An update from the editor: What a review of the pre-election Iowa Poll has found
Since election night, pollster J. Ann Selzer has examined the pre-election Iowa Poll’s methods and respondent sample for clues about why its results differed so widely from the actual vote.
The poll showed Vice President Kamala Harris leading by 3 percentage points. In the actual vote, President-elect Donald Trump won Iowa by 13 points.
Selzer’s review has taken the form of testing plausible theories against available data. To date, no likely single culprit has emerged to explain the wide disparity. But I wanted to walk you through what has been looked at and what relevant data shows.
Following the Register’s long practice, we already released the poll questionnaire. For transparency, we’re also releasing the poll’s full demographics, crosstabs and weighted and unweighted data, as well as a technical explanation from Selzer detailing her review.
What the review found
Here’s a brief summary of the theories tested and the findings.
- Were the demographics skewed? Selzer notes that this poll and other recent Iowa Polls have shown a higher response in the 1st Congressional District, in southeast Iowa, which leans more Democratic than the others. But the poll weights by congressional district, which would address that. Another common question she’s fielded: Did the poll include too many city dwellers, who tend to vote Democratic, and too few rural dwellers, who tend to vote Republican? In fact, the poll’s sample included 31% of likely voters who were city dwellers, a smaller percentage than in five comparable polls that identified likely 2024 voters, stretching back to 2021.
- Did the poll fail to detect the shift found nationally among men of color toward Trump? The Iowa Poll’s sample of Latino men (20) did go to Trump by a 4-1 margin. But the numbers are so small that they had little effect on the overall poll results.
- By ending interviews on Thursday, did the poll fail to capture late-deciders? Because exit polling was not conducted in Iowa, this is unknowable. Before publication, Selzer had flagged that Thursday was Trump’s best night. Per her usual practice, she calculated two-day rolling averages, but those looked flat.
- Was the poll’s weighting flawed? The Iowa Poll’s respondent base is built to give all Iowa adults an equal probability of being contacted by an interviewer. Then Census data is used to adjust demographic variances to ensure an appropriate cross-section of all Iowa adults. As Selzer explains in more detail, the weighting in this poll was minimal and yielded a final sample of likely voters that matched the percentage of the most recent poll.
- Should “recalled vote” from the 2020 presidential election have been included as a weighting factor? Selzer has not found voters’ recollections of their past votes to be highly reliable. The poll included more self-identified 2020 Biden voters than would seem to be proportionate. As part of her review, Selzer tested the recalled vote that respondents reported for this poll compared to the actual Iowa vote in 2020. The findings suggest a 6-point lead for Trump in 2024. So that may have played a role, but it’s far from Trump’s actual 13-point victory.
While no one factor has jumped out as determinative, it’s also possible that a combination of lesser factors played a role.
Other possible factors
Selzer also examined exit polling in neighboring Wisconsin, to see if the demographic patterns of turnout there differed significantly from the Iowa Poll response sample. She found some subgroups with differences but general alignment on party ID, age, sex and other breakdowns.
Some critics have accused the Iowa Poll of a Democratic bias. Statistician Nate Silver, who rates pollsters, did a calculation to determine historical bias toward one party or the other. Across 54 Iowa Polls, he found a negligible result, a 0.1% tilt toward Democrats, a smaller bias figure than for all but one of 25 top-rated polls.
Some factors are unknowable from the data in hand, such as whether the poll may have missed blocs of voters amid growing distrust of polling and the media. Courtney Kennedy, Pew Research Center’s vice president of methods and innovation, told Paste BN that she believes evidence supports a theory that Democrats participate in surveys more than Republicans and that weighted adjustments could help account for that.
What's next
Selzer has conducted the Iowa Poll on a contract basis for the Des Moines Register since 1997. For a couple of years now, she has discussed with me a timetable for winding down her work on the poll. She notified me over a year ago that she would not renew when her 2024 contract ended with publication of the pre-election poll.
“My invaluable assistant, Michelle Yeoman, and I have treasured our role in the Iowa Poll,” Selzer said. “My working relationship with the Register and its team of reporters and editors has always been fair, transparent and professional.”
When the Iowa Poll was founded on Dec. 12, 1943, editors said it was “inaugurated as a public service to the people of Iowa so that each citizen will know what the people of this state as a whole are thinking.”
In the decades since, the Iowa Poll has been a valuable resource for residents, advocates and lawmakers in shaping public policy in Iowa. Register editors are evaluating the best ways to continue surveys that will provide accurate information and insight about issues that matter to Iowans.
The Iowa Poll has been an important legacy indicator and we recognize the need to evolve and find new ways to accurately take the pulse of Iowans on state and national issues. Our intention is to reimagine how to best capture public sentiment and opinion among Iowans to provide valuable and accurate insight into legislative issues and politics.
Thank you for reading and for caring about the accuracy of information you consume.
(Unable to see the embedded documents above? Click here.)
Carol Hunter, executive editor, writes a weekly column for Des Moines Register readers.
This story has been updated to add video.