The States Swing Both Ways
If you're like me, you've probably been tripping on this New York Times data visualization since Election Day:
It depicts the change from the county-level presidential vote in 2020 to the county-level Senate vote in 2022. And it's pretty fascinating. In an election that was supposed to have a strong Republican swing, but ended up looking like just a mild red shift, we see that the shift was not at all uniform. Some states (including Colorado and Pennsylvania) swung hard left, while others (including Florida, New York, and Connecticut) shifted to the right, at least as judged by the presidential-to-senatorial votes.
Nate Cohn offered a smart interpretation of this, suggesting that, at least in part, these results were driven by states where the population felt most threaten by rapid changes in abortion access and in states where Republicans nominated election deniers for high office. To be sure, there were also important differences in candidate quality in different states.
But I was curious about the degree to which this midterm election was an outlier. Do we normally see states going off in different directions in the Senate elections that follow a presidential election?
To get a sense of this, I collected state-level presidential election returns from 2016 and 2020, and compared those to state-level senatorial election returns from 2018 and 2022. I was interested here in the Democratic share of the two-party vote in these contests. (I omitted states with top-two election systems, as well as those with strong third party candidacies. I also only included those 2022 contests in which we have at least 80% of the votes collected.) The data are available here.
The figure below shows the Democratic Senate vote shares from 2022 minus the Democratic presidential vote shares from 2020 in each state. And yes, consistent with the analysis in the Times, some states moved more into the Democratic column (to the right in the figure below) while others saw a more Republican vote. The standard deviation of these shifts is 3.31.
Now compare that to the shifts from the Democratic presidential vote share in 2016 to the Democratic Senate vote share in 2018:
The big difference is that almost every state saw a Democratic shift going into 2018, which isn't too surprising in a year in which Democrats gained 40 House seats. But the dispersal of these data points is still pretty striking, ranging from Utah Senate Democratic nominee Jenny Wilson doing nearly five points worse than Hillary Clinton did in that state, to Joe Manchin doing 24 points better than Clinton did in West Virginia. The standard deviation of these figures is 5.46. Take out Manchin and it's still an impressive 4.15, far greater than that of the other graph.
The lesson to take from this is that Senate elections may often diverge from presidential ones, and that divergence can vary a great deal from state to state. There was actually more divergence in 2018 than we're seeing this year. It's just that the overall trend was pretty strongly Democratic four years ago, while this year it's pretty close to zero.