Democratic presidential candidate Bernie Sanders told a crowd in Warren, Mich. about his fundraising strategies and how he differs from frontrunner Hillary Clinton. (Reuters)

This post has been updated.

Here is a party trick. Tell someone that you can predict the result of a Democratic primary in advance, by asking one question. "Ooh," the people around you will say, and give you money and attention. The question you ask? What percentage of the state is black?

If the state is more than 10 percent black: Easy call. Hillary Clinton wins it. Under 2.5 percent black? You're probably safe saying that Bernie Sanders will triumph. Anything in between, and you can guess -- but this should pretty much have you covered.

We had noted after New Hampshire -- which seems like four years ago by itself -- that the upcoming map looked favorable for Clinton because more states had more non-Hispanic white voters. We noted that most states had grown less white over time, and mapped the next states to vote.

Now, we can overlay a series of results from actual elections in those states.


The less-heavily white states backed Clinton. The whiter ones went for Bernie. Iowa's something of an outlier, we'll note. But the correlation is strong.

That chart is actually a little misleading, because it shows white voters, not black ones. Hispanics are less strongly supportive of Clinton than black voters. If you compare results to the percentage of the state's Democratic voters that are black, you can almost predict the margin of victory -- not just who will win.


(Where exit poll data from 2008 or 2016 was not available, we used black population density.)

There's no similar split on the Republican side. Donald Trump has been doing consistently well across demographic groups, while the Democrats are much more split.

Stepping back to look at the forest, this is why Clinton is likely to be the nominee. There are several big states still to come with large black populations. If your parlor trick is any guide, those are states that Clinton should carry.

The post was updated to clarify the data used in the second graph.