After writing that there's not much evidence that the Bain ads are reshaping the race, some were keen to remind me of the NBC/WSJ poll, which set the Bain narrative by releasing a battleground subsample showing Obama with a larger advantage in the battlegrounds than the rest of the country. I did not mention the NBC/WSJ poll in the prior post, mainly because there were plenty of reasons to doubt that Obama had a structural advantage in the battlegrounds. But as if those analytical questions weren't enough, CNN and Pew Research have released different findings: CNN found that the battlegrounds actually leaned more GOP than the rest of the country, while Pew Research shows the nation and battleground behaving exactly the same way. Equally important, Pew hasn't found any shift in the battleground states.

Despite these new numbers, you’ll see plenty of analysts continuing to cite the NBC/WSJ poll without consideration of contrary data points. This type of cherry-picking is a hallmark of horserace coverage: once a narrative (gender gap, swing state advantage, whatever) emerges, polls confirming the narrative receive outsized attention and other data points are ignored. To avoid confirmation bias, it's better to rely on averages or other means of assessing all available data.

At this stage, there’s just not much evidence indicating that negative advertisements are doing much to alter public opinion in the battlegrounds. There aren’t any polls providing a clear baseline for comparison (say, before v. after) and state polling hasn’t moved too much. That’s not necessarily surprising, given the small footprint of the Bain advertisements and entrenched views of the President. Romney wouldn’t be expected to lose many voters either, since he hasn’t yet persuaded very many swing voters himself—he seems to hold the same number of voters that McCain won in 2008—and nearly all of them disapprove of the President’s performance. Perhaps Romney’s favorability numbers would be first to move, but they might not even budge too far, since plenty of voters already have negative views. Even if the attacks did increase Romney’s negatives that will be tough to judge, given the infrequency of swing state polls from similar organizations releasing favorability numbers in a given state.

Without better evidence, analysts should retreat to a more defensible but perhaps frustrating position: the Bain attacks could be working, but we probably won’t be able to tell. Undecided voters don’t need to make up their mind in July. Instead, the Bain attacks only need to reinforce existing unfavorable perceptions of Romney among undecided voters or raise apprehensions that might ultimately be exploited by future stories or attacks. This might not show up in the head to heads in July, but it could in November if Obama reclaims a portion of the white working class vote or they don’t turn out for Romney.