Mapping the South for Obama

[I massively updated the middle part of this post after thinking about it on the ride home]

I was pretty startled by the two maps I saw at Strange Maps over the weekend. They showed a distinct correlation between cotton production in 1860 and Obama support in 2008. Where more cotton was picked 150 years ago, there were more Obama voters. There’s ways in which the correlation makes sense: slavery increased production on plantations, slaves were African-Americans, and, now, African-Americans voted almost unanimously for Obama (95%, according to CNN). But could it be that that band of Obama support snaking through the middle of South Carolina, Georgia, Alabama, and Mississippi was still predominately African-American, though it had been the center of cotton production in the waning days of slavery? Enter GIS and 2000 census data!

Before I continue, I have to make the caveats that I don’t really know what I’m doing here. I’m just playing around with some data and ArcGIS. Hit me back in half a year, and I’ll be much smarter.

Anyway, first, the vote results by county (poll results via Mark Newman, shapefiles courtesy US Gov’t). My map is a bit different than the one floating around the internet, since I think I broke down the colors differently. Simply put, where the colors are light, the support for Obama (blue) or McCain (red) was under 5%. Where the colors are bright, support was between 5% and 10%. Where the colors go dark, support goes over 10%:

How does that thin, thin band of blue match up with African-American populations in the South? Enter my second map, which charts the percentage of the population identifying as “Black” in the 2000 census:

There seems to be a strong correlation, at least visually. It’s interesting to note, however, that Obama failed to carry many counties that were at least 40% black. Some even went rather strong for McCain. As I don’t have voter demographic information, I can’t really speculate as to the reason. But one possible reason is that the black population is too young to vote in those counties.

But youth was also, apparently, a huge indicator of Obama support. Of the under-30 crowd nationaly, 66% went for the Democrat. So I decided to see if youth could predict anything about Obama support–perhaps it, too, would correlate strongly. This sets up the need for two maps, however, one of everyone under 30 in 2000, and one just of everyone ages 18-30 in 2000:

% of population under age 30 in 2000, by county

We can see that the same sort of band winds its way through the Deep South, though south Georgia and Louisiana, which are rather young, didn’t go for Obama so strongly. Let’s see how the demographics work out for the youth of today, the 18-30 demographic:

% of population aged 18-30 in 2000, by county

Now it’s just a visual mishmash. In the earlier version of this post, I had % of population under the age of 22, but that’s not really useful information. It’s not like those people all aged 8 years since 2000 and weren’t replaced by new youths. It made more sense that if I didn’t adjust the racial demographic data at all, I shouldn’t also adjust the age data–just so I can be consistent. But what do we see here, other than that there’s no one young in far west North Carolina? I’m not sure. In fact, it’s so visually messy, that the map ceases to be useful, and I might need to actually scatter plot this.

So let’s first look at Obama’s margin of victory against the population under 30:

I’m not (yet) a statistician, but this plot, to my eye, shows that Obama trended a bit younger than McCain, and that the younger the counties got (proportionally), the bigger margin of victory he got. But if this is the “young” demographic, what does the “youth” demographic show?

Is it showing the same sort of thing? I can’t tell. Maybe? If it is, it seems very subtle, and I don’t know how to suss out the information any better. It does seem like once 27% or more of the population was “youth,” they went for Obama more than McCain, but then there are those outliers in the super youthy counties, which all went for McCain. So if the youth plotting isn’t, actually, all that helpful, let’s plot the margin against racial makeup:

This, I believe, speaks for itself…

Anyway, I think there’s more to be said about the age thing. I just wish I understand the data better. And I wonder how much the data is being skewed by including Katrina. In 2000, Louisiana was very young and very black, as my later maps show. Yet it was decidedly not interested in sending Obama to the White House, unlike other black and young counties in the Deep South.

Tags: Barack Obama, GIS, John McCain, statistics, The South, youth