Are The Recent NYT Senate Polls Underestimating Democrats?

Take a look at the partisan makeup of the samples The New York Times used for their newest Senate polls.

Those who follow US politics more closely may know that there was a flurry of high-quality Senate polls by The New York Times/Siena that flew in this week, causing discussion, confusion, and, of course, panic.

While Democrats are favored to win the House even in the face of Republican redistricting, their Senate battle looks much more difficult. They have to flip four seats to win control of the chamber. Apart from holding battlegrounds in Michigan and Georgia, they have to flip Maine and North Carolina, as well as two out of four states that voted for Trump by over 10 points (Alaska, Ohio, Texas, and Iowa). As of today, the Senate is a toss-up, with VoteHub giving Republicans a 56% chance of winning control.

So what do the new polls say, and what might they have missed?

Latest NYT/Siena Polls Might Skew Republican

The polls are a mixed bag for Democrats. They are the clear favorites to flip North Carolina, where former Democratic Governor Roy Cooper leads his challenger by 7 pts. Maine, the most Democratic state on paper, is turning into a close race, as populist Democratic candidate Graham Platner leads Susan Collins (R) by just three points. Democrats in Ohio, Alaska, and Iowa are 2-3 pts behind, while the race between James Talarico (D) and Ken Paxton (R) is tied in Texas. This is troublesome for Democrats as they need to win at least two of the states where they are behind and carry all where they are leading or tied.

But these results also seem strange at first glance. Ohio should be a prime target for Democrats, with former Democratic Senator Sherrod Brown, who is well-known in the state, trying to reclaim his seat in the Senate. Yet, he is doing one point worse than the Democratic Candidate in Iowa, Josh Turek, a race long considered a stretch for Democrats. There is a simple reason for that divergence: Poll sampling.

Sienna/New York Times has recently started to weight their polls not just by party ID, but by recall vote, meaning they ask respondents whom they voted for in 2024. Since most midterms voters are higher-propensity voters who also vote in presidential races, this provides us with some useful data (In presidential elections, more non-voters turn out compared to the previous midterms and vice-versa). Election results, after all, don’t just change because people change their minds, but because some people don’t turn out. In 2024, Kamala Harris suffered from plummeting turnout among 2020 Biden voters. In the 2018 midterms, Democrats profited from the strong turnout of young voters.

In 2026, low-information Trump voters might stay home, and newly motivated, younger voters will likely vote Democratic. Most analysts expect the electorate to be much more Democratic as a result. The New York Times/Sienna polls do not. At least not by much.

The column on the right is the important one. It shows the difference between the sample electorate of the state poll and the state’s 2024 electorate. As you can see, the swing in the electorate of the samples varies slightly but decisively. The state with the unexpectedly good results for Democrats, Iowa, has a sample that’s 5.2 pts bluer (more Harris voters) than the 2024 results. Ohio, the negative surprise for Dems, only has a sample that’s 2.2 bluer than the state’s 2024 electorate.

Yes, states have different profiles and different economies, so different swings towards the Democrats are to be expected. Iowa is a farming state that has been suffering, but, in the data we have, this different election environment doesn’t show up.

Forecasters Predict Blue Electorate

According to The Argument, Democrats will do 8 pts better in Iowa in 2026 than they did in 2024, but they’ll do 11 pts better in Ohio. This does not align with the NYT’s sampled electorate.

Generally, Democrats are favored to outperform their 2024 results by 8 to 10 points, which would put the generic ballot at D+6 to D+8. A D+8 result (the same result as 2018) would likely require not just persuasion, but a bluer electorate. An electorate, the NYT/Siena polls are not working with in most of their polls. This leads me to believe that the NYT’s newest polls might be underestimating Democratic turnout and overestimating Republican turnout (except for Iowa, perhaps). If the electorate will be just 2-3 points bluer than the NYT/Siena polls suggest, Democrats could flip all six Senate seats and end up at 53 seats.

An indicator for better-than-expected results is the Democrats’ sizable advantage with highly motivated voters. In a recent Fox News poll of Maine, Graham Platner lost to Susan Collins by 3 points among registered voters. But among those “extremely motivated” to vote, he leads Collins by 9 pts.

We still have to keep in mind that turnout gaps are smaller in states that have high-profile races, as more voters will be aware of the election. A state with no big race will see lower turnout, with mostly motivated voters making their way to the ballot box. Democrats might therefore benefit if some races do not get nationalized too much and stay “under the radar”. Iowa or Nebraska, where Independent Dan Osborn is running, might be examples for that.

Can We Trust the Polls?

You should, however, also keep in mind that NYT/Siena is an “A” rated pollster, with outstanding methodology and transparency. It might just be that Democrats do unexpectedly well in Iowa while underperforming in Ohio. The best thing to do is throw these polls in the average and wait for Election Day. But can we trust the polls? Yes, but you should be aware of recent polling misses.

While polls famously overestimated Democrats whenever Trump was on the ballot, they didn’t in midterm years.¹ In 2018, most polls were spot on, and in 2022, they even underestimated Senate Democrats in some key races. In 2025, they heavily underestimated Democrats: Mikie Sherrill was forecasted to narrowly win the governor’s race in New Jersey by about three points. She ended up winning by over 14 points. A ridiculous polling miss. I expect midterm polls to be more accurate, but, given recent history, it appears more likely that they will underestimate Democrats rather than Republicans.

1. Interestingly enough, Polls overestimated Democrats the most in Iowa in 2020 and 2024 (by 8.2 points).