Sunday, November 30, 2008

Black states, white Republicans

++Addition++ Per Agnostic's cosmetic suggestions, the graph presented in a way more suggestive that the black voting percentage contributing to the increased Republican support among whites.


In discussing some examples of where individual and state-level electoral trends differ, I've made note of the fact that blacker states are more likely to vote Republican than less black states are. Given the overwhelming propensity of blacks to vote Democratic, this obviously suggests the black population's effect on the white vote in a state is stronger than it is on the entire state's electoral leaning (obvious though it is, I failed to present it in this way--Orwell's remark "To see what's in front of one's nose requires a constant struggle," appropriately describes me, except that my ignorance is never willful).

The direction of a state's white vote and the percentage of the electorate that is black correlate at a firm .55 (p=0)*. For each percentage point increase in black representation among a state's electorate, McCain improved his support among whites by three points.

Yet looking at all states dilutes the potency of the black population's effect on the white vote. Whitebread red (ie North Dakota) and blue (ie Vermont) states, where racial concerns are mostly abstractions, attenuate the relationship. If only the top 25 states by percentage of the population that is black are included, the correlation jumps to .72.

I speculate that in blacker states like Mississippi, whites are less likely to vote Democratic in part because doing so means supporting wealth transfer policies that primarily benefit blacks at the expense of whites. It's a less abstract moral exercise and more of a real world sacrifice than it is in say, brainy Vermont, where leftist quasi-socialism doesn't have anywhere near the same pathological consequences as it would down South. Also, in lily-white states, most welfare beneficiaries are whites who are presumed to be only momentarily down on their luck.

It's easy to conceive how this pattern could be accentuated further in the coming four years, depending on economic conditions and the agenda President Obama pursues.

* This does not include DC, an entirely urban outlier without the same state-level concerns voters in the rest of the country have. It bucks the trend, with whites voting 7 to 1 in favor of Obama.


Stopped Clock said...

Great work. I can identify some of the dots just by their place on the map: the bottom left is Hawaii and Vermont, the far right is Mississippi and Alabama, followed by (in some order) Louisiana, South Carolina, and Georgia. I'd like to identify the four "legs" with New England, the upper Midwest/Pacific NW, the Great Plains, and the interior mountain West respectively, but I know that it can't be that perfect a fit.

That Vermont stands out so sharply even from its neighbors (and from upstate NY) shows that something unique is going on there.

But a couple of dots puzzle me: it seems there was a 25% black state where McCain didn't win a majority of the white votes. Are you sure that isn't an error? Also, I'm surprised to find any zero percent black states. I'm going to guess that it's North Dakota and Idaho, but I cannot find any states that have less than .5% blacks, which would be the lowest possible amount unless you rounded everything below 1 down to 0.

Stopped Clock said...

OK, it's obviously Maryland. I wouldn't have thought there were that many blacks there but apparently there are.

Audacious Epigone said...


Yeah, it's Maryland. McCain won 49% of the white vote, but he still outdid Obama there (Obama is reported to have received 47% of the white vote in the state).

A state's black percentage is based on voter turnout, not Census population numbers. Not that they vary much (blacks vote at a rate about equal to their population representation, while whites vote at higher rates and Hispanics at much lower rates), but I figure black turnout gives a better picture of perceived black political influence in a state.

BGC said...

I think this is one of your best ideas - you should keep going on it!

agnostic said...

It's been a few weeks since I read it, but I think Red State, Blue State, Rich State, Poor State has some graphs on Republican vote share and percent pop that's black. Easy read, if you haven't browsed / picked it up yet.

Some pedantic points:

1) The graph should have % black on the horizontal and % Republican on the vertical. It'll make it easier to follow your argument that the former is the cause and the latter the effect, and that the response is flat for low % black, but steeper for high % black.

2) Since you're looking at percentages, the best-fit line is constrained to not go above 100% for both of the variables. So, if you're using a linear Pearson correlation (as in Excel), it doesn't know that, and reports a lower correlation since it's trying to fit a straight line through something that's pretty flat but then increasing.

A better statistic here is Spearman's rank correlation (rho), which you can calculate very quickly and for free using this awesome site:

Go to "descriptive statistics" in the right menu, then search "Spearman." Just copy and paste the two data lists, hit compute, and it gives you a correlation and two-tailed p-value!

Audacious Epigone said...


I'd been planning on reading the book sooner, but instead I put it on the Christmas list (I had to come up with a few things, after all). Does it look at the white vote, or only the total vote by black population?

Thanks for the Spearmen site. I use a Pearson correlation by default, since it's the 'standard' people are most familiar with, and it's easy to build predictive linear equations based on it.

Audacious Epigone said...


I wonder how to look at urban areas exclusively. DC is an outlier, and also the only entirely urban area with racial exit polling available. Urban whites are obviously skewed far to the left of the general white population, but are whites in black cities like Detroit or St. Louis more likely to vote Republican than whites in Seattle or San Francisco are?

I could look at House and mayoral races to get some idea, but that requires being familiar with every election considered. To hit close to home, in Kansas City, MO, for example, the current mayor is a Democrat, but he ran against another Democrat in the election who was the liberal choice. In the vast majority of urban contests, I'd be looking at a Democrat against another Democrat or a token Republican, without a racial breakdown. So I'm not sure what I'd be able to do to address whether or not this holds in urban areas, where the black presence is most tangible.

Audacious Epigone said...


New Jersey is pretty dense, and it demonstrates the broader trend--McCain actually did slightly better than Obama among whites there.

Audacious Epigone said...


The Spearman correlation is only .34. If you were tutoring me, what might you suggest as a plausible explanation as to why?

Thanks for the tremendous site, by the way. Very user-friendly and expansive in what it lets users do.