My apologies if this is a duplicate post-- my first attempt apparently did not go
through.
William Dickens wrote:
> It isn't just that Palm beach is large. Palm Beach had an enormous number of
>Buchanan votes even for its size and particularly when you take into account that
>places that voted more heavily democratic also tended to be less likely to vote for
>Buchanan. See the attached.
The attached is a deeply misleading analysis.
If the population of interest is "legal ballots, legitimately cast in Florida on
election day"
then the concept of "outlier" is meaningless -- the sample we have equals this
population
(except for the overseas ballots).
If the population of interest is "ballots that Floridians on election day meant to
cast", then
the analysis is unconvincing since the predictions for Buchanan votes are constructed
using an
extremely unrealistic model of political voting (that is, the prediction is
conditioned only upon how many of my neighbors voted for Bush and little else).
The point is that it is the former population formulation that matters in deciding the
outcomes of elections. The latter formulation requires a type of cosmic knowledge (as
Tom
Sowell might say) that we do not possess, and should not as social scientists pretend
to
possess. God knows that most pre-election polls were far off, and those were based
upon more
information than is presented here.
--
Jay Bhattacharya
Economist
RAND
1700 Main St.
Santa Monica, CA 90405
phone: (310) 393-0411 x6396
fax: (310) 451-7025
e-mail: [EMAIL PROTECTED]
