I think that explains some of the shift.

But there is another reason too: the party ID numbers have shifted  according to Rasmussen.

Whereas in April and May, Democrats outnumbered Republicans by 10.1%, the margin is now 2.5 points closer, or 7.6%.

Since Rasmussen and Gallup (which probably uses the same party ID numbers) weights according to this scale, it explains why the horse-race numbers have tightened.

Now, the big question is why have the party ID numbers changed. I think there are a few explanations.

1) The April and May numbers were inflated with the national obsession over the continuing Democratic primary race. Many Independents thought of themselves as Democrats while observing the race. Since the primary ended, they've gone back to Independent status.

2) Republican numbers have not gained; Independent numbers have. Not only might this be loss of interest, but many Independents who voted for Clinton called themselves Democrats in April and May but no longer do.

3) A small subset may be anger over FISA and defections from the left to Independent status.

If you look at party ID patterns over the last few years, this actually looks pretty normal. In July 2004, during the DNC, Dems outnumbered Republicans by 3.0%. After the RNC, it was 0.6%. On Election Day the margin was 1.6%, with turnout and a tiny defection advantage to Bush being decisive.

In 2006, party ID advantages to the Dems started small and ballooned to 6.1% as the campaign went on and the Dems went on the attack. In 2007, it fell back a bit as the campaign season dissipated.

But then, starting in February 2008 when McCain sowed up the GOP nomination, the Democratic numbers exploded. The 10.1% numbers were unusually high.

So what does this mean?

The DNC will boost the advantage again. It always does. The RNC, coming so close the DNC, probably will not dent it as much this year. The margin will probably stay at 7.6% for a while.

The thing to watch is the percentage of Democrats coming home to Obama. If, as usually happens, the undecided Dems come home to Obama, he will get to the 90% mark that Democrats need to win.

As for comparing this to other polls, the no callback method may explain why Ras/Gallup has a tighter screen than other pollsters. The result may be, ironically, that the trackers show Obama gaining in the fall after the DNC and the stand-alones show the race getting closer.

PREDICTION: Unless some outside event fundamentally transforms the race (and no, the ad wars and minor policy shifts won't change things much), I suspect that the party ID difference will stay at around 7 to 7.5, Obama will pull in more Democrats than he is getting now, Obama's better turnout operation will boost his numbers by a point or two, and he will win by about 3.5 or 4 points.