Home > News > 0-for-1
86 views 2 min 0 Comment


- February 8, 2008

Before the Iowa caucuses, I predicted that Mitt Romney would be the Republican nominee. Well, reality has conspired to render my prediction wrong. What happened?

My prediction was based on nothing sophisticated. As I discussed earlier, some research by Martin Cohen, David Karol, Hans Noel, and John Zaller suggests that the number of delegates a candidate eventually wins is strongly associated with the number of pre-primary endorsements the candidate received from political leaders of various kinds. So, rather than come up with a squishy ad hoc formula for predicting the nominee, I simply assumed that the endorsements leader would win. Call this a kind of “Moneyball” approach to predicting the nominations.

At the time, I deemed Romney the endorsements leader, based on some data Hans presented at GW earlier in the fall. Below is this data, presented as a graph of the number of endorsements each Republican candidate had received as of August or so. The gray bar captures all endorsements for a candidate. The black bar “weights” the endorsements so that high-profile endorsements (from governors or members of the House or Senate) count more.


I looked at this and saw that Romney and McCain were essentially tied in terms of weighted endorsements. And then — crucial mistake forthcoming — I let some non-Moneyball factors cloud my judgment. “McCain’s campaign is kaput! He’s got no money! He’s flying coach to New Hampshire and carrying his own bag! Etc. Etc.” (I did this even though Cohen et al. show that endorsements matter more than media attention, money, and pre-primary polls in predicting the number of delegates.) So I discounted McCain’s endorsements accordingly and assumed Romney would emerge victorious.

Ultimately, the Moneyball approach actually suggests that McCain was potentially a more serious contender than most observers believed at the time, myself included.