Re: Florida vote count weirdness

Date view Thread view Subject view Author view

From: Gordon Mohr (
Date: Tue Nov 14 2000 - 10:02:40 PST

Robert S. Thau writes:
> From eyeballing this chart, the vast majority of the net change in the
> totals came from a very few counties where the change was heavily
> imabalanced in favor of Gore, with the bulk in just two:
> Palm Beach - net change +643
> Pinellas --- net change +478
> Duval --- net change +168
> Gadsden --- net change +163
> Polk --- net change +129
> The highest changes in favor of Bush were
> Martin --- net change -105
> Seminole --- net change -98 ---- (FWIW, last county to report)
> No other county that I can spot had as much as a fifty vote net change
> in favor of Bush. Note also that there were counties as populous as
> many of these (Hillsboro, Miami-Dade, Broward(!)) where the shifts
> weren't nearly as big, less than 50 --- some of which may have to do
> with what equipment is used where, to be sure.
> Even so, I am at a loss to explain how you generate numbers this
> skewed by rolling fair dice of any description; am I nuts, or is this
> statistical evidence of irregularity in the initial vote count ("let's
> just not count these ballots here")?

I thought this odd too, and was beginning to suspect that smaller
counties simply did something different in their initial counts --
for example, fed through batches of cards in smaller numbers,
individually investigated cards which registered as errors, etc.

Turns out there's another possible explanation: smaller counties
are more likely to have upgraded their ballots and counting machines
to an optical-scan system which significantly decreases, by a factor
of 16, the incidence of uncounted ballots.

I've only seen this mentioned in Salon:

So, when you realize that Bush won more counties, but those counties
were both (1) smaller; and (2) less likely to have missed votes the
first time, the pattern seems explainable.

As such, the Bush campaign may have to request manual recounts in
Bush counties, but it probably will only lead to a notable boost
in counties where a punch-card ballot is used.

The Gore campaign is likely to keep finding votes through manual
recounts in punch-card counties they won. HOWEVER, I suspect
that those local poll-workers may have already, at the time of the
first *mechanical* recount, taken extra care to increase the
number of ballots that were registered -- for example shaking or
shuffling ballot groups with a high percentage of noncounts, to
loosen 'chads'. That could explain Gore's 1500-vote boost in the
first recount, but would mean that much of his potential gain has
already been achieved.

It's certainly a mess; these punch-card ballots should be banned,
along with the wacky lever-systems I've only seen on TV. Both
seem designed for fraud and confusion.

I have my doubts about anonymous/secret voting too. Is the threat
of voter intimidation/inducement greater than the threat of fraud?
And given that we can punish either intimidation or fraud after the
fact, but we can only confirm and correct some kinds of fraud
without anonymity, do we want to err on the side of allowing fraud?

I used one of the OpScan ballots here in SF this year -- it seems
much less prone to confusion and counting-mischief than punch-cards.
The names of the candidates/measures are on the ballot itself, and
the ballot is marked with a pen. There was plenty of white-space
around labels and marking areas.

- Gordon

Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Fri Nov 17 2000 - 16:56:43 PST