Up: Home page
Previous: Issue 18
Next: Issue 20
Voting matters - for the technical issues of STV
Issue 19, November 2004
(Complete issue in PDF, 135Kb.)
Editorial
Report by Steve Todd
On 9 October this year, New Zealand held a number of STV elections
using the Meek counting rules. Several problems arose which delayed
the final declaration of the results. It appears that the main problem
concerned reconciling the number of voting papers that were scanned
into the database with the number that were subsequently sent to the
STV calculator.
The realisation that discrepancies were occurring led the local
councils and district health boards (DHBs) affected, to call in the
Auditor-General's office to audit the entire process. While the
computer error was discovered and fixed within a few days, the auditing
process meant that it took four weeks to complete all the
vote-counting. In contrast, the program which actually performed the
count, i.e. the STV calculator, appeared to operate without mishap.
A lesser, but equally frustrating, problem was that the ICR technology
used to process the ballot papers was unable to read (with a high level
of confidence) a considerably higher percentage of the scanned
documents than was expected. This led to much more human intervention
than was expected, with a consequent increase in the time taken to
process the votes.
The Justice and Electoral select committee of New Zealand's parliament
intend to conduct an inquiry into what went wrong. A focus of the
inquiry will likely be on why the two Auckland-based companies
contracted to process the STV votes in the northern part of the
country, did so seemingly without a hitch, and in a timely manner,
while the Christchurch and Wellington companies contracted to conduct
the remaining STV elections (in respect of 7 of 10 councils and 18 of
21 DHBs) did not.
There has not yet been a full explanation of the problems encountered,
but there is a suggestion that the computer systems used by the
Christchurch and Wellington companies may not have been completely
compatible.
There were also widespread claims of voter confusion (said to have been
caused by having FPTP and STV elections on the same A3-size voting
documents), leading to many Informal (Invalid) votes (errors) and blank
votes (non-participation) being cast, that the select committee will no
doubt inquire into.
Informal votes in council areas using STV appear to have been no more
than usual - 1.08% in Wellington and 1.49% in Dunedin, for example.
However, in the remaining 64 council areas, that used FPTP, the
Informal rate in respect of their DHB elections was up as high as 10 to 12%.
A likely explanation for this will be poor voting-document design.
There was no bold distinction between FPTP and "tick-voting" for the
mayoral and council ward elections, and STV and voting by numbering the
candidates in the DHB elections. In fact, apparently due to printing
restrictions, the DHB elections were set out under the name of the city
or district councils they were associated with! This means that some
voters (who did not read the voting instructions carefully) carried on
tick-voting into the DHB election - more than one tick for the
candidates and the vote was informal.
On the brighter side, the actual ballot data is likely to be made
available in respect of most, perhaps all, STV elections and hence it
will be possible to `check' the counts by re-running them.
Voting matters
There are 3 papers in this issue:
- B. A. Wichmann: Tie Breaking in STV. This paper considers a method of
handling ties when a computer is used by considering all possible outcomes.
It is an unfortunate fact that breaking a tie by a random choice gives an
impression that the outcome might be random when this is rarely the case.
- J. Green-Armytage: Cardinal-weighted pairwise comparison. This
paper considers the election of a single candidate by adding information
to a Condorcet-style count on the strength of the preferences for candidates.
- B. A. Wichmann: A Working Paper on Full Disclosure. This paper attempts
to put together major concerns about this issue which have been raised
in previous issues of Voting matters. The paper was written before the
New Zealand election data became available and hence does not mention this.
Readers are reminded that views expressed in Voting matters by
contributors do not necessarily reflect those of the McDougall Trust
or its trustees.
Papers with citations
- B. A. Wichmann: Tie Breaking in STV. (p1-5, PDF 40Kb)
- J. Green-Armytage: Cardinal-weighted pairwise comparison.
(p6-13, PDF 70Kb)
[27, 1-21, PDF].
- B. A. Wichmann: A Working Paper on Full Disclosure. (p14-16, PDF 32Kb)
Main index
Author index