Tom Holbrook at UW-Milwaukee has an interesting set of data on the impact of Presidential debates since 1988. The typical debate over this period has moved the national polling trend support by an average of 2 points. Tom’s whole piece is worth a read, but I’m going to steal his data table and hope he won’t mind:
Note that Holbrook’s numbers describe the support level for one of the two candidates. Around here, we generally speak in terms of margins … if the Gallup tracker goes from Obama 48–McCain 44 to Obama 50–McCain 42, we’d call that a 4-point swing, whereas Holbrook would describe it as a 2-point swing. Not that there’s anything wrong with that, but it means that you’d need to double these numbers to be consistent with FiveThirtyEight nomenclature.
Still, even when you double these numbers, the swings they produced simply aren’t that large. The most lopsided debates I can think of over this period were the town hall debate in 1992, when Bill Clinton shined, and the first debate in 2004, when George W. Bush stunk. Those debates moved the numbers by 4 and 4.5 points, respectively.
If you moved outside of this period, you might come across some bigger swings — Ronald Reagan had a very large surge in the polls after his debate with Jimmy Carter in 1980, but those were somewhat unique circumstances: the only debate that year involving the two major party candidates (Carter had tried to duck Reagan earlier), and one that came just a handful of days before the election. And of course, Ronald Reagan could run laps around either Barack Obama or John McCain as a debater.
Barack Obama presently has a 2-3 point lead nationally, which means that if McCain wins the debate and gets an average-sized winner’s bounce, the race should draw back to a tie, or perhaps a very small Obama lead. The bar for McCain actually pulling ahead after the first debate is fairly high — Clinton-in-Richmond high.
This is one of those reasons why our model is making Obama nearly a 3:1 favorite in spite of a seemingly small lead in the popular vote. Once you get out of the convention period, voter preferences tend to have become a lot more stubborn, and even terrific or terrible debate performances don’t tend to alter them all that much.