Wedgewood Polls is — or was — a political polling firm whose survey results were picked up by the New York Times and folded into Nate Silver’s widely cited election averages. On Monday night, the firm posted a statement to X declaring that the polling industry needs major reform.
It needed reform, as it turned out, because Wedgewood had been fabricating its polls.
Not weighting aggressively. Not using a sloppy likely-voter screen. Making the numbers up, then releasing them as though real people had answered real questions.
Here is what the firm published for the 2026 Senate races, all of it fiction: Cindy Hyde-Smith up four in Mississippi. The Democrat up one in Kansas. James Talarico up three in Texas over Ken Paxton. Abdul El-Sayed up three in Michigan over Mike Rogers.
Look hard at those last two. Texas and Michigan are the races most likely to determine control of the United States Senate. A firm inventing results from whole cloth put out three-point Democratic leads in both, and the most influential averages in American politics absorbed them, blended them with legitimate surveys, and published the result as the state of the race.
The polling industry needs major reform. Our statement below: https://t.co/Sj8JLYgply
— Wedgewood Polls (@WedgewoodPolls) October 20, 2026
This is the second confession of the cycle. During Wisconsin’s Democratic gubernatorial primary, a polling outfit showed far-left candidate Francesca Hong ahead by nearly twenty points. The firm later acknowledged the poll had been cooked, described it as a social experiment, and shut down its polling operation.
Two firms. Two admissions. Both errors running the same direction.
Now stack that against what honest people in the field have been saying. Frank Luntz told CNN over the weekend that he got 2016 wrong because he trusted exit polls Trump voters had refused to participate in, that he’s skeptical of this year’s numbers, and that he doubts anyone is measuring turnout correctly. The Times’ own Jeremy Peters has reported that Michigan polling is oversampling El-Sayed’s credentialed base while missing lifelong Jewish Democrats in the Detroit suburbs who are breaking Republican.
Three problems compounding. Republicans don’t answer. Samples tilt toward those who do. And some of the surveys were never conducted.
The aggregators own a piece of this. The whole premise of averaging is that independent errors cancel out. Fabrications don’t cancel. They sit in the mean, dragging it, wearing the credibility of the real polls surrounding them. And nobody detected it — Wedgewood confessed, which is the only reason we know.
Thirteen days out, the obvious question has no answer: how many others haven’t confessed?
For the record, there are state and federal laws that prohibit intentionally fraudulent polling. That’s not protected by the First Amendment; it’s called “election interference” and it’s illegal. The Trump Justice Department needs to make examples of these firms and send some folks to jail over it. Enough, already.
