As though New Hampshire wasn't already overprivileged enough in the broken primary system, the state may be the one to tip the scales in the general election to either party. According to The Signal's elections model, which orders the states from most to least likely to go to the Republican candidate, a GOP win in New Hampshire gives the challenger 270 votes to Obama's 268. If the president wins, he carries the election with 272 votes to his opponent's 266.
Our model, which I developed with Yahoo Labs economist Patrick Hummel by analyzing data from the past 10 elections, gives Obama a 59.4 percent likelihood of winning in the Granite State. This number is slightly higher than our prediction in our first post about our equations last week because the Real Clear Politics average of presidential approval polls has increased from 48 to 49 percent. The most likely outcome is still that Obama will win by 303 votes, carrying Ohio and Virginia as well as New Hampshire. As we noted before, however, elections are just as subject to chance as football games, and if the contest were held 100 times, we'd expect the Republican to win about forty times.
What if Obama wins Ohio but loses New Hampshire? The math is easy enough to tally, but in fact this is something our model does not allow. That is because prominent research on presidential models demonstrates that the most efficient way to predict state outcomes is to rank them in the order that they fall from one candidate to the other, rather than consider them as 50 independent contests. Since Virginia is more likely to vote Republican than either Ohio or New Hampshire, for example, if it votes Democratic then we assume the other two did as well.
Of course, in reality the states do not line up like dominos. Instead, they are independent elections in which we can draw correlations from regional or ideological ties that lead some states to move in tandem. While it's very difficult to imagine scenario where the Republican wins Delaware, a reliably Democratic state, and loses Oklahoma, a staunch conservative bastion, we can easily imagine the dice falling in a way that gives Virginia to the Democrats while Ohio and New Hampshire go Republican. More work needs to be done to identify all these relationship with any precision, as the noted paper makes clear. We'll be launching a predictions game at The Signal later this year that we hope will help produce this data.
We should note that, while this model does not use prediction market data--that seems like cheating--its prediction of a 59.4 percent likelihood of an Obama victory is nearly exactly where the spread currently sits. And of course, there is still a lot of campaign left. If we had had this model ready to publish just a few weeks ago, it would have pointed toward a more likely Republican victory, as Obama's job approval ratings were significantly lower. While the model currently predicts a second term for the president, his position is precarious. Drop his approval rating three percentage points, to 46 percent, and New Hampshire flips columns and the Republican wins 52.9 percent of the time.
David Rothschild is an economist at Yahoo! Research. He has a Ph.D. in applied economics from the Wharton School of Business at the University of Pennsylvania. His dissertation is in creating aggregated forecasts from individual-level information. Follow him on Twitter @DavMicRot and email him at firstname.lastname@example.org.