Of all the things that have intrigued me about this election in the last couple of months, I have to say the one that leaves me most curious is the meta-battle about prediction going on between traditional journalists and statistical analysts. The flash point is probably the set of constantly-updated predictions by Nate Silver of the New York Times.
The stats guru side of the debate is buttressed in recent years by the likes of InTrade, a prediction (or betting) market. The basic idea of both, from the point of view of the lay observer, is that the stats types and the markets can help you predict a likelihood of winning, or can give you odds. It can't tell you who will definitely win (at least not in any reasonably close election), but it can allow you to put your money where your mouth is,.as it were.
But the response from traditional journalists has been entertaining (at least from the perspective of someone who used to teach stats and loves the book Moneyball). Check out replies from Kathleen Parker, Dana Milbank, and Joe Scarborough. From my point of view, they fundamentally miss the point, and I would bet that they never had to take INTR 202... or have forgotten it all from years of accumulated punditry.
The whole saga crystallized recently when Silver offered Scarborough (who said the election is a "toss-up") a $1,000 charity bet on the outcome. There's plenty to say about this, but one of the most compelling sets of post-mortems about this election, I think, will be the debate over how and whether statistical analysis based on publicly available data is increasingly likely to crowd out "gut instinct" or "horse race" journalism when it comes election time.
Something that I will be watching for. I confess to casting in on the side of the geeks over the impressionists on this one.
No comments:
Post a Comment