NANOG’s Board Mandate Has a Denominator

Summary

• NANOG’s comparable 2017–2025 Board-election record shows persistent but partial participation: 1,690 ballots across 6,092 annual eligible-voter observations, a pooled ratio of 27.74%. These are repeated yearly observations, not unique people.
• Partial turnout does not by itself invalidate an election inside a voluntary association. NANOG’s published denominators make its bounded internal mandate more auditable; they do not turn individual members into a representative sample of all North American operators.

The revealing number is 165

The most revealing figure in NANOG’s recent election record is not a candidate’s winning tally. It is 165. In the 2024 Board election, 698 people were eligible to vote and 165 cast ballots. Turnout was 23.64%, the low point of the nine elections for which NANOG’s public pages report a consistent pair of eligible-voter and total-ballot figures.

That result tempts two opposite overstatements. One says a Board chosen with ballots from fewer than a quarter of the eligible electorate must be illegitimate. The other says that because the election selected NANOG’s governing Board, the winners carry the electoral voice of the North American network-operator community. The first confuses participation with validity. The second enlarges a corporate electorate into a continental constituency. Neither proposition follows from the count.

The figure establishes something narrower and more useful. It tells members and outside readers how many people could vote under NANOG’s rules and how many exercised that opportunity. It supplies a scale against which claims about engagement can be tested. It does not say that the remaining 533 eligible voters opposed the winners, supported them silently, missed a deadline, trusted other voters, disliked the choices or faced a barrier. It does not say that 165 companies, autonomous systems or networks delegated authority to the Board. The public record contains no evidence for any of those translations.

Turnout is often asked to perform work it cannot do. A low number becomes a diagnosis of apathy, capture or exclusion. A high number becomes proof of representativeness. An individual vote becomes an employer’s vote, and an organizational election becomes a verdict delivered by an industry. But an election page can establish only what its rules and reported fields show: a defined electorate, a voting opportunity, a number of submitted ballots and an outcome within a specified institution. Motive, social composition and authorization by outsiders require other evidence.

NANOG’s own governing boundary helps keep the analysis exact. The current bylaws provide for annual voting on Board candidates and proposed amendments. They describe seven voting directors: six elected directors and the Executive Director. They also state that NANOG is not itself a network operator. Its Board governs NANOG’s corporate affairs. An election does not hand that Board operational control over a member’s routers, routes, customers or infrastructure, much less over systems belonging to nonmembers.

The denominator is therefore not ammunition against NANOG. It is the foundation for the strongest defensible account of NANOG’s authority. A voluntary institution can govern itself through the constituency and procedure its rules establish without claiming public-law sovereignty or universal industry representation. Publishing the size of that constituency and the participation within it makes the claim more credible because it makes the claim’s limits visible.

The central question is not whether every eligible member voted. Voluntary associations routinely act through the eligible members who choose to participate. The sharper question is what the recorded participation authorizes an observer to say. NANOG’s archive provides a useful answer, but only if twenty-one annual links are not mistaken for one continuous electorate.

Before membership, eligibility came through attendance

NANOG’s past-election archive links annual records from 2005 through 2025. Its breadth is valuable. Archive presence, however, does not guarantee that every page publishes the same fields. More fundamentally, it does not make the electorate on every page the same institutional population.

The early records concern the Merit-hosted NANOG arrangement and its Steering Committee. The 2005 election page tied eligibility to attendance at at least one NANOG meeting during the preceding two-year window. More than 1,800 people met that condition, 319 registered for the election and 260 actually voted.

Those numbers occupy different levels. “More than 1,800” describes the eligible population. The 319 registrants were a subset who completed an election-registration step. The 260 voters were those who submitted ballots. Dividing 260 by 319 produces 81.50%, but that is registration-to-vote conversion, not turnout among everyone eligible. Dividing 260 by 1,800 produces 14.44%, but the true eligible count exceeded 1,800. Whole-electorate turnout was therefore strictly below 14.44%, not exactly 14.44%.

The distinction is more than mathematical housekeeping. Registration conversion can show whether people who entered an election process completed it. Whole-electorate turnout shows the share of those entitled to participate who did so. Neither number establishes whether that electorate resembled all network operators in North America. Attendance during a rolling two-year window was not paid individual corporate membership, an employer delegation, an ASN roll or a continental sampling frame.

The 2006 page reports more than 1,800 eligible people and 108 who voted. Actual turnout was therefore below 6.00%, never an exact 6.00%. The 2007 page gives more than 1,700 eligible and 183 voters, putting turnout below 10.76%. These are strikingly low upper bounds, but they remain measures of participation in an attendance-linked Steering Committee electorate. They cannot be appended to the later corporate-membership series as though only the date changed.

The 2008 record reports more than 1,818 eligible and 202 voters, so actual turnout was below 11.11%. It also describes a design in which the voter’s identity and ballot choices were decoupled. That is NANOG’s description of its mechanism, not an independent security audit. Properly bounded, it is still useful: the page says the system sought to check eligibility and control ballots without attaching a named identity to the selections. Alongside the electorate threshold and the voter count, the explanation revealed part of the control surface rather than merely announcing winners.

The limit appears on the 2009 page. It reports more than 1,784 eligible people and publishes candidate and charter tallies, but it gives no unique total-ballot count. Candidate selections cannot be added to reconstruct turnout because one voter could make multiple selections. A charter tally answers a different question again. The supported conclusion is simply that the public numerator required for a turnout ratio is missing.

This first regime supplies an important historical comparison. Participation among the attendance-linked eligible population was partial, sometimes markedly so. Yet those figures alone do not show that the Steering Committee lacked authority under the arrangement that created it. The reviewed material supplies no majority-turnout condition. Eligible people could decline to vote, and the records do not reveal their reasons. “Apathy” would be an inference; so would “satisfaction,” “exclusion” or “consent.”

Most importantly, these figures belong to a constituency produced by recent attendance. They show why denominators matter and how carefully threshold denominators must be handled. They are not the opening entries in a single 2005–2025 turnout curve. Every use of “eligible” has to retain its institutional noun: eligible under which arrangement, for which office and according to what rule?

The transition changes the constituency, not just the name

The 2010–2011 transition matters to this analysis for one narrow reason: it separates two electorates. Treating it as cosmetic would corrupt the denominator study; allowing it to dominate would turn this into a different article.

The 2010 election page records 210 yes and 16 no on the question of transitioning NANOG activities from Merit to NewNOG. The page also presents Steering Committee candidate tallies. These figures do not form a common unique-voter measure. The 210–16 count belongs to a transition question, while the candidate figures belong to a contest in which selections cannot be presumed to equal persons. Combining or cross-dividing them would create a crisp-looking answer to a question the page does not answer.

Ashwin Mathew’s 2014 Berkeley dissertation independently describes the institutional shift from the Merit-hosted, attendance-linked arrangement toward self-governance through paid individual membership. It also records contemporary transparency concerns. That account is valuable because it shows that corporate membership was a different constituency mechanism, not merely a fresh label for recent attendees. It is historical interpretation, not current turnout evidence, an external audit of the elections or proof that one concern caused a later reporting change.

The 2011 results page illustrates both the new regime and the new measurement limit. It publishes Board candidate tallies and a membership-structure result of 79 yes and zero no. “79–0” is unanimity among the recorded responses to that proposition. It is not unanimity among all eligible members. The page supplies neither the eligible-voter total nor a unique total-ballot denominator, so it cannot show how many eligible people abstained from that question or from the election.

That distinction matters because “unanimous” can conceal a denominator. Agreement among every counted response is a legitimate description of the proposition tally. Agreement among the whole electorate would be a much larger claim. Without the eligible total and a unique participation count, the second claim is unavailable. A result should not acquire broader consent merely through an adjective.

The transition thus closes one quantitative comparison and opens the corporate period. Before it, attendance created eligibility for a Steering Committee vote in the Merit-hosted arrangement. After it, paid individual membership defined the constituency for corporate Board elections. The office, governance form and route into the electorate changed. The public pages do not provide a bridge that would make the two electorates statistically continuous.

A reporting gap is not a missing election

From 2011 through 2016, NANOG’s migrated pages make the winners easier to see than the size of the electorate. That is a disclosure finding, not a judgment that the underlying elections were absent, secret or invalid.

The 2012 page names elected Board candidates and records approved bylaw amendments. The 2013 page publishes winners and, among the bylaw material, a terminology change from “eligible voters” to “members.” The 2014 page identifies Board winners and term assignments. The 2015 page records winners and an approved bylaw outcome, while the 2016 page again gives elected candidates and a bylaw outcome. None publishes the paired eligible-voter and unique total-ballot fields needed to calculate comparable Board turnout.

Several interpretations go beyond that evidence. Calling the elections secret ignores the results that are public. Saying the elections lacked internal records treats a migrated webpage as a complete inventory of whatever the institution retained. Declaring the outcomes invalid assumes that public denominator disclosure was a constitutive rule, when the available pages do not show that. Better disclosure improves outside auditability, but absence of a field on a public archive page is not proof that an election failed its governing requirements.

Filling the blank would be equally unsound. Candidate tallies count selections and may let one voter contribute several entries. Bylaw yes/no totals concern particular propositions and can have their own abstentions or eligible groups. A winner list contains no unique voter count at all. Where the page lacks the numerator and denominator for the same election, there is no responsible annual turnout calculation.

This is useful institutional ignorance because it is specific. We know what the public pages disclose and which comparison fields they omit. We do not know whether the gap reflects archival migration, contemporary reporting style, election administration or another cause. We do not know why consistent paired-denominator reporting appears later, and the sequence alone does not prove that criticism triggered the change. The record supports a description of reporting evolution, not a causal account of it.

Winner disclosure and electorate disclosure perform complementary functions. Naming the people who took office reveals the outcome. Candidate tallies show the distribution of recorded selections, subject to the ballot method. Publishing the eligible population and the unique ballots shows engagement with the opportunity to vote. No one field decides legitimacy. Together, however, the fields make the constituency and the result easier to scrutinize without turning “the community” into an undefined source of authority.

An annual archive can look like a ready-made time series because its links form an unbroken sequence. Comparability lives deeper: in the field definitions, the office being elected and the rule that creates eligibility. NANOG’s archive contains a regime break, several ballot types and changing public formats. The defensible quantitative series begins only when the same pair of fields appears for the same kind of corporate election.

Nine Board elections form the comparable core

That core begins in 2017. From then through 2025, each annual Board-election page states both the number of eligible voters and the total votes cast. The ratio between those fields measures participation inside that year’s corporate membership electorate.

The 2017 election recorded 687 eligible voters and 197 ballots, or 28.68%. The 2018 election recorded 692 eligible and 192 ballots, or 27.75%. The archive position and canonical page identify this as the 2018 election; an apparent date typo in a page heading does not belong in the analysis. The 2019 election recorded 716 eligible and 189 ballots, or 26.40%.

The 2020 Board election recorded 553 eligible voters and 176 ballots, a turnout rate of 31.83% and the high point of this nine-year series. The same page reports a separate bylaw election with 583 eligible and 125 votes. Those electorates and ballots must remain separate. Combining them would not produce a fuller turnout measure; it would produce a number corresponding to no actual election.

The 2021 Board election had 538 eligible voters and 156 ballots, or 29.00%. The 2022 summary reports 721 eligible and 201 ballots, or 27.88%. Its candidate and bylaw abstention fields do not replace or modify the unique ballot total. An abstention within a question is not evidence that an eligible person submitted no ballot.

The 2023 summary gives 762 eligible voters and 230 ballots, or 30.18%. Its one recorded candidate abstention is not a count of nonvoters. The 2024 result then records 698 and 165, or 23.64%. Finally, the 2025 summary reports 725 eligible and 184 ballots, or 25.38%. Ranked-choice and bylaw fields on that page answer other questions and remain outside the Board-turnout numerator.

Year	Eligible voters	Board ballots	Turnout
2017	687	197	28.68%
2018	692	192	27.75%
2019	716	189	26.40%
2020	553	176	31.83%
2021	538	156	29.00%
2022	721	201	27.88%
2023	762	230	30.18%
2024	698	165	23.64%
2025	725	184	25.38%

This table describes nine elections. It does not establish a trend, statistical significance or a cause. The rise from 2022 to 2023 is not proof of successful outreach. The 2024 fall is not proof of disillusionment. The movement in 2025 is not proof of renewal. No voter-motivation survey accompanies the figures, and the pages do not connect annual variation to dues, communications, nominations, conference attendance or ballot design.

Changes in the eligible count are equally resistant to easy stories. Eligibility fell from 716 in 2019 to 553 in 2020 and 538 in 2021, then rose to 721 in 2022 and 762 in 2023. Membership churn, dues status, the date of the good-standing snapshot, administrative timing and personal preference are plausible hypotheses. None is established. The annual denominator states how many people were eligible for that election; it does not explain how the number was produced or why it changed.

The durable finding is persistent partial participation in the defined membership electorate. Over these nine elections, annual turnout never reached one third. The ratios also stayed within an 8.19-point band. That supports a precise statement about the observed series, not a psychological account of the electorate and not a claim about the wider operator population.

Three summaries answer three different questions

Across 2017–2025, eligible-voter observations sum to 6,092 and ballot observations sum to 1,690. Dividing the second total by the first gives a pooled ratio of 27.74%. It summarizes all recorded annual voting opportunities while weighting each year by the size of its electorate.

It does not mean that 1,690 distinct people voted out of 6,092 distinct people. The same person may be eligible and may vote in several years; other people may enter or leave the electorate. Without longitudinal identity evidence, both totals are sums of repeated annual observations, not a cohort. Treating the denominator as 6,092 unique members would convert arithmetic into an invented population.

The unweighted mean of the nine annual turnout ratios is 27.86%. It asks what participation looked like in the average year when every year receives equal weight. Because the electorates differ in size, the mean need not equal the pooled ratio. The twelve-hundredths-of-a-point difference between 27.86% and 27.74% is not an error; it reflects two summary conventions.

The third description is the spread. The low was 23.64% in 2024 and the high was 31.83% in 2020, a range of 8.19 percentage points. Range shows observed variation. It does not establish the direction of a trend, the statistical importance of the changes or their cause.

The 2020 high illustrates why ratios need raw counts. Its 176 ballots were fewer than the totals in 2017, 2018, 2019, 2022, 2023 and 2025. The rate was highest because the eligible denominator was substantially smaller. “Highest turnout rate” is supported. “Greatest engagement” would need a definition and evidence the annual page does not supply.

The 2024 low makes the inverse point. Its 165 ballots were not the smallest ballot count; 156 were cast in 2021. The rate was lowest because the 2024 electorate was larger. That arithmetic says nothing about whether nonvoters were indifferent, content, busy, unaware, excluded, dissatisfied with the candidates or making a deliberate choice. Nor does it show that the participating group was coordinated or captured.

Raw counts and ratios are valuable together. A percentage can conceal a changing electorate size. A ballot count can make a larger year appear more engaged while ignoring the larger pool that could have participated. Side by side, they answer a bounded question: how many could vote, how many did and what share did the latter represent? They do not answer who those people were or why they acted.

Every electoral number needs its own noun

An election page can contain many numbers without containing several turnout numerators. Candidate tallies count selections. A ballot may permit multiple candidates or ranked preferences. A bylaw total counts responses to a proposition and may involve a different eligible group. An abstention may occur inside a submitted ballot. None of these fields can replace a reported unique total-ballot count.

The problem recurs across the archive. Early Steering Committee voters could make multiple candidate selections. The 2009 page has candidate and charter figures but no unique ballot total. The 2010 page has candidate tallies and a transition question. The 2011 page has Board tallies and the 79–0 membership-structure vote. Summing, subtracting or cross-dividing those fields would manufacture a precise answer to an unreported question.

The recent pages reduce the ambiguity by labelling total votes cast, but they do not abolish the need for care. In 2020, the Board and bylaw elections have different eligible and participating totals. From 2022 onward, abstention or proposition fields belong to particular choices. A candidate abstention is not a nonvoter. Confusing the two would turn a ballot-level response into a claim about someone who never participated.

The 2025 materials provide another warning. NANOG’s rules allow ranked-choice variants and let the Board specify the voting method before nominations. Method determines how candidate preferences are expressed and tabulated. It does not alter the basic turnout calculation when the page provides 725 eligible people and 184 total Board ballots. Candidate percentages produced under different methods should not be compared mechanically, while the unique ballot-to-eligible ratio retains its limited meaning.

Good disclosure thus requires something like a compact data dictionary: eligible individuals, registered individuals, submitted ballots, valid ballots, candidate selections, rankings, proposition responses and question-specific abstentions are different entities. The distinctions protect interpretation just as ballot secrecy protects individuals. The 2008 description of identity-choice decoupling shows how eligibility checking and confidentiality can coexist; later denominator reporting permits participation to be assessed without revealing how any named person voted.

Every annual page from 2020 through 2025 also links to certified-result documents. The links extend the visible chain from summary to certification. Their contents were not inspected for this account, however, and they supply no additional quoted facts or corroboration here. What can safely be said is that the annual pages provide the links, not what each linked PDF contains.

Auditability is cumulative rather than magical. Written rules define the constituency, office and authority. Privacy measures protect the secret ballot while eligibility and duplicate-vote controls are maintained. Annual pages announce winners and, in the comparable period, show the eligible and participating totals. Certified-result links offer a further route for review. No one element proves a flawless process. Together they make NANOG’s internal governance more legible without requiring the disclosure of individual choices.

The strongest defence of a partial electorate

The most serious defence of NANOG begins with what kind of institution it is. NANOG is a voluntary, private association with a defined corporate membership and written bylaws. Its Board does not need a plebiscite of every network operator in North America to manage NANOG. It needs authorization through the constituency and procedure that govern NANOG itself.

The reviewed bylaws do not make majority turnout a condition of a Board election. Fewer than half of eligible members voting therefore cannot, by itself, invalidate a result. Voting is an opportunity, not an obligation. An eligible member may lawfully decline to participate. Without evidence of motive, that absence cannot be converted into opposition to the winners, rejection of NANOG, satisfaction with the status quo or proof of a barrier.

Nor should a low ratio erase the agency of those who did participate. Each submitted ballot was an exercise of a right available within the association, and the resulting choice matters for the institution that offered it. Insisting that only universal or majority participation can produce meaningful governance would let nonparticipation cancel participation even where the rules impose no such condition. The sounder position preserves both facts: the voters made a consequential corporate choice, while the nonvoters’ preferences remain unmeasured. That is a defence of the election without commandeering silence as endorsement.

This defence should stand in full before external limits are discussed. A common mistake is to treat every voluntary body as a state and every internal election as a referendum on popular sovereignty. NANOG’s Board election has a different purpose. It selects directors for a bounded corporate arrangement. Regular elections, published rules, identified winners, privacy protections, denominator disclosure and certified-result links can sustain meaningful accountability even when participation is partial.

The benefits of a technical institution are not limited to those entitled to vote. Milton Mueller and Kenneth Merrill Sowell’s entity-observer study offers the broader alternative that private, bottom-up institutions can create public goods beyond the private interests that organize them. That proposition is an important answer to the idea that a bounded membership must imply a narrowly valuable institution. Knowledge exchange, professional norms and voluntary coordination may benefit people who never joined or voted.

Public benefit is not sovereign authority. A conference can spread operational knowledge without receiving a public mandate. A trusted forum can help people coordinate without controlling their networks. A technical discussion can influence practice without binding outsiders. Benefits received by nonmembers do not prove that those beneficiaries elected the Board or delegated authority over their systems.

Keeping these concepts distinct protects NANOG’s legitimacy. Its strongest claim is concrete: members entitled to vote select elected directors to govern NANOG alongside the Executive Director under the corporate rules. Election pages and bylaws make that arrangement open to examination. A grander assertion that the Board embodies all North American operators would require evidence NANOG does not need in order to govern itself and that the reviewed record does not provide.

Turnout therefore calibrates representative rhetoric rather than automatically determining validity. Even 100% turnout would still be participation by the defined membership electorate, not a vote by every operator. Conversely, 23.64% turnout can produce an internally effective election when the governing rules do not impose another threshold. The participation share informs a claim about engagement; the electorate definition controls who had the opportunity to be represented.

The criticism “only 165 people voted, so the Board has no authority” aims at the wrong source of authority. NANOG’s corporate arrangement, not an arbitrary turnout threshold, supplies the Board’s internal mandate. The criticism “165 ballots cannot establish the preferences of a continent’s operators” is stronger because it identifies an extrapolation for which no evidence exists.

North America is not the sampling frame

The word “community” can make a constituency sound self-defining. It is not. A claim of representation requires a population, a method for including members of that population and evidence connecting the electorate to the whole. NANOG’s election pages do not provide those elements for all North American operators because that is not the purpose of its corporate election.

The denominator in the comparable series is the number of NANOG members eligible to vote in a specified Board election. It is not a count of operators, employers, networks or ASNs. Individual members may work across many types of organizations, but the annual pages do not provide a complete employer or ASN mapping. Even if an employer is publicly known, an individual ballot cannot be treated as employer-authorized without evidence of delegation. A person is not automatically a company vote; a company is not automatically one network; and a network is not automatically one ASN.

The record also lacks a complete annual demographic picture of voters and nonvoters. It does not map their geography, employer concentration, network type, job role or recurring participation. It supplies no operator-population census against which the membership electorate can be compared. It does not show how many eligible individuals came from the same organization, which organizations had no eligible members or whether any voter acted in a personal or institutional capacity.

No turnout ratio can repair those missing bridges. A higher rate would show only that more of NANOG’s eligible electorate participated. It would not prove that the electorate had become more representative of the operator population. A lower rate would not prove that operators withdrew confidence. The category error arises before division: when the denominator is not the continental operator population, the resulting percentage cannot measure representation of that population.

NANOG’s authority over operational systems is still more clearly bounded. Its Board may govern membership, meetings, finances, committees and other corporate affairs assigned to it. A Board election does not authorize direction of a third party’s routing policy, maintenance schedule, customers or security response. NANOG can be influential, and voluntary coordination can be valuable, without influence becoming control.

Precision does not require timidity. An institution can speak from expertise without claiming electoral representation of every person affected by its field. NANOG can describe what its Board decided, what its members voted upon, what entities discussed or what a specified process produced. “NANOG’s Board decided” names a verifiable actor and procedure. “North American operators decided” identifies a different constituency and requires different proof.

Denominator disclosure encourages that discipline. It places annual scale next to the corporate noun. The result can be described at its true altitude: an election within a private, voluntary technical institution whose work may have value far beyond its membership. That formulation neither dismisses NANOG nor inflates it.

What the record leaves unknown

The public pages make a reproducible turnout series possible, but they do not explain the series. The membership and good-standing reconciliation behind each annual eligible-voter count is not present. The reasons the denominator rises or falls remain unknown. Dues status, renewal timing and membership churn may be worth investigating; they are not findings.

The same limit applies to participation. The record does not show why eligible members voted or abstained. A missed deadline, limited information, trust in other voters, satisfaction, discontent, weak candidate differentiation, access difficulty and competing demands are mutually different possibilities. None can be selected without direct evidence. Describing nonvoters as apathetic is no more justified than describing them as content.

The annual sums also hide recurrence. There is no evidence showing how many people appeared in several electorates or voted repeatedly. Consequently, 6,092 and 1,690 must remain annual-observation totals. They cannot be presented as unique population counts or a nine-year cohort.

Nor do the reviewed sources provide complete demographics, employer affiliations, ASN connections, geographic distribution, job roles or network types for voters and nonvoters. They contain no defensible comparison with an operator-population sampling frame, no proof that an employer or network authorized an individual ballot and no basis for calling the electorate representative or unrepresentative in a demographic sense.

These unknowns identify possible future research without supplying conclusions in advance. NANOG could, if it chose, publish a privacy-preserving explanation of the membership snapshot used to calculate eligibility. It could survey voting and nonvoting reasons while acknowledging response bias. It could publish carefully aggregated constituency information if it first defined the wider benchmark and protected against identification. It could distinguish new, returning and continuously eligible entities without exposing names.

More stable annual field definitions would also help. Board and bylaw pages could consistently distinguish eligible voters, unique ballots, valid ballots, question-specific abstentions, tabulation method and certification status. The recent archive already publishes much of this information. Greater consistency would reduce the risk that a reader mistakes candidate preferences or proposition responses for people.

These possibilities are not retroactive accusations. Missing public denominators in 2011–2016 do not prove that administrators lacked the counts. Absence of demographic tables does not prove exclusion. Absence of a motivation survey does not prove avoidance. Accountability improves when unknowns are named and investigated, not when they are weaponized into unsupported claims.

The turnout record is also silent about operational outcomes. It establishes no causal connection between election participation and routing reliability, security, training quality, conference programming or network performance. A higher-turnout year may coincide with technical events, but coincidence is not even a measured correlation here, much less causation. Board turnout is an institutional participation measure, not a proxy for the health of the internet.

Disclosure strengthens the mandate by bounding it

The important evolution in NANOG’s public election record is not a steady rise in turnout; the comparable figures show no such simple pattern. The gain is the regular publication of eligible-voter and total-ballot fields, followed from 2020 onward by certified-result links. These disclosures let a reader reproduce participation rates, separate Board elections from bylaw questions and assess the scale on which the organization’s electoral mandate rests.

Disclosure also exposes the edge of the evidence. A denominator says nothing by itself about nonvoter motives. A certification link does not transform an individual membership into an operator census. A privacy mechanism does not establish demographic representation. Each feature answers a particular accountability question, and the temptation to make it answer all questions should be resisted.

Institutional confidence is often expressed through expansive nouns: the industry, the region, the community. Electoral confidence is better expressed through particulars. Who could vote? Under which rule? For what office? How many cast ballots? Which method governed the count? What authority did the office receive? Those questions do not weaken a technical institution. They replace atmosphere with an auditable chain.

For NANOG, that chain supports a robust internal conclusion. Its defined corporate electorate selects directors for its Board. The public 2017–2025 pages record annual turnout between 23.64% and 31.83%, an unweighted mean of 27.86% and a pooled annual-observation ratio of 27.74%. Participation was partial. The record nonetheless supports regular internal elections whose authority does not depend on pretending that nonmembers voted.

The same chain supplies an equally robust external limit. No reviewed source maps eligible members comprehensively to employers, ASNs, network types, geography or the wider operator population. No source establishes that individual ballots were cast under employer or network delegation. There is no evidentiary route from 165 ballots in 2024, 230 in 2023 or the nine-year total to the collective will of North American operators.

The apparent tension disappears once internal authorization and external representation are separated. Internal authorization asks whether a body acts through the constituency and procedure its own rules establish. External representation asks whether that constituency can stand for a larger population. NANOG’s denominators make the basis of its internal mandate measurable. They do not supply the sampling frame required for the second claim.

The 2024 low is therefore neither a scandal hidden in arithmetic nor a number to wave away. It is a precise fact about one Board election: 165 ballots from 698 eligible voters. It tells members how much of the available electorate participated and leaves the reasons for abstention open. It does not, by itself, defeat NANOG’s capacity to govern itself. It also refuses to make that internal electorate something the evidence has never shown it to be.

A voluntary institution need not claim a continent in order to possess a meaningful mandate. It can publish its denominator, defend the integrity of its internal election, acknowledge what remains unknown and describe its authority at the boundary its rules establish. NANOG’s archive demonstrates why that restraint is not weakness. Disclosure earns trust by making the mandate inspectable—and by making its outer limit impossible to mistake.

SEO and social metadata

• SEO title: NANOG Board Elections: What the Turnout Denominator Shows
• SEO description: NANOG’s 2017–2025 Board-election records show persistent partial participation, auditable internal authority and no basis for claiming representation of every North American operator.
• Social title: NANOG’s Board Mandate Has a Denominator
• Social description: Nine comparable elections reveal the strength of NANOG’s voluntary self-government—and the boundary between a membership vote and a mandate from a continent’s operators.

Image metadata

• Alt text: Editorial data visualization of 698 outlined ballot positions beside 165 filled ballots, with a separate line marking NANOG Board turnout from 23.64% to 31.83% across 2017–2025.
• Caption: NANOG reported 165 Board ballots from 698 eligible voters in 2024, the low point of the comparable 2017–2025 participation series.
• Accessibility description: A synthetic, high-contrast editorial illustration uses labeled ballot symbols and numerical annotations to separate eligible voters from ballots cast. Patterns as well as color distinguish the 2024 figures, and a compact nine-year scale shows the reported turnout range without depicting any real person or polling place.
• Synthetic provenance: AI-generated editorial visualization grounded in NANOG’s published eligible-voter and total-ballot figures. It depicts abstract ballots and denominators, not a historical photograph, real election scene or identifiable person.