Mitt Romney (The Washington Post) and Osama bin Laden (Federal Bureau of Investigation)

Public Policy Polling has a new poll (pdf) out of Ohio showing Obama with his biggest lead since May. Given how hard it would be for Mitt Romney to win the White House without winning Ohio, that's a big deal.

But a secondary finding in the poll has gotten a lot of attention. PPP asked voters who they thought deserved more credit for the killing of Osama bin Laden: Obama or Romney. 63 percent said Obama, 31 percent weren't sure, and 6 percent said Romney.

The results for Republican voters were even more astonishing. 38 percent said Obama, 47 percent weren't sure, and 15 percent said Romney. What the heck is going on?

My best guess is one dash psychological trickery, and a healthy portion of sampling error. There's a huge literature in political science on the extent to which surveys like this capture true public opinion, and another literature on the extent to which peoples' ideologies affect their interpretation of objective fact, and both do a lot to explain the Romney-killed-bin Laden finding.

The most directly relevant study here is Vanderbilt professor Larry Bartels's 2002 paper, "Beyond the Running Tally: Partisan Bias in Political Perceptions." (pdf) Among other data, Bartels looks at polling conducted in 1988, in which voters were asked several questions that were politically relevant, but had objective, factual answers: Did unemployment go up or down under Reagan (it went down)? Did inflation get better or worse during his presidency (it got better)? He found that of the 10 questions asked, seven showed "very strong evidence of partisan bias," and that in six of those questions, "Republicans and Democrats could not even agree on the direction of change between 1980 and 1988." Two exemplary cases were inflation and unemployment. Here's how perceptions of those two metrics varied by party:

Source: Larry Bartels

Bartels found the same results for assessments of changes under Bush I and Bill Clinton, and a later paper (pdf) co-authored with Christopher Achen, then at Princeton, found the same results in reactions to Watergate and abortion.

The only issues that didn't show this kind of bias were ones where voters disagreed on the proper policy. For example, Democrats and Republicans' assessments of how defense spending changed under Reagan didn't differ much, but their views on the appropriateness of high defense spending differed drastically.

In other words, voters have trouble crediting politicians they don't like for policy outcomes they do like. And killing bin Laden is a policy outcome they do like. And so partisan effects have led some Republicans to argue that Obama was not primarily responsible for killing bin Laden or, even more absurdly, that Romney was responsible.

What's more, correcting peoples' factual misunderstandings doesn't seem to help at all. Brendan Nyhan of Dartmouth and Jason Reifler of Georgia State ran experiments measuring whether partisans who read news articles with correct information that ran against their ideological views were likelier to hold the right factual beliefs. They found the opposite effect — correcting people, in other words, doesn't inform them, it creates a backlash.

Telling conservatives that there were no WMDs in Iraq made them more likely to say there were weapons, and telling them that the Bush tax cuts reduced revenue made them more likely to say they increased revenue. Same for liberals -- while conservatives and moderates were less likely to think Bush banned all stem-cell research after reading an article pointing out that he only banned federal funding of it, liberals' stated factual beliefs didn't change at all. So ream after ream of news articles wouldn't have done much to help any unfortunate souls who formed the belief that Romney killed bin Laden.

Psychologists call the phenomenon on display here "motivated reasoning," and those of you (which is to say, hopefully all of you) who read Ezra's New Yorker piece on motivated reasoning and the conservative turn against the individual mandate will be familiar with the idea. But the Romney-killed-bin Laden finding also fits in with the broader literature on polling generally.

The most prominent model of public opinion, that of UCLA's John Zaller, argues that people read news media and absorb the things they read that comport with their general worldview, and then, when polled, repeat back whatever information they encountered most recently. More aware voters are more partisan in what information they internalize, but they expose themselves to a lot more information, whereas less informed voters read less but absorb more of what they read, regardless of which party it supports. So informed voters hold consistent but extremely partisan views, and uniformed voters hold very inconsistent views with no clear partisan direction. What's at the top of an uninformed voter's head could be anyone's guess.

The Zaller theory suggests that we should be extremely skeptical about public opinion polling, and particularly about its use outside well-trodden territory like election polls. To apply it to this case, suppose someone polled by PPP had just read a news story about Navy SEALs attacking Obama for taking credit for killing bin Laden right before getting a call from a pollster. The Zaller model suggests that would lead them, if they were uniformed or a Republican partisan, to answer that Romney was more responsible, or that they were unsure, just since the anti-Obama story was at the top of their head. But it doesn't mean that the person actually thought Romney killed bin Laden.

Some have suggested that the PPP poll is indicative of epistemic closure among Republicans, that they have closed ranks around their own worldview so tightly that they won't let inconsistent facts, like Obama killing bin Laden, into it. The motivated reasoning literature suggests that may be part of what's going on. But the Zaller model should remind us that, unlike people who read this blog, most Americans only have a casual interest in politics, and don't have particularly consistent or well-thought-through views on most political topics. You can consider that troubling or not, but it does caution against inferring from the poll that there are thousands of Ohio Republicans holding fast to a strong belief that Romney, somehow, got bin Laden killed.