Polling experts hail ‘great night’ after correctly calling most key US midterm election races

Political polling experts hailed a “great night” for the industry after surveys accurately foresaw much of the Democratic blue wave that swept across the US in midterm elections.

In only a handful of US Senate races, including Texas and New Hampshire, key polls failed to gauge the scale of the Democratic victory over Republicans on Tuesday.

Ashley Koning, the director of the Eagleton Center for Public Interest Polling at Rutgers University, said: “Some of the most credible pre-election polls that we know of correctly saw the direction of every race that’s been called so far.”

“If anything,” she added in an interview, “the polls underestimated some Democrats this year.”

The accuracy of US polling has come into sharp focus in recent years, especially after Donald Trump outperformed polls in the 2016, 2020 and 2024 presidential elections.

In the lead-up to this week’s midterm elections, polls had indicated that Democrats were poised to win the House and might lead enough states to also take the Senate. On Tuesday night, Democrats flipped both chambers, with victories across the country.

The final New York Times/Siena Poll, published on Sunday, put Democrats ahead in Senate races in Iowa, Kansas, Maine, Michigan, Ohio and Texas, all seats they ultimately won.

In several of those states, however, Democrats outperformed their polling numbers, marking a reversal of a trend seen in recent years in which surveys have tended to underestimate support for Republicans, especially under Trump.

Diana Mutz, director of the Institute for the Study of Citizens and Politics at the University of Pennsylvania, described the pre-election polls as “pretty good”, saying they had generally captured the results without any major errors.

One of the biggest challenges for pollsters, she noted, was “knowing who is actually going to turn out to vote” – especially in midterm elections, which typically see a lower turnout than presidential contests.

Christopher Wlezien, a professor of government at the University of Texas at Austin, observed that the accuracy of the pre-election polls varied on the size of individual victories.

“Some were much closer than others,” he said. “But it was hard to do, because the big swings like this are hard to predict … you have to try to capture who’s going to vote, and that’s always tricky.”

The polling industry has also grappled with decreasing response rates. In September, the Pew Research Center reported that rates had fallen from 36% in 1997 to the low single digits in recent years. It also pointed to what it described as growing evidence that Republicans had become less likely than Democrats to participate.

In this cycle, there was “so much focus on whether polls were underestimating Trump voters, as they have done in presidential elections when Trump has been on the ballot, that there was significantly less conversation about whether they could be underestimating Democrats, as was the case”, said Christine Matthews, president of Bellwether Research and Consulting.

Polls “accurately forecast dramatic shifts toward Democrats among some of the voters who were drawn to Trump in 2024”, Matthews said, referring to Hispanic and younger voters.

Sean Freeder, who oversees the Public Opinion Research Lab at the University of North Florida, said he believed this was likely the first time in more than a decade or so that polls “underestimated the Democratic vote rather than the Republican vote share”.

“I see one of the major effects here being that this might this election have increased the risk appetite for pollsters,” Freeder said. “Pollsters have been burned in the past by reality, and specifically in one direction. Now that, all of a sudden, you have an election in which the other party was undercounted, now I think it’s possible pollsters would say: I’m going to trust my results. I’m going to go ahead and release it.”

Looking ahead to 2028 and beyond, David Paleologos, director of the Suffolk University Political Research Center, said pollsters need to make sure they are not always just asking: “Are you undercounting Republicans? Are you undercounting Trump voters?” but also: “Are you undercounting Democratic voters, too?”

“Pollsters were wary of overestimating Democratic performance because of misses in recent years,” according to Tim Servinsky Jr, who leads the quantitative work at the Center for Survey Research at Pennsylvania State University. “Ultimately,” he added, “the polls got it right, but they possibly underestimated Democratic enthusiasm in places like South Carolina, Texas and Georgia, and because it wasn’t considered competitive, we saw virtually no polling in Mississippi.”