How Wrong Can the Polls Be?

The latest round of New York Times/Siena polls has a lot of Democrats feeling optimistic, and it’s easy to see why: Mary Peltola up 7 in Alaska, James Talarico up 6 in Texas, Sherrod Brown up 3 in Ohio, all on top of a September national poll with Democrats ahead by almost 9 on the generic ballot. To be honest, I’m excited too, but I want to highlight some caution, because polling has had some pretty big misses lately, and most of them have been in Republicans’ favor. We’ve been burned before.

So let’s see what a miss would actually do. The slider below applies one polling error to every race at once (which is roughly how national polling misses behave) using the Senate averages from the 2026 Election Tracker and the generic ballot for the House. The grey ticks are past years’ misses.

If the polls miss the way they did in 2024, Democrats STILL lose the Senate. Alaska, Michigan and Maine all flip, Texas comes down to a fraction of a point, and Democrats end up at 49, while still comfortably winning the House. The House feels safer. It doesn’t flip red until the miss reaches almost 6 points, which is bigger than any recent cycle. The Senate only needs about 2.

On the other hand, 2018 is probably the most directly comparable year we have: a blue-wave midterm under Trump. That year the polls actually underestimated Democrats by about a point. A repeat of that flips Iowa and puts Democrats at 53 seats (and Kansas surprisingly almost flips too), which is a Fetterman-proof majority.

Overall, since my Labor Day prediction post, things have been looking up for Democrats, and I’ll do an updated prediction closer to the election.