Following on his earlier post on the Iowa caucus, here again is political scientist Charles Stewart.
*****
As with the Iowa caucuses, vote shares for the Republican candidate in New Hampshire distributed themselves geographically in 2012 in ways that are highly consistent with how they were distributed in 2008. The following graphs illustrate that persistence of support.
In the order of the finish, we start with Romney, whose showing in 2012 repeated that of 2008, just shifted up about 5 percentage points in every town:
(In all these graphs, the area of the circles is proportional to the turnout in the town. The diagonal line shows an equal vote share to 2008.) (In all these graphs, the area of the circles is proportional to the turnout in the town. The diagonal line shows an equal vote share to 2008.)
I show two graphs for Paul, because it appears that he picked up support in areas that had given strong support to Huckabee in 2008. But, first, the graph that compares Paul's support in 2012 with that of 2008:
The eagle-eyed will discern that the regression line through the scatterplot has a slope greater than one: 1.52 to be precise. This regression away from the mean is one sign that the distribution of Paul's 2012 geographic support is not just a function of his 2008 support, plus 20 percentage points. This is where Huckabee comes in. When we plot Paul's 2012 support against Huckabee's 2008 support, this is what we get:
A multivariate regression, with Paul's and Huckabee's 2008 support on the right-hand-side pushes the lagged Paul vote coefficient down to 1.01, with a coefficient on the Huckabee vote of 0.72. (The r-squared of this regression is .68.) The individual-level hypothesis this suggests-and that I hope someone will test directly-- is that Paul built his dramatic surge in 2012 by appealing to the Huckabee voters who were drawn to his (Huckabee's) limited-government message. (In a moment, we'll see where the voters who were drawn to Huckabee's social conservatism went.)
It is also interesting that when we examine the residuals of the regression I just mentioned, the towns of Hanover (Dartmouth College) and Durham (UNH) have negative residuals-5 percent in Hanover and 2 percent in Durham. Trying to avoid making an ecological fallacy, these findings make me a little skeptical about the "Ron Paul, Siren of the Youth" meme, and a little more interested in exploring the "Ron Paul, Siren of the Tea Party" story. (In other words, commentators who are puzzled about the apparent lack of an effect of the Tea Party in these first two contests are probably missing the Ron Paul story that's trying to bite them on the noses.)
Where did Huntsman do well? In the land of McCain, shifted down 20 percentage points, as the following graph that shows Huntman's vote plotted against McCain's:
The large-town positive outlier for Huntsman is Hanover, not a typical New Hampshire town, as already suggested. One of the small blips around the Hanover data point is Dixville Notch, which provides some commentary about the quality of that little hamlet as a bellweather. (It should be noted that this data cloud has the highest correlation coefficient of all the scatterplots displayed here, .73.)
Finally, there's Santorum, whose vote shares mirrored Huckabee's just as much as Paul's votes did:
Santorum suffered, relative to Huckabee, in Huckabee's strongest towns. However, these do not appear to be towns where Paul was particularly strong (in either 2008 or 2012). Rather, they are towns in which Gingrich did better. (This is one of the few cases in which I can discern a persistent pattern across the two election years that helps explain the geographic location of Gingrich's support. I guess the distribution of angry old men is pretty random in New Hampshire.)
My purpose in providing these graphs is to help illustrate the strong geographic continuities in electoral strength in the medium-term. The coalitional structure of the Republican party that was on display in 2008 was present in 2012. It shifted a bit, as the identities of the candidates changed (or not), but the persistence across the two years is striking.
These graphs help to make a distinction, lost in most of the instant commentary, between Romney's fast break out of the starting blocks and his success in broadening his electoral base since 2008. From what I can tell in these election returns, his front-running status is primarily a story of the other candidates, not Romney. Thus far at least, Romney has almost nothing to show from five years of presidential campaigning. Romney stands at the top of the heap right now because he has the traditional Wall Street/Country Club wing of the party to himself.
Related to this point, one thing that is clear from mucking around in the election returns from 2008 and 2010 is how the ideological terrain has shifted since 2008, and how Romney's showing this time seems much less impressive, in light of the fact that he is the front-runner. In 2008, he was competing with John McCain, Rudy Giuliani, and Fred Thompson for roughly the same ideological space. Getting rid of those guys, and adding John Huntsman, Romney does just about as well as he did four years ago. At the same time, the candidate field has gotten much more crowded on the right, and the aggregate vote of these candidates is growing. (Recall my analysis last week of the composite candidate I named "Santoperrymann," which was consistent with these comments.)
On to South Carolina!
[A note about data sources: I got the town-level data from the 2008 presidential primary from the New Hampshire Elections Division. Their wonderful web site that always loses when its usability is assessed, but wins in my book because of the data you can find there, if you keep poking around. Wrestling the town returns into usable format takes a little work, but it can be done. It builds character. I got the town-level data for 2012 from the Boston Globe, because official results haven't been posted. This appears to be the same data other news sources are using, only it's in very good tabular form. The analysis above uses returns from the towns that are listed as reporting all its precincts.]
