[Watch Whole Hog Politics live: Join us today at 9 a.m. ET at TheHill.com as Chris Stirewalt and host Bill Sammon break down this week’s political news and answer questions from a live online audience.]
We know that President Trump doesn’t like pollsters, or at least not the ones that produce unfavorable results about him. But it is also true that polls don’t really like him either.
Right now, that means that each week, Trump is plumbing new depths of unpopularity as voters recoil from his personality and policies. Even low-quality pollsters with high degrees of pro-Trump lean in their past surveys show the president underwater with voters. When not even Rasmussen Reports can get him over 45 percent approval, you know Trump is in the soup.
But it is also true that in each of his three presidential campaigns, the best national pollsters have significantly understated his share of the vote. In 2016, Trump won when everyone — including him — thought he would lose. In 2020, what was forecast to be a 10-point blowout for Joe Biden ended up being less than 5 points. And in 2024, the campaign closed with Trump trailing by less than 2 points, but he won the national popular vote by a point and a half.
The debate has raged over why, at a time when pollsters have had a good record in midterms and have been pretty much on the nose with the Democrats’ share of the presidential vote, Trump keeps beating expectations.
Republicans like the idea of the “shy” Trump voter, which is akin to another piece of favored right-wing lore: the Bradley Effect. This is the idea that voters who felt pressure to vote against the socially undesirable Trump, lied to pollsters rather than admit their preference. This is the “silent majority” in action.
Maybe there are instances of this taking place. Certainly the pro-Kamala Harris spot telling women to lie to their Republican husbands about voting for the then-vice president suggests this kind of thinking has a home in both parties. But a better explanation is that voters who favored Trump were just harder for pollsters to find. The largest bloc of Trump’s support in all of his elections came from white, working-class voters. We have lots of evidence to suggest that this is also one of the groups most unlikely to participate in surveys, some probably by choice or because of suspicion and others because they’re just too busy. By the time you’ve found enough of these voters to make up the 40 percent or so you need for a polling sample, you’ve scraped down to get an unrepresentative sample.
If only rabbits with bent ears like purple carrots, you won’t catch many straight-eared bunnies in a purple-carrot trap.
One of the best pieces of evidence for the purple carrot theory of the missing Trump voter came from before Trump was ever on a ballot. The biggest polling miss in the modern era until 2020 was in 2012, when an average of high-quality polls showed incumbent Barack Obama ahead of Republican nominee Mitt Romney by 0.6 points at the end of the campaign. Obama went on to win the national popular vote by about 4 points. I doubt many of those “missing” voters were lying to pollsters to say they were voting for the white businessman instead of America’s first African American president. More likely was that Obama did the best among the voters who are the hardest to poll: working-class Americans.
The 2012 miss didn’t get nearly the attention as the Trump-era errors, which is understandable since it was a matter of degree, not outcome. The expected result — an Obama victory — occurred, so whether it was narrow or comfortable was just for style points. But the error still sent a jolt through the world of public opinion research.
One of those shook loose was the Gallup organization, which had been one of the only pollsters in the closing days of the 2012 campaign to show a lead — albeit a narrow 1-point advantage — for Romney. Gallup wasn’t any kind of a wild outlier. The difference between Gallup and contemporaneous polls was statistically insignificant. But it was obviously enough of an embarrassment that Gallup announced in 2015 that it would, for the first time since the 1940s, not release polls on the upcoming presidential election.
That decision made good business sense because while an election poll can be wrong, a poll for a corporate client really can’t be. You can draw the wrong conclusion from a small sample of respondents about how the electorate is going to vote, but not about how people feel about the new packaging on Post Toasties or what attributes are most important to consumers in purchasing a new vacuum cleaner. Except for in very extreme cases, market research is unfalsifiable. Measuring subjective preferences and attitudes is very different from trying to forecast an actual election that will be contested, tabulated and reported.
A horse race poll can be accurate without being predictive, which is to say that races change in the closing days and margins of error sometimes overlap. But getting it “wrong” makes the people who sell breakfast cereal and vacuum cleaners wonder if your high-priced consulting services are all they’re cracked up to be. So Gallup, the firm that had essentially created the polling industry as we know it when George Gallup started doing public polls in the mid-1930s, got out of the horse race.
That was no great loss, since we have lots of good polling on presidential elections, much of it superior to Gallup’s offerings. But what Gallup kept was the jewel in its crown: its more than eight-decade long effort to track public opinion on presidential job performance … until this week.
The company announced “an evolution in how Gallup focuses its public research and thought leadership” that would no longer include assessing presidential job approval, as it had for presidents going back to Franklin Roosevelt. It was not lost on anyone that this announcement followed Gallup giving Trump a scorchingly bad score for the final quarter of 2025. Some Democrats cried foul and suggested Trump had pressured the company to stop.
That’s not a crazy suggestion, given how heavy-handed Trump is with corporate entities that displease him. With lots of money and free legal services from friendly firms, Trump has a long record of harassing lawsuits like the one he filed against Iowa pollster Ann Selzer and the Des Moines Register. Targets often prefer to make a deal rather than endure the cost and trouble of endless litigation.
But it is also possible that Gallup just wanted to get out of the business of doing politically problematic polls for public consumption. Like the horse-race decision of a decade ago, the presidential approval decision may have just been that juice wasn’t worth the squeeze anymore.
Whatever the reason, it’s too bad. There are other pollsters that do better, more detailed public opinion research, but one thing you can’t buy is history. Gallup’s public service was that it provided stable, comparable data for 15 presidencies so we could reliably compare American voters’ attitudes at different times.
That helped us put things in perspective and gave us a better sense of ourselves. It’s a shame that didn’t fit Gallup’s idea of “thought leadership” anymore.
[Programming alert: Watch "The Hill Sunday with Chris Stirewalt" — From your local city hall to the halls of Congress, we’ll cover the big political stories of the day — and the ones that might have been missed. And, as always, we’ll have expert analysis from our best-in-the-business panel of journalists. Be sure to catch us on NewsNation at 10 a.m. ET / 9 a.m. CT or your local CW station.]
