President Trump is very sensitive to what public opinion polls say about his presidency. He pretends not to be, but on many occasions when Trump-friendly pollsters show his approval at or above 50 percent, he tweets about it. He constantly insists that his approval rating among Republicans is at 96 percent, an invented figure, and nearly as often insists that polls are failing to capture his full support.

“Yesterday, I guess we had a 53 [approval] poll,” Trump said last September, referring to one of those Trump-friendly polls. “And a lot of people say add 10 points to anything. Anybody voting for Trump, you can add — anytime you get a poll, you can add 10 points or seven points or six points. Take it any way you want.”

The president doesn’t usually go into why those points should be added, but he appears to subscribe to what’s become known as the “shy Trump voter” theory: that Trump supporters are 1) unwilling to share their support, 2) interested in fooling pollsters or 3) not included in polling. Add those factors, and, boom, throw however many points you want on top of existing poll numbers.

For Trump, this theory has the happy side effect of making his support look larger than it otherwise would be. He’s not trailing former vice president Joe Biden; add X number of points, and, suddenly, you’ve got a tie.

For external observers, though, it raises an interesting question: To what extent, if any, are polls missing Trump’s support? Clearly, some state polls missed the mark in 2016, leading forecasts to assume he had less of a chance of winning than he did. Was it because Trump supporters weren’t included or were wary of talking to pollsters?

Research after the fact assessed the idea that pollsters were being misled.

“[T]he notion that Trump supporters were unwilling to express their support to pollsters was overblown, given the scant evidence to support it,” a report from the Pew Research Center stated. “A committee of polling experts evaluated five different tests of the ‘shy Trump’ theory and turned up little to no evidence for each one. Later, a researcher from Yale and Pew Research Center conducted separate tests that also found little to no evidence in support of the claim.”

But we can evaluate this idea another way. A number of recent polls at the state level have measured both the presidential election and Senate races. If people were wary of revealing support for Trump specifically, it’s safe to assume that Trump would trail the Republican Senate candidates.

For the most part, though, the opposite is true.

Senate GOP
Trump +5
Senate +4
South Carolina
Trump +5
North Carolina
Trump +5
Trump +5

The only state in which the Republican Senate candidate does better than Trump is in Maine, where Quinnipiac University has Trump trailing Biden by more than 20 points, meaning that it’s possible this is an outlier poll. Everywhere else, Trump is either matching or outperforming the Republican Senate candidates.

It remains possible that pollsters simply aren’t contacting voters who will turn out to vote for Trump. After all, if he is able to mobilize people who don’t vote very often, those voters probably won’t be included in polling because they might be less likely to be included in filtering for likely voters. But this begs the question to some extent because it assumes the existence of support that’s functionally unmeasurable.

It also remains possible that the “shy voter” theory remains unfounded and that, just as national polling accurately measured the outcome of the 2016 race, polling is largely capturing the race as it stands. That Trump is doing better than his party’s Senate candidates in several states, even when he nonetheless trails Biden.

For a president eager to present a picture of inevitable victory, though, the idea that there’s a few million votes evading the spotlight of public polling sounds awfully appealing.

Emily Guskin contributed to this report.