Swing State Spotlight: Georgia
I was going through some quick analysis of the swing states this election, and GA jumped out at me as having a clear (and unusual) pattern. This is less a post with a clear point I want to drive home and more a wander through some post-election analysis.
Here’s the graph that made me pause:
The Y axis here is presidential turnout as percent change compared to 2020, purely on raw number of votes. I was originally using this metric because I was digging into the “people who voted in 2020 stayed home” narrative, and I do realize that it’s not accounting for population change. More on that later.
Background
GA is interesting, because its turnout rate is overall quite high, due to implementing Automatic Voter Registration in September. According to the folks at the Center for Election Innovation and Research,
In the first four years after AVR’s implementation, the active voter registration rate increased by 20 percentage points to 98% of eligible Georgians in 2020, up from 78% in 2016
This is an enormous increase! The program that automatically registers people when they interact with the local DMV-equivalent has been very successful, and the whole state operates at a fairly high turnout level as a result.
This high baseline makes turnout analysis within the state fun, because you can fairly clearly say that differential turnout isn’t related to the strength of registration programs or efforts by parties. It’s also interesting in terms of the relationship between turnout and election results, because while there was a bit of a leftward shift in the state from 2016-2020, that shift has been far from decisive and didn’t overcome the national trend this year.
GA In This Election
Anyway, the thing that stopped me with this graph is that while raw turnout is up a bit from 2020, there are some clear spikes in turnout in heavily-Republican counties. The state as a whole beat their 2020 turnout record, so it isn’t a matter of turnout hanging on in some counties in an overall low environment, it’s more like truly dramatic high turnout numbers.
The GA election data hub has a very nice dashboard that I cannot for the life of me figure out how to download cleanly, so I just ran some spot checks on the counties that really popped, to ensure I wasn’t just picking up fully on population growth. From what I can tell, this 2020-2024 turnout change is more muted once you correct for population growth, but not directionally different. When you look at the state as a whole, for which I can get citizen voting eligible population figures, the total turnout as % CVAP goes from 66.8% in 2020 to 60.8% in 2024.
Black %
The obvious first thing to check is Black population percent, which highly but not perfectly correlates with Democratic vote share. This is pulled from the 5 year ACS numbers, and I’m using the “Black only” category (and implicitly ignoring “2 or more races”) compared to the total population of the county in that same ACS file. The ACS has an annoying thing where the 1 year files update faster but drop counties under a certain population threshold, which makes analysis like this difficult. The 5 year average covers all counties, although it does not reflect quite as recent trends. I think this is probably a fine tradeoff for this analysis.
It’s not clear to me that Black percent is a better explanation for the turnout differences than Democratic percent (and I did check white percent as well, just in case, which was not any more useful). I also ran some quick regressions (separately, because dem percent and Black percent are so correlated as to render a joint regression useless) and they have a roughly equal relationship to delta turnout. The counties with dramatic turnout increases are certainly more white than average, and the ones with similar or lower turnout losses are more Black, but it isn’t clear that’s the driving factor.
College Education
Okay, is it college education related? Maybe the turnout increases are in more or less educated Republican towns? I am using percent population with a bachelor’s degree or higher, per the 2022 5 year American Community Survey.
The trend is super muddy among the full set of counties, pulled down by the cities with mediocre turnout and high bachelors or higher percent. What if you split by R vs D counties?
This looks a lot cleaner, although still not definitive. Those high-education counties are messing the trend up a bit. I cut the Democratic vs Republican counties right at 50% vote share, and when I ran this with vote share cut at the state average it didn’t look different. I can make those high-education counties drop off the Republican graph if I arbitrarily adjust the vote share cutoff upwards, but that doesn’t feel particularly fair.
Income
I’m using median household income, similarly from the 5 year 2022 ACS.
This is, unsurprisingly, super similar to the bachelors percent graph, and it has the same behavior where it becomes a clearer trend if you split it up into D and R counties.
So What?
Great question tbh. I’m not sure! There’s clearly something going on with increased turnout in Republican counties this year, and it looks like that’s also related to the income and education in those counties. If everything I just ran holds up, it would indicate that while Democrats are still dominating very high income/edu counties, the higher-SES end of Republican counties are showing bumps in turnout compared to low-SES Republican counties. This would mean that Republicans are, hilariously, running the Democratic turnout strategy of driving up turnout among the high-SES end of their base in extremely safe areas.
I’m going to dig into this in other states, because I suspect it isn’t just a GA effect. The phenomenon of high turnout in deep-red areas definitely exists elsewhere, and I’ll be interested to see if the income/education correlation holds up. I will also at some point get a half-decent dataset of voting-eligible population by county, which will make this all easier to control for population growth.
Happy Holidays, more to come as I get around to it.
