By Jere­my W. Peters, The New York Times

MOORESVILLE, N.C. — It wasn’t the most obvi­ous spot for a flag that peo­ple usu­al­ly buy to make a big state­ment. But there it was, peek­ing out from the inside wall of a garage, the white “Trump 2020” let­ter­ing just vis­i­ble from the street in this sub­ur­ban Char­lotte neighborhood.

From the front porch, Tiffany Blythe, a stay-at-home mom, said that she and many of the peo­ple she knows would be vot­ing for Pres­i­dent Don­ald Trump in Novem­ber — but that many of them were ner­vous talk­ing about it. And that hes­i­ta­tion is why Blythe doesn’t trust the polls that are now fore­cast­ing loss­es this fall for Trump and oth­er Repub­li­cans in North Car­oli­na and beyond.

“I’m not buy­ing it,” Blythe said. “There are a lot of silent vot­ers, and more will come out before the elec­tion. I think a lot of states are turn­ing red from blue, but you don’t hear about that in the media.”

The belief that Amer­i­cans aren’t get­ting the real sto­ry about Trump’s chances for reelec­tion has tak­en hold among many of his sup­port­ers. For Trump loy­al­ists, it is an appeal­ing sto­ry and one with some valid­i­ty: The news media, which large­ly failed to antic­i­pate Trump’s vic­to­ry in 2016, are under­count­ing his vot­ers, many of whom are even more reluc­tant today than they were four years ago to declare them­selves in his camp.

Trump makes this argu­ment often; on Sat­ur­day evening, he told reporters that “we have a silent major­i­ty the likes of which nobody has seen.” One of his poll­sters, John McLaugh­lin, has even put a name to this sup­posed flaw in the data, pre­dict­ing that the “hid­den Trump vot­er” will prove the news media wrong.

But the idea that there are sub­stan­tial num­bers of Trump vot­ers who will emerge from hid­ing on Elec­tion Day, large enough to sway the out­come, is not sup­port­ed by the lat­est pub­lic opin­ion research — or by a prop­er under­stand­ing of what hap­pened in past elec­tions where the vot­er sur­veys were off, said poll­sters who work for Repub­li­can and Demo­c­ra­t­ic candidates.

This does not mean that Joe Biden’s lead, with the most recent nation­al polls putting him ahead by as much as 10 points, won’t tight­en. And pub­lic opin­ion experts said there is grow­ing evi­dence that Amer­i­cans across the spec­trum have become more skit­tish about shar­ing their polit­i­cal pref­er­ences out­side a trust­ed group of like-mind­ed peo­ple. But it would be a huge leap to con­clude that the country’s tense polit­i­cal dynam­ics are caus­ing peo­ple to lie to poll­sters in large enough num­bers to explain Trump’s poor standing.

“There are many peo­ple who are vot­ing for Trump who are in envi­ron­ments where it’s polit­i­cal­ly unten­able to admit it because he’s become so tox­ic,” said Whit Ayres, a Repub­li­can poll­ster. “But I’m still not con­vinced that not telling your busi­ness asso­ciate or the peo­ple in your Rotary Club or the peo­ple in your coun­try club is the same thing as not telling a pollster.”

The pos­si­bil­i­ty that Amer­i­cans are hid­ing their true inten­tions from poll­sters has pro­vid­ed an irre­sistible sense of intrigue to pres­i­den­tial elec­tions before, even though there are few con­firmed exam­ples where it made a dif­fer­ence. Polit­i­cal experts com­pare such spec­u­la­tion to the qua­dren­ni­al pre­dic­tions of a bro­kered con­ven­tion, which has not occurred since 1952.

In 2008, there were ques­tions about whether sup­port for Barack Oba­ma was being inflat­ed in the polls by peo­ple who didn’t want to say they weren’t sup­port­ing him. That did not happen.

Four years ago, some sug­gest­ed there might be a sim­i­lar phe­nom­e­non at work with Trump sup­port­ers who were too embar­rassed to reveal them­selves. And when Trump won by squeak­ing out vic­to­ries in a few bat­tle­ground states, his back­ers argued that shy vot­ers were a rea­son the polls missed his strength in those places.

“The idea that peo­ple lie, it’s an inter­est­ing the­o­ry, and it’s not like it’s com­plete­ly off-the-wall,” said David Win­ston, a poll­ster who works with con­gres­sion­al Repub­li­cans. “But it’s obvi­ous­ly a very com­pli­cat­ed thing to try to prove, because what do you do? Ask them, ‘Are you lying?’”

Win­ston said that many pro­po­nents of the the­o­ry about hid­den Trump vot­ers rely on what is known as the Bradley effect, named after Tom Bradley, the for­mer may­or of Los Ange­les who lost the 1982 Cal­i­for­nia governor’s race despite polling con­sis­tent­ly ahead of his white oppo­nent. Among polit­i­cal sci­en­tists, the the­o­ry that emerged to explain the gap between the polls and the elec­tion results was that white vot­ers were wor­ried about appear­ing racist if they did not say they were sup­port­ing Bradley, who was Black.

But some have ques­tioned the valid­i­ty of the Bradley effect, includ­ing Blair Levin, one of Bradley’s for­mer advis­ers, who has argued that Bradley lost because of a com­pli­cat­ed mix of fac­tors, among them a robust Repub­li­can absen­tee vot­ing cam­paign and an unpop­u­lar gun con­trol ini­tia­tive on the bal­lot, both of which turned out a surge of Repub­li­can voters.

If vot­ers were indeed afraid of voic­ing their sup­port for the pres­i­dent, Win­ston said, oth­er num­bers in the poll would reflect that, like see­ing an uptick in the per­cent­age of unde­cid­ed vot­ers rather than a rise in sup­port for Biden.

“It would not be peo­ple say­ing they are vot­ing for Biden,” he said, “but that they’re undecided.”

While the effects of a hid­den Trump vote are cer­tain­ly over­stat­ed by the president’s allies, that does not mean that no evi­dence exists that polls are miss­ing some of his vot­ers. A small per­cent­age of his sup­port is prob­a­bly being under­count­ed and has been in the past, pub­lic opin­ion experts said. And in states like North Car­oli­na, where the mar­gin of vic­to­ry could be nar­row, the under­count could make a dif­fer­ence between a poll being right or wrong.

“We assume the race will tight­en, and as that hap­pens, the size of the shy Trump vote could very eas­i­ly come into play,” said Neil New­house, a Repub­li­can who led Mitt Romney’s polling in 2012.

In 2016, New­house said that Trump tend­ed to score 2 or 3 points high­er in phone sur­veys when respon­dents were asked to press a but­ton to record their pref­er­ences rather than talk to a live per­son. In post­elec­tion polling, when he asked peo­ple if they had ever been unwill­ing to talk about their vote, 35% of Trump vot­ers said yes. And they tend­ed to be women from Demo­c­ra­t­ic-lean­ing counties.

New­house has picked up fur­ther evi­dence of such reluc­tance recent­ly. In polls he con­duct­ed late last month in North Car­oli­na and Iowa, he found that one-quar­ter to one-third of vot­ers answered “yes” when asked if they knew some­one who is vot­ing for Trump but would not say so to any­one but their clos­est friends.

“This total­ly con­firms the notion of ‘shy Trump vot­ers,’” New­house said.

But, he added, if polls are under­count­ing some Trump vot­ers — a group that tends to be unique­ly expres­sive and adamant about their sup­port for the pres­i­dent — no one can say by how much.

And in any case, poll­sters said they have cor­rect­ed one of the biggest mis­takes they made in 2016, when they failed to account for the high num­bers of vot­ers with­out col­lege degrees who turned out, many of whom vot­ed for Trump. And they are includ­ing a larg­er pool of pos­si­ble vot­ers in sur­veys — not just peo­ple who say they are like­ly to vote, as poll­sters often do — because they antic­i­pate his­toric turnout.

One vari­able that pub­lic opin­ion experts are still grap­pling with is how the polar­ized polit­i­cal cli­mate is affect­ing the accu­ra­cy of their work. Recent research has shown that con­ser­v­a­tives fear they are more like­ly than mod­er­ates and lib­er­als to be tar­get­ed for being hon­est about their polit­i­cal beliefs, although self-cen­sor­ship appears to be ris­ing among most Americans.

In a sur­vey con­duct­ed last month by the lib­er­tar­i­an-lean­ing Cato Insti­tute, 77% of con­ser­v­a­tives said they felt unable to share their polit­i­cal opin­ions because oth­ers might find them offen­sive, an increase from 70% who said so in 2017. Among lib­er­als, 52% report­ed those con­cerns, com­pared with 45% three years ago; among mod­er­ates it was 64% ver­sus 57% previously.

Inter­views with vot­ers like Blythe sug­gest the unease over polit­i­cal speech has become high­ly politicized.

“I see Democ­rats not pro­tect­ing our free­doms, and that includes free­dom of speech,” she said.

Still, Geoff Garin, a Demo­c­ra­t­ic poll­ster, said what real­ly explains Trump’s low stand­ing is not flawed method­ol­o­gy but rather the president’s inabil­i­ty to expand his coalition.

“The prob­lem for the Trump cam­paign is not the silent Trump vot­er but the dis­ap­pear­ing Trump vot­er,” Garin said. “And there are a lot more dis­ap­pear­ing Trump vot­ers than there are silent ones.”

The flip side of that dynam­ic — peo­ple who did not vote for Trump in 2016 and who have now become sup­port­ers — is bare­ly a factor.

“Among Clin­ton vot­ers, it is less than 5%,” Garin said.

Hen­ry Olsen, a senior fel­low at the Ethics and Pub­lic Pol­i­cy Cen­ter, has not­ed anoth­er warn­ing sign in the data for any­one hold­ing out hope that a hid­den pock­et of vot­ers will save Trump: There aren’t many peo­ple back­ing a third-par­ty can­di­date right now who could poten­tial­ly switch to the Repub­li­can tick­et. Olsen said in an email that while some con­ser­v­a­tive vot­ers often park them­selves with the Lib­er­tar­i­an can­di­date in the months before an elec­tion, many even­tu­al­ly come home to the Republicans.

But the Lib­er­tar­i­an Par­ty can­di­date, Jo Jor­gensen, is reg­is­ter­ing minus­cule sup­port in the polls now com­pared with the high sin­gle-dig­its that Gary John­son, the party’s 2016 nom­i­nee, was see­ing at this point four years ago.(STORY CAN END HERE. OPTIONAL MATERIAL FOLLOWS.)Still, among Repub­li­cans in North Car­oli­na, none of these facts seem to dent their con­fi­dence or con­vince them that they should wor­ry about polls show­ing Trump los­ing to Biden — or oth­ers that have the state’s Repub­li­can sen­a­tor, Thom Tillis, behind his Demo­c­ra­t­ic oppo­nent, Cal Cunningham.

Tim Moore, the Repub­li­can speak­er of the North Car­oli­na House of Rep­re­sen­ta­tives, said, “Here’s the thing I won­der: If you think about how divid­ed we are as a coun­try, how polar­ized the two sides are at the nation­al lev­el when it comes to the pres­i­den­tial race, how many peo­ple are tru­ly undecided?”

Moore believes that vot­ers who say they are unde­cid­ed today will deci­sive­ly break for Trump and help him win the state.

In Blythe’s sub­ur­ban Char­lotte neigh­bor­hood, many peo­ple who answered the door were will­ing to say they sup­port­ed Trump for reelec­tion. But when asked for their names, few of them agreed.

“I’d like to keep my friends,” one woman said.

At anoth­er res­i­dence down the street, a man who opened the door bare­ly enough for a reporter to hear him say he was vot­ing Repub­li­can joked, “Don’t burn down my house.”