Alabama Megathread 4: A New Hope

[Adam Griffin]

Here's a few maps. The first 3 should be self-explanatory. The fourth map uses the benchmarks (margins) for each county that NYT put out yesterday afternoon for a very close race, and then compares the final margins in each county to that. It's worth noting that as a whole/statewide, NYT's benchmarks were quite good - and even held up well in most of the counties.

Green means Moore's margin was less than expected/Jones' was greater than expected.

Red means Moore's margin was greater than expected/Jones' was less than expected.

Some very clear patterns but not inherently tied to any partisan or racial patterns in the state:

[Adam Griffin]

Here's another map showing Jones/Moore vote share (two-way) by USDA region, which I think is somewhat relevant for most of Alabama given its history. Sadly, it makes the Glorious Relative Rebound in the steel counties in NW Alabama look non-existent.

[Adam Griffin]

Raw turnout change between 2014-2017:

[Adam Griffin]

To complement the original overall turnout change map I shared earlier: here's one showing raw turnout change for Dems between 2014-2017. Statewide, the Democratic raw vote increased by 57%. In Baldwin and Shelby Counties, Democratic raw vote increased by 160% (!).

[Adam Griffin]

And here's the GOP map; I've combined all 3 into one.

Raw turnout (2017 compared to 2014)

Total: +14%
DEM: +57%
GOP:  -14%

[Adam Griffin]

Isn't it what you'd expect though, when you go from R+28 to D+1.5?

Because that's about what you'd expect when you go from a 28% Republican win to a 1.5% Democratic win.

No offense, but you're essentially answering the question "why'd we win?" with "because we won".

[Adam Griffin]

Here are my initial calculations on the white vote. Thankfully, AL records registrations by race, which made my starting position more accurate than it would have been otherwise - not sure if they release voter turnout by race and county in the weeks following an election, but if so, that'll help make this projection more accurate.

These two maps show the same thing: the one on the left was the first one I made with 5-point gradients, but I decided to break it down into 1-point gradients on the right to make the differences between counties more visible (at the expense of making every county where whites were <15% D colorless)

[Adam Griffin]

Here are my initial calculations on the white vote. Thankfully, AL records registrations by race, which made my starting position more accurate than it would have been otherwise - not sure if they release voter turnout by race and county in the weeks following an election, but if so, that'll help make this projection more accurate.

These two maps show the same thing: the one on the left was the first one I made with 5-point gradients, but I decided to break it down into 1-point gradients on the right to make the differences between counties more visible (at the expense of making every county where whites were <15% D colorless)

Thank you!

Very interesting to see some of Jones' worst counties actually didn't have his worst performances with whites. Likewise, some counties he actually won gave him worse numbers with whites.

And Lee county being Jones' second best performance with whites is surprising and should worry Republicans.

'Tis a common misconception about many Deep South areas in states like AL & GA: people take one look at the heavily Republican counties (particularly those along the 35th parallel/TN state line) and assume "OMG those wretched redneck f[inks]ks" and then take a look at the counties that are narrowly won by one party or another in the southern portions of those states and assume they're relative beacons of tolerance or places where Democrats do better across the board.

In reality, the far northern counties in these states (like where I live) regularly post 25-30% for Democrats among whites (sometimes more) even in the present day, but simply have no base of minority voters to inflate vote share. Inversely, those competitive counties further south tend to have fairly evenly-divided racial demographics where virtually every black votes Democratic and every white votes Republican.

Basically the gradual reduction in white vote share for Democrats that you can see across the eastern half of the US as you move from north to south fizzles out just below the 35th parallel and the bottom drops out, giving you all of those Deep South counties where white Democratic support is in the single digits.

Case in point: Obama's share of the white vote from 2012:

[Adam Griffin]

College towns have been showing some interesting trends - especially college kids apparently starting to turn out more and more.

I'm really interested in the fact that white folks who live in the Black Belt are apparently more willing to vote for Democrats than those from neighboring counties.

This is a phenomenon I've observed across multiple elections in the rural Deep South. I initially considered that even small discrepancies in non-white support/turnout calcs could be having a disproportionate impact on white D projections in counties where whites are a small percentage (this definitely happens in the SW w/ Latinos). However, in the case of GA specifically, we have voter registration and turnout data by both race and county, and the trend still showed up even when maximizing black support. I ran some basic regression analysis awhile back and - to simplify - racial polarization among rural Southern whites tends to intensify substantially when blacks are between 25-55% of the population.

My amateur sociological guess is that when Southern whites aren't living alongside any meaningful number of blacks, they don't feel "threatened" politically (ergo, more willing to vote D); when they are beyond the point of being able to counter black political power, they're also more likely to vote Democratic due to a familiarity with black/Democratic local governance, and an understanding that the world isn't going to end. In between those two extremes exists an environment where the two groups can both wield political power and take it away from each other.

I will point out that 2 of the Black Belt counties in the AL projection were almost certainly being skewed by the initial concern I had in the first paragraph - both heavily-black counties (the heaviest-black 2 counties, IIRC). My calcs for Greene and Macon County suggested that 42 & 46% of whites there voted for Jones. I subsequently revised them using more detailed data.

[Adam Griffin]

Here's a map that shows the extent of the turnout discrepancy and where. Basically, this just shows the percentage-point difference between Democratic and Republican turnout relative to 2016. So statewide, Democrats pulled 92% of their 2016 numbers in this race; GOP 49%. That's a 43-point difference in turnout (92-49) between the two parties relative to 2016.

Both maps show the same thing - the one on the right eliminates numbers and uses starker colors.

[Adam Griffin]

Here's a map that shows the extent of the turnout discrepancy and where. Basically, this just shows the percentage-point difference between Democratic and Republican turnout relative to 2016. So statewide, Democrats pulled 92% of their 2016 numbers in this race; GOP 49%. That's a 43-point difference in turnout (92-49) between the two parties relative to 2016.

Both maps show the same thing - the one on the right eliminates numbers and uses starker colors.

Thanks Fmr Pres Grif!

I always enjoy your insightful posts and maps, especially regarding politics in Southern States, even though I frequently don't comment.

This map definitely shows the "Tale of Two Turnouts" at a County level, which so far seems to jive pretty closely with precinct data that I've looked at in three Counties of Alabama thus far (Jefferson, Baldwin, and Calhoun County.

One item which I'm also looking at, which doesn't reflect in County level numbers/maps are were their particular places within Counties where turnout collapsed or surged for candidates relative to previous elections, which of the course part of the reason that I used '16 US PRES as the baseline, both because precinct level result is readily accessible for all Counties in AL for the '16 GE, which can be easily matched against '17 data, and also since it is a Federal Election it's a way of testing both the transferability of the Trump brand to other candidates, as well as potentially areas where support for Trump is weakening.

One other thing that I'm looking at are the "Cross-Over" votes by precinct, since not only are these worth the equivalent of two votes (Double book accounting style), but also this shows where the Republican Coalition is Alabama is losing support, which one can't really ascertain simply by looking at County level results.

That's a very good point: the map more accurately displays shifts in vote share rather than turnout alone. I'd say that the bulk of the shift relative to 2016 is likely based on turnout rather than persuasion, however, but we'll likely see more data in the coming weeks that'll give us a definitive answer. Certainly the "2:1" dynamic at work wrt turnout/persuasion means that even a relatively small number of defections would have an outsized impact.

If I had to guess, I'd say persuasion (i.e. cross-over) was a bigger than average effect in North AL, while turnout played a bigger role in urban areas.

[Adam Griffin]

GOPland: Didn't get a majority - illegitimate!

[Adam Griffin]

The winner among write-ins by county:

[Adam Griffin]

Roy Moore's Jewish lawyer voted for Doug Jones, raised money for his campaign