Updated to add, here Andrew does it more thoroughly.
Using data from Andrew Gelman for county-level voting results (admittedly incomplete ones) here are some graphs about race, region, and the vote for Obama.
Advertisements
This is officially an award-winning blog
Advertisements
21 comments
November 16, 2008 at 5:51 pm
kid bitzer
hmm…
looks like there are a lot more dots in the lower-right hand graph.
i conclude that there are more counties in the south!
November 16, 2008 at 6:03 pm
grackle
It looks like the dots in the upper right are getting away. Can anybody do anything? Stop them!
November 16, 2008 at 6:06 pm
Ahistoricality
I think I find the left side of the graphs more interesting: the range of support for Obama is quite broad when there are almost no African Americans, almost randomly distributed.
November 16, 2008 at 6:11 pm
eric
the range of support for Obama is quite broad when there are almost no African Americans, almost randomly distributed.
Just at a wild guess (I’m not being sarcastic, though it may sound as though I am late on a Sunday) I’m going to suppose that this is where Gelman’s variable of income comes into play.
November 16, 2008 at 6:38 pm
kid bitzer
i’m still not focusing on the main point here, but another thing that strikes me is just the difference in the range of the x-axis between graphs.
in the south, there are some counties that are nigh on 90% black. in the west, looks like the highest gets up to 15%. that’s pretty wow.
i still might have preferred that the graphs have a constant x-axis scale, ‘cept i suppose that would have squeezed all of the info into an indecipherable mush on the left hand, in the non-southern graphs.
November 16, 2008 at 6:40 pm
CG
A line of best-fit Obama vote going down the middle of the scatters would be helpful. That way, we could see if the the Obama vote in the South really bottoms out in counties which are about 30% black, or if that minimum occurs at 10% or 20% or so.
November 16, 2008 at 6:46 pm
kid bitzer
interesting that there is a county in the south that has zero % black population, and went for obama almost 90%
austin tx? is that you?
November 16, 2008 at 7:57 pm
todd.
Just at a wild guess …
Can we not get a 3d or heat-map plot with both variables?
November 16, 2008 at 9:33 pm
Matt W
the range of support for Obama is quite broad when there are almost no African Americans, almost randomly distributed.
Well part of this is that the range isn’t so broad when there are African Americans — everywhere but the West you can see a distinct line below which there are no points, and that’s surely partly because all the African Americans are voting for Obama. It’s interesting to look at the slope of that floor, though; egregiously eyeballing, it looks like in the NE it’s from [0,20] to [20,60], in MW it’s from [0,10] to [30,40], in S it’s from [0, 5] to [60,50]. That makes the slope about 2 in the NE, 1 in the MW, .75 in the South. Not entirely sure what that means. It kind of looks like the MW has the most pronounced effect where no counties with a few black people voted for O. as strongly as some counties with some black people, although the S seems to experience it some too (and it might look more like the MW if the x-axes were the same).
Another thing is that some of the counties involved are going to be pretty statistically insignificant, though my favorite insignificant county went for McCain 151-8. Ayup. Anyway, kid, Travis County was 64% for Obama.
Some of these counties are pretty
November 16, 2008 at 9:34 pm
Matt W
Some of these counties are pretty
That’s an editing error. I am sorely lacking in firsthand evidence that these counties are pretty.
November 16, 2008 at 10:07 pm
Ahistoricality
Looking at it again, I also notice that the Northeast has a real upper limit on Obama votes in the absence of African American voters that isn’t present in the other regions. Hmm.
November 17, 2008 at 12:07 am
Brad
Looking at it again, I also notice that the Northeast has a real upper limit on Obama votes in the absence of African American voters that isn’t present in the other regions. Hmm.
I am not sure I buy that. The peak county in the West is ~80% at a 0% black population, while the peak county in the NE is ~75% or so. The odds of those being drawn from the same distribution are pretty good, I would think (especially since both are likely to be small population counties).
That said, I find it really annoying that the West does not go to 100%.
November 17, 2008 at 2:07 am
Ahistoricality
You’re right about 0%, but at near-0% African American population, the West goes over 80%, a level the Northeast doesn’t reach until almost 20% African American population.
November 17, 2008 at 7:35 am
wu ming
there’s also the issue of county size affecting the data, especially out west where counties are larger than some new england states. i wonder what this might have looked like had western counties been as compact as they are in the eastern half of the country. perhaps breaking it down by census tract would correct for that?
November 17, 2008 at 8:39 am
Matt W
there’s also the issue of county size affecting the data, especially out west where counties are larger than some new england states.
Geographically, but by population it’s the reverse. I think the tiny sample size would outweigh any gain you got from eliminating heterogeneity within those spacious counties.
November 17, 2008 at 9:53 am
Race, region, and Obama | Red State, Blue State, Rich State, Poor State
[…] but . . . Eric Rauchway merged our provisional county data with Census numbers on %black and made some graphs, which I played with a little to get the […]
November 17, 2008 at 10:18 am
eric
Note Gelman’s post above.
November 17, 2008 at 11:01 am
Barbar
You’re right about 0%, but at near-0% African American population, the West goes over 80%, a level the Northeast doesn’t reach until almost 20% African American population.
Eh, at near-0% African-American population the Northeast has some counties that look to be over 70%. Not sure if this is a really meaningful difference.
November 17, 2008 at 12:12 pm
john theibault
Lots to chew on in Gelman’s charts.
I agree with ahistoricality that the narrower range of Obama votes in Northeast counties with few African-Americans really stands out in Gelman’s second set of charts. Not only are there almost no counties with >75% pro-Obama votes in counties with 0-5% African-Americans, there are also almost none with <25%, unlike in the other three regions where there are come counties voting near 90% and 10% respectively.
And I wonder what county that is in the lower center of the Northeast chart, where non-black voters went only about 25% for Obama in a county with about 50% African-Americans. That’s some polarization for you!
Aside from a few outliers like that, the “Racial Harmony and/or Democratic Party Solidarity” index seems to move pretty straight from the West to the Northeast to the Midwest to the South. I wonder how similar the charts for 2004 would look — and whether that would tell us whether racial harmony or democratic party solidarity is the more powerful force here.
On the other hand, there are a lot of counties in the South that seem to have a Northeast/Midwest voting profile, with above 50% of the non-black voters going for Obama in a county with more than 50% African-Americans. Are they New South places like the Research Triangle or not?
November 17, 2008 at 12:55 pm
CG
I think wu ming is onto an important point. Large counties like in the West probably moderate the data somewhat, while small counties (incidentally, the smallest counties are in the “upland south” which Obama famously performed poorly in) will polarize the data. The county that includes both Oakland and Livermore will have an overall median % African American and % Obama voters coming from two disparate regions in the same county, while two adjacent suburban Louisville counties might show sharply segregated vote/race profiles (70%AA-60%O or 5%AA-30%O or something). This could contribute to the “harmony”/Democratic party solidarity that we are seeing in the West and not seeing in the South.
November 17, 2008 at 1:15 pm
More race, region, and Obama. « The Edge of the American West
[…] from here and […]