Experts Question Why Polls Were So Wrong In The Wisconsin Primary

Democratic Socialist Francesca Hong entered primary day as the prohibitive favorite to become the Democratic Party’s nominee for Wisconsin governor.

Yet, she wound up narrowly losing to Milwaukee County Executive David Crowley instead.

Crowley beat Hong by 0.4 percentage points, a margin of about 3,000 votes, in a race that polling had shown Hong leading by about 20 points for weeks.

Despite the slim margin, Hong ultimately conceded the race, and now Crowley will face Republican Congressman Tom Tiffany in November.

The gap between expectation and outcome sent pollsters, analysts, and commentators searching for an explanation, and their answers vary.

Marquette Law School polled the race twice, catching it at two different stages of collapse. The first, taken July 8-16 while Crowley sat out, had Hong at 26%, Barnes at 15%, and Lt. Gov. Sara Rodriguez at 11%. Rodriguez dropped out the next day. Crowley climbed back in July 18, and by the time Marquette polled again July 22-27, Hong had jumped to 38%, Barnes had ticked up to 16%, and Crowley had 7%. Barnes quit three days after that.

Both polls had small samples, margins of error above four points, and undecided voters in the 30s and 40s. Neither one could have caught what came next. Hong's 2021 posts calling for canceling Thanksgiving resurfaced in early August. She defended the posts on CNN, then reversed course on Politico, calling Thanksgiving her favorite holiday. The reversal revived old criticism of her past comments on her own "proximity to whiteness."

By August 3-6, when State Navigate polled a settled field, Hong still led, 44% to Crowley's 22%, with Sen. Kelda Roys at 11%. Kalshi had her as a 95.5% favorite as late as the Monday before the primary. By Tuesday afternoon, that number had fallen to 58%. Hong finished at 39.4%, close to what the late polls predicted. Crowley was the miss. The polls undercounted him by a wide margin. Roys took 7.5%, and former state official Joel Brennan took 4.7%, enough combined that their absence could have changed who won.

Pollster Nate Silver posted his confusion Tuesday night on X. 

"Hong might come back, but wtf is up with these primary polls?" he wrote.

Silver followed up, noting that many modern polls function as statistical models rather than raw snapshots, writing that "primaries are very hard to model from first principles, whereas general elections are relatively easy given very high partisanship and usually predictable demographic relationships."

Political scientist Larry Sabato pointed to a similar failure in Michigan days earlier, where Democratic Socialist Abdul El-Sayed defeated Rep. Haley Stevens in the Senate primary. Sabato wrote, "What amazes me is that just a few days ago, the same polling train wreck happened in Michigan. Few learned anything."

Pollster Frank Luntz said he agreed with Sabato's assessment.

CNN's Harry Enten suggested pollsters oversampled young progressive voters, inflating apparent support for socialist-aligned candidates in both states. 

“When you kind of look at the polling, you go, ‘Hey, wait a minute, maybe these polls had too many younger folks in them.’ And I think that speaks to something larger potentially, which is all these pollings, which are sewing progressivism on the left, the far left rising, seeing Democratic socialism, especially in the case of Wisconsin, rising, you got to look at these results and say, ‘Hey, wait a minute. Maybe, Johnny, maybe these polls are overestimating how much that Democratic socialism is rising within the Democratic Party,’ because in two key statewide races in a row, what we've seen is the lefties being vastly overestimated. And in the case of Wisconsin, the Democratic socialists being vastly overestimated by 20 points.”

The polls may have given Francesca Hong a coronation, but Wisconsin voters had other plans.

David Crowley’s comeback underscores the cost of allowing fashionable ideology, activist energy, and faulty assumptions to replace a serious reading of the electorate, and once again, the polling industry has egg on its face.