In Iowa State Senate District 35 — a seat Donald Trump carried by 21 points in 2024 — a new poll showed Democrat incumbent Mike Zimmer leading Republican challenger Joe Stutting by three points, 46% to 43%. The pollster who produced those numbers then did something pollsters almost never do: he told everyone why they were wrong.
Nuttycombe identified two problems buried in his own sample. First, as he put it, "LOTS of Rs unhappy with Trump are lying about their 2024 vote." Second, "Rural Iowa Rs ignored our survey." In a district Trump won by 21, the respondents who showed up didn't reflect the electorate that actually lives there. The people answering the phone skewed Democrat. The people hanging up skewed Republican.
So State Navigate ran the numbers again. This time they weighted the responses so that recalled 2024 votes matched the actual results — the ones where Trump dominated. The new numbers flipped the race entirely. Stutting led Zimmer 47% to 45%. In the governor's race, Democrat Rob Sand's comfortable 9-point lead evaporated into a dead heat: 45% to 45%.
A 5-point swing. Same respondents. Same questions. Different weighting assumptions. That's the distance between "Democrats are surging" and "Republicans are winning."
Nuttycombe acknowledged the core problem with unusual candor. "Republicans, particularly rural Republicans in Iowa, are very unresponsive to surveys, at least for now," he wrote. The implication is straightforward: if rural conservatives won't pick up the phone, every poll that doesn't account for that silence is sampling a fictional electorate.
State Navigate published both versions of the poll — the standard weighting and the adjusted version — precisely because the gap was too large to ignore. District 35 covers Clinton, Camanche, DeWitt, and Maquoketa in eastern Iowa. It flipped to Democrats in a January 28, 2025 special election, making it the first state legislative seat to change hands during Trump's second term. That flip made it a bellwether. The question is whether the bellwether was ringing true or ringing hollow.
Reuters published its own analysis identifying six reasons Democrats may fall short of a midterm landslide. Reason number five: "Polls have historically understated Republican support." The wire service noted the largest misses came when Trump was on the ballot — and while Trump isn't on the ballot this November, his voters are. Reuters also flagged that Republicans outspent Democrats by $188 million in Senate battleground advertising through the end of September, with Trump's MAGA Inc. super PAC holding $416 million in cash as of early September. The infrastructure to turn silent Republicans into actual voters is funded. The polls measuring them are not.
Echelon Insights analyst Patrick Ruffini has raised the same alarm about response bias distorting midterm surveys nationally. When one side's voters systematically avoid pollsters, every topline number inherits the distortion. It doesn't matter how sophisticated the methodology is if the raw data is missing the people who decide elections.
The rebuttal from polling defenders is predictable: weighting is standard practice, every survey adjusts for demographics, and State Navigate was being transparent by publishing both versions. Transparency about producing a misleading number is still producing a misleading number. The unweighted version — the one with the Democrat lead — is the one that gets the headline. The corrected version gets a footnote. Nuttycombe published both. Most pollsters publish one.
Ann Selzer's infamous 2024 Iowa poll showing Kamala Harris leading Trump by three points in a state he won by more than 13. That poll was off by 16 points. Selzer retired from political polling afterward. The methodology that produced a 16-point miss didn't retire with her.
The Reuters/Ipsos congressional ballot currently shows Democrats favored 43% to 35%. That 8-point gap is the number cable news will cite between now and November 3. State Navigate just demonstrated what happens when you re-weight a single Iowa district to match reality: a 3-point Democrat lead becomes a 2-point Republican lead.
Multiply that correction across 40 to 50 competitive House races and a handful of Senate contests where Democrats need a net gain of four seats for a 51-49 majority.
The only poll that matters is on Election Day. Nuttycombe said so himself.