With the conventions behind us, the presidential election has entered the last lap of the race. For the next two months, both candidates will be going full bore. While some surprises are undoubtedly in store, we can start to serious ask the question: who will win?
There are several ways to answer this question. There are plenty of knowledgable pundits who are willing to answer question. I am not one of them. I could not even be bothered to watch the conventions, although I wish I had to just to have seen Clint Eastwood’s speech. Nor can I say I’ve read a lot of predictions from other people about the election. In other words, I’m lazy.
Partly because I’m lazy, but mostly because I’m easily convinced by numbers, I really like Nate Silver’s FiveThirtyEight blog at the New York Times. For those who have not heard of Mr. Silver, he is a statistician whose forecast predicted the 2008 election to an incredible degree of accuracy. I like Mr. Silver’s approach because it dispenses with the partisan rhetoric. Sure, numbers can be used for partisan purposes, but unlike pundits, Mr. Silver’s business is predicting, not pontificating. More importantly, I feel like get most of out my ten free New York Times articles when I read, say, a detailed account of the electoral geography of Arkansas.
So What is Mr. Silver saying? Well, right now, he gives President Obama an 75% chance of re-election. Those are what I call betting odds (the real betters are giving Obama about a 65% chance of victory). Obama is still riding the post convention bounce, but throughout the summer, Mr. Silver has given Obama above a 60% chance of keeping the White House. That’s not to say that he does not predict a close race – with Obama winning around 51% of the vote. With election season in high gear, more polling data should give Mr. Silver more material to refine his forecast.
Of course, not everyone agrees. Although Mr. Silver draws many variables into his model, its core is polling data. Basically, Mr. Silver is trying to predict how people will act by asking how they will act. In the language of economics, he treats voter’s decisions as exogenous to his model. The potential problem is that people may change they minds before election day. Because it relies on polling data, the FiveThirtyEight model cannot predict how people are liable to change their minds.
An alternate way to forecast the election that addresses this potential issue is to eschew polling data and instead build a model based on factors that influence how people vote. In such models focusing on “the fundamentals,” voters’ preferences become endogenous to the model – they are determined by the other factors the modeler includes. Several political scientists have done this, basing their models on economic factors. Their reports give the president a less favorable outlook. They suggest the race will be extremely tight – one model giving President Obama just over 50% of the vote and the other only 47%.
So who’s to believe? I’m inclined towards the fundamental models because they ignore somewhat fickle voter opinion. There is nothing to say voter preferences will not change before election day. At the same time, the fundamental models are based, not on a theoretical model of voter choice, but historical correlation between candidates’ vote shares and outside conditions. Thus these models miss some of the intricacies of individual races that the FiveThirtyEight model may capture. The difference between these models is the difference between a weather report from a climate scientist and a meteorologist – one looks a long term data and the other looks which way the wind is blowing.
My inclination is that the election will be tighter than the FiveThirtyEight model is currently predicting. Although Mitt Romney has not been a spectacular campaigner, the winds are blowing against the president. It might just take one more bad economic report or campaign mishap for the race to tighten substantially. That’s not to count President Obama out, but I’d give him 60/40 odds of winning. Fortunately, we don’t have long to wait and either way, election handicappers will have another datapoint when they gear up from campaign 2016.