Back in 2004, a colleague and I created something we called the "pocketbook model" to predict the outcome of presidential elections. It considers just five simple variables: unemployment, income growth, inflation, whether we're at war, and third-party candidates. We ignored poll data, approval ratings, personalities, campaign spending, debate performances, advertising, and even the names and positions of the candidates.
Instead, we started with the assumption that most Americans generally reward the incumbent party with their vote when the economy is good, and punish them when it's not. And it turns out this "pocketbook" model has a pretty good track record. If you look back at all the presidential elections since 1916, the model correctly predicts the winner of the popular vote in 21 of the 24 contests.
More recently, the model predicted the outcome of the last three elections to within a couple percentage points, mostly using information that was available months in advance.
In the three cases the model didn't work, voters seem to have had other things on their minds. In 1952, 1960, and 1976, the economy was doing well, but the incumbent party lost all three elections. The Cold War likely handed Eisenhower the White House in '52, Kennedy's made-for-TV charisma narrowly beat out Nixon in 1960, and disgust with Watergate ushered in Carter in 1976.
So what does that tell us about 2012? Right now, the pocketbook model suggests chronic unemployment and sluggish growth will make President Obama a one-term president. If things improve enough between now and Election Day, he might still have a shot. But at this point, there doesn't seem to be anything on voters' minds big enough to trump the economy.