The final polls show a blue surge. Can we trust them?

Democrats gained ground right to the end in the final New York Times/Siena University polls of the 2026 midterm campaign. The party is ahead in all seven states likeliest to decide control of the Senate: Michigan, Kansas, Texas, Maine, Ohio and, more narrowly, Alaska and Iowa.

A substantial Democratic lead in Texas? Kansas? Anything like these results would have been unfathomable just a few months ago.

In today’s polarized era, there’s no precedent for a party to fare this well in so many states that are generally safe wins for the other side. The only plausible comparisons are the Democratic victories in 2006 and 2008, but the recent polls outpace even those. The average of all polls shows Democrats leading in the same states as the Times/Siena surveys, giving the party the edge in more solidly Republican Senate seats than it managed to flip in 2006 and 2008 combined.

On the other hand, there’s precedent in the Trump era for polls to overestimate Democrats, and it’s reasonable to wonder whether the polls might be overestimating Democratic strength yet again. This is a real possibility. President Donald Trump’s supporters are relatively unlikely to respond to polls, which introduces the ever-present risk of a polling misfire. And there are signs that this is a serious problem for pollsters again this cycle.

Nonetheless, Democrats lead by enough that they could win the Senate even if the polls underestimate Republicans by as much as they did in 2016 or 2020. And if there were ever a time when American voters might offer such a decisive verdict on the president, this could be it.

Historically, quagmire abroad and high prices at home are the ingredients for a failed presidency. Since the U.S. attack on Iran, Trump’s approval ratings have fallen into the mid-30s — below the long supposed “floor” in his support. It’s made him the most unpopular president heading into a midterm election since Harry Truman in 1946.

For the first time, a significant number of Republican-leaning voters disapprove of his job performance. He’s much less popular than he was ahead of previous Trump-era elections, including those when Republicans were clobbered.

His approval ratings are at least a net 10 points worse than they were in 2025 before decisive Democratic victories in New Jersey and Virginia. They’re a net 12 points lower than they were heading into Election Day in 2018, when Democrats won the national vote by nearly 9 points in a “blue wave” (though they were unable to flip key red-state Senate races).

With Democrats again needing to win reliably red states to take the Senate, it’s absolutely possible that Democrats will be disappointed once more. To keep those Senate seats, Republicans merely need to get Republican-leaning voters to show up and vote Republican, as they usually do. It wouldn’t be a shock if that happens. This is still Texas and Kansas we’re talking about.

But this seemingly simple task is much more challenging for Republicans today than it was in 2018. Back then, Republicans could win merely by holding voters who supported Trump. His approval rating was over 50% in all of the crucial red states where Democrats fell short. This time, they won’t have that essential advantage. His approval rating is under 40% across the five red states included in this poll (Kansas, Texas, Ohio, Alaska and Iowa).

The president’s unpopularity may simply make it impossible for Republicans to compete in blue or purple states. The latest polls show Democrats making their largest gains in Maine and Michigan, where Troy Jackson and Abdul El-Sayed now hold meaningful leads.

Here’s a tour of the findings by state — and a look at the question of whether the polls might be off once again.

At the beginning of the year, Texas and Kansas didn’t figure prominently in Democrats’ path to control of the Senate. They haven’t won a Senate seat there in a generation (or three).

Now Texas and Kansas look like two of the Democratic Party’s best red-state opportunities. That’s true not only in Times/Siena polls, but also in the average of other recent polls.

A core problem for Republicans: Ken Paxton in Texas and Sen. Roger Marshall in Kansas are two of the party’s weakest candidates of the cycle. They’re the least popular candidates tested in these polls, and their favorability ratings have fallen since the last Times/Siena polls in September, despite tens or hundreds of millions of dollars in ad spending on their behalf.

Just 33% of Texas voters say they have a favorable view of Paxton — who has faced investigation, indictment and impeachment — and 54% have an unfavorable view.

The ratings for Marshall, a former practicing physician, aren’t much better, with 39% favorable and 50% unfavorable. He has faced a deluge of criticism after a Times report that he sued hundreds of patients over minor medical debt. And he hasn’t helped himself with a string of missteps on the campaign trail, like saying, “This is one of the best economies I’ve ever seen in my lifetime.”

These favorability numbers are poor in any circumstances, but they’re especially bad for Republicans in a red state. Sen. Ted Cruz, R-Texas, almost lost to Beto O’Rourke in 2018, but Texans had a favorable view of Cruz and approved of Trump’s performance. With neither condition in place, why wouldn’t Democrats get over the top this time?

In contrast, Democrats have put forward two of their strongest candidates of the cycle in James Talarico in Texas and Adam Hamilton in Kansas. They have wide appeal and both lean into their religious training. Hamilton is senior pastor of America’s largest United Methodist congregation; Talarico is a seminarian. They boast among the highest favorability ratings of any candidate in our polls.

Of the red states in the poll, Kansas is the only one that has made a meaningful shift toward the Democrats over the last month. Hamilton has become much better known, with the share of voters without an opinion of him falling to 6% from 36%. As more voters have gotten to know him (and Marshall), he’s pulled into a clear lead.

Republicans in Texas have an additional problem: crumbling support among Hispanic voters.

The poll finds Talarico with a 38-point lead among this group, 65-27, an enormous swing from 2024, when Hispanic voters were evenly split in the presidential race, according to Times Voter Trends. This kind of swing would be enough to turn Texas into a toss-up even if Paxton weren’t such a flawed candidate. Indeed, the poll finds Republican candidates ahead by less than 1 point across the state’s U.S. House races while the Democrat Gina Hinojosa leads by 1 point for governor.

Democrats made their largest gains since September in Maine and Michigan, the only two blue or purple states among the core Senate battlegrounds.

Together with the result of our last Minnesota poll, the Democratic surge in these purple-to-blue states suggests that the national political environment is simply too inhospitable for Republican candidates trying to compete for the support of Democratic-leaning voters, even when they may have relatively favorable candidate matchups.

The largest Democratic gains came in Maine, where Jackson led by 6 points, up from a 3-point deficit in September. Jackson remains relatively unpopular with Maine voters — and less popular than the Republican incumbent, Sen. Susan Collins. But the gap between the two narrowed somewhat, with the share of voters expressing a favorable view of Collins falling to 48% from 51%.

The story is somewhat different in Michigan, where El-Sayed now holds a double-digit lead. Unlike in Maine or Minnesota, this is not a state where Republicans have a clear candidate quality advantage. Republican Mike Rogers’ favorability rating is negative; Democratic El-Sayed’s numbers are positive.

The Michigan results follow a string of recent polls showing El-Sayed making steady gains over the last month, as he consolidates the support of Democratic-leaning voters who started out skeptical of his candidacy. His lead now matches his party’s across the state’s U.S. House races; in September, he underperformed Democrats on the generic ballot by 6 points.

The poll was taken almost entirely before Haley Stevens, El-Sayed’s Democratic primary opponent, made a stunning eleventh-hour endorsement of Rogers. The poll showed almost no sign of remaining dissent in the Democratic ranks. El-Sayed held a 96-2 lead among voters who said they voted for Harris, up from 91-5 in September.

Black voters, who offered tepid support to El-Sayed in September, now back him, 93-4. Rogers did not receive the support of any respondent in the city of Detroit (among 43 people).

Voters over age 65 have also swung toward El-Sayed, though he continues to run well behind the Democratic candidate for governor, Jocelyn Benson, among this group.

At the beginning of this election cycle, these three states seemed like the Democratic Party’s best opportunities.

Ohio still looks like a plausible part of the Democratic Party’s path of least resistance to the Senate, with Sherrod Brown leading by 6 points. But at least in these polls, Iowa and Alaska seem less favorable toward Democrats.

Unlike in Kansas or Texas, Republicans have put forward conventional, scandal-free candidates in those two states. They’re not necessarily popular, to be clear. Rep. Ashley Hinson, R-Iowa, is the most popular Republican of any tested in the polls, but voters were split on whether they had a favorable or unfavorable view of her, 43-43. Nonetheless, even she trails narrowly in the poll. Her opponent, Josh Turek, is viewed more favorably than any candidate of either party across these seven surveys.

Mary Peltola was the only Democrat to lose ground since September in Times/Siena polling. The Alaska poll was almost entirely completed before the Times reported that operatives aligned with the Peltola campaign had orchestrated an effort to place another Dan Sullivan on the ballot to siphon votes from Republican Sen. Dan Sullivan. (Alaska is a challenging state to poll, so we started earlier than in the other states.)

Over the history of the Times/Siena poll, there have been two previous elections when the final polls showed everything seeming to break toward Democrats.

The first was in 2018. Our final polls showed Democrats gaining across the board, including in races that had barely been on the radar. Democrats surged by double digits in just a few weeks to build startling leads over longtime Republican incumbents in districts that supported Trump in the previous election. Democrats even led in Georgia’s 6th, which had badly disappointed them in a special election one year earlier when Jon Ossoff lost.

In the end, the final Times/Siena polls were dead-on — including where Democratic strength was most surprising.

The second time was in 2020. Joe Biden steadily gained in our polls in October, as Trump’s poor debate performance and subsequent COVID-19 hospitalization reignited the possibility of a landslide win.

In the end, the final Times/Siena polls in 2020 underestimated Trump by 5 points (Biden won, but by much less than our polls suggested). The only accurate polls we conducted were taken a full year ahead of the election, before the pandemic kept people indoors and contributed to Democrats responding to surveys at much higher rates than Republicans.

This cycle, Democrats are once again responding to Times/Siena polls at a much higher rate than Republicans. Across these final battleground polls, registered white Democrats were about 30% likelier to take the poll than registered white Republicans. That’s actually a bit higher than in the final polls in 2020.

In recent days, there have also been warnings from other pollsters. The Washington Post released the result of a poll finding Hamilton up around 10 points in Kansas, but the pollsters did not trust the result. Registered Democrats were more than three times as likely to cooperate with the survey as registered Republicans (though it is worth noting The Post did not observe a similar problem in earlier polls of Maine and Michigan).

The persistence of this kind of nonresponse bias introduces the risk of a systematic polling error like those in 2016 or 2020. That said, nonresponse bias doesn’t make a systematic polling error inevitable. Our 2022 final polls were on the money even as white Democrats were 28% likelier to respond than white Republicans. Similarly, our accurate 2018 polls had a much higher response from Democrats.

There are two factors that help explain why it’s possible for the polls to be right, even when Trump’s supporters are less likely to respond.

One is turnout. There’s a lot of evidence suggesting that Trump’s supporters aren’t just less likely to respond to surveys, but also less likely to vote in midterm elections. This would help explain the basic pattern of polling error during the Trump era: Polls have a pretty good record in midterms but not in presidential elections.

The second factor is a technique called “weighting,” in which pollsters make statistical adjustments to give more weight to voters from groups underrepresented in the survey responses. This gives pollsters a chance to produce reasonable results, simply by giving more weight to registered Republicans, self-identified Republicans or the voters who say they voted for Trump who did respond to the survey.

Weighting is not a panacea. That’s because nonresponse bias can run much deeper than what can be observed and corrected in the data. A poll might be weighted to have the right number of Trump supporters but still have the wrong kind of Trump supporters — and particularly the kind of Trump supporter who might be voting Democratic.

Imagine, for instance, if the Trump voters who responded to a pollster were likelier to be the kind who supported Nikki Haley in the 2024 primary. A pollster might then have too many of the kinds of Trump supporters who would readily defect to Democrats. The Post believed it had this kind of issue in Kansas: Even after weighting, the results still seemed too good for Democrats.

Despite its imperfections, weighting has kept the survey industry afloat. In elections like 2018 or 2022, it was sufficient. For many pollsters in 2024, it probably prevented another misfire like those in 2016 or 2020. As far as I can tell, many public pollsters — maybe even most public pollsters — would have shown Kamala Harris with a clear lead over Trump in 2024 if they had used the weighting techniques that were common back in 2018 or 2020. Instead, they’ve adopted heavy-handed weighting techniques that can bludgeon even the worst samples back into the ballpark.

You can see why it’s hard to make any assumptions about the polling error this cycle in today’s Times/Siena polls. This summer, we adopted a new set of weighting techniques to improve our polling. When we reevaluated our 2024 polls with this new weighting approach, we saw that our polls would have been about 1.6 points better on average. The difference felt meaningful, but it was fundamentally modest.

This time? Well, look for yourself. Here’s what our final polls this year would show if we had used the weighting from prior cycles:

— Michigan: D+16 (today’s actual published result is D+11)

— Alaska: D+6 (actual D+1)

— Iowa: D+4 (actual D+1)

— Kansas: D+10 (actual D+5)

— Maine: D+5 (actual D+6)

— Ohio: D+11 (actual D+6)

— Texas: D+9 (actual D+7)

On average, that’s a difference of more than 3 points.

What does all of this mean for 2026? Well, you can choose your own adventure. More than anything, it illustrates how hard it is to predict how the polls might err.

Maybe it implies that we’ve adopted weighting techniques that can handle the problems that caused the Times/Siena poll to overestimate Democratic strength in the past, especially in white working-class states. If so, maybe we’re dead-on.

Or perhaps one could argue that the Times/Siena polls will underestimate Democrats. Our polls were accurate in 2018 and 2022, after all, so maybe one should trust the old approach in a midterm election. If so, maybe the Democrats will do even better than our polls suggest.

Or maybe the extraordinarily blue results on our old methodology and the increased effect of weighting suggest that nonresponse bias is worse than ever. If so, polls could seriously underestimate Republicans once again. There’s no reason to assume that 2016 or 2020 is the worst-case scenario.

We’ll find out soon enough.

This article originally appeared in The New York Times.