Update: New graphs here.
Following what we did for 2000 and 2004, we took the Pew pre-election poll data and broke it down by state and income. The numbers are noisy in some of the states (for example, New Hampshire had 15 respondents in the lowest income category and only 11 in the highest category), so we piped the data through a hierarchical model to get a more stable estimate for the voting patterns among rich, middle-income, and poor within each state. Here’s what we got (red and blue states are those McCain or Obama would’ve won):
Update: My original maps were slightly wrong–they were based on unadjusted poll numbers. I re-posted the corrected maps 6.28pm.
The most striking pattern is our estimate that Obama would’ve won almost all the states, if only low-income voters were counted. This is our best estimate for each individual state; however, given the uncertainty in these estimates in small states such as Idaho and North Dakota, we wouldn’t claim with certainty that there was such a clean sweep.
Among rich voters, Obama won in California and some northeastern and midwestern states–“blue America,” if you will. (I don’t really know what to say about New Mexico, but there is such a thing as sampling variability!)
The five income categories I used in the analysis are: 0-20,000; 20-40,000; 40-75,000; 75-150,000; over 150,000. The graphs above show the estimates for the highest, middle, and lowest of these five categories. I assume the numbers represent family income (as reported by the survey respondent).
P.S. I don’t know what sort of technical detail people want on this blog–I got enough complaints on my last couple of posts–so I’m posting the R code for the fitted model here.