Per commenter FreeDem’s request, here are some other ways to interpret Ron Paul’s primary performances in each state, building on the statistical models estimated here.
The first exercise is to simulate what Paul’s percentage of the vote would have been in each state had they all voted at the same time with the same “institutions” (candidates on the ballot, primary vs. caucus, closed vs. open, and so on). For this exercise, I assumed that all calendar and institutional characteristics were at the median for the nation as a whole. Thus, I assume a primary not a caucus, a closed primary rather than an open or semi-open one, three other candidates on the ballot, a ballot date after Mitt Romney had built substantial momentum with his Michigan victory on Feb. 28 — but before his virtual mathematical clinching on Apr. 2, and no home state advantage for any candidate.
Under these conditions, Paul would have obtained 13.9% of the vote in the average state. Here is how he would be forecast to do in each state, in alphabetical order:
| State | pred12 |
| Alabama | 7.39% |
| Alaska | 15.37% |
| Arizona | 8.70% |
| Arkansas | 13.77% |
| California | 12.17% |
| Colorado | 10.81% |
| Connecticut | 15.94% |
| Delaware | 12.48% |
| District of Columbia | 12.17% |
| Florida | 10.14% |
| Georgia | 12.20% |
| Hawaii | 12.30% |
| Idaho | 11.53% |
| Illinois | 9.68% |
| Indiana | 15.93% |
| Iowa | 19.70% |
| Kansas | 13.73% |
| Kentucky | 14.82% |
| Louisiana | 10.51% |
| Maine | 19.48% |
| Maryland | 13.58% |
| Massachusetts | 17.73% |
| Michigan | 16.80% |
| Minnesota | 12.71% |
| Mississippi | 6.53% |
| Missouri | 12.76% |
| Montana | 14.80% |
| Nebraska | 11.26% |
| Nevada | 10.09% |
| New Hampshire | 31.74% |
| New Jersey | 11.75% |
| New Mexico | 12.29% |
| New York | 17.28% |
| North Carolina | 12.63% |
| North Dakota | 15.54% |
| Ohio | 12.80% |
| Oklahoma | 16.45% |
| Oregon | 15.12% |
| Pennsylvania | 26.43% |
| Rhode Island | 27.08% |
| South Carolina | 16.29% |
| South Dakota | 15.40% |
| Tennessee | 13.42% |
| Texas | 15.57% |
| Utah | 5.39% |
| Vermont | 31.41% |
| Virginia | 24.61% |
| Washington | 15.81% |
| West Virginia | 12.54% |
| Wisconsin | 13.85% |
| Wyoming | 13.27% |
The rank order of states here is of course exactly the same as in the table of residuals from my earlier post, but the presentation of the numbers is perhaps more intuitive.
The next analysis looks at why Paul improved from 2008 so much in some states and little if any in others. In my last post I found that Paul’s electoral performance in 2012 is predicted by state opinion ideology, with more left-wing states giving more support to Paul. However, 2008 LP support is not at all correlated with Paul support in 2012. I inferred that Paul received a great deal of support from voters who would otherwise go Democratic and from liberal Republicans and independents, particularly since the association between ideology and Paul support strengthened conditional on an open primary.
However, I did not control for how “pro-Ron Paul” a state was in 2008. FreeDem reports seeing some odd state-level differences between Paul’s 2008 success and 2012 success when eyeballing the data. Using my analysis of the 2008 returns, I now try formally testing whether Ron Paul’s support in 2008 was correlated with his support in 2012. I regress the log of Paul support in 2012 on the institutional variables, Democratic-Green vote share in 2008, and the log of institution-corrected Paul vote share in 2008. Here are the results:
. regress lnrp caucus12 open12 closed12 ocb clinch home demgr08 lnrp08 Number of obs = 49 F( 8, 40) = 9.67 Prob > F = 0.0000 R-squared = 0.6592 Adj R-squared = 0.5911 Root MSE = .31484 ------------------------------------------------------------------------------ lnrp | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- caucus12 | .5744784 .1465334 3.92 0.000 .2783233 .8706335 open12 | .0946405 .1444686 0.66 0.516 -.1973414 .3866224 closed12 | -.1744016 .1367005 -1.28 0.209 -.4506836 .1018805 ocb | -.1192173 .0418184 -2.85 0.007 -.2037355 -.0346991 clinch | -.1881422 .1700441 -1.11 0.275 -.5318141 .1555297 home | .0832717 .2090593 0.40 0.693 -.339253 .5057963 demgr08 | .0091053 .004799 1.90 0.065 -.0005938 .0188044 lnrp08 | .4681954 .1376803 3.40 0.002 .1899331 .7464578 _cons | -.5945686 .6079726 -0.98 0.334 -1.823327 .6341899 ------------------------------------------------------------------------------
Now I do find evidence that Paul’s support in 2012 depended on the strength of the Paul constituency in 2008, controlling for everything else. Although the statistical significance is strong, the substantive import is in a sense strikingly low. Since the dependent and independent variable are both in logs, we can interpret the coefficient on Ron Paul support in ’08 — 0.468 — as an elasticity. In other words, for every 1% more votes Paul got in 2008, he got just 0.5% more votes in 2012. That tells us that while Ron Paul’s base turned out again for him in 2012, his additional success was probably not related to picking up a bunch of semi-libertarian marginal voters who might have gone with him had his campaign been stronger in 2008. Instead, he reached a whole new constituency. Paul more than doubled his vote share between 2008 and 2012. That mostly did not come from mobilizing more libertarians. He got the libertarians to vote for him again, but most of his new votes came from people we would not normally think of as libertarians.
Who exactly were these new Paul voters? We can find out more, I think, by looking at individual states where Paul improved or did not improve between elections. Here are some states where Paul did badly in both years (adjusted vote shares from 2008 and 2012 in parentheses): Alabama (2.9%/7.4%), Utah (3.0%/5.4%), Mississippi (1.9%/6.5%), Arizona (4.4%/8.7%), and Illinois (5.4%/9.7%). These are states without many libertarians and also without many social and foreign policy moderates and liberals. We could throw in West Virginia (2.9%/12.5%), Florida (3.8%/10.1%), New Jersey (4.9%/11.8%), Delaware (4.2%/12.5%), and California (6.3%/12.2%) as also being of this type.
Then we have states like Idaho (10.8%/11.5%), D.C. (7.8%/12.2%), Montana (7.8%/14.8%), North Dakota (7.1%/15.5%), South Dakota (8.6%/15.4%), and Washington (8.2%/15.8%), where Paul did well in 2008 but did not improve much in 2012. He seemed to hit a ceiling in places like these. My hypothesis is that these states have a lot of libertarians but not many social and foreign policy moderates who are willing to take a Republican ballot. In all the states I just mentioned except perhaps D.C., social conservatives tend to dominate the state Republican party. Washington may be a partial exception, but only a partial one. Washington certainly does not have the tradition of moderate Republicanism that, say, Massachusetts and Connecticut do.
Then there are states where Paul underperformed in 2008 and then improved significantly in 2012: Connecticut (4%/15.9%), Indiana (4%/15.9%), Iowa (3.8%/19.7%), Kentucky (3.8%/14.8%), Massachusetts (3%/17.7%), Oklahoma (4%/16.5%), Rhode Island (6.1%/27.1%), South Carolina (4.3%/16.3%), and Vermont (6.4%/31.4%). New York (6.7%/17.3%) is a state where Paul went from slightly above average to well above. What at least some of these states have in common is a lot of former Republicans and independents who are moderate to liberal on social and foreign policy issues. (I can’t explain Oklahoma, but South Carolina’s improvement might have to do something with a bounce from New Hampshire, and Kentucky with Rand Paul’s prominence.) In Vermont, especially, I will bet my bottom dollar that most Ron Paul voters were centrists or even progressives ticked off at wars, banker bailouts, and civil liberty infringements, but not necessarily signing onto Paul’s whole economic program. His outstanding improvement there may also have something to do with spillover from New Hampshire.
Finally, there are only a few states where Paul did very well in both years: Maine (6.9%/19.5%), Michigan (7.2%/16.8%), New Hampshire (11.2%/31.7%), and Pennsylvania (7.8%/26.4%). These seem like states with both quite a few libertarians and lots of moderates and progressives willing to take a Republican ballot. While economic insecurity might be another hypothesis for the increase in Paul support, that wouldn’t explain why Paul did so poorly in Nevada (10.1% adjusted), New Jersey (11.8% adjusted), or Georgia (12.2% adjusted), all of which have suffered more than the national average from the bursting of the housing bubble.
PILEUS 
A blogging machine today!
“Instead, he reached a whole new constituency. Paul more than doubled his vote share between 2008 and 2012. That mostly did not come from mobilizing more libertarians. He got the libertarians to vote for him again, but most of his new votes came from people we would not normally think of as libertarians.”
I suspect the voters may not think of themselves as rigid libertarians, nor would we think of them as libertarian, but I wonder if they probably have libertarian-ish views, or just voters won over by a libertarian candidate.
There’s a lot of polling and analysis from the Cato Institute recently (example: http://www.cato.org/publications/policy-analysis/libertarian-roots-tea-party) about the libertarian roots of the Tea Party and the strength of libertarianism within the Tea Party and the GOP. But your analysis would suggest that Ron Paul was mobilizing not just libertarians from 2008, but social and foreign policy moderates, perhaps even independents or Democratic-leaning voters too.
My own back of the envelope look at the numbers from Cato and the like is that their Tea Party libertarian voting block is only about half of all libertarians they identify in surveys, the other half does not identify as part of the Tea Party nor do they identify as Republican. A horde of independent or even Democratic leaning independents who refuse to affiliate with the Tea Party or GOP? Sounds like the sort of social or foreign policy moderate who might still be willing to vote for Ron Paul.
Sort of commissioning my own national survey, I’m not sure what other methods are around for trying to look at this hypothetical group of Independent/Democratic supporters of Ron Paul who may or may not be non-TP/GOP libertarians. Maybe using precinct data analysis from this fall for the upcoming gay marriage and marijuana referendums in states like Maryland, Maine, Minnesota, Colorado, and Washington.