As polls have turned against him, Donald Trump’s allies have speculated that there is a shy tribe of supporters who are reluctant to declare themselves and whose invisibility in the polls would mean Hillary Clinton’s advantage is being overstated.
The problem for the New York businessman? Pollsters have looked far and wide and deep into their methodology and they say they have struggled to unearth evidence of the shy Trump voter theory advocated among others by Kellyanne Conway, Mr Trump’s campaign manager.
The debate is just one of a legion of questions hanging over the accuracy of opinion pollsheading into a razor-thin contest. Recent opinion polling mishaps in the UK, Israel and Spain have raised the stakes for pollsters heading into November, as has Mr Trump’s heavy emphasis on polls in his public statements — at least when they are going his way.
“It is a multidimensional challenge,” said Michael Dimock, president of the Pew Research Center.
Faced with an unconventional candidate such as Mr Trump, pollsters have been mindful of the so-called Bradley effect — named after a mayoral candidate in Los Angeles for whom polls misstated support. However, Douglas Schwartz, director of the respected Quinnipiac University Poll, said the primaries earlier this year had proved the theory of the shy Trump supporter wrong.
On balance, polls had been accurate in the primaries in gauging support for Mr Trump, he said. “My thinking is if it didn’t show up in the primaries it’s not going to show up in the general,” Mr Schwartz said. “The Trump voters don’t seem shy. At the rallies they don’t seem afraid to say how they feel.”
Trump battles tax affair claims
If there are questions about the accuracy of this year’s polls they hinge on specialists’ perennial struggles to predict who will actually turn out and vote leading up to November 8. “Really identifying who is going to be motivated to show up is one of the bigger challenges in election polling,” Mr Dimock said.
Among the big questions in this election are the effect of Mr Trump’s statements onLatino voters, who tend to turn out in smaller numbers than other groups but may be motivated by his divisive rhetoric on immigration. On the other hand, could Mr Trump inspire a larger than normal share of white, working-class voters to turn out at the polls?
Mollyann Brodie, head of public opinion research for the Kaiser Family Foundation and president of the American Association of Public Opinion Research, said the “likely voter” models that analysts use are one of the most contentious areas of debate among pollsters.
Even poll aggregators and those averaging poll results this year were making their own assumptions about the reliability of different polls and other factors in their models. “I definitely think it is a time for poll watchers and audiences to be looking for a variety of sources of information,” she said.
Michael McDonald, associate professor of political science at the University of Florida, said the US was not necessarily a “one size fits all electorate”, and that different states could see lower or higher turnouts than previously.
On top of this come methodological questions that dog opinion pollsters. A mounting share of polls is now being conducted online, for example, using an array of techniques, and the profession is divided over how reliable they are likely to prove.
While some pollsters favour telephone polls, response rates have been tumbling in recent decades, in part perhaps because people may be reluctant to answer calls from an unrecognised number, Mr Dimock said.
Mr Trump’s anti-establishment message also carried its own polling challenges as did the disenchanted voters most likely to support him. “If you despise politics and a pollster calls you up and wants to ask 20-30 questions about politics you are probably going to hang up,” said John Della Volpe, director of polling for the Institute of Politics at Harvard.
That applied equally to the challenges presented by millennials and the US’s growing Spanish-speaking Latino population, with many pollsters struggling to reach them in the first place or surveying voters only in English.
When Harvard’s Institute of Politics polled young voters earlier this year it found much heavier support for Mrs Clinton among those Latinos who chose to answer questions in Spanish rather than English, for example, said Mr Della Volpe. Just 1 per cent of the Spanish speakers said they would vote for Mr Trump versus 9 per cent of the English-speaking Latinos.
For all the recent ups and downs, the polls this year had also been remarkably consistent in pointing to a Clinton lead — with polls taken since Mrs Clinton’s first debate with Mr Trump last month underscoring that message. The RealClearPolitics website’s polling average has given Mrs Clinton a lead since the end of July.
Some experts play down the likelihood of a major polling upset, pointing to the US’s long history of presidential election polling. The sheer volume of polls this year in the US also carried its own value.
“With so many different folks producing results there would have to be such a fundamentally different electorate showing up on election day than anyone conceived of for there to be a major surprise,” Ms Brodie said.