John Minchillo / AP

What are the tipping points in the states likely to  decide the 2016 presidential election?

Bellwether counties in swing states show that the demographic gulf  between the Democrats’ more urban coalition and the Republicans’ base of rural and blue collar whites is poised to grow ever larger in 2016.

To pinpoint the factors that could decide the places at the fulcrum of the struggle between Hillary Clinton and Donald Trump, The Atlantic, working with Polidata, an election analytics firm, examined voting trends in several key swing states to explore how their vote has shifted in recent elections.

In Ohio, Florida and Colorado, the analysis compared the county-by-county vote in 2004, the last time a GOP presidential nominee carried the state, with the results during President Obama’s victories in those states in 2012; in North Carolina, it compared the results between 2008, when Obama won the state, and 2012 when Mitt Romney took it.

This analysis found a clear pattern that suggests the stark demographic divides of the 2016 race will accelerate and intensify the patterns of voter allegiance already visible in recent elections. Looking across the four states, the greatest increases in Democratic votes consistently emerged from among the largest counties, places that almost universally had a high minority population, above-average numbers of college graduates or both. Simultaneously, over the same period, Republicans typically deepened their advantage in smaller, more heavily white counties across these states.

The tumultuous 2016 race is on track to deepen both of these trends. With Donald Trump facing widespread rejection by minority voters and underperforming literally every Republican nominee in the history of modern polling among college-educated white voters, he faces the risk of voters further retreating from the party in the most densely populated metropolitan areas across the swing states.

But with his continued support among white blue-collar and non-urban voters, he may hold, or even expand, the Republican advantage beyond the metros. This history suggests the formula for Hillary Clinton is to run up the score in the most populous places. Trump’s hope in these states rests more on death by a thousand cuts — small increases in margin across many smaller counties that overcome erosion in the metro areas. That was largely the formula that allowed Romney to tip back North Carolina in 2012 despite persistent deficits in the biggest places.

The maps below capture these changes. For Ohio, Florida and Colorado, they measure the increase in the total number of Democratic votes by county from 2004 to 2012. In each case, the map divides the states into three bands: counties where the Democratic increased their total vote by more than the state average; counties where Democrats saw an increase but one that was less than the state average; and counties where the Democrats won fewer votes. In North Carolina, the map tracks the same changes for the Democratic vote from 2008 to 2012.

Ohio

Florida

Colorado

North Carolina

Despite these states’ many differences, these results show some strikingly similar patterns. One is that Democrats have relied overwhelmingly on the most populous places for their gains. In Ohio and North Carolina, over half of Democratic gains came from their improvement in just the three most populous counties; in Florida and Colorado, half of the gains came from just the five largest counties.

In Ohio, the state’s most populous counties — namely those surrounding Columbus and Cincinnati — saw vast increases in Democratic votes that overwhelmed the losses elsewhere between John Kerry’s defeat in 2004 and President Obama’s victory in 2012. Democrats enjoyed their biggest increase in Franklin County of Columbus: it provided 60,000 more votes for Obama in 2012 than it did for Kerry.

Hamilton County, which includes Cincinnati, provided 20,000 more votes for Democrats, ultimately swinging from red in 2004 to blue in 2012. And while Cuyahoga County in Cleveland was an exception, with a decrease of almost 1,300 Democratic votes, the county still provided the Democrats with the most votes in the state in 2012, nearly 450,000 votes—and that Democratic loss of 1,300 votes was minor compared to the GOP’s loss of 31,000 votes in that county in the same year. In all, the three Ohio counties with the greatest increases in Democratic votes from 2004 to 2012 contributed an additional 90,000 votes to the Democratic Party. That surge overwhelmed smaller counties where Democratic support declined.

Similarly, in Florida, the counties where Democrats gained the most were Orange County, which includes Orlando, and Miami-Dade County. In Colorado, the list was topped by Denver County, Arapahoe County (outside Denver) and El Paso County (Colorado Springs). While El Paso County still swung to the Republicans, the GOP’s increase in votes between 2004 and 2012 equaled about a quarter of that of the Democrats.

The counties where Democrats recorded their greatest gains reflect their increasing reliance on a coalition that revolves around college-educated whites and non-white voters. Whites represented less than their statewide share of the population, for instance, in 13 of the 18 counties in Ohio where Democrats gained the most from 2004 to 2012 and 10 of the 16 in Florida.

Likewise, the counties with the greatest growth in Democratic votes almost all had elevated shares of college graduates. Of the 18 counties in Ohio with the greatest gains for Democrats, 15 exceeded the state average in the share of college graduates. Florida saw that same trend — in the 16 counties where Democrats gained the most votes, 14 ranked above the state average in college graduates. In Colorado, eight of the 11 counties with the greatest Democratic growth at least as many college graduates as the state average.

When Democrats did gain in heavily white counties, it was almost uniformly linked to education. Just five of the Ohio counties where Democrats gained most from 2004 to 2012 ranked above average in their white population; four of those five also exceeded the average in four-year degrees. In Florida, just six of the 16 counties where Democrats gained most were above average in white population — and all six were also above average in bachelor’s degrees. Likewise, of the 11 Colorado counties where Democrats made the biggest gains, in six counties the white population exceeded the state average; there, too, all six also had at least as many bachelor’s degrees as the state average.

The counties where Democrats lost votes since the last GOP victory represented a demographic bookend to the ones where they advanced. In Ohio, whites exceeded their share of the state population in 38 of the 51 counties where Democrats lost votes from 2004 to 2012. In Florida, whites were over-represented in 15 of the 20 counties where Democrats lost ground. In Colorado, the six counties where Democrats lost ground from 2004-2012 all exceeded the state average in their share of whites.

The counties that lost Democratic votes were also far less likely to be college-educated. Of the 51 counties that lost Democratic votes in Ohio, just 12 had an above-average share of the population with college degrees; the same was true of just three of Florida’s 20 counties that lost Democratic votes, and just one of Colorado’s six counties that lost Democratic votes. All of these results underscore the class inversion among whites that has reshaped the electoral competition.

North Carolina

North Carolina is the one state on this list that tilted from blue to red. But from 2008 when Obama carried the state, to 2012, when it went to Romney, it largely followed the same demographic tracks as the other swing states that flipped in the opposite direction. The key difference in North Carolina was that Obama lost more ground in smaller, mostly white counties than he gained in the big, more diverse counties.

In all, from 2008 to 2012, Obama increased his vote in 39 of North Carolina’s 100 counties. In 18 of the 25 counties where Obama gained the most, minorities exceeded their average share of the population; college graduates exceeded their statewide representation in five of the remaining seven.

But Democrats lost ground in the same sort of counties where they retreated from 2004 to 2012 in Ohio, Colorado and Florida. From 2008 to 2012, Obama lost votes in 61 North Carolina counties. Of those, more than two-thirds had more whites and fewer college graduates than the state average. The difference between North Carolina and the other states is that the volume of Republican gains outside of the metro areas slightly exceeded the quantity of Democratic gains inside of them. For Obama it was that death by a thousand cuts: in the North Carolina counties where he ceded ground from 2008 to 2012, the largest loss was less than 2,000 votes; the average loss in those counties was just 558 votes.

With Clinton’s prospects dim of recapturing much ground in rural and small-town North Carolina counties, her key to her success in the state is further widening the Democratic advantage in the big metro areas, like Charlotte and Raleigh––a priority that has been clearly in evidence from the travel schedule that she, Vice President Tim Kaine, and other top surrogates like Michelle Obama has followed in the state.

In North Carolina, as in Ohio, Colorado, Florida and the other major swing states, the election will likely unfold as a test of strength between metropolitan areas breaking for Clinton and non-metropolitan areas rallying to Trump.

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