The tracking polls showed Obama remaining at elevated levels, but the Washington Post threw a wrench into a clear assessment of the race:
 
The Washington Post showed Obama leading by 1 point among likely voters but 6 points among registered voters, a bounce of 3 points among likely voters and 7 points among registered voters. As a result, the gap between registered and likely voters actually widened after the DNC, presumably because Obama only swayed the views of unlikely voters without convincing them to turnout on his behalf. In this regard, the poll is much different than yesterday’s CNN poll, which actually showed Obama making big gains among likely voters but not the broader universe of registered voters.

It’s hard to know what to make of all of this, except to dismiss it as static. Rasmussen showed an Obama bounce and continues to show Obama’s approval rating at 52 percent, despite applying a likely voter model. Gallup shows an Obama bounce and continues to show Obama’s approval rating, despite simply being a poll of registered voters. So most data points point toward a bounce among both registered and likely voters, with CNN and the Washington Post splitting on whether Obama's bounce was disproportionately among likely or registered voters. Indeed, our relatively small sample of post-convention polls shows Obama leading by an average of 3.3 points among likely voters and 5.5 points among registered voters, which is about the same gap as prior to the convention. Given just how few polls there are in either column, there won’t be a clearer picture of the likely voter/registered voter gap for quite some time. If you want more on the Washington Post poll, you should read this article by Mark Blumenthal. It's possible that different metrics for what constitutes a "likely voter" contributed to the difference between CNN and the Washington Post, but the opacity of the two likely voter screens makes it difficult to be sure about exactly how.

Despite the ABC News/Washington Post poll, Obama leads by an average of 3.9 points in polls conducted since the DNC, 48.7 to 44.8 for Romney. Part of the downward pressure was due to a downtick in Obama's support in Rasmussen and Reuters/Ipsos, although Obama's approval rating held at 52 percent in Rasmussen and Obama gained an additional point in Gallup's tracking poll. A new PPP/DailyKos/SEIU national survey showed Obama leading by 6 points among likely voters, which partially counter-acted Obama's weak showing in the ABC News/Washington Post poll.

SurveyUSA shows Obama leading in Florida by 4 points. That’s actually down from their last survey, which showed Obama up by 5 in the Sunshine State. But a 4 point Obama lead is quite consistent with SurveyUSA’s other post-DNC polls, which show Obama up by 10 in Minnesota and 16 in Washington. What do all of these numbers have in common? They’re all very near Obama’s 2008 performance.

While SurveyUSA’s initial returns paint a consistent picture of the post-convention landscape, Gravis Marketing is deeply confused. They manage to show Obama up 4 in Ohio but down 5 in Virginia. I haven’t talked very much about this pollster, mainly because it just doesn’t appear especially credible. Their website is amateurish and they’re not primarily a polling firm, but instead a robo-caller that does some polling on the side, perhaps once they realized that their unused robocalling devices could be used to conduct cheap polls. Given that they don’t release cross-tabs and provide results down to a tenth of a percent, it’s hard to take them seriously. Even so, RealClearPolitics has elected to include them in their averages.

PPP’s poll of Arizona is a little unusual. While their prior two polls showed whites at 71 and 75 percent of registered Arizona voters, today’s poll of likely voters showed whites at 83 percent of the electorate. I don’t think that the switch to likely voters accounts for a shift of that size, and whites were 75 percent of the Arizona electorate in 2008.

Speaking of PPP, poll watchers should observe that their likely voter model is unusual. They open the interview by saying: “This is a short survey about this fall's election. If you don't plan to vote in the election this fall, please hang up now.” PPP then comes up with “target ranges” for the composition of the electorate. In contrast, most polls survey the universe of adults or registered voters and weight their sample to census figures. That allows them to construct a representative base, and from there they exclude the “unlikely voters” without making their own assumptions.  By sending non-voters packing, PPP loses the ability to weight their poll to the broader universe of all adults or registered voters and forces them to make a subjective call about the composition of the electorate. There’s some chance that the value of their “hang up” method as a likely voter screen outweighs the cost of subjectively weighting the sample, but it’s highly unusual and risks missing shifts in the electorate

Odds and Ends                                                     
--Just a few days after the Romney campaign started spending in Wisconsin, the Obama campaign decided to air advertisements in the Badger State. The size of either side’s buy is unclear, but it appears that Wisconsin enters the fall as a true toss-up state.

--A GOP source with access to private GOP polling told ABC News that Romney is down by 5 points in Ohio and 4 points in Virginia. Obama leads by 2 points in Wisconsin, while Colorado, Iowa, and Nevada are all tied up. Who knows if it’s true.