In general, the election models for the 2026 midterms predicted a blue wave, but undersold the massive blue tsunami America saw on election night. Very few people predicted Texas, Kansas, South Carolina, North Carolina, Mississippi, Maine, Iowa, and Ohio all flipping senate seats blue. Texas, Kansas, Iowa, Ohio, Nevada, and Georgia also flipped gubernatorial seats. So, what did they get wrong?
Isaac Saul of Tangle (Center) said, “Right now, the generic ballot appears to favor Democrats by +13. The polling average (which we all had a hard time accepting) was only +9, underestimating Democrats in 14 of the 15 key Senate races, and the average polling error across those states was 4.4 points.”
The Cook Political report, which The Washington Post (Lean Left) called “Washington’s favorite political oracle,” was right on the money in 2024. This time, they lowballed Democrats. On October 30th, 2026, they predicted that the Democrats would pick up 2–6 Senate seats and 12–25 House seats; Democrats are currently projected to pick up 7 Senate seats and 31 House seats.
Lakshya Jain of The Argument (Lean Left) had an election model that put Democrats at +11 nationally, only two percentage points off from the actual result. He referred to the polls as an inverse of 2016, adding that pollsters didn’t seem to actually believe their results, and “they were so scarred by 2016, 2020, and to some extent ‘24 that they were like, if we miss low on Democrats, so be it, but the industry reputationally cannot take another cycle of missing out on Trump voters.”
Marc Novicoff at The Atlantic (Left) made similar points, noting that it’s hard to get people to pick up the phone to participate in polls. “The adjustments they made this cycle were in service of maximizing the accuracy of their surveys, yes, but they also had another goal: not getting burned again,” he added. “In the long history of electoral polling, errors have not tended to favor one party over the other. Then, from 2016 to 2024, they did. That they no longer do is perhaps one more sign that the Trump era is ending.”
Did pollsters adjust their results too far towards Republicans?
GQR Insights & Action Democratic pollster Anna Greenberg said that the polls weighted conservative respondents more heavily to compensate for the assumption that Democrats are more likely to answer polls generally. Although they didn’t quite predict how many young people, and young men in particular, would vote Democrat, she implied it was a great night for the polls. “Certainly, the polls were directionally correct. I think they anticipated a blue wave. I think they probably did not anticipate the tsunami,” she said.
Before the election, polling modeler Nate Silver (Lean Left) criticized a poll from Marist College which appeared to have made an implausibly large adjustment against Democratic chances in the Maine senate and gubernatorial races. He theorized that Marist was afraid of publishing optimistic numbers for Democrats and “reverse-engineered its likely-voter model to produce a result that was more aligned with the consensus in Maine.” Both candidates ended up outperforming Marist’s poll.
Daron Shaw, a GOP pollster and the Republican partner on the Fox News Poll, and Pollster David Paleologos, the director of the Suffolk University Political Research Center, echoed that the polls were directionally correct but undercounted Democrats. Chris Anderson, the Democratic partner on the Fox News Poll, said, "This is sort of a seismic election and you can’t model for a seismic election. We definitely didn’t undercount Republicans, that’s for sure."
Who did well?
Jain argued that moderates tended to win the day. Moderate Republicans, even those that lost like Sen. Susan Collins out of Maine, outperformed expectations. While progressives that won, he said, largely came in under expectations. Democrats won big in areas with more fiery Republicans like Ken Paxton in Texas and Loren Bobert in Colorado. This aligns with extensive data that voters are more likely to vote against the other side than for their own.
It’s possible the progressive wins were a result of a “vote blue no matter who” attitude, but progressives themselves are more likely to see the wins as a mandate to change the status quo and a rejection of a moderate Democratic party.
Jain also noted that young people were the biggest factor here and when we talk about young men, they’re more likely to be disengaged and, when they do vote, to focus on pocketbook issues than women, who are more likely to factor social issues into their vote. Perhaps among rising housing and healthcare costs, a populist left economic candidate was an appealing option—or at least more appealing than Republicans who support Trump’s war in Iran and his tariffs driving up costs.
What’s next for polling?
Polling via phone call used to be considered the most accurate polling method, but now people are less likely to pick up the phone, particularly for unknown numbers, due to a surge in spam calls. Some states have made efforts to prevent spam calling, such as requiring callers to physically type in phone numbers. But it may be hard to get people to start answering the phone again, now that text messaging is a more convenient option in many cases.
These factors leave room for innovation, such as using deliberative groups, a method spearheaded by AllSides, that allows for convening dozens of virtual focus groups at once without a moderator. The organic conversations that come out of these groups could perhaps clue pollsters into what personalities and demographics are going unaccounted for in their models.
Clare Ashcraft is a media analyst and product manager of AllSides Roundtables. She has a Center bias. Reviewed by product manager Evan Wagner (Lean Left) and news and bias assistant Johnathon Held (Lean Right)
.png.webp)
