Up: Issue 10
Previous: Paper 4
Next: Paper 6
Voting matters - Issue 10, March 1999
A review of the ERS97 rules
B A Wichmann
Recently, I was asked to interpret the Newland and Britton 3rd
edition rules (referenced as ERS97) with some specific examples and
therefore read the rules carefully for the first time. I think I was largely
successful in interpreting the rules correctly, but was surprised at a
number of features of their presentation.
Over the last 20 years, I have been involved with the specification of
programming languages for the International Standardization Organization
(ISO). The requirements here are again for precision and clarity. ISO have
adopted drafting rules for standards which I think are very helpful and are
not far removed from the style of the presentation of section 5 of ERS97.
There are a number of detailed differences in which I prefer the ISO
approach. These differences are as follows:
- Separation of normative (requirements of the standard) and
non-normative text. In ISO, the model election would be a non-normative
annex. In fact only sections 5 and 6 are normative.
- In ISO, defined terms would appear before the main text. In ERS97, the
Glossary in section 6 appears after their usage in section 5.
- In ISO, notes are non-normative and laid out in a manner to make this
clear. The note in 5.6.2 in ERS97 is clearly normative (and uses
shall, as in ISO standards).
It seems to me that the adoption of the ISO drafting rules would be a
worthwhile undertaking if any revision of the rules was contemplated.
Indeed, I see no reason why a suitable revision should not be proposed to
ISO as a standard, since it would allow other organisations (in any country)
to use it by reference. Currently, many organisations contain rules for STV
in their constitution which is unsatisfactory when the rules themselves are
very old - a method of reference would be useful in such contexts.
My major and perhaps controversial comments on the rules arise from my
desire to see it formulated more closely as an algorithm, rather than as a
description. In trying to interpret the rules, one is necessarily performing
a function like that of executing a computer program. Since the main purpose
of the rules is surely to aid Returning Officers, then the computer program
approach is helpful. Of course, I am not suggesting that computer
terminology should be used, but merely that the style should allow for
conversion into a program in an obvious manner.
My specific points arising from the above computer perspective, and from
other analysis are as follows:
- There is no provision for conducing a count with the aid of a computer
or by an entirely automatic process. Since computer programs of both types
are routinely used for this purpose, this is a major fault. Note that the
Church of England rules make specific provision for this,
including the certification of appropriate programs by ERS. Breaking ties by
lot needs a different wording allowing for the use of a pseudo-random number
generator.
- I think the wording associated with checking and records should be
separated by being in a paragraph after the corresponding actions. (This is
not straightforward as some paragraphs are a mixture.)
- As I see it, only those paragraphs which are needed for reference
purposes need be numbered. This would reduce the apparent complexity of the
rules. Currently, the whole of section 5 needs to be read to determine what
use is made of each part of the rules.
- Section 5.5 (completion of the count) is not referenced at all, since
it is invoked when appropriate conditions are satisfied. This is not
algorithmic in the conventional sense, indeed, in computer terms could be
seen as 'interrupt-driven'. I think this section should be used
explicitly.
- The calculation of the quota and the recording of transfers appears to
give the impression of undertaking computations to one hundredth of the
vote. However, this is not achieved, since that accuracy requires that the
transfer values are computed to a greater accuracy. Indeed, if p
votes are transferred, then there is a truncation error of at most
p/100, which implies that transfer values should be computed to about
(number of digits in total votes)+1 digits. I do not believe that an
arithmetic approximation which can lose a whole vote is acceptable since
the voter could reasonably equate the loss to his/ her vote. Unfortunately,
the rules depend upon (number of papers)×(transfer value) in
hundredths of a vote, so it is difficult to increase the accuracy without
complexities elsewhere. Hence I conclude that this problem is inherent in
this type of rule and could be seen as a defect in ERS97.
- The rules mention coloured forms, but the colouring is not apparent in
the copy of the forms in the example - the solution is to print the
'beige/blue/ green/white/pink/yellow' on the forms, so that photocopying
them retains the information (or so they can be photocopied onto the correct
colour paper).
- Not all uses of the defined terms appear in bold in the rules. I would
suggest that the uses of a defined term uses a different font (say,
italic).
- Paragraphs used in more than one place should be given a name and
referenced by name (as with the sections 5.3 and 5.4).
- A batch is a set of bundles each having the same transfer value, not a
type of bundle as given in 6.1.
- The definition of stage of the count is ambiguous, or perhaps depends
upon the layout to parse.
- The definition of transfer value should have 'deemed elected' rather
than 'elected'.
- The statement that for small elections counting slips are not required
should be made once at the start, rather than each time slips are
mentioned.
- The second sentence of 5.6.4 is confusing. Surely the point is that an
auditable record of the count should be kept? If all recording forms are
optional, then why are counting slips specifically mentioned in 5.1.3,
5.3.12 and 5.4.3?
- The term 'formally excluded' (in 5.5.2) clearly means exclusion without
the application of the rules associated with exclusion, although this is not
explicitly stated.
I have attempted to reformulate the rules along the lines that ISO would
use, but I do not regard the result as at all satisfactory. My attempts were
based upon a minimal change to the wording, but it appears that a more
radical approach is needed.
A few issues have been noted by others that I should also add for
completeness:
- a) Conventional practice appears to be that the transfers values are
not included in the result sheet. I do not like this, since the values are
hard to reconstruct and are available.
- b) The handling of withdrawn candidates is not mentioned in section 5
of ERS97, although it is surely a possibility with all elections (and is
noted in section 2.2).
- c) A minor ambiguity has been noted in the rules. (I hope to report
fully on this in the next issue of Voting matters.)
Conclusions
Is any 'improvement' to the wording needed? I think the rules should be
readily usable just from the booklet. In this regard, the model election and
examples given are very helpful. However, they do not cover all the
situations that can arise. Moreover, for the model election, the actual
papers are not included (not unreasonable for 785 voters, but this means
that this single long example cannot be re-worked completely by the reader).
Also, the explanations given are not always adequate. For instance, in
Section 8.2 it is said that, because the surplus could change the order of
the last two, it 'must be transferred', without any hint being given that it
is required to look at whether the next two or more to go out are definite,
in which case it must not be transferred. In the particular instance the
action taken is correct, but that is not the point. How to decide that it is
correct is not fully stated as it should be [3].
Of course, the fact that ERS runs courses in conducting an STV election is
very helpful as is the large number of people that have had such training
and can pass on their skills to others.
Hence I conclude that improving the wording is not that vital, but it would
be a shame not to consider the ISO approach if a revision was produced in
the future.
References
- R A Newland and F S Britton. How to conduct an election by
the Single Transferable Vote. ERS 3rd Edition. 1997.
- Church of England, General Synod. Regulations for the
conduct of elections by the method of the single transferable vote. GS930. 1990.
- I D Hill. Private communication.
Up: Issue 10
Previous: Paper 4
Next: Paper 6