Will the Polls Be Wrong Again? Pollsters Weigh In

The 2016 and 2020 presidential elections — in which many respected pollsters were wildly off — left in their wake questions over whether polling was irrevocably broken, or still a barometer of the nation’s pulse that could be corrected with some fixes in methodology.
Eight years ago, national polls were predicting an easy victory for Hillary Clinton over Donald Trump, right up until Election Day. Nearly nine in 10 polls overstated Clinton’s support among voters that year, according to a Pew analysis done post hoc.
In 2020, polls showed President Joe Biden with a strong lead over then-President Trump. Just before the election, Biden led by 8.4 points in FiveThirtyEight’s average and 7.2 points in RealClearPolitics’ average. However, he won the popular vote by less than 4.5 points—just enough for an Electoral College victory.
If polls are off by a similar margin this year, Vice President Kamala Harris’s narrow lead in many averages could mean Trump is actually ahead and heading for an electoral landslide.
But that’s a big if. Many reputable pollsters claim they’ve fixed their polling errors and are now better able to capture Trump’s support, both nationally and in the battlegrounds.
As to why past polls have been so off base, some pollsters believe Trump supporters opt out of surveys due to distrust in institutions like mainstream media and polling organizations. Scott Keeter, senior fellow at the Pew Research Center, told Newsweek that this mistrust may explain why Trump’s support is often underrepresented in poll results.
“Trump supporters may have less trust in the institutions sponsoring these polls. If they choose not to participate because of this distrust, their support isn’t accurately captured,” Keeter said.
He suggested this lack of participation contributed to the polling errors of 2016 and 2020, while midterm polling in the 2022 midterms showed greater accuracy. The difference is that Trump was not running for office in 2022.
“Presidential elections attract a more diverse voter base, which adds variability,” he said. “This is why we can’t assume the accuracy seen in 2022 will hold for 2024, especially with Trump on the ballot.”
Despite the improved performance of midterm polls, many Americans remain skeptical that polls can accurately capture public political preferences, especially in an election that experts describe as “the tightest in recent years.”
In response to past errors, pollsters have refined their methods, increasingly using online and text message surveys. Polling has evolved considerably from 2016, when most polls were still conducted via landline, even though at that point less than half of U.S. households had a working landline phone. In fact, by 2022, 61 percent of U.S. polling organizations that conducted national surveys in 2016 had shifted to new methods. This evolution in polling continues.
Berwood A. Yost, director of the Center for Opinion Research at Franklin & Marshall College, explained that modern polling now involves a blend of online, text, and other techniques to reach a broader audience, with methodology tweaked to reflect changing communication norms.
“Pollsters have adapted over the past eight years, moving away from phone surveys and incorporating text messaging, mail, and online surveys to reach those who might otherwise avoid polls. But it’s still experimental—these adjustments aim to better capture voices that might otherwise go unheard.”
Over the past two decades, the rise of cellphones, text messaging and online platforms has transformed survey research. Traditional methods, like live-operator phone polls, are nearly obsolete, with fewer than a quarter of Americans still using landlines, according to CDC data.
Reliance on mobile phones, however, introduces a new challenge: Americans increasingly avoid answering unknown calls.
“Now that caller ID reveals who’s calling, fewer people answer unknown numbers, and even fewer are willing to discuss personal views with strangers,” said Spencer Kimball, director of Emerson College Polling and associate professor of Political Communication at Emerson College.
Unlike some traditional pollsters, Kimball said, Emerson uses a mix of techniques, including text-to-web surveys, interactive voice responses on landlines, email outreach and opt-in panels. This combination helps them reach a more representative audience, particularly younger voters. Some pollsters have also adopted mail methods known as address-based sampling.
Pew found that adjusting poll samples based on multiple factors can improve accuracy. One popular adjustment is called “recalled vote,” which asks voters whom they supported in 2020. Some pollsters believe this adjustment helps avoid underestimating Trump’s support, while others worry it could introduce errors if people misremember past votes.
Well regarded polling outfits like CNN/SSRS, The Washington Post and Monmouth University are using this technique.
“We don’t know if today’s polls are more accurate than four years ago. In 2016 and 2020, many new Trump voters emerged, challenging pollsters due to their lack of voting history. Pollsters are trying to improve, but only after the election will we know if these adjustments worked,” Keeter said.
Despite those new methods, it’s uncertain if polls will fully capture voter preferences, given the unpredictable dynamics of this presidential election cycle.
“The question of whether polls miss certain voters remains unanswered,” said Yost, adding that in 2016, national estimates of Trump’s vote share were fairly accurate. The underestimation was larger in 2020 due to unique factors like the pandemic.
Still, Yost is optimistic about new data refining polling methods, which he believes will better capture voters who were missed in previous elections.
“We’ve made changes, like managing ‘don’t know’ responses differently in vote-choice questions,” he said.
“By not encouraging ‘don’t know’ responses, we aim for more accurate estimates, as seen in 2022. Additionally, we’ve expanded outreach through text messaging to recruit participants.”
While pollsters have made changes with the hope of improving accuracy in the 2024 election, they are also contending with a collapse in media trust. Yost is calling for greater transparency in presenting results and providing clear explanations of poll methodologies to help explain to the public how national polling works, and how it doesn’t.
“That’s one of the biggest changes we’ve made,” he said. “We’re also mindful of polling limitations, especially in close races. No poll with 700-1,000 participants can precisely determine the state of a race within one or two points.”