ARIN's Two Turnout Numbers and the Cost of an Inconsistent Record

Summary

• ARIN's archived 2024 results repeatedly report participation by 862 General Members through 702 Voting Contacts and calculate 43.91 percent turnout. The same page later reports that 959 General Members cast ballots through 789 Voting Contacts. Both sets cannot describe the same undifferentiated act of participation.
• The smaller denominator is internally consistent with the candidate percentages: Peter Harrison's 401 votes equal 46.52 percent of 862. The larger figure may represent a broader ballot state or distinct election event, but ARIN does not label the difference, publish a reconciliation, or identify which 97 organisations and 87 contacts enter only the second total.
• Nothing in the discrepancy alone establishes a wrong winner, fraudulent ballot or missed quorum. It does establish a publication-control failure. ARIN should preserve a versioned election ledger, define every counting state, reconcile organisation and contact totals, and attach corrections without silently replacing the historical record.

One page, two accounts of participation

ARIN's archived announcement of 7 November 2024 appears designed to close the election record. It names the winners for the Board of Trustees, Advisory Council and Number Resource Organization Number Council. It lists candidate totals and percentages. It describes the nomination and election process, says all online votes were confirmed, records that quorum was achieved and states that the process and results were verified under the bylaws. This is the page a member, journalist, researcher or future board should be able to treat as the authoritative account.

The page nevertheless gives two incompatible participation totals. Beneath each contest it says: "Voting Participation: 862 ARIN General Members (via 702 Voting Contacts); 43.91% of total eligible General Members." Farther down, a section headed "2024 Voter Statistics" says there were 1,963 eligible organisations as of 17 October and that active participation consisted of 959 General Members casting ballots through 789 Voting Contacts. The difference is 97 member organisations and 87 contacts.

This is not a rounding issue. Eight hundred and sixty-two divided by 1,963 is 43.91 percent, so the displayed percentage belongs to the smaller figure. Nine hundred and fifty-nine divided by the same eligible population is 48.85 percent. A reader cannot make the two totals converge by applying ordinary rounding, removing a fraction of a vote or changing the presentation of a candidate tie.

It is possible that both numbers originated in real system states. Some organisations may have opened or submitted a blank ballot that counted for membership participation but not for a contest. The NRO NC contest may have had a different eligible population. A corrected export may have arrived after the contest tables were prepared. But those are hypotheses. ARIN's page does not assign either definition to either figure. The record therefore fails at the point where verification should begin: it does not say what each number counts.

Candidate percentages reveal one denominator

The contest tables allow a limited reconstruction. Peter Harrison received 401 votes in the Board election and the page displays 46.52 percent. Dividing 401 by 862 produces 46.52 percent when rounded to two decimals. Ron da Silva's 344 votes become 39.91 percent on the same base. Chris Tacit's 307 become 35.61 percent. The Advisory Council percentages also use 862: Kathleen Hunter's 574 equals 66.59 percent, and Chris Woodfield's 516 equals 59.86 percent.

That consistency matters. It shows that 862 was not merely a stray sentence copied beside one contest. It was the denominator used to calculate every published candidate percentage. The Board and Advisory Council tables therefore describe a population of 862 participating member organisations, not 959. The NRO NC table lists totals and percentages that likewise use 862.

Candidate percentages in a multi-seat approval election are not shares that sum to 100. Each participating organisation can support more than one candidate, so a candidate's percentage measures the fraction of the participation denominator that selected that candidate. This design makes the denominator especially important. Without it, the percentage looks like a vote share when it is actually a reach measure across organisational ballots.

The arithmetic does not tell us why 959 appears later. It does tell us which total generated the published result table. Any reconciliation must start there. If 959 organisations submitted some kind of ballot, ARIN needs to explain why 97 were excluded from every candidate-percentage denominator. If 959 was erroneous, the later statistics block needs correction. If the two figures refer to different election instruments, each instrument needs a name, eligibility rule and count.

This is the useful boundary of independent calculation. Public arithmetic can test internal consistency; it cannot inspect the private election database. A credible institution should not force outsiders to guess the missing data model.

A Voting Contact is not a voting organisation

ARIN's record also contains two contact totals: 702 and 789. The distinction between contacts and organisations is legitimate. ARIN gives the vote to an eligible General Member organisation. That organisation acts through a designated Voting Contact. One person can serve as Voting Contact for more than one organisation, and therefore submit separate organisational ballots. Counting people and counting member organisations answers different questions.

The smaller pair implies that 702 contacts represented 862 organisations, an average of about 1.23 organisations per contact. The larger pair implies 789 contacts represented 959 organisations, an average of about 1.22. The ratios are close, but similarity does not make the totals interchangeable. It suggests that the unexplained addition is distributed across both units rather than caused by one contact suddenly representing 97 organisations.

This distinction protects against two opposite errors. Counting only contacts understates the formal electorate when one authorised person lawfully represents multiple legal members. Counting only organisations can overstate the diversity of human participation when corporate groups or service providers centralise voting authority. Both measures belong in a complete report, alongside a distribution showing how many contacts represented one, two, three or more organisations without identifying individual voters.

ARIN's page deserves credit for publishing both units. The failure is that it publishes two versions of each. Because organisation and contact totals move together, the discrepancy is unlikely to be a simple label swap. Eight hundred and sixty-two is not 789, and 959 is not 702. Nor can the gap be explained by saying that one figure counts contacts and the other members.

A reconciliation should preserve the relationship between the units. ARIN should state whether the same contact submitted every organisation's ballot in one session, whether blank ballots count toward participation, and whether a contact's partial completion can leave some represented organisations in a different state. These are operational details with governance consequences.

Eligibility was a dated status, not a permanent population

The result page reports 2,023 General Members in total as of 9 September 2024 and 1,963 eligible organisations as of 17 October. The footnote defines an eligible organisation as a General Member in Good Standing with a properly registered Voting Contact linked to an ARIN Online account by the update deadline. Sixty General Members therefore fell outside the election-ready population between those two reported categories.

This is a coherent distinction. General membership alone is not enough if dues are outstanding, contact data is missing or the authorised person lacks the necessary account link. A fixed eligibility date prevents the roll from changing during voting. It also means that each published number must carry its date and rule. A year-end membership total cannot replace an October voter roll.

ARIN's 2024 annual report illustrates the problem. It says there were 1,663 General Members as of 31 December 2024, substantially fewer than both the September and October figures on the election page. That difference may reflect the participation-based conversion of membership status, year-end cleanup, a changed reporting definition or another administrative event. The annual report and election archive serve different purposes. Neither should be silently used as the denominator for the other.

The electorate is therefore a sequence: service organisations that could request General Membership; organisations holding that status; General Members in good standing; those with a valid Voting Contact; those appearing on the certified roll; contacts who authenticated; organisations for which a ballot was initiated; organisations for which a ballot was submitted; and ballots included in each contest. A turnout figure selects two stages from that sequence.

ARIN disclosed enough to identify several stages but not enough to connect them. The two turnout totals occupy unidentified positions near the end. Until those positions are labelled, the public cannot know whether it is comparing submitted ballots with counted contest ballots or an accurate figure with an obsolete one.

Quorum does not resolve the discrepancy

The process summary says quorum was achieved because ballots were cast by more than five percent of the total eligible General Membership. Both 862 and 959 are far above five percent of 1,963. The inconsistency therefore does not threaten quorum on the published numbers. It would remain satisfied even if a substantial number of ballots were removed.

This is important because governance disputes often jump from any numerical defect to invalidity. A wrong participation total can matter without changing the legal outcome. Quorum is a threshold question: did enough eligible organisations participate for the election to proceed under the bylaws? Once the record is hundreds of ballots above the threshold, a 97-organisation dispute does not produce a close-call quorum case.

But a comfortable threshold is not a licence for a careless record. Turnout supports claims about representation, member engagement and institutional health. It may determine whether General Members retain status under participation rules. It shapes outreach plans and historical comparisons. Researchers may use the official percentage to assess whether governance reforms increased participation. A number can be immaterial to the winner and still material to accountability.

The same point applies to candidate ranking. The published margins between the last elected and first unelected candidates should be assessed using vote totals, not participation denominators. Chris Tacit received 307 Board votes and William Charnock 273, a gap of 34. The discrepancy does not tell us how any additional 97 organisations voted, whether they were permitted to select Board candidates, or whether they submitted blank ballots. It therefore cannot prove that the ranking would change.

The proper conclusion is narrower and stronger: validity has not been disproved, while auditability has been reduced. ARIN can maintain confidence in the declared winners and still acknowledge that its turnout publication needs correction.

Blank ballots are a plausible explanation, not a published one

ARIN's membership model gives blank ballots unusual significance. Current guidance explains that submitting a ballot, even a blank one, can count as participation for maintaining General Member status. That creates at least three potentially valid participation measures: organisations that submitted any election ballot; organisations that made at least one selection in a contest; and organisations included in a particular contest denominator.

If 959 organisations submitted something and 862 made at least one candidate selection, the two numbers could coexist. The 97-organisation difference would represent blank or otherwise non-contest ballots. The contact difference could reflect the people who submitted them. That would also explain why candidate percentages use 862 rather than 959, if ARIN defines those percentages over nonblank participating organisations.

Yet the official page does not say this. It uses the phrase "cast ballots" for the 959 figure and "Voting Participation" for 862. Ordinary readers would expect cast ballots to be at least as specific as participation, not a broader administrative status. The process summary also says quorum was achieved as ballots were cast by more than five percent, without identifying which count supplied the quorum test.

There are further complications. A ballot may be blank in the Board contest but contain Advisory Council selections. An organisation might vote in Board and AC elections but not the NRO NC election. A system might count one election package, three contest ballots or each represented organisation differently. "Blank" is not a single state unless the ballot architecture is described.

ARIN should not rely on outsiders to infer this explanation from later membership guidance. If blank submission is the answer, the correction is simple: name the 959 as election packages submitted, identify 862 as organisations casting at least one valid candidate vote, disclose contest-specific blank counts and state which measure satisfied quorum and participation requirements. A short reconciliation would turn speculation into evidence.

Multiple contests may have different populations

The 2024 event selected three trustees, five Advisory Council members and one NRO NC representative. Although ARIN presents the contests together, they need not have identical eligibility or completion states. A voter may skip one contest. Rules for the NRO NC may differ from rules for ARIN corporate bodies. A ballot interface may permit partial submission.

The published tables nevertheless use one denominator, 862, across all three. Amy Potter's 312 NRO NC votes are shown as 36.19 percent, which again equals 312 divided by 862. This uniformity is convenient but raises a question: does 862 mean organisations that submitted the combined election package, organisations with any valid selection, or a fixed denominator imposed across contests regardless of abstention?

If abstainers remain in the denominator, percentages measure support among all package entities. If they are removed, percentages measure support among organisations making at least one selection in that contest. Both approaches can be defensible. They communicate different levels of reach. The page should say which it uses.

The larger 959 figure might include organisations that participated in a governance status process but were not eligible for every contest. It might include a preliminary count before duplicate organisational states were consolidated. It might include ballots submitted after an early export and later confirmed. Again, these possibilities show why a data dictionary matters. They do not supply one.

A modern online election should be capable of producing a contest matrix without compromising secrecy: eligible organisations, authenticated contacts, represented organisations, submitted packages, blank packages, contest entities, contest abstentions, invalid or quarantined ballots and counted ballots. No candidate choice needs to be linked to an organisation. Aggregated state counts are enough to reconcile turnout.

ARIN's current presentation compresses that matrix into one unexplained line and then contradicts it. The cost is not only confusion about 2024. It makes later year-to-year comparisons vulnerable to hidden definition changes.

Certification verifies a process, not every sentence on a webpage

The result says the eligible list was certified by ARIN's President and CEO and that a trustee confirmed the review. It says the President, that trustee and General Counsel confirmed adherence to approved procedures and the vote tally. These are meaningful controls. They identify responsible officials and distinguish roll certification from tally confirmation.

They do not automatically certify the later prose and statistics block. Election systems often produce several outputs: a vendor tally, an eligibility export, a participation report, a communications draft and a web publication. A correct tally can be inserted into a page with an incorrect explanatory number. A reviewer can verify winners while missing a denominator copied from a different report.

This is why institutional assurances should be attached to artefacts. A statement that "the results" were verified leaves open whether officials reviewed candidate totals only, the turnout calculation, the contact mapping, the web page, or all of them. A signed certification should name the file or hash, export time, contests, included states and reconciliation performed.

The publication discrepancy is therefore not evidence that the named officials failed in every duty. It is evidence that the chain from verified system output to public historical record lacks a visible final control. The election may have been administered correctly and communicated inconsistently.

That distinction should make correction easier. ARIN need not reopen ballots or cast suspicion on candidates merely to explain the page. It can retrieve the certified exports, identify the source of each number, publish a reconciliation and preserve the original text in an amendment history. Treating every correction as a threat to the whole election encourages institutions to remain silent. Separating tally assurance from publication assurance permits honest repair.

The 97-organisation gap has no public identity

An audit does not require publication of voter names or choices. It does require a categorical account of the difference. Ninety-seven organisations are large enough to demand a reason but small enough to describe through aggregate states. Were they blank ballots, NRO-only voters, late confirmations, duplicate records, test entries, organisations represented by contacts who did not complete every ballot, or a typographical error?

The public record provides no answer. Nor does it show whether the 87 additional contacts map only to those 97 organisations. Because a contact may represent several members, a reconciliation should be many-to-many aware. Simply subtracting totals does not prove that the larger set contains the smaller set.

Version timing may matter. The election closed on 1 November and results were posted on 7 November. The eligible roll was dated 17 October. A vendor could have issued a preliminary participation export and a final one. ARIN could have revised the page after publication. The Vault now describes archived material as final and unchanged, but that does not reveal whether the page had prior public versions or whether the inconsistency existed from first publication.

This is where archived captures can assist, but they cannot replace ARIN's own ledger. A web archive may show when a sentence appeared. It cannot prove what the election system contained, whether a draft was corrected internally or which officials approved each output. The authoritative reconciliation must come from the institution holding the records.

ARIN should publish a compact difference table: state name, organisation count, contact count, inclusion in quorum, inclusion in membership-retention participation, inclusion in each contest denominator and reason for exclusion. Such a table would answer the governance question without exposing a single vote choice.

Turnout claims should be downgraded until reconciled

Institutional legitimacy is partly a claim about who participated. If ARIN says 43.91 percent of eligible General Members voted, it is making a factual representation about the reach of the election. If its own page also supports 48.85 percent, the claim lacks a stable base. The responsible response is not to choose the more flattering figure. It is to downgrade confidence in any turnout conclusion until definitions are reconciled.

This downgrade does not extend automatically to every fact on the page. Candidate totals are internally coherent. Winners are clearly identified. The election dates, candidate slate and oversight roles are documented. Evidence should be graded at claim level, not used as an all-or-nothing badge.

The distinction is especially important in an organisation that presents itself as transparent and community governed. A transparent archive is not merely a collection of documents. It is a system in which a diligent outsider can reproduce central claims from defined inputs. Here, the outsider can reproduce 43.91 percent from 862 and 1,963. The outsider cannot discover why the page later says 959.

ARIN should therefore avoid using 2024 turnout as evidence of increased engagement, successful membership reform or broad regional endorsement until it publishes the bridge. It may still report both numbers with qualifications. For example: 959 organisations submitted an election package, of which 862 cast at least one candidate vote. If that is accurate, it is more informative than either percentage alone.

Legitimacy improves when an institution narrows a claim to what its evidence can bear. A correction would not show weakness. Leaving a visible contradiction unexplained while continuing to cite turnout would.

Membership retention raises the stakes

ARIN's General Membership rules connect voting participation to continued status. An organisation can maintain that status by participating at least once within a defined period, and a blank ballot can count. That makes the participation ledger more than a communications statistic. It can influence which organisations remain in the future electorate.

If the 959 figure is the membership-retention count and 862 is the contest denominator, the distinction should be explicit because the two measures serve different legal and governance purposes. A member may satisfy the obligation to participate without selecting a candidate. That choice preserves formal membership while withholding electoral endorsement. Treating it as candidate-election turnout would overstate support; omitting it from retention records could wrongly alter status.

The discrepancy also creates a temporal feedback loop. Today's participation record helps define tomorrow's General Membership. Tomorrow's eligible electorate then becomes the denominator for the next turnout rate. An unexplained counting change can therefore affect trends even when it does not alter a 2024 winner.

ARIN should document the status effect separately from ballot counting. Each organisation should be able to see privately that its participation was credited, while the public receives aggregate totals. The election system should produce a signed retention event distinct from the secret candidate selections. That separation protects ballot secrecy and permits audit of membership status.

The larger lesson is that governance data has downstream consequences. A line on a result page may feed annual reports, board dashboards, outreach targets and eligibility processes. Publication controls should reflect that use. The institution needs a canonical source for each metric, not manual copying among documents with subtly different definitions.

Until ARIN explains the two figures, members cannot know whether the public contradiction is cosmetic or connected to a status ledger. The absence of evidence of harm is not evidence that the distinction is unimportant.

Regional representation cannot be inferred from either total

ARIN publishes voters by subregion, which can help test participation across the United States, Canada and the Caribbean. But geographic interpretation depends on the same denominator discipline. A count of organisations, a count of contacts and a count of ballots can produce different regional pictures, especially when one contact represents members registered in several places.

Neither 862 nor 959 measures all network operators, resource holders or Internet users in the ARIN service region. The electorate consists of eligible General Member organisations. Service Members that did not opt into General Membership, downstream customers, policy entities and affected users are outside that denominator. A high percentage would still be corporate participation, not a regional plebiscite.

The contact structure adds another layer. The location of a Voting Contact may differ from the legal address or operations of the represented organisation. A multinational group may centralise voting in one office. Caribbean participation should not be measured by assuming every contact's location equals the organisation's operational constituency.

This does not make subregional reporting impossible. It calls for explicit units. ARIN can publish eligible and participating organisations by registered service subregion, then separately publish the number of unique contacts and concentration bands. It can show how many organisations had no valid contact without naming them. It can compare rates over time only after holding definitions constant.

The inconsistent 2024 total weakens any regional analysis because readers cannot determine which population the subregional file reconciles to. If it sums to 862, what are the other 97? If it sums to 959, why do contest tables exclude them? A checksum should be part of every downloadable table.

Representation claims are strongest when the route from member record to aggregate is visible. Geography cannot rescue an unstable base count.

A correction must preserve history

ARIN's Vault warns that archived information is published in final form and not changed. That preservation policy protects against silent historical revision. It should not prevent correction. The standard response to an error is an attached amendment that leaves the original visible, identifies the date and author of the correction, and explains the effect on prior conclusions.

Silently editing 959 to 862 would remove evidence of the publication failure and make archived captures appear inconsistent with the official site. Silently changing 862 to 959 would be worse because candidate percentages would no longer reproduce. Deleting the page would damage the election archive. None is necessary.

A correction notice should answer five questions. Which number is accurate for each defined purpose? What system report produced each published figure? Did the discrepancy affect quorum, candidate totals, ranking, membership-retention credit or subregional tables? Which officials reviewed the reconciliation? What control changed for future elections?

If both numbers are valid, the notice should relabel them rather than choose one. If one is wrong, it should state the cause, such as a draft export or copy error. If records no longer permit a full answer, ARIN should say so and identify the retention gap. A candid unknown is more trustworthy than unexplained precision.

Preserving history also means publishing machine-readable snapshots. A CSV of aggregate contest states, a human-readable certification and a cryptographic digest can coexist. The digest does not reveal votes; it proves which file officials signed. Future corrections can reference the original digest and a new one.

The goal is not to punish a web editor. It is to restore a reliable institutional memory. An election archive should survive staff turnover and changing software without asking readers to trust whichever paragraph appears last.

What a canonical election ledger should contain

A canonical ledger need not expose secret ballots. It should record aggregate transitions and responsibility. The first section should identify the certified eligibility snapshot: date, rule version, total General Members, exclusions by reason, eligible organisations, unique Voting Contacts and concentration bands for contacts representing multiple organisations.

The second should describe authentication and submission: contacts who authenticated, organisations available to them, ballot packages opened, packages submitted, blank packages, incomplete packages, quarantined submissions and confirmed organisational ballots. The third should provide contest-level counts: eligible organisations, entities, explicit abstentions, blank selections, counted ballots and candidate totals.

The fourth should reconcile outputs. Every organisation-count state should have a definition and a checksum relationship. Every public percentage should name numerator and denominator. Contact totals should never be presented as a substitute for member totals. If membership-retention credit uses a broader state than candidate participation, that state should have its own label.

The fifth should document assurance: election vendor, system version, export time, certifying officials, exceptions, incident log and signed result artefacts. A public summary can omit security-sensitive detail while confirming that an independent reviewer inspected it.

Applied to 2024, the ledger would make the discrepancy easy to resolve. The 862 and 959 counts would either occupy distinct rows or one would be marked superseded. The 702 and 789 contact totals would follow the same mapping. Candidate percentages would point to the 862 row. Quorum and status retention would point to their actual rows.

This is not bureaucratic excess. ARIN already holds most of the data. The improvement is to preserve definitions and relationships at publication time rather than reconstruct them after a challenge.

Independent review should test the publication chain

ARIN's internal certification roles are identifiable, but a mature election should also receive review independent of candidates, staff who operate the system and trustees whose colleagues are being elected. Independence need not mean outsourcing every task. It means that someone with access and no stake in the result can reproduce the critical totals and report exceptions.

The reviewer should test three layers. First, eligibility: sample or fully reconcile the certified roll to the applicable membership and contact rules. Second, tally: verify that confirmed ballots produce the published candidate totals without linking choices to identities. Third, communication: recalculate every percentage, compare every narrative total to the signed aggregate and confirm that downloads sum to the headline.

The 2024 contradiction likely would have been caught by the third layer. That is useful because it shows why technical vote verification is not enough. Public legitimacy depends on the record members can inspect, not only on a private assurance that a vendor counted correctly.

The independent report should be short and specific. It should identify the artefacts reviewed, materiality thresholds, unresolved exceptions and whether each exception could affect winners or only reporting. A statement that everything was "verified" without scope invites overreading.

Review also protects ARIN officials. When a discrepancy emerges, they can point to a documented chain rather than defend an institution-wide claim. If the problem occurred after certification, the correction can isolate it. If it originated in the system, the reviewer can identify the relevant control.

No governance structure eliminates error. Good structure makes error detectable, bounded and repairable. ARIN's 2024 page currently detects the error for us by contradicting itself; the institution should complete the repair.

The winners and the record are separate questions

It is fair to ask whether focusing on turnout distracts from the elected candidates. The answer depends on how the claim is framed. The published vote totals establish a ranking under the announced method. The unexplained participation gap establishes uncertainty about the denominator. Those findings can stand together.

The Board election placed Peter Harrison first with 401, Ron da Silva second with 344 and Chris Tacit third with 307. The Advisory Council and NRO NC tables likewise identify clear winners. Nothing in the two turnout statements assigns additional candidate choices or demonstrates that excluded ballots exist. It would be irresponsible to recalculate winners from imaginary votes.

It would be equally irresponsible to say that because winners cannot be overturned from public evidence, the contradiction does not matter. Elections create office and an accountability record. Members need both. A correct result with an unreliable explanation is better than a wrong result, but worse than a correct and auditable result.

Separating the questions produces a proportionate remedy. Do not annul the election on this evidence. Do not accuse candidates or contacts. Do require ARIN to reconcile the totals, disclose impact and improve publication controls. If reconciliation later reveals a tally issue, that new evidence can be assessed under the applicable challenge rules.

This approach avoids two familiar institutional failures: catastrophising every error until officials become defensive, and minimising every error until records lose value. The 97-organisation gap is a bounded, answerable problem. Treating it that way should make a public answer possible.

What members should be told now

ARIN should issue a notice beside the archived result. It should state that the page contains two participation totals, that candidate percentages were calculated using 862, and that a reconciliation has been completed or is under way. It should avoid implying fraud or blaming voters.

The notice should then publish the state definitions. If 959 submitted election packages while 862 cast at least one candidate vote, say exactly that. If a late report was copied by mistake, identify it. If the larger total includes blank ballots credited for General Membership retention, show the count and confirm those organisations received the proper credit. If the discrepancy affects a voters-by-subregion file, replace that file through a versioned amendment.

Members should also receive a private way to confirm that their organisation's participation status was recorded, without revealing candidate choices. This matters for organisations whose General Membership depends on participation. Disputes about status should have a documented appeal path and should not require a member to disclose its ballot.

For future elections, ARIN should announce denominator definitions before voting opens. The result template should be generated from the canonical aggregate rather than manually populated. Automated checks should recalculate percentages and reject a page where narrative totals do not match tables. A named reviewer should sign the final public artefact.

Finally, ARIN should report the correction in its next annual governance review. The lesson is broader than one number: election assurance extends from voter eligibility through public archival communication. Closing that loop strengthens rather than diminishes the legitimacy of the elected bodies.

The cost of leaving the record inconsistent

The immediate cost is uncertainty about 97 organisations. The longer cost is cumulative. Future reports may copy 43.91 percent or derive 48.85 percent. Researchers may build a trend line with incompatible definitions. Board presentations may claim a participation increase or decline that is partly a reporting change. Members may question whether their own ballot was counted in one total but not another.

An unexplained contradiction also changes the burden of trust. Instead of ARIN supplying a coherent record, every reader must choose a theory. Supporters may select the explanation most favourable to the institution; critics may select the most damaging. Neither has the data. The absence of reconciliation becomes space for polarisation.

The cost is especially high because the error is easy to see. No specialist access is required. The two statements sit on the same page, and the percentages can be checked with elementary division. If a visible inconsistency persists, members may reasonably wonder how less visible exceptions are handled.

The opportunity is equally clear. ARIN can demonstrate the behaviour expected of an accountable registry: acknowledge a bounded problem, publish the source definitions, preserve the prior record, assess impact and change the control. That response would provide stronger evidence of institutional maturity than an unqualified claim that the process was verified.

The 2024 election may well have selected its winners correctly. The public record does not currently earn the same confidence for turnout. Corporate authority rests on rules, but durable legitimacy also rests on the ability to show how facts were produced. ARIN's two numbers are not a verdict on its election. They are a test of whether the institution will make its own history auditable.