How Trump can turn the tables on the pollsters  

President Trump has a problem with pollsters. But rather than berate them as “negative criminals” and call for their investigation, the president should pursue a more rational, record-based case against them — if, that is, he is able to tone down the hyperbole and focus on the facts.

Trump assailed pollsters in bombastic terms following the release of surveys indicating his approval rating has plunged to levels seldom matched by a U.S. president three months into his term. Trump specifically attacked the New York Times-Siena College poll, which had reported that, by 54 percent to 42 percent, respondents disapprove of the president’s “handling” of his job. He also singled out for scorn the results reported in a Washington Post-ABC News-Ipsos survey that showed only 39 percent of respondents approve of his performance. 

There were plenty of other unfavorable assessments, too, released to mark the passing of 100 days of Trump’s second term.

CNN reported the president’s approval rating was underwater, at 41 percent to 59 percent disapproval, and declared the result “the lowest for any newly elected president at 100 days dating back at least to Dwight Eisenhower — including Trump’s own first term.” 

And a Marist College survey, conducted for PBS and NPR, reported 45 percent of respondents said Trump deserves an “F” for his job performance so far. About half that number (23 percent) said Trump merits an “A.” Just 42 percent overall approve of Trump’s job performance, according to Marist. 

Were he to choose a more focused, accountability-driven response, Trump could readily point to flawed poll results reported in the 2024 presidential campaign. Given that many of them erred in estimating the election outcome, he could pose a simple question: What makes them believable now?

Trump, in short, could argue pollsters have persistent credibility problems and point to vulnerabilities exposed in recent elections.

Traditionally, pre-election polls have been regarded as representing what George Gallup, a founding figure of survey research, called an “acid test” of the soundness of polling techniques. “The only justification of an election forecast,” Gallup once declared, “is to test polling methods.”

The late Philip Meyer, a well-regarded journalist, educator and past president of the American Association for Public Opinion Research, essentially agreed with Gallup’s characterization, noting: “A poll is an estimating device while an election is an exact measurement of the real world. By holding the device against the reality, we can learn how well the device is working.”

The reality revealed in the three most recent U.S. presidential elections was that pollsters were unable to measure with accuracy Trump’s popular backing. They collectively understated support for Trump in each election, despite having made methodological adjustments following the 2016 and 2020 campaigns — adjustments specifically intended to measure Trump’s popular backing more accurately.

Just one pollster — AtlasIntel, a Brazilian firm — accurately projected Trump’s winning all seven states where the election turned in 2024. Those swing states were Arizona, Georgia, Michigan, Nevada, North Carolina, Pennsylvania and Wisconsin.

The final New York Times-Siena College pre-election poll showed Trump ahead in just one swing state, Arizona. His Democratic rival, Kamala Harris, led in four of those states, and the race was tied in the other two. Some of the errors were not close: The Times-Siena poll missed the final outcome in North Carolina by 5.3 points and in Nevada by 6.1 points. 

The Washington Post developed a poll-based election model which, as the 2024 campaign closed, showed Harris leading narrowly in four swing states and holding a lead of 2 points nationally. The Post claimed its model “only includes polls that are transparent about how they are conducted … and use methods that have demonstrated accuracy.” 

CNN’s polling, which is conducted by the SSRS research firm in suburban Philadelphia, was especially off-target. Its final pre-election polling gave Harris leads of 6 percentage points in Wisconsin and 5 points in Michigan — polling errors of nearly 7 points and 6.5 points, respectively. CNN’s polling showed Harris carrying four swing states: Arizona, Michigan, Wisconsin and North Carolina. Had she won them, Harris would have been elected president.  

CNN’s performance was especially striking because the network had announced modifications to its polling methods following the 2020 campaign. That year, CNN’s final poll erroneously indicated Joe Biden led Trump by 12 percentage points. Biden won that election by just 4.5 points.

In 2024, the Marist/PBS/NPR poll estimated Harris led Trump by 4 percentage points at the end of the campaign. Trump won the popular vote by 1.5 points — which meant the Marist poll was off by 5.5 points. According to data compiled by RealClearPolitics, no national pollster in 2024 misfired as badly as Marist.

And no statewide poll in 2024 was more dramatically in error than the once-well regarded Iowa Poll, conducted by J. Ann Selzer for the Des Moines Register newspaper. Selzer’s final pre-election poll reported that Harris was ahead by 3 points in what has become a decidedly Republican state — a finding enthusiastically embraced by Trump foes, as it suggested Harris was on her way to the Oval Office. 

Trump went on to win Iowa by 13 points, implying a 16-point error for Selzer’s poll. 

Before taking office, Trump sued the Iowa Poll for fraud, claiming the erroneous poll “was no ‘miss’ but rather an attempt to influence the outcome of the 2024 presidential election” by projecting “a false narrative of inevitability” about Harris’s candidacy. 

Instead of suing and haranguing pollsters, Trump might well turn survey results back on them. For example, a Marist survey last year reported that 61 percent of respondents said they had little or no trust in polls; 38 percent claimed to have a great deal or a “good amount” of confidence in them. It would not be too outlandish to argue that pollsters’ approval ratings, at least according to Marist, are worse than Trump’s. 

Or, the president might point to a recent Emerson College poll, which indicated that voters, by a margin of 48 percent to 47 percent, would choose Trump over Harris in a rematch. 

The data are plentiful for a dispassionate, critical assessment of recent polling performance. Of course, dispassionate assessment has never been Trump’s favored approach. 

W. Joseph Campbell is a professor emeritus at American University in Washington, D.C. He has written seven solo-authored books including “Lost in a Gallup: Polling Failure in U.S. Presidential Elections.”