Nate Sivler has a post arguing based on correlation of word frequency with past State of the Unions that Obama re-branded himself in the State of the Union to be more Clintonian. Only including rhetorical aspects, while excluding policy aspects that vary greatly based on the particular problems that a president faces, Nate finds that the correlation between Obama's 2010 speech and past Clinton speeches is even greater than when he includes policy aspects. He also says that "What's just as striking, however, is how dissimilar Obama's State of the Union was to any predecessor apart from Clinton." Nate is correct about the lack of correlation with other speeches being an interesting finding. However, his broader point about Obama re-branding himself to be more like Clinton is suspect. If we are looking to this data for evidence that Obama re-branded himself we would expect that Obama's 2010 speech would have a higher correlation with Clinton's past speeches that Obama's 2009 speech has with Clinton's past speeches. This is simply not the case. Obama's 2009 speech was more Clintonian than his "re-branded" 2010 speech by this measure.
Obama's 2009 Speech has a higher correlation with Clinton 1994 (.67) and Clinton 1998 (.6) than his 2010 speech has with Clinton 1994 (.65) and Clinton 1998 (.54). This is based on the correlations that Nate himself reported for rhetoric, but not policy words.
While I agree with Nate that it is interesting that Obama's 2010 speech is dissimilar to other past presidents except Clinton (in a way that his 2009 speech was not I would add), this is insufficient to overcome the blatant contradiction between Nate's argument about Obama re-branding himself and the directionality of the change between Obama 2009 and Obama 2010. The correlation between Clinton and Obama dropped between 2009 and 2010, not the other way around. The idea that Obama re-branded himself does not stem from the data.