The recent murders at a Colorado Springs Planned Parenthood center, and the political arguments that have followed, remind us once again of the connection between political polarization and terrorism.

We are used to the idea that Democrats support legal abortion while Republicans oppose it, but this has not always been the case. This came up here a couple of years ago, but I think it’s worth returning to the topic now that it is in the news again.

To start with, here’s the trend in public opinion from the National Election Study. For each election year we show the average opinion on abortion on 1-4 scale (where 1 = abortion should never be legal, and 4 = abortion should always be legal) among self-declared Democrats, Independents, and Republicans:

I made the graph a few years ago, but I don’t think the polarization has gone away. Sample sizes are small in some years, so I wouldn’t recommend trying to interpret every jump in those lines, but the basic pattern is clear: no partisan polarization on abortion before 1992, lots since.

Yair Ghitza and I then broke these data down by ethnicity to see where the polarization was happening. And here’s what we found:

These lines show estimated regression coefficients predicting abortion attitude from partisanship (that is, predicting that response on the 1-to-4 scale from a party-identification variable that takes on the values -1, 0, 1 for Republicans, Independents, and Democrats), using a hierarchical model to get stable estimates for each year and for each of four ethnic groups.

As you can see, almost all the polarization is occurring among whites. (Also some among the “other” category, but they represent only a small fraction of the electorate.)

We then continued and broke down the whites by income and education:

Polarization is concentrated among upper-income and well-educated whites.

And now we can return to the news story about the killer:

He was an independent art dealer with a degree in public administration from a Midwestern college . . . He was generally conservative, but not obsessed with politics.

This is N=1, and of course non-college-educated people can perpetrate political violence as well. But I find the statistics helpful in understanding partisan polarization more generally.