Last night, Nate Silver wondered aloud what the predictive value of CNN's post-debate poll might be in determining a potential (likely inevitable) change in the national polls.
Since he could find no study on the subject, he decided today to conduct one himself. The chart below represents his efforts (which I will explain in a moment):
Nate Silver compares the predictive value of CNN's snap reaction polls, post-debate, with changes in head-to-head polling.
What you see above represents the margins by which either the incumbent (I) or the challenger (C) won the debate in CNN's instant reaction poll and the subsequent change in the head-to-head polls.
As you can see above, and as Silver notes, the most comparable margin happened in 1992, when Bill Clinton "won" the debate by 42 points and gained a 4.1 bump in the national polls.
However, what you also see is that there are many instances in which the "winner" of a debate actually lost ground in the polls. Here Silver represents this correlation statistically:
Note that while there is a mild positive correlation, it is "not statistically significant."
Silver projects a 2.2 point bounce for Romney in the polls based on this analysis – or, rather, notes that this correlative study suggests as such.
A win for Romney in the near-term? Yes. A game-changer?
Not by a long shot.