Just to follow up on my previous post about Tuesday's Colorado Republican caucus, I've played with a bit more county-level data and found a few interesting correlations. None of these are perfect, and a lot of things are moving simultaneously, so to find the one thing that caused Santorum's upset is a fool's errand. Nonetheless, these are suggestive.
I also encourage you to read Sean Trende's county-level analysis of New Hampshire's and Florida's primaries (he finds some similar things to what I found about Evangelical voters) and Nate Silver's interesting essay today.
Anyway, some graphs after the jump. All bivariate relationships are statistically significant at the p ≤ .05 level.
Turnout in the Colorado Republican caucus (as a percentage of active Republican registrants) was 8.2 percent, ranging from 6.1% in Adams County to 19.9% in Costilla County. And it looks like Santorum did better in the counties with higher turnout. The counties that were more enthusiastic about Romney were less enthusiastic about voting.
More conservative counties (as measured by the 2008 general election vote) were much more likely to vote for Santorum. This correlates highly with the Evangelical vote share.
The 2010 Party Splits
Romney's caucus vote correlates highly with the 2010 primary vote for U.S. Senate candidate Jane Norton. Norton, you may recall, was the candidate with the solid resume, lots of insider support, and a huge monetary advantage whom the activist base nonetheless didn't like, and she lost to a Tea Party-backed conservative. Sound familiar?
At the suggestion of Kevin Ingham and Craig Hughes, here's how income related to the Romney vote: