Political Demographics

Nate Silver had a really beautiful analysis over the past few days of the demographic tendencies of American voters. In his first piece he noted that the vast majority of voters looked like swing voters, i.e. between a 25% and a 75% chance of voting Democratic, if you just look at non-political demographic factors, like age, race, sex, income, education, religion, location, etc. The second piece, however, took a look at what happens if you include political demographics, that is to say, party and ideology. The answer is, not shockingly, that there are a lot fewer swing voters: indeed, just 26% nationally. And in that piece he included a chart of the percent of each state's 2008 electorate estimated to be Democratic base voters (>75% chance of voting for Obama), Republican base voters (<25% chance of voting for Obama), and swing voters (those in the middle). [Note: Hawaii and Alaska excluded] I, given my predilection for this sort of thing, made a bunch of maps.

The first map is a fairly obvious one: the net margin of base voters in each state. So if a state has 40% Democratic base voters and 33% Republican base voters, that's a +7 state. Obviously, darker blue is more Democratic, darker red is more Republican, white is a tie:

  

This map looks basically a whole lot like the actually 2008 map. We can see some differences, though: Kentucky wasn't any toss-up (neither was West Virginia), and Michigan and Wisconsin were both pretty solidly pro-Democratic. So we can make a map of the difference between this demographic base-voters margin and the real 2008 election-day margin:

Notice that now we're really not talking at all about how the state turned out in the 2008 election. Rather, darker blue here indicates that we think Obama "overperformed" the demographic fundamentals in a state. This difference would come from winning the swing vote or from having a stronger base than your opponent. And as for the pattern...? There's more blue than red here, which I suspect is not an accident in that he won the election. There are a few pockets of red, but not presenting a particularly cohesive pattern. What does that mean? I have absolutely no idea. If I had to guess, I'd say that maybe this kind of thing indicates the potential for improvement by each party. You can think of it, I'm guessing, sort of like FIP: if your party did much better in a state than the demographic fundamentals suggest it should have, you'll probably regress to the mean next time. Conversely, if you didn't do as well as you could have given the underlying demographics, you might be able to make up a lot of ground there fairly easily. One would like to have a historical record of this, which there obviously isn't, but that's all I can think of.

Finally, we can just make a map of the percentage of swing voters in each state. Darker purple means more swing voters, lighter purple means fewer swing voters:

There's one obvious pattern: New England is where it's at for swing voters. Unfortunately for my region (Rhode Island has the most swing voters in the nation! [Is that good or bad?]), we're not actually a swing region, because there are a whole frickin' lot more Democratic base voters than Republican base voters. Maybe we'd get more attention in a national popular vote scenario, then. It's intuitive to me that Mississippi is a really, really low-swing voter state: about 40% of the state is black, and the other 60% is really, really, really Southern whites. There appears to be some sort of swing-vote farmer culture in Iowa/Minnesita/North Dakota, but it doesn't carry very much into the surroundings.

One way to try to discern a pattern in the divergence between the base voter margin and the actual election results is to zoom out a little, and look at regions instead of states. Specifically I'm using the 538 regions, which you can see on the following map:

The lesson of this map is somewhat more instantly evident. Obama underperformed in the "middle" of the country, while overperforming everywhere else, especially his home two home regions, North Central and Pacific (which includes Hawaii, even though this data set doesn't). So anyway, there you have it. There's my little graphical presentation of this data provided by Nate.