I’m going to start out a bit abstract, so hang with me here.
The easiest way to conduct a poll is to call up a bunch of people, ask if they’re going to vote, and then who for. This generally gives you the wrong answer. Why? Because, for a variety of reasons, you usually don’t wind up with a sample that looks like the electorate. That’s why most public pollsters who know what they’re doing also ask for things like race, age and gender then weight their sample to what they expect the electorate to be. One explanation for why polling was generally accurate in 2016 except in the Midwest is that an education gap appeared for the first time and Midwest pollsters didn’t know they had to weight for education. The highest quality polling, mostly done in well-financed campaigns isn’t really polling as we think of it; it’s modeling. The campaign will identify many discrete voting blocks, poll how they intent to vote, and then assemble the expected electorate using each of those kinds of voters.
The reason why this is done is because voters don’t move as a whole. Different types of voters move in different directions. That’s something we all intuitively know, but often forget to take into account when thinking about elections. Consider the special elections that have taken place since November. They’ve been notable not for flipping any chamber, or even making too much of a difference in terms of who holds the seats, but for showing a general trend of Democrats doing a lot better than would be expected. While low turnout special elections are generally a problem for us, we’ve been, in general, overperforming presidential toplines. That’s big news, and suggests good things for us in the coming midterms. However, despite there being a full 31 special elections completed so far, we don't have many indications for how some important constituencies are voting.
To start with, not all of those races have been contested. 7* races lacked a Republican and 4** lacked a Democrat, although one of those for each party came from getting shut out of runoffs. Of the 20 remaining races, 3*** of them took place in very heavily democratic districts where we had no room to grow, and little ability to fall, so they’re not as interesting. I suppose you could look at turnout, but I’m not convinced comparing special election turnout to the turnout in different special elections is a valid move, especially for races that are essentially already decided. That leaves us 17 interesting races. While not as good of a haul as 31, it’s still enough to show us some trends about different constituencies.
Specifically, I’m going to break the districts down by race . I should note that the figures for race are not exact. They are based on the 2010 census, and in some cases are approximated from precincts that were split to make the districts. “Other” means all races other than than black or white.
district |
white vap |
black vap |
other vap |
swing |
MN-HD-32B |
97 |
0 |
3 |
23 |
MT-al |
89 |
0 |
11 |
14 |
nh-hd-carroll 6 |
98 |
0 |
2 |
11 |
ct-hd-68 |
93 |
1 |
6 |
-23 |
ct-sd-32 |
94 |
1 |
5 |
8 |
nh-hd-hillsboro 44 |
91 |
2 |
7 |
-1 |
ny-ad-09 |
87 |
3 |
10 |
39 |
ok-hd-28 |
76 |
6 |
18 |
45 |
ia-hd-89 |
85 |
7 |
8 |
35 |
ga-sd-32 |
78 |
8 |
14 |
0 |
ks-04 |
84 |
9 |
7 |
20 |
TN-HD-95 |
82 |
9 |
9 |
12 |
ct-hd-115 |
70 |
11 |
19 |
1 |
de-sd-10 |
71 |
17 |
12 |
4 |
va-hd-85 |
67 |
21 |
12 |
-6 |
sc-hd-84 |
71 |
22 |
7 |
14 |
va-sd-22 |
70 |
27 |
3 |
1 |
In case you’re wondering about that 18% in OK-HD-28, it’s mainly Native voters who tend to vote like the white voters in that part of the state. The 19% other in CT-HD-115 is heavily but not exclusively Hispanic.
The first thing you might notice is that there haven’t been many elections with a significant black population. This does make a certain amount of sense. Many of the the districts we won uncontested were black majority. Still though, we wound up with a set of elections that seem uniquely poorly suited to telling us how black voters are voting, including a special election in the second least black congressional district in America. Even for the districts with a large black population, other factors make things harder to figure out. VA-SD-22 had a major third party candidate, and precinct splitting elsewhere makes comparisons to presidential results difficult. Overall, this table has a lot to say about, say, white voters in the midwest, but less to say some other important voters, particularly minority voters. That’s where SC-05 and its large number of black voters comes in.
Tonight there will be two special elections. Frankly, the finals results GA-06 will be kind of boring. I mean, sure finding out the winner will be exciting, and a million thinkpieces will be launched as soon as AP calls the race. But we already basically know what the result will look like. The race will basically be tied, and the voting patterns are pretty established. Even though it’s pretty clear who the winner will be, SC-05’s final results will tell us more about the current state of electoral politics. It’s a race that involves 5 major kinds of voters, 4 of which we don’t really know what their voting patterns will be like
urban Black voters
- Where they are important for 2018: Our coalitions in the midwest, the south, and the mid-Atlantic are highly dependent on them. Also Nevada.
- How things are looking: 2016 was a bad year for black turnout, and Patrick Ruffini has seen some warning signs in the GA-06 primary results. To the extent municipal elections are predictive of anything, signs have not been great. Also, the blacker districts have not had as rosy results for us so far.
- Where they’ve been represented so far in specials: A few precincts in Wichita. There would have been a few more without precinct splitting.
- Where they are in SC-05: Rock Hill mostly. I suppose Gaffney as well, although that’s pushing the definition of urban. This category of voters is not actually that plentiful in SC-05.
rural black voters
- Where they are important for 2018: There are very few black voters in rural areas outside the south, but in it, especially in the deep south, the black belt is vital to us.
- How things are looking: Same as listed above, but with the added caveat that for every other race, we performed worse among rural voters in 2016. It’s unlikely that our margins will fall, but it’s conceivable that turnout may.
- Where they’ve been represented so far in specials: Functionally nowhere. The only contested race that had a rural black population, VA-SD-22, also had a third party candidate who made analysis hard.
- Where they are in SC-05: Basically everything south of Rock Hill is rural and has a significant black population, so quite a lot actually.
southern urban white voters
- Where they are important for 2018: Nashville, Atlanta, NC, and New Orleans mostly. There aren’t that many, honestly. There was a huge white flight in the civil rights era, and old patterns have stuck. Many who stayed are fairly liberal.
- How things are looking: There’s not really much to be concerned over or celebrate so far, but making gains in the medium to small southern cities would be really nice.
- Where they’ve been represented so far in specials: Nowhere.
- Where they are in SC-05: Again, just Rock Hill and maybe Gaffney if you’re being generous.
southern suburban white voters
- Where they are important for 2018: Let’s talk bigger than 2018. When the suburbs across the country started moving towards us in the 1990s, the southern suburbs stayed put, which is basically why the south is so Republican right now. During the Obama years, the diversification began to activate a leftward pull. 2016 was important because the white suburbs moved left too, even in famously conservative “fiery pit of hell” type places like Shelby County, AL. Doing better in the white southern suburbs could win us most of the southern states.
- How things are looking: So far there’s not too much to go on, but things are looking optimistic. You’ve probably heard all about GA-06 by now, but the other two southern suburban districts went well for us. It’s hard to say if that’s white voters, POC, or both driving the change, or if the change is even real.
- Where they’ve been represented so far in specials: GA-SD-32, SC-HD-84, and TN-HD-95.
- Where they are in SC-05: We’ll get the suburbs of booming Charlotte but also the old-school, more static outer suburbs of Columbia.
southern rural white voters
- Where they are important for 2018: Most of the south.
- How things are looking: That’s tricky. These voters took a hard, hard right during the Obama years and were Trump’s base of support in the primaries, so you’d expect them to stick with him. Looking back to Nixon, they stuck with him right up until the impeachment. During the waning years of the Bush administration, when the entire country was turning to the Democrats the fix things, rural southern whites mostly didn’t. But most of them have voted Democratic in their lives, so we also just don’t know.
- Where they’ve been represented so far in specials: Nowhere. That Oklahoma district wasn’t in the southern part of Oklahoma.
- Where they are in SC-05: A lot of the district is southern rural whites. Upstate is very white, and downstate has a heterogeneous mix of whiter and blacker rural areas.
For some of these demographic groups, this may the be only chance we get to see how a large number of them are voting before regular elections. That’s why tonight GA-06 may eat up more of my attention, but I’m going to be looking to SC-05 much more in the coming days.
*VA-SD-09, VA-AD-71, AL-HD-67, CT-HD-07, LA-SD-02, AL-HD-58, and CA-34. CA-34 was the district they got shut out of a runoff
** LA-HD-08, LA-HD-42, LA-HD-92, SC-SD-03. We got shut out of the runoff in LA-HD-92, sort of. What actually happened is complicated but not interesting.
***NY-SD-30, CT-SD-02, and PA-HD-197, the last of which was so Democratic that when a residency issue caused us to turn to a write-in campaign, the only candidate on the ballot, a Republican, came in 3rd.