Every time a trailing candidate becomes the winner after a batch of big-city absentee ballots reports late, a portion of the electorate assumes the fix was in — and Wisconsin, more than almost any other battleground state, has spent the last six years supplying the raw material for that assumption, usually without any actual evidence of wrongdoing behind it.
Key Points
- In Wisconsin’s August 2026 Democratic gubernatorial primary, a Milwaukee County USB-stick mix-up delayed roughly 28,000 absentee ballot results, temporarily distorting the reported margin between David Crowley and Francesca Hong.
- Officials attributed the delay to human error — five of nine memory sticks held audit files instead of election results — and re-downloaded the correct data before finalizing the count.
- Crowley won the certified count 65,131 to 62,739; Hong conceded after the Associated Press called the race near 3 a.m.
- No named source, audit, or court record has produced evidence that the error changed any vote total; the public record shows a procedural delay, not a demonstrated reversal.
- This episode fits a well-documented Wisconsin pattern — Madison’s uncounted ballots, Green Bay’s duplicate mailings, the 2020 recount — in which administrative mistakes, not tampering, drive election-night confusion.
How Wisconsin’s central-count system produces late-night surprises
Wisconsin is one of a shrinking number of states that allows municipalities to count absentee ballots at a single central location rather than at each polling place. Thirty-nine municipalities across fourteen counties, Milwaukee chief among them, use this method, which means absentee results and Election Day results simply do not arrive at the same time. State law compounds the delay by prohibiting any processing of absentee ballots — not just counting, but opening envelopes and running them through tabulators — before polls open on Election Day itself. In a jurisdiction like Milwaukee, where tens of thousands of absentee ballots stack up for same-day processing, that legal architecture guarantees a lag between the early Election Day returns and the fuller, more Democratic-leaning absentee tally that follows hours later.
That structural lag is precisely what produces the “red mirage, blue shift” pattern familiar from 2020 and repeated in subsequent cycles: an early lead built on in-person votes erodes or reverses once absentee batches from Milwaukee and Madison are folded in. It is a known, well-documented feature of Wisconsin election mechanics, not an anomaly invented to explain away a bad result. Understanding this mechanism matters because it supplies the innocent explanation for late shifts before anyone reaches for a more sinister one — the count moves late because the law requires it to move late, and because Milwaukee’s ballots are processed centrally rather than distributed across hundreds of small precincts.
What actually happened in Milwaukee County in August 2026
In the August 11, 2026 primary, David Crowley — the Milwaukee County executive who briefly dropped out of the governor’s race before Governor Tony Evers persuaded him back in — faced Francesca Hong, a Madison state representative and self-described democratic socialist, for the Democratic nomination. Milwaukee’s central count ran into trouble that night: election officials discovered that five of the nine USB memory sticks used to transfer tabulator results contained audit data rather than election results, because, as Milwaukee Election Commission director Paulina Gutierrez explained, someone had pushed the wrong button during the export process. Staff had to travel back to the counting location, re-download results from five machines, and return to the county office before the tally could be finalized, delaying roughly 28,000 absentee ballot results well past midnight.
The delay mattered because those absentee ballots leaned toward Crowley’s base in Milwaukee County, while Hong’s strength ran through Dane County and other college-town precincts. Early in the evening, with Milwaukee’s absentee batch still outstanding, some running tallies showed a tight or even Hong-favoring margin; once the corrected Milwaukee results were folded in, Crowley’s position solidified. The Associated Press called the race for Crowley near 3 a.m., with a final certified margin of 65,131 to 62,739, and Hong called Crowley to concede. Milwaukee County Clerk George Christensen acknowledged the mistake bluntly — “human error happens” — while stressing that the correction process itself was transparent and that the underlying tabulation was secure.
Weighing the “rigged count” narrative against the documented record
This is where the evidence has to be weighed honestly rather than assumed. The claim that Milwaukee’s late data reload “destroyed” Hong or unfairly manufactured Crowley’s win rests on a real, confirmed technical failure — the wrong files were on those USB sticks, and that fact is not in dispute. But the leap from “there was a processing error” to “the error changed the outcome” is not supported by anything in the public record. Officials on the ground, from both the city clerk’s office and the county election commission, stated plainly that there was no sign the counted ballots had been tampered with, and framed the re-download as an abundance-of-caution correction rather than a cover-up. Hong’s own concession, delivered directly to Crowley after the AP call, is itself a data point: the candidate with the most direct incentive to contest an unfair reversal did not do so.
At the same time, it is fair to note what the public record does not contain. No independent forensic audit, court finding, or released incident report has directly confirmed that the pre-correction and post-correction Milwaukee tallies were identical at the batch level — the assurances came from officials, not from a released technical reconciliation. That gap is real, but it is a gap in documentation, not a documented contradiction of the official account. Absent a specific, sourced finding that the totals changed improperly, the stronger and better-evidenced position is that this was what Wisconsin election administrators have called it repeatedly over the past several cycles: a costly, embarrassing, entirely human mistake, corrected in view of the public before certification.
Live coverage of the Wisconsin Democratic primary abruptly ended as Milwaukee County officials revealed a tabulation error involving absentee ballot results with five of nine USB drives containing audit logs instead of vote totals
Officials blamed human error delaying the final… pic.twitter.com/sVhtMLkYqX
— NewYork-Insight (@NewYork_Insight) August 12, 2026
A recurring pattern, not an isolated scandal
Milwaukee’s USB mix-up did not happen in a vacuum. Wisconsin’s election administration has produced a string of comparable errors in recent cycles: Madison’s clerk’s office left nearly 200 ballots uncounted in the 2024 general election, an omission the bipartisan Wisconsin Elections Commission itself called a “profound failure”; Green Bay mailed duplicate absentee ballots to more than 150 voters ahead of a spring election, triggering a state probable-cause finding and the clerk’s resignation; and Milwaukee itself recounted roughly 30,000 to 34,000 ballots in November 2024 after tabulator doors were found improperly sealed, a recount officials again described as precautionary rather than corrective of any fraud. Even the 2020 presidential recount, paid for by Trump’s campaign in Dane and Milwaukee counties, turned up human-error mistakes — such as 386 absentee ballots never opened — without altering the outcome.
The pattern that emerges is consistent: Wisconsin’s central-count system, its legal timing restrictions, and its reliance on manual data transfer between tabulators and county offices create real opportunities for administrative error, and those errors surface publicly almost every cycle. What has not surfaced, across any of these episodes, is documented proof that an error changed a certified outcome. For voters trying to judge the next late-night ballot drop headline, that history is the most useful guide available: treat the mechanical delay as expected, treat the official explanation with appropriate scrutiny, and reserve judgment about “stolen” results for cases where someone actually produces the missing audit trail rather than simply pointing to the hour on the clock.
Sources:
wisconsinwatch.org, thevotingnews.com, pbs.org, foxnews.com, politifact.com, nbcnews.com
