Since I posted on a really bad study that’s outside of my area of expertise the other day, I thought I should make it up to you by posting on what I think is a good study by Gelman et al. that’s also outside of my area of expertise today. Plus, with a title like “Rich state, poor state, red state, blue state: What’s the matter with Connecticut?” (via the Columbia University stats blog, where Gelman is a contributor) it has to be blogged (the entire study can be read at that link). I mostly read it to learn about the methodology (interesting use of regression, if you’re interested in that sort of thing), but unless you understand regression, the analysis is a bit heavy, so I’ll just talk about the results. Fortunately, those are pretty cool too, especially if you dislike David Brooks.
Here’s the idea behind the study. There are conflicting opinions about how income affects voting patterns: the belief that liberals are rich elitists while Republicans are mostly low and middle-income folk, for example, or the belief that Republicans are rich, fat, oil barrens screwing over the little guy while poor people vote Democrat. To figure out which one is true, there are several levels at which you could analyze income and voting data. Gelman et al. picked four (from p. 5 of the manuscript):
- Aggregate, by state: to what extent do richer states favor the Democrats?
- Nationally, at the level of the individual voter: to what extent do richer voters support the
- Individual voters within states: to what extent do richer voters support the Republicans,
within any given state? In other words, how much does context matter?
- Counties within states: to what extent do richer counties favor the Democrats, within any
When they aggregate by state, they find that, as many have noted, wealthier states vote Democrat, and poorer states vote Republican. However, when they looked at individual voters nation-wide (using data from the National Election Study), they found that wealthier individuals tend to vote Republican. Richer individuals also tended to vote Republican at the state level, too. So at the first three levels of analysis, the results show why there’s so much confusion about how income affects voting. Richer states vote Democrat, and it’s easy to infer from that fact that richer individuals vote Democrat, but at least on a national scale, richer individuals tend to vote Republican. The county-data, however, is where things begin to get confusing. Here are the graphs from the counties nationwide, just counties in southern states, and just counties in non-southern states (I know it’s kinda small, but you can see them nice and large on p. 12):
Those are regression coefficients, but all you really need to know is that anything above the horizontal line is positive, and means that counties with higher incomes are tending to vote Republican. Notice that the nationwide graph (left) shows that higher income counties vote Republican (except in 2000), and that the middle graph shows that this effect is strongest in southern states, but disappears in nonsouthern states (graph on the right; if you want to see graphs for individual states, look at p. 13-15). Hmm…
Something strange is going on here. Why is there a relationship between income and voting in the counties of some states, but not others? Why the south (O South, your torrid suns!)? Well, here’s a graph using 2000 presidential election results that helps to explain what’s going on (in Gelman’s post, as well as the paper):
The three states, Mississippi, Ohio, and Connecticut were picked because the first is poor and voted for Bush, the second has a medium average income, and the vote was very close in 2000, and the third because it is a wealthy state that voted for Gore in 2000. What the graph shows is somewhat surprising. In poor states, poor people vote Democrat and wealthy people vote Republican. The same is true in a medium-income state, though less so than in the poor state. In the wealthy state, however, income has very little effect on voting: both the rich and the poor vote Democrat and Republican. That’s what’s wrong with Connecticut: it’s wealthy enough for income not to affect how people vote. And that’s what’s going on in the south. Since southern states tend to be very poor, income matters more. Gelman et al. offer some reasons for why this pattern exists, including that in poor states, economic issues are more salient, causing peopel to vote based on economic interest rather than social issues and things like the war, but really, who knows?
Let this study be a lesson to you, then. Anytime you want to overgeneralize based on data from one level (e.g., which states vote for which presidential candidates), think again. Trends in human behavior are often much more complicated than they first appear.