Gov. Matt Bevin’s (R) apparent loss in his reelection bid — the race hasn’t yet been called — probably doesn’t augur doom for President Trump’s reelection bid in the state next year. Bevin won in 2015 by about nine points, less than a third of the margin by which Trump won the state a year later. By now, if you’ve paid any attention to Tuesday’s results, you know the other complicating factors: Bevin’s unpopularity, broad indifference among Kentucky Republicans to his leadership and so on.

But there is one way the results this week mirror presidential politics. As Yphtach Lelkes of the University of Pennsylvania noted on Wednesday morning, the results of Bevin’s reelection bid were far more polarized on urban-rural lines than he had seen in 2015. As it turns out, that split looked a lot more like 2016, a presidential election year.

The results didn’t look much like 2016, of course. If we map out the shift in the vote by county from 2016 to 2019, there was a universal shift in Democrats’ direction.

If we compare the 2019 results to the 2015 race, though, the pattern is a bit different. The eastern part of the state voted much more heavily Democratic; the western part, more Republican. But notice the coloration of the counties themselves. That reflects county population density, a proxy for the urban-rural split. In some of the darkest counties on the map, there was a big shift to the Democrats.

In fact, the vote shifted more rapidly to the Democrats the more densely populated the county was. (Notice that on this and following graphs, the horizontal scale is logarithmic. Circles are scaled to the total vote in 2019.)

The way Lelkes visualized that was by comparing the percentage of the two-party vote earned by Bevin (that is, the number of votes for Bevin as a percentage of all votes for him and his Democratic opponent) and population density in counties.

In 2015, that relationship looked like this, with more densely populated counties being more likely to vote Democratic.

In 2019, the same comparison looked like this.

The point is that diagonal line, showing the trend as population density increases. It’s much steeper in 2019 — meaning that the relationship between how densely populated a state was and how much it did or didn’t support Bevin was much stronger.

Notice that the dots, which represent different counties, drop down between 2015 and 2019. That reflects the drop in Bevin’s share of the vote — and why he’s now trailing. The dots drop faster on the right side of the graph, showing how more densely populated counties moved away from Bevin more significantly. (This, of course, is also what the line represents.)

Compare Bevin’s fate with Trump’s in 2016.

Which line does the one from 2016 more strongly resemble, that of Bevin in 2015 or Bevin in 2019?

If we insert that graph into our animation, the answer becomes obvious. The urban-rural polarization in 2019 looks a lot more like 2016 than 2015, contributing to Bevin’s apparent loss.

Since the 2016 election, we’ve seen a pattern of particularly suburban areas voting more heavily Democratic. What happened in Kentucky is probably related but interesting in its own right: Turnout relative to 2016 was higher in more heavily Democratic areas.

Which is another reason that 2019 won’t look like 2020: Next year, the vote in Kentucky will presumably be at presidential-year levels.

It is, however, one reason Bevin almost certainly won’t serve another four years.