If you peruse the comments in today's polling thread, you will see various sorts of proclamations about when John McCain's convention bounce will become meaningful (that is, when it can be described an actual shift in electoral preferences, rather than a temporary bounce). Some of these are exceptionally specific: "If Barack Obama isn't at 47 points in the Gallup Daily tracker by 2:51 PM on Wednesday, September 10th, then we're all d00mEd!!!".
Among the many difficulties that we face in this unique election cycle is figuring out exactly how long one should expect a convention bounce to last. Under ordinary circumstances, bounces are actually fairly persistent, lasting for perhaps several weeks:
Of course, we don't know what two bounces will look like when laid down on top of one another (with a one-week gap intervening). Our best guess was something like this:
That is, the Republicans would still be getting some residual benefit from having had the last convention for perhaps as many as two or three weeks from today.
Intuitively, that feels somewhat wrong to me. Most conventions are held over the summer, when the news cycle is much slower, and the convention gets to linger for longer as the last thing on voters' minds. This does, however, raise an important point: political time is relative rather than absolute. If it feels like the Democratic Convention was a month ago -- well, in political time, it might as well have been a month ago, since Sarah Palin and the Republican Convention displaced it as the first thing that voters will recall when they think about the election.
What I am saying, then, is that we should evaluate the robustness of the Republican bounce by how well it holds up to the currents of political time, rather than any specific date on the calendar. Specifically, I would want to see how the bounce holds up to the next major development of the campaign, particularly if it is a pro-Obama development. For example, let's say that Colin Powell endorses Obama tomorrow morning. I might expect a fairly strong reaction to this in the polls, not because the endorsement is all that important unto itself (most endorsements aren't), but because it displaces the GOP Convention as the most recent event of the campaign -- it pushes political time forward. And if the polls didn't move in reaction to such an endorsement, I'd think Democrats would have reason to worry.
On the other hand, if the next couple of weeks are relatively newsless, I would not necessarily expect the bounce to fade all that quickly, and I would not necessarily be worried (as a Democrat) if it didn't.