Georgia's Very Own Megathread! (v2)

[Florida Man for Crime]

52% AA!?

That sounds good to me!

Now, the question is whether the AA % continues to be higher in the 2nd week and Dem turnout accelerates this week as the end of early voting approaches (as it has often done in the past). If so, then I think Dems can feel good. If not, then, well... it seems like there has been a lot of GOP turnout in the first week, so it will be worrying if we now simply revert to the trend of the previous week.

[Florida Man for Crime]

Oh wow, the Black % actually went slightly up today. I was afraid it might go down at least for today and tomorrow (before going up late in the week).

Looks like there is probably/hopefully going to be a very good surge for Abrams over the week's early vote.

[Florida Man for Crime]

Voters today in GA:

White --- 89224 --- 57%
Black --- 48136 --- 30.8%
Hispanic --- 2214 --- 1.4%
Asian --- 2026 --- 1.3%
Other --- 14921 --- 9.5%

This would normally be a day (I think) when Republicans would do a bit better on the weekend, but instead Black turnout already seems to be going up. Hopefully it continues to go up over the course of the week - as I gather from what Griff has previously said, that has been the case in the past.

[Florida Man for Crime]

Of those to vote so far...

65% of those who voted on Election Day in 2014 are white (27.8% of voters)
61% of those who voted early in 2014 are white (38.6% of voters)
51% of those who didn't vote in 2014 are white (33.6% of voters)

So white cannibalization as compared to relatively more new non-white voters. Good.

Well, except now that I think of it, there is probably a higher % of Other (who are actually White) among those who didn't vote in 2014?

If so, then probably not quite as good as that would seem at first glance.

[Florida Man for Crime]

Tuesday's electorate was 56% female, compared to 56% on Monday.  

Code:

Female	821969	54.6% (+0.2)
Male		646477 	42.9% (-0.3)
Unknown	38064        2.5% (0.0)

There is a lot of focus on race for obvious reasons, but the Female vs Male turnout is pretty remarkable.

Most polls generally have, (at the very most, and often less) Females making up something like 53% and Males 47% (at the least, and very often more). Anyway, so if the turnout is that much skewed towards females, it seems like a potential source of error in polls where they could be underestimating Dems.

Whereas this is a gap of more than 10 points.

BTW, in the TargetSmart GA data (which does not include this latest day, the Gender split in share of votes cast so far is 55.67% Female  and  44.31% Male. That is taking out most of the unknowns through modeling based on their names and such.

And that is an 11.36% turnout gap.

[Florida Man for Crime]

BTW, the Gender gap in terms of turnout being so large seems to be pretty much unique to GA. You also do not see it in other southern states (at least not to nearly the same extent) like NC and FL. And in TN, there is actually more male turnout, over-performing the share of male registered voters. So it is probably not solely felon disenfranchisement (although that makes a difference), but seems to be especially strong of a thing in Georgia in particular, more than anywhere else.

In most states you have much more even splits, although there is generally a bit of Female over-performance.

[Florida Man for Crime]

BTW, the Gender gap in terms of turnout being so large seems to be pretty much unique to GA. You also do not see it in other southern states (at least not to nearly the same extent) like NC and FL. And in TN, there is actually more male turnout, over-performing the share of male registered voters. So it is probably not solely felon disenfranchisement (although that makes a difference), but seems to be especially strong of a thing in Georgia in particular, more than anywhere else.

In most states you have much more even splits, although there is generally a bit of Female over-performance.

Using which data? Exit polls?

Using the TargetSmart early vote data. That is voter file data on early voting from pretty much every state.

https://targetearly.targetsmart.com/index.html

So it should be as accurate for other states as what people put down on their voter registration forms, plus with modeling to fill in most of the unknowns. And even with that, you don't see the same gender turnout split in other states, at least not to the same extent.

[Florida Man for Crime]

Using the TargetSmart early vote data. That is voter file data on early voting from pretty much every state.

https://targetearly.targetsmart.com/index.html

So it should be as accurate for other states as what people put down on their voter registration forms, plus with modeling to fill in most of the unknowns. And even with that, you don't see the same gender turnout split in other states, at least not to the same extent.

Do all of these states record gender data? Seems silly to ask, but most of what Georgia collects can't be found universally in other states. Depending on that, I'd be skeptical of their figures, since their modeling in a lot of these races obviously isn't indicative of the actual preferences of voters.

For what it's worth, in their modeling, they also have LA at 55/45. I don't see any other states there where I feel a reasonable comparison to GA can be made along these lines (though FL is 54/46 currently), but to the extent that the gender gap is lower in other Southern states, it almost certainly has to do with black population share, lower black male turnout compared to black female turnout and/or incarceration and probation rates in general.

Recording age and gender on voter registration forms is either universal or nearly universal.

Race data is the main difference between GA and most other states, but gender is standard.

[Florida Man for Crime]

I can't wait to see the demographics on the voters from today. This is going to be good, I think.

[Florida Man for Crime]

17% "other" today sounds good - that should mean a lot of new registrants.

[Florida Man for Crime]

And here's the GA Votes numbers like I did yesterday (where each group's %s show their share of that vote type's electorate, rather than their share of the 3 types):

Full-size image

Yeah, that is a much better way to do the crosstabs, thanks. Very counterintuitive the way he does them on georgiavotes.com

[Florida Man for Crime]

Here is my analysis of GA Early Vote data using the TargetSmart data, which now seems to be complete. See the Early Voting thread in the Congressional Elections board for more info and similar analysis for other states.

But I am particularly interested in GA, so wanted to post this here to see what you all think. I am unfortunately more pessimistic than I would like to be about GA at the moment for Abrams.

Do you see reason to really disagree with my analysis? how concerned should we be about the #s among Infrequent Voters that Rs seem to be getting?

Code:

GA --- All Voters  ---           44.4% D --- 50.1% R --- 5.5% UNK ---                          D+4.2 Swing in Partisanship Margin from 2014
GA --- Super Voter ---           39.2% D --- 58.2% R --- 2.6% UNK ---                          D+3.4 Swing in Partisanship Margin from 2014 ---              (R+0.4 Change in Swing Since Oct 31)
GA --- Freq. Voter ---           46.7% D --- 48.6% R --- 4.7% UNK ---                          R+0.1 Swing in Partisanship Margin from 2014 ---              (D+0.1 Change in Swing Since Oct 31)
GA --- InFreq. Vt. ---           47.3% D --- 43.0% R --- 9.7% UNK ---                          R+4.2 Swing in Partisanship Margin from 2014 ---              (D+2.3 Change in Swing Since Oct 31)
GA --- Never Voted ---           46.7% D --- 39.5% R --- 13.8% UNK ---                          D+0.2 Swing in Partisanship Margin from 2014 ---              (D+1.7 Change in Swing Since Oct 31)

Georgia is one of the few states where looking closely at the early vote data makes me less optimistic about Dem chances than I otherwise would be.

There *is* a 4.2% overall swing in Partisanship of early voters as compared to 2014. However, unfortunately Georgia is very different from most other states in that the Dem swing has come basically entirely from Super Voters. This should be pretty surprising, since turning out low propensity voters has been the #1 focus of the Abrams campaign. But it looks like a lot of low propensity Republicans are also turning out. That means Dems are at risk of having cannibalized the election day vote in Georgia, and I am pretty worried about the election day vote. I suspect that, and given that African Americans who are eligible to vote were already very well mobilized compared to most other states, and given felon disenfranchisement laws that disproportionately effect African American males, there are just not enough additional low-propensity Dems to secure a victory by itself, unless low propensity White Republicans don't turn out.

One other way to spin things onto the bright side for Dems is that Dems have improved among Infrequent and New voters by D+2.3 and D+1.7 since Oct 31 (compared to the same time in 2014). The optimistic hope would be that this is a forerunner of huge election day turnout among Dem-leaning Infrequent and New voters. But overall, I would be pretty worried that Republicans seem to have done a pretty good job of turning out their Infrequent voters in Georgia, and that too much of the Dem early vote turnout may be cannibalization of Election Day super-voters, as opposed to actually bringing in new voters.

The only other optimistic take I can see for Dems is that maybe a lot of the additional Modeled R partisanship voters are not really Rs any more. If a lot of those are actually former Rs in the North Atlanta suburbs, but they swing en masse to Abrams, then things could go better than the data seems to indicate. If Abrams does win, I think it will be more because of a large swing among white voters in the North Atlanta suburbs towards Dems, as opposed to just Abrams turning out the base.

Regretfully, I am thinking that despite Abrams doing well in the polls, I may need to rate GA-GOV as tilt R for my final prediction, though I have not made a final decision yet. But this is really one state where it will all come down to election day turnout, and most likely also to a runoff.

[Florida Man for Crime]

Just a reminder that TargetSmart whiffed BADLY in 2016. They were literally missing multiple states by double digits. Don't take their "modeled partisanship" garbage seriously.

It is not supposed to model the vote, it is supposed to model whether people generally consider themselves as D or R. Of course that will also fluctuate just like Party ID fluctuates.

I remember that. Clinton was supposedly ahead in Florida by 7 at one point in a TargetSmart poll.

It is not a poll, it is early vote data. Regardless of whether it overestimates D or R partisanship (I am sure it does one or the other), it is consistent methodologically over time, so the comparison to 2014 should be valid. And that is showing more of the Dem gains coming from voters with better vote history, which is a bit disconcerting.

I am thinking at this point Abrams' best bet is for it to go to a runoff, and then in the runoff Rs lose intensity (while nationally all those Dem small donors who have been donating millions will suddenly start pouring it all into her campaign), while Abrams still has her ground game to turn out voters. Yes, I know that historically Rs have tended to do better in runoffs, but I think that is probably the better shot at this point. I do think it is still possible for Abrams to win (even without a runoff), but if so it looks like it will be more as a result of swinging suburban whites over to her side as opposed to just turnout by itself. The polls out of GA-06 and GA-07 are encouraging on that point.

[Florida Man for Crime]

Good luck to all of you today. Cheers

I would love for Abrams to win. Good luck!

[Florida Man for Crime]

This article has some estimates apparently from the Abrams campaign (internal estimates) on what % of the White vote Abrams got in various counties:

https://www.theroot.com/nobody-should-be-talking-about-beto-and-gillum-in-2020-1831061434

You must be logged in to read this quote.

[Florida Man for Crime]

And here are my initial projections/modeling that sorts out the "others" into one of the other four categories:

Wow, whether one looks at your approximate assignment of the "Other" or at the numbers including "Other," those are some very impressive numbers on the non-white turnout. Better than Presidential, in a midterm year!

On the other hand, that does mean that there is probably only fairly limited room for improvement in 2020. So GA is not outright impossible for the Dems to win in 2020, but getting that last 1% is not going to be easy.

[Florida Man for Crime]

https://medium.com/@tombonier/targetsmart-analysis-shows-a-younger-and-more-diverse-georgia-electorate-fb38bdb6224e

[Florida Man for Crime]

https://medium.com/@tombonier/targetsmart-analysis-shows-a-younger-and-more-diverse-georgia-electorate-fb38bdb6224e

I think I may have mentioned this here a few days ago, but I'd point out that I take issue with Targetsmart's racial modeling for GA. Besides the fact that GA actually releases its racial data, their modeling seems to reduce the black electorate relative to the SoS turnout totals. That almost certainly wouldn't be the case, as you'd expect the real share of the black electorate to be larger than the SoS reported figure (again, because of the "other" category containing voters across all racial lines).

Their modeling is great to have in situations where this data isn't available from the state (which is almost every other state), but it seems to be wonky here at best. They may just be using their formulas without incorporating the state's data, but it's almost guaranteed that there aren't tons of people lying about being black on their voter registration forms.

I assume their first-time stats are more on point (I see their black/white shares of the first-time electorate - 50/35 - match up perfectly with my estimation of the "other" category, which tend to be one and the same), but when in doubt, go with the SoS totals for race, gender and age (and in the case of the latter two, there's little need for interpretation due to the dreaded "other" dynamic).

Voter file race modeling incorporates actual hard race data like in GA when available (unless TargetSmart is doing something really weird and unusual, which I doubt). I don't really see the inconsistency between this and the #s you posted. From their graph, they appear to have about 31% of the electorate being African American, which is about what you have and what the SOS has. The difference, if there is one, in comparison to your estimates is that they may be assigning a bit fewer of the "other" to be African Americans. But this is a pretty marginal difference at most.

[Florida Man for Crime]

Catalist analysis of GA-GOV. There is a lot of really great stuff here. Looks like it hasn't been posted here yet, although it is now a week old:

What Happened in the Georgia Gubernatorial Election?