The five late governor polls examined here showed Rob Sand ahead. The Senate surveys disagreed. Against the returns available Wednesday morning, several surveys captured the direction of the election, while others understated Democratic support or put the Senate candidates in the wrong order.
That is a more useful conclusion than declaring polling vindicated after 2024. This year’s early comparison looks much better than the extraordinary miss in the final 2024 Iowa Poll. But other polls in 2024 were closer, and the final 2022 Iowa Poll was close to the results in both statewide races examined here.
What “right” means
A survey can identify the eventual winner yet miss the size of the victory. It can also show a tiny lead for the eventual loser without establishing that either candidate had a secure advantage. We compare the published Democratic-minus-Republican margin with the election margin for the same race. Positive differences mean the poll was more Democratic than the result; negative differences mean it was more Republican. Table differences are expressed in percentage points (pp). We make no additional allocation of remaining undecided responses; each poll’s published treatment of leaners is retained.
This is a preliminary audit, not a final scorecard. The Associated Press candidate totals displayed by Iowa Public Radio, checked at 10:24 a.m. Central on November 4, put Sand ahead by 10.98 points and Josh Turek ahead by 4.11. Both widgets displayed “Updated Nov 4, 2026 at 8:47 AM” without a time-zone label. AP estimated 99% of votes counted; that estimate is not certification. These are unofficial listed-candidate denominators. Write-in totals are unavailable, and completeness has not been independently established. Additional counting or corrections can change the comparisons. Sources: governor returns and Senate returns.
Governor: the direction was clear; the margin varied
All five likely-voter governor surveys reviewed here showed Sand ahead. AtlasIntel/PollsterGraph’s 54.1% to 44.0% was closest to the current margin. YouGov and NYT/Siena also showed clear leads, while Rasmussen and co/efficient each showed Sand ahead by five points, roughly six points smaller than his current advantage.
| Poll | D / R shares | Poll margin | Difference from current margin |
|---|---|---|---|
| AtlasIntel / PollsterGraph Oct. 28–Nov. 1 | 54.1% / 44% | D +10.1 | -0.9 pp |
| YouGov Oct. 26–29 | 53% / 46% | D +7.0 | -4.0 pp |
| NYT/Siena Field dates not independently verified | 51% / 43% | D +8.0 | -3.0 pp |
| Rasmussen Reports Oct. 27–31 | 50% / 45% | D +5.0 | -6.0 pp |
| co/efficient Oct. 28 | 50% / 45% | D +5.0 | -6.0 pp |
These polls told a broadly consistent story about who was ahead, but “they got the winner right” would conceal meaningful differences in their estimates of the race. Rasmussen’s governor figures are separately published in its governor toplines.
Senate: divided signals, not a clean sweep
The six Senate reports divide evenly on which candidate led. AtlasIntel, YouGov and NYT/Siena put Turek ahead. Rasmussen, co/efficient and GrayHouse put Ashley Hinson ahead. AtlasIntel’s full-response margin of 4.1 points is almost identical to the current 4.11-point lead. YouGov and Siena showed a one-point Democratic advantage. The other three surveys were about five to six points more Republican than the current returns.
| Poll | D / R shares | Poll margin | Difference from current margin |
|---|---|---|---|
| AtlasIntel / PollsterGraph Oct. 28–Nov. 1 | 50.5% / 46.4% | D +4.1 | 0.0 pp |
| YouGov Oct. 26–29 | 50% / 49% | D +1.0 | -3.1 pp |
| NYT/Siena Field dates not independently verified | 48% / 47% | D +1.0 | -3.1 pp |
| Rasmussen Reports Oct. 27–31 | 46% / 47% | R +1.0 | -5.1 pp |
| co/efficient Oct. 28 | 46% / 48% | R +2.0 | -6.1 pp |
| GrayHouse Oct. 28–31 | 46% / 48% | R +2.0 | -6.1 pp |
The displayed 0.0-point AtlasIntel difference is rounded; the underlying Senate difference is about −0.01 point. A one- or two-point polling lead should not be read as a confident winner prediction. Nor does AtlasIntel’s close match in one race establish that its approach will outperform other methods in future elections.
2024: a major miss, but not every poll missed alike
The final Des Moines Register/Mediacom Iowa Poll, conducted by Selzer & Co., put Kamala Harris ahead of Donald Trump, 47% to 44%. Trump instead won by 13.21 points using all valid candidate votes and write-ins in the certified statewide canvass. That is a 16.21-point Democratic overstatement of the margin. The original Iowa Poll report remains the source of the published 47–44 result.
Yet Emerson’s final Iowa survey showed Trump ahead 53% to 43%. Its raw margin was 3.21 points more Democratic than the certified result, substantially closer and on the correct side. We use its initial responses, not its separate allocation of undecided voters. One spectacular miss should not stand in for every survey conducted that year.
2022: a closer same-office comparison
Comparing 2026 governor and Senate polls with 2022 surveys of those same offices is more direct than comparing them only with a presidential race, though the candidates and electorate still differ. The final 2022 Iowa Poll showed Kim Reynolds ahead of Deidre DeJear 54% to 37%, and Chuck Grassley ahead of Mike Franken 53% to 41%, as reported by KCRG.
Reynolds’s certified margin was 18.52 points, versus 17 in that poll. Grassley’s was 12.18, versus 12. Those are Democratic margin overstatements of 1.52 and 0.18 points. Sources: the legislative governor canvass, printed pages 99–101, and the Senate canvass.
A second example prevents the 2022 comparison from resting on one firm. The final Cygnal survey for Iowans for Tax Relief Foundation was fielded October 26–27, 2022, and showed Reynolds ahead by 19 points and Grassley by 11.2. Its margin differences were 0.48 points toward Republicans for governor and 0.98 toward Democrats for Senate. Its governor question named only Reynolds and DeJear, while the Iowa Poll included Libertarian Rick Stewart, so the questionnaires were not identical.
What this comparison cannot tell us
These are individual late surveys, not a polling average or an exhaustive census. The 2026 selection includes the six likely-voter reports already documented in Iowa Decides’ late pre-election coverage. It is a review of that set, not every poll conducted before the election. An additional GOTV report is excluded because unresolved version and methodology questions prevent a reliable comparable pre-election estimate. Its published governor topline favored Lahn, so the five-poll agreement must not be generalized to all polling. Field dates differ: these surveys measured opinion at different points before Election Day, not identical election-day snapshots. This does not establish an industry-wide improvement, a national pattern or a causal explanation for a miss.
Turnout modeling, who answers surveys, weighting, late decisions and ordinary sampling variation can all contribute to a gap. The toplines alone do not tell us which factor mattered. A published margin of error generally describes uncertainty in a candidate’s share, not the uncertainty in the difference between two candidates, and it does not cover every source of polling error. See AAPOR’s guidance on polling accuracy.
The cautious lesson is practical: the late surveys were useful evidence of Sand’s advantage and a competitive Senate race. Their precision varied. The record from 2022 and 2024 likewise contains both close estimates and misses. Polls are better read as uncertain measurements than as promises of an election result.
Methods and source notes
- All 2026 reports describe likely-voter samples. AtlasIntel: October 28–November 1, n=1,219, online recruitment, reported ±3 points. YouGov: October 26–29, n=2,020, text invitations to online interviews, ±3.1 points, leaners included. Rasmussen: October 27–31, n=901, phone and online, ±3 points. co/efficient: October 28, n=1,020, mobile text and landline, ±3.05 points. GrayHouse: October 28–31, n=601, text-to-web and live telephone, ±4 points.
- Siena’s November 1 announcement directly confirms the reported shares and a ±4.7-point margin of sampling error, including a design effect of 1.45. Full Iowa response categories, sample size and fieldwork details remain unverified here. Its row is an announcement-level comparison, not a complete survey audit.
- AtlasIntel funding disclosure and YouGov commissioning sponsor/first-publication date remain unverified. Rasmussen’s Senate article has a dateline inconsistent with its fieldwork; the official October 27–31 toplines were documented before the election. These limitations previously kept some reports outside the site’s numerical polling charts. Including them here for transparent comparison does not remove those chart eligibility restrictions.
- We retain published rounding and response treatment. AtlasIntel uses its full-response table, not its valid-vote model; GrayHouse uses its initial ballot. Some rounded totals differ from 100. No reported margin is a winning probability. As a sensitivity check, recalculating every poll and result among only the two major-party candidates leaves the central conclusions unchanged: a split 2026 Senate picture, a very large Selzer miss in 2024 and much closer 2022 estimates. This adjustment is a separate calculation, not the pollsters’ published result.
- 2022 and 2024 result denominators include valid write-ins but exclude overvotes and undervotes. The 2026 denominators include only the candidates listed by AP. Consequently, current-to-historical comparisons are provisional, not a perfectly harmonized final ranking. The 2024 certified margin may differ slightly from reports that omit write-ins.
- Historical examples are illustrative: final Selzer and Emerson in 2024, final Selzer and Cygnal in 2022. Original Selzer pages were not directly accessible during this review; the 2024 shares were corroborated in indexed syndicated reporting, and the 2022 shares in KCRG. Emerson, Cygnal and certified results were checked directly.