The polls are a bit of a mess right now, but the sources of disagreement seem a little clearer today. A big polling duel might be shaping up for November: Gallup and Rasmussen v. World.

Before delving into details, let's not forget that the big picture is quite clear. If we simply ignored trendlines or the characteristics of the firms, an average of polls would show Obama clearly ahead nationally, probably by about four points, and clearly ahead in the big East Coast battlegrounds by a similar margin. Would there be a relatively big spread in the polls? Yes, although the polls do cluster around the median, even though there are a few outliers. But the devil is always in the details, isn't it? The problem is that we're not just interested in who would win the election if it were held today, we're also interested in the trend, since we're at a potential inflection point coming out of the convention when changes could be especially significant. Once you start looking at the details to catch a glimmer of the trend, things get unclear.  

The key issue is that Obama has faded in the Gallup and Rasmussen tracking polls, while most of the other polls conducted over the same period have a much older baseline (August). This wouldn't be an issue if the new polls with an old baseline showed a tight race like Gallup and Rasmussen, but instead they show Obama performing at a comparable level to the other post-convention polls. So the question is whether we should interpret the balance of recent polling as evidence of a tightening race or a continuing Obama bounce.

On paper, the trackers are well-suited to judging movement. But Gallup and Rasmussen essentially alone in clearly showing evidence of a close race. And it's worth recalling that there is a broader set of pseudo-tracking polls that don't show similarly clear evidence of movement, like Reuters/Ipsos, PPP/DailyKos/SEIU, and YouGov/Economist. The RAND American Life Panel also shows Obama holding his post-convention bounce, and while it isn't a traditional poll, it has (what I consider to be) a promising methodology that should be taken very seriously, especially for judging movement in the race. Even if you ignored the non-Gallup/Rasmussen trackers, there is a defensible compromise position: Obama still has a four-point lead, but Obama’s bounce was probably larger than suggested by an initial wave of swing state polling adulterated by automated firms.

Tomorrow, a wave of polls could show a close race and start adding allies to Gallup and Rasmussen's side. But at this point, it’s hard to escape the sense of Gallup and Rasmussen (and Gravis Marketing, if you care) versus World. In battleground state polls conducted by Rasmussen and Gravis since the DNC, Romney actually leads in five of the nine surveys. But Obama leads in every other non-partisan survey, including some eye-opening leads in big battleground states like Virginia and Ohio. Nationally, Gallup is the only survey of registered voters showing Romney within three points of Obama and Rasmussen is the only survey to show Romney with a lead of any kind. This isn’t to say that Gallup and Rasmussen are wrong, just to say that it’s a little harder to argue that their movement is representative of other polls when their top-line results aren't representative.

With Rasmussen, there are plenty of plausible explanations. Between the absence of cell phone polling, internet panel constituting just 15 percent of their sample, and party-weighting, it's not hard to envision how they could systemically underestimate Obama's standing. In a certain respect, it's surprising that Obama ever took a lead in Rasmussen's tracker after the DNC, since it required Obama to hold an unrealistically large advantage among independent voters. 

The real hold-up, then, is Gallup. It's important to emphasize that, at least as I understand them, Mark Blumenthal's methodological indicts of Gallup shouldn't prevent them from tracking along with the other polls, even if they result in a GOP-lean. Gallup also has a very large sample, which makes it tough to completely ignore their voice in pointing toward a faded bounce. To put it differently: can we really believe that Obama's maintaining his post-convention bounce if he's lost a net-6 points in Gallup over the last week? Given the other polls, there's certainly a case. But it's hard to ignore one of the better indicators of movement at our disposal. A similar situation unfolded in early August and ultimately the other polls moved into line with Gallup, not the other way around.

Regardless of where you come down on the Rasmussen/Gallup v. World fight, there's no question that yesterday's polls were relatively representative of the battle, with Gallup and Rasmussen showing a close contest, and just about everyone else pointing toward a fundamentally different election.

Nationally, Pew Research showed Obama up by 8 points among likely voters—and it is worth remembering that Pew Research also has extremely large samples, an extremely good track record, and polled over roughly the same dates as the most recent Gallup tracker. However, it is also worth remembering that Pew Research generally shows a larger gap between registered and likely voters than indicated by today's poll, and that their likely/registered voter gap can fluctuate between now and Election Day, at least if last cycle's poll are any indication. On the opposite end of the same question, AP-GfK showed Obama up by just 1 among likely voters, but they actually found Obama leading by a 10 point margin of registered voters—even more than Pew. Who’s right? History suggests the right answer is likely to be a little closer to Pew’s result than AP-GfK’s, and the balance of polls show a 2 or 3 point gap between registered and likely voters. But either way, AP-GfK, Pew, and just about every other poll seem to disagree with Gallup's assessment of the race.

The story was the same at the state-level. With the exception of a Rasmussen poll showing Romney ahead in New Hampshire and Obama’s one point lead in Colorado (discussed in the previous post), the state polls were extremely favorable to Obama. The biggest news was probably Virginia, where new CBS/NYT/Quinnipiac, We Ask America, and Fox News surveys showed Obama leading by 3 to 7 points with at least 49 percent of the vote. These returns are highly consistent with previous surveys from NBC/Marist, PPP, and the Washington Post. 

Trailing in Virginia is bad enough for Romney (more on this later today), but when coupled with two surveys by Marquette University and Quinnipiac showing Obama leading by 14 and 6 points in Wisconsin, Obama's electoral position starts to look extremely strong. Realistically, Obama’s not up 14 points in Wisconsin—he’s heading to Milwaukee and Madison is the second most saturated market in the country. But the two polls suggest that Wisconsin is probably more like Lean Obama than toss-up, at least for now, and the combination of Wisconsin and Virginia puts Romney in a tough spot. At that point, just about any other state could put him over the top. And speaking of other states, Fox News also found Obama leading by 7 and 5 points in Ohio and Florida, results that, again, are quite consistent with the other live-interview surveys conducted since the DNC. Those two states are important to Romney's chances, by the way.

Gallup is a credible strike against the view that Obama's bounce is persisting, but they look more like an outlier than a leading indicator of a tightening race. Perhaps tomorrow's NBC/Marist polls will throw a wrench in the Gallup/Rasmussen v. World fight, or maybe the wrench will come next week; it's only a matter of time. But for the moment, the balance of polls points toward a clear lead for the president--perhaps four or five points nationally and in the big eastern battleground states that hold the keys to victory in the Electoral College.