Renters make good Democrats, and other demographic observations

You've probably noticed that it takes certain kinds of congressional districts to elect certain kinds of Representatives. At an instinctual level, you can probably guess that a mostly non-white district in a northeastern city is going to elect a liberal Democrat, and a mostly white district in the rural south is going to elect a conservative Republican. Things get a little more complicated with, say, a middle-class suburban district in the west; that's the kind of place where candidate strength, fundraising, and what the direction the national winds are headed all play a role. But there a lot of other variables that go into shaping a district's makeup, and those variables can tell us something about a district's political potential.

Why talk about this now, with the next election more than a year and a half away? Partly because now is the time when targets get picked and candidates get recruited; competitive races don't usually just pop up out of sheer will but require a lot of groundwork. But more importantly, the Census Bureau finally graced us last month with demographic information for the nation's congressional districts. Although the most recent Census has been in the books for several years now, things got slowed down by the redistricting process (which, of course, relies on the Census' initial population figures); they had to wait until the new district lines were finalized to be able to calculate new district data.

With access to that data, finally, I initially planned to write a piece about the various superlatives in congressional districts (whitest districts! poorest districts! best educated districts! and so on). That's interesting trivia, of course, but by itself doesn't tell us much about how we can reshape the House in 2014 and in future years, so I also decided to pinpoint Republicans in the districts with the demographic categories that seemed most hostile to them (say, for example, the five congressional districts with the highest percentage of African-American residents that are still represented by Republicans).

Rather than put up dozens and dozens of tables, though, that left me wondering: which variables actually matter the most? Which particular demographic categories are most strongly related with whether a district tends to elect a Democrat or Republican? That way, we could focus on only a few most important categories. So, with that in mind, I calculated correlations for each of the categories in the Census' release, factored against the percentage the Democratic candidate for the House got in each district. Some of the results are predictable, but others were a total surprise. Here's a chart of the characteristics that had the strongest positive and negative relationships with Democratic share of the House vote:

Some of the other things that you'd think might matter turn out not to matter much at all. For instance, the correlation coefficient on median household income is only 0.02, meaning no relationship in any direction. (With correlations, 1 or -1 means a perfectly corresponding relationship within the data, while 0 means nothing but random noise.) It's tempting to think of the Republicans as the "party of the 1 percent" and to think of all the Democrats representing blue-collar districts in the cities, but stop and think about the number of affluent suburban districts that elect Democrats, or the number of abjectly poor areas in the Appalachians that elect Republicans.

Follow over the fold for full discussion on why these factors might matter, and the promised lists of Republicans vulnerable according to these criteria ...