Today the Washington Post/ABC News released a survey showing Obama over majority support among registered voters (51% Obama, 45% Romney). But as Romney's pollster Neil Newhouse (a partner in the firm Public Opinion Strategies) pointed out in a blast email, the poll asked about a few of Romney potential liabilities just prior to the vote question. This goes against polling best practices, and it's possible the survey shows elevated Obama numbers as a result.
In the full questionnaire, respondents were asked a few questions about Romney's wealth, his 14% tax rate, whether he cut or create jobs, and whether his Mormon religion is a reason to support or oppose him. Then came the general election vote question. Respondents were primed to think of some of Romney's potential liabilities, without similar questions about Obama's personal liabilities. However, it's worth noting a series of Obama job approval ratings on the economy, the deficit, taxes, and others also appeared before the vote. To avoid all these potential complications, the horserace should come quite early in a questionnaire.
Now, to be clear, the individual questions about Romney's liabilities were balanced. They were not the message-testing questions typical in internal campaign polling. And I don't think (nor did Newhouse mean to imply) that the poll is deliberately skewed. Lastly, this is not a "push poll," a term referring to negative persuasion calls masquerading as polls (but frequently--and incorrectly--used to describe any negative message testing).
However, I do think Romney is now facing a likability problem nationally, even as he excels in the Republican primary. In this particular poll, over half of voters say the more they learn about Romney, the less they like him (a question that appeared before the series of questions on Romney's liabilities). Familiarity is breeding contempt, as I wrote about here, and pollster.com tracking shows Romney's unfavorables continue to surge while Obama has opened up a larger lead. But I'm not fully convinced this particular poll adds as much to the discussion as some might think. And it's also a good reminder to watch out for methodological quirks--no matter which candidate benefits.
[Disclosure: I've worked with Neil Newhouse on a series of bipartisan research on Walmart Moms.]