[From Bill Powers (2000.12.19.0445 MST)]
Bruce Abbott (2000.12.18.1105 EST)--
The "subjective bias" in the counting machines reflects the voting standard
their use assumes: it is a vote if, and only if, there is a hole there.
That is the standard they were designed to meet, and they do so with great
accuracy. It is also the standard voters were warned in the polling places
had to be met if their votes were to count.
You're idealizing the way these hole-punch machines actually work. When you
push with the stylus, there is resistance until you push hard enough, and
then (if you do push hard enough) the resistance suddenly gives way and the
stylus goes all the way in and bottoms out. However, this does not
guarantee that "there is a hole there," because nothing guarantees that the
chad drops completely away. If the chad is now a flap held by one or two
corners, the hole can look open on visual inspection from the top, yet
close up again when the card slides into the counting machine. Also, when
you pull the card out to look at the holes (assuming you really do that --
but who really does?) all you see is a pattern of punched and unpunched
rectangles, with no indication of which _should_ be punched. All machines
of this kind that I have used have a fixed booklet under which the card
slides, so all you see is the printed booklet on one side, and a column of
guide holes on the other side, at the right. The card itself is not
visible. I don't think a voter inspecting the card could possibly tell what
presence or absence of a hole in any given position meant.
No, I'm saying that if a dimpled chad counts as a vote, then to be
consistent, a ballot with a dimpled chad for one candidate and a hole for
another candidate competing for the same office would have to count as a
double-vote and rejected on that ground.
I would expect that to happen, since observers from both parties were
present (in Illinois, the law required it) and could challenge any judgment
of a valid or invalid vote. This was not a case of Democrats deciding which
ballots were valid or invalid and just chucking out the votes for the other
guy. There are a lot of hawk eyes on both sides watching every ballot
during a manual count.
The system is not based on trust.
Incidentally, these punch-card systems require that the card be slid under
the ballot booklet and the top edge, when it appears, be lifted enough to
slide over two pins that then engage registration holes in the card. If the
voter stops feeding the card in when it first stops, the top edge of the
card could be butting against the registration pins. This would displace
_all_ the holes relative to their proper positions, and all votes would be
other than what was intended. When the card was pulled out, there would be
no indication that this had happened. There is no printing on the cards to
indicate the meaning of any given punch hole.
Do you know that by far the most typical experience with recounts is that
they make little difference -- only a few votes one way or the other,
tending to cancel out?
That's easy to say when there is a single uniform voting system and the
candidates are essentially tied, but it's clear that the most errors
occurred in places where the cheapest machines were located. If that
happened to coincide, for whatever reason, with places where the most votes
for Gore were cast, then one would NOT expect a recount to favor both
candidates equally. Picking up intended votes from apparent no-votes would
favor the candidate for whom the most votes were being cast.
Anyway, you're overlooking the _extremely small_ difference in the votes
for the two candidates, so close to a tie as to be statistically
indistinguishable. The difference was "only a few votes one way or the
other." Even if the errors were random, a recount could easily have either
increased Bush's lead, or swung the election the other way. Gore, who was
behind, was willing to take the chance that Bush's lead might increase.
Bush, who was ahead, was not willing to take the chance that his lead might
decrease or vanish. The chances of either outcome of a recount, I would
guess, would have been around 50:50.
I really would like to know how you would be talking
now if the situation had been reversed, with Gore ahead and the Bush team
calling for a recount only in highly populated Republican strongholds,
precinct workers finding votes in dimple chads, and uncharacteristically
large (compared to typical recount experience) gains for Bush being
announced from partial recounts in progress withing Republican held
precincts. Think about it. Wouldn't you have been screaming that Bush was
trying to steal the election? (Sound familiar?)
In my case, probably, especially if I were as drunk as some of those
protesters looked. But such paranoid objections overlook the fact that both
parties, as mentioned above, keep a sharp eye on the other side during
manual recounts and do not let a single disputable decision get by without
loud objections. If both sides agreed that there was a clear indication
that the voter _tried_ to poke a hole through the card, then of course the
suspicion of cheating by one side would be groundless. And as far as I
could tell, that is how it worked: both sides had to agree. Also, my
impression was that people were trying to be impartial even if they favored
one side. People do tend to rise to the occasion, and a national election
is a somewhat solemn occasion.
I like your reasoning (surprised?), but would like evidence that there were
problems with the machines. Thus far no one has brought forth a machine
which demonstrates the problems that have been alleged.
I wonder whether you're ever actually used a punch-card ballot. I don't
think your imagination is giving you an accurate picture of how they
actually work. There are problems with _all_ punch-card machines, of the
nature I described above.
And it isn't the
counting machines that were the problem, if there was one, it was the
punch-card ballot device, which supposedly made it difficult in some cases
to punch out a chad. I am unwilling to reason backward from the result to
the supposed defect, and then use the supposed defect to explain the result.
If there was such a problem, it should be easy to demonstrate.
The problem was demonstrated: hanging chads, pregnant chads, and all the
similar stuff that opponents of the recount were loudly making fun of. And,
if anyone had thought to look, displaced holes due to failure of registration.
[This from Rick Marken]>> My guess is that most of
the uncounted votes in those counties were from the cheap punch
card machines in the Democratically inclined precincts of these
majority Republican districts.
You made that one up, Rick. I think that the same machines are used
everywhere within a given county.
Is the Brookings Institute going to recount every county in Florida? In the
counties using the punch-card system, what will be counted as a vote?
Unless the answer to the first question is "yes" and a uniform standard is
used in the recounting, the "actual" count will be no more meaningful than
any of the previous "actual" counts.
Don't forget that a statistical fluctuation of about 0.00004 of the total
would be enough to throw the result the other way. If that happened, Gore
should not be given the Florida electors. Neither should Bush if the
opposite happened. Either the election should be declared inconclusive, or
the electoral votes should be split. The fact is that it's impossible to
deduce the will of the voters in Florida. The signal was drowned in noise.
Everything else
aside, this election has provided great stuff for my statistics class to
mull over in the Spring semester -- like how a giant spike for Buchanan in
Palm Beach turns out on analysis to be much ado over nothing.
So you say. I have yet to verify that. I think you need to get confirmation
of your analysis from an independent source before you start spreading it
around. The plot I saw on TV suggests otherwise: if the Buchanan spike was
ten units, I didn't see any other spikes on the county-by-county plot that
looked as high as three units. Of course I didn't get a really good look.
Is there a county-by-county tabulation anywhere?
This could really go on forever. I agree with you that we have more
important things to talk about, now that all arguments are moot.
Best,
Bill P.