Perot in 1992 and the coalition nature of American politics

I am sitting here wondering what is wrong with me, and why people don't like me.
But there is no answer to a question like that, so I will ignore it.

Instead, I will talk about what is on your mind: Ross Perot, and where exactly his support came from. As discussed in the last post, the evidence that Perot gained votes predominantly from one party or the other is, from the evidence I have, not apparent.

One thing about the diagram yesterday is it looked suspiciously like a much earlier plot I had done: Obama's margin versus high school graduation rates. States that went for McCain tended to have either low or high graduation rates.

(I hope you are following my intuition, because I am actually not, so connect the dots for me. After all, connecting dots is what this blog is all about).

I decided to plot Perot's total vote (not his margin, obviously) against high school graduation rates. (Using data from the 1990 census)

If any of you have been paying attention to my futile quest to link together demographic data with election margins, this data should really stick out, because...there is actually a pretty strong correlation here. I actually ran Perot vs. college, and then the same diagrams for Bush and Clinton, and none of them were very conclusive. But this diagram shows that states with high High School graduation rates had a pretty meaningful tilt towards Ross Perot. In fact, there is something about the way I made this diagram that conceals this: the cluster of states just inside the upper right hand corner is mostly New England states, which have two things in common with prairie/mountain states: lots of high school graduates, and a liking for Ross Perot. Notice that there is not much else they have in common though: Massachusetts and Idaho are stereotypically the polar opposites of national politics.
Notice down in the lower left, low High School and low Perot support states. Now, if you can remember back 18 years, the Southern states were divided between Bush and Clinton. Arkansas, Tennessee and West Virginia were Clinton states. Alabama, Mississippi and South Carolina were Bush states. But all of them have low High School rates, and low Perot rates.

One of the ways that I look at the American "two party system" is as a "two coalition system". The coalitions are made of many different regional groups, with both official power structures and differing demographies. Perot carved up a big part of that coalition for himself in some parts, but not so in others.

This is still relevant, because the current coalition that makes up the Republican Party has two major geographic bases of support: the Prairie/Mountain states and the South/Appalachia. But these two groups have very different demographies, and different cultures and politics.

Perot "taking" votes in 1992

There are only two possible ways for a Democratic candidate to win an election: ACORN, or a third party "steals" them from the Republican Party.
Dog food isn't very tasty, neither is it nutritious.
Anyway, one of the pieces of CW thrown around about the 1992 election is that Clinton won because Perot peeled off votes from Bush. This is also disputed, but since the ballots, with their '2nd choice' bubbles, are all sequestered in a vault under Mt. Rushmore for 99 years, we won't know for a while who people would have voted for if Mr. Perot was not in a race.
But, we do have a technology that can hint at it. And that technology is...scatterplotting! Of course!
First, let me apologize that the new version of openoffice done gone and thrown my Y-Axis labeling right in the middle where it confuses things. I updated to karmic koala because I was trying to get Mario Kart's sound to work right, and it just kind of happened...
Anyway, back in 1992, George Stephapolous has to rent a motel room to call Clinton and tell him when a bimbo is erupting, because they don't have cell phones, and they have no idea what KARMIC KOALA is. But they do know who Ross Perot is.
The above diagram has basically no correlation between how much of a margin Clinton had, and what total percent Perot got. If Perot was mostly taking votes from conservatives, he would be getting a lot more votes in Nebraska, where there are plenty of conservatives, than in Massachusetts, where they are not quite as many. And yet Perot got 23% of the vote in both, even though Nebraska was 18 points against Clinton and Massachusetts was 18 points for him. Of course, looking at the plot a bit closer shows that there may be a little bit of a lean towards Perot in more conservative states: so maybe there was a few states where it made a difference. Over all, though, there doesn't seem to be much evidence either way from this data.

As promised: Canada oh Canada

One good thing about Canada is, there is only 13 demographic units to enter data for.
Of course, I am still looking for good sources of data, and for good sources of interesting data. Those lacking, I just looked at what wikipedia could tell me, and decided to look to see if gdp per capita and land area were correlated:
There does seem to be some sort of trend here, or more than one. Also, the two big outliers are both very small territories. So I don't know. Also, yesterday I promised that we would get a plot that looked like a dragon's head. And of course this doesn't look like a dragon's head, but it does look somewhat like a pair of pliers. Or, as they say in Canada: "spanners".

I keep on telling myself I will diversify, and then don't: ERS data on

I really do want to do more international stuff, but I tend to look at the US data, because I know where to start, and I know where the good data can be found. But some day, I will try to figure out how the Canadian census data works, and you will be able to compare Alberta to Nova Scotia all you want.

Another thing is, when I started this, I wanted to look for interesting shapes. But, truth be told, most data looks about the same: a big smudgy diagonal line. Such as this:
Rural and urban poverty go up together, and rural poverty is almost always higher: although by differing margins. While there seems to be the expected grouping in poverty along regional lines, the ratios themselves seem to be caused by different factors. Massachusettes, Nevada and Indiana all have higher urban poverty than rural poverty, but for very different reasons.

I will still be out there looking for an interesting shape. Life span versus miles of highway amongst Canadian provinces looks like a DRAGON HEAD! Maybe!

Oregon, unemployment, and college

Once I get on a run, I tend to run with it.
Especially if it is something that I have data on, and can just cut and paste to different documents.
In the nation, there isn't much relation between college and unemployment, so lets look at the same data in Oregon.
(BTW, for those of you who don't know, I am an Oregonian. Mostly. Also, Oregon gives a good cross section of demographics.)Overall, there isn't a lot of direction in this diagram, although the four urban counties with most of the educated people and where much of the work goes on, are all grouped together in the middle.
So, again, another inconclusive result.

Oh, and also: 2010 is going to happen soon. It already did in London.
2010 means a number of things, including a CENSUS. However, the census results won't be out for two years, or so.

The same thing, with Oregon:

So, to make up for my terrible downturn in daily updating, and because I got interested,

I decided to look at the last plot I did, but for Oregon and its counties. Luckily, the census has growth figures by county, and the Department of Labor has a handy tool for looking at county level data

So the same graph for Oregon gives us:
There is a little bit more shape to this, but admittedly, not much. And again, although its not very clear, there is the same 5/6th pattern as we had last time: the only area where we don't see any points is high-growth, low-unemployment. Hood River and Benton Counties come close, though!
Another problem with this, as I have said for my Oregon diagrams before, is that bit all Oregon counties are equal in population. Especially noticeable in this, with Harney County, population 8,000, just sitting down there in the corner. I am tempted to do this graph with, for example, only the 10 or 15 most populous counties, and see what results I get from that.

Unemployment versus growth: once again, the easy explanation isn't the best one.

One piece of obvious conventional wisdom that I had been carrying around was that unemployment was higher in areas with high growth: that areas with high unemployment were areas with large influxes of population, and they therefore had high frictional unemployment, or had "oversold" themselves to potential workers. It seems like a good argument, and there are certainly a few data points to support it.
But you, my astute readers, know about "seems like a good argument" and "a few data points to support it". The actual scatter plot of the data is, as could be expected, scattered.
And, as is also often the case, there is a "three quarters" effect in here, although not a distinct one. Actually, It would be more a "five-sextet" effect. If we divide unemployment into high, medium and low, and growth into high and low...all of the five sextiles are occupied, except for "high-growth, low-unemployment". Utah, Texas and Colorado have high growth and medium unemployment, but there is nothing to the left of them.
Of course, in a normal economy, this graph might look different. So lets hope that we have a normal economy so I can find out! Also, I will have a job, and might not have time to scatterplot.
Incidentally, I found out this data several weeks ago, and just didn't bother to make a graph and post it. I actually have lots of stuff like that that I am sitting on!