Since the first presidential debate last week, Obama has led in just two national surveys: YouGov/Economist survey and Rasmussen. It's not unusual or unlikely for the occasional poll to show a 3-point Obama lead while an average of national surveys points toward Romney in the lead by one or two points. But Rasmussen and YouGov/Economist aren't just typical surveys.
Unlike every other survey, Rasmussen weights for party-ID. Unlike nearly every other major survey, the YouGov/Economist poll is conducted online. (And it has a decent track record and has polled for the entire campaign season, making it a rarity among online surveys). So a dissent by YouGov/Economist and Rasmussen raises the possibility that their differences are due to methodological distinctions. If that's the case, then it's worth considering whether their methods might explain their different results.
One explanation is that the Rasmussen and YouGov/Economist polls are missing a shift in Romney's direction, but might come back toward the average later. Another explanation is that telephone-based surveys weighted by census variables are missing Democratic-leaning voters who aren't feeling like talking about politics after Romney’s debate victory. This is quite common after conventions, which often result in temporary surges in the willingness of partisans or leaners to respond to telephone pollsters. It's less common after debates, but this wasn't exactly a typical debate.
Not only do panel-based internet surveys have higher response rates than telephone polls, but they can note which of their panelists are responding or not. According to Doug Rivers, the president of YouGov/Polimetrix, past Obama supporters were 4 percent less likely to respond to their most recent survey than past Romney supporters. Among respondents that YouGov was able to recontact, there was very little movement following to the debate. If YouGov/Economist is right and partisan response rates are driving Romney's numbers in the national telephone polls, the telephone surveys might not catch up until a new event—perhaps a better performance by Obama in the next debate—causes Democratic-leaners to return to their telephone and voice their support for the president.
Rasmussen works much differently, but its decision to weight by party-ID irons out many changes caused by shifts in partisan response rates—including after big events like debates or conventions. So the relative stability of the Rasmussen tracker suggests that what's moving the other polls might just be party-ID. On the other hand, Obama holds a small lead among independent voters in today's Rasmussen poll.
Weighting by party-ID has major drawbacks, of course (there's a reason no other surveys do it). Sometimes shifts in partisanship aren't just due to changes in response rates, but instead reflect genuine shifts in partisan affiliation. This was the case after the RNC in 2004. If that's what's happening now, then Rasmussen could find itself to the left of the poll average if the post-debate shift was large enough.
And although the RAND American Life Panel isn't exactly a poll, it too shows Obama maintaining a slight lead one week after the debate. It uses a correlate of partisanship to correct for response rates: whether respondents selected Obama or McCain in 2008. Does that explain the difference between RAND and the other surveys? It's possible. Either way, it's another unique poll standing to the left of the more traditional telephone surveys.
Even if YouGov/Economist was right, it wouldn't explain the gap between the battleground state and national surveys, since almost all of the battleground state surveys were conducted by live telephone interview, including many of the best surveys for Obama. And the Reuters/Ipsos survey is also conducted online, but it shows Romney holding onto a lead.
So what does this tell us about Romney's lead in the poll averages? Has there been a real shift toward Romney? Or are Democratic response rates sagging after a demoralizing debate? It might be a little of both and, unfortunately, there isn't a great way to tell. We'll have to see whether YouGov/Economist and the telephone surveys converge, or if YouGov/Economist insists on Obama's resilience between now and November. Nonetheless, it's worth flagging an interesting moment when methodology might correlate with a difference in the polls.