On Monday, Democrats outperformed Hillary Clinton in two state legislative special elections in Minnesota (senate district 54 and state house district 23B). These races didn’t generate the flashy headlines that some others have—neither seat changed hands and Democratic overperformance was below average—but they’re part of a broader pattern. On average, Democrats have been outperforming Hillary Clinton in special elections.
So far, most quantitative analyses of special elections (including my own) have focused on this Democratic overperformance and what it means for the midterm elections, but that’s not the only interesting part of the story. How the overperformance is spread out is also notable. Specifically, Republicans are underperforming their baseline by larger amounts in the Trumpiest districts. To see this, first look at how much the Democratic candidate in the special election overperformed (or underperformed) Hillary Clinton.
It’s not difficult to see the trend in this graphic (G. Elliot Morris generated a version of this graphic last week). In state legislative and congressional special elections, Democrats have seen bigger overperformances compared to 2016 in redder districts. (Note that the circles are sized based on the number of votes for major party candidates).
Maybe more interestingly, they’ve also outperformed Clinton in areas where Trump outperformed Mitt Romney.
In other words, Democratic overperformance has been, in general, larger in areas where Trump beat Romney. I don’t have demographic data for all these districts, but it’s worth noting that in general Trump outperformed Romney in areas with high concentrations of non-college educated (often rural) white voters.
When you feed both variables—Trump’s victory margin and how he compared to Romney—into a model that predicts Democratic overperformance in specials, both have a statistically significant effect. The fit isn’t perfect; for example, a lot of the variance in how much special election results differ from the baseline goes unexplained in the model. But the basic pattern is simple: Republicans are, in general, underperforming in red districts and districts that have voters who swung from President Obama in 2012 to Trump.
Turnout could explain this underperformance. There’s evidence that some Trumpy Republicans—rural, white, blue collar voters—failed to turn out for Ed Gillespie in Virginia’s gubernatorial race or Roy Moore in Alabama’s Senate race, and that Republican turnout may be down more generally. (It’s worth nothing that Moore may be a somewhat special case, since he was credibly accused of having improper sexual contact with teenagers while he was in his 30s, and a more generic Republican may have seen better results.) That would be consistent with a picture of an electorate where Democrats are excited to vote against Trump and Republicans aren’t ecstatic about the president.
Part of the issue may also be vote share. The 2012 presidential results are, on a district-by-district level, a better predictor of recent special election results than the 2016 presidential election. And, as David Shor has suggested, that could mean that Democrats are winning back a number of voters who preferred Obama over Romney but voted for Trump rather than Clinton.
And these explanations aren’t mutually exclusive: Both turnout and vote share typically matter in elections. It’s possible that Democrats are both turning out their voters and bringing some Trump-converts back into the fold.
So should Republicans panic? Are Trump-friendly, blue-collar voters going to abandon Republicans? It’s way too early to say, and special elections alone won’t answer these questions. But if the GOP does end up having a broader problem with turnout, it could hurt them in the midterm elections. As Kyle Kondik points out, there are two open seats in Minnesota (the 1st and 8th congressional districts) where Trump won over Obama voters and a fall-off among that group could hurt the GOP.
This scatterplot shows Trump’s share of the two-party vote (Trump did better on districts in the right, Clinton did better on the left) and the difference between Trump and Romney’s two-party vote share (higher points indicate a greater Trump overperformance, and lower points indicate an underperformance) for every seat with a retiring congressman. (Red dots represent a retiring Republican, and blue dots represent a Democrat). The data can be found here and here.
The graphic doesn’t include all retirements; Pennsylvania retirements are omitted because the details of the map are up in the air. But the point is that some vulnerable districts have Trump-Obama voters in them. The Minnesota 1st and 8th and the New Jersey’s 2nd are all examples of places with numerous Obama-Trump voters. And, as Dave Wasserman points out, a district doesn’t have to be chock-full of non-college-educated whites for a drop in their turnout level to make a difference.
Obviously, this isn’t a foregone conclusion—Monday’s special elections were in Minnesota, after all—but the broader pattern here suggests that Republicans could have problems in key races if they don’t hold onto voters that Trump grabbed in 2016.
Source : http://www.weeklystandard.com/trump-voters-may-not-show-up-in-2018.-and-some-may-become-obama-voters-again./article/20115751040