[E-voting] interesting note on voting error rates from New Scientist

Justin Mason jm at jmason.org
Wed Oct 27 00:08:00 IST 2004

Hash: SHA1

Found in New Scientist vol 184 issue 2469 - 16 October 2004, page 6 --

Sorry -- they have for-pay archives, so I'll just snip a little bit under
fair use ;) -- here it is:

  The accuracy of a voting system is often assessed by what is called the
  "residual vote". This is the difference between the number of voters who
  turn up at polling stations and the total number of votes allocated to
  the candidates. Voters can still choose to spoil their ballots, but if
  one system regularly produces a higher residual vote than another its
  accuracy may be questioned.

  According to this criterion, the most accurate way to record votes is to
  use optical scanning machines. These work in a similar way to
  photocopiers, and register a voter's pencil mark on the ballot by the
  amount of light it absorbs. These systems produced a low average
  residual vote of around 2.1 per cent during presidential elections from
  1988 to 2000, according to a study to appear in the Journal of Politics
  by Stephen Ansolabehere and Charles Stewart of the Caltech-Massachusetts
  Institute of Technology Voting Technology Project.

  Touch-screen voting machines have a far higher residual vote of 3.0 per
  cent. These machines register a residual vote when a voter activates the
  machine but then fails to cast a vote. Experts attribute the high
  residual vote on these machines to their sometimes confusing or annoying
  interface, which require voters to navigate a menu and touch the screen
  to register their vote for their preferred candidate. Punched cards have
  a residual vote of 2.9 per cent.

this is aside from the other errors an EVM can make.  But it's a good
figure indicating of how "confusing" a voter finds the equipment to be.

Looking at the data from that E-Voting pilot in Dublin (
<http://lists.stdlib.net/pipermail/e-voting_press/2004/000012.html>), it
seems the residual rate might indeed have been about 2.4%-2.9% in that
case, comparable with the residual vote rates of touchscreens in the US.
hard to tell without a more careful look, though.

Also, another snippet from a sidebar on the same article:

  Europeans tend to be puzzled by America's enthusiasm for mechanical and
  electronic voting systems.  [jm: hah!]   What can be wrong, they ask,
  with writing an X in a box on a ballot paper?

  The answer has to do with counting the votes rather than casting them.
  First, there's the logistical challenge, especially in the US. On
  election day, an American voter may be asked to cast dozens of votes,
  from choosing a presidential candidate to deciding funding for local
  after-school programmes. Hand-counting ballots would take days or weeks
  in many US counties, and because many different electoral races are
  recorded at the same time making piles of papers for each candidate is
  impractical.  [jm: hardly an issue in Ireland.]

  Then there is the problem of accuracy. Stephen Ansolabehere and Charles
  Stewart of the Massachusetts Institute of Technology recently studied
  the results of New Hampshire elections over the past 40 years. The state
  keeps records not only of certified votes but also of figures for
  recounts after each close race. The pair found that counting by hand
  allocated an average of around 2 per cent of votes to the wrong
  candidate. Optical scanning, which automatically detects marks on paper,
  misallocated only 0.4 per cent of votes, presumably when ballots had
  been misfed into the machine or had jammed.

  When votes are recounted, the error rate falls, often to as low as a few
  hundredths of a per cent. Yet in many countries, the margin of error for
  counting by hand is higher than the level at which recounts are
  triggered, and this places the validity of some results in doubt,
  Ansolabehere says.

Go optical scanning!  Unfortunately Ireland's unlikely to go there, but at
least it provides a way to say "look, this is a much better high-tech
system, we're not just Luddites".

- --j.
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Exmh CVS


More information about the E-voting mailing list