Will Democrat James Talarico defeat Ken Paxton in the 2026 Texas U.S. Senate race?

This race is still a test of whether party loyalty beats candidate doubts, and the newest evidence says loyalty is winning back ground. In the September Texas Southern University poll, only 84 percent of Republicans backed Paxton. In a different poll, the UT/Texas Politics Project survey released October 13, 92 percent of Republican likely voters said they would vote for him; the two polls use different samples, so this is a comparison across pollsters, not a measured trend. His favorability among Republicans, 69 percent, still trails Talarico's 92 percent among Democrats. Late-deciding partisans usually come home, and this looks like the start of that.

Money now cuts both ways. Talarico reported raising $64 million in the third quarter, the largest such haul of any Senate candidate in history per Politico, but Axios reported that Elon Musk gave $158 million in September alone, much of it to a rescue effort for Paxton that began in late August. A challenger's fundraising measures enthusiasm; a megadonor's rescue measures how worried the party was.

Talarico is back in public view after nearly two weeks off the trail, including a CNN town hall reported by Axios, which removes the absence penalty we applied on October 12. Paxton still has not agreed to debate. A frontrunner avoids debates; a candidate who has just regained the lead in his own party's eyes may simply see no reason to take the risk.

The reference class is Texas Senate races since 1990: Democrats lost all 13, and the last Democrat elected was Lloyd Bentsen in 1988. Big money has not broken the streak. Beto O'Rourke raised nearly $45 million more than Ted Cruz in 2018 and still lost by 2.6 points, and Colin Allred outraised Cruz by about $21 million in 2024 and lost by 8.5.

The closest analogue is O'Rourke in 2018, a charismatic Democrat against a polarizing Republican in a Trump midterm. The analogy breaks in three ways that favor Talarico: the national environment is now worse for Republicans than in 2018, the Republican nominee is struggling to unify his own party, and the seat is open, so no incumbent advantage protects the Republican.

Start from the Texas prior. Zero Democratic wins in thirteen tries since 1990, smoothed with the rule of succession, gives about 7 percent. That rate counts years in which no Republican nominee trailed in October polling, so it is weak evidence here and the polls carry most of the weight, which is why our number sits far above it.

What changed since October 12: four polls released on October 13 and October 14. Political Wire reported a Siena College/ReconMR poll with Talarico ahead 48 to 44 and a Quantus Insights poll tied at 47. The UT/Texas Politics Project poll, fielded September 25 to October 5, had Talarico ahead 50 to 47, and Marist, fielded October 8 to 11, had him ahead 47 to 45 among likely voters, with his registered voter lead down from six points last month to four. Two of our signposts fired. The down signpost, Paxton's share of Republicans above ninety percent in a quality October poll, is met by the UT poll. The up signpost, a third quality October poll with Talarico ahead by 4 or more, is met by the Siena/ReconMR poll.

Arithmetic, our calculation: 59, minus 8 for Republican consolidation, plus 3 for the third poll showing a lead of four, plus 1 because Talarico is back on the trail, gives 55. As a cross-check we rebuilt the number the way we did on October 12: the seven polls since late September, including the Fox News likely voter poll released October 2 with Talarico ahead 51 to 49, with the four newest counted twice, average a Talarico lead near three points. Subtract about two points for the Texas polling misses of 2020 and 2024 and a little for remaining Republican homecoming, and a race this close with a polling error of about six points lands in the low 50s. Marist's finding that Texas likely voters prefer Republicans by five on the generic congressional ballot also says Talarico is running on crossover votes that can erode. We split the two methods and land at 54 percent, with a range of 42 to 65.