President Trump delivers remarks on expanding health coverage options for small businesses on Friday, June 14, in Washington. (Jabin Botsford/The Washington Post)

There is one poll question I keep coming back to when I think about President Trump’s 2020 reelection campaign. It’s the one in which pollsters ask whether people would “definitely not” vote for him.

This is an especially bad number for Trump. National polls generally show a majority of people (51-56 percent) say they wouldn’t (Fox News polls being the exception). And now a poll also shows that number is remarkably bad for Trump in a surprising place: Texas. The University of Texas/Texas Tribune poll shows that, when asked if they would vote to reelect Trump, 43 percent responded “definitely not”, and another 7 percent say “probably not.” That’s half the state intending not to vote for a Republican president — in Texas.

Those numbers are apparently starting to register in early 2020 general election polls — including Trump’s own internal polls — which show Trump struggling to climb out of the low 40s in key states and sometimes trailing by double digits. He’s also running neck and neck with Joe Biden in Texas.

But just how reliable is this question? And what does history tell us about whether these people are actually completely off the table for Trump?

The first thing we can say is that Trump’s struggles on this poll question are unprecedented. Pollsters have asked this question dating back to Jimmy Carter, and nobody has ever seen such a high “definitely not vote for” number as Trump.

The highest this number has ever gotten for Trump’s predecessors was 53 percent. That was for George H.W. Bush one time in October 1992. But asking this question on the eve of the election is kind of cheating. People know who the opponent is and have largely decided, so you might as well just ask them to choose between the candidates.

Apart from that, the highest a politician has ever climbed on this question was Barack Obama, who some polls showed reaching the high-40s on the “definitely not vote for” question in 2011. Obama hit 49 percent in a September 2011 McClatchy-Marist poll and 47 percent in a June 2011 Washington Post poll. The number who said they would definitely vote for him was significantly lower in both polls.

As you might remember, Obama went on to win reelection in 2012 despite those poor early numbers. What’s more, he took 51 percent of the vote. That suggests either that 1) he won all the voters that were open to voting for him in that September 2011 McClatchy-Marist poll, 2) people open to voting for him turned out at higher rates than those dead-set against him, and/or 3) he converted some people who had sworn off supporting his reelection.

There are precisely zero examples, though, of an incumbent president actually taking more of the vote than one of these “definitely not vote for” polls allowed. That would seem to suggest that Trump might top out at between 44 percent and 49 percent of the popular vote, depending upon which pollster you trust.

The example above is important, though. There are perhaps other reasons why Obama would have the second-highest “definitely not vote for” number in history, but it likely has a fair amount to do with the fact that he is our most recent former president. Over time, people have been writing off incumbent presidents earlier and earlier.

A June 1979 CBS/New York Times poll, for instance, showed 58 percent of Americans were disinclined to vote for Carter’s reelection, but just 30 percent said they were “definitely” not voting for him. Ronald Reagan was particularly unpopular during the middle part of his first term, but his “definitely not vote for” number never crested 33 percent. This number has crept up over the years, as our country has become more and more polarized and there are fewer and fewer swing voters.

But there is a difference between creeping up to the mid-40s — a share of the vote which each major party is virtually guaranteed in national elections these days — and climbing 5, 6 or even 10 points higher. And Trump trailing by double digits is one thing; having a majority of the country say there is no situation they can fathom in which they’ll vote for him is quite another. If these polls are accurate and these people actually follow through, Trump will need to rely upon third-party candidates taking from the Democratic nominee, another perfect electoral-college storm, or both.

It’s foolish to predict elections this early. But this question is the rare one in which people are asked to entertain a whole host of possibilities and allowed to offer a nuanced, middle-ground position — which is often an attractive one for poll respondents. They’re not taking it. A huge majority of those who say they dislike Trump and/or disapprove of his job performance are also saying they can’t see a circumstance in which they’ll vote for him.

He could plausibly win without any of their votes, but you — and he — wouldn’t want to count on it. It will apparently be up to him to convince people who have written him off early to take a second look.

If he does it, it might be an even greater political trick than the one he pulled off in 2016.