A crowd gathered outside President Trump's rally in Greenville, North Carolina.Rachel Jessen

Buried beneath the blustery bravado of Donald Trump’s openly racist attacks on four Democratic congresswomen of color were clear signs of electoral anxiety.

Trump insists he is producing great results for the country, especially on the economy. And yet, at the price of provoking great backlash, he moved in an unprecedented manner this week to portray four nonwhite Democratic representatives as fundamentally un-American, not only ideologically, but also racially and ethnically.

In so doing, Trump has telegraphed that, ahead of 2020, he hopes to focus at least as much on the jagged divide of “Who is a real American?” as on the traditional question incumbent presidents seeking reelection highlight during generally good economic times: “Are you better off than you were four years ago?”

That choice may reflect the convergence of inclination and calculation. Trump’s instinct is to center his politics on cultural and racial conflicts that pit Americans uneasy about the nation’s changing identity against those who welcome or accept it. But Trump also faces clear evidence that he may be unable to build a winning coalition with just the voters satisfied with his performance in office. That’s evident even with an economy that’s booming, at least according to measures such as the low unemployment rate and the soaring stock market.

The latest such evidence comes in a new study released today by Navigator Research, a consortium of Democratic research and advocacy groups. The report, provided exclusively to The Atlantic, examines a group that many analysts in both parties believe could prove to be the key bloc of 2020 swing voters: Americans who say they approve of Trump’s management of the economy but still disapprove of his overall performance as president. And it shows Trump facing significant headwinds among that potentially critical group, partly because of the divisive language and behavior he’s taken to new heights, or lows, since last weekend—tweeting about the congresswomen and encouraging his supporters to attack them as well.

“The main takeaway from this analysis is that while some Americans might be giving Trump positive marks for his economic performance, they are strongly held back by three things: the values that they have, the views they have on other noneconomic issues, and some very real concerns about Trump’s character and temperament,” says Bryan Bennett, an adviser to Navigator Research.

This conflicted group looms so large over 2020 because about half (or even slightly more) of voters express support for Trump’s management of the economy, but only 40 to 45 percent of them give him positive marks on his overall performance. That difference could be the tipping point between a coalition that places Trump close to the comfort zone for presidents seeking reelection—support from about half of Americans—and one that leaves him trying to secure a second term with positive marks from a much smaller circle. The only presidents since 1952 who sought reelection with approval ratings below 50 percent—Gerald Ford, Jimmy Carter, and George H. W. Bush—all lost.

The conflicted voters, if they break for Trump, bring him “in range” to win, says the GOP pollster Gene Ulm: “He’s incredibly close. Can I predict that he’s going to win Michigan, Wisconsin, Pennsylvania? No, I can’t do that. But he is within spitting distance of that 47 [or so] mark he needs to win when you look at these chunks of people.”

These voters consistently register as a substantial group. Since April, polls from CNN, Quinnipiac University, and ABC/The Washington Post have found that between 16 percent and 19 percent of Americans who approve of Trump’s handling of the economy still disapprove of his overall job performance. That’s a very high disparity by historical standards: The NBC/Wall Street Journal poll has found that Trump’s approval rating among Americans who say they are satisfied with the economy is running 16 to 20 percentage points lower relative to the approval ratings of George W. Bush and Barack Obama.

The Navigator research, based on the cumulative results of all six national surveys the group has conducted since the 2018 election, defines this group slightly more narrowly: Its results put about 10 percent of Trump’s voters, or 6 percent of all voters, in this camp. (The reason: Navigator, compared with most public polls, records a slightly lower approval rating for Trump, both overall and on the economy.)

The added value of the Navigator research is that, by gathering results across several surveys, it provides a polling sample large enough to look more deeply than the public surveys themselves have done at the attitudes and characteristics of these divided voters. In other words, the results present a sharper picture.

Nearly two-thirds of the conflicted voters are men, compared with just under half of the overall electorate. Compared with all voters, they are also more suburban (60 percent), slightly wealthier (70 percent earn $50,000 or more), better educated (45 percent have college degrees), and somewhat younger (56 percent are under 50 years old). Fully one-third of them, more than might be expected, are nonwhite.

Less surprisingly, these voters are loosely rooted in their political views. A much higher share of them than the electorate overall identifies as independents or moderates (about 45 percent in each case). In 2016, these voters reported preferring Hillary Clinton over Trump by a seven-point margin, though fully 30 percent said they voted for a third-party contender—much higher than the population overall. But in 2018, these same voters broke sharply against Trump: They backed Democratic candidates in House elections by a resounding 20 percentage-point margin.

In Navigator’s polling, the economy emerges clearly as Trump’s greatest advantage. Though Democratic strategists such as the pollster Stanley Greenberg and the super PAC Priorities USA believe they can dent Trump’s edge by arguing that the economy still isn’t delivering for many families, for now these conflicted voters give the president a crushing 55 percentage-point edge over congressional Democrats when asked which side they trust more to handle the issue. That’s an imposing lead on the concern that most academic models from political scientists and economists consider the biggest factor in deciding presidential elections.

But on every other front, Trump faces headwinds. In the surveys, these voters prefer congressional Democrats over Trump to handle taxes (by nine points), immigration (by 10 points), and health care (by 34 points).

The conflicted voters also return negative verdicts on key measures of Trump’s character. Three-fourths of them say he is looking out for himself, not the country, compared with about three-fifths of the electorate overall, Navigator found. It recorded a similar disparity when it asked whether he is bringing more corruption to Washington or reducing it. “There are a lot of character things, but you also see policy [resistance],” says Bennett, the associate director of polling and analytics at the Hub Project, a liberal advocacy organization. “That does speak to the fact [that] these voters do have a nuanced view of Trump: They are willing to give him credit, but they are holding back from giving him full approval because of questions about other policies and his character.”

Navigator didn’t track other measurements of Trump’s personal behavior, but results from Quinnipiac University’s national poll in March trace some of voters’ doubts confronting Trump. Among voters who said they approved of his economic performance, fully one-third still said he is not honest; about one-fifth said he does not have good leadership skills or care about average Americans; and 44 percent said they do not consider him a good role model for children.

Those numbers notwithstanding, Ulm doesn’t consider the obstacles Trump faces with these voters insurmountable: “He’s creeping up with them.” Like most Republican strategists I’ve spoken with, he sees three keys to Trump converting conflicted voters: focusing their attention on his economic record, soothing their concerns about his behavior and rhetoric, and painting Democrats as ideologically extreme. “When you separate Trump out on it, and you look at the commentary on these voters, they are by no means liberal voters,” Ulm says. “You have a lot of people [who] like everything he’s doing, but would never have him [over] for dinner.”

With his openly racist and xenophobic attacks on four Democratic congresswomen, Trump this week has elevated the third goal at the price of impeding (if not incinerating) the first two. Still, Ulm sees little risk for Trump with these conflicted voters in deploying such language. “I don’t think it affects moving them one way or another,” he says. “So much of how he talks … is just built into his stock price.”

But with polls showing that a clear majority of Americans consider the remarks racist and un-American, Democratic strategists predominantly believe Trump is narrowing his audience by alienating voters otherwise receptive to his economic record. “Is some of Trump’s racial insensitivity baked in? Sure,” the Democratic pollster Andrew Baumann says. “But this is so much more explicit than he’s been before. I think it will have an impact on some of those conflicted voters. I don’t think the bottom is going to fall out and he’s suddenly going to drop to 35 percent approval, but I think it just drives home all of those concerns these folks have about his conduct and elevates that portion of the equation above the economy.”

Where both sides agree is that the key measure for Trump next year will be the share of voters who approve of his overall performance, not the (for now) wider group that gives him good marks on the economy. “They have to become approvers,” Ulm says. For incumbents, he adds, “job approval is your vote—it’s almost like religion.” The latest NBC/Wall Street Journal poll dramatically underlines that point. It found that in 2020 matchups against any of the four major Democratic contenders—former Vice President Joe Biden and Senators Bernie Sanders of Vermont, Kamala Harris of California, and Elizabeth Warren of Massachusetts—Trump drew support from a minuscule 5 percent or less of voters who disapproved of his overall job performance as president. (Trump led each Democrat by at least 83 percentage points among voters who approve of his performance.)

That’s actually in line with recent history: Exit polls found that Obama carried only 3 percent of voters who disapproved of his job performance in 2012, and Bush won just 6 percent of them in 2004. The notion, common especially in White Houses seeking reelection, that large numbers of voters may dislike the incumbent but vote for him anyway—because they dislike the challenger even more—is simply not supported by electoral evidence. Bush and Obama, in fact, each lost a higher share of voters who approved of their performance than they won among those who disapproved.

The central role of the president’s overall approval rating in deciding his reelection fate underscores the conundrum that Trump so vividly demonstrated this week. Across a wide array of public and private polls, he’s not consolidating nearly as much support as previous presidents with voters satisfied with the economy, many of them financially comfortable, suburban, and college-educated. (Those were the same voters who turbocharged the Democratic sweep of affluent metropolitan-area congressional districts in November.) The CNN and ABC/Washington Post polls show Biden winning just under one-fifth of voters who say they approve of Trump’s economic performance, a much higher level of defection than Bush or Obama suffered among the economically satisfied.

By all evidence, those defections are driven at least partly by the divisive confrontations Trump constantly stirs over race, gender, and culture. And yet Trump feels compelled to keep fueling those fights, as he repeatedly did this week, in part because the fires he has already lit may have permanently repelled too many of the voters satisfied with his economic record. He has to double down on stirring turnout from his base through racial and cultural strife to offset his underperformance with swing voters alienated by exactly that behavior. It is as if Trump is on two diverging roads: He has already moved so far down the path of centering 2020 on American identity that he can no longer realistically cross back to focusing it primarily on the economy. He fights over American identity not only because he likes to, but also because, by this point, he must.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.