Can presidential debates change the outcome of a race?

In light of our analysis of presidential debates, I thought it would be of interest to the class to see how presidential debates have affected campaign outcomes over the past 50 years. Gallup Poll provides the following graphic which illustrates how the gap of support among voters for the election winner and his opponent changed from before the first debate to after the last debate:

Note: Gallup did not have comparable pre- and post-debate registered voter figures on presidential preferences, so it is hard to interpret the influence of the debate. Gallup says that the 1980 Carter-Reagan debate likely did not have a significant impact on the outcome of the race.

Thus, there have been two debates from between the 1960 and 2004 elections which have shifted voter support from one candidate to another: the 1960 Kennedy-Nixon debates and the 2000 Bush-Gore debates. We have already analyzed the Kennedy-Nixon debates, but I wanted to learn why the 2000 debates resulted in such a huge shift of support from one candidate to another. Gallup's polls after each debate show that Bush's performance steadily garnered support among voters:

What is interesting to note is the discrepancy in support from registered voters from 1-3 days after a debate to 4-7 days after the same debate. While other factors may account for this difference, it appears that media coverage of a debate in the days following it can significantly affect the public's perception of it. Why else would there be a 5% swing from October 4-6 to October 8-10? It would be interesting to examine the results of the debates from the past two Obama elections and see if a similar result occurs, or if social media, such as Twitter, has made the immediate reaction to a debate more permanent.