Programming note! Our final podcast before the election is today, Oct. 29, 2026. David and I will take all your questions about the midterms live at 2:00 PM Eastern. Please join us! Paying members can ask questions in the chat, though all readers can follow along. Link to the stream is here.
And be sure to log on to the Strength In Numbers website or the Substack app on Nov. 3 at 6:45 PM ET to join us for our election night livestream with the teams from The Downballot and FiftyPlusOne on Substack Live!
Just click below to add a reminder to your calendar:
Apple | Google | Office 365 | Outlook | Outlook.com | Yahoo
At this point, you must be getting whiplash from all the different takes speculating about what type of voters the pollsters are missing this year. Liberals swear that forcing samples to represent the results of the 2024 election undercuts Democrats, much as polls missed in 2025 in New Jersey and Virginia. The president and speaker of the House of Representatives, meanwhile, have launched an apparent campaign to discredit the polls, and the chairman of the Federal Communications Commission has called them tools of voter suppression and disenfranchisement. Some Republican strategists have been arguing for months that the polls are undercutting them.
After the misses of the last decade — which were small in absolute size, but mattered a lot in close races — can we even trust the polls?
For my part here at Strength In Numbers, I have written several different investigations into the polls. I wrote in September about why polls might be underestimating Democrats — and then, this month, why they might be overestimating them. Then last Thursday, I showed again that polls may be understating Democrats because of a particularly wonky choice pollsters are making about how to make their polls politically representative of the electorate. And yesterday, in a final installment in this “how could the polls be wrong?” series, I showed that polls might not have enough Trump voters from red zip codes, which would inflate support for Democrats on the national generic ballot by about a point.
So which is it?
Well, unfortunately, if you are after a concrete answer, I can offer little in the way of a prediction about the direction of error. Instead, my focus has been on all the ways the polls could be off, each with generally a small impact on national toplines. The goal of my work has been to inform you about how pollsters generate their numbers — the assumptions inherent in the “data-generating process” behind the toplines you see on your social media feeds and the news.
So in this short post today, I want to give you two simple charts that summarize all my work on potential polling error this cycle. One chart lists the possible ways the polls could go awry, and another — for the nerds who want numbers — shows those factors alongside my estimate for how much each could impact national polls.
This article is free to read if you register for a free subscription to Strength In Numbers. While I paywall a few stories a week for paying supporters of this site, I also send 2 free emails a week, plus our weekly podcast, to everyone on the email list.