The Secret of Santorum's Success?

I've been mostly focusing on the Colorado results from last night's Santoruming, which, I'll admit, caught me by surprise. Those seeking to understand what happened here should begin with the knowledge that the Colorado GOP is far from monolithic. Yesterday, Micah Cohen at fivethirtyeight ran a nice geographical profile of the state's Republicans (featuring the occasional blatherings of Kyle Saunders and myself). Cohen notes that the business Republicans in the Denver metro area don't necessarily vote like the libertarian Republicans on the Western slope, who don't necessarily vote like the Evangelicals in Colorado Springs. And last night's map showed some geographic diversity, with Romney doing well in the Denver suburbs -- but not by enough -- and Santorum taking just about every other part of the state (with the exception of the sparsely-populated northwest).

So how did Santorum end up taking the state, when Romney dominated it four years ago and when Romney had the backing of an impressive array of Republican leaders, including the state's most recent Republican governors and U.S. senators and the current speaker of the statehouse?

I was curious about the role of Evangelical Protestants in the state. Here's a telling scatterplot, looking at the percent of religious adherents in each Colorado county who are Evangelicals and Santorum's caucus vote share:

And yes, that regression line is statistically significant at the p ≤ .05 level. I know the data are a bit noisy -- chalk that up to caucus results in small counties, where sometimes only a few dozen people actually turn out to vote.

Below is a multivariate regression equation in which I use the percentage of Evangelicals, Catholics, and Mormons to predict Santorum's county-level share of the caucus vote in Colorado. I also include a variable measuring whether there was a Tea Party organization in the county in 2010. The only predictor that is statistically significant is the Evangelical percentage:

    santorum |      Coef.   Std. Err.      t    P>|t| 
-------------+---------------------------------------
    evangper |   .2715796   .1125573     2.41   0.019
 catholicper |   .1328817   .1060132     1.25   0.215
   mormonper |  -.2968235   .1923018    -1.54   0.128
  teaparty10 |  -2.368753   3.282964    -0.72   0.473
       _cons |   32.19836   5.826821     5.53   0.000

Meanwhile, here's the same regression predicting Romney's share of the caucus vote. The Evangelical variable is just on the edge of significance, suggesting that the more Evangelicals, the lower the Romney vote. Conversely, the more Mormons, the higher the Romney vote. No, I cannot explain why the Tea Party variable is positive and significant here. This may be spurious, or it may indicate that Tea Partiers who were not Evangelicals backed Romney:

      romney |      Coef.   Std. Err.      t    P>|t|

-------------+---------------------------------------

    evangper |   -.193941   .0985893    -1.97   0.054

 catholicper |  -.0887819   .0928572    -0.96   0.343

   mormonper |   .5840437   .1684377     3.47   0.001

  teaparty10 |   5.884603   2.875557     2.05   0.045

       _cons |   32.35636   5.103729     6.34   0.000

Again, these are county-level data, which may not be the best level to examine such questions. Nonetheless, what does this mean going forward? Is this evidence that Evangelicals will not accept Romney due to his religion? Or that Romney supporters lack the enthusiasm to seal the deal? I noticed this interesting tweet from David Leonhardt this morning:

Streak continues: Romney has won every state Hillary did in '08 (NH, FL, NV) and lost all those she lost (IA, SC, MN, MO, CO).

Here's the important difference: she lost all those other states to Obama. Romney lost those other states to two different people. If there were one non-Romney sucking up all the votes and delegates, that would be a serious threat, but at least so far, that's not happening.

Romney still has what his rivals lack: the money and institutional backing to be competitive everywhere at the same time. We'll see evidence of that on Super Tuesday. Gingrich and Santorum may continue to make him bleed, but I still don't see how either of them deprives him of the nomination.

Update: More telling scatterplots here.