Forecasting the Election

John Sides, Lynn Vavreck, Seth Hill, and Ezra Klein created a model for Klein's Washington Post blog to forecast the election outcome based on just 3 variables - incumbency, economic growth, and presidential approval. Since incumbency won't really "vary" for the 2012 election, there's effectively only 2 variables here. The model is similar to others like the one put together by Alan Abramowitz in advance of the 2008 election and the one put together by Nate Silver last November.

This new model is a bit more simple than Silver's model (the main difference is that it leaves Romney's ideology out of the equation) and it is quite a bit more simple than Abramowitz's model (insert math here).

The other interesting difference is the conclusions to be drawn from the Klein model. To put it most simply, Klein's model yields some pretty happy results from the Obama campaign's perspective. This is because Klein's model seems to count incumbency as more of an advantage than the other two. Just to give a sense of this, if we plug in some conservative (or even pessimistic) assumptions about the state of the economy (say 1.6% growth) and the President's job approval (say 46%), the President is a 78.4% bet to be reelected. Even at 0% growth and a 44% approval rating, the President has a 45.6% shot at re-election in the Klein model. The comparable numbers from Silver's model are 40% and 19%. Obviously, more optimistic scenarios (say, 3% growth and a 51% approval rating) make the President a very solid bet to win in either model (95.3% in Klein's model and 71% in Silver's model).

Are these models useful and, if so, which is likely more accurate? If you're having trouble sleeping at night because you have to know the outcome of the November election right now, these models are not going to help you. Even if they are done perfectly, they are only probabilities. And as someone who drafted Albert Pujols with the second overall pick in my fantasy baseball draft because he was a "very good bet" to be a fantasy monster, I would like to remind everyone that an event that has a 5% likelihood of happening like say, Albert Pujols hitting no homers in April, does happen ... roughly 5% of the time. But these models can tell us something useful. One useful piece of information is that, even though the models don't totally agree on the absolute likelihood of Obama being reelected under various circumstances, they do vary in roughly the same way. In other words, presidential approval and economic growth seem to have the same effect across these models as they move up and down.

As for which of them is most accurate, we really won't even know that after the election because (as these models demonstrate by providing probabilities) there are other variables that can and will affect the outcome. But I will say that Klein's model seems slightly closer to reality to me. One reason Klein's model seems like a better forecast is that Obama has some built in advantages in some of the key swing states like Ohio. I discussed this in more detail a couple of weeks ago and Chris Cillizza wrote about the problems Romney faces in the electoral math just yesterday. In addition, the incumbency effect that seems to be weighted more heavily in Klein's model really does matter. The President is, after all, the President. He has a really cool backdrop (the White House, the presidential seal, Air Force One, etc.) everywhere he goes and he has real presidential kind of accomplishments he can point to. As Mel Brooks would say, it's good to be the king.

Right now, Obama is ahead and Romney needs something to change whether it is the economy, some other intervening variable, or maybe a "game changing" VP pick. And since I have yet to see a single historical example of a VP pick delivering victory to a presidential candidate, I'm gonna rule that third option out.