Wednesday, March 31, 2010

Education and the elections by region: the Hispanic Southwest.

Now I set about the continuation of my task to separate the relationship between education and elections down into regions.
One of the first things to be mentioned about this is that separating the United States into regions is never a cut and dried task. Parts of California obviously have much more in common with parts of Oregon than they do with the Arizona/Mexico border, and yet I will be grouping all of California and Arizona together. I think that most of the groupings I made make sense, but I might have to look at the results closely before I decide. I am grouping states together based not just on demographics, but also on the number of counties that state has, which can skew results wildly. Texas and California have similar demographics, but it doesn't show up in a scatterplot because the 200 small, rural, white counties that Texas has make it look like they are quite different.

So I put together California, Nevada, Arizona, and New Mexico, and this is what I got:

The first thing to notice is that in opposition to the national map, this area has only two counties over the 30% mark that voted for McCain, and both of those were only barely over the 30% mark, and only narrowly voted for McCain. This might be significant, or it might be caused by the area's larger counties. In an another state, where Orange County would be two or three counties, one of them might be much more wealthy, educated and Republican. But because of California's large counties, Orange County combines many different demographics.
At least, that is one theory.
Other than that, We have kind of the expected points: highly educated urban areas in the upper right, minority communities (mostly Hispanic) in the lower right, and rural, white counties in the lower left.
As I have pointed out before, counties make a bad unit sometimes. I specifically marked San Diego and Los Angeles Counties, because while they are not outliers, it is important to remember that most of the population in these data points is in only a few of the data points.

Monday, March 29, 2010

Incidentally: the ACS versus census census data

After doing my gigantic plot, I found out that there was actually newer census data than the 2000 census data I was using.

That data comes from the American Community Survey (ACS), which is done pretty much continuously, and generates new estimates of key demographics also pretty much continuously. Some of these 2008 numbers are different than the 2000 numbers!

However, having made this mistake, I will justify on it on several grounds: first, the census data, while older, is more complete. The census really does try to count every single person in the country --- while the ACS just estimates it. Also, the ACS doesn't seem to try to take a picture of every single county, which is one of the purposes of my plots.

Although, I am certainly waiting for the 2010 census data to come out.

The same, but different: How McCain did

So we looked at the total situation, we looked at the Obama states, and now it is time to look at the McCain states.


Once again, we have an almost random-looking assortment of dots, that when properly examined, show the breadth and depth of American politics and demographics. And, as I've said on the last two posts, a few things can be deceptive about this diagram. For one thing, the big dense ball below the 20% mark on the left: mostly rural counties across the south and east. Again, some of this may be clearer when I break down by regions.

Other than that:

In states that McCain won, McCain did pretty good in most counties, as well. Which is a fairly intuitive result. However, above the 30% mark, the counties started to even out, and above 40%, Obama won more counties than McCain.

What is especially interesting to me is the counties above the 35% mark. On the left, they seem to mostly occur in the south: Alabama, Georgia, Tennessee, and Texas. The two Kansas cases are somewhat nominal, since they occur pretty close to the line. On the right side, we have a few Southern middle class African-American counties (DeKalb, Fulton in Georgia), a few college counties (Clay in South Dakota, Boone in Missouri, Gallatin in Montana) and a few resort communities (Blaine in Idaho, also Gallatin in Montana). Oh, and also Travis County Texas, of course.

So, what this tells us is that while the demographics of the middle-class and upper-class south are still probably pretty Republican, those same groups of the mountain west and great plains may be becoming more liberal. Is the rest of the northwest following Oregon and Washington solidly into the Democratic camp? Will the south stay the south?

And to those questions, I don't have an answer.

Saturday, March 27, 2010

Breaking down the megapost: Obama's states, and college graduation rates.

I said I would break down the plot of counties in 49 states by region, but before I did that, a simpler way to break them down is simply by states that Obama won versus states that McCain won. This also didn't involve much extra work, just some cutting and pasting, and "voila!"


This shows the margin vs. college numbers in the states that Obama won, that represent a great deal of America's population, but less than half of its counties. (Quite a few of America's counties are in the south and east).
Much of this graph looks like the larger graph, especially the right side: many of the counties that Obama won biggest in are also in states that Obama won. The top of the graph looks much the same, as well. Of America's counties with over 50% college graduation rates, all of them are in states that Obama won. At the 40 to 50 percent mark, the situation is not so clear cut, although Obama still leads there. From 30 to 40, the situation becomes less muddy, but it is only at 20 to 30 that the situation becomes more muddy.

Since (as I have said many times) Obama's success was based on him managing to win both an urban/minority vote and the "educated urban/suburban" vote, the place on the left side, above 40 shows his weakest spot. Highly educated counties that voted for McCain are located in Indiana, Ohio, Colorado, New Jersey and Virginia (and possibly Wisconsin). If suburban, educated (and white) voters do deflect from Obama in 2012, those are all likely places that it could happen. Although I believe some of those places are more likely than others.

The bottom left quadrant seems to be mostly rural counties in Urban states: while Obama might have some trouble in suburban Florida, he has a bigger problem in rural Florida. However, these rural areas in urban states make up a relatively small share of the electorate.

The bottom right quadrant is also interesting: although some of the low education, high Obama counties are still there, (and are mostly black, Hispanic or Native American areas), many of them aren't. Much of Obama's low-education support comes from rural counties across the deep South, or Hispanic or Native American areas inside otherwise conservative states. So those areas don't show up on this diagram.

Wednesday, March 24, 2010

I forgot how easy Oregon was: time for a break

After all that time where I was punching in data from all those southeastern states with lots of repeating counties...
I forgot how easy it is to put in data for Oregon's 36 counties.
So, as a little sidetrack from what I was doing:

Demographic data, as I have said sooooo many times, usually does not paint a clear picture. One of the clearest and most unfortunate pictures I have seen painted is that Hispanics are becoming America's new underclass. (I believe that America does have a class system, although I know many people disagree). Unfortunately, armed only with census data and open office, I can only diagnose the problem. Fixing it would take effort. And acknowledgment.
Ah, and after giving my speech, I will make a technical comment: high school rates also seem to have a lot to do with urbanization, (inversely), but here we see that urbanization doesn't seem to have a large impact: Multnomah and Clackamas counties are two of the largest counties, but they are far ahead of the rural, Hispanic counties in the upper left corner.

Tuesday, March 23, 2010

The 2008 election and college graduation rate: every county in the country.

I have done many plots in the past, showing the connection between Obama's margin in the 2008 election, and college graduation rates. After having done a number of these, I thought it would be worthwhile to go the whole distance, and get the data for every one of the United States' 3300+ counties, and see what it revealed.

As I wrote in the summer of 2008, the window of attack available on Obama during the campaign was narrow, and McCain's campaign chose not to portray Obama as a populist demagogue appealing to uneducated minorities. Instead, they did the more acceptable thing of painting Obama as a "celebrity" favorite of the Latte-sipping crowd. And even within the liberal camp, this was kind of accepted as truth: Obama was the favorite of the young, the college educated, and the urban, as well as of the poor and minorities.

But what does the data say? By looking at county-level data, we have a tremendous amount of datapoints to look at. This many data points deals with much (but not all) of the limitations of working with only two variables. One of the biggest problems with looking at college graduation numbers is that they often occur in urban areas that also have large numbers of poor, minority voters. Did New York County (also known as Manhattan) have a wide margin for Obama because it is one of the best-educated counties in the country, or because it is one of the most ethnically diverse? With enough datapoints, the answer emerges. (Although, of course, the answer is the somewhat predictable, and somewhat disappointing, "both"). Of course, looking at county level data has its own problems: my data includes Loving County, Texas, population 56, and Los Angeles County, California, population 10,000,000, and both are given equal weight. This is especially a problem because certain areas of the country, (mostly in the south and east), have lots and lots of little counties, whereas many of the Western states have much fewer counties. For that reason, I will be breaking these figures down into regions in upcoming posts.
Also, of course, the fact that as a counties college educated population goes up, that it becomes more politically liberal, doesn't "prove" that it is the college educated people that are becoming more liberal. However, with 3300+ data points, it certainly is a hard case to argue against.

With those caveats in mind:


The first thing this shows is what every student of American politics should know: there are few fast and easy rules of American politics. There are counties that Obama won that are above the national mean for college graduation (25%), counties below it, and the same is true for McCain.
However, Obama did seem to do better than McCain amongst the most elite-educated counties. Of the 11 counties where more than 50% of the people are college graduates, McCain won only one: Douglas County, Colorado, and that was a relatively narrow victory. Above the 40% line, and even the 30% line, the situation is a bit more murky, although Obama still seems to be ahead. This is especially interesting because college graduation rates often coincide with affluence, meaning that these counties should not be all that difficult for Republicans to win, based on economic self-interest. Also, the high-education counties that McCain did win by large margins tend to cluster in one region of the country: the south, with some in the midwest. (Which will be explained further when I break this down by region).

Of the low education counties that Obama won, the most extreme examples, located in the lower right, tend to be heavily minority. They include Hispanic areas (Starr and Zavela, in the Texas border region), Native American counties (Shannon and Buffalo, in South Dakota), and African-American counties: (The Bronx, Baltimore City). There actually are some low-education white areas that Obama won as well: Eliot County, Kentucky and Anaconda/Butte, Montana, for example. Also, large chunks of the upper midwest are not highly college-educated, are white and rural, but tended to go for Obama.

This diagram also reveals some of the political constraints that are put on Obama's policy decisions. Obama won through a coalition of some of the most educated, affluent people, and some of the least educated and least affluent people. Even if we are cynical enough to discard the idea of being President for all the people, Obama (and the Democratic Party), have to somehow manage to keep a coalition that includes Pitkin County, Colorado (one of the country's wealthiest counties) and Baltimore City happy. There are often large differences in economic interests, worldview, and values between the high and low-educated counties of the US.

The Republican party also has the same problem: it certainly wants to win back those affluent suburban counties, but it has to at the same time keep its base (that gigantic sea of data points below the 20% mark), which is largely white, rural and low-education, happy.

And this is one way that 2012 will be fought.

Monday, March 22, 2010

The master plan is done, but it will take some time to tease it all out:

I have actually been making a scatterplot every day, I just haven't been posting them here.

But finally I put them all together:
I will be explaining what this is about, and then breaking it down, in upcoming entries.

Wednesday, March 10, 2010

Education in the Midwest:

It took me a while to put this together, but I put together a plot of high school and college rates in the Midwest (Minnesota, Wisconsin, Iowa, Illinois, Indiana, Michigan & Ohio).

I figured that these states make a good cohort, since they have relatively small geographical counties, usually without significant geographical boundaries between them, and they have a wide mix of agriculture and industry. Of course, as with any grouping, they aren't a perfect fit, but I think they work well enough.

(The picture should probably be clicked on, its a big picture).
After doing all this work, there isn't a lot of new information in this picture. Unlike in the Western states, there are quite a few counties with less than 10% college graduates. I think this might be an artifact of the small geographic size of midwestern counties: Western counties, because of their large geographic size, pretty much have to have certain facilities (such as hospitals) in every county. Midwestern counties, since there are no geographic barriers, don't. That is one possible explanation, at least.
Other than that, there is not a lot of surprises: big metropolitan areas tend to hang to the left, and the top right has four elite counties, three of which are college counties, and the other of which is a wealthy suburb of Indianapolis.

Monday, March 1, 2010

Race and education in Alabama

One thing I have wondered about is the generally low levels of education, both high school and college, in Southern states. I also know that many Southern states have high African-American populations, and this is a group that often is (statistically speaking) weak in education.

I bet the reader doesn't have to think very far for the various uncomfortable conclusions that could possibly be derived from looking at this.

But actually, when you look at things closely, it is better. Especially in this case:

Alabama is a good test bed for this, because it has a large number of counties, and they vary from 1 to 70% African-American. And across these counties, as you can see, the education tends to be fairly uniform: and for that matter, uniformly bad. Jus' sayin'. There are counties in Alabama between 60 and 65% high school graduation rate that are almost totally white, as well as counties with those numbers that are majority black. The one thing that is absent is any counties in the upper right: there are no majority African American counties with high high school graduation rates. Although Montgomery and Jefferson do come close.

What is interesting about this is there could be an assumption that is actually backwards. The reason why African-Americans, in the US, have lower education rates might actually be an artifact of the fact that many of them live in the rural south, where EVERYONE has low education rates. At least in part.