As I’m sure you know by now, I’m interested in differences between rich and poor. Higher-income people are more likely to vote Republican, and we’ve seen this in many different subgroups of the population. Among whites, among blacks, among religious attenders, etc., the poorer voters among these subgroups are more Democratic and the richer ones are more Republican.
This got me wondering: What are the subgroups of the population for which this isn’t true? Or, more generally, how do rich and poor differ in their voting patterns, in different subgroups of the population.
Here’s we found, courtesy of the 2000 and 2004 Annenberg surveys. For each group, we’re looking at Republican share of the two-party vote intention among people in the upper third of family income, minus Republican share … among people in the lower third of family income:
(Click on any of these graphs to see larger versions.)
A striking pattern. The differences between rich and poor are much larger among conservative, Republican groups than among liberal, Democratic groups. At the very bottom of the graph above, you see a few groups where richer people are more likely to vote Democratic. All of these are groups that are mostly liberal and Democratic.
To look at it another way, we made a graph showing the different subsets, plotting rich-poor voting differences vs. average Republican vote for the group. Separate graphs for 2000 and 2004:
Pretty consistent, I’d say. Now we have to think about what this all means.
P.S. The graphs were made by Dan Lee (who’s not the same as Daniel Lee, who made a bunch of other graphs that I’ve been posting recently). I’m a lucky guy to have been able to work with two different Dan Lees.
I don't understand what the X & Y axes are in the graphs below the jump. Does the X axis tell me that in 2000, Blacks and Jews were the least Republican of your groups? And the Y axis, that in 2000, among the 'Very Liberal' and 'Non-citizens', the share of the top 3rd who voted Republican was 15%-20% larger than the share of the bottom 3rd? How's that work for non-citizens? Vote fraud?
Marcel:
In the lower graphs, the x-axis is the Republican-ness of the group (as measured by % favoring the Republican candidate for President, minus % favoring the Democratic candidate for President), compared to the U.S. average. The y-axis is the difference in Republican vote preference, comparing people in each group who are in the top third of U.S. family income, comparing to those in the bottom third.
And, yes, it's vote preference, not actually votes. Which is why non-citizens can be included, I guess. I'll have to go back to the surveys and check that.
Seems like it maybe saying something about the social-conservative/econ-conservative distinction. If we say that the large differences in rich/poor voting in a group signals some kind of polarization in that group then it seems reasonable to say posit that the poor in groups like Catholics or Mormons and to a lesser degree Asians, Blacks, and Jewish are socially conservative but not particularly econ-conservatives.
Those highly varying half dozen are an interesting subset.
I did a lot of work on the "lost" 2002 midterm exit poll, crunching the raw data, and one thing I found was that a lot of the ideological terms used in the questions weren't being understood by all the respondents the way the designers of the poll wanted them to be understood. For example, as I recall, a large fraction of blacks said they belonged to the "Christian right."
The "Tyler Perry demographic" — black church ladies — tend to tell pollsters that they are "conservative" but that doesn't mean they are conservative in the voting booth.
Prof. Gelman,
What is "Orthodox"? Is that Greek/Russian/Eastern Orthodox Christian?
Prof. Gelman,
When you guys sliced this up, did you set the cut points for each third by the overall income levels of the US population, or by the income levels of each subgroup?
Ceolaf: Yes on Orthodox, I think. In answer to your second question, the cutoffs are upper and lower thirds of U.S. family income, not income levels by subgroup.
In general, I think it would be more informative to report on each racial group separately.
For example, blacks of every single subgroup vote so overwhelmingly Democratic that including them just adds more noise than value to this type of graph.
The truth is that what people are mostly interested in is what factors account for why whites are so divided between blue and red.