How Well Did the Polls Predict the Election? | Newswise – Newswise
Sign up for the wires and see archived wires
Browse experts available to comment on breaking news
Request an expert contact, get responses directly to your inbox
Find an expert by topic in a comprehensive database
Press conferences, expert Q&A and pitches in a live, virtual setting
A reported drop in Republican turnout could account for the more modest margins that were predicted in some Democratic victories, said political scientist Brian Schaffner
Newswise — In the lead-up to the midterm election, many people were wondering if they could trust the polls. The polling industry, after all, had had some high-profile misses since 2016, including pre-election polls that pointed to a Kamala Harris victory over Donald Trump in 2024.
But the polls turned out to be right in predicting a blue wave this year, even if the wave was better described as a tsunami.
Brian Schaffner, Newhouse Professor of Civic Studies in the Department of Political Science and Tisch College of Civic Life, said that the undercounting this time around likely stems from an age-old problem for pollsters: Many more people say they will vote than actually do. He spoke with Tufts Now about why some polls did better than others (including the poll he and his students produced in the Tufts Public Opinion Lab) and what surprised him about the election.
People probably aren’t thinking that the polls were too far off, because they correctly pointed in the direction of a Democratic win. However, they were off by several percentage points on average. Most of the polls were showing that Democrats might win the House vote nationally by 7 or 8 points, and it ended up being 13 points.
In races where the polls were clearly pointing to victories for Democrats, like in Georgia and North Carolina, those victories definitely happened, but they were bigger than the polls were suggesting.
I suspect most of the slippage here comes down to identifying likely voters. The hardest thing for election polls is trying to figure out who’s actually going to vote. Pollsters want to make sure the sample of voters looks Republican enough for what they think the election’s going look like, but who they think is actually going to vote is a judgment call. From preliminary data, Republican turnout really dropped off significantly in this election. I would imagine that accounts for most of the underrepresentation we’ve seen of the Democratic victories.
It’s always been hard. But I don’t think we’ve had a midterm election during the modern era of polling when a president’s approval has been this low. In a typical midterm election, pollsters might expect that even if Republicans are not answering polls as much, they will probably still show up to vote. This time, a lot of Republican voters were maybe feeling especially discouraged and wanted to sit this one out.
Most pollsters figure out who their likely voters are and then adjust that group to look like what they think the electorate will look like. What we do, along with YouGov, is first weight the data to look like American adults, and then apply a likely voter model that lets people’s traits and how they’re answering questions—Are you registered? Have you already voted? Will you definitely vote?—affect what the partisan composition of our likely voters looks like.
Our likely voter model reflected what actually happened on election day, which is that Democrats were more likely to vote.
When you have a big, one-sided election like this, the polls are often going to understate it, just because it is so out of whack with recent history. Our poll, for example, showed a 12-point House margin for Democrats. You look at that and you think, when’s the last time that happened?
Also, there’s a tendency in the industry to be very aware of what other polls are showing. If pollsters get a poll that doesn’t look like the other polls, they might change the way they’re weighting their sample, or they might decide not to release the poll. There’s a term for it, called “herding.” Pollsters can be too careful in that way. If you’re going be wrong, you don’t want to be wrong all alone.
On election night, I worked with other pollsters and elections experts for CBS News to help decide when to call individual elections. Throughout the night, I had to focus on the minutiae of which counties had come in in any given state. Only after we made our last call of the night at 2 a.m. and I was walking back to my hotel did it really set in for me, the enormity of this.
Democrats winning in South Carolina, in Kansas, potentially in Mississippi—seats that you never would have thought Democrats had a chance of winning any time in the near future. Much less things like an African American woman being elected governor in a Deep South state like Georgia. If anyone says they weren’t surprised, they’re lying.
It would be really interesting to look at the swing voters—people who voted for Trump in 2024 and swung over and voted for a Democrat. It mostly seems to be driven by the economy, but let’s see what those voters look like.
Also, the question of turnout. Who stayed home? And how easy will it be to re-energize them in two years? What do people want from this divided government? What do they want Democrats to do?
I suspect there’s a wide range of views, and it will be important for Democrats to figure out how to navigate that. And of course, the election’s already three days old—it’s time to start thinking about 2028 and who the parties are going to nominate.
https://now.tufts.edu/2026/11/09/how-well-did-polls-predict-election
Connecting Research and Experts with Journalists
Unlock Your Access to Newswise Research News including Embargoed News and Expert Pitches
Used only to deliver research news. Unsubscribe anytime.
Journalists use Newswise as a source for research news, experts, ready-to-use content and story ideas. Media relations professionals can connect with reporters and share their organization’s news with a wider audience. Public readers discover the latest research news in science, medicine, social sciences, environment, technology, factchecks and business news from the world’s most credible universities and research organizations. More than 7,000 email wires go to journalists from more than 2,400 media outlets around the globe.
2026 Newswise, Inc