Statistical Modeling, Causal Inference, and Social Science

Instead of three income levels, poor middle rich, can you treat income as a continuous variable and note the point at which each state "flips"? I'd be interested in seeing a single choropleth map of red white and blue hues showing California and NE as deepest blue, Texas and the West as deepest red, and other states in between as shading into paler hues or white.

My colour adaption of Robert Vanderbei's famous 2004 "Purple America" map is similar.

Thanks for posting these.

I'd love to see a chart showing which states are most polarized along this metric. E.g. MS is; WY isn't.

Is there a scatter plot visualization that is more respectful of the population or electoral votes of the states? The fitted line could be. The effect appears much less apparent in the top few states.

What happens if you adjust for each state's cost of living and/or tax rate? (Living on $50K in NJ is very different than living on $50K in KY.)

How do you associate votes with income? Where do you get your data? When I vote, I don't disclose my income.

Where's Alaska and Hawaii?

Does the same relationship hold if you consider only white voters at the three different income levels?

Similar question: are you defining "middle-income" as only band 3 in your paper?

In the south the "rich/poor" distinction masks a "white/black" distinction. In the southwest, it masks a "white/latino" distinction. Do you have the data to show white rich, white middle, white poor, black rich, black middle, black poor, etc.?

Another interesting feature is that in the poor states, the rich voters are overwhelmingly Republican. In the poor states, the rich and the poor voters are most polarized; voting behavior is highly correlated with income. This correlation is much weaker in rich states.

This all just points back to Andrew's point that interactions matter a lot. You can't generalize to say rich voters vote one way without looking at their state.

Derek: Can you weight your map by population in those counties? So you can see why some states went one or the other despite what appears to be the relative side it seems to be on?

Perhaps a consequence of majoritarian political institutions....

http://weblamp.princeton.edu/rppe/events/apr2003conf/Iversen.pdf

To answer some of the above questions:

1. We discuss cost of living here and here.

2. Data come from pre-election polls and exit polls; see our paper.

3. Alaska is northwest of Washington state; Hawaii is southwest of California.

4. The patterns are similar but not so strong if you only look at white voters.

5. The estimates are based on a model fit to all the data, then we take estimates for the lowest category (0-16th percentile), middle (33-67th), and highest (95-100th).

6. We've looked at the data without the blacks but have not looked at each ethnic group separately.

Did I get disenfranchised, again?
Denny, Alaska

Nice maps. You should add the missing states though... :) Our artic and tropical paradises are important too.

It would be interesting to elect the president based upon who won more states (or electoral votes from those states) across all income ranges.

No. Wait I meant we should use the popular vote. Why should someone else vote be worth more or less than mine.

Why should Wyoming's 500-600 thousand people get 3 votes and New Jersey's 8.5-9 million people get 15 votes? That's absurd!!! In NJ that's 1 vote for every 580 thousand people or so.... in Wyoming they are getting 1 vote for every 180 thousand????

Look at the places Bush won across all income levels. Tell me that's a majority of the country. Pleazzzzee.... These are the states with the lowest populations and Texas. That add's up to what? Homefield advantage and an advantage in the electoral college.

Denny,

States #49 and 50 are on the scatterplots.

Jason,

I'm with ya on the electoral vote imbalance. But Bush did win the popular vote in 2004.

This voting pattern raises an interesting paradox: Obviously, we now know that the GOP's economic policies creates more poor people (witness the last eight years, QED), which means that their policies undermine their electoral base. But the real stinger is that the Democrats, by creating more rich and middle income people (witness Clinton's eight years, QED) are ALSO undermining their own electoral prospects.

Do you guys have a similar analysis based on education level?