Conventional wisdom suggests that if unemployment remains high, President Obama's reelection chances will suffer considerably. As Binyamin Appelbaum wrote in the New York Times, “No American president since Franklin Delano Roosevelt has won a second term in office when the unemployment rate on Election Day topped 7.2 percent.”

But political scientists say unemployment, per se, isn't the metric to look at. Disposable income is. "If there were 10 percent unemployment and everyone were fairly compensated for lost income, the consequence of that would be very different than if no one was compensated," says Doug Hibbs, whose "bread and peace" model predicts election outcomes based on real, per capita growth in disposable income, as well as military casualties.

In practice, of course, unemployment and disposable income growth tend to be highly correlated, and the core insight is the same: if the economy's weak, Obama's weak too. Yale economist Ray Fair's model, which uses three variables — real per capita GDP growth during the first three quarters of the election year, inflation during the last term, and the number of quarters in the last term where real per capita GDP growth exceeded 3.2 percent — predicts that if last quarter's meager 1.3 percent GDP growth rate keeps up through next year, with inflation unchanged and no good growth quarters, Obama will get around 48 percent of the vote. That's still close, and the margin is within the model's standard error of 2.5 percent. But it's not good. And if we have a full-on double-dip, with -0.5 percent average growth next year, Obama only gets 46.7 percent — outside the standard error, and only a point more than McCain got last time around. In other words, it'd be a blowout.

That said, there are a few reasons for the Obama administration to be hopeful.

For one thing, both Hibbs’s and Fair's models emphasize that economic activity closer to an election matters more. Using the per capita GDP growth rate from only the first three quarters of the election year works better than using the growth rate for the two years before the election, Fair says. "A graph of economic conditions in the first 10 quarters of a presidency versus the vote in the next presidential election will show no statistical relationship worth talking about," Columbia professor Robert Erikson, who's worked extensively on election forecasting, tells me. "It is the degree of economic growth between now and election day." So if the economy starts to turn around next year, Obama will reap the benefits, despite weak growth this year and the last.

Forecasters also say that there's always a possibility that this election will be the one that breaks the models. Erikson says it's possible that, because of the unusually long duration of the downturn, voters will still blame the Bush administration, and Obama will suffer less. Hibbs figures that the unusual nature of the contraction will be reflected in economic data that traditional models already consider, but concedes that, given how few presidential elections we have upon which to build models, there's no way of knowing for sure. The evidence does suggest Obama will bear an outsized share of the blame relative to his actual ability to affect the economy, but again, this election could be different, Hibbs says.

And then there's the nomination process. It could be that in all the elections incorporated into current models, the parties nominated relatively mainstream, electable candidates, which could mean the models will fail if an unusually extreme candidate is nominated. "I'm conditioning on the last 20-some odd elections, and each time the party didn't nominate some extremist," Fair says. "They didn't nominate people like [Rep.] Michele Bachmann ... and if they did do that, then my equation would probably be quite a bit off." The nomination of someone like Bachmann or former Alaska governor Sarah Palin would be "a gift from Tea Party heaven," Hibbs quips. "All bets would be off from the normal thinking."