GA Early Voting Update 11/01
Sienna/NYT released their Governor # and curiously it’s pretty close to AJC (Kemp leading 50-45). But obviously their Senate # looks a little different. We’ll see who is right. One thing you can say about Abrams is that like a Trump or Obama she does have some followers who are outside the traditional likely voter screen, so I won’t be surprised if she does better than expected. And looking at the exit poll from ‘18, when she came close to forcing a runoff, she was only getting ~25% of the white vote and already had some weakness with black men (Kemp got 12%). So many of the narratives that try to explain her weaknesses this time were already present when she came close last time. And most polling shows her at or above 25% of the white vote. My great theory on ‘18 is that Kemp already got almost everyone that was persuadable.
One of the early voting mystics I consult said that Dems usually lose M/Tu of this week before winning the end but that doesn’t seem to be the case with yesterday’s returns. About 150k ballots came in yesterday and it looks like we won them by about +6k. We also really keep overperforming with new voters. My #’s for new voters are up to a +41k win, which overall for Any Early Voter + Any ‘18 voter who has yet to vote puts us at +21k out of 4.1mm total. And using some turnout and party support models for any voter who hasn’t voted yet puts us at a potential electorate of 4.6mm with Dem support at 50.3% or about a +25k margin. On Saturday when I ran this same # we were losing by -100k.
4.6mm seems a little high to me, but this projection keeps getting better because the NEW voters who didn’t vote in ‘18 that are turning out have been and are getting better for Dems. Most of the votes that Republicans are banking are from reliable ‘18 voters. One caveat on this 4.6mm projection - the roughly 1 million voters who haven’t voted yet (out of 6 million total) who are considered most likely to turn out are only about 40% Democratic and we would expect to lose their votes by -206k. These are the day of election surge voters that are likely waiting. The roughly 1 million voters that are LEAST likely to turn out are about 60% Democratic, but even still we are only counting on winning this group by +11k (out of ~58k who actually turn out) so that’s already factored in.
But it does show something about turnout. With nearly a billion dollars being spent, it seems like someone could run a smarter turnout operation that actually goes after some of these voters, but most turnout operations don’t seem quite that targeted. Still - of the voters who have voted early who are in the highest decile for Dem support, 2.8% had modeled turnout less than 50%. For the lowest decile of Dem support, only 0.8% had modeled turnout less than 50%. And it’s repeated by race - 5.1% of black voters who have shown up were considered unlikely to vote vs only 3.3% of white voters, and by age - 17.5% of age 18-29 early voters were modeled as less than 50% chance of voting vs only 2.3% of age 65+.
Just looking at what’s left - to get to the 4.1mm Early+18 number, Republicans still need 399k of the “Strong Republicans” to actually show up and 391k of the “Likely Republicans”. 50% of the most likely Strong Republicans in the electorate are in and only 38% of the Likely’s are in. For the Democrats, those numbers are 246k Strong Dems left to vote (61% already voted) and 362k Likely Dems left to vote (44% already in). So you’re definitely seeing the outlines of what should eventually be a strong election day for Republicans materialize, or at least what would need to be a strong election day.
One thing to look at is that of the 2018 voters who haven’t shown up yet, 503k voted in the Republican primary this year and 211k voted in the Democratic primary. And of those primary voters, 82% of the Republican voters voted in person and 76% of the Democratic voters also voted in person. So I continue to believe that the ‘20/‘21 turnout patterns where we just saw incredibly massive election result disparities depending on the mode of voting (Trump -220k, Perdue -359k on election day) will return closer to the mean of what we saw in ‘18 and before when Kemp won election day by -94k votes. As recently as 2014 Michelle Nunn got 49% of advance in person, 42% of mail and 43% of election day. Not much difference back then.
One final thing to look at today is Asian and Hispanic voters. It’s generally thought that Asian voters are trending more the Democrats way and the reverse is true of Hispanics (although I should note, the few precincts in the state that show high enough Hispanic population % that they are actually likely to have a majority of voters or close to it, Clinton, Abrams and Biden all received very similar %’s and were in the high 60’s). That said, 29% of Asian voters who voted in 2018 have already turned out this year (vs 43% for black voters and 39% for white voters) while only 22% of Hispanics from ‘18 have turned back out. That’s maybe positive for Democrats, or maybe just noise.
That said, because there are so many new voters who are Asian or Hispanic, thanks to a disproportionate share of non-18 voters, the TOTAL early vote for Asians is already at 48% of their ‘18 turnout (for voters who remain on the voterfile). For Hispanics this number is 35%. For black voters it’s 51% and for white voters it’s 48%. That should serve as a reminder of how dynamic the electorate is and that it’s always doing a lot of changing below the surface even if the topline number might not change that much.
One final thought - I know a lot of private polling shows a recalled 2020 Presidential vote that adds up almost to 100%, often with Trump near 50%. Pollsters are rightly scared of under-counting “shy Trump” voters (although I rarely meet Trump people in real life who are shy about it). Square that with the fact that 8% of early voters (so far) didn’t bother voting in the 2020 presidential election, and contrary to what you might believe about them, modeling suggests we’d carry these new voters by about 11%. If we end up with a 28% black electorate (it’s 30% right now FWIW but expect it to come down by election day), which is similar to the ‘18 and ‘20 electorates, AND 8%+ of voters are new, it just seems impossible to me that 50% of the people that turned out voted for Trump. The recalled vote in the Sienna/NYT is 43 Biden-41 Trump and when you factor in new voters, refusals to say etc, that seems more in line with what we’re seeing.