2.2 Local authority STV elections – October 2004

2.201
Triennial elections for local authorities were held on 9 October 2004, along with those for District Health Boards (DHBs) and Licensing Trusts. A number of local authorities and licensing trusts chose to base their elections on the single transferable vote (STV) system. The majority of local authorities, however, retained the first past the post system (FPP). All DHB elections were conducted on an STV basis.

2.202
The New Zealand STV-based electoral system requires voters to rank candidates on the basis of preference. A computer-based system then analyses the results and determines the successful candidate(s) once the manually cast votes have been digitised for processing.

Background

2.203
The majority of elections were completed successfully over the weekend of 9 and 10 October. However, the STV elections associated with vote-processing company electionz.com and its sub-contractor Datamail encountered major processing problems.

2.204
In the 3 weeks leading up to 9 October, the Electoral Officers (EOs) had progressively passed validly cast votes to electionz.com and Datamail for processing, but during the weekend of final counting a number of these votes were found not to have been correctly entered into the STV calculator.¹ The problem occurred during the “transformation” stage when manual voting papers were converted to computer-readable, digitised data for input into the calculator.

2.205
Seven local authorities, 18 DHBs, and one licensing trust were affected, all of which had contracted electionz.com to process the votes cast. The immediate effect was that EOs were unable to finalise the results of the elections they were each responsible for.

2.206
In response, electionz.com and its contractor, Datamail, reviewed and enhanced their procedures and controls to allow the vote count to be completed.

Our involvement

2.207
We were approached by 6 of the 7 local authorities affected, the 18 DHBs, and the one licensing trust, with support from their relevant EOs. We agreed to provide assurance that these enhanced procedures and controls were operating as they should, and that all validly cast votes received by electionz.com were being processed effectively into the STV calculator².

2.208
This represented a unique challenge for the Office of the Auditor-General, which has no formal or statutory role in local authority elections. Nor had the Office previously been asked to become involved in either an STV or an FPP election.

2.209
There were limits to our assurance. For example, we could not provide assurance over the work or judgements of the EOs in providing the votes cast in their elections to electionz.com.

2.210
Our work involved:

• observing the recovery process once the initial issues were identified;
• assessing the reasonableness of the improved processes and controls implemented by electionz.com and Datamail; and
• verifying the subsequent results.

2.211
Throughout the project, we liaised regularly with a “stakeholders group” assembled by the Department of Internal Affairs, and a grouping of affected EOs.

The impact of our intervention

2.212
The first request for assurance was received from Wellington City Council on 12 October. The other local authorities, the DHBs, and the licensing trust, quickly followed with requests of their own. From 14 October we were on-site at the premises of electionz.com/Datamail in Petone until the final outstanding results were cleared on 5 November.

2.213
It was agreed with the “stakeholders” that the enhanced system needed to be trialled for accuracy on 3 local authorities and one DHB. This process was time-consuming and was not completed until 28 October – a full 14 days after we arrived on site and 16 days after Datamail’s recovery programme began. The results of the trials had to satisfy both the affected EOs and ourselves as independent assurance providers.

2.214
Once the enhanced processes and controls had been trialled and proven, the results of the other 21 organisations were cleared in the following 8 days.

2.215
Clearly, the Local Electoral Act 2001 did not envisage an independent party verifying the actual results of validly cast votes; nor did the public entities affected expect to have to ask for our involvement. Of necessity, that involvement added to the time taken to clear the election results.

2.216
However, the key influence on timing was the failure of a data processing system pivotal to determining the results. This failure occurred in the compressed time of an election weekend, and was exacerbated by the fact that it happened at the end of the voting period and not in the weeks leading up to 9 October.

2.217
Revised procedures and controls had to be introduced across 2 organisations, which meant a lengthy recovery period – much to the frustration of EOs, public entities, candidates, and public alike.

2.218
If existing processes and controls had operated accurately in the first place, there would have been no need for third-party intervention.

Our observations

2.219
The review of the 2004 local authority elections currently being undertaken by the Justice and Electoral Committee (refer following section) may deal with this issue in more detail.

2.220
While the Select Committee may identify ways to improve the staging of STV-based elections, the events we dealt with related to processes and controls. They did not concern the underlying foundation of an STV voting system.

2.221
Deriving the results of an STV-based election requires the use of computer systems, including an STV calculator. A comprehensive control environment is required to ensure that vote counting processes work as intended. Attention should be given to the processes and the controls over both aspects of the vote count – the physical handling of voting papers and the computer systems on which the count is reliant.

2.222
From our involvement in the events following 9 October, we conclude that:

• the roles and responsibilities of EOs and counting contractors should be clear, well understood, and formally documented;
• the complete “end-to-end” counting process should be formally documented; and
• information system controls over vote processing should be aligned with “best practice” in such areas as change control, data completeness checks, systems testing, and the identification (and recording) of contingency procedures.

2.223
As a corollary, EOs should:

• recognise that, in using a vote count contractor, they need a good understanding of that contractor’s approach in order to monitor the voting process;
• understand the key steps taken by that contractor, including receiving voting papers, their “transformation” into digitised data, the input of that data into the STV calculator, and reconciliation and control checking; and
• consider the need for, nature of, and extent of any verification of the proposed system before the vote count begins.

2.224
Timing is significantly affected if third parties have to become involved in the election process, either to resolve specific issues or to provide external confidence in the results finally released. Hence, it is important that all parties recognise the impact of a failure in the physical or information systems environment of an election. EOs and contractors should therefore consider the risks associated with the vote count, and implement strategies to reduce risk where necessary.

2.225
The heightened tension of the voting process requires good communication between EO and contractor. This allows any problems in the process to be addressed and the information needs of stakeholders to be met effectively, including inquiries from candidates, the public entity involved, media, and the public.

The Select Committee inquiry

2.226
The terms of reference for the current inquiry by the Justice and Electoral Committee extend well beyond issues associated with the failure of electionz.com and Datamail systems. They include a review of key elements of electoral systems, the impact of voting papers covering both FPP and STV-based elections, and the extent and adequacy of voter education.³

2.227
Committee members, however, have already met some of the key parties involved in the vote processing failure that delayed the 2004 election results. Their investigation should help to ensure that future STV-based elections do not encounter the same problems that local authorities and other public entities experienced last year.

2.228
We are aware that other interested parties, including the Electoral Working Party of the Society of Local Government Managers (SOLGM), are assessing the issues surrounding this matter.⁴ This may also help to ensure that the problems encountered do not recur.

2.229
The Office of the Auditor-General found itself in an unprecedented situation during the 2004 local authority elections, in performing the role of independent verifier of the vote count. It is our hope that the inquiry and reviews of the Select Committee and others will ensure that we have no need to repeat such a role.

1: The “STV calculator” takes the digitised data and, based on preferences cast, calculates the electoral winners. The number of votes not processed correctly varied from authority to authority. For background on the STV elections and some of the key concepts, see www.stv.govt.nz.

2: The operation of the STV calculator was not in question, and it was not our role to review it. It had been developed before the elections, and was subject to separate, independent certification as required by section 19AB of the Local Electoral Act 2001.

3: The full terms of reference for this inquiry are located at www.clerk.parliament.govt.nz/Programme/Committees/Submissions/je2004elections.htm.

4: We note that SOLGM has an existing publication Code of Good Practice for the Management of Local Authority Elections and Polls.