One thing Mr. Sanders has going for him are the anti-Obama Democrats.
Mr. Sanders has generally fared poorly in the South and Appalachia, but he has tended to do better in a surprising spot: areas where there are large numbers of the old registered Democrats who vote Republican in presidential elections, but nonetheless find themselves trapped in a Democratic primary thanks to a closed or semi-closed system.
These conservative voters appear to be choosing Mr. Sanders in big numbers. You can see the traces of it in the stark increase in Mr. Sanders’s support when you cross from Georgia into the Florida Panhandle, a state with a closed primary and party registration. You can see it along the borders of Oklahoma, and along the North Carolina border as well. It shows up in another way: the large numbers of voters who are voting for “uncommitted” or a minor candidate.
It’s not entirely clear whether these voters actually support Mr. Sanders. The exit polls in Oklahoma showed Mr. Sanders winning big — 59 percent to 24 percent — among the large number (28 percent) of voters who wanted the next president to change to less liberal policies. That doesn’t necessarily mean they don’t support Mr. Sanders; perhaps they think of liberalism in cultural terms, like racial issues or guns. But it does raise doubts. Mr. Sanders won easily among the 17 percent of voters who trusted neither candidate in an international crisis.
The model accounts for this with a surprising variable: Barack Obama’s share of the vote in the 2012 Democratic primary. That’s not a typo. As president and running without a major opponent, Mr. Obama won just 57 percent of the vote in Oklahoma in 2012, thanks to these conservatives who still vote in Democratic primaries.
The pattern is good news for Mr. Sanders in West Virginia and Kentucky, with the model putting him over the top there, even though Mr. Sanders has lost almost all of the counties bordering these two states. Without this variable, Mrs. Clinton would be favored in both states.