How much can we believe the polls?

descriptions off, selected

captions settings, opens captions settings dialog

captions off, selected

This is a modal window.

Beginning of dialog window. Escape will cancel and close the window.

End of dialog window.

The Hill's Headlines (PM) - September 2, 2026

The Hill's Headlines (PM) - September 2, 2026

▪ What to make of the polls

▪ Hegseth under rising scrutiny

▪ Google catches a break

▪ Schwartz’s first test as CDC director

Thank you for signing up!

Subscribe to more newsletters here

Most election observers seem to agree that about eight states are in play to decide control of the Senate this November. But looking at some recent polls, one might think solidly red states like South Carolina and Arkansas are gettable for Democrats too. 

Democrats have been riding high in recent weeks as they’ve developed a steady lead in the generic congressional ballot and are increasingly hopeful about flipping both the House and Senate on Election Day. They’ve also been boosted by President Trump’s falling approval rating, reaching some of the lowest numbers he’s ever seen.

His approval rating currently stands at 39.4 percent in the Decision Desk HQ (DDHQ) polling average, and some recent surveys have placed the number as low as the low 30s.

“I think anytime you see numbers this bad for an incumbent president ahead of a midterm election, it just bodes poorly for the president’s party,” said Geoffrey Skelley, the chief elections analyst for DDHQ.

The big question, though, is just how poorly it will bode this November. History is strongly on Democrats’ side in their battle to win control of the House, but the party is still fighting an uphill battle to win back the Senate. Or is it?

Polling has consistently shown a few states that Trump comfortably carried in 2024 have become key battlegrounds this year, including Texas, Ohio and Alaska.

If Democrats have a legitimate chance to flip those seats, that would suggest a much bluer national picture than 2024, one that most analysts agree is likely given the backlash the sitting president’s party often faces during midterm cycles.

But along with the polls showing Texas state Rep. James Talarico (D) a few points ahead in the Lone Star State and Iowa state Rep. Josh Turek (D) a few points behind in the Hawkeye State have been surveys suggesting Democratic candidates in ruby red areas are in striking distance, too.

In the aftermath of South Carolina’s GOP runoff last month, a pair of polls sponsored by Democratic nominee Annie Andrews’s campaign showed her tied and leading by 3 points, respectively, against Sen. Darline Graham (R-S.C.). A poll from July sponsored by Democrat Hallie Shoffner’s campaign showed her behind Arkansas Sen. Tom Cotton (R) by just 3 points, while another poll not associated with any campaign found her leading by 3 points.

Similar results have been seen in certain House races. One released Tuesday sponsored by House Majority PAC, which works to elect Democrats to the House, showed Democrat Leela Gray leading Rep. Anna Paulina Luna (R-Fla.) by 5 points in a district that Cook Political Report rates as “likely Republican.”

As Trump carried South Carolina by nearly 20 points and Arkansas by more than 30 points, these results create a somewhat hazier picture of what the electoral map actually looks like.

General distrust in polling has only increased following the revelation of a couple fake polls that were released this cycle and big primary polling misses in the Michigan Senate race and Wisconsin governor’s race.

Experts said the key is knowing how to read the polls and not making too much of any single result.

Generally, polls released by a campaign or sponsored by a partisan group are viewed with additional skepticism beyond independent public polling.

Charles Franklin, the director of the Marquette Law School Poll, said polls released by campaigns can provide useful data, but added that the public must keep in mind the motivations behind a poll’s release.

“A campaign only releases a poll if it serves their interest, if it advances their narrative, makes them look in better shape,” he said.

Franklin noted that past analyses have found campaign-sponsored polls tend to overestimate a result by about 5 points in their candidate’s favor. Even if that shift is implemented in some of the aforementioned polls, all of them would still be considered competitive.

Franklin advised against focusing too intently on a single poll, instead taking all polls on a race together to get an idea of where a race stands. If a poll, partisan or not, yields results significantly different from the rest of the surveys, that poll may be an outlier.

That may be the case in the Hendrix College poll that showed Cotton down by 3 points. (It was famously true of a poll showing........

© The Hill