How David Crowley won the undecided voters that Wisconsin media pundits treated as an afterthought

David Crowley did not “beat the polls” so much as he won the part of the electorate that the news coverage treated as background noise. That news media error was not only mathematical. It was narrative.

Crowley won because the race was volatile, late-deciding, and structurally unfinished until the final stretch. Francesca Hong’s support was real. Marquette’s final poll had her at 38%, and she finished around 39%.

The miss was Crowley. Marquette had him at 7% in that final survey, while 34% of Democratic primary voters were still undecided about two and a half weeks before the primary. Charles Franklin’s post-election explanation was blunt: Hong’s number held, but undecided voters moved heavily to Crowley.

That means the better question is not, “How did Hong collapse?” It is, “Why did so many analysts assume the undecided vote would behave passively?”

HOW CROWLEY WON

Crowley became the consolidation vehicle. After withdrawing July 8 and reentering July 18 with Governor Tony Evers’ backing, he occupied the practical “can win in November” lane at the exact moment many Democratic voters were being forced to reassess risk, electability, and party unity.

AP reported that Crowley rejoined the race less than a month before the primary, with Evers campaigning with him in the final week, and Crowley arguing that he was the stronger nominee against MAGA Republican U.S. Representative Tom Tiffany.

The departure of Mandela Barnes from the race mattered because his exit released voters who were not automatically Hong voters. WisPolitics reported that after Barnes left the race, Crowley and Hong both appealed to his supporters, and one poll found Barnes’ voters almost evenly split on second choice: 27% for Crowley and 26% for Hong.

That does not mean Barnes delivered Crowley the win. But it supports the view that Barnes’ departure did not simply clear the field for Hong. It reopened the field for voters who were progressive, pragmatic, Black, Milwaukee-connected, union-oriented, anti-Tiffany, or simply uncertain.

Hong also had a ceiling problem that most news coverage understated. In the July 22-27 Marquette poll, Hong dominated very liberal voters but was much weaker among moderate and conservative Democratic primary voters.

Among moderate and conservative primary voters, she had 19% first-choice support, while 42% were undecided. Even with leaners included, she had 27%, while 16% remained undecided. That warning sign was mostly ignored by political pundits. A candidate can lead a fractured primary and still be vulnerable if the remaining undecided electorate is ideologically different from the committed base.

The general-election argument also mattered. Earlier Marquette polling found Hong trailing Tiffany by 3 points among registered voters and tied with him among likely voters. The same survey showed Barnes running better against Tiffany, while Crowley was not tested because he had dropped out when the poll was conducted.

That left a vacuum. Once Crowley reentered, voters who were worried about November had no fresh public head-to-head data. So the race became susceptible to cues from Evers, other Democratic figures, late spending, and negative attention around Hong.

WHERE THE MEDIA FAILED

The first failure was treating a poll as a verdict instead of a measurement. Polls are snapshots with margins of error, field dates, undecided voters, turnout assumptions, and candidate lists that may become obsolete.

The July Marquette poll that showed Hong leading was conducted in a campaign that had already been scrambled by withdrawals and reentries. One earlier Marquette survey did not include Crowley because he had dropped out before reentering after the survey was completed. News coverage that treated Hong’s lead as destiny converted voter uncertainty into election certainty.

The second failure was ignoring what the poll itself was saying. A race with 34% undecided voters is not a settled race. It is a race waiting for a late sorting event. The fact that Hong’s final vote share closely matched Marquette’s final number should have tempered the “polls were completely wrong” narrative after the election. The larger miss was the allocation of undecided voters, especially after Barnes’ exit, Evers’ intervention, and the final electability push.

The third failure was confusing “candidate strength” with “media salience.” Hong was the most interesting story. She was a democratic socialist, younger, had progressive energy, national ideological implications, possible history-making candidacy, and a sharp contrast with Tiffany.

Crowley’s story was less well defined from a TV news packaging standpoint. He left the race, returned, needed consolidation, and depended on voters making a late practical choice. The media system tends to reward the cleaner narrative, even when an evolving narrative is more electorally important.

The fourth failure was overfitting national templates. Some coverage framed Wisconsin as another test of progressive momentum after other state races. That was not irrelevant, but it risked making Wisconsin voters secondary characters in a national Democratic Party drama.

AP described the race as part of a broader struggle between progressive and moderate Democrats, but the actual outcome also turned on Wisconsin-specific factors. Those included Evers’ role, Milwaukee County, Tiffany’s profile, the open primary, candidate withdrawals, and late-deciding voters.

The fifth failure was post-election self-dramatization. The same system that overstated Hong’s inevitability now benefits from overstating the shock of her loss. “How did everyone miss this?” is content. It keeps pundits central to the story.

But the responsible correction is narrower and less dramatic. Hong’s support was real. Crowley’s late consolidation was undermeasured. Undecided voters were mishandled. And the Democratic primary campaign changed faster than the public narrative did.

As for coverage by the news media, there is also professional laziness in treating polls as content objects rather than evidence of conditions reflecting a moment in time. But the deeper problem is institutional.

Political media has built a business model around prediction, while maintaining the rhetorical escape hatch that “polls are only snapshots” when those predictions fail. That creates a predictable cycle. Inflate the horse race, imply certainty, act surprised by an unexpected outcome, then produce another round of analysis about the surprise.

Crowley won, in part, because undecided and displaced Democratic voters broke late toward the candidate who became the clearest anti-Tiffany, Evers-backed, general-election vehicle. The news media failed because it centered on the poll leader, not the unresolved Wisconsin electorate.