Vote swings are (approximately) uniform across the nation, so there’s no point in focusing on median America

by Andrew Gelman on March 8, 2012 · 3 comments

in Campaigns and elections

I got the following in the email:

Bush-Obama America

There are 272 Bush-Obama counties across the country. George W. Bush carried these counties in 2000 and 2004. In 2008, Barack Obama won them. Now where do they go in 2012? I [David Dent] am launching a blog to focus on these pivotal counties to the outcome of the Presidential Election on November 6. After Barack Obama turned these counties blue in 2008, I began a journey to explore the state of the American Dream in Purple America, which is the subject of my next book. We begin presenting views of these slices of America, with the mission of sharing the voices of the everyday people who will decide the election. On April 2, we will expand to the presentation of video portraits of at least three Bush-Obama counties a week. The blog’s journey will take you to more than 40 American states to meet a diverse group of people who represent a dynamic and transformative element of America.

As qualitative reporting, this might be just fine—-I have no idea—-but for the goal of understanding American voters and the election, there’s no particular reason to focus on these swing counties. Given recent political history, we can expect the vote swing from 2008 to 2012 to be approximately uniform, which means you could look just about anywhere for the anticipated 4%-or-so swing toward the Republicans this year.

{ 3 comments }

Dynotec March 9, 2012 at 4:55 am

County to county swing from election to election has a standard deviation of about 5 points. Saying “Oh, it’ll be 4 +/- 10″ isn’t very interesting.

And it’s not random, swing exhibits a *lot* of regional auto-correlation that’s well correlated with ancestry/race/culture/region. And once you restrict only to white vote, you see much more variation and regional autocorrelation than before.

These trends, far from being irrelevant, are the *basis* of long-term political shifts in the US.

Andrew Gelman March 9, 2012 at 10:00 am

Dynotec:

I agree that there is some variation and it is interesting (indeed, I’ve made graphs and written about such variation) but there’s no need to focus on counties that are near 50%. A county that moves from 30% R to 40% R (say) is just as worth looking at.

Bill Bishop March 9, 2012 at 11:42 am

We mapped the flippers back in 2008. Map is here: http://www.dailyyonder.com/them-changes-counties-switched-parties-2008/2008/11/13/1758 It is interesting (visually anyway) that a bunch were in the Midwest. We have the full list of counties if anybody wants ‘em.

As I recall the standard deviation of the R vote in 2008 increased from 2004 — and the divide among counties increased within most states.

Comments on this entry are closed.

Previous post:

Next post: