Four weeks from the midterms, the headlines have settled on one story: Hispanic voters have dumped Donald Trump, the 2024 gains are dead, and Texas and Florida are about to come unglued. Reuters/Ipsos has Hispanic approval at 24 percent, the lowest of either term. NBC News/Telemundo has it at 27 percent approve, 69 percent disapprove, with Latino registered voters preferring a Democratic Congress 59–35. Nate Cohn averaged recent high-quality polls at roughly Democrat +35 among Hispanic likely voters, against a Harris edge of about 7 points in final 2024 surveys. That is a reversal. It is not a reason to set the House forecast on fire.
Start with what 2024 actually was. Trump did not win Latinos. He cut a generational Democratic margin and took something like 45 to 48 percent of the Hispanic vote, up from about a third in 2020. Pew had him near 48 percent to Harris’s 51. The press treated that as a realignment. It was a price-driven swing in a bad economy, concentrated among men and newer voters, sitting on top of a bloc that still leaned Democratic. Reversion toward the old pattern was the base case the day after the election. Discovering it in October and calling it a collapse is how you get paid for a story you should have written in pencil.
The approval numbers are also being used as if they were a ballot. Job approval in the mid-20s is ugly. It is not the same as a congressional vote, and it is not the same as what his own voters are doing. Pew’s April survey had only 22 percent of Hispanic adults approving, down from 36 percent in February 2025. Among Latinos who actually voted for him in 2024, approval was still 66 percent. That is a 27-point drop from the start of the term. It is also a majority. Equis, which oversamples this electorate for a living, has spent the cycle finding a narrower Democratic lead in competitive districts than in the national Latino sample — on the order of 11 to 13 points, not 24. About a third of his 2024 Latino voters told Equis they were disappointed or regretted the vote. That leaves two-thirds who did not.
The issue mix explains the drop without requiring a conversion story. Cost of living is the top complaint, and Trump’s ratings on prices have cratered. Reuters has tied the latest inflation spike in part to the war with Iran. Interior immigration enforcement is the other bruise: NBC/Telemundo found about two-thirds saying deportations have gone too far, and roughly a quarter saying they know someone detained or deported. Border security still polls better than the rest of the agenda. That is a coalition under stress, not a coalition that joined the other party. Democrats’ own favorability with these voters remains mediocre. Equis has had them roughly split in competitive House seats. People can hate the grocery bill and still not trust the alternative.
Geography is where the overreaction gets expensive. Latinos are about 16 percent of eligible voters. A large swing among them moves the national popular vote a few points and decides a specific set of districts. Competitive House seats are more Latino than the country, Equis put them around 19 percent by citizen voting-age population, and a dozen of Cook’s competitive seats are at least 20 percent Hispanic. South Texas, a couple of Miami-area seats, the Central Valley, Colorado’s 8th, Arizona’s 6th: that is the list. A Hart Research survey of Latinos in 17 battleground districts found Democrats up 58–35, and about 11 percent of 2024 Trump voters in that sample moving to the Democrat. Enough to flip or tighten several seats Republicans drew on the assumption that 2024 was the new floor. Not enough to “transform the midterm map” into a Democratic landslide.
The raters have already priced a Democratic House. Cook’s late-September range was a gain of 5 to 15 seats. Quantitative models sit higher, some around a net of 20, with a wide band. Democrats need about three. The Latino reversion makes the Texas and Florida end of that range more plausible than it looked after redistricting. It does not add a second wave on top of the generic-ballot environment, weak presidential approval, and the usual midterm penalty for the party in power. Those were already doing the work.
Two mechanical limits keep getting skipped. Midterm turnout among Latinos usually lags presidential turnout, and the voters cooling on Trump are the ones most likely to sit it out. A poll of adults is not a poll of people who will cast a ballot in a House race. Subgroup samples are also noisy. The same Economist/YouGov series that feeds the panic showed Hispanic approval at 37 percent in June and 36 percent in late August. The story was “gains” then and “collapse” now. Both pieces ran.
Call the shift what it is. Republicans lost ground they won on prices, and enforcement alienated a slice of the 2024 crossover vote. In a few heavily Hispanic districts, that can be the difference between holding a gerrymander and handing it back. Nationally, Latinos remain a Democratic-leaning bloc that swung, and is swinging part of the way back. That is a correction. The coverage is treating it like a funeral.
