Election model watch: Is Obama quietly losing?

August 2, 2012

Given the latest GDP data and Obama's approval ratings, the Wonkblog election model now predicts that Barack Obama will win in 81 percent of possible scenarios, a pretty strong vote of confidence. But not all political forecasters agree. Douglas Hibbs, whose "Bread and Peace" model is one of the more notable presidential election models in political science, is now predicting (pdf) that Obama will lose to Romney.

Hibbs's model relies on just two variables: growth in inflation-adjusted disposable income during a president's term, and the number of American combat deaths during the term. The latter variable only proved consequential in three elections: 1952 (at the height of the Korean war), 1968 (at the height of the Vietnam war), and 2004 (as the Iraqi insurgency was ramping up). Currently, combat deaths are few enough that Hibbs's model does not include them as a factor for the 2012 race. But the economy hasn't been Obama's friend, and this has been reflected in the disposable income numbers. Hibbs's model now estimates that Obama will get 47 percent of the two-party vote, while Romney will get 53 percent:


Source: Douglas Hibbs

If Hibbs is right, then Obama is set to lose by a pretty comfortable margin, closer to George H. W. Bush's loss in 1992 or John McCain's in 2008 than Hubert Humprey's in 1968 or Gerald Ford's in 1976.

Hibbs's model has definite shortcomings. Disposable income is a less standard way of tracking the state of the economy than GDP, which is why we included the latter in the Wonkblog model. Further, Nate Silver has argued that Hibbs's model in particular performs well for the years from which it extrapolates -- that is, the elections from 1952 to 1988 -- and very poorly for elections before and after. Here's the difference between Hibbs's predicted results and the actual ones:


Credit: Nate Silver

In particular, the model predicted that Al Smith would win in 1928, Thomas Dewey would win by a landslide in 1948, and that Al Gore would have won comfortably in 2000. In sum, Silver alleges, the model does worse than just positing that each party will get 50 percent of the vote.

So take Hibbs's findings with a grain of salt. And also note that these models get less important the closer we get to the election, because polling will get more accurate. But the model does serve to validate the conventional wisdom that the weak economy is Obama's greatest liability.

Comments
Show Comments
Most Read Business
Next Story
Brad Plumer · August 2, 2012