Yesterday’s polls sent a pretty clear message: the race is close.
The national surveys split between Obama and Romney while the battleground state polls all showed a 1 or 2 point race. Compared to last week’s numbers, the national polls were better for Obama. The president performed near or above his standing in the prior poll of each survey, and ABC/Washington Post and Politico/Battleground provided Obama with a lead, even though neither pollster has a reputation for providing him with particularly strong results. On the other hand, talk of a big Obama comeback is still premature, even if it's eventually confirmed by more polls. Obama continues to trail and hasn't made gains in Gallup, Rasmussen, or the UPI/Cvoter survey.
Perhaps the most interesting state result came from Suffolk, which showed a tied race in New Hampshire, a state that has looked quite good for Obama throughout the summer and September. The post debate polls haven’t looked great for Obama, but the Rasmussen poll actually represented a 3 point improvement compared to a post-DNC poll and the previous WMUR poll showed Obama performing unrealistically well, so the state of the race isn’t especially clear. Given the state’s Democratic-lean for most of the summer, it would be nice to see a little more data to confirm Suffolk’s finding, or perhaps just add a bit of confusion.
One thing on Iowa: the state hasn’t been polled much since the debates and the two post-debate polls show a tight race. But he two polls were conducted by Rasmussen and ARG and both pollsters showed Obama doing better in Iowa than the other battleground states. Given these two poll's modest GOP-lean, it’s quite possible that other surveys will show Obama with a clearer lead.
Although the national and state numbers both pointed toward a pretty clear depiction of a tight race, the unfortunate polling news of the day was a USA Today/Gallup poll showing Romney with a 5-point lead in the battleground states. In a misguided response assured to add fuel to the fire, Obama pollster Joel Benenson released a memo accusing Gallup of a flawed likely voter screen that unduly excluded Democratic-leaning women.
Here’s what I think of battleground subsamples: they’re outright uninformative and contribute to bad media coverage. Dozens of polls survey thousands of voters across the battleground states every week and provide a far more complete picture than a subsample of a few hundred respondents drawn from a national survey.
And what of Benenson’s criticism of Gallup’s likely voter screen? Benenson argues that Gallup’s question about whether voters know where they’ll vote screens out new and transient voters who haven’t yet familiarized themselves with how they’ll vote on Election Day. Intuitively, this argument makes sense, although it’s unclear whether removing this criterion would meaningfully move the numbers. That said, this might help explain why the gap between Gallup’s likely and registered voter model closed by Election Day, perhaps as voters went online and learned about where they were supposed to vote.