IP: Re: Counting away (after a while)

[Dave Farber <farber () cis upenn edu>]

> Date: Wed, 13 Dec 2000 12:12:30 -0800
> From: Brad Templeton <brad () templetons com>
> To: Dave Farber <farber () cis upenn edu>
>
>
>
> Those counts will be interesting for historians, but I wonder what
> good will come of them?  If they show for Gore, or if some counts
> show for Gore and some for Bush, the market-destroying fight will
> resume, with calls for Bush to resign (that he won't heed, or will at
> least fight as hard as Gore fought this battle) and a real crisis.
>
> On the net, where I feel I should be surrounded by people with
> scientific training, why do I feel alone in saying that this election
> was a tie?
>
> My mathematical training says that when two results are equal to
> within the accuracy of the measuring device, they are to be viewed
> as equal.  It's a tie.  You don't strain the measuring system past
> its limits to look for the "real winner."  There is no real winner.
>
> In fact, I find it amusing to think that the discipline of working
> with approximate results is also sometimes known as "fuzzy math."
>
> The system has no way to deal with a tie, which is a shame.  So I am
> not surprised at the non-scientists attempting to eke a "true" winner
> from the results.  But the rest of us should know better.
>
> We should also know that the error is not simply one in the mechanical
> measuring devices.  We're uncovering a new type of error which we might
> call "political error."   That's the error bounds which arise from arguing
> over definitions, and doing it in courts.   In any given election, it's
> clear that the final totals can be manipulated, within certain bounds,
> through the application of legal and definitional arguments.  All elections,
> looked at this closely, are full of irregularities.  Some are physical
> like how hard a hole is punched or whether a checkbox got outside the
> bounds.   Some are procedural like whether ballots conform to the law.
> Some are ethical like whether the wrong people had access to ballot
> applications.
>
> We didn't even see all the issues in this one, partly
> because you'll never see all of them, and in part because some of them
> were not politically appropriate to push.  (ie. Gore couldn't support the
> Seminole suit or others like it, and Bush couldn't push other things
> once they took their 'accept the first count' stance.)
>
> The political and procedural margin of error is very large in some countries,
> and it is smaller in the USA but not non-zero.
>
> So our fuzzy math tells us the result in Florida (and indeed the USA) was
> a tie, the difference being within our margin of error.  Bush beat Gore
> in the same sense that sqrt(2) * sqrt(2) is less than 2 on my calculator.
>
> Flipping a coin would have been as much of a result as the processed we
> just witnessed.



For archives see: http://www.interesting-people.org/