< Back to table of 'First Report - December 2004 Contents'
Part 2 – Work of the Commission
2.9 Other Issues
Review of Hand-counted Voting System
A comparison of the performance of the current paper system and the chosen electronic system was carried out for the information of the Commission and the results are reported at Appendix 2L PDF.
Eight issues were selected for comparison on the basis that they were considered relevant to the Commission’s work. The electronic system was found to be superior on three issues (removal of randomness, prevention of unintentional invalid vote and removal of subjective judgment) and the manual system was also found to be superior on three issues (avoidance of malpractice, opportunity to cast blank vote and transparency and legitimacy).
Of the remaining two issues (relating to major and minor counting errors) it was not possible to say which system was better but it was thought likely that, if the trustworthiness of the electronic system could be established, it would prove superior.
Rules for the Counting of the Votes
A recurring issue which arose in various forms in the course of the Commission’s work relates to a particular aspect of the rules for the counting of the votes.
Under the current rules, the method of distributing surplus votes involves a degree of variability that could, in theory, affect the outcome of the count in question and/or of subsequent counts. This variability takes two forms.
The first arises when the proportions in which a given surplus is to be distributed have been determined and the next step is to transfer X surplus votes to candidate A, Y votes to candidate B, etc. The question is: Which of the X physical ballot papers should be transferred? The answer is: take X votes from the top of the bundle of papers that have second (or next available preferences) for candidate A6.
The second element of variability arises with surpluses that accrue on the second or subsequent counts where, rather than having the whole of an elected candidate’s votes examined to establish what proportion of their total votes favours each remaining candidate (and then distributing the surplus accordingly), the proportion of the surplus to be assigned to each remaining candidate is determined on the basis of their share of the continuing preferences in the parcel of votes last received by the candidate whose surplus is being distributed.
Although capable of eliminating this and other minor anomalies in the calculation of proportional distributions of surpluses (i.e. in this case by calculating the proportions in the whole of the elected candidate’s votes which would be assigned to each remaining candidate), the chosen electronic system has been designed to reproduce the manual system in exactly its present form.
There is a more accurate method of counting, known as the “Gregory method”. This always gives the same result and is in fact the system used in elections to Seanad Éireann. The reason it is not used in other public elections is, presumably, that it is time consuming to implement by hand. However, electronic systems of counting are perfectly suited to handling this method.
The Gregory method of counting could be easily implemented in the electronic system and would increase accuracy of the process and thereby enhance the value of electronic voting by allowing the system to be used to its full capacity. Such a change would, more importantly, be more democratic in that it would allow every relevant preference to be taken into account. Any change to the counting rules would require an amendment to the law governing the various types of elections.
This and some further issues regarding inaccuracies arising from the current counting rules and process are discussed in the report at Appendix 2L PDF.
Verifiable Audit Trail
The Commission’s terms of reference required it to consider the chosen system, which does not include a verifiable audit trail. However, while the issue of an audit trail therefore does not fall within the Commission’s terms of reference for consideration, the Commission sought to establish the feasibility for audit within the system, not least as a useful indicator of “accuracy” as contained in the Commission’s terms of reference.
The issues surrounding an audit trail in the context of the chosen system are accordingly set out in a report at Appendix 2M PDF which concludes that the chosen system is not feasibly capable of modification to provide a full audit trail without unrealistic cost.
The consequences of this are that while it is possible to independently verify that the votes recorded on ballot modules are those uploaded onto the PC at the count centre, there is no method of validating that the votes stored on ballot modules are those which were originally entered by the voters using the voting machine at the polling centre. The report also outlines a number of alternative voting systems which claim to offer verified voting.
The Commission has noted, in the context of its review of secrecy and accuracy issues, that the existence of an audit trail is principally an indicator rather than a determinant of accuracy and that in certain respects it is inconsistent with the competing requirement of secrecy of the ballot.
Tallying at Elections
Under electronic voting as currently proposed, the opportunity to “tally” likely results through observation of the ballot papers being counted is lost. A feature of the chosen system which may be used to address, in part, the loss of the tallying function is the facility to publish subsequently the votes cast at an election.
Although this loss of the tallying function enhances the secrecy of the ballot, it also reduces transparency to a level below that which exists in the manual system as regards visual verification of the accuracy of the counting process. Conversely, the proposed remedy for this reduction in transparency, namely, the publication of the votes cast, can have the effect of diminishing voter secrecy in certain circumstances, i.e. where voters deliberately record “signature” voting patterns in lower ranking preferences to make their votes identifiable.
6The justification for this procedure is that the mixing of the votes at the beginning of the count ensures that the physical papers that are transferred are a random and therefore representative sample of the whole bundle. The problem is that there is a not insignificant degree of variability in the extent to which this is so. To the extent that it is not so, this create potential variability in subsequent counts in which, on an elimination or on a surplus transfer, these transferred ballot papers are examined for preferences for continuing candidates.
< Back to table of 'First Report - December 2004 Contents' |
