Keith Srakocic / AP

CLEVELAND—The tumultuous 2016 presidential race appears poised to realign the states at the tipping point of American politics.

Since Bill Clinton’s first victory in 1992, Democratic presidential candidates have consistently run better in the aging, predominantly white and heavily blue-collar swing states clustered in the Rustbelt than in the younger, more diverse, and increasingly white-collar swing states arrayed across the Sunbelt. That pattern, in fact, has largely shaped presidential races since passage of the Civil Rights Act of 1964 fractured the century-long Democratic control of the “solid South.”

But now public polls and private assessments alike show Donald Trump running more strongly against Hillary Clinton in several of the key Rustbelt battlegrounds than in their Sunbelt counterparts.

With blue-collar whites providing the core of Trump’s support, the Rustbelt has emerged as his most—and perhaps only—plausible path to an Electoral College majority. Simultaneously, Democrats are increasingly viewing Sunbelt states that not long ago were considered safely Republican as the closest thing to a firewall for Clinton, largely because of the resistance to Trump among minorities and white-collar whites. “This shift was probably coming anyway because of the changing demographics of the Sunbelt, but Trump radically accelerates it on both ends,” said long-time Democratic strategist Paul Begala, a senior adviser to the pro-Clinton Priorities USA political action committee.

The trend was underlined last week by the release of a flurry of NBC/Marist College/Wall Street Journal polls that showed Trump even with Clinton in Ohio and just narrowly behind in Iowa, while lagging by virtually identical margins of six to nine percentage points in Virginia, North Carolina, and Florida in the Southeast, and Colorado in the Southwest.

Political strategists in both parties generally rank those four states, plus Nevada, as the five true swing states in the Sunbelt. New Mexico, a sixth Sunbelt state, was competitive during the George W. Bush years, but both sides now place it safely in the Democratic camp. Together, the five most competitive Sunbelt states offer 72 Electoral College votes (led by Florida with 29).

Likewise, strategists identify five Rustbelt states as key battlegrounds: Ohio, Iowa, Wisconsin, Pennsylvania, and Michigan. (The final state that both sides have treated as competitive in recent elections is New Hampshire, which is geographically distant from both clusters, but demographically and economically more closely resembles the battlegrounds in the Rustbelt than the Sunbelt.) The five Rustbelt battlegrounds offer 70 Electoral College votes (led by Pennsylvania with 20).

Just gaining ground in the Rustbelt won’t be enough for Trump if he can’t contain his losses in the Sunbelt. Even if Trump flipped all five of the Rustbelt swing states—each of which backed President Obama over Mitt Romney in 2012—he would still lose the election if he can’t win any of the five Sunbelt swing states, assuming no other states changed hands.

The shifting balance in these two key groups of swing states reflects, above all, the changing demography of each party’s electoral coalition. The working-class white voters underpinning Trump’s support represent a much larger share of the vote across the Rustbelt than the Sunbelt, whether using data from the exit polls or the Census post-election survey (which consistently identifies non-college whites as a large share of the electorate than the exit polls do). The Census results identify non-college whites as at least 54 percent of 2012 voters in all five Rustbelt states—and no more than 44 percent in any of the Sunbelt states. And except for Florida, the Rustbelt states, as a group, are also generally older, at a time when Republicans are dominating among whites older than 45.

Contrasting dynamics are molding the Sunbelt states. Diversity is the most important. The non-partisan States of Change project has projected that from 2008 to 2012 the minority share of eligible voters will rise by more in each of the Sunbelt swing states than in any of the Rustbelt battlegrounds, with the biggest increases registered in Nevada (a stunning 7.3 percentage points), Florida (4.5 points), and Colorado (4.3 points). In each of those states except Colorado, the States of Change model projects that the white share of eligible voters will dip below 70 percent in 2016. By contrast, the model projects that whites will still comprise at least 80 percent of eligible voters this year in all of the Rustbelt swing states except Michigan (where it will dip only to 78 percent).

Democrats have also benefited because college-educated whites, who have been grown warmer toward the party since the 1990s, generally comprise a larger share of the total white vote in the Sunbelt than Rustbelt battlegrounds. While the Census found that whites holding at least a four-year college degree represented only a third or less of all white voters in 2012 in Ohio, Iowa, and Michigan, they comprised at least 45 percent of all whites voting in Virginia and Colorado. (College whites clustered at 36 percent of the total white vote in Pennsylvania, Wisconsin, and Nevada and slightly above that in Florida and North Carolina). This dynamic may prove especially important this year with most national polls showing Clinton even or ahead of Trump among college-educated whites—a group no Democratic presidential nominee has carried in the history of modern polling dating back to 1952.

It’s possible that this evolving geography could tip other states from both buckets  into the competitive category. When the Trump campaign recently identified 17 states as 2016 battlegrounds in a private briefing for a group of congressional Republicans first reported by the Wall Street Journal, it included reliably Democratic Minnesota in the Rustbelt as a state it hoped to contest. It also identified Republican-leaning Arizona and Georgia in the Sunbelt as diverse states it would need to defend. (Somewhat incongruously, it also included red-tinted Missouri and Indiana from the Rustbelt as states it thought it might need to protect.) But at this point few Democrats—including inside Clinton’s campaign—say they believe any of those will become truly competitive.

Obama’s two victories signaled growing Democratic strength in the Sunbelt, but despite some early party fears, displayed no erosion in the Rustbelt: Compared to the 1992 to 2004 period, Obama improved on the Democratic performance in both groups of states. This year, though, the polling raises the possibility that Trump could advance in the Rustbelt and retreat in the Sunbelt, leaving Clinton potentially more reliant on the latter than the former. That would mark a major reversal.

In the six elections since 1992, Democrats have won the five Rustbelt battleground states 27 times out of 30 chances. Pennsylvania, Michigan, and Wisconsin are all bricks in the Democrats “blue wall,” meaning the party has carried them in each election since at least 1992. Democrats also carried Iowa five times (losing only to W. Bush in 2004) and Ohio four times (with Bush narrowly carrying it both times).

By contrast, over the past six elections, Democrats have carried the five Sunbelt battleground states just 13 times. Over that period, Nevada is the only state in this group that Democrats have carried at least four times; they have won Florida and Colorado three times each, Virginia just in President Obama’s two victories, and North Carolina only in 2008.

Perhaps even more strikingly, the average share of the vote that Democrats have carried since 1992 in each of the five Rustbelt swing states is greater than their average performance over that period in any of the five Sunbelt swing states (although the gap is tiny between Ohio, their weakest performer in the former group, and Florida, their best in the latter).

In fact, these patterns trace back much further than 1992, when Bill Clinton’s first victory signaled the emergence of a Democratic coalition that has carried the popular vote in five of the past six presidential elections. The Rustbelt has been relatively friendlier terrain for Democrats than the Sunbelt in almost all elections since the passage of the landmark civil-rights laws of the mid-1960s triggered the realignment of the white South to the GOP.

Before the civil-rights era, for instance, the Democratic presidential nominee won a higher share of the vote in North Carolina than in Iowa in every election from 1932 through 1960. But since then, the only Democrat to run better in North Carolina than Iowa was southerner Jimmy Carter in 1976 and 1980. Likewise, Democrats have run better in Iowa than in Virginia in every election since 1964 except 1980. Over that same period, only Southerners Carter (in both 1976 and 1980) and Clinton in 1992 ran better in North Carolina than in Wisconsin. No Democrat has won a higher share of the vote in Virginia than in Wisconsin since Adlai Stevenson in 1956. Perhaps most important, Democrats have captured a higher share of the vote in Ohio than in Florida in every election since 1964 except for three involving Southern nominees: Carter in 1976, Clinton in 1996 and Al Gore in 2000.

Although it was not as affected by the racial polarization that reordered politics in the South, the West reveals a similar story: The Democratic presidential nominee has performed better in Wisconsin than in Colorado in every election since 1960, and better in Iowa than Colorado in each contest since 1976.

Today, though, it’s common for Democratic strategists to privately assess that Florida will be an easier state for Clinton to win than Ohio and maybe even Pennsylvania, and that Virginia, Colorado, and even North Carolina might be easier than Iowa and possibly Wisconsin. Democrats appear especially focused on North Carolina. Charlotte served as the site of Clinton’s first joint event with Obama, and has also received more advertising than any other market except Cleveland, according to NBC tabulations.

Although public polls have, inevitably, produced somewhat dissonant results, they largely reinforce those private verdicts.

Across the Rustbelt, the most recent polls by Marist and Quinnipiac Universities both show Trump and Clinton tied in Ohio; the latest Marist survey gives Clinton a slim three-point lead in Iowa while a Monmouth University poll put Trump narrowly ahead there. The Marquette University Law School poll released last week gives Clinton a four-point lead over Trump in Wisconsin. Results in Pennsylvania have diverged, with Quinnipiac, which generally produces better results for Republicans, showing Trump slightly ahead and Marist giving Clinton a comfortable nine-percentage point advantage. Although there has been little public polling in Michigan—a state where Democrats have averaged almost 51.6 percent of the vote since 1992—that’s generally considered a more distant reach for Trump.

In the Sunbelt, recent polls by Marist, Fox News, and Monmouth all show Clinton leading comfortably in Colorado, and surveys from the first two organizations give her a solid lead in Virginia. Polls by Marist and the Democratic group Democracy Corps also show her ahead in North Carolina. Monmouth recently showed Clinton narrowly leading in Nevada (Democracy Corps had her tied.) The greatest divergence has been in Florida, where Quinnipiac placed Trump ahead, but Marist, Democracy Corps, and Project New America, another Democratic group, have all showed Clinton leading.

In some ways, Democrats have been living on borrowed time in the Rustbelt. Even as the GOP has extended its inroads among working-class whites, every Democratic presidential nominee since Al Gore in 2000 has exceeded his national share of the vote among non-college whites in all five of the Rustbelt battlegrounds, the exit polls have found. But the appeal to those voters—especially non-college white men—of Trump’s insular and combative message on trade and immigration could end that act of political levitation: Both the Marist and Quinnipiac surveys, for instance, found Trump leading Clinton by about 20 percentage points among non-college-educated Ohio whites. Clinton will also be challenged to match the enormous African American turnout that Obama generated in Rustbelt states such as Ohio. (Trump, though, could provide a powerful incentive: The new Marist polls found him attracting literally zero support from blacks in both Ohio and Pennsylvania.)

Across the Rustbelt, the deciding factor may be whether Trump partially or entirely offsets the ground he’s likely to gain among blue-collar whites by retreating with their white-collar counterparts. As the veteran conservative electoral analyst Henry Olsen told me earlier this year, “If you do better, for instance, in rural Wisconsin but significantly worse in Madison and small suburbs of Green Bay and the exurbs of Milwaukee ... you don’t do overall better.” Indeed, while the recent Marquette poll showed Trump holding a solid ten-point advantage over Clinton among non-college whites in Wisconsin, it also showed her leading him by 16 percentage points among whites with at least a four-year college degree; that was far greater than Obama’s three-point lead over Mitt Romney among Wisconsin college-educated whites in 2012, according to exit polls.

White-collar whites also loom as a pivotal factor in the Sunbelt swing states: Trump will face difficult odds if he can’t run competitively in places like Northern Virginia, the Denver suburbs of Jefferson and Arapahoe counties, the I-4 corridor from Orlando to Tampa in Florida, and the white-collar neighborhoods in Charlotte, Raleigh, and Research Triangle Park across North Carolina.

Minority turnout will likely be the other critical dynamic in these states. Democrats feel especially bullish on Virginia and North Carolina because most of the minority voters there are African Americans, who most party leaders view as more reliable Clinton supporters than Hispanics, both in their likelihood to vote and their affection for her. Despite all the antipathy among Hispanics toward Trump, a surprising number of Democratic strategists privately worry that their enthusiasm for Clinton appears modest. Yet Trump’s fierce attacks on undocumented immigrants have provided Democrats so much fuel that even most Republican strategists are bracing for elevated turnout—and widening deficits—among Hispanics in Florida, Nevada, and Colorado. That’s one reason the Sunbelt could prove much tougher terrain than the sort of Rustbelt state where Republicans will begin gathering for their convention today.

Atlantic assistant editor Leah Askarinam contributed.

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