Frank Luntz sat on CNN this weekend and told its lefty audience something they did not want to hear. Asked whether the polls are underestimating Trump voters, he said they don’t answer the phone — and then he did something pollsters almost never do.
He admitted he blew it. He got 2016 wrong, he said, because he believed the exit polls. Trump voters walked away from the exit interviewers, refused to participate, because they’d concluded the whole apparatus was arrayed against them. He invoked the cadets he teaches at West Point, who can’t lie, cheat, or steal, and said he holds himself to the same standard. He’s skeptical of the current numbers. And what he’s really watching is turnout, which he isn’t convinced anyone is measuring properly.
That’s a man with no reason to help Republicans telling a network that has spent a year promoting a blue wave that its central premise may be unmeasurable.
Now do the arithmetic, because a three-point error is not a rounding error — it’s the entire election.
Take the current RealClearPolling Senate averages at face value and Democrats finish at 53-47. Shift them three points toward Republicans and the chamber ends 54-46 the other way. Same polls. Same races. One adjustment inside the historical margin of miss, and the outcome inverts completely.
The House is murkier. The averages currently produce a Democratic majority around 235-226. Apply the same three-point correction and it becomes a coin flip, with either side scraping together a narrow majority.
?Frank Luntz *DUMPS COLD WATER* on CNN viewers’ dreams of blue wave?
Luntz warned Trump voters "don’t respond" to pollster phone calls.
He admitted he "got 2016 wrong" because he trusted the exit polls.
Luntz asserted he is "skeptical" the current polling is accurate.
Moreover, he cast doubt on whether pollsters are using the right measurements and asking the proper questions to determine turnout.
Luntz has been a renowned pollster for decades, so this is meaningful coming from him.
He knows firsthand how polling can miss Trump voters.
Since there are many close races, underestimating Republican turnout by a small margin could make all the difference.
It sounds like there’s a real chance of that happening.
— Jason Cohen ?? (@JasonJournoDC) October 18, 2026
And the three-point figure isn’t pulled from the air. Harry Enten — also on CNN, also no friend of this administration — walked through the Midwest numbers earlier this month. Republicans beat their September polling by nine in Iowa, six in Ohio, and five in Michigan in 2024. In 2018, Trump’s first midterm with Trump not on the ballot, they beat it by six, six, and eight. Across 2016, 2018, 2020, and 2024, the GOP outran its September numbers by at least five points in all three states every single time.
Luntz explains why: the people being missed stopped participating a decade ago and never came back.
But here is where the comfort stops and the work begins, because Luntz’s actual warning cuts the other direction too. The reason these voters are hard to sample is that they’ve been told for years not to believe any of it. That skepticism is earned. It is also dangerous, because a voter who dismisses the polls as rigged is one short step from concluding his own ballot doesn’t matter.
It does. The polling error only exists if the people being undercounted actually show up. Republicans will win the midterms handily – IF GOP voters cast ballots. Do so early.
