Ann Selzer is stepping back from polling elections after conducting the Iowa Poll for the Des Moines Register for the last 27 years, she announced on November 17.
Top Iowa Republicans reacted by accusing Selzer of “skewing” her numbers or publishing “fake news polling”—all because her final Iowa Poll was way off the mark.
The attacks on Selzer—not from MAGA randos, but from the heart of the GOP establishment—reflect a broader Republican strategy to discredit mainstream media outlets like the Des Moines Register. They also validate unhinged behavior like President-elect Donald Trump’s call for Selzer to be criminally investigated.
ANATOMY OF A “BIG MISS”
No one disputes that this year’s final Iowa Poll was, in Selzer’s own words, “the biggest miss of my career.” Whereas the survey found Vice President Kamala Harris leading Trump by 47 percent to 44 percent among likely voters, Iowans who cast ballots favored Trump by 55.7 percent to 42.5 percent for the Democratic nominee.
Selzer’s analysis of what went wrong found “no likely single culprit” that could explain the 16-point discrepancy. Some might fault her methods, particularly her decision not to adjust results by partisanship or recalled vote from the last election. That makes her surveys more susceptible to response bias, since Democrats are now more likely than Republicans to participate in telephone polls. Weighting by recalled 2020 presidential vote would have produced a 6-point lead for Trump instead of a 3-point advantage for Harris.
Others have criticized Selzer’s use of random digit dialing (rather than registered voter lists) to find respondents, or her decision not to weight by education. Those longstanding features of her methodology have often served her well—including in 2016 and 2020—but not this year.
For Iowa Republicans in the age of Trump, a poll can’t be wrong due to bad luck or honest mistakes. It must be a malicious plot.
ERNST MOCKS “FAKE NEWS POLLING”
Hours after the Register announced Selzer’s departure on November 17, Senator Joni Ernst crowed, “Donald J. Trump’s Election Day results put Ann Selzer’s fake news polling to shame.”
“Fake news” is one of Trump’s favorite attack lines. He doesn’t mean the media got something wrong; he insinuates the journalist lied on purpose.
In the real world, nothing is more valuable to non-partisan pollsters and news organizations than an accurate survey.
Ernst of all people should know Selzer’s not in the “fake news” business. The Iowa Poll was the first to show her with a significant lead over Democrat Bruce Braley in the 2014 U.S. Senate race. Selzer’s survey from September of that year found Ernst 6 points ahead, and the final pre-election poll showed the Republican up by 7. Those surveys looked like outliers at the time, but Ernst ended up beating Braley by about 8.3 points.
Maybe Iowa’s junior senator hoped taking a swipe at Selzer would help her get back into Trump’s good graces. Since being snubbed at the Republican National Convention, she has tried to ingratiate herself by campaigning for Trump around the country, going to Mar-a-Lago for election night, and preparing a report for the president-elect’s so-called “Department of Government Efficiency.”
If Ernst’s cheap shot at Selzer had been a one-off, I wouldn’t bother writing about it.
But there was a coordinated effort to impugn Selzer’s integrity, led by the man who’s been running the Republican Party of Iowa for a decade.
“THE JIG IS FINALLY UP”
Iowa GOP state chair Jeff Kaufmann has aggressively pushed the narrative that Selzer has manufactured findings in order to sabotage Republicans. Before the election, he accused her and the Register of “spreading propaganda.” After Trump’s lopsided victory, he wasn’t satisfied with posting taunts like a meme of himself smiling at the Iowa Poll’s freshly-dug gravesite.
In a guest column for the Washington Examiner, Kaufmann declared, “For too long, we have let Ann Selzer use her polls to influence races.” Using cherry-picked examples, Kaufmann claimed “Ann Selzer has a trend in her polls — and it’s to boost Democrats.” He repeated the charge at X/Twitter on November 19: “No one should trust the Des Moines Register polling again. This is not a rare occurrence. It is typical for this hyper-partisan paper to poll favorably for Democrats.”
As Kaufmann surely remembers, Selzer was the only pollster to find a comfortable lead for Ernst at the end of the 2014 campaign. And that’s not the only time her pre-election survey looked better for Republicans than than the polling average.
Going into the 2016 election, most published surveys predicted a close presidential race in Iowa. In contrast, Selzer’s final poll had Trump leading Hillary Clinton by 6 points. He ended up winning by 9. That same Iowa Poll indicated that Senator Chuck Grassley was 23 points ahead of Democratic opponent Patty Judge. Grassley was re-elected in 2016 by a margin of 24.4 points.
Seven of the last eight public polls before the 2020 presidential election found another tight race in Iowa, ranging from a 3-point lead for Joe Biden to a 2-point lead for Trump. Then there was Selzer, who had Trump ahead by 7 points—by far the closest to his 8-point victory.
Instead of giving Selzer credit for accuracy in 2020, Kaufmann’s op-ed accused her of “manufacturing fake support for a Democratic candidate” that year. His evidence: Iowa Polls from September and October showed Democrat Theresa Greenfield ahead of Ernst. (So did many other surveys at the time.) Selzer’s final pre-election poll showed Ernst leading Greenfield by 4 points—which wasn’t far off from the senator’s winning margin of 6.6 points.
The last Iowa Poll from the 2022 cycle showed Grassley leading Democratic challenger Mike Franken by 12 points, which was dead on.
Nate Silver, the founder of FiveThirtyEight.com, gave Selzer an A+ rating for years. When he reviewed 54 of her Iowa Polls this year, he calculated a “mean-reverted bias” of 0.1 percent toward Democrats. Silver’s analysis found every other pollster had a larger lean toward one of the parties.
None of it mattered to Kaufmann. After the Des Moines Register announced Selzer was closing the book on election polling, the Iowa GOP chair claimed she was lying when she said she’d decided more than a year ago not to renew her contract. And he again suggested she had used her work to promote one side: “Another miss from Selzer! After years of skewing polls in favor of the Democrats, the jig is finally up.”
Kaufmann wasn’t just speaking for himself. He was expressing the official Iowa GOP position. The state party’s X/Twitter account reposted his attacks, as did prominent Republicans like Iowa’s RNC committeeman Steve Scheffler. This storyline dovetails with longstanding efforts by Republican strategists and operatives to denigrate legacy media in Iowa, including the Des Moines Register and Cedar Rapids Gazette.
Naturally, Trump took the hostile rhetoric to another level.
“AN INVESTIGATION IS FULLY CALLED FOR!”
Trump has repeatedly described journalists as “enemies of the people,” and he’s not ready to let bygones be bygones when it comes to this year’s Iowa Poll. When he shared the news of Selzer’s departure with his Truth Social followers, he claimed she “knew exactly what she was doing” when she released a “totally Fake poll that caused great distrust and uncertainty at a very critical time.”
Trump characterized the poll as “possible ELECTION FRAUD” by Selzer and the Register, adding, “An investigation is fully called for!”
Publishing a poll isn’t election fraud or any other kind of crime. Anyway, why would an inaccurate poll benefit Harris? It could just as easily hurt her campaign by making Democrats complacent and inspiring more Trump fans to come vote on election day.
Was the president-elect serious? His team didn’t walk it back. Dan Morrison reported for USA Today, “Asked if Trump was calling for a criminal investigation of the pollster, spokesman Steven Cheung replied: ‘President Trump was very clear in his Truth Social post.'”
Before and after November 5, legions of MAGA activists excoriated Selzer on social media, claiming she was dishonest or had been paid off to swing the election to Harris. Some called for the pollster to be jailed.
Republican Congressional candidate Kevin Virgil breezily declared the poll to be “propaganda,” even though he admitted he had never heard of Selzer before this month.
In an interview with Gray TV’s political director Dave Price, Selzer said the West Des Moines police recently advised her to be on guard because of threats circulating on social media. The warning prompted her to take “all available precautions” for her safety. It’s a reasonable concern, given how many others have faced death threats after being accused of election fraud.
Top Iowa Republicans could have said what everyone in this field knows: Selzer’s a professional whose last snapshot of the Iowa electorate was very wrong.
Instead they keep claiming the Iowa Poll is part of a deliberate media effort to undermine Republicans. Kaufmann’s likely goal is to generate more favorable news coverage for his side by “working the refs,” as GOP operatives around the country have done for decades.
But smearing Selzer crossed a line, especially now that the pollster and Iowa’s leading newspaper are on Trump’s very long list of enemies he wants to punish or prosecute.
11 Comments
But smearing Selzer crossed a line, especially now that the pollster and Iowa’s leading newspaper are on Trump’s very long list of enemies he wants to punish or prosecute.
The Republicans have become experts at locating and twisting wedge issues to rile the mob and ignite crude, malicious, often dangerous responses. It’s all a game.
And it’s not only bonafide Rs. When the WNBA player Brittney Griner was jailed in Russia, the rightwing racists/nationalists rained terror down on her in social media posts. Less for her “crime” and more that she was a public figure who used her celebrity to push back on racism and police brutality by staying in the locker room during the national anthem — plus (of course), that she is gay.
Kaufmann is a cut off the MAGA block. No appreciation of Selzer’s years of bringing the public’s opinion to Iowans via the Register. Or, as they say, polls are a snapshot not a movie. But to impugn Selzer’s integrity is shameful by people no longer capable of shame. Ernst is a small-time, small minded shyster who uses her position as a U.S Senator to leverage her career, which means sitting next to J.D. on the campaign plane and knocking a few doors in Pennsylvania — and chiming in when it pleases the king. Rumor was she wanted to be Secy of Defense. Ironically, she’s as (un)qualified as Hegseth, except it’s testosterone that flows in his veins.
Gerald Ott Tue 19 Nov 2:35 PM
Selzer Poll an embarrassment to this Iowan
Selzer Poll made Iowa the butt of many jokes. DMR downward spiral continues. I had already voted for Kennedy prior to the poll release. Her poll didn’t pass the “smell test.” What a way for someone to end their career.
ModerateDem Tue 19 Nov 3:40 PM
Thank you, Laura
The vilification being hurled at Ann Selzer is both irrational and disgusting. But as the old saying goes, consider the source. In this case, sources.
PrairieFan Tue 19 Nov 3:49 PM
Laura missed an important point
Every serious poll comes with an estimate of the margin of error. The last poll of Seltzer found Harris with 47% of the votes and Trump with 44% of the votes, with a margin of error of plus or minus 3.4 percentage points. Now, hold on, we’ll do a bit of maths. To compute the probability of the election result given the poll result, we need to assume a sampling distribution for the poll’s estimate, which is typically a normal distribution centered at the poll’s reported value (44%) with a standard deviation (95%, unless mentioned otherwise) derived from the margin of error. Given these figures and the mathematical principles of statistics, the probability that Trump’s true level of support is 55.7% or more is approximately 7.67 × 10⁻¹², which is extremely close to zero.
Pardon the technicalities, but statistics are maths.
In short, in layman’s terms, Seltzer’s poll results are way too far off from the election results to be attributed to bad luck in a properly executed poll. That is why questioning the professionalism or the ethics of the pollster is warranted.
By omitting the mention of the margin of error, the article by Laura is misleading.
Karl M Tue 19 Nov 5:20 PM
No need to "pile on"
The Selzer/Iowa Poll was the most outrageous thing to come out of Iowa in years. At this point with the election over there is certainly no need to “pile on” and President Trump should just let it go.
union50702 Tue 19 Nov 7:25 PM
Margin of Error; Level of Mendacity
Good point, Karl, about margin of error. But I think the point is that MOE is based on assumptions just as much as it is based on statistics.
If 800 random people are sampled the result should match reality as closely as the MOE suggests but will only do so 95% of the time. Maybe Selzer just managed to fall into the 5% error slot.
Or maybe the 800 people were not really random. Maybe some voters are more inclined to own up to their choices than others or more inclined to spend several minutes answering questions.
There is no need to bash a pollster as successful as Selzer has been.
Also, . . . .what Iowa politician would not know who Selzer is? Where has Virgil been all these years?
IowaVoter Tue 19 Nov 8:12 PM
The bigger question
The bigger question is, why is the President-elect going after a pollster, even one who failed to estimate a margin of error. Trump has so many important things to do, and yet he wants a pollster to be prosecuted. Like today, he nominated Dr Oz as Head of Medicaid and Medicare. I am not making this up.
Karl M Tue 19 Nov 8:39 PM
Karl, even with perfect sampling
1 out of every 20 polls will be off by outside the margin of error, just by chance.
The margin of error is a 95 percent confidence interval, assuming a perfectly representative sample.
Clearly the last Selzer poll did NOT have a representative sample, because there were way too many Biden 2020 voters.
Laura Belin Wed 20 Nov 5:51 PM
Laura, there is no perfect sampling
Statistics rely on random sampling, which means that any Iowa voter had an equal chance of being selected for Seltzer’s survey. A perfect sampling is slightly different, it would mean that the sample reflects the Iowa voters’ population in all characteristics – which is practically impossible to do
Your conclusion is correct: Selzer’s poll did not have a random sample of the Iowa voters, there was bias in the sample. In that case, it is often impossible to estimate the margin of error.
One thing I wish I knew, is how many calls Selzer had to place in order to get one person answering the phone. Maybe it is time to acknowledge that cold calling people is no longer a way to obtain a random sampling. There are probably entire generations who never answer cold calls.
Karl M Wed 20 Nov 8:25 PM
Spam calls
Karl is probably right. People don’t answer cold call any more. I get so many spam calls, I click on the call, but when there is a delay or other indication that it is a spam call, I hang up. I know others who will not answer a call from a number they don’t recognize, because of so many spam calls.
I think election polling should stop: first, because of the inaccuracies discussed in this post and the comments, a second, because the drive me nuts.
Wally Taylor Thu 21 Nov 9:41 AM
Less than one percent of poll calls are answered
Per the chief pollster of the New York Times, in 2022: “Who in the World Is Still Answering Pollsters’ Phone Calls?
Response rates suggest the “death of telephone polling” is getting closer.”
Karl M Thu 21 Nov 2:15 PM