On Fri, 26 Nov 2004, Doug Henwood posted:
> Philadelphia Inquirer - November 24, 2004
>
> Final Tallies Minus Exit Polls = A Statistical Mystery!
> by John Allen Paulos
What's cool about this post is that Paulos is the very model of amiable respectability on such matters -- the bestselling author of books on mathematics (how rare is that?) So it's news when he says that Freeman's study is so obviously respectable. Maybe someone like him can get on TV to say it.
Also interesting is Paulos's impliciation that the questions Freeman raises can't be answered (contra the mystery pollster) until Mitovsky et al. release their full data -- and that they could if they did.
I had previously thought the full data couldn't adjudicate anything since the exit polls were continually marked to market, so that the earlier, provisional results were discarded. Paulos implies that all the provisional data and the records of their changing might still exist. Teixeira's analysis implies the same thing: that the exit polls aren't simply replaced with the returns, but reweighted until they conform to the headline win/loss results of the returns -- which can only be done through a mass of calculations that any respectable pollster would have to keep. And which, if released, would provide a record of all the divergences.
Michael