepare her better for the bearpit.

Lunchtime reading

The Best Kind of American by Kim Ghattas: “Why, in 1984, did Iran have its Lebanese proxies kill Malcolm Kerr, the president of the American University of Beirut, when he arrived at its campus on a sun-kissed promontory that symbolized the city’s debonair charm?”

Cecily “Chips” Gemmell, Churchill’s last surviving secretary: She worked under Winston Churchill for four years from 1947 and described him as a “loveable Teddy bear.” A demanding boss, she still said there was plenty of laughter.

Share this article:

Facebook

Twitter