Alabama Part II

I wrote a post yesterday pointing out that the real story of Doug Jones’s upset in Alabama was the inordinate amount of white support for the Democratic candidate. In support of this position, I used the 2008 and 2012 exit polls to show that black turnout was no different in 2017 than those years, and yet those years saw the Democrat lose by over 20 points while this election saw the Democrat win by 1.5 points. Some basic math and reasoning tells us that the difference between losing by over 20 points and winning by 1.5 points was not high black turnout but rather white voters supporting the Democrat by much higher margins than they typically do.

The post did not really have a point beyond that. There is no clear reason why this helps any of my political preferences. I just thought it would be worthwhile to tell the truth about the Alabama outcome in the face of nearly unanimous misreporting about what happened.

Because people who are real into politics often follow it in the same way normal people follow TV shows (i.e. with huge emotional investment in the drama, characters, and plot development), this post attracted a lot of weird ire. Most of the weird ire was just that: strange outbursts with not even an attempt at making a substantive point. But there was one argument that some brought up that is worth addressing here.

The argument goes like this: It is not right to compare the 2017 election to the 2008 and 2012 elections because the latter two were presidential elections. If you want to really figure out whether black turnout or different white voting behavior was the primary cause of Jones’s victory, you should use midterms or other special elections as comparisons.

This argument poses some practical difficulties because no such data exists for comparison. But you can fiddle with the exit poll data to simulate what a normal low-black-turnout midterm election would look like and see that, relative to such an election, it is still the case that the change in white voting behavior was by far the largest factor in Jones’s victory.

To do that, I created the following baseline scenario. For this scenario, I took the 2017 exit polling data and I subtracted 5 points from the black share of the electorate and added 5 points to the white share of the electorate. This simulates what would have happened if black turnout had been massively lower. For this scenario, I also place white Democratic support at 15% (this was what percent voted for Obama in 2012) and I place black Democratic support at 96% (this was what percent voted for Jones in 2017).

baseline

As you can see, in the baseline scenario, Doug Jones only gets 34 points from black and white voters combined.

From this baseline scenario, we can simulate what would happen if black turnout increases to the level we saw in the Jones election. Holding all else equal, increasing the black share of the vote from 24% to 29% and decreasing the white share of the vote from 71% to 66% gives us this outcome.

bt

Jones’s point total rises 4 points from 34 points to 38 points.

From the baseline scenario, we can also simulate what would have happened if white voters had swung towards Jones in the magnitude they did, but black turnout had remained unchanged. We do this by increasing the percent of whites voting for Democrats from 15% to 30% (Jones got 30% of the white vote), holding all else equal.

ws

Jones’s point total rises 10 points from 34 points to 44 points.

Finally, we can bring in the actual 2017 data, which both combined high black turnout and the white swing together.

ep

As you’d expect, Doug Jones gets 48 points under this scenario, which is 14 points higher than under the baseline scenario.

So the total difference from the baseline scenario is 14 points. The change in the white vote is responsible for 10 of those points and the change in black turnout is responsible for 4. This means the change in white vote was 2.5 times as important as the high black turnout.

Of course, all I have done here is simulate a low-black-turnout, normal-white-GOP-support election by fiddling with the numbers in a spreadsheet. You could pick different values if you want and get somewhat different magnitudes. But there is no plausible baseline in which changes in black turnout contributed more to the Jones victory than changes in the white vote.

  • Neoliberals have created a political version of the Magical Negro whose purpose is to save Whitey.

  • Jason Legg

    Is ‘clear’ doing a lot of work here, or you you think there’s no import to the error for what I presume are your preferences?

    “The post did not really have a point beyond that. There is no clear reason why this helps any of my political preferences. I just thought it would be worthwhile to tell the truth about the Alabama outcome in the face of nearly unanimous misreporting about what happened.”

  • Nicholas LeCompte

    “For this scenario, I also place white Democratic support at 15% (this was what percent voted for Obama in 2012) and I place black Democratic support at 96% (this was what percent voted for Jones in 2017).”

    Gee, you think maybe in Alabama, just maybe, the fact that Obama is black may have screwed up this analysis? That maybe Obama got an unusually low share of the Alabaman white vote in 2012 for some reason? I’m sure the perfidious legacy of neoliberalism explains why Bill Clinton was so much more popular there.

    “But there is no plausible baseline in which changes in black turnout contributed more to the Jones victory than changes in the white vote.”

    This statement is banally true when you look at 2008 and 2012. And I realize the methodological problems in analyzing the demographics of midterm voters. But it seems to me that the only fair basis for comparison is the last statewide election between two white candidates. Your “simulation” has a major confounding variable that didn’t exist in the special election. It is ironic to see you commit some classic neoliberalism here: a plethora of numbers and charts mask a basic qualitative misunderstanding.

    And Matt, Jesus: you’re smarter than this. Even if I uncharitably assume worse intentions than you have, you’re smarter than this. And you’re smart enough to know that the ire drawn was not “weird,” but because you have a well-documented history of ignorance when it comes to this sort of thing.

  • Andrew

    Well, even though Obama won the popular vote in 2012 by only 1.8% more than Hillary did in 2016, he outperformed Hillary in Alabama by 5.5 points.

    If Obama’s blackness uniquely hurt him in Alabama, it’s odd that a white candidate did significantly worse 4 years later.

    It seems unlikely that Obama received an unusually low percentage of the white vote in Alabama relative to Hillary. It’s unfortunate that we don’t have 2016 exits to compare too, but it’s hard to imagine Hillary did better than Obama did among Alabamian whites considering she performed even worse there overall.

  • Suede

    Are you suggesting black voters love Neoliberalism…you know, the ones living in extreme poverty in Alabama due to it?

  • crown of ivy

    witness the condition of the american left: present data which challenges an empirical assumption about the relative weights of various factors contributing to a recent electoral result, and people will scold you like a misbehaving child. i’m very disappointed in you, young man!

  • Scott Garren

    Discussing white voters as a homogeneous group is silly. There is no reason to believe whites “swung” to Jones. What happened is Democrats showed up to vote and Republicans did not. Democratic turnout was close to 2016 levels and Republican turnout was far lower. Policy matters. We must offer a clear vision that offers real, credible, authentic hope for poor, working class and middle class people.

  • SoloCup

    I think this is the right take.

  • Jack

    The big change in the vote was that half of the Trump 2016 voters did not vote in this election, while nearly all the Clinton 2016 voters did.
    If you go through the data, a maximum of ~70k white voters changed from R in 2016 to D in 2017, while at least ~600k white R voters from 2016 did not vote in 2017.
    Which is the bigger change, 600k white voters abstained?
    BTW, VA showed similar pattern of high D and low R turnout. If this continues through 2018, then D will win the House.

  • Proust

    More simply put, because of the closeness of the race, it’s fair to say both groups of voters were important in getting Jones elected.

  • Lewis

    Hey Matt, Alabamian here. I wondered if you could do something about urban vs rural. I followed the liveblog on 538, and it seemed that turnout in Birmingham and other more liberal cities and counties was strong, while Nate Silver noted that turnout was relatively weak in the Republican white counties. So even though, as you showed in your other post, the white share of the electorate wasn’t super different than in other elections, it seems that different whites were showing up to vote.

    It would also be interesting to compare this to Roy Moore’s last election, when he ran against a democrat for supreme court. I recall he won narrowly.

  • avutunnat

    Why no mention whatsoever of 2004 data, when Obama was not on the ballot? Isn’t that the control? If you bother to look, the share of white voters decreased from 73% in 2004 to 66% in 2017. That doesn’t necessarily mean lower turnout (ratio of registered voters to actual) of whites, it just means the turnout for non-whites increased by a greater margin.

    In any case, that 7% swing is determinative. If the 7% of the 2017 electorate is non-white, nearly all 7% voted for Jones. If the 7% is white, 5% would have gone to Moore and 2% to Jones. It stands that if non-white voters’ relative turnout reflected the last general election without Obama at the top of the ticket that there would have been a 5% swing to Moore, deciding the election in his favor.

    The takeaway here is that non-whites turned out in unusually high numbers in proportion to whites, representative only of two past elections where a minority was at the top of a Presidential ticket. Also, despite there not being a minority at the top of the ticket, whites did not do proportionally worse than when faced with minority to vote against. Whites voted nearly 80% for R (Bush) in 2004, but only 70% for R (Moore) in 2017. That 10% differential is easily explained by Jones being more appealing (conservative) than John Kerry and the pedophilia allegations against Moore.

    It is at once true that non-white voters turned out in a historically high proportion to whites, and whites did not vote R in the same proportion as they had in recent elections. If either whites or non-whites were represented in the election as in the past, the result would have been different.