A number of you have written in to ask how the polls ended up doing in the 2026 midterms. I think we know that they underestimated Democratic strength. But how much? How consistently? What I’ve assembled below is just a provisional, back-of-the-envelope read. Others will do this more comprehensively, looking across all polls, comparing them with the final certified results and then adding additional fancy numbers stuff. But I wanted to come up with a provisional read.
I took 12 Senate races that by the widest measure appeared to be contested. I took the final poll average in each from Nate Silver’s Silver Bulletin and I compared them to the current posted results on The New York Times results page. Votes are still being counted, but margins won’t change much. With the poll average and the final result I rounded to the closest whole number before comparing them to find the amount of error or miss.
As I mentioned last night, the data analysts I trust think Scott Colom is more likely than not to win the Mississippi Senate race. A short time ago I noticed that DecisionDeskHQ has more votes reported than most reporting sites which mostly use the data feed provided by the AP. DDHQ is a several year old upstart elections service which has aimed to compete with AP’s data service. They’re legit. They provide data to RCP and many other sites, but not as many as the AP. I’d still like to see these numbers confirmed by another source. But I’m pretty sure they’re real. They now show Colom opening up a lead of 1,864 votes. Add to this that the remaining votes appear to be in Colom-friendly areas. We can’t be sure of the outcome yet. But his odds look good.
NBC News currently projects that Democrats will hold 258 seats in the House in the next Congress — an 81-seat margin. Let’s assume that’s accurate; in any case, it won’t be far off. If that holds, that will be the largest majority either party has held since 1990 — 36 years ago. It’s important to note that this was in the pre-1994 U.S. political system in which Democrats still held large majorities based in part on incumbency in southern districts that were already strongly Republican. That changed decisively in 1994, and U.S. politics has never been the same. The last time Republicans won a majority that large was, I believe, in 1928.
I want to again thank everyone who joined us for last night’s election night video-cast. We had a great time (with an assist from the American electorate) and we hope you did too. It’s not quite as exciting as a rerun. But if you’d like to catch a few minutes of it just to see what it was like you can check it out here. We plan to do more of these. The written word remains our primary medium. But there’s a lot more that even a small organization can do with live coverage of events through video. If you joined us, we’d love to hear your feedback.
Just before 2 a.m., I’m trying to make sense of the Senate results. If I’m making sense of this right, Democrats have clearly won 54 senate seats, appear likely to win 55 and could still win 56.
The final uncertain seat is in Mississippi. The Times calls it a toss-up with a 61% chance of Republicans winning it. But election watchers I trust seem to think it’s just a bit more likely that Colom will win. To be clear, they’re only talking about likelihoods, and it’s a bit more likely, in their view, that Colom will win. To be fair to the Times, these are models that are fully devised in advance. And this is extremely close. When it’s that down to the wire, it’s down to picayune details, idiosyncrasies of particular counties that may just be hard for a model to fully capture. The people I’m following suggest that in the key counties where votes remain the last ballots tend to lean Democratic.
Thank you to everyone who joined us for tonight’s livecast. We had about a thousand people watching through most of the evening from 8 PM to around 12:30 or so. We really enjoyed it and we hope you did too. We’re going to be doing more of these.
As I’ve noted a few times, polling is just a predictive artifact. None of all those polling questions matter now. But one thing is worth observing simply to think clearly about the unpredictability of polling error. It will take a while for numbers folks to crunch all the numbers. But simply look at the Senate races it’s clear that while some polls were accurate quite a number of them significantly underestimated Democratic strength – South Carolina, Texas, Iowa and a number of others. This isn’t an exhaustive list. I’ve been on our video-cast for almost five hours. I’m a bit fried. But again, polling error doesn’t just underestimate Republicans. It goes both ways. It’s unpredictable.
Something went wrong. Please refresh the page and/or try again.
The United States of America? “Over.” Christianity? “Gone.” Trump? A Democratic “sleeper cell.” English? We may not be allowed to speak it under the regime of “Woke 2.”
The Republican Party’s generational defeat in the midterms Tuesday inspired some all-time, classic crashouts among MAGA influencers and GOP operatives on X. There was infighting, finger-pointing, and, as the night wore on, tears.
