Sullivan notes "The bounce is over in the Gallup and Rasmussen tracking polls, even though they don't yet register the fallout from Romney's 'fuck-you, 47 percenters' speech. It shows a tightening of the race again."
First, Rasmussen is a joke, and a bad one at that. So forget Rasmussen. Second, the dead bounce is a shift of 1 point, which, within a standard margin of error of 3, statistically means nothing. Third, as Sullivan is right to point out, the quantifiably significant shift has come in partisan enthusiasm--up by 20 points for Democrats and only 9 for Republicans.
Once you bundle those factors, are you justified in concluding that we're witnessing "a tightening of the race again"?
National, head-to-head matchup polls have been perpetually problematic in this presidential campaign, in that they've driven a vastly distorted perception of a close race, which it isn't (Nate Silver's latest odds give Obama a 73% chance of winning, with 303 electoral votes). I rarely cite these wretched things, although yesterday I cited the NBC/WSJ poll's national number because Obama's lead fell well outside the margin of error--and that lead (5%) is now approaching a statistically meaningful watershed: an 8% lead is unbeatable, because its national distribution reflects sufficient leads in critical swing states. More important than Obama's 5% national lead, however, was his 50% job approval rating. If accurate, then it's hard to see how Obama could lose (assuming a conscientious turnout, of course).
And naturally most important is the simple consistency of Obama's likely-voter leads in battleground states. The polling there is all that matters; national polling is fun but nearly always pointless.
So, when we ask if we're seeing a "tightening of the race," what we really mean is: How's are things in Ohio?