Wednesday, June 5, 2013

Where the Republicans are Really Red, and Democrats are Really Blue

I've been meaning to post about the release of Boris Shor and Nolan McCarty's state legislative data. With ideal points of every state legislator in the country over the past two decades, this is one of the most important datasets on state politics, legislatures, and parties to be released in a long, long time.

Anyway, one of the many, many things we can learn from this dataset is the location of the most ideologically extreme state parties. The map below categorizes the Republican Party in each state based on the median ideal point of its Republican legislators over the past two decades. The redder the state, the more ideologically extreme it is.
Now, this is something you would not glean just from looking at presidential vote results. California has been a reliably blue state for several decades now, but its Republican Party is, by this measure, the most conservative state Republican Party in the country. Colorado's GOP is pretty far right, as well.

Conversely, here's the map of state Democrats. The bluer the state, the more ideologically extreme it is.
Again, California pops up -- it has the most liberal Democrats and the most conservative Republicans in the land. Colorado and Washington are similar, coming up in the most extreme categories in both parties.

Note, however, that these graphs don't really tell us why a state might have extreme or moderate legislators. The Republican parties of the deep South are probably very conservative because they represent some of the most conservative voters in the country, while the extremism of parties in more moderate states may be more an indicator of the strength of party organizations in screening out moderates in primary elections.

Still, this sort of data is extremely useful if we want to understand the political context of a state party and the politicians who emerge from it. Barack Obama was an Illinois state senator (yep, he's in the dataset -- ideal point of -.695), although they're far from the most liberal group of Democrats in the country. George W. Bush got along famously well with Texas Democrats when he was governor, but those are among the most conservative Democrats in the land. How has Chris Christie survived in blue New Jersey? Well it probably helps that his state's Republicans are moderate by national standards, keeping pressure off him to veer rightward. And it's probably not a coincidence that California, with more than a tenth of the country's population, hasn't produced a viable presidential candidate in more than three decades. Apart from Schwarzenegger (who was elected under pretty unique circumstances), the only people that get elected there are the reddest Republicans and the bluest Democrats.

Update: Oh, here's something else that's interesting. Since we tend to assume that state parties are strong due to state-specific characteristics (state laws regarding primaries, redistricting systems, political culture, etc.), we'd probably think that where one party is extreme, the other is, as well. Nope. If you scale the median ideal points I described above in the direction of greater ideological extremism, here's what you get:
The relationship is negative. California's the big outlier here, with both parties being strong. Otherwise, the typical pattern is for one party to be pretty extreme while the other is pretty moderate. And that pattern shows up more or less where you'd expect it. For example, New York Democrats are very liberal because they can be, while New York Republicans are moderate because they have to be. Flip that around for Mississippi.

(Notes: Maps made with Stata using spmap. Updated for higher resolution.)


  1. Very interesting. However, I find the terms "liberal" and "conservative" problematic. Since the issues that make up those two ideologies are not morally or logically associated, but rather only politically related, it leaves a lot to be desired. It would be interesting to know how those charts look on various issues such as gun rights/control, regulation of industries, religious issues etc. That would tell us which groups are strong in each state. I would guess that they would vary more from state to state in those measures than the degree to which they vary from "liberal" or "conservative." For instance, I bet Mississippi is strongly religious but not particularly economically "conservative." And I understand Vermont is very pro-gun but the least religious state in the country.

    1. The terms "liberal" and "conservative" are admittedly inadequate here. Generally, you can describe people (particularly legislators) who vote consistently Democratic as "liberal," and those who regularly vote Republican as "conservative," but that's a fairly recent feature in U.S. politics, and one that may break down somewhat at the state level, as you describe. I'm using it as a convenient shorthand for strong party behavior.

    2. You can pull some of this information from the raw NPAT responses, subject to the usual response rate, response bias, dogs and cats living together problems.

      I have them for 99/00, and did a quick and stupid principal components on all the issue areas where there were at least a few questions, then predicted the first recovered factor. Like I said, quick and stupid. Each of them is based on few enough responses that the scores are probably more noise than signal, but what the hell. Also there's no good bundle of "economic" issues.

      Anyway, both parties in MS are really conservative on abortion, guns, and environmental regulation. On social issues and education, the D's are conservative for D's while the R's are towards the R center. On health care, both are towards their party's national center.

      In spite of its reputation and minimal firearm regulations, Vermont legislators of both parties tended to have pretty "liberal" scores.

      Overall, the individual issue scores were correlated highly enough (in the .6 to .7 range with only a few notably lower than that) that looking at them issue by issue doesn't really add a whole lot to the story.

  2. This comment has been removed by the author.

  3. I tried clicking the link to the data and got overwhelmed. I am obviously curious about the white states. Mainly, I'm curious about Louisiana. As a Louisiana resident I would say we should be INCREDIBLY red and maybe just a little bit blue, but the map doesn't detail that as such. Is that because our issues tend to be different? I would say geographically, I could agree with that... maybe. But a good deal of our legislation (dealing with education, abortion, etc.) is unfortunately religiously oriented, therefore I would assume again that we'd be incredibly red. We also tend to not help poor people in this state, or generally, the social welfare of any Louisiana citizens. Again, I would assume that we'd be really red.

    1. I just reread the top of the blog, and I realize that it's been taken over the last 20 years. We have had a major shift of 'party politics,' in this state in the past 20 years, which would be the only logical reasoning for why the state of Louisiana is white...

    2. Morgan, I just checked the data, and even in 2010 (the most recent year available), Louisiana has among the most moderate parties. Now, keep in mind that these figures are based on the actual legislative voting behavior of the state legislators. What they're voting on is not factored in. So you may be right that they're voting on very right wing legislation, but they're not voting in rigid party blocs when they do so.

  4. I wouldn't say it's just conservative republicans that are unable to change. Liberals seem eternally preoccupied with culture wars that go back at least as far as the 1960s. The leftist mind in the western world is still at war with the 1950s image of suburbia, where a woman's place was in the kitchen and everybody was white, despite the fact that racial and sexual politics have changed enormously in the last half century. Conversely, the right is still at war with the 1960s image of the counter-culture, even if free love and socialism fell out of favor on the left decades ago. Conservative T-Rex and liberal brontosaur alike are both dinosaurs.

    irene of Skagway AK

  5. The regression line is not a very helpful way of understanding the data in this plot. The r squared is low. But, another way, the significance of the hypothesis that GOP ideology rank order is in a linear function of Democratic ideology rank order, or visa versa, is not very well supported relative to the hypothesis that as a first order hypothesis they a basically independent of each other, with some support for non-linear second order tweaks to the core assumption that they are independent at some level.

    Descriptively, I think it would be more fruitful to conceptualize the Republican data set as a fairly narrow range of variation from about 0.4 to 1.2 with the outlier of the exceptionally weak and moderate GOP parties in the Northeast and Hawaii.

    Meanwhile, the Democratic party is a much bigger tent with a range of ideology values ranging in an approximately uniform distribution in a range from about -0.5 to 1.5 (although always more liberal than the local GOP) a range of ideological variation that is about three times as great as the GOP range excluding outliers.

    The data kind of look like a decent fit to a hypothesis in which each of the two major state parties in any given state have an ideology rating determined by that particular party's past historical trajectory, starting out at the same place and then random walking from the respective starting points for each party, with second order "pulls" at each random walk state (i.e. each election cycle) from Congress and also from the aggregate influence of the other states that are part of the same regional culture (e.g. New England, Mormon West, Plains, South, Pacific). California and New York are plausibly outliers because these states are so large and inward looking in their politics that their random walks were basically immune to influences from other states in the region (although probably not to Congress) their potential to random walk to greater extremes is less restricted. In this kind of model the random walk of the state Democratic parties would be longer than the random walks of the state Republican parties because their most recent point of clear ideological definition is more recent. The modern Democrats ideologically emerge at all sorts of different spots from a last definition around the time of the New Deal, while the GOP has less variation since it has only been drifting apart from a starting point around the time of the Goldwater campaign.

    You'd have to tweak the parameters and do a Monte Carlo analysis to get patterns similar to those seen, but this analysis in less cold and mathematical terms basically implies that state political parties are genuinely best conceived of each as one of many individual non-profit organizations that share a historically remote a common origin, rather than as a mere ideological label. In other words parties are a product of the evolving group of people who choose to participate (to some extent at random) in that state party's political machine, rather than a mere generic, national ideological label that is synonymous with "the more liberal party" and "the more conservative party" in a manner indifferent to who happens to get involved in it. This reflects the experience of those involved that the personal priorities of those who choose to get involved at a serious level in a given state do influence the policies that a particular state party seizes upon.