(This article was originally published at Statistical Modeling, Causal Inference, and Social Science, and syndicated at StatsBlogs.)
After I posted this recent comment on a blog of Steven Pinker (see also here), we had the following exchange. I’m reposting it here (with Pinker’s agreement) not because we achieved any deep insights but because I thought it useful to reveal to people that so-called experts such as us are not so clear on these issues either.
AG: I noticed your article on red and blue states and had some thoughts. . . . The short summary is that I think that your idea is interesting but that, as stated, it explains too much, in that your story is based on centuries-long history but it only fits electoral patterns since the 1980s.
SP: Though the exact alignment between red and blue states, political parties, and particular issues surely shift, I’d be surprised if the basic alignments between geography and the right-left divide, and the issues that cluster on each side of the divide, have radically changed over the past century. (Obviously if you define “red” and “blue” by the Republican and Democratic parties, they’ve changed radically, but that isn’t what I was referring to.) Certainly the difference in rates of violence go back centuries (I show this in Better Angels, using data from Randolph Roth), and at least qualitatively, some of the other differences (hawk-dove, corporal punishment in schools, capital punishment, religion in public life, liberal attitudes on race and sexuality) go back at least several decades, probably longer. This is all an interesting empirical issue. Do the GSS data go far enough back, particularly disaggregated by region, to see how many of the signs of the correlations have flipped?
AG: Regarding politics, the Democras have been on the left and the Repubicans have been on the right since the late 1960s on just about every issue. But the red state blue state division that you see in the maps has really been happening only since the 1990s. And if you set aside the south and racial issues, the Democrats have been to the left of the Republicans since 1932. Just to take a couple examples, California used to be a solid Republican state and New York State used to be around the national median. So, to the extent the right/left divide is reflected in voting in presidential elections, yes, the basic alignments have change a lot.
Regarding correlations in issue attitudes, we have a paper in the American Journal of Sociology from a few years ago. We used the National Election Study. Corrleations among issue attitudes have been gradually increasing along with partisanship. My guess is that correlations between ideology and geography were lower 40 years ago.
I believe you when you say that certain geographic patterns go back centuries, but issue attitudes depend on so many other things. For example, in the two world wars, the east coast was interventionist and the midwest was isolationist. That makes sense based on ethnic sympathies (lots of people of English descent in the east and German descent in the midwest) and also economics of international trade. It’s not necessarily anything to do with deep attitudes toward the military or violence. That’s just one example. My point is not that it’s a bad thing to look for deeper connections, just that if every issue is complicated by particulars, you won’t necessarily get the right answer even if you look at lots of issues at once, if you try to map voting (“red and blue states”) or issue attitudes to coherent mindsets.
SP: Putting aside the red (blue?) herring of Democratic vs Republican parties, I still get the feeling that there’s a principal component of more-or-less left-right opinion (across issues), and a correlation with a North+Coasts vs. South+Mountain-States geographic vector, across a century or more—though I’m prepared to believe that there is a lot of variance that isn’t accounted for by either, and that both the variance accounted for by the LR component and the magnitude of the correlation have shifted over time. Mind you, this is based on qualitative historical readings by Fisher, Nisbett, Sowell, and others—I don’t have the kind of quantitative data that would settle it.
The real point, I suppose, is that neither Pinker nor I are experts here. But sometimes the experts get confused too. See this post from a few years ago, where I discuss the celebrated twentieth-century partisan reversal and argue that an interpretation of this pattern by actual experts, published in the American Political Science Review, was flawed. One reason that there’s an opening for quasi-outsiders such as Pinker and myself to advance hypotheses here is that there remain lots of data sources that have not been well integrated into existing historical writing.
Please comment on the article here: Statistical Modeling, Causal Inference, and Social Science