Okay, so there's been this stark difference between the Generic Ballot and the results we've been seeing in Special Elections. The Generic Ballot has been hovering around 6-8 points for the Dems, but we're seeing an average of a
17 point shift left in the Special Elections. The weird thing is that the generic ballot and Special Election data with a large n are both generally good predictors of election results. (Provided the special elections aren't too far away from said election)
This has been commented on by a few folks like Cook and Silver, but there hasn't been a lot of concrete explanation. I usually pour over the Morning Consult crosstabs (because I'm a dork), but today was the first time I looked at the rural/suburban/urban subs.
It looks like Suburban Voters are currently
+12 for the Democrats (Silver handicaps MC as 3 points to the right, so that's up to a 15 point advantage...though that's not exactly how handicapping works for subs). That would explain what we're seeing. The Special Elections have largely been in regions with large suburban areas (plus, typically, large rural areas). Between Trump's approval rating softening a bit with rural voters (though still very strong), the increased energy on the left, and the dramatic shift for Suburban voters- that 17 point shift left starts to make sense.
I know this isn't the most mindblowing revelation, but it does help when trying to figure out which way the winds are actually blowing for the midterms.