In the Democratic presidential contest, there have been a number of clearly defined splits between the two candidates. Hillary Clinton is supported more heavily by older voters and by women; Bernie Sanders, by younger voters and men. But the most important split between the two has been race. In most states, the numbers of young or old voters and the numbers of men and women are pretty consistent. But the number of black voters in particular varies widely -- and that's largely what's responsible for Clinton's essentially insurmountable lead in the race.

We can even visualize that disparity. Here's the density of the black population in states that have voted so far. (We cheated and moved Nevada in a bit.)


And here's Clinton's margin of victory (or loss) in each county against Sanders.


If we take all of the counties that have voted on both sides, we can compare how candidates do in counties that are more- and less-heavily white, black or Hispanic relative to the overall average. (The Census Bureau counts race and ethnicity separately. References to white and black population density are the percentage of the population in each county that was white or black and not Hispanic.)

In other words: How much support does Hillary Clinton get in counties that have a higher percentage of white voters than average, and what support does she get in counties with a lower percentage?

The answer, as you might expect, is that she does much better in less-white counties. Sanders does better in more-white counties. But we did the same analysis for Donald Trump.


We'll get to Trump in a second. Let's dive a bit deeper into Clinton's numbers.

If we plot Clinton's support in each county versus how white it is, there's a pattern: Her support falls as whiteness goes up.


(The dashed line is the trend.)

And as the density of the black population increases, so does her support.


Remember: This isn't voters. This is people. You can see how clearly the relationship works by clicking through the images below. (Hit the buttons.)

All of that was a set-up for those Trump numbers, which almost certainly caught your eye. Trump does better in less-Hispanic counties -- which certainly fits with your likely expectations.

But it's not that simple. The graph contrasting population density with support here is less clear-cut than Clinton's above.


As it turns out, this is mostly a function of the fact that a lot of heavily Hispanic counties have been in Texas -- which Trump lost to Ted Cruz overall.

You can compare the density of the Hispanic population to his support. Note the very tip of Florida, the only county where Trump lost in the state. It's where Sen. Marco Rubio is from -- but it's also heavily Hispanic.

(We checked numbers for the other candidates, too. Cruz does better with more-densely Hispanic counties for the same reason that Trump does badly with them: They're in Texas.)

The split on blacks for Trump is more interesting. Trump does better in counties that have a higher density of black residents. And again, we're not talking about voters. In the Deep South, where Trump did particularly well, the percentage of the Republican primary vote that was black was low: 4 percent in Alabama, 6 percent in Georgia and Mississippi.

But while the correlation isn't terribly strong, it exists.


We knew that Trump did very well in more-heavily black states, but now we learn that he did well in more-heavily black counties, too -- but not so well that it pops out at you the way Clinton's maps did. Note the counties along Mississippi River, though, heading up from Louisiana. That region, as it did for Clinton, backed him strongly.

He did well in Appalachia, as well, in eastern Ohio and western Virginia, but the link between white density and his support wasn't as strong as it was in places where a lot of black people live.


We don't yet have enough data to know why areas with a larger black population might lean toward Trump. It seems safe to assume, though, that if the choice in November is between him and Clinton, areas with large black populations are not likely to vote Trump.