Wednesday, February 06, 2008

Missouri for Obama - unless you are AP

AP made a bad call on Missouri last night but otherwise the exit polling held up well. I was pissed in fact that they don't share this like they used to. When I saw the CA exit poll data from CNN, which very oddly leaves off the key total indicating the winner but include the female and male components and population percentages which should give you a good "exit poll winner" after a couple of math hoops and the assumption that males and females are voting in proportion to their state population.....hmm... but is this reasonable assumption?

I think folks make way too much of the challenges of exit polling. In general exit polling is good, but sometimes it is not. In 2000 the exit polling for Florida was fine - people *thought* they'd voted for Gore when in fact many ballots were spoiled due to double vote for Buchanon and Gore. This spoiled butterfly ballot factor clearly lost the race for Gore (crystal clear by the way if you missed the later analysis) even though it was not contested by him in the infamous Supreme Court decision which actually did not make a difference in the outcome.

So -Exit polls are good, but not always correct. Welcome to math.