After dramatically underestimating support for Donald Trump and other GOP candidates in the last two presidential elections, pollsters this year are attempting to improve the accuracy of their surveys in a number of ways. The effectiveness of the measures is a concern for strategists and pollsters in both parties.
Tom Bonier’s initial response is skepticism when he learns that polls in Ohio’s Senate race show a tie. After instance, the 8-point setback that President Biden suffered in Ohio two years ago was far larger than the polls had predicted. Mr. Bonier, a seasoned Democratic strategist and data guru, believes that pollsters are once more overstating support for the party.
Mr. Bonier also has a second thought: perhaps pollsters have learned from their prior errors and are now accurately identifying his party’s gain in popularity. Because of the 2016 and 2020 polling errors, Mr. Bonier claimed that “everyone feels a little bit burnt and gun-shy.”
One issue pollsters face as they work to increase their accuracy is that they haven’t identified the primary contributors to the mistakes that led to many Americans being shocked by Mr. Trump’s victory in 2016 and his closer-than-expected finish in 2020, which the professional association of public opinion researchers determined to be the biggest polling miss in 40 years.
According to several pollsters and academics, a particular group of Republicans—those who are most dedicated to supporting Mr. Trump and his political philosophy—are refusing to participate in their surveys. Others contend that while these voters do respond to polls and are present in the pollsters’ data, they are underrepresented in the outcomes due to stale assumptions about the make-up of the electorate in the country.
In an effort to solve these issues, pollsters are experimenting.
Lee Miringoff, director of the Marist College Institute for Public Opinion, said that the Marist Poll, among other organizations, sends text messages to voters inviting them to participate in online polls in an effort to target hard-to-reach voters. The poll’s other methods, such as calling voters on their cellphones or landlines with live interviewers, are meant to reach a larger sample of randomly chosen respondents than the text messages do. Respondents contacted by text message and phone interviews are both included in the Wall Street Journal survey.
Democratic pollster Celinda Lake claimed to have tried with several strategies to help people feel more at ease speaking to interviewers by establishing a sense of trust. She questioned people in a Montana political poll if they had witnessed the football game between the Montana State and University of Montana’s Bobcats and Grizzlies, or the “Bobcats-Grizzlies.”
Including “questions that telegraph that you’re local, and not liberal,” as Ms. Lake put it, was her strategy for boosting participation. She stated that additional testing was required but that she believed the method provided a sample that more accurately matched the make-up of the Montana voting pool.
According to some pollsters, even if they have the right proportion of voters from a particular group—for instance, suburban voters or Democrats—they may only be speaking to those who are the happiest to take part in polls and who don’t necessarily reflect the full range of opinions within the group. Some people, for instance, think that too many Democrats who are very involved were included in polls as a result of the issues in 2020.
Republican pollster Christine Matthews stated that to allay this worry, some in the field are looking up the voting history of their poll respondents, which is a public record, to determine whether the individual regularly votes in primary elections. An excessive number of Democratic primary voters might lead to an overestimation of Democratic fervor while undercounting “soft Democrats,” who could be less fervently supportive of the party’s nominees.
In a similar vein, Republican pollster Bill McInturff stated that he is attempting to ensure that he obtains a representative sample of opinions from each geographic category, including urban, suburban, and rural regions.
Mr. McInturff monitors whether the proportion of, for instance, urban Democratic voters or rural Republican voters is consistent with how those groups voted in the 2020 election when he conducts surveys for NBC News with Democrat Jeff Horwitt. As a result, he added, “the Biden margin in urban counties will not be overrepresented, and the extremely huge Republican majorities in Trump, rural areas will be more completely represented.”
Another unanswered subject that pollsters and scholars are now debating is why national polls, despite their flaws, have outperformed state polls in this election cycle’s statewide contests for the Senate and governorships. The American Association for Public Opinion Research determined that to be the case in their studies for the years 2016 and 2020.
In the meanwhile, a lot of experts in the industry point out that Georgia’s state-level polls in 2018 and the Senate runoff elections in 2021 both performed well overall. That implies that surveys are more reliable when Mr. Trump isn’t on the ballot, which is the case this year, according to many.
While the final polls performed well in the contest for Virginia governor in 2021, they performed substantially worse in the race for governor of New Jersey. Patrick Murray, the director of the Monmouth University Polling Institute, claimed that his polls “blew it” in the New Jersey race by failing to pick up on the significant opposition that Democratic Gov. Phil Murphy faced that nearly brought to the failure of his reelection campaign. He said that he is currently refocusing his survey in order to avoid focusing too much on the outcome of an election, or the “horse race.”
Mr. Murray now polls respondents on how supportive they are of each contender separately rather than pitting them head-to-head. This offers him an idea of the degree to which each candidate is supported, as well as the percentage of voters who have decided not to support a certain candidate, but it does not provide a head-to-head comparison that may be seen as predicative.
According to Mr. Murray, a pollster’s role is to “tell a story” about why people are supporting certain candidates or choosing not to support a particular candidate. “Polls shouldn’t be used to predict the future, but rather as a way to understand what voters’ thoughts are about.
There are hints that polling is more reliable during party primaries, where the voter pool is smaller and more homogenous in some respects since primaries often draw the party’s most ardent supporters while less ideological individuals frequently abstain. Pre-election surveys properly predicted that John Fetterman, a Democrat, would win by a margin of more than 30 points in this year’s competitive Pennsylvania primary for the Senate race. This year, they correctly predicted the outcomes of Republican Senate primaries in Ohio, Arizona, North Carolina, and Pennsylvania, frequently within a narrow margin of victory.