DAO Governance for Accountability in Canadian Public Administration

Accountability stands as a foundational principle in democratic public administration, yet its operationalization remains contested across jurisdictions and policy sectors. Mark Bovens (2007) defines accountability as a social relationship in which an actor holds answerability and enforceability obligations to others. In the context of public governance, this encompasses traceability of actions (transparent record-keeping), sanctionability (mechanisms to penalize wrongdoing or enforce rules), and contestability (avenues to challenge and review decisions). Mulgan (2000) emphasizes that accountability is "an ever-expanding concept" essential to maintaining legitimacy in public institutions. Traditional public institutions rely on oversight bodies, audits, and hierarchical controls to achieve these ends. Gaps remain. Complex bureaucracies can obscure who is responsible for decisions, and formal channels to contest decisions are often cumbersome or lacking entirely.

In this context, blockchain-based Decentralized Autonomous Organizations have emerged as a novel model addressing some of these accountability challenges. DAOs are organizations run by member consensus with rules encoded in smart contracts on a blockchain. A DAO is fundamentally characterized by three enabling features: transparency and contestability, sanctionability, and open participation. Transparency and contestability mean that all proposals and decisions are visible to the public, deliberation forums are accessible, and members can amend decisions through follow-up votes in a continuous process of challenge and improvement. Sanctionability refers to the built-in capacity to enforce consequences, such as slashing tokens or blocking malicious actors from future participation. Open participation means that membership is accessible to all who hold the token, without gatekeeping by a central authority.

Decisions in a DAO are executed automatically by code once approved by voters, creating a tamper-proof, transparent ledger of every action. In principle, this algorithmic governance model makes all transactions and votes visible to members, theoretically strengthening accountability through transparency and traceability. Each decision is recorded immutably and linked to the stakeholders who voted on it, clearly tracing responsibility for outcomes. Moreover, by automating rule enforcement, DAOs introduce built-in sanctionability in the form of smart-contract constraints and penalties (Wright and De Filippi, 2015; Hassan and De Filippi, 2021).

While the theoretical benefits of DAOs have led to international experimentation, Canada’s public sector has lagged in exploring blockchain-based governance. The Treasury Board of Canada Secretariat’s "Canada’s Digital Government Strategy" (2022) and "Digital Standards Playbook" (2023) omit DAOs and related distributed governance models. Other jurisdictions are advancing: the United Nations Internet Governance Forum 2023 coalition launched a pilot DAO for public-sector governance, and Wyoming has passed legislation enabling DAO formation. Canadian public policy has not yet grappled with this possibility, and no empirical study has compared the de facto governance patterns of DAO systems with those of Canadian public institutions.

The omission reflects a broader gap in the comparative governance literature. Existing DAO research focuses on whether these systems deliver on their democratic promises (Azouvi et al., 2019; Gilson and Bouraga, 2024) or on their technical architecture (Hassan and De Filippi, 2021). Existing public administration research examines institutional accountability through established frameworks (Bovens, 2007; Mulgan, 2000). No empirical study bridges these literatures by measuring the same governance properties across both system types and asking whether the systems face similar pressures despite their architectural differences.

The DAO literature’s default framing positions decentralized governance as an alternative to representative institutions: direct participation replaces hierarchical delegation, and encoded rules replace administrative discretion. Empirical evidence complicates this framing. Participation rates in major DAOs reach only 1 to 2 percent of token holders (Liu, 2023). Delegation concentrates effective voting power in specialized agents (Polkadot OpenGov). Traditional institutions, when motivated by crisis, can achieve distributional reforms comparable to those promised by technological alternatives (the Canada Council’s COVID-era Gini reduction). These patterns suggest a bounded thesis: DAO and traditional governance may face similar pressures while exposing those pressures through different visibility properties.

This study tests that proposition. The primary research question is: Do DAO governance mechanisms produce governance patterns materially different from traditional public institutions, or do they face similar pressures with different transparency properties? A secondary question follows: What does the answer imply for hybrid governance design in the Canadian public sector?

The focal case is grant-making at the Canada Council for the Arts, a federal Crown corporation distributing approximately $300 million annually. The comparators are Polkadot OpenGov (on-chain referendum governance) and Gitcoin Grants (quadratic funding). The study develops and applies a measurement framework (TI/C/S) that operationalizes auditability and author-specified adaptive capacity across both system types, uses a descriptive temporal analysis of the CCA’s COVID-era distributional shift, and examines delegation and participation dynamics in the DAO comparators. The contribution is a systematic cross-system comparison that identifies parallel governance pressures despite divergence in institutional design, and that locates the policy-relevant distinction in the transparency infrastructure surrounding those pressures.

Modern DAOs instantiate the concept of digital democracy by distributing governance power among stakeholders through blockchain technology. Beck, Mueller-Bloch, and King (2018) define a governance framework for blockchain-based systems characterized by decision rights (who can vote), transaction logs (immutable records of all actions), and token incentives (rewards for participation and penalties for violation). In a DAO, token-holders or members collectively propose and vote on decisions, and the agreed outcomes are executed automatically via smart contracts. This design is meant to boost participation and reduce hierarchical gatekeeping in decision-making. On-chain enforcement mechanisms ensure that approved decisions are executed without human intervention, making rules tamper-proof. The blockchain ledger provides a single, incorruptible source of truth about proposals, votes, and fund flows, making it possible to audit who did what and when.

The ideal of a democratically governed, transparent DAO can diverge sharply from reality. Emerging research and early DAO experiments reveal significant pitfalls alongside their benefits. Participation inequality is one major concern. Azouvi, Maller, and Schmid (2019) found that ostensibly democratic token-voting in DAOs often concentrates power in the hands of a few large stakeholders; participation rates in major DAO communities reach only 10 to 20 percent. Liu (2023) reports that in contemporary DeFi DAOs, average participation rates drop to 1 to 2 percent of token holders, consistent with Azouvi, Maller, and Schmid’s documented patterns. Gilson and Bouraga (2024) confirm that voter turnout remains below 25 percent in many DAOs even after several years of operation, indicating that low engagement is a structural feature rather than a growing pain. Low turnout combined with unequal token distribution produces de facto plutocracy. Lalley and Weyl (2018) and Buterin, Hitzig, and Weyl (2019) proposed alternative voting mechanisms like quadratic voting to mitigate inequality, but adoption remains limited in existing DAOs. Chohan (2023) situates these governance failures within the broader institutional context, arguing that DAOs face the same collective action problems as traditional organizations and that blockchain infrastructure does not inherently resolve them.

Beyond participation inequality, pseudonymity complicates accountability. Rong and Mao (2023, Belfer Center), examining CityDAO, document how anonymous participation initially seemed to empower diverse stakeholders, yet ultimately the project concluded that identity disclosure was necessary for greater accountability, trust, and legal enforceability. This finding highlights a fundamental tension in decentralized systems: anonymity protects privacy but undermines the ability to hold individuals accountable.

Historical cases show that "code as law" requires flexible safeguards and strong meta-governance. The 2016 collapse of The DAO after an exploit diverted millions of dollars exposed the brittleness of purely code-governed systems. Rather than relying on smart contracts to recover stolen funds, the community chose a hard fork of the entire blockchain to reverse the exploit, revealing that governance ultimately requires collective human decision-making and the authority to override code (DuPont, 2017). Wright and De Filippi (2015) and Crepaldi (2019) theorize this through the concept of "secondary rules," distinguishing between operational rules embedded in code and meta-governance rules that govern how the code itself can be amended.

The CityDAO case underscores similar challenges. While active anonymous participation brought enthusiasm and innovation, the project experienced a "vetocracy" problem: high coordination costs, passive-aggressive behavior in deliberations, and decision paralysis. Like The DAO, CityDAO concluded that robust governance requires mechanisms beyond blockchain code, including identity systems, formal dispute resolution, and structures that make participation less cognitively burdensome. These cases illustrate that well-designed DAOs benefit from strong meta-governance, mechanisms for emergency intervention, and methods to reduce coordination costs, whereas poorly-designed DAOs devolve into paralysis or capture.

Given these lessons, exploration of DAOs in public administration has focused on domains where their transparency and automation benefits are greatest. Proposals target public procurement, aiming to reduce discretionary bias through immutable ledgers and open bidding, and participatory budgeting, intended to strengthen democratic legitimacy through direct citizen voting on budget allocations. Rikken et al. (2022) provide the most direct treatment of DAOs as citizen participation mechanisms in Public Administration Review, arguing that decentralized autonomous structures can formalize participatory governance while acknowledging that meaningful adoption requires resolving legal status, identity verification, and integration with existing administrative hierarchies. Van Kerckhoven (2023) surveys the emergence of DAOs in public policy across jurisdictions, concluding that the regulatory environment remains the primary determinant of adoption speed and that most experimentation has occurred in jurisdictions with permissive blockchain legislation.

Van Vulpen and Jansen (2023) advance a community-centric DAO design grounded in Ostrom’s (1990) principles for governing commons, proposing that DAOs succeed as governance instruments when they incorporate graduated sanctions, conflict resolution mechanisms, and boundaries that define a stable membership community. Their framework is directly relevant to the institutional ecology question this study raises: whether governance mechanisms designed for anonymous, globally distributed, financially motivated participants can function within a nationally bounded, professionally interconnected cultural funding body. The Ostrom-inspired design principles suggest that successful transplantation requires explicit adaptation to the stakeholder community’s characteristics rather than a generic governance template.

Public sector experiments with DAOs remain nascent and incremental. Zug, Switzerland piloted blockchain voting for municipal referenda. Wyoming passed legislation enabling DAO formation and registration, paving the way for legal recognition. No government has yet deployed a DAO for core administrative functions. DAOs are better understood as complementary tools that augment existing institutions.

To systematically compare governance systems, this research introduces the TI/C/S Accountability Toolkit, operationalizing Bovens’s (2007) three-dimensional accountability framework through four rubric dimensions. Three dimensions capture auditability: Traceability (transparency and immutability of records), Contestability (formal mechanisms to challenge and review decisions), and Sanctionability (enforcement and durability of penalties). A fourth dimension, Adaptive Capacity, is introduced in this study to capture the institutional cost and speed of modifying governance rules in response to novel conditions. The adaptive-capacity scores are author-specified rubric judgments anchored in observable institutional features, not directly observed empirical quantities. The inclusion of adaptive capacity is a deliberate design choice: auditability-only frameworks structurally favor systems engineered for transparency (blockchain-based systems) over systems engineered for discretionary flexibility (traditional bureaucracies). Adaptive capacity introduces a dimension on which the ordering reverses, producing a mixed comparison that reflects the trade-off between visibility and responsiveness. Each dimension is scored on a 0-to-1 scale from publicly inspectable institutional features, making the framework replicable as a scoring protocol while leaving component-level judgments open to reviewer disagreement.

To evaluate accountability through these lenses, we employ a comparative methodology. We operationalize each dimension as a composite rubric index (ranging from 0 to 1) based on publicly inspectable features of each system’s governance structures. The auditability branch is reported separately from the four-dimension overall score so that author-specified adaptive-capacity judgments do not disappear into a single scalar.

The analysis uses the TI/C/S Accountability Toolkit, a framework that operationalizes institutional accountability through four dimensions: Traceability (observability and durability of records), Contestability (capacity to challenge or revise decisions), Sanctionability (enforceability of rules and consequences), and Adaptive Capacity (speed and discretion in rule modification). The TI/C/S toolkit translates Bovens’s (2007) accountability framework into rubric-scored institutional features, enabling systematic comparison of traditional and blockchain-based governance systems while preserving a distinction between observed quantities and author-assigned scores. The paper combines qualitative inquiry, quantitative analysis, and the TI/C/S framework to evaluate how DAO mechanisms affect accountability. The research employs four integrated components:

The Canada Council for the Arts was selected as the focal case study for three reasons: (1) Substantive fit: its grant-making process is a quintessential public governance function involving transparent criteria, peer review, and published outcomes, making it well-suited to comparative analysis with DAO mechanisms; (2) Organizational readiness: the Council publishes comprehensive data on grants, enabling empirical analysis; (3) Data accessibility and generalizability: the Council’s grant data is publicly available and large enough to enable robust statistical inference (58,277 grants), and the Council’s governance challenges are representative of broader issues in public administration.

Inequality Metrics: We compute multiple inequality measures to assess the distribution of funding and voting power. The Gini coefficient is calculated as:

where $x$ is the sorted vector of grant amounts or voting power. We also compute the Herfindahl-Hirschman Index ($\text{HHI} = \sum s_i^2$), Theil T-index ($T = \frac{1}{n} \sum \frac{x_i}{\bar{x}} \ln \frac{x_i}{\bar{x}}$), Atkinson index at multiple inequality aversion parameters ($\varepsilon = 0.5, 1.0, 1.5$), and Palma ratio (top 10% share / bottom 40% share). All estimates include 95% bootstrap confidence intervals calculated from 5,000 resamples using the percentile method with seed $= 42$.

Temporal Analysis: The CCA’s COVID-era distributional shift is analyzed as a descriptive temporal sequence. Three segments are defined: pre-COVID (2017–2019, Gini stable at 0.657–0.660), pandemic response (2020–2021, Gini declining to 0.611 and 0.551), and reversion (2022–2023, Gini rising to 0.602 and 0.620). With seven annual observations and three segments, the design supports descriptive trend characterization; causal attribution remains outside the design. The sequence detects a level shift coinciding with the pandemic-response period. Policy design changes (broadened eligibility, simplified processes), budget expansion (68 percent increase), and applicant behavior shifts occurred simultaneously. The overlap among these mechanisms is analytically important: distributional shifts in public grant-making cannot be attributed to a single policy lever with aggregate data.

Statistical Tests: For the Polkadot voting-power inequality subsample (N = 20 referenda), we use exact small-sample non-parametric inference rather than normal approximations. Mann-Whitney U tests use exact p-values for approval-status comparisons in referendum duration. For the delegation analysis, Polkassembly vote totals were retrieved for the full binary-outcome population (N = 1,564). A separate 105-referendum systematic sample, using every 15th referendum from offset 0, was retained as a verification bridge because it was the first live-audited branch. Spearman rank correlations are computed against referendum order to test temporal trends in delegation share across the June 2023 to March 2026 observation period. Bootstrap confidence intervals for the mean delegation share are computed from 5,000 resamples using the percentile method with seed = 42.

TI/C/S Operationalization: Each dimension score is computed as the unweighted mean of $k$ observable components:

For the CCA, traceability components include grant amounts published ($c_1 = 1$), recipient names published ($c_2 = 1$), program criteria published ($c_3 = 0.5$), deliberation records ($c_4 = 0$), and rejection data ($c_5 = 0$), yielding $\text{TI} = 0.500$.

Adaptive Capacity Operationalization: Adaptive capacity is an author-specified rubric score. It is scored as the unweighted mean of four components: (1) rule modification speed (time from decision to implementation of a governance rule change), scored inversely on a 0-to-1 scale where faster modification scores higher; (2) authorization requirements (number of veto points or approval stages required), scored inversely; (3) scope of discretion (whether the decision-making authority can modify rules, criteria, and resource allocation simultaneously), scored directly; and (4) documented crisis response (whether the system has visibly modified governance rules in response to an exogenous shock within the observed period), scored as binary. For the CCA, these author-assigned components yield: rule modification speed (0.90, ministerial/administrative authority permits changes within weeks), authorization requirements (0.80, hierarchical approval but no external consensus process), scope of discretion (0.95, eligibility, process, and funding levels modified simultaneously during COVID), documented crisis response (1.0, COVID-era redesign documented). CCA adaptive-capacity rubric score: 0.913. For Polkadot, the components yield: rule modification speed (0.40, minimum 15 days for approved referenda, median 28 days for contested), authorization requirements (0.35, proposal, deliberation, quorum, on-chain vote required), scope of discretion (0.50, parameters modifiable through governance but protocol upgrades require runtime changes), documented crisis response (0.50, governance has responded to exploits and parameter adjustments, though no single event comparable to COVID-era institutional redesign). Polkadot adaptive-capacity rubric score: 0.438. For Gitcoin, the components yield: rule modification speed (0.60, round parameters set by team with community input between rounds), authorization requirements (0.55, team authority for operations, governance process for protocol changes), scope of discretion (0.60, round matching caps, eligibility, and mechanism modifiable between rounds), documented crisis response (0.50, platform migration from centralized rounds to Grants Stack/Allo Protocol, showing architectural adaptability). Gitcoin adaptive-capacity rubric score: 0.563.

The framework scores institutional design features, not stakeholder experience. Different weighting schemes (privileging outcome accessibility, counting informal appeals, measuring lived experience) would narrow the auditability gap but do not eliminate it. The adaptive-capacity rubric reverses the ordering: the CCA scores highest (0.913), followed by Gitcoin (0.563) and Polkadot (0.438). This mixed ordering is the intended design contribution; it reveals the trade-off between auditability and responsiveness rather than a unidirectional superiority claim.

The TI/C/S framework scores auditability and adaptive capacity, the institutional infrastructure of accountability. It does not measure governance quality in a broader sense. Several dimensions that bear on governance performance fall outside the framework’s scope, and the scores should be interpreted accordingly.

Expert judgment quality. The CCA’s peer review process embeds professional expertise accumulated through years of engagement with the arts sector. Reviewers apply contextual knowledge about artistic trajectories, regional cultural ecosystems, and the developmental stage of applicants. The TI/C/S framework scores the opacity of this process (low traceability, low contestability) without assessing whether opacity serves an epistemic function by protecting candor and enabling nuanced evaluation. A governance system that scores low on contestability because expert judgment is insulated from political pressure may, under certain conditions, produce better decisions than one that scores high because every decision is subject to popular vote.

Asymmetric risk exposure. The framework measures whether sanctionability mechanisms exist, not whether decision-makers bear consequences proportional to their decisions. In the CCA, peer reviewers face no financial downside from funding underperforming projects. In Polkadot, token holders vote their economic interest, but that interest is the token’s market price rather than the quality of funded work. A framework that measured the degree to which decision-makers are exposed to the downside of their decisions (Taleb, 2018) might produce a different ordering, or might find that all three systems score poorly on this dimension.

Institutional longevity and survival selection. The CCA has operated since 1957. Polkadot OpenGov and Gitcoin Grants have operated in their current form for two to five years. The comparison selects on surviving DAO implementations; the base rate of DAO failure is not computed here and would contextualize the scores of the two DAO comparators. Ding et al. (2023) provide the most comprehensive survey of DAO governance structures and their durability.

Stakeholder welfare. The framework does not measure whether governance arrangements produce outcomes that serve the interests of the relevant stakeholder community. Grant distribution equality (measured by Gini) is a distributional metric, not a welfare metric. A perfectly equal distribution of funding to projects of random quality would score well on inequality measures and poorly on cultural impact.

These exclusions are methodological choices, not oversights. The TI/C/S framework was designed to score a specific set of institutional properties that are publicly inspectable and comparable across systems. The excluded dimensions require different data, different methods, and in some cases different epistemological commitments. Readers should treat the TI/C/S scores as author-assigned measures of institutional visibility and formal accountability infrastructure, and should not infer from a high score that a governance system produces superior outcomes or serves its stakeholders well.

The Canada Council distributed $2.23 billion across 58,277 approved grants between 2017 and 2024. The overall Gini coefficient is 0.619 (95% CI [0.613, 0.625]), indicating substantial inequality in funding allocation. This figure exceeds Canada’s income Gini of approximately 0.32, suggesting that arts funding is considerably more concentrated than national income. The Theil index of 0.826 (95% CI [0.794, 0.862]) and Palma ratio of 7.80 (top 10% receives 7.8 times the bottom 40%) confirm this concentration. The HHI of 0.00011, by contrast, indicates that no single recipient dominates, suggesting that concentration is driven by a wider distribution across mid-tier recipients rather than monopoly by one or a few actors.

The COVID-19 pandemic period (2019–2021) is treated as a descriptive temporal sequence in equity-oriented policy design. Between 2019 and 2021, the Gini coefficient declined from 0.660 to 0.551, a reduction of 16.5 percent. The shift coincided with deliberate policy choices by the Canada Council: emergency funding programs broadened eligibility, simplified application processes, and targeted individual artists. The 68 percent increase in total funding (from $272M to $457M) during this period makes disentangling policy design effects from budget expansion effects difficult; the subsequent reversion of Gini from 0.551 to 0.620 when funding contracted suggests budget scale was a significant driver. Grant counts increased 26 percent (from 7,866 to 9,904), total funding rose 68 percent (from $272M to $457M), and median grant size grew from $15,590 to $25,000.

By 2023, inequality had reverted to 0.620, suggesting that structural factors driving concentration reasserted once emergency measures were withdrawn. The reversion deserves explanation: the COVID-era Gini reduction was driven by specific, temporary mechanisms: broadened eligibility criteria that admitted applicants who would normally fall outside program scope, simplified application processes that reduced barriers for smaller organisations, targeted emergency funding for individual artists, and a 68 percent increase in total funding (from $272M to $457M) that allowed more grants without reducing awards to established institutions. When emergency funding was withdrawn, eligibility criteria narrowed, and total funding contracted, the structural features of arts funding (large institutions with recurring multi-year grants, peer review processes that favour established organisations with track records, and program designs calibrated to institutional rather than individual applicants) reasserted. The reversion suggests that the COVID improvement reflected a temporary expansion of the funding pool rather than a durable change in distributional architecture. Sustaining the lower Gini would have required permanent changes to program design, eligibility criteria, and funding levels. This finding identifies both the potential and the limits of traditional institutional reform: equity improvements are achievable within existing structures, while persistence requires sustained institutional commitment rather than emergency-driven expansion.

The reversion exposes a tension between two governance properties that the DAO literature has not adequately distinguished: adaptive capacity and rule durability. The CCA’s COVID-era response provides the basis for a high adaptive-capacity rubric score: the institution rapidly modified rules, criteria, and resource allocation within existing administrative structures. That capacity is a governance property, and one that smart-contract-encoded systems sacrifice by design. A DAO-governed grant system operating under the same conditions would have required a governance vote, quorum achievement, and potentially a protocol upgrade to broaden eligibility or redirect funding at comparable speed. The trade-off is material: encoded rules prevent silent reversals (durability) but constrain rapid adaptation to novel conditions (optionality). In a domain characterized by periodic fiscal shocks and shifting political priorities, as Canadian public administration is, the value of each property depends on whether the governance environment is stable enough to favor durability or volatile enough to favor adaptive capacity. The COVID episode suggests the latter.

The reversion also shows a political problem that institutional design alone cannot resolve: governments do not sustain equity-focused budgets when emergency conditions end. A DAO can automate distribution rules, but it cannot prevent governments from changing the rules or reducing appropriations. The obstacle is political commitment, and encoding rules in a smart contract does not solve political commitment failures. This limits the case for DAO adoption to the narrower claim: DAO mechanisms can improve the transparency and auditability of grant-making, which may indirectly support political accountability by making policy reversals more visible and raising their political cost.

The COVID-era temporal analysis measures redistribution, not transparency, contestability, or sanctionability. A complete comparison between traditional and DAO-based policy would require measuring whether emergency-era policy changes also improved accountability scores using the TI/C/S framework. Without this measurement, the temporal analysis shows equity gains but does not establish accountability equivalence.

The top 10 recipients account for approximately $120 million over the 2017–2024 period, led by the National Ballet of Canada ($24.1M), the Canadian Opera Company ($17.6M), and Les Grands Ballets Canadiens de Montreal ($12.9M). These major arts institutions dominate funding, while smaller independent artists and emerging organizations receive substantially smaller amounts.

To assess how DAO-based governance could be applied to the Canada Council’s grant-making process, we constructed a detailed scenario workflow combining elements of both traditional accountability and blockchain-enabled features. The proposed model would function as follows:

Integration of DAO mechanisms into the CCA’s grant process would substantially alter the three accountability dimensions:

Traceability: A DAO model would raise CCA traceability from 0.50 to approximately 0.80 or higher. All proposals, scores, votes, and decisions would be recorded on an immutable public ledger. Reviewer anonymity could still be preserved to reduce pressure on assessors, though that choice would narrow accountability for the quality of scoring decisions. Public reasoning would therefore need to include score rationales alongside final outcomes. That would improve visibility while adding work for reviewers and narrowing the space for candid feedback.

Contestability: A DAO model would substantially increase contestability from 0.12 to approximately 0.70-0.90. Community members could propose reconsideration votes, and a later consensus could reverse an earlier decision. That would formalize a process that is currently ad hoc. Repeated votes on the same decision would also create pressure toward gridlock and decision fatigue, especially when many proposals move through the system at once.

Sanctionability: DAO integration would allow automated enforcement of grant compliance through smart contract penalty mechanisms such as fund blocking or milestone gating. In a public sector context, those mechanisms face legal constraints. Canadian administrative law requires proportionality in penalties and opportunities for appeal, so blanket automated enforcement would sit uneasily with procedural fairness. Sanctionability in a hybrid model would increase to approximately 0.60-0.80 if smart contracts handle routine compliance while human oversight remains in place for penalty decisions.

Overall, a DAO-based model could increase CCA auditability from 0.238 to approximately 0.55–0.70 on the three-dimension auditability mean, depending on implementation details. Whether this auditability improvement translates into improved governance outcomes is the open empirical question this study identifies but does not resolve. The gain in auditability comes with trade-offs against adaptive capacity: greater transparency can chill peer review, expanded contestability can slow decisions, and automated enforcement can collide with the proportionality and procedural fairness requirements of Canadian administrative law. A hybrid model that selectively adopts auditability mechanisms while preserving the CCA’s adaptive capacity would need to balance these dimensions rather than maximizing auditability alone.

The four-dimension comparison produces a mixed ordering. DAO systems outscore the CCA on all three auditability rubric dimensions: the largest gaps are in contestability (CCA: 0.113 vs. DAO average: 0.967) and sanctionability (CCA: 0.100 vs. DAO average: 0.917). The author-specified adaptive-capacity rubric scores the CCA above both DAO systems (0.913 vs. 0.438 for Polkadot, 0.563 for Gitcoin). No system achieves coordinate-wise dominance across all four dimensions. The four-dimension overall scores (CCA: 0.406, Polkadot: 0.809, Gitcoin: 0.849) compress the gap substantially compared with the auditability-only scores (CCA: 0.238, Polkadot: 0.933, Gitcoin: 0.944). The compression reflects the trade-off the framework was designed to capture: systems engineered for transparency sacrifice responsiveness, and systems engineered for discretionary flexibility sacrifice visibility. The auditability mean is reported alongside the four-dimension overall to preserve comparability with the three-dimension analysis in the robustness checks.

A scenario analysis estimates how the CCA’s auditability scores would change under selective adoption of DAO-inspired mechanisms. Each scenario is computed through explicit component substitution: the CCA’s current TI/C/S component scores are modified at the component level (e.g., "deliberation records: 0 → 0.75 under Scenario B"), and the composite is recomputed. The improvement ranges in Table 4 reflect the lower and upper bounds of defensible component-level rescoring. A reader who disagrees with a specific component assignment can substitute their own value and recompute; the method is transparent and replicable by construction.

The predicted improvements in CCA auditability range from +0.08 to +0.22 depending on which combination of mechanisms is adopted. Scenario C (Community Signaling), which most closely resembles a DAO model, yields the largest improvement (+0.12 to +0.22) because it strengthens both contestability and transparency. Each scenario carries a trade-off against adaptive capacity: formalizing appeals (Scenario C) introduces procedural requirements that slow decision cycles; encoding milestone compliance (Scenario D) constrains the discretionary flexibility that produced the CCA’s COVID-era response. The counterfactual analysis measures projected auditability gains without netting out the adaptive capacity cost, which would require empirical observation of the mechanisms in operation.

All inequality measures confirm high CCA inequality relative to both DAO systems. The Gini coefficient of 0.619 is substantially higher than Polkadot’s 0.079 and Gitcoin’s 0.285. The Atkinson index across multiple inequality aversion parameters (ε = 0.5, 1.0, 1.5) also shows consistent ordering: CCA exhibits the highest inequality at every level of inequality aversion. Bootstrap stability analysis reveals that the coefficient of variation of the Gini estimate is less than 0.3 percent, indicating robust point estimates across resampling. The CCA upper tail is fat-tailed rather than thin-tailed: a separate tail audit yields Hill exponents between 1.68 and 1.82, and trimming the top 1 percent of grants lowers the Gini to 0.561. The comparative ordering is unchanged, but the headline Gini should be interpreted as tail-sensitive rather than sufficient on its own.

Unit incommensurability caveat. The Gini comparisons measure different units: grant amounts (CCA), voting power (Polkadot), and donations (Gitcoin). These are resource concentration measures within each system’s primary allocation mechanism, not comparable measures of the same underlying quantity. A national ballet company receiving $24 million and an individual artist receiving $5,000 represent legitimate variation in the scope, cost, and institutional requirements of funded activities; the CCA’s high Gini may reflect appropriate heterogeneity rather than inequity. The cross-system Gini table (Table 5) is illustrative of how resource concentration patterns differ across institutional types. It should not be read as evidence that Polkadot or Gitcoin distribute resources more equitably than the CCA in any normatively meaningful sense. The within-system temporal analysis of CCA Gini (Table 2) is the methodologically sound comparison, because the unit (grant amount in CAD) is consistent across years.

Weighting scheme definitions: Equal assigns 0.25 to each of the four dimensions. Auditability-Only uses the original three-dimension equal-weighted mean (adaptive capacity excluded). Adaptability-Heavy assigns 0.4 to adaptive capacity and 0.2 to each auditability dimension. Visibility-Heavy assigns 0.4 to traceability and 0.2 to each remaining dimension. Compound Hostile (4-dim) applies the workflow compound_hostile component-rescoring scenario to the three auditability dimensions, then adds adaptive capacity at equal weight.

Sensitivity analysis across four positive weighting schemes plus one compound hostile rescoring branch shows that the ordinal ranking is robust but the gap magnitude is assumption-sensitive. The CCA’s four-dimension positive-weight score ranges from 0.406 (equal weighting) to 0.508 (adaptability-heavy), and its compound-hostile score is 0.384. DAO positive-weight scores range from 0.735 to 0.944, while the compound-hostile branch lowers Polkadot to 0.580 and Gitcoin to 0.611. Under the four-dimension framework, coordinate-wise dominance no longer holds: the CCA outscores both DAOs on the author-specified adaptive-capacity rubric. The CCA still ranks third under every positive weighting scheme and under the compound hostile branch, but the distance between systems compresses substantially when adaptive capacity is weighted at 0.4 (CCA at 0.508 vs. Polkadot at 0.735, a gap of 0.227 compared with 0.696 under auditability-only scoring). This compression reflects the trade-off between visibility and responsiveness that the four-dimension framework was designed to capture. Under the auditability-only branch, the workflow component-rescoring and evidence-extraction audits remain valid: both DAO systems outscore the CCA on every auditability dimension, and the ordinal ranking survives all tested adversarial scenarios. The four-dimension comparison adds the rubric-based claim that the CCA’s governance advantage lies in a dimension that the original three-dimension framework did not measure.

A separate adaptive-capacity rescoring pass treats Appendix A4 as an interrater-style problem rather than a settled measurement. Under a moderate hostile scorer, the CCA-first adaptive-capacity ordering survives (CCA 0.863, Gitcoin 0.625, Polkadot 0.438). Under a deliberately DAO-favorable stress scorer, the ordering changes (Gitcoin 0.775, CCA 0.675, Polkadot 0.650). The adaptive-capacity dimension therefore should be read as a transparent trade-off lens. It is plausible under the author’s and moderate-auditor scoring, while less invariant than the auditability branch.

Polkadot and Gitcoin score 0.93–0.94 on the published auditability branch, while the Canada Council scores 0.24. The gap is large under the stated rubric and robust across tested weighting schemes. The Council’s own record complicates the inference that should be drawn from it.

During the 2020–2021 emergency redesign, the Council’s Gini fell from 0.660 to 0.551 while grant counts and median awards rose substantially. The change came from administrative rule changes: broadened eligibility, simplified processes, targeted emergency funding. No new technology was required. The CCA displayed what might be called administrative optionality: the capacity to exploit an unforeseen stressor by rapidly modifying rules, criteria, and procedures within existing institutional structures. That capacity is itself a governance property, and one captured by the author-specified adaptive-capacity rubric.

The later return toward pre-pandemic concentration (Gini 0.620 by 2023) matters as much as the temporary improvement. Emergency procedures, broader eligibility, and additional funds were discretionary measures tied to a specific political moment. Once that moment passed, the structural features of arts funding reasserted: large institutions with recurring multi-year grants, peer review processes that favor established organizations, and program designs calibrated to institutional rather than individual applicants. The reversion separates two facts that should not be collapsed into one: the Council can produce a more equal distribution through internal redesign, and that redesign will loosen once the surrounding political conditions change.

DAO-style mechanisms address this durability problem by encoding rules in infrastructure that resists discretionary modification. Encoded rules, however, trade adaptability for persistence. A smart-contract-governed grant system would have required a governance vote, quorum, and potentially a protocol upgrade to achieve the same rapid adaptation the CCA executed administratively during the pandemic. Durability prevents silent policy reversals but constrains the system’s capacity to respond to novel conditions. Which property is more valuable depends on the volatility and predictability of the governance environment. In a domain characterized by periodic fiscal shocks and shifting political priorities, the cost of reduced adaptive capacity may exceed the benefit of encoded rule persistence.

The binding constraint, as the COVID episode illustrates, is political commitment rather than institutional design. Encoding rules in a smart contract does not compel a government to sustain funding levels. It can make a reversion more visible and raise the political cost of reversing commitments without public justification. That contribution is bounded. The case for DAO adoption in public administration rests on the narrower claim that improved visibility supports political accountability, not the broader claim that encoded rules solve political commitment failures.

The relationship between transparency and political commitment has empirical grounding outside the blockchain context. Fisman and Gatti (2002) find, across 50 countries, that fiscal decentralization reduces corruption by approximately 0.1 per unit increase, a result they attribute to the accountability effects of bringing decision-making closer to observation. Diamond (1999) argues that democratic consolidation depends on institutional mechanisms that raise the cost of norm violations, not on the elimination of political self-interest. Both frameworks support the narrower claim: institutional transparency can constrain political discretion by making deviations from stated policy visible. Neither supports the stronger claim that transparency mechanisms compel sustained political commitment. The CCA’s COVID reversion is consistent with both: the policy reversal occurred despite being observable in public data, suggesting that visibility alone is insufficient without political constituencies willing to enforce the norm. DAO mechanisms may lower the monitoring cost for those constituencies without guaranteeing their mobilization.

Polkadot’s delegation pattern constitutes this study’s most consequential empirical finding for the comparative governance literature. In the full binary-outcome population, the mean delegation share is 91.7 percent (95% CI [90.8%, 92.5%]), with a median of 99.1 percent and a positive temporal trend across the June 2023 to March 2026 observation period (manual Spearman rho = 0.607, N = 1,564). Effective voting power is exercised by specialized delegates rather than individual token holders.

This representative drift reframes the comparison between DAO and traditional governance. The standard framing in the literature positions DAOs as direct-democratic alternatives to representative institutions: stakeholders vote directly rather than delegating to elected officials. The Polkadot data complicate that framing. Delegation is the rational response to high information costs in any governance system (Downs, 1957). When the cognitive burden of evaluating each proposal exceeds the expected influence of a single vote, rational participants delegate. That dynamic operates in traditional democracies, corporate governance, and DAO governance alike.

The analytically relevant question is whether delegated governance in DAOs differs structurally from representative governance in traditional institutions. Three properties distinguish DAO delegation from traditional representation. First, transparency: delegation in Polkadot is a public, on-chain act, whereas the influence of advisors, lobbyists, and party structures on traditional representatives is substantially opaque. Second, reversibility: token holders can reassign their delegation at any time with minimal friction, whereas replacing elected officials requires waiting for an election cycle. Third, granularity: token holders can delegate to different specialists on different proposal tracks, whereas citizens cast a single vote for a bundled representative platform.

These properties do not eliminate the representative structure; they make it more inspectable and more responsive to principal dissatisfaction. The governance advantage of DAOs over traditional institutions may lie less in the elimination of representation than in the quality of the principal-agent relationship within the representative structure: lower monitoring costs, lower switching costs, and higher-resolution delegation choices.

Gitcoin’s participant dynamics add a complicating dimension. With 76.9 percent of participants in any given round being new users and only 4.4 percent returning from previous rounds, Gitcoin’s governance community is transient rather than stable. Sustained democratic governance requires a persistent community with accumulated institutional knowledge and relational trust (Ostrom, 1990). Whether a system with extremely high turnover per cycle constitutes a governed community or a recurring crowd is an open question that the accountability scores alone cannot resolve.

The delegation concentration finding and the Gitcoin churn finding together suggest that DAO governance faces many of the same structural pressures as traditional governance: concentrated decision-making by specialists, low participation by the broader membership, and difficulty sustaining engaged community over time. The advantage lies in the infrastructure’s capacity to make these dynamics visible and measurable rather than in eliminating them.

For public sector organizations like the Canada Council, the comparison points toward a bounded hybrid model. The strongest candidates are the mechanisms that increase visibility and contestability without displacing statutory oversight.