CNN’s Magic Wall map of US presidential results is seen on a mobile phone on November 7, 2024. (Beata Zawrzel/NurPhoto via Getty Images)

An oracle has gone badly wrong. The most impressive results were given by a little-known company from Brazil. A nagging problem has resurfaced, and some media critics have become profane in their assessments.

Such was the case with the 2024 presidential election polls. Their collective performance, while not stellar, it was improved from four years before. Overall, polls signaled a close result in the race between former President Donald Trump and Vice President Kamala Harris.

That’s what the election produced: a modest victory for Trump.

With votes still being counted in California and several other states more than a week after Election Day, Trump got it 50.1% of the popular vote to Harris’ 48.1%a difference of 2 points. That margin was closer than Joe Biden’s 4.5-point victory over Trump in 2020. It was closer than Hillary Clinton’s 2016 popular vote victorycloser to Barack Obama’s 2008 victories and 2012.

In addition, there were no errors among the national pollsters as dramatic as CNN’s estimate in 2020 that Biden led Trump by 12 points.

This time, CNN’s latest national poll said the race was at a standstill — a result anticipated by six other pollsters, according to the data compiled by RealClearPolitics.

The most striking discrepancy this year was The Marist College Surveymade for NPR and PBS. Harris was estimated to hold a 4-point lead nationally at the end of the campaign.

Iowa’s Big Miss “Oracle”.

In any case, a sense lingered among critics that the Trump-Harris election has resulted in yet another poll embarrassment, another entry in the catalog the failure of the poll in the presidential electionwhich is the subject of my latest book, “Lost in a Gallup.”

Comedian Jon Stewart gave harsh voice to such sentiments, saying of pollsters on his late-night show on election night: “I don’t want to hear from you ever again. Ever. … You don’t know shit and I don’t care about you.”

Megyn Kelly, a former Fox News host, also denounced pollsters, stating on her podcast the day after the election: “The poll is a lie. They don’t know anything.”

Two factors seemed to encourage such derision — a widely discussed poll of Iowa voters released the weekend before the election and Trump’s analysis of the seven states in which the result returned.

The Iowa poll injected shock and surprise late in the campaign, reporting that Harris had taken it a 3 point lead in the state over Trump. The result was likened to a “bomb” and them implications it seemed clear: If Harris opened up an advantage in a state with the partisan profile of Iowa, her prospects for winning elsewhere looked strong, especially in the Great Lakes swing states of Wisconsin, Michigan and Pennsylvania.

The survey was conducted for the Des Moines Register by J. Ann Selzera veteran Iowa pollster with an outstanding reputation for opinion research. One comment in The New York Times in mid-September, Republican pollster Kristen Soltis Anderson declared Selzer “the oracle of Iowa.” MSNBC’s Rachel Maddow praised Selzer’s pre-election polls for their “uncanny predictive accuracy.” Reviews published in June by data guru Nate Silver gave Selzer’s polls an A-plus grade.

But this time, Selzer’s survey missed dramatically.

Trump won Iowa by 13 points, which means the poll was off by 16 points — a stunning divergence for a perfect poll.

“Even the mighty were humbled” by Trump’s victory, Times of London said of Selzer’s survey failure.

Celery said afterwards she will “review data from multiple sources with the hope of finding out why this (discrepancy) occurred.”

It is possible, other pollsters have suggested, that Selzer’s reliance on telephone polling contributed to the poll’s failure. “Telephone polls alone … will not reach low-propensity voters or politically disillusioned non-white males,” Tom Lubbock and James Johnson wrote in a comment for The Wall Street Journal.

these days, few pollsters rely solely on the telephone to conduct electoral polls; many of them have opted for hybrid approaches which combines, for example, telephone, text and online sampling techniques.

The surprise sweep of cradle states

Trump’s sweep of the seven hotly contested states certainly contributed to perception that the polls failed again.

Conformable RealClearPoliticsHarris had slim leads late in the campaign in Michigan and Wisconsin, while Trump was slightly ahead in Arizona, Georgia, Pennsylvania, North Carolina and Nevada.

Trump won them all, a result no poll had predicted — except AtlasIntel of Sao Paulo, Brazil, a “little-known” firm as the New Republic noted.

AtlasIntel estimated that Trump was ahead in all seven swing states by a margin that closely matched the vote results. In none of the swing states did the AtlasIntel polls deviate from the final vote by more than 1.3 points, an impressive feat.

AtlasIntel did not respond to email requests we sent asking for information about its background and survey technique. The company describes himself as “a leading innovator in online surveys” and says it uses “a proprietary methodology,” without revealing much about it.

The founder and chief executive is Andrei Romanwho earned a doctorate in government from Harvard University. Roman turned to X, formerly Twitter, after the elections to post a graphic that promoted AtlasIntel is “the most accurate poll of the US presidential election.”

There was an outpouring of praise from the surveyors recalling a kind that appeared periodically since the 1940s. Then it was polling pioneer George Gallup placed two-page advertisements in the journalism publication “Editor & Publisher” to assert the accuracy of its presidential polls.

Again underestimating Trump’s support

A significant question facing pollsters this year—their great known unknown—was whether changes to sampling techniques would allow them to avoid underestimation Trump’s support as he had it in 2016 and 2020.

Misjudging Trump’s support is a nagging polling problem. The 2024 election results indicate that the shortage persists. By margins ranging from 0.9 points to 2.7 points, the polls underestimated Trump’s support in the seven swing states, for example.

Some polls have misjudged Trump’s support by an even larger margin. CNN, for example, underestimated Trump’s vote by 4.3 points North Carolinaby more than 6 points in Michigan and Wisconsin as well as Arizona.

Results that erred in the same direction suggest that adjustments to sampling methodologies were inadequate or ineffective for pollsters in trying to reach Trump supporters of any type.

This article is republished from conversation under a Creative Commons license. Read on original article.