WASHINGTON — National and most swing-state public opinion polls show a steady lead among likely voters for former Vice President Joe Biden, but President Donald Trump disputes those surveys because he says they miss "shy Trump voters."
Trump insists he has a much larger base of supporters than the polls show, a group he calls "the silent majority" — a term he borrowed from President Richard Nixon — to describe voters who don’t tell pollsters and reporters that they’ll vote for his reelection.
Pollsters, however, reject or downplay Trump’s critique, saying that they don’t believe people lie to pollsters when they answer questions and that they have adjusted their polling methods to ensure they include a large enough sample of likely Trump voters in their surveys.
"The idea of a shy Trump voter is something Trump supporters tell themselves to make themselves feel better. There is no such thing," said Republican pollster Whit Ayres. "The idea that they're going to be lots of those voters missed is just not likely to occur."
Still, stung by Trump’s 2016 upset of Democrat Hillary Clinton, even some experienced political analysts say they have a nagging worry that maybe there are undercurrents, especially among white, working-class voters, that somehow are being missed.
Kyle Kondik, managing editor of Sabato’s Crystal Ball, a political newsletter, at the University of Virginia Center for Politics, said he can’t point to anything wrong that the polls are doing but added "a little voice in the back of my head" wonders if pollsters "are way off" on white voters.
Noting mounting evidence in polls in many of the competitive states that Biden’s doing a lot better with white voters than Clinton, Kondik said, "I do worry that it is possible that the polls collectively are just maybe missing some of the kind of rural Trump vote."
Yet a study by pollsters found no evidence for "shy Trump voters." The American Association of Public Opinion Research’s in-depth report on 2016 election polling found it was the most accurate estimation of the popular vote in eight decades.
Some state polls had samples that underestimated Trump supporters, especially working-class white men and women who did not got to college, in Wisconsin, Pennsylvania and Michigan, where Trump won by less than a combined 80,000 votes, the report said.
"The polls in the upper Midwest and some of the state polls were clearly off, and they were off, primarily, because there were too many college-educated whites and an insufficient number of non-college-educated whites" in the pollsters’ samples, said Ayres.
"But most pollsters have adjusted for that now," he said.
Many pollsters now include a category for non-college-educated whites and college-educated whites, said Gary Langer, whose firm Langer Research Associates produces ABC News polls. It is one of the standards of the public opinion research association.
Some pollsters are taking into account the partisan "sorting" of America, in which people tend to live in communities where others are like them and share many of the same views, especially on politics, through polling by geography.
After experimenting in polling for the 2019 Kentucky governor’s election, Marist College's Institute for Public Opinion will change how it polls in states for NBC News.
It will incorporate geographic checks for urban, suburban, exurban and rural areas in drawing representative statewide surveys and rely more heavily on random dialing to ensure better coverage of the voting population, Marist and NBC News said.
Don Levy, director of the Siena College Research Institute, said polls for The New York Times are based on calls to randomly selected names from voter registration lists divided into geographic areas in each state with demographic characteristics that match all voters in those areas, he said.
Langer, who employs the long-standing polling practice of using random digital telephone interviewing with robust cellphone samples, urges news outlets, readers and viewers to take a close look at how polls are done.
"You see a number and a percentage sign," Langer said, "and you'll want it, you grab it, you run with it. And I like to say that running with numbers is like running with scissors. It's easy to get hurt."
He added, "We have to stop and say, ‘Well, where were these numbers obtained, what methods were used, are we getting full disclosure on the methodology, on the questions on the analysis, to make sure that these are quality data in which we can reliably depend."
Editor's note: Early versions of this story incorrectly the three states that Donald Trump won by less than 80,000 votes in 2016. The three states are Michigan, Pennsylvania and Wisconsin.