Milwaukee — Wisconsin’s Aug. 11 Democratic gubernatorial primary produced a near-upset that exposed limitations in pre-election polling and underscored the impact of late-campaign shifts. Every pre-primary poll since March had shown Francesca Hong with a commanding lead, often by 20 points or more, but unofficial returns put Milwaukee County Executive David Crowley narrowly ahead with about 39.8% to Hong’s approximately 39.3%.
Where the Marquette poll missed the mark
The Marquette University Law School poll, conducted shortly before the primary, showed Hong at 38% and Crowley at roughly 7%. Marquette poll director Charles Franklin said the survey accurately captured Hong’s base support and broad age and ideological splits among likely voters but did not predict a late unification of undecided voters behind Crowley.
"If you look at the margin between the poll and the result, then this is way, way off, because we had Crowley at 7%... But what the poll was not helpful for was knowing that those folks would all converge on one candidate: David Crowley," Franklin told the Milwaukee Journal Sentinel.
Gov. Tony Evers, in an interview with the Journal Sentinel, suggested the discrepancy reflected low engagement during much of the campaign rather than outright polling error. "There was nothing going on except for occasionally the opponents talking about something," Evers said. "It's not that I thought the polls were wrong. I just think there was so little accomplished in the first part of the campaign."
Late changes and an unplanned poll
The dynamics shifted in the closing weeks. Marquette’s second-to-last poll was carried out before Sara Rodriguez withdrew and before Crowley re-entered the race. Those developments prompted Marquette to run an additional, unplanned survey that included Crowley, but the new poll still substantially undercounted the support Crowley ultimately received.
Political observers say the episode highlights how quickly undecided voters can reshape a primary when the field changes and when voters seek an alternative late in the contest. The Marquette results appeared to capture the ideological and age divides within the Democratic electorate but failed to anticipate how undecided or late-deciding voters would coalesce.
Practical takeaways for campaigns and pollsters
- Small early leads can evaporate: A candidate favored in steady polling is still vulnerable if the electorate remains fluid.
- Polling after field changes matters: Last-minute entries or withdrawals can reshape decision-making among undecided voters.
- Undecideds are not monolithic: Their late choices can converge on one candidate and dramatically alter outcomes.
Analysts noted that the Marquette poll did correctly identify where Hong was weaker — among older and more moderate Democratic voters — but it did not show that those voters would rally around a single alternative. That aggregation of votes largely explains the gap between the poll’s snapshot and the actual returns.
Numbers at a glance
| Measure | Marquette Poll (pre-primary) | Unofficial Primary Result |
|---|---|---|
| Francesca Hong | 38% | ~39.3% |
| David Crowley | 7% | ~39.8% |
For Wisconsin voters and political operatives, the episode serves as a reminder that polls are point-in-time measurements, not predictions of inevitability. When races have low sustained engagement and a sizable undecided bloc, late developments — endorsements, withdrawals, or renewed campaigns — can have outsized effects.
The narrowness of the margin — with unofficial returns showing Crowley ahead by less than a percentage point — also means recount possibilities and post-election scrutiny of turnout patterns and vote totals could follow, depending on how the returns are certified. In the meantime, the result reshapes the Democratic primary picture in Wisconsin and raises fresh questions for pollsters ahead of the fall campaign.
As Wisconsin moves toward the general election, campaigns, pollsters and voters will be watching whether this round of polling misses prompts methodological changes, more frequent re-surveys after field shifts, or greater emphasis on tracking undecided voters and late-deciding cohorts.
