Bryon Shafer and Amber Wichowsky kindly reply to my earlier post on their article. I had written, apropos of the linkage between the nominations process and the general election:
bq. There is one aspect of the nominating process that may have affected Obama’s vote share in November (although it was not necessarily crucial to his victory): mobilization, or get-out-the-vote efforts. Obama was clearly able to convert his organization from winning delegates to winning voters…This particular aspect of campaign strategy may have been one important link between the two stages of this presidential election.
When I wrote that, I had in mind GOTV only as a campaign strategy. That is, I was thinking that the Obama campaign no doubt learned a lot from developing and implementing a mobilization strategy — particularly in caucus states — and then could build on that work in the general election.
Shafer and Wichowsky do me one better and actually look to see whether an Obama primary victory is associated with higher turnout in that state. This seems to be the case; the opposite seems true for McCain. I have a couple thoughts in reply. To echo my thoughts on their first post, I’m still not sure why we would expect any relationship between primary outcomes and turnout, at least in theory. One could presume that a primary victory energizes supporters, who then are more likely to vote in November. This would be particularly true if the candidate maintained a strong organizational presence in a state through the general election. Ultimately a better test would be to see whether a candidate’s primary victory was associated with higher turnout among that candidate’s supporters.
A second thought is purely methodological. In this post and in their article, Shafer and Wichowsky measure change over time by first ranking states and then measuring whether states moved up or down in ranking. I am quite wary of rankings (see here and here). I’d prefer the change in the level of turnout, 2004 vs. 2008. One could also norm turnout by subtracting the average for each year, to account for any secular trend.) Weighting by Electoral College votes seems unnecessary, and perhaps ill-advised, for testing for this particular relationship.