I’ve been thinking for awhile about doing a project on political polarization in the contemporary US, since the level of it is starting to remind me of other democratic systems in the process of destabilization (specifically, Spain leading up to 1936), and I’m thinking that it’s time.

One easy way to start would be to take the county-by-county election returns (or precinct by precinct, if we really wanted to break it down) from each of the last several presidential elections (I’m thinking 2000, 2004, 2008, 2012, and 2016) and measure the difference between the top two finishers. That is to say, I’m focusing on margin of victory here as a proxy for polarization (or as its outcome, if you want to think about it that way). I could then track, for each county, how that margin has changed since 2000 relatively easily and plot a linear trendline.

Then, mapping the slope of the trendlines (yes, I know this assumes a linear model), I think we could see some interesting results. My hypothesis is that the margins will have increased, particularly in the deep blue states (though also likely in the deep red ones).

Any suggestions from peoples with experience either with linear models (@rsharp, @katealamode) or political mapping (@rvanheertum) is totally welcome.


That’s interesting, but I might use other metrics like atttitudes about the other party, attitudes about opposing candidates, how many people split their ticket, etc. Some of these attitudinal figures should be available in poll form, but Pew has certainly been tracking them for a while in various forms. Just a thought on an alternative. As to your approach, I’m not sure how close elections are really speaks to how partisan people are. Gore vs. Bush was extremely close, but the argument there was there was little difference between the two candidates (though how wrong those who argued that turned out to be!). The same could be said of Nixon/JFK in 1960 – there were clear differences, but a lot of similarities as well (as opposed to, say, LBJ vs. Goldwater four years later).


@rvanheertum, I guess the question that I was thinking of, rather than the one I said, was about whether or not the polarization between rural and urban America has been getting stronger (my hypothesis is yes). The presidential data was always kind of an “I know it’s available and pretty easy to get in crunchable formats,” so is kind of a proof of concept. Correlating with congressional elections is harder because of gerrymandering and other less+nefarious forms of shifting boundaries. But still interesting to also do.

The Pew data was actually a thought for a part 2 of this, by the way. I’m also curious to see if one can measure the correlation between distrust in government as a whole, tea party voting, and population density.


Oh, then that makes sense. The further back you can go would be helpful, if you are planning on running a time series. You could also code the differences into ranges, which might make the results more robust. I would think some control variable would be helpful, though I’m not sure what the data set looks like.

You must be logged in to join this conversation. Login below or create a new account.