I agree in the general case.
In fact, I fully expect that (for example) an IAT would find stronger associations between "food stamps" and black people than between "epic" and black people, but would not find stronger associations between "food stamps" and white people than between "epic" and white people, and if I did not find that result I would have to seriously rethink my belief that "food stamps" is a dog-whistle in the particular case under discussion; it's not unfalsifiable at all.
But I can't figure out any way to falsify the claim that "by finding enough 'code words' you can make any criticism of Obama racist," nor even the implied related claim that it's equally easy to do so for all texts. Especially in the context of this discussion, where the experimental test isn't actually available. All Eugene_Nier has to do is claim that arbitrarily selected words in the article you cite are equally racially charged, and claim -- perhaps even sincerely -- to detect no difference between the connotations of different words.
I wouldn't actually use IAT to find these kind of connections - I would look at the use of phrases in other contexts by other people, and I would look at the reactions to the phrases in those contexts.
To take a historical example from Battle Cry of Freedom: The Civil War Era by James M. McPherson: in the 1862 riots against the draft, one of the banners that rioters carried read, "The Constitution As It Is, The Union As It Was". That this allusion to the Constitution is an allusion to the legality of slavery under said Constitution is supported by...
Here's the new thread for posting quotes, with the usual rules: