ELECTIONATE SEPTEMBER 25, 2012
Every day that goes by without a shift in Romney's direction or an event that could plausibly induce such a shift is a lost day for the Romney campaign. This isn’t about who wins the news cycle; whether Obama refers to Middle Eastern violence as a bump in the road doesn’t matter. It’s about events that could reshape one of the most stable races in modern electoral history.
The race is likely tighten, if for no other reason than because Romney’s still short of 47 percent, which, in my view, is probably his floor given the president’s disapproval rating. But the margin is somewhat less important than whether Obama falls beneath his reelect number. If likely voter surveys show Obama around 49 percent with Romney behind by a discernible margin heading into early voting, the president is going to get reelected whether Romney’s at 43, 45, or 47 percent. If Obama does fall clearly beneath 49 percent, then we’re looking for Romney to exceed 47 percent, which to date has looked like his floor based on Obama’s disapproval rating, but also his ceiling in national polls.
Today wasn’t good enough for Romney, as surveys showed Obama holding between 48 and 50 percent of the vote nationally and in the critical battleground states. The Politico/GWU/Battleground survey showed Obama at 50 percent of the vote, while Rasmussen and Gallup didn’t show any movement in Romney’s direction. Romney’s best news came appeared to come from an Angus Reid poll showing Obama up by just 2 points, but it represented a 6-point shift in Obama’s direction since their prior survey and it was the first Angus Reid poll to show Obama with a lead of any kind since the onset of the general election campaign.
If you only considered today’s battleground polls, you might conclude that Obama leads by at least four points with at least 49 percent of the vote in every battleground state—that’s 347 electoral votes. That’s mainly due to Obama’s standing in a Civitas survey that showed Obama leading by 4 points in North Carolina (Obama now leads Romney in an average of post-DNC polls in North Carolina), but ARG polls also showed Obama leading by 5 to 7 points in Iowa, Florida, and Nevada. I don’t give much credit to ARG, but these results are generally better for Obama than their previous national and state-level surveys since the DNC, so it’s still a bit of good news for the president. For good measure, Rasmussen and WeAskAmerica found Obama up by double digits in Wisconsin and Michigan, results that seem a lot more reminiscent of 2008 than 2004.
The outlier of the day was undoubtedly Susquehanna/Tribune-Review, who showed Romney within two points in Pennsylvania. If they hadn’t said anything, I would have just chalked it up to a bad sample and never taken a closer look. But instead, Susquehanna published a bizarre defense of their methodology, which amounted to asserting that the race will be close... because they so. Susquehanna used the bizarre language of 2004 versus 2008 “turnout models,” as if pollsters should be making subjective decisions about the composition of the electorate. We don’t need a poll to tell us whether the race would be close if the electorate looked like 2004 (it would be), we need polls to take representative samples to help us determine whether the electorate will look like 2004.
So what happened? They took a wildly unrepresentative sample and defended it by asserting that they were using a “2004 turnout model.” Their respondents certainly didn't resemble the 2004 electorate: 7 percent of the voters in the Susquehanna poll were ages 18-29, but 17 percent were ages 18-29 in 2004; 36 percent were over age 60 in the Susquehanna poll, compared to 24 percent in the 2004 exit polls; 50 percent were conservative, compared to 34 percent in 2004; 87 percent were white in the Susquahannah poll, compared to 77 percent in 2004. So let’s not kid ourselves: they took a quick poll and missed young and non-white voters and then either didn’t weight their sample or only lightly altered it based on their premonition that this would be a close election, rather than the responses of voters.