Audience members listen as U.S. Democratic presidential nominee Hillary Clinton speaks at a Women for Hillary campaign finance event in Washington, DC, U.S. October 5, 2016. (REUTERS/Brian Snyder)

There's an anecdote from a Wall Street Journal story by Reid J. Epstein last month that captured a lot of attention. In it, a Democratic candidate in Iowa recalls approaching a house to speak with a voter, encountering her husband working in his garage. The man tells the candidate that he and his wife aren't interested in talking to a Democrat -- but when the candidate, who calls later and speaks with the wife, learns that she is planning to vote for Hillary Clinton.

"There’s a lot of men who answer the door and speak for their wives," the candidate, Claire Celsi, told Epstein. "When I do get to talk to a woman, they say [Donald] Trump is unacceptable."

This is one story, but it captures something that skeptics of Trump's have suggested in the past. The Trump campaign frequently talks about how polls underrepresent his support, arguing that there's a chilling factor from some politically correct stigma about announcing that you agree with his policies. Polling hasn't actually found much of an effect to that extent, even when comparing live-caller polling -- that it, polls where you have to speak with a real person -- with online polling. When Morning Consult and Politico looked at this last week, they found a slight effect, mostly among those with a college education.

No one, though, has spent much time looking at the scenario above: women nervous about stating their support for Clinton. (Why not? In part, certainly, because Clinton's leading in the polls and hasn't repeatedly suggested that there's a polling flaw that's undercounting her votes.)

New data from the University of Maryland explores that question.

Respondents were asked to evaluate the extent to which they agreed with the statement, "There are Americans who will vote for [Trump or Clinton] in the presidential election even though they are not prepared to state that openly." Overall, more people agreed with the idea that a hidden vote existed for Trump, though that differed by party. Republicans were far more likely to say there was a hidden Trump vote than a hidden Clinton vote; Democrats were only slightly more likely to say the same.


There was also a difference by gender. Men were slightly more likely to say there was a hidden Trump vote and slightly less likely to say there was a hidden Clinton vote than were women. The difference was small: women were 5 percentage points more likely to say they strongly or somewhat agree with the idea that Clinton voters may be unwilling to state their support openly.

The university was generous enough to provide us with a breakdown by gender and party. (The sample size for independents was too small to be significant.) There, a wider gulf opens up.

Hidden_Gen

Republican women were 8 percentage points more likely than Republican men to say that there is hidden support for Hillary Clinton. They were 5 percentage points more likely to say they felt that way strongly, and 3 percentage points more likely to say they felt it somewhat. Democratic women agreed with the idea more than Democratic men, too, though to a lesser degree.

This is suggestive, but it doesn't really answer the question. The problem with there being a vote that isn't detectable by pollsters is that it isn't detectable by pollsters, making it either frustratingly elusive but likely (if you're sympathetic to the idea) or a bit like Bigfoot or UFOs (if you aren't). There are certainly scenarios in which individuals may not want to be open about their political feelings. Whether or not those situations add up to a significant number of votes on Tuesday is a question that's probably impossible to answer yet.