Where the 2004 candidate bases ended up in the 2008 New Hampshire Primary

This regression produces a number of coefficients that represent how much support was transferred from one candidate to another. For example, if the coefficient between John Kerry and Hillary Clinton is .73, that means that for every percentage point that Kerry had in a district in 2004, Hillary tended to pick up .73 percentage points in 2008. (…)

Clinton
.73 Kerry
.66 Lieberman
.18 Clark
.12 Edwards '04
-.05 Dean

Obama
.93 Dean
.45 Clark
.38 Edwards '04
.15 Lieberman
.09 Kerry

Edwards (2008)
.37 Edwards '04
.37 Clark
.24 Kerry
.15 Lieberman
.00 Dean

This Wesley Clark group is interesting for a number of reasons. For one thing, it seems to be the fastest-growing Democratic demographic: if you look at the cities where turnout increased the most from 2004 to 2008, they tended to be districts where Clark did well. For another thing, the Clark voters do not end up quite where you'd expect them to. You might assume that Clark voters tend toward Clinton because of the strong relationship between the two politicians, or perhaps with Richardson because of their shared foreign policy experience. But in fact it's Obama and Edwards that get most of these votes.

It's not especially easy to define a Clark voter, because they tended to be all along the demographic and ideological spectrum in 2004. But perhaps the common thread is not ideology but one's source of information. These Clark numbers look a lot like a typical Daily Kos straw poll. A broad-based, idiosyncratic, but fast-growing group that tends to prefer Obama and Edwards to Clinton? This might be the netroots in action.