Did Obama receive >25% of white vote in any Mississippi or Alabama counties?

[Adam Griffin]

It'd be likely in the northern counties in Alabama's case or in Birmingham/Montgomery. If someone can get me a spreadsheet with the county-by-county votes and Census data/racial turnout, I'll make a map for them like this.

[Adam Griffin]

Unfortunately, Alabama does not keep turnout by race on file, so when doing a map like this, you have to combine Census data with voter registration statistics to help paint a larger picture. I built a formula for use with Alabama just like I did with my previous Georgia maps (in fact, I used the same formula). Unfortunately, I am not too happy with the first attempt. Either exit polling in Alabama is subject to a reverse Bradley effect or I will need to tweak this further. My formula for the state as a whole suggested Obama got 18% of the white vote in 2012, which seems quite a bit off from the 10% or so that is believed to be the case. However, I do believe that each county in relation to one another is by and large sound.

[Adam Griffin]

Yeah, you know...the more I look at the 2008 numbers and the likely racial composition of AL in both 2008 and 2012 (based on Census figures and the average discrepancy between Census and turnout in Southern states that do keep racial turnout data), I really think there was some sort of reverse Bradley effect at play in 2008 exit polling. It just doesn't add up, the whole "Obama got 10% of the white vote in 2008".

Even the active voter rolls in 2012 in AL were >70% white; it's virtually guaranteed that the AL electorate would have been even whiter than the active rolls. Black turnout was likely 90% of the roll number, which would put it somewhere between 23-24%; non-white, non-black turnout was likely around 3%.

73.5% white
23.5% black @ 95% D = 22.3
3.0% other @ 60% D = 1.8
Total = 24.1

Obama's 2012 share = 38.4
38.4 - 24.1 = 14.3
14.3/73.5 = 19.4%

You can tweak this all around the margins, but it's certainly not going to cut Obama's share of the white vote nearly in half. In some cases, it'd actually increase it. I think I'll stick with the original version of my above map. Interesting.

[Adam Griffin]

Yeah, you know...the more I look at the 2008 numbers and the likely racial composition of AL in both 2008 and 2012 (based on Census figures and the average discrepancy between Census and turnout in Southern states that do keep racial turnout data), I really think there was some sort of reverse Bradley effect at play in 2008 exit polling. It just doesn't add up, the whole "Obama got 10% of the white vote in 2008".

Even the active voter rolls in 2012 in AL were >70% white; it's virtually guaranteed that the AL electorate would have been even whiter than the active rolls. Black turnout was likely 90% of the roll number, which would put it somewhere between 23-24%; non-white, non-black turnout was likely around 3%.

73.5% white
23.5% black @ 95% D = 22.3
3.0% other @ 60% D = 1.8
Total = 24.1

Obama's 2012 share = 38.4
38.4 - 24.1 = 14.3
14.3/73.5 = 19.4%

You can tweak this all around the margins, but it's certainly not going to cut Obama's share of the white vote nearly in half. In some cases, it'd actually increase it. I think I'll stick with the original version of my above map. Interesting.

It is likely that black turnout was higher in Alabama than white turnout.

Unless I'm not catching something here, that still wouldn't make up much of the difference. That's definitely possible and not necessarily contradictory to some of what I said (as usually, the biggest difference between the active rolls and the electorate is caused by a lack of non-white, non-black turnout), so let's say black turnout equaled 26% of the electorate, non-black non-white turnout equaled 3% of the electorate and whites were 71% of the electorate.

71% white
26% black @ 95% D = 24.7
3.0% other @ 60% D = 1.8
Total = 26.5

Obama's 2012 share = 38.4
38.4 - 26.5 = 11.9
11.9/71 = 16.8%


To get to 10% D for whites as quoted in the exit polls, you'd need something like this...

67% white
30% black @ 99% D = 29.7
3.0% other @ 70% D = 2.1
Total = 31.8

Obama's 2012 share = 38.4
38.4 - 31.8 = 6.6
6.6/71 = 9.3%

I just don't think that is reasonable. Black turnout as a share of the electorate being 30% (when it was only 30% in GA in 2012, in a state where the voter rolls were 5-6 points blacker) doesn't seem feasible, nor does it seem feasible that blacks would literally be capable of going that Democratic (or even the non-white non-black group reaching 80% D).

[Adam Griffin]

I feel like Blacks in Alabama would be more than 95 percent D, more like 98-99. Would that make a difference?

It would make some difference, and like I outlined above, if black turnout as a share of the electorate were as high in Alabama in 2008 as it was in Georgia, then that would explain it. My own calculations in Georgia suggested that black Obama support dropped from 98% in 2008 to 95% in 2012. There wouldn't be a lot of wiggle room for that dynamic in Alabama if blacks voted 98-99% D in 2012 (exit polls claim that AL blacks, too, voted 98% D in 2008). I just don't buy the premise that AL blacks are somehow more organized or likely to comprise as large of a share of the electorate as they are in Georgia, which has a black population five points larger than AL's. I have the 2012/2013 stats on-hand for each state, so let's take a look:

Georgia 2013 Census Black Pop: 31.4%
Georgia 2012 Registered Black Voters: 29.9%

Alabama 2010 Census Black Pop: 26.6%
Alabama 2012 Registered Black Voters: 26.7%

Alabama's voter rolls are in better shape with respect to black representation, for sure, but I fail to see how this could translate into the kind of surge we would have needed to see in 2008 and 2012 to explain "10% white Dem". For what it's worth, 2008 AL exit polls claim that 29% of the voting block was black. That'd put it within striking distance of the "10% white Dem" territory, but it's a hard pill for me to swallow to believe that whites were actually under-represented (the same exit poll claims only 65% of voters were white, despite more than 70% of the registered voters being white then) and blacks over-represented at the polls in either election by such a large amount.

So basically, yeah, even though I'm in no formal or scientific position to do so, I guess I'm questioning the accuracy/margin-of-error of the AL 2008 exit polls with respect to race.  

[Adam Griffin]

Perhaps I've been vindicated! Apparently CBS did do exit polling in 2012 in AL, and showed Obama getting 15% of the white vote.

http://www.cbsnews.com/election-results-2012/exit.shtml?state=AL&race=P&jurisdiction=0&party=G&tag=dataDisplay;3060

[Adam Griffin]

And here's Mississippi. Three counties on here are denoted with red dashes; these are counties that had extensive prison populations and/or the resulting adjustments were statistically nonsensical (i.e.: suggesting a majority of whites voted for Obama; heavily black counties in all cases). Their numbers were adjusted to take this into account, but take those three with a particularly large grain of salt.

The overall statewide formula produces a result in which Obama received 11% of the white vote in 2012.

[Adam Griffin]

The exit polls generally count Hispanic as a separate category from White. This would lead to a difference of a few points versus the census figures in AL and MS, and a larger difference in GA, wouldn't it?

In all of my calculations, I broke down the projections by race, including Latino, Asian, Nat-Am, "Other" and so forth. Georgia in particular was quite easy, since turnout data by race in all of these categories is kept on file with the SoS. In MS & AL, I had to project but I used two tables of Census data (the one that includes non-Hispanic white as "White" and the one that shows Hispanic/Non-Hispanic by origin) that helped me split apart each group along the same lines.

[Adam Griffin]

This was the (very rough) calculation I did a while ago. Looks pretty similar to Griffin's



These were the top 6:
34%   Jefferson
25%   Marshall
21%   Benton
22%   Oktibbeha
22%   Lafayette
22%   Noxubee

I just backed into these numbers from the black% in the county, the county result, and the overall racial turnout rates in each state. The formula isn't perfect because black and white turnout would vary by county. Also, I used a constant 96-4 for the black vote in every county and 71-27-2 for the Hispanic vote in every county, and then backed into these white vote numbers.

This formula did lead to Obama getting a negative % of the white vote in a few counties, but I think it's probably pretty close.

I ran into the negative numbers in several counties as well. Half of them were prison counties, so adjusting for that variable took care of most of the anomalies; in the rest, I simply had to make an educated guess. I'm intrigued at how you came to <10% in Hinds County, though...?

[Adam Griffin]

The exit polls generally count Hispanic as a separate category from White. This would lead to a difference of a few points versus the census figures in AL and MS, and a larger difference in GA, wouldn't it?

In all of my calculations, I broke down the projections by race, including Latino, Asian, Nat-Am, "Other" and so forth. Georgia in particular was quite easy, since turnout data by race in all of these categories is kept on file with the SoS. In MS & AL, I had to project but I used two tables of Census data (the one that includes non-Hispanic white as "White" and the one that shows Hispanic/Non-Hispanic by origin) that helped me split apart each group along the same lines.

In AL, Latino + other are 7.1% in the 2013 Census estimates. They were only 3% of voters?

That is my estimate, yes. I'm using, again, estimates from what I know about turnout habits and citizenship by race in Georgia. In the past two presidential elections around these parts, Latino turnout was 1/2 or less of its share of the population, and in AL, Latinos are a huge bulk of that "Other". In addition, Asian and Native-American turnout rates are often even worse. When you factor in the percentage of Latinos who are undocumented in an average Southern county (30-50%) and consider who's an adult, you often wind up with a CVAP number that is at or under 40% of what it would be on paper (total population).