What’s up with bad – and even fake – polls in the 2026 midterms? We asked an expert.

Washington | Oct. 29, 2026, 5:00 a.m. ET | 8 min read Political polls are just a “snapshot in time, not a predictor” of how an election will turn out, the axiom goes.During the 2026 midterm election cycle, that caveat seemed especially apt in some key statewide primaries – including for Michigan and Wisconsin – which turned out quite differently from how the polls had suggested.In another twist, “polls” that were widely cited and, in one case, incorporated into The New York Times’ average of major polls turned out to be fake. The two most prominent fake pollsters ultimately acknowledged what they had done, calling it a “social experiment” to expose a lack of media vetting and oversight of polls. Now, on the eve of the Nov. 3 midterms, polling is again in the spotlight. As things stand, the Democrats are on track to retake control of the U.S. House, and could win a majority in the U.S. Senate. Or so suggest the real polls, as cited by experts. Umit Bektas/ReutersThe U.S. Capitol is reflected in the Capitol Reflecting Pool in Washington, Oct. 27, 2026. History also points to a strong Democratic performance in November, as the president’s party almost always loses seats in the midterms – sometimes a lot. But, given the polling anomalies this cycle, perhaps more surprises are in store.The Monitor turned to veteran pollster Charles Franklin, director of the Marquette Law School Poll in Wisconsin, for his perspective. Dr. Franklin has been regularly involved in national polling since 1990 and joined Marquette in 2012, where he has conducted more than 100 polls of Wisconsin and the nation. Here are excerpts of our conversation, lightly edited for clarity.What are the origins of today’s problems with polling? The biggest one that was starkly displayed in 2016 is what we call “nonrandom, nonresponse.” In English, what that means is, the people that won’t talk to you, by definition, are not represented in the poll.Twenty, 30 years ago, Democrats and Republicans alike hated telemarketing calls, and wouldn’t answer the phone or take a poll. But there wasn’t a partisan difference, and so even though people didn’t all respond to the polls, it didn’t really distort the Democratic and Republican balance.What we saw in 2016, with President [Donald] Trump’s arrival on the scene, is that it seems pretty clear that he motivated a set of people who, in the past, had not voted very often. It’s important not to exaggerate, but maybe that made up 4% to 6% of the electorate that he mobilized. But those people who were so positive toward Donald Trump also were very negative toward government, the media, and pollsters, and so they were exactly the kind of people that didn’t want to talk to pollsters. Annie Rice/APPresident Donald Trump speaks at a campaign rally at Lubbock Memorial Civic Center, Oct. 27, 2026, in Lubbock, Texas. In the past, because they also didn’t vote, they didn’t affect the polls very much. But once they became voters and declined to do the interviews, then that introduced a new bias in the polling and left those voices unrepresented in survey results.The irony, of course, is that by not participating, they actually make their own candidate look less good in the polls, and that helped tip the polls to show Trump losing, when in fact, as we know, he won by less than a point in Wisconsin, Michigan, and Pennsylvania in 2016. But that was a systematic polling error across virtually all of the polling.What have pollsters done to address this problem? The polling industry has spent the 10 years since then trying to develop ways to reach those voters – not just Trump voters but, more generally, hard-to-reach actual voters. ... We’re getting better, but the thing that keeps pollsters awake at night is, of course, you don’t know much about the people that won’t talk to you.So it’s hard to know whether we’ve sufficiently overcome that problem, or whether it might be a problem now. The difference that we’ve seen in the three presidential races starting in 2016 versus the midterm elections is that in 2018 and 2022, the polling was actually pretty accurate.In my particular case, 2016 was our biggest miss. But in 2018, with the same methodology, we were within 2 points of the actual outcome. That demonstrates that without any change in the polling methodology between those two years, the change in the nature of the electorate, and therefore the nature of who’s a respondent and who’s especially a nonrespondent, changed because, as we pretty much know now, President Trump motivates these folks to participate. But when he’s not on the ballot, they drop out.What went wrong with your polling on the Wisconsin Democratic primary for governor? For months, democratic socialist Francesca Hong was way ahead in the polls. She ended up losing narrowly to David Crowley, a moderate Democrat. What went wrong is not, I think, wrong, but that the campaign actually matters. You had one candidate drop out. Then, you had another’s campaign implode because of false finances. Then, the guy that dropped out got back in, and then, yet another candidate dropped out. Nam Y. Huh/APDemocratic gubernatorial candidate Francesca Hong makes a heart gesture to supporters after speaking during a primary election night watch party, Aug. 11, 2026, in Madison, Wisconsin. All that happened in the last three weeks and four weeks of the campaign, starting July 8. So the front-runner, Francesca Hong, the democratic socialist, in our polling was getting 38% of the vote, and that put her well ahead, though obviously well below 50%. It turned out that she got 39.3% of the vote. The story of that race is just a bizarrely chaotic endgame to that campaign, with those people dropping out, getting back in, and the governor, who had said he would not endorse in the race, endorsing, and then campaigning with Crowley enough to put him over.Were there other factors that affected polling? Yes. The people who were most undecided in our polls were older voters, over 50, and that’s unusual because usually those folks make up their minds early. [These were] the Democratic voters who were less favorable to democratic socialism, and so you have the combination of a high turnout group that is undecided and does not have a positive view of democratic socialists.[Our polling] showed where the undecideds were, and we showed that those undecideds were not the sort of folks you would expect to vote for a democratic socialist candidate. What we did not have evidence for is that those folks would necessarily go to David Crowley.How do you explain the fake polls – Median Strategies and Wedgewood Polls – that were initially taken seriously this cycle? [Median Strategies] popped up when they put out numbers for the Los Angeles mayor’s race, and then the mayor retweeted it. The LA Times then investigated and immediately found that it was fake – and did a pair of front-page stories about it. Amanda Villegas/APLos Angeles Mayor Karen Bass speaks during a campaign event, Oct. 19, 2026, in Los Angeles. So that’s, on the one hand, an example of a failure, because the polls were out there and they got enough attention that the mayor would retweet it. But it is also an example of the actual media quickly looking into it and quickly coming to the right conclusion that this is bogus. What about the so-called Wedgewood polling that was incorporated into the New York Times averages of major polls? How did the Times and another polling aggregator, FiftyPlusOne, figure out that Wedgewood was fake? Both outlets sent questions to the alleged pollster, asking about their methodology and who they were. The trouble was, apparently Wedgewood just plain lied about it.What’s the next step beyond essentially taking their word for it that they’re acting in good faith? It’s too late this cycle for that to make much difference. But looking ahead to 2028, when we’re going to have a gazillion primaries in both parties, this is going to be a big deal then. I’m sure all of those polling aggregation sites are thinking hard about what they can do differently.How do you explain the rise of fake polls? Have we had this in past cycles? We have, but in a somewhat different way. The short answer is, the rise of social media. And the fact that you can get a lot of attention on the hyperactive social media channels means that there’s an incentive to get clicks by putting out something that you know the social media folks are anxious to hear about, right?Oh, I need a new poll of Maine. Well, I’ll give you one of those. There’s a big thirst and demand for polling, and it extends to races that the major polling organizations may not see as important enough to do a lot of polling in. Another factor in this cycle is the rise of prediction markets, like Kalshi and Polymarket, where people place bets on certain outcomes, be it sports or politics or just about anything. How are these markets affected by polls? We have reasonable evidence that the markets move with polls, though it’s not clear that a single poll moves them very much, if at all. Take the Texas Senate race, where there was a lot of polling. One poll didn’t move the market much one way or another. But then there was the Michigan primary, where there was not much polling. That was the case where one result could move that market.You’re talking about Democratic state Sen. Mallory McMorrow, who dropped out of the U.S. Senate primary after the fake Wedgewood poll showed her in single digits, right? And that affected prediction markets? Right. She was struggling to move out of third place, and so when this fake poll came out saying she was doing even worse than expected, that had an impact. Julia Demaree Nikhinson/APMichigan state Sen. Mallory McMorrow speaks during a news conference with Abdul El-Sayed (standing left) at the Spirit of Detroit shortly after Dr. El-Sayed was declared the winner of the Democratic nomination for the U.S. Senate in Michigan, Aug. 5, 2026, in Detroit. She released a statement saying her fundraising had “dried up.” That’s an example of those fake numbers actually having an impact on the campaign.Is polling done online or still by phone? We are mostly online now. How do you get through to people? For the state sample, we draw from the state’s list of registered voters. That’s a public document of who’s registered to vote. We send an email and a text to that sampled person and say, “This is the Marquette Poll. We’re conducting a public opinion poll. We’d like you to participate. Here’s a link to complete the poll online.”We then follow up with a reminder, both by email and text. If after those four attempts to reach someone we still haven’t heard back, then we do still phone them. That means we end up with about 5% to 10% of the sample that is still reached by telephone. But that’s compared to 100% of the sample that was all telephone prior to 2022, when we switched.We do run into suspicion about clicking on a link, and I’ll get anywhere from one to four or five phone calls or emails from people we’ve sampled saying, “Hey, I got this message. Is this really you?” Deepen your worldviewwith Monitor Highlights.Politics with respectGet political stories with respectful analysis.There‘s a world of new ideas in everyBooks newsletter.There‘s more to life, enrich yours withCulture & Learning weekly.Follow humanity‘s discoveries withScience & Nature stories in your inbox.Gain a spiritual perspectivefrom the stories in your inbox.Want to understand the deeper impact of critical events? Learn the Monitor‘s insight.Already a subscriber? Log in to hide ads. The trust issue must be huge in polling these days. Well, it is. I actually tell the folks that reach me, “You know, hey, I’m glad you took the required IT training, and you paid attention to it.” ALREADY A SUBSCRIBER? LoginReal news can be honest, hopeful, credible, constructive.The Christian Science Monitor was founded in 1908 to lift the standard of journalism and uplift humanity. We aim to “speak the truth in love.” Our goal is not to tell you what to think, but to give you the essential knowledge and understanding to come to your own intelligent conclusions. Join us in this mission by subscribing.