There is an intersting post over on Monkey Cage purporting to show that there is no effect of income on voting pattern in blue states, but there is in red states.
By way of evidence, we get two graphs - one for Maryland(blue/rich) and one for Texas(red/poor) - with counties plotted by income vs. George Bush’s 2000 vote share. Helpfully highlighted are the two relevant data points at either extreme. For Maryland it’s Baltimore vs. Montgomery, for Texas it’s Collin vs. Zavala.
Click over and have a look at the graphs. For Maryland, you will notice a pattern that is essentially flat. There is little, if any, interaction between Bush’s vote share and county income. Maryland voters gave Bush just over 50% support regardless of income.
For Texas there is a kind of pattern, but it’s a little more complicated and a lot less salient than the commentary paints it. The first thing you notice is that income distribution in Texas is a lot more uniform than in Maryland. Almost all counties are clustered between 25k and 45k. Within this cluster, there does seem to be a pattern of richer areas voting for Bush more than poorer areas, but there’s a huge amount of variation, and nearly all of them have rates of support greater than 50%. However, it is clear that no counties that gave Bush less than 50% support are in the upper income groups. Vote shares of less than 50% are made up exclusively by counties below the 30k range, and that is indeed striking. Also striking, however, is that support for Bush levels out at around 75%, and few of the counties where people earn more than 50k on average are very high above this line. Nearly all of Bush’s strong support, in other words, comes not from the richest counties, but from the upper part of that “cluster” between 25k and 45k. Once you get much past 45k, the pattern is the same flatline as Maryland’s, albeit support is in the lower 70s rather than the lower 50s.
I’m not entirely sure what the post’s point is, but it seems to be that the relevant “red-blue” divide is actually an income divide. If that’s the point, then it’s interesting for being half-right and half-wrong. It’s half-right in that there does seem to be an effect of income on voting pattern, but probably only at the lower margin. That is, the extremely poor tend to vote Democrat, but once you get out of the lowest brackets income ceases to be explanatory for voting pattern. (Remember - both Texas and Maryland show no interaction between income and Bush’s vote share above around 35-40k.) The conclusion is half-wrong in the sense that for the remaining voters - that is, those counties with average incomes above 30k - “Texas” and “Maryland” DO seem to be explanatory variables. The relevant counties in Texas gave Bush more than 70% support; in Maryland it was just over 50%. In other words, once you get above the “low-income” threshold, then “state” takes over from “income” as the lion’s share of the explanation. So in that sense it’s the opposite of what seems to be claimed. It isn’t that in poorer states - where “income matters more than in rich states” - the poor vote Dem and the rich vote Republican, it’s that the extremely poor vote Dem regardless (one assumes) of location, and over and above that other factors play a role.
Of course, it’s impossible to tell anything of substance from Maryland (chosen because the Brooks column the post is responding to used a Maryland-based example) and Texas (no explanation for choice given aside from it being a “red state”) alone. To conclude anything about voting patterns by state, we obviously need more than two data frames. Those will apparently be supplied by the authors’ upcoming book, and perhaps the argument is more convincing there (there’s a wild thought - cherry-picked data being less convincing???). Extrapolating from the data given, though, it seems difficult to imagine that the authors will be able to sustain their income-based explanation. Though income probably does play a role in the manner I noted (as a “threshold”), the pattern seems likely to much more complex that this commentary would suggest.