In 2024, Republicans outperformed their September polling in Iowa by nine points. In Ohio, six points. In Michigan, five points. CNN's senior data reporter Harry Enten looked at those numbers on CNN News Central and arrived at a conclusion his own network probably didn't want to hear.

"Fool me once, shame on you. Fool me four times, I just have to ask the question, could it happen again?"

Enten walked through the receipts. In the 2002 midterms, the GOP was underestimated by three points in major statewide races. In 2014, three points. In 2016, when Donald Trump shocked the world, three points again. In 2020, the underestimation ballooned to five points. And in 2024, back to three points. That's four of the last six election cycles where polls systematically shortchanged Republican performance.

"We've had really good Democratic polls coming out from Ohio, Michigan, Iowa," Enten told viewers. Then he dropped the context: "You look at this and you go, you know what, maybe we should just hold on a second here because we've seen this dance before."

2018 showed the same thing. Republicans outperformed September polls in Iowa by six points and in Michigan by eight points. The states where Democrats are banking on a midterm surge are the exact states where polling has been most wrong about Republican strength.

Enten then applied the math to November 2026. "If history holds in terms of the GOP outperforming their September polls, well, there's a very, very strong chance that Republicans hold on the United States Senate." He noted that even a two-to-four point polling error in the GOP's direction could produce a 53-47 Republican Senate.

This is CNN's own analyst telling CNN's own audience that the data CNN keeps showing them has been wrong in the same direction, in the same states, for over two decades. Pollster Frank Luntz has described the current voter sentiment landscape as "a messier picture" — which is the polite version of admitting the models still can't find Trump voters.

Watch Enten's entire warning to Democrats for yourself

The polling industry had six years after 2016 to fix the problem. They tried new likely-voter screens. They adjusted for education. They weighted for social trust. Trump outperformed again in 2024 anyway. Whatever the polls are missing, they're still missing it.

Every cycle, the same states show the same gap. Iowa. Ohio. Michigan. The numbers move in the same direction every time. Enten didn't offer a theory for why. He just laid out the data and told his audience to hold on.

Four of the last six elections. Same direction. Same states.