From: William Langham (firstname.lastname@example.org)
Date: Thu Nov 16 2000 - 00:17:06 EST
This is a bit long, but provides a good run down on computers and
---------- Forwarded message ----------
Date: 15 Nov 2000 20:55:25 -0000
Subject: CPSR Answers Computer-Based Voting Technology Questions
Computer Professionals for Social Responsibility (CPSR), a public
interest organization that focuses on the benefits and risks to society of
computer technology, offers the following answers to frequently asked
questions about computer-based voting technology.
Q: Why do vote counting systems produce different totals when the ballots are
recounted? Shouldn't machine counts and recounts of ballots produce
repeatable, reliable results?
A: Many people have wondered why a computerized vote-counting system
would have any significant inaccuracies. Some have publicly speculated that
such variation must be the result of deliberate human action. Some people
believe that computerized counts will always be more accurate than human
counts, because of inevitable "human error."
However, computerized vote-counting systems are complex, prone to
several kinds of error. Well-designed vote-counting systems minimize these
errors. Some systems, particularly older systems, are not so well-designed,
and are more prone to error. To illustrate this problem, we will describe
some reliability problems with the oldest type of computerized
ballot still in use, the Vote-O-Matic(tm). This system was once very popular
and is still used in many places, including 15 Florida counties: Broward,
Collier, Dade, Duval, Highlands, Hillsborough, Indian River, Lee, Marion,
Osceola, Palm Beach, Pasco, Pinellas, Sarasota, and Sumpter.
The following describes reliability problems associated with one phase of
the elections process: gathering ballots and running them through readers.
Problems may occur in other phases, included materials design and printing,
polling place administration, voter education, and vote tally software. By
focusing on this one phase, we do not imply that the other phases are
trivial. Conducting elections is demanding work, in all phases. Also, this
paper focuses on errors. Elections frauds certainly have arisen in the
history of American politics, but to our knowledge no fraud has been alleged
in the ballot counting process for this election. Some level of error is
inevitable when counting Vote-O-Matic ballots, however.
CPSR has been studying Vote-O-Matic-type vote counting systems for over ten
years. Experts, including CPSR's own project personnel, have concluded that
the Vote-O-Matic system has inherent accuracy limitations. Furthermore,
careful manual counting of Vote-O-Matic ballots should always be more
accurate than machine counts.
The Vote-O-Matic system uses as a ballot the Hollerith punch card, also
known as a "computer card." This once-common card is roughly 3" by 7", with
small rectangular holes. For Vote-O-Matic cards, each hole in the card
represents a vote for one candidate (or in favor or against a ballot
measure). The ballot is counted by feeding it, short-side first, into a
reader. (The card is made with one corner clipped, so that the correct end
of the card is fed in first.) The reader has lights and sensors. When a hole
passes over the sensor, light shines through, and the hole is read as a
Hollerith cards were used for the 1890 census, and millions and millions of
critical activities between then and the 1970's. Thus, one might expect the
Vote-O-Matic system to be extremely reliable. But important differences
between the standard Hollerith card and the Vote-O-Matic card make the
Vote-O-Matic far less stable and reliable. There are three main problem
- ballot reader machines
- what happens when a ballot reader reads a ballot
The ballots use essentially the same card size and hole positions that IBM
adopted in 1924 soon after they bought Hollerith's company. However, the
cards are not the same. Hollerith's approach was to punch a hole in a solid
piece of paper. Vote-O-Matic cards are pre-punched. Each square "chad" is
held in place by a small wad of paper fibers at each corner. The vote then
makes a hole by pushing the chad out with a round stylus. However, sometimes
a chad will be partly punched out or will snag on something and be pulled
out, creating what is known as "hanging chad."
Hanging chad can be attached at one, two, or three corners. Chad attached at
one corner are usually torn off by the card reader or in handling. Chad
attached at two corners are also often torn off, unless the two corners are
on the side of the chad that is fed first into the card reader. Then, often,
the chad will be forced back into the hole, only to flap open again later.
chad attached at three corners are also usually forced closed by the card
reader. Handling the cards can also change the status of hanging chad. Some
studies have been done on chad, but there are many independent variable and
complicating situations, so the preceding is a generalization.
These pre-punched cards are also reportedly sensitive to changes humidity.
The reasons have not, to our knowledge, been studied, but it is likely
because the chad loses and gains moisture faster than the bulk material.
Thus taking a box of Vote-O-Matic cards from an air-conditioned room to a
humid evening to another air-conditioned room will have unpredictable
effects. It may take the cards some time to settle down after the ordeal.
The pre-punches also make the cards less rigid than a normal Hollerith card,
and thus more prone to bending. Bent cards often cause problems during
reading. The trailing edge of the card is uneven, because of tabs from where
the write-in tab was detached. The faces of the card are not as smooth as a
regular card, again due to the pre-punches.
* Ballot Reader Machines
So far as we know, there are no longer any manufacturers of Hollerith card
readers. High-speed card readers have a lot of precision parts. Existing
readers must be periodically rebuilt, but many companies no longer exist and
the remaining manufacturers, so far as we know, no longer offer maintenance
contracts on the units. Elections is about the last market left for
Hollerith card readers. Elections companies buy up equipment from counties
as they move away from Vote-O-Matic systems, and sell it jurisdictions still
Elections aren't a particularly hard life for a card reader, since a reader
is only used for a few days a year. Still, the readers eventually need to be
rebuilt, which elections companies do with a dwindling supply of spares,
hangar queens, and whatever rebuild protocols they devise.
Still, some parts age more on calendar time than with use. As the readers
age, they become less reliable and more prone to error and breakdown.
* What Happens When a Ballot Reader Reads a Ballot
Ideally, a stack of ballots is sucked one-at-a-time from the input hopper to
the output hopper of a card reader, each being counted accurately. However,
sometimes two cards are sucked through. This is probably because
pre-punching makes small ridges on the bottom of the card, and an identical
pattern of small troughs on the top. The ridges tend to get caught in the
troughs. Also, feed mechanisms have to be engineered with consideration of
the air cushion between the cards, as one moves relative to the other. This
air cushion will not have the same properties for Vote-O-Matic cards as for
normal cards, due to surface roughness. For whatever reason, misfeeds
Hanging chad can flip open and closed. Detached chad can become stuck in the
feed path, increasing double feeds and misfeeds. Detached chad can jam two
cards together, increasing misfeeds. In some machines, detached chad can jam
over the light or sensor, causing holes to not be read until the chad blows
out of the way. Detached chad can migrate from one card to the next.
Chad that was not detached before, but merely buckled or only detached on
one corner (which counts as "not an open hole" in many jurisdictions) can
catch on other cards and become hanging chad or be torn loose.
The read process can be quite traumatic to a Vote-O-Matic card.
Q: Is counting ballots by hand more or less reliable than counting them by
A: A human count of
Vote-O-Matic cards should almost always produce a significantly more
accurate result than automated reading. People cannot count cards as
quickly as a card reader, but a card reader is much more limited than a
person in how it can handle and read a card. Any damage a card has sustained
can confuse a card reader or cause it to malfunction. People are better able
to deal with such problems.
Unfortunately, reading a Vote-O-Matic card by machine changes the card.
Cards that have had one or more trips through a high-speed card reader will
appear different to a human reader than they would have when freshly punched
by the voter.
Erik Nilsson, an election technology analyst for CPSR, believes that the
Vote-O-Matic system should be replaced. "For a quarter century, election
experts have been calling for the Vote-O-Matic system to be retired. The
results of the 2000 election show that it is now time move beyond this
Q: Would Internet voting solve this problem?
A: Internet voting is often suggested as a solution to election counting
problems, but has many problems of its own, for example:
- If people voted from home, it would be very difficult, perhaps
impossible, to assure that those who vote are who they say they are.
Someone could vote for one of their family members, for example.
- If people vote from home rather than in a polling place, vote secrecy and
privacy could be compromised. Elections in many democratic societies,
including the U.S., are based on the promise of secret ballots, where only
the voter knows who he or she voted for (unless he or she chooses to tell
- A home-based Internet-based voting system would favor people who have
computers and Internet connections at home. Such amenities are not
possessed by all citizens in the U.S.
- Purely electronic ballots leave no paper trail, so electronic subversion
of voting records could be difficult or impossible to detect.
- Voting from home could destroy the sense of shared civic responsibility
and pride that most people clearly feel when they go to an actual polling
place to vote.
On the other hand, Internet voting could offer the following advantages:
- Customized presentation of voting choices, for example voter-selected
- Reliable vote tabulation
- Access for the disabled, and rural
- Can handle large numbers of voters
Computers, of course, can and will be used in elections. One approach that
could provide the advantages without many of the disadvantages would be to
provide Internet terminals in polling places. Voters would come to the
polling place and identify themselves, as they always have. Vote-O-Matic
and other outdated, unreliable systems would be replaced by more current
technology. Each polling place would have a "manual" backup system on
site, for when the network connections or computers fail (as they surely
will) or when a voter is simply unable to understand how to use the
A home-based Internet voting system is completely out of the question until
access to the Internet in the U.S. is universal. Until such a time,
adopting a home-based Internet voting system would be unconstitutional.
Today we are far from universal access. For example, in some urban poor
districts, 14% of households lack even basic phone service, much less
Internet connections. On some Indian reservations, the percentage of
phone-less households is even higher: 40%.
Computer Professionals for Social Responsibility was founded in the early
1980s by computer scientists and engineers who were concerned about the
use of computer technology in military applications, particularly the Strategic
Defense, or "Star Wars", Initiative. In the mid-1980s, the organization
branched out to include other issues, such as electronic privacy, freedom of
speech, and the use of computer technology in elections.
For further information, please visit CPSR's website:
http://www.cpsr.org/issues/voting.html or contact the CPSR office
at 650-322-3778 or email@example.com.
Susan Evoy * Managing Director
Computer Professionals for Social Responsibility
P.O. Box 717 * Palo Alto * CA * 94302
Phone: (650) 322-3778 *
This archive was generated by hypermail 2.1.2 : Fri May 31 2002 - 23:55:01 EDT