Emory political science professor Alan Abramowitz seems to have a mathematical election model that works. Abramowitz's system has correctly predicted the popular vote winner within two percentage points for every presidential election since 1988. This year, it's predicting an Obama win: 54.3 percent, versus McCain's 45.7 percent.
The model isn't perfect, of course, but it does factor in a wide range of variables such as GDP, a party's time in office, and recent polls. "While factors outside of the model, such as rising partisan polarization and resistance to an African American candidate by some white voters may result in a somewhat smaller popular vote margin for the Democratic nominee," Abramowitz writes, "the combination of an unpopular Republican incumbent in the White House, a weak economy, and a second-term election make a Democratic victory in November all but certain."
If you're skeptical of models, check out the Iowa Electronic Market's trading index for the presidential election. For decades, it's been a much better predictor of presidential wins than Gallup polls. As of today, the market's predicting a 54 percent win for the Democrats, versus 45 percent for the Republicans. it could be a coincidence that those numbers are so close to Abramowitz's, or it could be that investors are reacting to his model's predictions. A third option: it could be that Obama actually is going to win by ten points.