There is talk among Republicans, and some trepidation among Democrats, that Donald Trump could benefit from a silent vote. Although these voters aren't captured by polls, the privacy of a voting booth or a mail-in ballot will allow them to vent their anger and resentments.
The theory holds that in some circles it's not respectable to publicly support the inflammatory New York billionaire, but it's easier in private.
This is a variation of the so-called Bradley effect: In several instances over recent decades, white candidates have outperformed polls when running against a black opponent. Something similar was at work in the U.K. referendum on exiting the European Union. To the surprise of financial markets and bettors, the "leave" camp won the June vote, which was interpreted as an expression of discontent with the elites.
Similarly, in the U.S. presidential election, former House Speaker Newt Gingrich, a Trump supporter, claimed that "the average citizen will not tell pollsters the truth."
A silent or secret Trump vote would be a big deal: Hillary Clinton is up by several points in the latest surveys, not a comfortable margin.
There are, however, reasons to question whether Trump will outperform the polls.
This wasn't the case in the primaries. The Republican nominee underperformed the final polls in about as many states as those where he outperformed.
In Iowa, Ann Selzer, the pollster for the Des Moines Register and Bloomberg Politics, lost her unblemished record when she showed Trump with a small lead. He narrowly lost to Ted Cruz. Sure, that all was within a margin of error and Iowa's contest is a caucus, where turnout is low. And a week later, in New Hampshire, the final polls suggested Trump would win by about 15 points; he won by 20.
That trend was erratic. In one of his most important closing victories in Indiana, he finished well ahead of the final surveys. But in South Carolina and Michigan, though he won, he underperformed the last polls. In Wisconsin, the closing polls suggested Cruz would beat Trump by about seven points; the Texan won by 13.
The Bradley effect gets its name from the 1982 California governor's race, when the Democratic nominee, Los Angeles Mayor Tom Bradley, who was black, was running ahead of his white Republican opponent. When Bradley ended up losing, there was speculation that some voters didn't want to admit to polltakers that they wouldn't support a black candidate.
Trump has aroused racial feelings more than any national candidate since George Wallace 40 years ago, but a lot has changed. In 1982, there had not been a black governor anywhere since Reconstruction, and of course a black man has won the last two presidential elections, carrying California twice by landslides.
The Brexit vote did shock the establishment because the final polls pointed to a victory for the "remain" camp. Yet only a few weeks earlier, most of the surveys showed a win for "leave." In the interim, a pro-EU member of Parliament, Jo Cox, was assassinated. Some analysts speculate that the tragedy created a short-term backlash against the exit movement that dissipated by Election Day.
The U.S. presidential contest should be a "change" election, which ordinarily would work to the advantage of the Republican candidate in the last eight weeks. In 1980, Ronald Reagan surged at the end, demonstrating that he was up to the job. It's doubtful Trump has that capacity, so he needs a silent vote to offset his high negatives.
Even if this hidden strength adds a few points for Trump, the Democrats have a powerful offset. The Clinton campaign can rely on better data and analytics, as well as a better get-out-the-vote operation than President Barack Obama had in 2012. That may be worth at least two points.