Part 4: The errors and why they happened

General Election 2023: Independent review of counting errors.

4.1
As outlined in Part 1, the focus of our work was on why the errors identified by the Electoral Commission occurred and why they were not identified by the processes designed to detect those errors.

4.2
Although there are some electronic processes for verifying information about voters and votes, and for collating information about votes, the process relies heavily on paper ballots, manual counting of ballots, manual data entry of results, manual checking of results, completing dual vote investigations, and manual extraction of votes that need to be removed from the count.

4.3
In part, this is because of the prescriptive processes set out in the Electoral Act 1993, which reflects the less technologically advanced society of the time the legislation was passed and amended (in 1956 and 1993 respectively). An electoral system that relies on manual processes to such a degree is vulnerable to mistakes, especially when people are doing manual processes when tired and under significant time pressure.

4.4
During the 2023 General Election, official result processes were completed on the morning the official election results were announced. This was because of delays in closing the electoral rolls and completing vote extractions and final result processes. Work that would normally be spread across two days was instead done in one morning.

4.5
Organisational culture, processes, and management decisions can create conditions in which errors are more likely to happen, where people are more likely to take shortcuts or ignore safeguards, and where controls may not operate as intended. Our review looked at the immediate causes of mistakes and the factors that we consider contributed to the mistakes going undetected.

4.6
In this Part, we describe:

What were the errors?

4.7
The errors in the official results arose because of votes being misplaced, data entry errors, and one instance of a ballot box being misplaced and not counted. Our review and subsequent investigations also identified that a number of apparent dual votes had not been removed from counts in keeping with the instructions given by the National Office.

4.8
Candidate vote errors were identified in 15 electorates (at 17 voting places) and party vote errors were identified in six electorates (at eight voting places):

  • A ballot box from the advance voting period was misplaced and not counted.
  • Ballot papers were incorrectly sorted during the official count and consequently not correctly counted.
  • Data entry errors resulted in votes being incorrectly allocated to candidates and parties.
  • A data entry error was made when advance voting results and election day results were transposed, resulting in errors affecting candidate and party vote counts.
  • A data entry error meant election day results were incorrectly entered as both advance voting and election day voting results, resulting in errors affecting candidate and party vote counts.
  • A data entry error meant special votes cast during advance voting were incorrectly entered as special votes cast on election day, resulting in errors affecting candidate and party vote counts and reducing the total number of votes.
  • A data entry error meant one more candidate special vote was entered than had been counted for the party vote, resulting in errors affecting candidate and party votes.

The process and instructions for managing ballot boxes were not followed

4.9
When this issue was first identified, media reports described the ballot box as having been "forgotten" or "left at the Commission's East Coast Office".26 We explain here in more detail what we understand happened.

4.10
The Electoral Commission Operations Manual advises that if a voting place expects to have more than 2000 ordinary electorate votes, it will need more than one ballot box. Each ballot box represents a "phase".

4.11
When a ballot box is removed and stored securely at electorate headquarters, the number of ballot papers used is reconciled against the ballot papers counted.

4.12
The Electoral Commission told us that for one voting place, two ballot boxes were prepared - one for advance voting and one for election day. During advance voting, both ballot boxes were deployed to the voting place at the same time, instead of just the one prepared for advance voting. The two ballot boxes were removed to electorate headquarters at the end of advance voting, and a third ballot box was then prepared and provided for election day. This box was labelled with the same details as had been attached to one of the ballot boxes used during advance voting.

4.13
Ballot papers from the two ballot boxes used during advance voting were counted during the early count on election day. The third ballot box was counted in the voting place after voting had closed. Late on election night, the error was identified when EMS (the system used to record the results) could not accept three separate ballot box counts because it had been set up to accept the results of only two ballot boxes. Verbal instructions were given to merge the two physical ballot boxes used during advance voting with all ballot papers from the first two boxes put into one ballot box, and for all documentation to be redone so there would be only one set of documentation and one result recorded for the count of advance votes. The count results were merged in EMS, but the two ballot boxes (with ballot papers) and the paperwork were not. We were told that the error was difficult to identify because the ballot boxes were not properly labelled.

4.14
During the official count, only one of the two ballot boxes that should have been merged was counted, and this was not picked up by the checks that should have occurred in the electorate or at the National Office. The check to compare votes counted during the official count against the early count was not done and the discrepancy of 611 votes was not identified. When this error was later identified and the votes were re-counted, a variation of nine votes was identified, increasing the discrepancy to 620.

4.15
Our review found that electorates had different ways of managing how ballot boxes were moved between secure storage areas and count rooms. This depended on the storage space available or the preference of the electorate manager.

4.16
The movement of ballot boxes between electorate headquarters and voting places is recorded in the materials tracking module in EMS. Count results are recorded in EMS, matched to specific ballot box information, but this is not reconciled with the materials tracking module. Doing this would help increase the likelihood that all ballot boxes issued to voting places are accounted for in the count results.

Quality assurance checks for the count were not done properly

4.17
As outlined in Part 3, the Electoral Commission had quality assurance checks in place. However, the checks did not pick up all the errors in the results. From the work we have done, we consider this is for two reasons:

  • the quality assurance checks were not done properly; and
  • the quality assurance checks were not as effective as they should have been.

4.18
For some errors, the electorate manager did not carry out checks to confirm that data was entered into EMS correctly, or assumed that someone else had done it without evidence to support this assumption.

4.19
Similarly, we were told that quality assurance checks carried out at the National Office did not detect errors because either the checks were not carried out or were not performed as rigorously as they should have been.

4.20
We were told that it is possible that some electorate managers placed undue reliance on the quality assurance checks that would be carried out at the National Office, while some National Office staff might have made assumptions about the rigour applied by electorate managers in their reasonableness checks. This resulted in a layer of quality assurance checks not operating as intended.

Quality assurance checks were insufficient

4.21
We reviewed the official results processes and controls. Effective controls are well designed, operate consistently, and prevent significant risks materialising. They need to be reviewed regularly to ensure that they remain effective. When controls to detect or prevent errors are manual, there is always a risk of human error. Reducing or eliminating the risk of human error in the results process requires a range of strategies.

4.22
We expected to see clearly documented quality assurance processes that described the checks that needed to take place, the sequence in which these checks should occur, and a system for proving that they had been done.

4.23
Our review identified vulnerabilities in current processes and controls. These vulnerabilities include manual processes (physical vote counting and manual data entry), with mitigating manual controls that are not designed effectively (data checking does not require anyone to verify that the data has in fact been checked against source material). Quality assurance checks are not operating effectively and rely on human judgement, and reasonableness checks are not well understood.

Result-checking processes were not well understood or well documented

4.24
During our interviews with electorate managers, we were told that results instructions were not consistently followed, and that data entry checks and reasonableness checks were not always carried out. We were told that this was, in part, because people did not fully understand what they were expected to do.

4.25
In our view, some of the processes and controls for recording results and for checking that data was correct were ineffective during the post-election period.

4.26
We were told that EMS reports used for reasonableness checks did not make it easy to identify errors quickly because they are large, complex reports that are difficult to interpret.

4.27
There was no system for documenting and verifying that data entry checks or reasonableness checks had been carried out. This meant that electorate managers and National Office staff assumed that checks had been done, when sometimes they had not, or had not been done as consistently, thoroughly, or extensively as they should have been.

4.28
The Electoral Commission relied on the experience, skill, and diligence of key electorate staff, such as electorate managers and post-election managers. However, the variability in skill and experience of staff in those roles affected how well processes were understood and implemented. This is likely to have reduced the effectiveness of manual controls in the results process.

4.29
The Electoral Commission has relied on individuals with prior election experience for checking the incoming electorate results. Within the National Office, electronic scripts had been used during previous elections to address challenges with the volume and complexity of information coming out of EMS and help staff to interrogate the results data more easily. An employee who knew how to do this left after the 2020 election, and the process used to interrogate data was not documented. In the 2023 General Election, the Commission did not have staff with the capability to design and deliver these steps, and this additional checking was not done. This created a gap in the effectiveness of the National Office's quality assurance checks, which was not fully understood at the time, and which was exacerbated by the limited time to complete the checks manually. The Electoral Commission told us it will seek to address the gap in electronic data analysis by adapting reporting from EMS.

4.30
We looked to see whether there was a clear and structured escalation and resolution process for potential errors identified by quality assurance checks in the National Office. Our enquiries found there was no structured and clear process for escalating matters and a lack of clarity about what sorts of problems should be escalated. There needed to be clarity about what was expected. Potential errors should be recorded and assigned, and how they have been resolved should be clearly documented. This would provide assurance to the Electoral Commission Board that matters affecting quality have been appropriately resolved. Reporting documented errors and resolution processes to the Board would have allowed a more systematic view on quality issues.

Key processes were not documented

4.31
A factor which contributed to the challenges experienced was that the Electoral Commission had not fully documented the quality assurance data checks that National Office staff were expected to perform. Without adequate documentation, it is difficult to see the end-to-end processes, identify controls to manage risks, identify inter-dependencies, or target assurance activities. When processes are not documented and experienced people are unavailable or stretched across several tasks at once, as occurred in the 2023 General Election, mistakes are more likely to be made and go undetected.

4.32
This is not the first time that this has been raised as a risk. A review of the 2020 General Election recommended that the Electoral Commission fully document these manual quality control processes. During our review, we identified a need for the Electoral Commission to complete the Voting Services' Process Manual to fully document the reasonableness checks (including data checks) that are supposed to be done by people in the National Office. The Commission has since said that this work has been completed.

There was no system to verify that checks had been done

4.33
The rigour and consistency applied to the physical count of ballot papers did not extend to the process of recording the official count results.

4.34
Some electorate managers told us that they signed off final result certificates for their electorates, indicating that reasonableness checks had been completed, without doing the checks or verifying that they had been done. This was inconsistent with guidance in the Electoral Commission Operations Manual. There was no checklist to record what had been done.

4.35
Similarly, there was no system at the National Office to record that reasonableness checks had been done.

4.36
The Electoral Commission's executive leadership team and the Board were given verbal assurance on 3 November that official result processes and quality assurance checks were complete. However, count errors had gone undetected at the electorate level and were not resolved by the National Office before official results were finalised.

Our recommendations for improving vote counting and checking processes

4.37
Processes need to be standardised with detailed but clear standard operating procedures. Process implementation and resulting error rates need to be monitored to ensure that processes are consistently followed and quality is not compromised. This means that errors can be detected early and promptly corrected.

4.38
In our view, the Electoral Commission needs more effective processes to ensure that tasks are completed as expected. It also needs more supporting guidance for tasks, especially those that are complex or rely on judgement. This would help ensure that tasks are well understood and that all necessary steps are taken, which in turn can reduce the likelihood of errors.

4.39
We consider it is important to continue to separate key activities such as data entry and data entry checking. Different count teams at voting places and electorate headquarters should also be maintained and "one-up reviews" for key tasks should continue. For example, an electorate manager should not both check that the data has been entered correctly and carry out quality assurance checks. Fresh eyes are more likely to pick up mistakes. This needs to be supported by accountability measures to provide evidence that tasks have been completed (either workflow approval or documented evidence). Double-entry of data, followed by a check (manual or automated) could also be considered to provide additional assurance.

Technology could be used to strengthen the vote counting process

4.40
For repetitive, high-volume tasks, data validation can improve quality and reliability by reducing or preventing errors, saving time, enhancing the user experience, and reducing fatigue.

4.41
There are likely many technology-enabled options to support or replace the manual counting of votes, but the risks associated with these options and the research to support the use of these technologies was beyond the scope of this review. The Electoral Commission could consider whether technology (with appropriate security controls) could provide additional robustness to the manual vote counting process. Using technology to automate repetitive or manual tasks can help reduce the likelihood of errors caused by human oversight or fatigue. The technology could include, for example:

  • scanning ballot papers received, replacing the need for a physical count. This would enable voting information to be recorded and votes to be totalled, reducing balancing and other data entry errors; and
  • scanning count results into the electronic record system rather than manually entering them. This would not replace the physical count, but would replace the data entry process.

4.42
Technology like this can be useful when processes involve significant time constraints and large volumes of information or data. We recognise that this would have resourcing implications and that the Electoral Act 1993 would likely need to be amended to allow this.

4.43
When full automation is not possible, technology-enabled controls can support complex manual tasks (like quality assurance checks) and reduce errors. Examples of controls include having the right information accessible in the right format, systems that are user-friendly and easy to navigate, and/or built in data validation.

4.44
The Electoral Commission could also explore enhancing EMS to make it easier to identify more obvious errors, such as when duplicate results are entered for election day and advance voting. During the Port Waikato By-election, the Electoral Commission piloted a new, more user-friendly Electorate Detail Report to identify potential duplicate data entries.

Continuous improvement

4.45
In our view, as part of continuous improvement, the Electoral Commission needs to analyse errors, identify gaps or vulnerabilities in systems, processes, and controls, and prepare an improvement plan to address them before the next election.

4.46
In the Port Waikato By-election, the Electoral Commission trialled a new checklist to guide the electorate manager through the key checks that needed to be carried out. The Port Waikato electorate manager provided feedback that the checklist worked well.

4.47
Although the checklist worked well in the by-election, a by-election is typically much less complex than a general election. The Electoral Commission could consider how processes might be implemented more broadly in the next general election.

4.48
The Chief Electoral Officer told us that the Electoral Commission has now mapped key election processes. We recommend extending this to identify automated and manual controls that manage risks in the processes, and testing these for effectiveness.

Assurance

4.49
It is important to have a formal process to provide the executive leadership team and the Electoral Commission Board with assurance that all exceptions identified are either below the error tolerance or have been resolved. There was no such formal process during the official count process in the 2023 General Election.

Recommendation 1
We recommend that the Electoral Commission review all vote counting and quality assurance checks and controls to address gaps and vulnerabilities.

Enrolment and voting patterns and events put pressure on processes and people

4.50
Humans make mistakes. Enrolment and voting patterns and events in the 2023 General Election created extra workload and delays in processing enrolments and votes. This in turn put pressure on the time available to carry out post-election checks. In our view, these escalating pressures created a situation where mistakes were more likely to occur and to go undetected.

Special votes and enrolments during the voting period were higher than expected

4.51
As we describe in Part 2, in 2023 fewer people voted early and significantly more people chose to enrol and vote closer to election day. More people cast special votes than ever before, and special votes take significantly longer to process. This extra work exceeded the Electoral Commission's projections that had informed the number of staff it employed, meaning there was not enough staff to keep up with enrolment processing.

4.52
During the 2023 General Election, election workers used an electronic roll to look up voters to direct them to the appropriate ordinary vote or special vote queue and to update enrolment details for voters changing their address within the same electorate. The electronic roll had two outages on election day, and some election workers were uncertain about what to do. Although this might not have been a significant factor, we were told that when the outages happened some election workers directed people to the special vote queue. They were asked to cast a special vote when this might not have been necessary.

4.53
We were told that the increased number of special votes, combined with delays completing enrolment processing, resulted in late apparent dual votes that needed to be investigated. As we described in Part 3, special votes and dual vote investigations are time-consuming to sort, process, and count. The high number of apparent dual votes being identified late put pressure on completing all post-election processes that are meant to follow the dual vote investigations. Instead, some had to happen concurrently rather than sequentially (see Figure 2). This also delayed clearing the apparent dual votes report, which we discuss in more detail in paragraphs 4.61 to 4.74. The official results should not be finalised until these processes have been completed and votes that are not eligible to be counted have been extracted.

4.54
Another factor contributing to quality assurance checks not being done properly in the post-election period arose due to mistakes made when reconciling ballots issued against ballots returned. The information submitted from voting places through the reconciliation application is worked out as a cumulative total. This information is provided in a report to the electorate headquarters, recording ballot papers issued for only that day. There was some inconsistency in what was reported from voting places. This meant that electorate headquarters staff spent more time trying to verify the accuracy of the reconciliation information to ensure that ballot paper totals were correct. We were told that in the post-election period, this became a key focus of the quality assurance checks undertaken, potentially reducing the time available to carry out quality assurance checks on the accuracy of other results.

Enrolment processing took longer than expected

4.55
About 143,000 more people enrolled or made changes to their enrolment details in the advance voting period and on election day in 2023 than in 2020 (a 46% increase). These enrolment activities also took place later in the voting period than in the previous election. The Electoral Commission overestimated the proportion of digital enrolments. Only 47% of enrolments were digital and 53% were paper enrolments, which take longer to process. The Commission had allocated five days after the election for processing enrolments and responding to queries about voter eligibility, but this took 11 days.

4.56
For the first time, enrolment processing took place in three centres instead of being spread across regional offices. The Electoral Commission had 22 permanent staff and 130 additional processing staff in Auckland, Wellington, and Christchurch. At peak, they processed about 20,000 enrolments a day. We were told that with hindsight, although processing teams might have been larger in these centres, there had been a loss of critical experience that would have been helpful with the volume of work that eventuated.

4.57
The Electoral Commission did redeploy 10 trained staff to help with enrolment processing. Although extra temporary staff were available to be called on to do enrolment processing, they were not considered suitably trained. When the delays in enrolment processing became apparent, the enrolment team did not have capacity to train extra staff at short notice.

4.58
We were told that the enrolment team did not have good visibility of incoming enrolment volumes from voting places and electorates. We note that this data was recorded as part of the daily reconciliation process at voting places (for advance voting and on election day) and entered into EMS. This information could be viewed in a dashboard, but the Electoral Commission told us that the dashboard understated demand by about 6%.

4.59
Despite this information being available and a potentially useful indicator of incoming enrolment volumes, we were told that the enrolment team did not use it to understand and manage the flow of work. The enrolment team used the numbers from the enrolment forms scanned and digitally entered as these came into the processing system. The Electoral Commission told us that some electorates did not scan and upload enrolment forms in a timely way.

4.60
We question whether the Electoral Commission was quick enough to use the tools at its disposal to identify the increase in enrolments and special votes, and to deploy extra resources where they were needed. We were told that it was apparent to staff early on that enrolment volumes were high and that this would affect other processes. That knowledge does not appear to have been widely communicated or the implications understood.

A condensed timeframe to complete post-election processes

4.61
The increased volume of enrolments and the delay in closing the electoral rolls condensed the time available to complete all post-election processes into the third week of the post-election period.27 Those post-election processes were supposed to happen in the second week, so the delay further limited the time needed to respond to any issues that arose when completing third week tasks, such as conducting reasonableness checks across all results and confirming, finalising, and publishing the results.

4.62
As a result, at 5.22pm on the day before the official results were announced, the Electoral Commission issued instructions to electorates to:

  • make a decision about apparent dual votes based on the information held and extract any votes that could not be confirmed, as is required under section 176 of the Electoral Act, so that the final status of each voter's vote could be entered into the electronic system used to compile the master roll; and
  • apply the qualification status of the special votes returned from the Registrars of Electors or enrolment team so that votes could be either included or excluded from the special vote count.

4.63
The latter instruction was needed as different parts of the Electoral Commission were working together to resolve clerical errors, and there was no further time to solve all remaining items.

4.64
Although there were working documents that recorded some of the queries and work completed within the Electoral Commission, there is no system to record or track how many special vote decisions were questioned by electorate managers or what changes were made after these discussions.

4.65
The instruction to resolve remaining apparent dual votes meant that decisions needed to be made based on the information available at the time. The Voting Services team at the Electoral Commission had a record of the apparent dual votes that were being resolved when the instruction was issued. Many apparent dual votes are found to be due to clerical errors, such as the same voter being accidentally marked off an electoral roll twice. Apparent dual vote investigations are complex and time-consuming to complete. In some electorates, due to the shortened timeframes, not all investigations could be completed before the deadline.

4.66
Section 176 of the Electoral Act requires any apparent dual vote to be removed, so the Electoral Commission's instructions to rely only on the information from the enrolment team and the master roll does not seem unreasonable in the circumstances.

4.67
However, the instructions were not completely followed, and the Electoral Commission was not aware of this at the time. Although Voting Services had a monitoring process, it did not identify that some apparent dual votes had not been extracted.

4.68
The Electoral Commission told us that electorate managers ran out of time to complete dual vote investigations and did not extract remaining apparent dual votes that had not been resolved by the deadline. Some electorate managers might have extracted dual votes that they had resolved but were unable to update the records after the deadline.

4.69
When official results were announced, there were 892 apparent dual votes on the apparent dual vote reports for all electorates that might not have been resolved and extracted in keeping with the instructions given. In electorates where the vote count margins were narrow, this could have affected the candidate result. Further investigation by the Electoral Commission provided more certainty and reduced the numbers that might have affected the candidate result.28

4.70
The Electoral Commission told us that even if the instruction to remove the apparent dual votes had been followed completely, it would not have resulted in a different outcome in any electorates. The Commission has notified the Chief District Court Judge and the three judges who oversaw judicial recounts and provided a briefing. The judges acknowledged the information provided. They noted that there was nothing they could have done differently based on the information given to them at the time of the recount, that they no longer had a role to play, and that they have not asked the Commission to take further action. The Commission has also notified the Minister and the Secretary for Justice.

4.71
The shortened timeframe to complete dual vote investigations was unprecedented, and there was no standard operating procedure to check that the dual vote investigations had been resolved. This introduced a risk that could have undermined the integrity of the official results.

4.72
Several electorates and the team processing votes received from overseas completed vote extractions and data entry just before midnight on Thursday 2 November, the night before official count results were announced. The extraction process is manual, and involves election workers identifying ballots that are ineligible to be counted and physically removing them. This was done late at night under great time pressure.

4.73
This in turn put significant pressure on the remaining reasonableness checks29 that needed to happen before the official count was finalised. A final quality assurance check that would usually be completed over two days had to be completed within hours. In our view, these factors contributed to election count processes and quality assurance checks failing to detect and prevent the errors in the official results.

4.74
We acknowledge that there are several prescriptive legislative requirements that dictate timing (return of the writ) and process for an election. In this instance, because there was a statutory deadline and fixed resources, the trade-off between cost, quality, and timeliness was made in favour of timeliness. The Electoral Commission needs early information to be able to make decisions on the types of trade-offs it may be facing, and to communicate this.

Other systemic factors might have contributed to the errors

4.75
In the course of our work, we also observed some systemic factors that might have contributed to the errors made. We have outlined these and made some wider recommendations for the Electoral Commission to consider.

Guidance could be improved

4.76
There were manuals and other guidance for Electoral Commission staff and election workers to help them run the election and to count the votes. The Commission also had a comprehensive programme for training staff. However, aspects of the guidance and training could be improved.

4.77
As already noted, the National Office did not have its own processes for the vote count fully documented at the time of the election. The Electoral Commission told us that it has now revised and completed its Voting Services Process Manual. Keeping resources up to date and fit for purpose can help ensure that processes are well understood and followed consistently.

4.78
Detailed but clear operating manuals will reduce human error and ensure that individuals understand what they need to do, who they are accountable to, and what they are responsible for. This is particularly important for the Electoral Commission because turnover for electorate and National Office staff is high. There are large numbers of staff involved in the count, and they need to understand their roles and who is accountable for quality checks.

4.79
Role-specific election processes for election workers are documented in a suite of operational manuals and personal instructional manuals. The Electoral Commission might want to consider whether improvements could be made to the language and format of manuals that are used by a largely temporary workforce with considerable variation in educational and work backgrounds. Consolidating and making the materials available digitally for voting places and electorate headquarters might make it easier for people to locate the specific guidance they are looking for.

4.80
We consider there is benefit in continuously reinforcing why people are being asked to do certain tasks, through training, manuals, and briefings. In our view, they need to better understand the potential impact on the integrity of the count if tasks are not carried out.

4.81
The Electoral Commission told us that it needs to update its operational manuals to:

  • ensure that the authority for certain tasks is clear and appropriate;
  • explain the importance of quality assurance checks and how electorate managers should carry them out; and
  • provide more guidance on data entry checking and troubleshooting (for example, to cover an eRoll outage).

Training on post-election processes needs to be reviewed

4.82
Training helps to ensure that people understand the election processes and what is expected of them in their role. We looked at the training that election workers were given. There was some variation in the training format and delivery.30

4.83
The Electoral Commission provided election worker training in a variety of ways, including in-person training and self-directed online learning. Training was staggered at different times for different cohorts of regional staff.

4.84
Electorate managers received their initial training in February 2023. We were told that a lot of information was provided in a short time, and feedback surveys were generally positive. One electorate manager commented that it would have been good to have more training on the key roles that they would need to recruit for and what specific skills would be needed.

4.85
Regional trainers helped train large numbers of short-term staff, including advance and election day voting place staff, count staff, and those involved in scrutiny processes. More comprehensive training was provided to logistics managers, recruitment and rostering managers, official count process leads, and post-election managers. Regional managers and regional advisors also provided on-the-job coaching.

4.86
Within electorates, training could be variable. One electorate manager told us that evening training sessions for election workers were carried out in electorates based in cities, but this was not feasible in remote rural electorates. We were also told that some people were not recruited in time for training sessions and only had access to online materials.

4.87
Electorate managers had three days' training on the post-election period in September 2023. This included online training, PowerPoint presentations, and practical exercises. Training is best delivered closest to the time when it will be needed. However, we were told by some electorate managers that the post-election training was too close to the election when they were focused on preparing for election day.

4.88
People responsible for specific post-election processes completed the same online training, and in-person presentations, and practical exercises as the electorate managers. People processing special votes had a half-day's training and people responsible for counting ordinary votes were given less training – about three hours. All were given a manual to read, and started work a few days before the post-election period to prepare for their role.

4.89
Some electorate managers told us that, with hindsight, they did not feel as well prepared for the post-election period as they were for the pre-election period and election day, and that more training on post-election processes would have been beneficial. One electorate manager commented that on election day, "the penny dropped that the election goes beyond the election day." Another electorate manager commented to us that they would have liked more specific training on how to do reasonableness checks, how to manage count tallies, and how to prioritise post-election activities.

4.90
We consider that particular thought should be given to the adequacy of training on post-election processes. People in key roles need to be supported with sufficient, targeted, and timely training to ensure that they have the knowledge to fulfil their role. Key messages should be re-enforced, such as accuracy over speed. Other support mechanisms might need to be put in place, particularly for people who have not worked in elections previously.

Recommendation 2
We recommend that the Electoral Commission review and update standard operations manuals and instructions, to improve the clarity of information about quality control activities and why they are important and to clarify accountability and responsibility for carrying them out.
Recommendation 3
We recommend that the Electoral Commission complete the end-to-end description of the election process and inter-dependencies of activities, and identify controls that support the election process.

It was difficult to recruit election workers

4.91
The Electoral Commission recruited about 22,000 people to work during the election period. This is a significant undertaking. Traditionally, the Commission has relied on retirees and students, who have the flexibility to do temporary work. We heard from regional and electorate managers and staff in the National Office that it was difficult to recruit enough workers. We were told that the temporary nature of the work, the modest pay, and the timing of the election might have affected recruitment. Some electorates were harder to staff than others; South Auckland and lower North Island electorates had more vacancies than other electorates.

4.92
The Electoral Commission told us that it assessed recruitment needs by using dashboard reports from the recruitment system combined with feedback from electorate managers about how recruitment was progressing. The Commission said that where gaps were identified, it would deploy National Office staff to assist. However, we did not see a comprehensive enough process for senior managers or the Board to quickly identify staffing gaps or the actions taken to resolve them.

4.93
The Electoral Commission introduced an online recruitment system in late May 2023. The Commission told us that this system was initially used internally and was expanded to manage external recruitment for the election.

4.94
The online recruitment system did reduce some recruitment risks and manual steps, for example by allowing the Electoral Commission to easily do Ministry of Justice criminal record checks on all applicants. It was also used for other employment administrative tasks, as well as to enable access to training and other work systems (although there were some challenges with this).

4.95
Some users told us that the design of the online recruitment system was too complicated and time consuming. Some National Office managers, electorate managers, and regional managers that we spoke to told us that some potential applicants for the 2023 General Election found the registration of interest and subsequent application process confusing. We were told that this could have discouraged some people from applying and might have particularly disadvantaged older applicants, who can be less familiar with using online application processes.

4.96
The timeframe needed to attract candidates and complete the full recruitment process meant people were recruited later than originally planned, which delayed their access to the online training and learning materials.

4.97
The online recruitment system was integrated with some other Electoral Commission systems. Permissions depended on the access initially given to a person, depending on the role they were appointed to. The intention was to grant the access that the Electoral Commission anticipated that people would need across different information technology systems.

4.98
Some people worked in more than one role during the election period or were delegated duties that were not originally envisaged as part of the role they were initially hired for. This created challenges and delays in arranging access to the systems they needed to use for both training and work activities.

4.99
The integration of the recruitment system and other information technology systems did not anticipate subgroups of user needs, and extra steps were needed to provide the access that people needed. This meant that some staff were unable to access eLearning before face-to-face training, either due to the extra time to get through the online recruitment process or delays in gaining access.

4.100
The Electoral Commission's recruitment strategy for key roles has to be designed to employ the right people with the right skills to the right place at the right time. Regional managers told us that they had difficulty recruiting electorate managers with prior experience and recruiting appropriately experienced people to other critical roles, such as post-election managers and official count process leads. This was particularly so in rural electorates.

4.101
From what we understand about post-election processes, the work requires good attention to detail. Some electorate managers said that, with hindsight, they wished they had understood earlier the type of skills that would be called for when hiring post-election managers and official count process leads.

4.102
Thought should be given to the timing of recruitment for post-election managers and official count process leads. Electorate managers would benefit from receiving guidance about the skills required for these roles which differ from pre-election and election day work.

Recommendation 4
We recommend that the Electoral Commission review the personnel requirements for elections, the process for recruiting and training election workers, and planning for contingencies (such as staff unavailability, system outages, and fatigue).

People worked long hours

4.103
We were told that election staff were working long hours for extended periods in the lead-up to election day and in the post-election period. The Electoral Commission told us it had a Health Safety and Wellbeing Lead and had mitigation steps in place to support staff, including, for example, mandatory dinner breaks with food provided. Despite this, people still ended up working long hours. From election day, enrolment teams were working 12 to 14 hours seven days a week to process enrolments and do enrolment checks. One electorate manager reported that they had been working for 17 hours by the time they had to carry out their quality assurance checks, and that other people had been working 70-hour weeks for three months. Tired people are more likely to make mistakes.

4.104
We were told that National Office staff were also stretched during the official count period. Four staff were involved in completing quality assurance checks while also providing field support and preparing for the Port Waikato By-election. The need for the Port Waikato By-election shortly after the 2023 General Election meant the Electoral Commission had to plan the by-election and make changes to count systems and reports at a critical time in the election period. Not being able to separately resource and manage the by-election reduced the attention that staff were able to give to quality assurance checks.

4.105
Staff fatigue, particularly in the post-election period, is a significant operational risk that the Electoral Commission needs to manage carefully. Specific consideration should be given to rostering to ensure that staff responsible for final count and quality assurance checks are adequately rested. This might require building in additional resource and quality assurance for the post-election period.

4.106
It is important for the Electoral Commission to have contingency plans that enable electorates and the National Office to "scale up" quickly, cover absences or staff shortages, and manage fatigue over an extended election period.

4.107
In our view, the Electoral Commission needs to carefully review resourcing to ensure that enrolment and post-election work can be completed in a timely way and avoid pressure on the later processes that are safeguards to providing a correct official count.

There were challenges with information technology equipment and systems

4.108
We were also told about challenges accessing enough information technology hardware and systems to do the work efficiently and within timeframes.

4.109
In particular, several electorate managers said that there were not enough laptops for electorate headquarters staff. This caused delay and inefficiency, which in turn affected processing times for official count activities.

4.110
The complexity of the recruitment system and challenges in the induction process, particularly for staff working in more than one role, resulted in people sharing passwords to access the systems they needed to complete their work. People sharing passwords and requesting blanket access to multiple systems to work around permissions issues introduced other security risks.

4.111
Improvements could be made to information technology systems to better manage material handling, particularly ballot boxes. Currently, tracking ballot boxes is not integrated into EMS. Integrating these systems would be one way to ensure that all ballot boxes issued to voting places are tracked and counted.

4.112
It is vital that the Electoral Commission's systems are stable, robust, and secure. We have been told that two of the Commission's three systems that are critical to support running elections are at "end of life" and that it is difficult and expensive to maintain and make changes to them. We understand that work has been commissioned to better understand the remaining capability of these systems.

4.113
We encourage the Electoral Commission to consider its information technology as part of assessing organisational and election risks. As part of this, the Commission should consider potential security risks associated with how the various systems, including the online recruitment system, are configured.

Recommendation 5
We recommend that the Electoral Commission review hardware requirements, to ensure that electorates have sufficient technology to complete tasks required of them.
Recommendation 6
We recommend that the Electoral Commission review the information technology systems that support the election process to ensure that they remain fit for purpose for both electorate and National Office functions.

26: McConnell, G (2023), "Electoral Commission says a ballot box was missed in official count", at stuff.co.nz.

27: Post-election processes included scrutinising eligibility to vote, processing special votes, investigating apparent dual votes, extracting post-writ changes (such as ineligible voters, those eligible for party vote only), and completing the count of all ordinary and special votes.

28: The Electoral Commission has carried out further analysis which identified 321 apparent dual votes that could have affected candidate counts across all electorates. There was one electorate where there was a small margin between the leading and second candidate (even after judicial recount), and where the apparent dual votes report showed that up to 60 apparent dual votes might not have been removed. This was of particular concern to us. Subsequent analysis by the Commission has identified that, at most, 24 of these votes may have impacted the candidate vote.

29: We describe these in paragraphs 3.33-3.41 and 4.29.

30: The format and delivery varied depending on the features of that electorate - such as the time it would take to travel to attend training in a geographically spread-out electorate compared with an urban electorate, and when
people were recruited.