Once considered a firewall for Hillary Clinton, Nevada has sharply turned into a tight and unpredictable contest for the former secretary of state as senator Bernie Sanders steadily gains support from critical voting blocs. (Alice Li/The Washington Post)

If you were to scoop up a handful of Democratic primary voters from any one of the early primary states, they'd generally contain a pretty similar demographic mix, for the most part. 2016 so far is shaking out like the electorate in 2008: Slightly more women, generally older, a more liberal electorate now than then, and a mix of educational attainment, but not huge differences.

Until you looked at race. In some states, like Iowa and New Hampshire, everyone would be white. In others, like South Carolina, most of the voters would be black.

Quinnipiac University released a new poll on Wednesday which shows how those various groups look at the two remaining Democratic contenders. The biggest gulfs between Hillary Clinton and Bernie Sanders fall along ideological, age and racial lines.

Since ideology and age don't vary as much between states, let's set them aside for a second. Let's instead focus on race.

Iowa and New Hampshire, as we've noted before, are far whiter than most other states -- including Nevada and South Carolina, the next two states to vote in the 2016 contest. A new CNN/ORC poll in South Carolina shows that Hillary Clinton has a gigantic lead among non-white voters in the state, and therefore a gigantic lead overall. South Carolina is the first brick in what's become known as Clinton's "firewall" -- the states with big non-white populations where she should have a big advantage.

How much less white is South Carolina than Iowa (where Clinton and Bernie Sanders tied) or New Hampshire (where Sanders won by a mile)? This much.

Which would lead one to suspect that Nevada, which is also much less white than Iowa and New Hampshire, might look similar. CNN/ORC also polled there -- and it doesn't.

There wasn't a large enough non-white population for CNN to break out separately, but its polling director (former Postie Jennifer Agiesta) noted in in her analysis. "Although the pool of potential caucus-goers in Nevada is more racially diverse than those who participated in Iowa or New Hampshire," Agiesta wrote, "the racial divide among likely caucus-goers isn't nearly as stark as among voters in South Carolina, with both white and non-white voters about evenly divided between the two candidates."

About evenly. But why?

It appears that it's in part the composition of that non-white electorate. South Carolina is heavily non-white because it is heavily black. Nevada is heavily non-white in part because it is heavily Hispanic. (As we will always note: Hispanic is not considered a racial identity by the Census Bureau, but an ethnic one. So it identifies white Hispanics and non-Hispanics, the latter being represented in the large graph above.)

Both Quinnipiac and Gallup have broken out candidate favorability by ethnic group, allowing us an interesting comparison. Quinnipiac has two big buckets: white and non-white. Gallup looks at white, black and Hispanic. (All of the figures below relate to Democrats and Democratic-leaning independents.)

On net, non-white voters look at Clinton much more favorably than do white voters. Among white voters in Quinnipiac's data, Clinton has a fairly lousy plus-41 net favorability, which means that she's viewed 41 points more favorably than she is unfavorably. Among non-whites, the split is +72. That's actually about the same as Sanders's split among non-whites -- but among whites, he sees a +77 favorability.

Gallup's data includes a lot of people who said they didn't know enough about the Vermont senator to offer a rating. So the net favorabilities are lower, given that the percentages viewing him positively and negatively are lower.

We can adjust those numbers though, figuring out what percentage of those with an opinion view him positively or negatively. And what we find is that black Democrats view him with some antipathy -- but Hispanics view him about as positively as do whites.

Which may explain the gulf between the non-white voters in Nevada and South Carolina.

The bad news for Clinton is that this may undermine the strength of that firewall. In general, non-white voters have her back, but that may depend to some extent on those non-white voters being black voters. In states with a large Hispanic population, it's not clear (at this point/from limited polling) that she can count on the same level of support. Which could be important in a state like Texas, were a third of the Democratic primary vote in 2008 was Hispanic -- though so far, according to polling from Public Policy Polling, Clinton still leads in Texas thanks to her support from black voters. As we've seen in Nevada, though, that can change quickly.

Or put another way: Clinton's firewall may be missing some bricks.