Statistical Modeling, Causal Inference, and Social Science: What does the 2004 Election Say About the Red/Blue Paradox?

What does the 2004 Election Say About the Red/Blue Paradox?

Do we still see an (income) paradox in 2004? Let's first look at the state level. A quick correlation between median family income and percent Republican vote shows a -0.41 pearson's (and -.46 spearman) correlation. Both are significant. So at the state level, it looks like lower income states are voting for the Republican candidate and higher income states are voting for the Democratic candidate.

What about the individual level? Let's look at the exit polls.

R D
<15K 36% 63%
$15-30K 41 58
$30-50K 48 51
$50-75K 55 44
$75-100K 53 46
$100-150K 56 43
$150-200K 57 43
>$200K 62 37

So it looks like the paradox is alive an kicking. So do we still believe it's an aggregation problem? Is the paradox only alive in rural areas, but dead in urban areas? More to come...

Sam Cook

Aleks commented:

From these excellent visualizations
(http://www-personal.umich.edu/~mejn/election/) it seems to me that this paradox should be tackled with a mixture model.

Sam Cook

Andrew commented:

Aleks,

I'm not sure why a mixture model would be needed at the county level--although I see why it's appropriate for individuals, who are divided into Reps, Dems, and others.

Regarding the maps: they're pretty, but I don't like the stretching of the areas in order to be population-proportional. This stretching introduces visual artifacts. I'd prefer, e.g., a map with a red dot for each 10,000 R voters and a blue dot for each 10,000 D voters.

The key flaw, I believe, is everyone's implicit assumption that a map has to be fully filled-in with color.

Sam Cook

Aleks commented:

There is a visualization for your preferences too! http://www.averdevelopment.com/election/ Still, observe the rather distinct edges between densities in some areas.

I'd nevertheless check the utility of a latent variable describing the "type" of each county. I'm slightly more used to mixture models in such cases because of interpretability.

The question is whether the vote aggregations are purely due to correlated demographics, or whether there are additional county-level or state-level effects.

Aleks