The Georgia Sixth Race Will Be Won (or Lost) in Cobb County
For those that would rather watch than read, here is the video version of this post:
At long last, the Sixth District runoff is almost here. The in-person early vote has started. After only six days, the runoff numbers surpassed the primary and were on pace for turnout in line with the 2016 general.
If you read further into that Nate Cohn Twitter thread, he looks at the makeup of the electorate. Currently, it doesn’t appear that this massive turnout is helping either side. Rather it appears to be cancelling each other out. As is always the case, take these analyses with a grain of salt. This is all modeled and predicted behavior based on the voter file and extra information Nate Cohn has. However, I tend to agree that this turnout doesn’t favor either side yet. In fact, higher turnout here could help the Handel campaign.
I took some time to dig into the results from the primary to try to understand how turnout affected the outcome. I’ve heard statements like “the Ossoff campaign was a turnout machine,” but I like looking at the details. And those details throw a bit of cold water on the idea that the turnout helped the Ossoff campaign. Even so, there are large nuggets of hope for Ossoff.
Here’s the summary table for the primary results. This is old news as of now, but there are some interesting takeaways just from this.
43.47% turnout is fantastic for a special election and Cobb County running away with 47.46% turnout is an interesting tidbit.
Ossoff’s worst performance was in Cobb County. This was mostly expected given that this part of Cobb County is strong GOP territory. However, Handel’s performance didn’t keep pace in Cobb. Take a look at the difference between Cobb and DeKalb. They’re near opposites in the totals. Cobb was about 41% Ossoff/59% Non-Ossoff and DeKalb 59/41. Handel’s performance was 16.61% in DeKalb and 18.46% in Cobb. There was an 18 point swing away from Ossoff from DeKalb to Cobb. Of those 18 points, Handel only pulled away two.
In Cobb County, Handel had 31.4% of the non-Ossoff vote. Compare that to 42.5% in Fulton and 40.1% in DeKalb. Cobb County saw the highest turnout, the lowest Ossoff vote, and Handel’s lowest share of the non-Ossoff vote. What all this means is that Handel needs to watch out for Cobb County. There was definite excitement in Cobb, but the excitement turned into votes against Ossoff but not for Handel. She’ll need to work on getting those voters from Cobb to come back out and vote for her after having not done so in the primary. There’s a lot packed into the precinct results, so I’ll break this down piece by piece and also show where Ossoff struggled in the primary.
This is a chart of every precinct in the Sixth split into % Ossoff (blue), % Handel (orange), and % Other (gray). Each point is a precinct, and the size of the circles represent the numbers of registered voters (as of the primary) in each precinct.
There are a few takeaways here. First, the long tail of low turnout, high % Ossoff precincts is drastic. There is only one precinct with less than 35% turnout where Ossoff received less than 59% of the vote. It’s never certain if those who didn’t vote would’ve voted for Ossoff by the same margin, but the “turnout machine” probably left votes on the table here by not pushing the turnout in these precincts.
The negative correlation (.33 r-squared [and as a reminder, r-squared measures how well data fit a model from 0 (no fit) to 1 (perfect fit), here’s Wikipedia if you want more on that]) between turnout and the Ossoff share of the vote is impacted greatly by these precincts with less than 35% turnout, but the trend is there. It does flatten out at the highest turnouts, but, at a minimum, the higher turnout didn’t automatically result in higher returns for Ossoff.
Second, this chart would indicate the highest turnout was probably impacted by those voting for someone other than Handel or Ossoff. The % Other vote has a decent positive correlation (.33 r-squared) with turnout. The trendline for Handel hits a peak at about 47%. At turnout higher than 47% the vote share for Handel slightly decreases. At the same time, the share of votes for all other candidates continues to increase leaving that little bit of white space between the orange and gray points.
Given that the highest turnout came from Cobb County and Handel’s lowest share of the Non-Ossoff vote was also in Cobb County, it’s obvious that most of these precincts were located in none other than Cobb County.
The interactive map above is currently styled by turnout from light green (low turnout) to dark green (high turnout). If you’re zooming in to the west part of the district, you’ll notice there are some empty spaces there. These are split precincts that distort the turnout percentage given that not all of the registered population was able to vote in the Sixth District. I’ve left those out as I’m most interested in turnout here.
You can see here that the hotspots for turnout were in north DeKalb and almost all of Cobb County. There are, of course, other scattered spots with high turnout, but the consistency in Cobb County is impressive.
Let’s break this down even further.
Ossoff —
The above chart is the same information as the first, but I’ve only included the Ossoff numbers and then colored them by county. You can see the distinct county differences with DeKalb giving Ossoff the largest share of the vote. The low Fulton points filtered in the Cobb points are mostly the North Fulton precincts where Handel did quite well. It is interesting to note that Cobb is the most contained county with only one precinct having less than 35% turnout. Beyond this, the graph doesn’t show any particularly new insights as compared to the original, but it has added meaning when plotted on the map.
These images show a filtering of the precincts. The first image is the whole district shaded by vote share (light to dark). The second is precincts with higher than average turnout. The third is precincts that have both above average turnout and above average vote share. For reference, the shading of the precincts is based on the filter not the original, so the colors will change from picture to picture.
The darker blues show the hot spots for Ossoff in the streaks up the middle of DeKalb and Fulton counties. The second image that includes precincts in the top 50% of turnout almost reads as a negative of the best Ossoff precincts. Both of the streaks through DeKalb and Fulton are mostly absent. This is made even more evident by the last image which filters down even further to only include precincts that were above average % Ossoff. Of the 201 precincts in the district, only 29 had above average turnout and above average % Ossoff.
Even more than just the number of precincts, the number of voters this includes is important. There were 433,665 registered voters in the precincts included (again ignoring those split precincts). These maps are shaded by registered voters. Similar to above, the first is all precincts, the second is the top 50% of precincts by registered voters, and the third is the second filtered by the top 50% of precincts by %Turnout.
Notice the dark shade of Cobb County. Then compare the second map here to the second turnout map. There is a decent similarity between these. Again this map cuts away significantly at the best Ossoff precincts. Similarly, the last map does away with almost any of Ossoff’s top 50%. In fact, here’s the last map filtered down to Ossoff’s top 50%:
There are only 13 precincts here comprising about 18,000 registered voters in DeKalb, 13,000 in Fulton, and 2,500 in Cobb. Of the largest precincts with the best turnout, Ossoff only did well in precincts totaling about 7.9% of potential voters. I will say that there is little to no correlation between the vote share or turnout with size of the district, but it is important to remember that the number of precincts is less important than the number of voters and votes.
All of this together should give Ossoff supporters both pause and hope. If this was the best of the Ossoff “turnout machine,” then there has to be concern as to what can be done to stop the united GOP front that coalesced behind Handel. But to counter that, there is obvious potential here. There were definitely votes left on the table for Ossoff. If the maps filtered by % Ossoff can be filled in with more precincts, then he has a chance to win.
Non-Ossoff —
As the Non-Ossoff maps essentially show the opposite of the Ossoff maps, these are mostly included for reference. One point I want to make here is the similarity between the second and third maps. Other than lower DeKalb, the maps are nearly identical showing again that positive correlation between turnout and % Non-Ossoff.
But the real intrigue is when you split this vote into % Handel and % Other.
Handel —
A few things to notice in this Handel chart. First, the grouping of Fulton County precincts at the top-center of the chart are the north Fulton precincts that Handel won. Next, the correlations are interesting. DeKalb (.50 r-squared) and Fulton (.27 r-squared) have a slight upward correlation. Obviously, Fulton does start decreasing at the top end, but I believe this is more an effect of those north Fulton precincts. Without those, the quadratic r-squared increases to .47. It’s not fair to say they’re outliers, so they’ll remain in. And it might be the case that the Ossoff “turnout machine” really took hold in Fulton and not anywhere else.
Either way, compare both DeKalb and Fulton to the near flat line of Cobb County and an r-squared of .08. Even with the high turnout, Handel only broke 25% in one of the Cobb precincts. And once again, her share of the vote started decreasing after after about 47%. Mapping these numbers again gives further insight.
In similar fashion to Ossoff, Handel’s strongest precincts in North Fulton disappear when you look at the top half of turnout. Even so, she holds on to more precincts in that final filter. And these precincts were large enough to have an impact. Here’s the map filtering down to the largest precincts:
Handel did better in the larger precincts with higher turnout. These 28 precincts left comprise nearly 20% of the voting population. Of course, the best Handel precincts didn’t get any higher than about 42%, but we have to assume much of the other vote is coming her way. The issue for her is that the area with consistent high turnout (Cobb County) didn’t vote for her.
Other —
If I were Handel, comparing this chart and maps to her own would make me a little nervous. The Fulton and DeKalb trends are very similar to Handel’s, but this isn’t the case in Cobb County. The % Other in Cobb jumps to the top of the chart here and actually has a decent upward trend. As has become obvious, Cobb has consistently high turnout, but looking at this final map filtered down to the largest precincts it becomes clear that Cobb not only has consistent turnout, it also has consistent turnout in some of the largest precincts voting strongly for someone other than Handel and Ossoff:
In fact, the precincts in the top half of turnout comprised of about 95,000 Cobb registered voters, 76,000 in Fulton, and 48,000 in DeKalb. Cobb influenced turnout more than any other county and saw the highest average of %Other. Of Cobb precincts in the top half of turnout, Other averaged 41.2% compared to 18.7% Handel and 40.1% Ossoff. 41 precincts are present in this final map comprising about 30% of the registered voter population.
Even with all that, these final three charts show just how divergent Cobb was as compared to Fulton and DeKalb and why this is important for Handel.
This first chart plots % Other on the x-axis and % Handel on the y-axis. We would assume that as the % Other vote increases that the % Handel would increase as well. She was the highest GOP vote getter, but as we can see, this is mostly true for both DeKalb and Fulton counties yet Cobb doesn’t follow the same pattern. The same holds true when we plot % Handel against % Non-Ossoff.
This chart may be the best evidence that Cobb is following a different pattern than the rest. We would expect even higher correlations in this chart than the last given that as the % Non-Ossoff vote goes up the second-highest vote getter should be going up as well. Those voting for someone other than Ossoff voted for Handel more than anyone else.
This again holds true for the Fulton and DeKalb counties (.69 and .79 r-squared respectively), but the blob of orange that is Cobb County only has a .14 r-squared. For counties other than Cobb, % Handel and % Non-Ossoff were well correlated. Handel had minimal impact on the total Non-Ossoff vote in Cobb. Compare that to this final graph:
This plots % Non-Ossoff against %Other. The correlations are much more substantial (.9 DeKalb, .82 Fulton, and .73 Cobb). This is partly to be expected given that Handel was only about 38% of the total Non-Ossoff vote, but the difference between the Cobb numbers is the most interesting.
Cobb voters showed up with the highest consistency against Ossoff. But that excitement against Ossoff didn’t rally behind Handel. This part of Cobb County is of course mostly covered by Judson Hill’s former seat in State Senate District 32. And that’s seen in his strong showing in Cobb County. Hill did endorse Handel shortly after the primary.
It’s likely that the Republican vote will coalesce behind Handel, but you have to think that the excitement shown for Hill and candidates other than Handel won’t translate so easily to showing up on June 20 for Handel.
Final Questions
There are two major questions to be answered in the runoff:
- Will Cobb County’s consistent turnout for other candidates in the primary translate to high turnout for Handel?
- Will the Ossoff “turnout machine” be able to increase turnout in his best precincts and reverse the negative correlation from the primary?
It’s impossible to answer these questions now. The early vote turnout is substantial, but as mentioned before, looking at who has already voted doesn’t present any obvious advantage for either candidate. However, Ossoff was less than 4,000 votes away from breaking 50% in the primary. If the answer to question one is a “no,” then Handel probably loses. But even a “yes” isn’t a guaranteed victory. Holding everything else equal, a “yes” answer to both questions probably means an Ossoff victory, but that’s not certain. Ossoff is working against history in this district, and early vote turnout on a general election pace might mean all of this is thrown out the window.
Either way, on Election Day I’ll be watching the Cobb County returns with the most interest. My gut tells me that the efforts of the candidates in DeKalb and Fulton will cancel each other out, and this race will be won (or lost) in Cobb County.
