The One Poll Showing Talarico Ahead of Paxton Comes From a Democratic Operative's Shop
Like most Texans, I first encountered Texas Public Opinion Research the way you are supposed to encounter it: secondhand and already vouched for. The Texas Tribune ran the headline that James Talarico led Ken Paxton by 5 points in the US Senate race, and the Talarico campaign promoted it within the hour. The organization behind it carried the word "nonpartisan" in every description, so the number arrived as a finding rather than as a claim. I had never heard of the outfit before that morning. What made me curious was not the politics of the result but its loneliness, because every other public survey of that race had been describing something entirely different.
Set the sponsor question aside and place the polls side by side. The New York Times and Siena found the race tied at 47% each. The University of Texas Politics Project found Paxton at 43% and Talarico at 42%, which it correctly called statistically indistinguishable. Quantus found Paxton ahead by 2, A2 Insights found Talarico ahead by 2, and Texas Pulse found him ahead by 3. Average those five and you land within half a point of a dead heat. Then TPOR arrives at Talarico +5, the friendliest number in the field, and that is the one that becomes the headline. An outlier is not a lie, and outliers are ordinary. An outlier treated as the consensus is a different problem, and it is a problem about labels rather than about arithmetic.
So I read the release, and the poll does not support the headline even on its own terms. The American Association for Public Opinion Research advises that a candidate generally must lead by 1.5 to 2 times the stated margin of sampling error before the lead is statistically meaningful. TPOR reported a margin of 3.4 points. Divide 5 by 3.4, and you get 1.47, below even the lower boundary of that rule. Undecideds stood at 14%, nearly three times the size of the gap being reported as a lead, and TPOR’s own follow-up found those undecideds leaning toward Paxton by roughly 19% to 10%. Push them mechanically, and the margin falls under 4.
TPOR called its finding a widening lead, and this is where the release passes from optimism to fabrication. In May it had Talarico at 47% and Paxton at 44%. In July it had Talarico at 45% and Paxton at 40%. Talarico did not gain a point in that stretch; he lost two. The gap grew because Paxton lost four while undecideds doubled from 7% to 14%. A lead that widens because both candidates are shedding support is not momentum in any sense a reader would recognize from the word. It is an electorate becoming less settled, which is the opposite of what the coverage conveyed.
The crosstabs are stranger still. Between the May and July waves, TPOR’s statewide margin moved 2 points. In those same seven weeks, its Latino margin moved 50 points, from Paxton +4 to Talarico +46, its noncollege margin moved 30 points toward Talarico, its college margin moved 29 points toward Paxton, and its white margin moved 19 points toward Paxton. Those are not findings; they are cancellations. Real electorates do not lurch 50 points among Hispanics while barely moving overall. Samples do that when subgroups are small, or weights have changed, which is why professional standards require pollsters to publish unweighted subgroup sizes. TPOR did not publish them, nor did it disclose its recruitment source, mode, languages, likely-voter model, weighting targets, or design effect.
What it did publish was an invitation. The release states that crosstabs are available upon request, which is the industry’s way of signaling that nothing is being hidden. I know of two media outlets that took TPOR up on it, and neither received a response, not from TPOR and not from Slingshot Strategies, the firm that fields the surveys and whose address appears as the press contact. Curious, indeed. An organization that wants its number treated as a public good has an odd way of behaving when the public asks to see the work. The offer of transparency turns out to be part of the product rather than a description of it.
Which brings me to the part of the inquiry that changed the piece I meant to write. TPOR’s materials say it operates “without allegiance to any political position or special interest,” and its director said at launch that he hoped it would become a “go-to neutral source of reliable information.” Those are not moods; they are factual assertions about institutional structure. “No allegiance to any special interest” describes relationships that either exist or do not, in the way that “grass fed” is checkable while “the best beef in Texas” is not.
Here is what the relationships look like. TPOR's director is Luke Warford, former Chief Strategy Officer of the Texas Democratic Party and the party's 2022 statewide nominee for Railroad Commissioner. That alone would prove nothing, because a résumé is not a conflict. Warford is also, right now, a founding partner of the Agave Democratic Infrastructure Fund, an organization that says its purpose is to help Democrats win in 2026 and that funds research and messaging to determine what Democrats in Texas should be saying to voters. Agave has paid almost $200,000 to Slingshot Strategies. Slingshot advertises that it works with candidates, progressive organizations, and issue advocates, and that it powers separately branded public polling initiatives. TPOR is a separately branded public polling initiative.
State the principle plainly, because a principle nobody can state is a principle nobody can apply. My objection is not that Warford has a Democratic past. If it were, I would have written an indictment of Marist, of Quinnipiac, and of half the university survey centers in the country, and any competent defender would say so in his first paragraph. My objection is concurrency. Warford has a Democratic present, running simultaneously and performing the same functions. One organization exists to elect Democrats. The other claims allegiance to no political position. Both do polling, messaging, and research through the same consultancy, under the same director, in the same week, about the same electorate. The standard is not purity of biography; it is simultaneity of control. A former strategist can run a neutral shop. A sitting strategist running an active electoral fund cannot run one at the same time and call it neutral, any more than a judge holding stock in the defendant becomes impartial by promising to be fair. One last thing to note is that Warford launched Agave in January 2024 and TPOR just six months later, not because Agave wasn’t successful but likely because it needed the services of a ‘nonpartisan’ polling firm it controlled.
Notice why the label is worth paying for rather than merely nice to have. “Nonpartisan” performs one function: to move a survey out of the folder marked “Democratic research,” where a reporter treats it as an interested party’s claim, and into the folder marked “public opinion,” where he treats it as a finding. That reclassification is the entire value proposition, and in July it worked perfectly. The outlier became the headline, and from there the number traveled into the fundraising materials of a pro-Talarico super PAC pitching donors. The loop closes neatly. Democratic-adjacent money produces neutral-labeled research, the label secures the coverage, the coverage becomes evidence in a fundraising appeal, and the appeal produces more money. The label is not a byproduct of the operation; it is the mechanism.
Let me concede the strongest objection completely, since the concession costs nothing. Even if we assume that every TPOR number is accurate as arithmetic. The misrepresentation remains what it was, because the challenged claim is not “these results are wrong,” it is “this organization has no partisan allegiance.” Still, the record is worth knowing, because a neutral label asks readers to extend confidence and it is fair to ask what they got for it. TPOR had Trump up 8 in Texas in 2024, and he won by roughly 13.7. It had Jasmine Crockett leading Talarico by 1 in the January primary, and Talarico won by roughly 6.2, meaning the poll named the wrong leader. It published Paxton over Cornyn by 8 in March and again in April, and Paxton won the runoff by approximately 27.7 points. Across the completed contests, the average distance between the published margin and the eventual result runs to roughly 9 points. As a pollster, Warford leaves a lot to be desired.
The legal question is narrower than my critics will pretend. Texas does not regulate bad polls, and it should not. It regulates false representations about the characteristics, source, and sponsorship of goods and services, which is ordinary consumer protection law and applies to the sale of expertise as readily as to the sale of a used car. The Attorney General’s office has twice advanced this exact theory, that ideologically committed work may not be marketed as objective civic expertise, in its actions involving the Center for Tech and Civic Life and Institutional Shareholder Services. In both, the alignment had to be inferred from conduct. Here nothing needs inferring, because the director publishes it himself on the website of the other organization he runs. In my opinion, TPOR is ripe for a Texas Deceptive Trade Practices Act investigation.
Two quick answers before I close. Someone will call the neutrality language puffery that no reasonable person credits, but puffery is opinion about quality, and this is a verifiable statement about relationships, which is the category the law treats as actionable when false. Someone else will say this chills speech, and that takes one sentence: nobody is being told what to publish, only what to disclose about who paid.
Skin in the Game: How Real Money Exposes the Political Polling Cartel's Methodological Corruption
Let me begin with a small confession that ought to matter to any reader weighing what follows. I have never placed a bet on a prediction market. As an American citizen, I am not legally permitted to wager on Polymarket, and I have not done so. What I have done, every day for years now, is cite prediction markets in my writing because they are the single…
So here is the offer that should settle it. TPOR may operate as a Democratic polling shop, which is lawful and arguably overdue, since Texas Democrats plainly need better information than they have been getting. It may not be that and simultaneously sell Texans the opposite. Drop the label or drop the fund. Either choice is honorable, and the refusal to choose is the only thing here that is not.
If you enjoy my work, please subscribe https://x.com/amuse/creator-subscriptions/subscribe
Alexander Muse is a Fellow at the John Milton Freedom Foundation and publishes daily political analysis at amuseonx.com. Primary sources cited in this piece are linked inline; campaign finance figures are drawn from FEC filings, polling data from publicly released crosstabs, and legal claims from filed pleadings. Corrections are posted to the original URL with a dated changelog. Readers who identify errors are invited to contact the author directly. Each op-ed edited for grammar and clarity using Ai in a sponsored partnership with Grammerly. Data provided in a sponsored partnership with Polymarket.
Democrats have no concept of partisanship like fish have no concept of water.
This poll, like all the polls between a Democrat and a Republican, is not the least bit believable when it shows the race this close. THIS is how polls work from the left: deception, lies, and exaggerations so that it APPEARS to be a close race. I live in Texas, and I absolutely scoff at this total nonsense.