Fivethirtyeight studied every election since Trump -- 99 special elections plus regular state elections in NJ and Virginia -- and checked whether there were any strong predictors of whether Trump voters would support a Democratic candidate.
They found that while there is a small advantage to Democrats running in working-class districts that voted Trump in 2016, there were so many exceptions that no clear generalizations could be drawn. There was even less order in patterns of voters in suburban vs urban districts.
Fivethirtyeight's conclusion is that it's hard to use the elections since 2016 to know where Democrats should concentrate their fire to flip GOP seats in the 2018 midterms.
But they don't try to analyze whether policies played any role in the outcome of these elections -- for example, did Democrats fare better when they adopted right-wing policies that attracted DCCC endorsements, or when they ran on a Warren/Sanders left-wing alternative? To me, that's the main question we need to answer in planning tactics to take Congress away from the GOP in 2018: how do we frame a progressive alternative and win elections with it? Taking seats from Republicans by running Democrats who vote for Republican legislation isn't much of an accomplishment.
But there’s a massive caveat: The relationship in each case is quite weak.8 You can see that in how wide the variation is in Democratic overperformance. In areas that shifted toward Clinton, Democratic margins have been up to 36 percentage points better than the partisan leans of their districts would lead us to expect — and they’ve been as much as 37 points worse. In places where the median household income is less than $50,000, Democrats have run ahead of their presidential candidates by as much as 61 points and run behind by as much as 37. On average, Democrats are doing better in working-class areas than in suburban ones — but the dozens of examples to the contrary make a blanket statement like that almost worthless (and certainly not something you should base a midterm prediction on).
In short, special elections aren’t really telling us whether the 2012 or 2016 map is a better picture of current American partisanship. Our best guess is that the 2018-and-beyond map will be a hybrid of the two. Some of the voting patterns we saw emerge in 2016 may stick around, but the 2012 map still holds plenty of sway. We’ve already seen some reversion to the mean in “Trump country.”
Be Skeptical Of Anyone Who Tells You They Know How Democrats Can Win In November [Nathaniel Rakich/Fivethirtyeight]
(via Naked Capitalism)