A Polling Firm Faked Its Numbers. The Press Reported Them as Fact.

Median Strategies called itself an independent polling and research firm. On August 17, it admitted the polls were invented.

The firm withdrew every survey it had released and said the operation was a “short-term social experiment examining how purported polling information could enter and spread through the political information ecosystem without independent verification.” It declined interviews, named no one involved, and disclosed no funder. Then it shut down.

The fabricated data was not obscure. Median Strategies had shown Los Angeles Mayor Karen Bass leading Councilwoman Nithya Raman by nearly twelve points, citing 560 voters surveyed and a 4.1 percent margin of error for a survey that never happened. Bass posted the number, then deleted it. In Wisconsin, the firm put Democratic candidate Francesca Hong more than twenty points ahead in the primary for governor. The Los Angeles Times reported the fabrication on August 17. By then, newsrooms across the country had already carried the fake results as legitimate news.

The elections were real. Hong lost. David Crowley won the Wisconsin Democratic primary on August 11 by less than a single point, after dropping out in early July and re-entering ten days later with Governor Tony Evers behind him. For weeks, the public record around these races included numbers a firm has now confirmed it made up.

The lesson is not that one firm behaved badly. It is that nothing stopped it. A number can be conjured from nothing, fitted with a fake margin of error, and moved into wide coverage before a single editor asks to see the crosstabs. The apparatus that tells Americans what they think has no verification layer worth the name. It ran the invented numbers the same way it runs the real ones, because from the outside the two look identical.

The fake firm is only the clearest case. The polling industry that signs its name misses too, and it misses by margins that decide elections. This month, surveys showed Abdul El-Sayed comfortably ahead in Michigan’s Democratic Senate primary before he won it by about a point and a half, and showed Francesca Hong leading in Wisconsin before she lost. The American Association for Public Opinion Research put the average error in 2024 presidential polling at 3.3 points, and the midterm cycles before it ran worse. A fabricated poll and a poll that is merely wrong do the same work. Neither one measures what a population believes. Both move it, and neither is corrected until the ballots are counted.

Texans live downstream of that apparatus on the one question that matters most here. They are told, at regular intervals, that they do not want independence. The figure cited is almost always low, and it almost always comes from the same polling industry now explaining why its own product cannot be taken at face value. The Texas Nationalist Movement has argued for years that a poll like that is not a readout of what Texans want. It is a tool for managing what they think they want, a number repeated until the people who might act decide they are alone.

When a pollster with a real record measured the question, the result did not match the narrative. SurveyUSA, graded A-plus by FiveThirtyEight for accuracy across roughly nine in ten of the elections it has called, found in 2022 that 60 percent of Texans would support Texas “peacefully becoming an independent country along with other conservative states.” That poll is from 2022, and it remains the highest-quality public read on the subject. The lower numbers Texans hear repeated come from instruments no one is auditing.

The Texas Nationalist Movement has argued for two decades that the future of Texas is not a matter of opinion research. It is a matter of consent, and consent is measured at a ballot box, not in a phone bank. The movement’s central demand is a binding statewide vote on independence. A referendum is the one instrument that cannot be fabricated. It has a paper trail and a result no anonymous firm can withdraw on a Monday afternoon.

That is the difference between a poll and a vote. A poll can be invented by people who decline to give their names. A vote is cast by Texans who give theirs.

Median Strategies showed how little it takes to invent what a population is told about itself. The pollsters who merely get it wrong reach the same place by accident. Texas will keep being told what it wants until the day it decides to answer the question in the only way that cannot be faked or managed.