# The Surprising Power of the New Hampshire Primary

As you probably know, there are lots of models for predicting who will win the presidential race this year. To summarize: most of the models predict a small Obama victory, but a few predict a Romney victory. Most likely, it's gonna be close.

Dylan Matthews rounds up a few of the forecasts here. Most are fairly ordinary, but I was pretty intrigued by the model from Helmut Norpoth and Michael Bednarczuk of SUNY Stony Brook. Unlike every other forecaster on the planet, they don't include any economic variables at all. Maybe the economy is great, maybe it sucks. They don't care. Their model has two inputs: (1) how well you did in the New Hampshire primary, and (2) whether you're an incumbent running for reelection.

This is....eccentric, to say the least. Not the incumbent thing: most forecasters recognize that, historically, incumbents usually win if their party has been in the White House for only four years. But the New Hampshire primary? According to the authors, it tells us a lot. An incumbent who gets serious competition (Carter 1980, Bush 1992) is in trouble. Likewise, a challenger who doesn't clean up against the opposition is also in trouble. And out of all possible primaries, New Hampshire has the most predictive power. This all gets translated into a formula, of course, and here it is. Note that 3rd and 4th terms are Democrat-centric because the current incumbent is a Democrat:

Here's how the arithmetic plays out:

• .445 (65 - 56.7) - .138 (63.2 - 47.7) + .366 (53.7) - .333 (48.8) + 48.2445
• .445 (8.3) - .138 (15.5) + .366 (53.7) - .333 (48.8) + 48.2
• 3.69 - 2.14 + 19.65 - 16.25 + 48.2
• = 53.15% popular vote margin for Obama

As it happens, I don't believe this for a second. Lop off a couple of points because Obama is black and I think you're closer to the final outcome. My best guess is that Obama will win about 51% of the popular vote.

But who knows. The authors claim that their model explains 89% of the variance in party shares of presidential votes, and that's pretty good. On the other hand, it's wise to be cautious. The simpleminded rule that incumbent parties win after four years in office and lose after eight or more years correctly predicts 13 out of 15 elections since 1952 (the outliers are 1980 and 1988). That's an 87% success rate.

Still, I thought this model was pretty fascinating. I have a hard time believing it, but if Obama really ends up winning 53% of the vote in November, my hat's off to the authors.