This piece by Sharon Begley makes me rethink my desire for Newsweek to go bankrupt. It’s about a study of done and how to tell which ones are likely to be right:
The best predictor, in a backward sort of way, was fame: the more feted by the media, the worse a pundit’s accuracy. And therein lay Tetlock’s first clue. The media’s preferred pundits are forceful, confident and decisive, not tentative and balanced. They are, in short, hedgehogs, not foxes.[….]
If there are three possibilities (say, that China will experience more, less or the same amount of civil unrest), throwing darts at targets representing each one produces a forecast more accurate than most pundits’. Simply extrapolating from recent data on, say, economic output does even better. But booking statistical models on talk shows probably wouldn’t help their ratings.
I went to a panel discussion by some poly sci professors the Friday before the election where they carried on about the Bradley effect and the vagaries of polls and how it wasn’t clear at all what would happen on Tuesday. My friend who was with me got worried and asked “is it really so up in the air?” I said “No, the election will go more or less exactly how Nate Silver is predicting it will go.” And it did.
Full disclosure: I found this article via Howie Kurtz.