As an initial definitional question, we do have to define what we're calling the 'suburbs' here--and for simplicity I'm basically going to include the whole Democratic-trending part of the Atlanta area, including all of DeKalb and Fulton Counties--even though it includes Atlanta
There are plenty of suburbs in DeKalb and Fulton, and Biden made huge gains there. Unfortunately, I don't have the ability to exclude Atlanta-proper from the historical data. But we can go back at the end and take out DeKalb and Fulton and see if it's a different story
As defined here, these Democratic-trending Atlanta metropolitan area counties include a majority of the Georgia population and voted Biden 60, Trump 39. It voted Clinton 54, Trump 41. So that's a net-8 point swing.
According to the state's vote history data, the composition of the electorate by race/ethnicity in 2016-->2020 is as follows:
White 53.5 --> 50.1
Black 31.5 --> 31.2
Asian 2.4 --> 3.7
Hispanic 2.9 --> 3.9
Other 1.5 --> 2
Unknown 8.2 --> 8.7
So we have the same basic story as the statewide data, though with the Black share of the electorate holding up somewhat better and a larger decline in the white share of the electorate. This is mainly attributable to demographic changes
In these counties, the white share of registered voters fell from 48.1 to 44.4 since 2016--or about the same pace as the overall electorate. So that tells you that most of the decline is due to demographic shifts, not changes in differential turnout by race
So then there's the big question: how much did this shift contribute to Biden's victory, compared to shifting attitudes?
The demographic data alone can't answer that question. You need to know how these folks voted, and so I'm going to go back to the Times/Siena poll
If we take the two Times/Siena polls together, here's how voters broke by voter file racial group (here Biden share of major party vote, to fix undecideds):
White 43
Black 92
Hispanic 60
Asian 61
Unknown 61
Other 82
Many are pretty small samples! But I think it's reasonable
If you take those figures along with the actual racial turnout data we just mentioned, then you'd estimate that Biden won 61.8% of the major party vote in these counties. That's pretty close to his actual 60.7% share, so I think this is more than good enough
If you hold the Times/Siena results by race constant, then you'd get these results under different electorates (by race)
2016: Biden 61.2
2020: Biden 61.8
So that's a .6 pt shift due to simply due to topline changes in the racial composition of the electorate
If so, then about 15% of Biden's gains could be attributed to topline changes in the racial composition of the electorate (as Biden gained about 3.9 pts of major party vote share, of which .6 pts could be attributed to the decline in the white share of the electorate)
Now that said, it's important to note that 'topline changes' are absolutely not the only changes in the composition of the electorate.
It's possible, for instance, that new white residents were more supportive of Biden than older white residents. That's not captured here
As a result, we can't say--based on this data, at least--that 85% of Biden's gains were due to vote flipping, just because 15% were due to shifts in the topline racial composition of the electorate
That said, the Times/Siena poll can help us here as well. We know whether our respondents voted, based on the same vote history data. So we can see whether the folks in a given racial group who voted in '16 are much different than those who voted in 2020
The Times/Siena data doesn't show any meaningful difference between those with a 2016 or a 2020 record of voting, within each racial group. Here, a positive number means Biden did better among 2020 vt than 2016
White -.4
Black -.8
Hispanic +.6
Asian +1.5
Other -1.4
Unknown -1
You could read into those tiny changes if you want, but I'm just calling it a wash. There's certainly no obvious evidence that the voters who turned out in 2020 were much different than 2016.
I don't think this is hugely surprising, given the turnout data by party that we see elsewhere in the country, like NC/FL/NV--which also have fast changing populations. But of the three, GA certainly has the most Dem trend so perhaps this might have gone differently
Anyway, I don't think this constitutes definitive proof. But if the Times/Siena data were taken as authoritative on this question--and I don't think it can be--then you probably would have to attribute most of that 85% shift to voters actually changing their mind
And since this wasn't clear, this implies that the 'new' white/black voters in ATL were actually somewhat more supportive of Trump than those who voted in both 2016 and 2020.
That said, small differences; small samples. Not at all definitive
One interesting thing about this election is the extent that the 2016 post-mortems and subsequent arguments for how Democrats should win--by basically everyone!--don't necessarily look great in retrospect.
There were basically two major diagnoses for Clinton's win--and two main arguments for how Dems should win going forward. Neither is how Biden pulled it off
One theory was that Trump won by flipping white, working class Obama voters, and therefore Dems needed to lure them back--maybe with a populist economic pitch.
I think that explanation for Trump's win was accurate,
but Biden had very, very limited success with Obama-Trump vote
The thing that's most dispiriting about the 'vote dump' charts (which purport to show irregularities, but just show large Dem. cities reporting), is that it's in such complete bad faith that there's no way the electoral process could be reformed to guard against it going forward
Take mail voting, for instance. If you wanted to restore the credibility of the electoral process, you could eliminate no-excuse mail voting on the grounds that it's no longer credible to a wide swath of the electorate, even if you thought their concerns were completely wrong
The 'vote dumps,' OTOH, are an inevitable artifact of how jurisdiction reports their votes in batches, rather than updating their tally vote by vote. There's really nothing you can do to avoid this. Taking issue with it just means you don't believe election results, period
It was clear by 3AM or so on Election Night that we were probably headed to Biden at 306, and the 2020 gods have pulled out every stop to keep things even vaguely interesting for as long as possible
At the time, yes, I thought Biden was pretty clear favored. In retrospect, I was wrong about that—and I was against/rejected the AP call. At the time, I thought Biden would win by 40-50k in the end
One thing that's fairly unique about election analysis--and that rubs people the wrong way, I think--is the emphasis on the components of change from one election to the next
Take a football game. If a few weeks ago, Seattle loses to football game, 42-35, and then a few weeks later, Seattle beats the same team 35-28, with Wilson throwing 5 TDs, the headline is probably about Wilson throwing 5 TDs and the offense winning them the game
In electoral analysis, that's definitely not how we'd cover it. We'd say that the Seattle defense made huge strides and/or that the opponent's offense fared worse. Wilson would almost be taken as a given
One election looms over the Georgia runoff: the 2008 runoff, when the GOP won a runoff election by 15 points after leading by just 3 points on Election Day.
I really don't think this should, well, loom over our analysis
For starters, some of the mythologizing about the 2008 runoff is wrong. Many analysts have blamed a precipitous decline in black turnout, but I don't think that's what happened
By my estimate (since for whatever reason the GA SOS didn't publish it, as they usually do), the Black share of the electorate in the GA special was 27.7 percent Black--it was down a bit from the general (29.9 IIRC), but still quite healthy (and indeed, higher than 2020!)
Let's take a look at the turnout data so far in North Carolina, where counties worth one-third of the electorate have now updated their vote history data
These counties lean a little bit left, with a disproportionate chunk of the white liberal vote (Wake, Buncombe, Durham all in). So I think these numbers could be rosier than the final tallies for Democrats statewide, but I think the patterns will largely hold up
Let's start with party registration.
In these counties, 77% of Democrats turned out v. 82% of GOP. That's a 7 pt increase over 2016 in both cases.
As a result, the electorate by party reg in these counties is D+5.8, v. D+7.4 among all registered voters.