by Dan Burns

Aug 16, 2026, 6:30 AM

How did the polls miss so badly on Flanagan/Craig?

I’m an early to bed early to rise type, and while I’m neither wealthy nor wise I do enjoy excellent health. I went to bed the night of the primary election, before any results started being reported in Minnesota, kind of confident that Lt. Gov. Peggy Flanagan, for whom I’d voted earlier that day, would win, but by no means highly confident. So when I woke up around 2AM, as often happens, I had to get up and see before there was any chance I’d get back to sleep. Of course I was delighted that Peggy had absolutely crushed it, but also a little irked that I’d had to deal with nervousness during the day because of bad polling. (I was also pleased that Trina Swanson won here in MN-08. But mildly disappointed that we don’t get to enjoy the hilarity of Mike Lindell in the general.)

The polls for the Wisconsin race were off by roughly 20 points—about the same amount they were off in Minnesota (though in that case, they underestimated the progressive Flanagan), and in the Michigan race last week. Clearly, something is rotten in the state of cellphone autodialers. Polling primaries was tough business even 30 years ago when everyone had a landline and calls from strangers were not 95 percent spam, because one has to gather data and then estimate how many people are actually going to turn out. That can vary wildly depending on the state, the party, whether it is a competitive race, the national environment, and so on.

Incidentally, the vaunted prediction markets also completely flubbed this one. They couldn’t even sort out who was likely to win when votes started coming in and the political pros could guess with increasing accuracy who was likely to win based on where the remaining votes were located. They were quite obviously just swinging wildly back and forth with the headlines. It turns out markets are not some infallible distillers of Truth from the celestial ether.
(The American Prospect)

Actually, that remark about Minnesota isn’t entirely correct. Two SurveyUSA polls for KSTP showed the race (supposedly) being tight, but a PPP poll for the Democratic Lieutenant Governors Association, which backed Flanagan, had her up by 14. So we’re really only talking two polls from the same pollster that stunk it up.

My initial thought was that here in Minnesota the progressive equivalent of Trump’s “hidden voters,” so-called because they’re not the type to ever bother with responding to pollsters on the phone or online, and who don’t show up for every election, had a lot to do with it. But that doesn’t explain Wisconsin and Michigan. Also, Trump’s hidden voters are only a few percent, just enough to have put him over the top in 2016 and 2024. The misses here are way higher than that.

I think the Prospect article is pretty much right when it goes on to note that overall turnout was much higher than the pollsters probably expected. As a general rule they tend to try to play it safe, especially these days when it’s clear just how hard, if not impossible, really representative samples are to get. We’ll see how they do in November.