Credit for the victories for Joe Biden, Raphael Warnock, and Jon Ossoff in Georgia is deservedly given to the organizing efforts to register and mobilize voters. I was proud to do what little I could to advance those efforts. There has been a lot of interesting analysis of turnout in the runoff relative to the general election, which was undeniably critical.
There’s still a question that I haven’t seen answered:
Why weren’t vote percentages for Raphael Warnock and Jon Ossoff for Senate, and Daniel Blackman for Public Service Commision (PSC), all the same? Who chose to vote for one of the Democratic candidates but not all of them (and why)?
The trivial answer is that some voters split their votes between Democrats and Republicans, and/or cast a ballot but didn’t vote in all races (undervoted). The county-level voting data, which I obtained here can give us some clues about how these voters were distributed around Georgia, which I think sheds some light on the answers.
Senate races
First, undervoting was not an important factor in the relative results of the Senate races; the numbers of votes cast in the two races (Ossoff/Perdue and Warnock/Loeffler) were identical to within 52 votes, or 0.001%! Curiously, the undervote rate for both races was much greater (about 0.16% lower than the total # of votes cast) but was almost identical in each case. I’m not sure why; either there is some rate of random user error, or of people deciding they just wanted to vote in one race but not the other, but at equal rates. What this means is that the difference between Warnock’s percentage and Ossoff’s percentage of the vote is effectively the same (to within less than 0.16%) as the % of Warnock/Perdue voters. (Undervoting in the PSC was a bigger factor, but more on that later.)
Some have raised the question whether racial identity was an important factor. Obviously, race is strongly correlated with party choice in Georgia; did it influence some to vote for one Democratic Senate candidate versus the other because of a shared racial identity? County-level data show no support for that idea; the difference in % votes cast for Warnock and Ossoff in each county do not correlate with the % of the population identifying as Black or African American in that county. (This is of course merely one of the many dimensions of racial and ethnic identity that are important to Georgia voters, which I was pleased to see getting attention this election cycle.)
Some predicted that the competitive primary between Kelly Loeffler and Doug Collins in the Nov. 3 election would leave Collins voters bitter and unwilling to vote for Loeffler. While it may have affected turnout, voters in counties where Collins outperformed Loeffler did NOT show large numbers of Perdue/Warnock voters. In fact, the reverse was true – the greatest proportions of Perdue/Warnock voters were in counties where Collins underperformed.
So what are these counties with the greatest percentages of Warnock/Perdue voters? The top 7 are Fulton, Cobb, Oconee, Forsyth, Clarke, DeKalb, and Cherokee. FIve of these are Atlanta urban/suburban/exurban counties and the other two are around Athens (home to the University of Georgia); they are distinctive within the state for having a high % of highly educated, high-income residents. Indeed, the level of Warnock/Perdue voters correlates well with college graduate % based on county-level Census data:
If I were to speculate about these Warnock/Perdue voters’ motivations, they may have considered themselves moderate Republican or Independent voters in the past, may see the GOP policies as aligned with their financial interests, but find Trump and the explicitly racist and fascist tendencies of the current Republican party distasteful. In fact, these (I assume) would be the same types of voters that led Perdue to out-perform Trump in Georgia on November 3. Indeed, we can find a mild correlation between how much Warnock ran ahead of Ossoff and how much Perdue ran ahead of Trump, in each county. It shouldn’t surprise us that some of those Perdue-Biden voters on Nov. 3 came back as Perdue-Warnock voters on Jan. 5. I believe (don’t have numbers to support) that some also were further turned off by Perdue’s craven support for Trump’s post-election lies and voted for both Democrats for Senate.
Overall I think the small difference in votes for Warnock and Ossoff doesn’t originate in the Democratic base, who were motivated and informed to vote for both Senate candidates together in the runoff. In other words, the campaigns and GOTV operations did their jobs well! The ticket-splitters are presumably voters who are turned off by the GOP at the national level but may still press the button for individual Republicans. Which leads to the other runoff race...
Public Service Commission
If we look at the PSC race, as I mentioned before, there was a more significant degree of undervoting. Overall the number of votes cast in the PSC race was about 49,000 (or 1.1%) lower than in either Senate race (1.25% lower than the total votes cast). For comparison, the number of votes cast in the Perdue/Ossoff Senate race was 0.9% lower than in the Presidential race on Nov. 3, so I guess this is a typical degree of drop-off for a down-ballot race. Undervoting was correlated at the county level with lower % of college graduates.
Some (overseas absentee) voters weren’t eligible to vote in the PSC race, and there have been rumors that the PSC race was improperly left off of some absentee ballots. I do not know whether eliminating undervotes would have given Blackman enough votes to win, as there is not a very clear correlation between undervotes and Democratic-voting counties.
Interestingly, in spite of undervoting, the GOP PSC candidate, Lauren “Bubba” McDonald, ran ahead of both the other GOP candidates in terms of absolute number of votes – some voters explicitly chose to be represented by Democrats in the US Senate and by a Republican in the state PSC. While both Daniel Blackman and the various GOTV groups did great work to promote the PSC race, the amount of resources directed to messaging on this race was a small fraction of that for the Senate races, so I would expect that for many voters this was a “generic Republican” vs. “generic Democrat” situation. My interpretation is that we can’t take it for granted that Democrats will win in Georgia even when there is a monumental turn-out effort. We’ve got more work to do!
Thanks for reading this, and I will be interested to see any comments.