The Annenberge National Election Survey has just released the results from their poll. In it, they did a very large national survey (n=8314) with a sub-sample of battleground states (n=3418) and of "persuadable" voters in those battleground states (n=832). Samples this large allow them to analyze that last category--persuadable voters in the battleground states. Such an effort is all to the good, since, as they point out, this is where the campaign is currently focused. Ruy Teixeira and Andrew Sullivan argue that the data show trouble for Bush.
That's true, but only to an extent. First, the diffence in Bush's approval rating between the persuadables and all battleground state voters (44% vs. 47%) is not statistically significant. Second, these persuadables are a notoriously ficke and disengaged bunch. They have lower levels of education (54% with a HS diploma or less vs. 49% of all battleground voters), they are less likely to be registered to vote (68% vs. 79%), less likely to follow the campaign closely (54% vs. 69%), and less likely to have voted in 2000 (59% vs. 70%).
Given this, I'm not sure what sorts of conclusions we can draw about what these persuadables will do in November or what impact it will have on the election.