Date: Thu, 16 Nov 2000 03:36:35 -0800 (PST)
From: [Not authorized to distribute]
To: rusin@math.niu.edu
Subject: your election statistics article

Dave,
     Earlier this evening as I was scanning through an on-line news"paper,"
your name jumped out at me from the screen, so I decided to take a look at
your article at http://www.math.niu.edu/~rusin/recount/index.html
It seemed a decent presentation of the statistical issues and may help a
few people grasp the statistical issues involved who otherwise would not.
     I have some doubt, however, about the validity of a claim you made
in limitation #1 under "Analysis of Palm Beach County" and Footnote [3].
It is a wonderful thing that punched cards in general and pre-perforated,
manually punchable card specifically are obsolete.  It is similarly awful
that elections still use the latter.  Although it has been 15-20 years
since I last had to deal with the pre-perforated cards, I have not at all
forgotten how flimsy they are.  They are flimsy before use and become
even more so when punched.  You claim that hanging chips--I had never
encountered the term "chads" until I saw it in the newspapers this past
week--would disappear quickly during successive passes through card
readers or human hands.  While that claim is probably true for the chips
that were loose already at the start, the problem is that new chips can
be torn loose at their perforations just by handling or by being run
through a card reader.  Worse yet is that these cards have a relatively
high failure rate each and every time that they go through a card reader.
Such failures most commonly include frayed edges and major tears, but
sometimes involve torn dividers between the punchable spots/blocks on the
cards.  These cards are not durable like the cards you probably remember
putting through a keypunch machine.  My guess is that the error rate (in
the original total number of ballots cast) due to card damage might
decrease from the first to the second tally, but would then increase
again thereafter.

[deletia -- djr]
==============================================================================
Date: Thu, 16 Nov 2000 12:57:17 -0800 (PST)
From: [same]
To: rusin@math.niu.edu
Subject: Re:  your election statistics article

>Thanks for your detailed comments about the punched cards. Some huge 
>number (maybe a third of the electorate) of voters in the US still use
>these. I suspect there will be a lot of pressure to get rid of them
>after this. When I used punched cards in the 70s to write computer
>programs, we didn't give much thought to misread cards; there were
>plenty of other bugs to dominate the process... But you're right, they

     How true.  But we were not using the perforated cards, which only
had 40 columns, for programming and normally not even for data.

>are flimsier now, and there are certainly more of them to worry about.

     Maybe my writing wasn't too clear just before bedtime last night.
It's not that the cards have become more flimsy over the years, but
that the perforated cards are already very flimsy and flexible.  Each
of the 40 columns, which were at the same positions as all of the
even-numbered (or maybe it was odd-numbered--I no longer remember)
columns of an 80-column card, is effectively like a little hinge in
the card because the card is weakened there by the column of perforations.
I dealt with them so rarely that I don't recall whether unpunched holes
would tend to become punched through flexing of the cards, but the
grosser problems I described before were very, very frequent.  A stack
of that kind of card tended to have jams on *every* pass through a card
reader.  For one thing, it was difficult to get the cards to lie flat,
so a stack of them tended to be "fluffy", puffed upward.  Weight on them
helped, so one trick was to put several inches of blank, normal cards
on top of the perforated cards to hold them down, so the card reader
could grab the last few hundred perforated cards with some chance of
not jamming or missing them.
>
>By the way our local voting (I think that means county-wide) switched
>to filling in ovals with markers this year, but the ovals are
>printed on similarly thin cardboard. I wonder how many votes jammed
>or got mutilated by the machines? 
>
     I've been wondering about that, too.  We had two ballots apiece
done that way, one for statewide initiatives and one for candidates
and local initiatives.  Each sheet arrived in a normal letter envelope,
neatly folded in three places to fit the envelope, and they were to
be returned inside a couple of other more or less normal-sized
envelopes.  Their machines must be pretty good to deal with the folds.

>Thanks again for writing. I hope you don't mind if I quote a bit of
>your message to people who want to know how tabulations can be in error.
>
     No problem.  I noticed while writing to you last night that you
had updated your page again.  I took a peek at it and decided most of
it hadn't changed, so I didn't go through the whole thing again.  But
since you're updating it frequently for now, I'll probably take a
closer look later.