Over the past week, eye popping, double-digit leads for Democrats in House polling has raised a question that has haunted pollsters and forecasters throughout President Donald Trump’s campaigns and time in office: whether polls are underestimating Republicans.

Over the past week, a few surveys have found Democrats with double-digit leads on the generic ballot, which serves as a measure of which party respondents across the country prefer, in general, for Congress. Specifically, a recent Quinnipiac survey found Democrats with a 12-point lead while an Economist and YouGov poll found Democrats with a 15-point lead on the generic ballot.

While these polls are not the first to find Democrats with a big lead this cycle, two mid-September surveys both found Democrats with a 12-point lead, they have, in part, spurred on the raising of a perennial question: Can people trust the polls?

This issue has become a feature of the Trump era of American politics because of some now infamous polling misses in 2016 and 2020, with surveys underestimating support for Trump in both instances. Even in 2024, when the polls were largely predictive of election outcomes, what polling error did exist downplayed Trump’s voters. 

In 2026, however, there is less anxiety among pollsters about underestimating Trump’s support. This is largely because Trump himself is not actually on the ballot. In the two midterm elections during the Trump era when Trump wasn’t running, the polls were largely predictive of election outcomes. In 2018, for example, House polling was actually more accurate than the historical average of 5.9%, and in 2022 the polls largely predicted what happened, a very tight battle for the House with a slight Republican edge.

John Culverious, a political scientist and pollster at the University of Massachusetts Amherst, said in an interview with The Caucus that “it varies by pollster” but that by and large pollsters are less concerned about undercounting Trump’s support than they were in 2020 or 2024.

This is because of the quality of the polls in previous midterms and despite Trump’s latest midterm message of “Pretend that I am on the ballot.” Culverius explained that this attempt from Trump to recreate his previous success is unlikely to succeed because it hasn’t worked for presidents in the past.

“The problem isn’t the message. The problem is that the message doesn’t work. It didn’t work for Obama. It didn’t work for Trump and it largely didn’t work for Biden. Every president would love to have the electorate that they won,” Culverius said. 

This doesn’t, however, necessarily mean that the best polls for Democrats will end up being predictive of the final outcome on election day. It’s part of the reason that websites like Real Clear Politics maintain a polling average. Culverius also cautioned about trying to translate generic ballot polling into a specific number of seats gained or lost by a given party. 

“The House picture is hard to think about in aggregate because, what's the difference between, 20 or 30 or 40 House seats? I don't quite know, but also in a typical midterm election, you'd expect Democrats to win the House,” Culverius said.

While some models that do forecast a number of seats for each party have been fairly accurate in the past, the relative dearth of polling in specific House seats, as well as the recent mid-decade redistricting cycle, complicate the issue.

There are a few key trends Culverious, and other pollsters are watching, especially as polls switch over from registered voter models to likely voter models, which are typically based on asking voters whether or not they plan on voting.

The first is male voters, and whether or not men, and especially white men without a college degree, say they plan to vote this year. Depressed turnout among this group, especially in rural areas, could signal the bottom falling out for Republicans, according to Culverius.

For Democrats, Culverius said, the groups to watch are unmarried women and voters in small cities and suburban areas. A surge of these people voting in the midterm electorate would likely mean a boost for Democratic turnout. 

The last group to watch is Hispanic and Latino voters, which Culverius identified as a key segment of voters to watch in terms of persuasion in 2026. Especially in states like Texas, Latino voters on the whole have swung dramatically between elections and appear to be souring on Trump and the Republican Party ahead of the midterms this year.