New Zealand 2023 General Election

The Auditor-General and the Justice Committee both record that the errors came to light when a journalist queried the accuracy of the party vote after the 3 November 2023 declaration; the Commission's own quality assurance had not detected them. The amended results of 9 November 2023 listed three voting places with party-vote data entry errors, 15 with candidate-vote errors, one electorate where special-vote data was entered incorrectly and five voting places that recorded election-day votes as advance votes. Corrections added 693 party votes and 708 candidate votes, taking the party-vote total from 2,850,527 to 2,851,220. The Justice Committee records that the Auditor-General's review considered the 22 errors the Commission had by then identified.

Two advance-voting ballot boxes were merged in the results system on election night but not physically merged, and the paperwork was not reconciled. During the official count only one of them was counted. The check that compares the official count against the election-night count was not done, so a discrepancy of 611 votes went unnoticed; on recounting, the figure rose to 620.

Electorate checks were either not done or not done with the required rigour, and some electorate managers signed final result certificates without completing or verifying the reasonableness checks. There was no system to evidence that checks had been done, no fully documented standard operating procedures for the National Office steps, and a final quality assurance process that would normally take two days was completed in a few hours on the morning the result was announced.

Section 176 of the Electoral Act 1993 requires apparent dual votes to be removed. On the evening before the declaration the Commission told electorate staff to resolve outstanding cases on the information available and extract the rest. When the official results were announced, 892 apparent dual votes remained on electorate reports that might not have been handled as instructed. Because the Commission was unaware, the issue was not before the judges at the judicial recounts; it later briefed the Chief District Court Judge and the three recount judges.

The marae's chief executive was a candidate in the election. The Commission stated it was aware of the candidate's involvement with marae management during venue selection, obtained an agreement that the candidate would enter the voting place only to vote, assigned a designated contact person and briefed marae staff on the election advertising rules. The Committee nonetheless heard evidence that the venue was not a neutral polling place and that its alignment with a particular party and candidate was apparent, which it said the offering of food linked to voting exacerbated. It recommended the Government encourage the Commission to better monitor connections between polling sites and candidates, avoid premises with direct connections to candidates or parties, and consider requiring polling sites to be policy- and viewpoint-neutral. The Commission plans to review its guidance on political neutrality in selecting voting places, and the Committee agreed that review is necessary.

The New Zealand Social Media Study, led by Mona Krewel at Te Herenga Waka Victoria University of Wellington, fact-checked more than 4,000 Facebook posts by parties and party leaders across the five weeks before election day. It reports that fake news averaged 2.6% of posts per week against 2.5% in 2020, peaking at 3.8% in the final week, while half-truths rose from a 2.5% weekly average in 2020 to 4.8% in 2023, peaking at 6.5%. The tier here is set by the sourcing rule rather than by doubt about the study: the figures rest on a single research group's own account of its unpublished coding, with no second study or official finding pointing the same way.

The Justice Committee records that during the 2023 election cycle the Commission worked with online social media providers and established escalation mechanisms to report breaches of the election advertising rules and to report misinformation or disinformation about the election process. The Commission corrects and, if necessary, escalates concerns about false information on enrolling or voting, and at this election published videos and information addressing matters it expected to be the subject of incorrect information. Section 199A of the Electoral Act makes it an offence and a corrupt practice to knowingly publish false statements intended to influence voters on polling day or the two days before; the Commission received two complaints under it for the 2023 election, neither of which required further action. The Committee recommended extending the prohibition to the entire voting period, given the growth in advance voting.

On 21 November 2023 the Commission referred three individuals to Police for enrolling and voting while on the Corrupt Practices List, which disqualifies a person from enrolling or voting for three years after conviction for a corrupt practice; it referred a fourth on 12 December 2023. It cited section 118 (false statement for enrolment purposes) and section 227 (voting while on the list) of the Electoral Act 1993. Separately, on 27 September 2023 it referred Rhema Media to Police over election programmes broadcast on radio outside the election period under section 80I of the Broadcasting Act, and made a further referral over signs promoting Te Pati Maori published without a promoter statement under section 204F of the Electoral Act. In every case the Commission stated the matter was with Police and that it would not comment further; the cited sources record referrals of alleged breaches, not findings of guilt.

The Commission stated this when it reissued the results on 9 November 2023, the Auditor-General confirmed it in his May 2024 review, and the Justice Committee recorded it in November 2024. Turnout was unchanged at 78.2%.

Applications for judicial recounts in Nelson, Mt Albert and Tamaki Makaurau were lodged on 8 November 2023. The Nelson decision issued on 10 November and the Mt Albert and Tamaki Makaurau decisions on 15 November. In each case the count was carried out by counting staff in the presence of the Judge, the Returning Officer, scrutineers appointed by the candidates and Electoral Commission staff. Each judgment records minor variations in the numbers but no overall difference to the outcome, and each Judge made an amended declaration of the result under section 180(10) of the Electoral Act 1993. That includes Tamaki Makaurau, where the official count had left a margin of 4 votes.

The Commission's further analysis, reported by the Auditor-General, identified 321 apparent dual votes that could have affected candidate counts across all electorates, and in the one narrow-margin electorate that concerned the auditors it concluded that at most 24 votes may have been involved. No independent recount of those electorates was carried out on this ground, and the recount judges said there was nothing they could have done differently on the information they had.

A submitter to the Justice Committee maintained that the marae breached the treating rules in section 217 of the Electoral Act, which applies to giving food, drink or entertainment for the purpose of corruptly influencing how a person votes. Another submitter argued the offence should be removed as disproportionately affecting Maori. The Commission received complaints about the provision of food and other activities while the voting place was operating; it had advised marae management that food and activities could be provided to everyone coming to the marae as manaakitanga, and on the information it received did not consider the test for treating was met. Some members of the Committee recorded that they found the Commission's stance unconvincing. The Committee noted that the offence and the case law do not let the Commission give the certainty parties and candidates want, and recommended the offence be reviewed for clarity. No prosecution is recorded in the cited sources.

The Auditor-General traced the identified errors to misplaced ballot papers, manual data entry mistakes and checks that were not performed, and stated that his findings should not be taken as criticism of any individuals. He described a process that relies heavily on paper ballots, manual counting, manual data entry and manual checking, with electronic processes used to verify voter information and to collate and record results in the Commission's election management system. His review examined quality assurance, not any allegation of intent, so it cannot speak to matters outside that scope.

The New Zealand Social Media Study coded Facebook posts published by parties and party leaders. It does not measure reach or persuasion, does not cover other platforms or non-party accounts, and says nothing about the administration of the vote. The Justice Committee heard submitters both urging stronger action on disinformation and warning that such action could impinge on free speech, but recorded no measurement of any effect on voting. No fetched source documents specific false claims about the 2023 count or any Commission rebuttal of them.

The Commission's practice on referral is to state that the matter is with Police and to make no further comment, so its own releases carry the allegation but not any outcome. Neither the Auditor-General's review nor the Justice Committee's inquiry reports the disposition of the 2023 referrals. The releases do not say which electorates the four individuals voted in, so the referrals cannot be related to any particular electorate margin.