The theory goes like this: Tired of facing hostility for their beliefs, supporters of President Trump decide against sharing their political views with others. That shyness leads them even to reject calls from polling firms or to answer questions incorrectly, with the result that some level of support for the president isn’t captured. And, voilà, suddenly the polls are wrong.

Trump loves this theory, of course. He’s taken to suggesting one should simply slap on a few percentage points to any poll of how he’s doing, as he did back in February.

In general, that’s not how polling works. Polls are conducted with specific statistical boundaries that make it hard simply to decide your support is being undercounted by a random amount. Of course, polls also don’t consistently show you have 95 percent approval, either.

Poll inflation aside, it’s also not demonstrably the case that Trump’s support isn’t being fully captured. But on Wednesday morning, the hosts of Trump’s favorite entertainment program, Fox News’s “Fox and Friends,” hosted David Brody — himself a Trump favorite — to explore this theory of “stealth Trump voters.” That is, people who are too intimidated to tell family members or even random people over the phone that they support Trump.

Brody said he had spoken with several such people for an article at the website Just the News. Fox’s Steve Doocy asked who these “stealth voters” are.

“They’re all over the map,” Brody said. “This is anecdotal. They’re out there. That’s — we know that for sure. You can’t put it in a test tube. It’ll be a surprise on Election Day as to how many and where they show up potential battleground states.”

Kudos to Brody for at least acknowledging these are anecdotal stories. It’s not really fair, though, to move from “these are anecdotal” to “therefore there are lots of them and therefore they might lead to a surprise on Election Day.”

“But here’s what we know when I was talking to many of them — and believe me, my inbox was filling up. We talked to three; there are so many more that we’re filling up my inbox,” Brody continued. “But here’s what they told me. They said, look, if a pollster calls my house, I’m not talking to them at all.”

One of the fascinating dynamics that surrounds political polling is that people misunderstand the nature of a randomized sample. Pollsters often hear complaints from individuals that their polls can’t be trusted because the complainant has never been called for a poll — which, of course, doesn’t mean anything at all. There are lots of reasons someone might not be included in a poll, depending on the type of poll being conducted.

But notice, too, what’s happening here: Brody and Doocy are extrapolating from three people who say they wouldn’t answer a poll question if called to a broad misrepresentation in polling. These aren’t even people who say they didn’t talk to a pollster; they’re people who say they wouldn’t.

Brody then offers data to prove his point.

“And this is exactly what we’ve seen, at least from a trend line standpoint with Internet polls,” he said. “We know that President Trump does better in Internet polls than he does when people are just calling the home. And so it’s very interesting.”

In his article, he points to a study conducted by the polling firm Morning Consult in December 2015. That study suggested Trump supporters might be more willing to indicate their support in an online poll than when talking to a live interviewer. FiveThirtyEight saw a similar effect in August 2016 but that the effect didn’t exist in July of that year.

Where does it stand now? There’s no significant difference between the two methodologies in general election polls conducted this year. Some live-caller polls show Trump with a smaller deficit against former vice president Joe Biden. Some online or automated polls show Biden with a wider lead. The trend lines for the two are similar.

In fact, an interesting study conducted by the Pew Research Center suggested polls conducted with an online methodology might be overstating Trump’s approval numbers. It found a significant number of respondents might simply whip through the poll agreeing with everything — including with Trump’s performance as president. That’s approval ratings, though, and not political support.

To bolster his point, Brody pointed to two other bits of evidence.

The first was an academic study that found those unwilling to share their views of which 2016 presidential candidate they supported were more likely to support Donald Trump. During the Fox News interview with Brody, the network flashed a data point from that study: 54 percent of those individuals backed the president in that election.

But this wasn’t a measure of misrepresenting political views to pollsters. It was about people being unwilling for a variety of reasons to share their views with people they knew personally. It was also the case that 27 percent of the people included in that study ended up voting for Hillary Clinton, substantially narrowing any effect of the “stealth voter” concept.

The other bit of evidence for Brody’s point that he and Doocy discussed was, as you might expect, the results of the 2016 election. Trump’s win despite polling in several key states has often and broadly been used to argue for the inaccuracy of polling in general, including by Trump. But, as we’ve written, that’s overly broad. The 2016 polling nationally did a good job of capturing the results of the race. That Trump overperformed in states like Wisconsin and Pennsylvania was obviously hugely significant for the outcome of the election, but that was far more a function of flaws in how the polls were conducted than any “shy voters.”

There are, as Brody says, people who feel social pressure to deny support for Trump. There is not, despite Brody’s claims, evidence that this significantly affects how Trump fares in polling. There is also not evidence that these voters will shift the results of the election this year.

There is one person who undoubtedly enjoyed that segment, though: “Fox and Friends” superfan Donald Trump.