Clinton in Brooklyn, which is not in the South. (Andrew Renneisen/Getty Images)

There’s a weird dynamic at play in the Democratic primary contest. Hillary Clinton has taken great pains to question whether or not Bernie Sanders is actually a Democrat, noting that he’s served in Congress for decades as an independent. Sanders, on the other hand, is critiquing Clinton for appealing to less-liberal voters. In other words, Clinton hammers Sanders for being outside the party and Sanders attacks Clinton for straying from liberal ideology.

Sanders’s most common articulation of Clinton’s non-liberalness is by dismissing her biggest wins as coming from more-conservative areas. He’s repeatedly waved away Clinton’s wins in the South as being from a more conservative area, as he did during last week’s debate in Brooklyn. “No question about it,” he said on Thursday. “We got murdered there. That is the most conservative part of this great country. That’s the fact.”

Depending on how you divvy up the “parts of the country,” Sanders is correct. Each year, Gallup tracks the number of people in each state who identify as conservative or liberal and ranks where the gap between the two is widest. In 2015, Alabama was the most conservative state in the union, and Sanders’s home state of Vermont the least conservative. If we consider “the South” as defined by the Census Bureau (as on the map below), it is the most conservative part of the country.

That blurs a point that’s often been made in response to Sanders: He’s won a lot of conservative states, too, like Idaho, Oklahoma and Wyoming. Clinton has done well in a number of very-conservative states, but so has Sanders. In fact, if you take Vermont out of the mix — a very liberal state that backed its senator by a wide margin — there’s essentially no correlation between the conservativeness of the state and the margin of victory for either candidate.

Anyway, the conservativeness of the state is different than what Sanders is implying. He’s implying that Clinton won because those voters are more conservative. The data above, considering the states on the whole, doesn’t differentiate between Democrats and Republicans. It’s the whole state. So it’s worth considering if Sanders’s specific point is correct.

It’s clear from exit polling that Clinton has done better with more moderate voters nationally.

She wins with every group overall, including very liberal voters (of whom there are more than in years past). And generally, as the gap widens between the number of moderate voters in a state and the number of very liberal ones, Clinton does better.

But not always. In Oklahoma, for example, Sanders won big despite it being a state with more moderate primary voters. Oklahoma drags the overall trend downward, as Vermont did in the graph above.

Regardless, the average Southern state for which we have exit poll data has a differential between moderate and very liberal voters that’s about twice that of states in the Midwest.

There may be a reason for that. Those Southern states are also states with a large percentage of black voters. The most obvious indicator of the outcome of a Democratic contest has been how many black voters there are in the state. Where there are more, Clinton does better.

In March, we took data from the General Social Survey to point out that white voters were more likely to identify as liberal than black ones. After that piece, Jacobin Magazine’s Matt Karp used the same data to consider how black voters felt about the issues at play in the primary.

We can look at both of those things regionally, too, of course. Considering only those people who identify as Democrats (strongly or weakly), we can see how each region compares in its willingness to, first, identify as liberal versus conservative and, second, respond to a central question of Democratic politics: How big a role the government should play in problem-solving.

Southern Democrats were less likely to identify as liberals since 2000 — but northeastern Democrats were about as likely to identify as “conservative.” (We’ve noted before that Democrats are far more likely to call themselves moderate or conservative than Republicans are to call themselves moderate or liberal.) What’s more, there’s no real difference between Democrats by region on the question of government’s role in society.

Another way to look at it: Elected representation. VoteView tracks the partisanship of members of Congress over time, mostly by considering that same question of government’s role. Southern Democrats in the House were once more liberal than other Democrats and then, until fairly recently, much more conservative. But as the South has moved to the right and Democrats have lost seats there, Democrats from southern states have grown increasingly similar to their colleagues from other parts of the country in terms of ideology.

So here’s what we know. We know Clinton does better with people who identify as more moderate, and we know that she did very well in a part of the country where Democrats are both slightly less likely to identify as liberal and much more likely to be black. All of this gets tangled together: Are Democrats in those states more moderate because they are more black?

We also know that the conservativeness of the state on the whole bears no relationship to the results of the Democratic primary.

Sanders’s goal in suggesting that Clinton is the choice of conservative voters is a bit different than Clinton’s goal in suggesting that real Democrats support her. Clinton wants to inspire Democrats to rally to her side more than they already are. Sanders, on the other hand, wants voters to believe that he’s the preferred candidate among Democratic voters, an argument that necessarily means that he must somehow explain away getting demolished in the South.

Sanders could dismiss the results there more easily with a more direct argument: He got demolished in the South because there are more black voters. That’s impossible to argue with. But one can also see why it’s not his preferred argument.