Experiment 2 · Aave DAO

the participation gap

DAOs promise decentralised democracy — but most members never show up. This story measures the gap between who can vote and who actually calls the shots on-chain.

Setting the Stage

DAOs optimise for proposal passage, not downstream participation — most “governed” members never vote. There are 950 closed Snapshot votes on aavedao.eth. Let’s see who actually shows up to decide them.

Permanently Low Turnout

Across all 950 proposals, low turnout is a permanent fixture, not a passing phase. While a few outlier votes spike, the last-12-month median is just 126 unique wallets per vote. The crowd is consistently absent.

Does the Topic Matter?

Treasury votes draw the most individual wallets because people naturally track the money — a median of 1,305 wallets, versus 530 for “other” topics. Yet even the highest-turnout categories mobilise only a tiny fraction of the eligible community. VP turnout stays flat regardless of topic.

Fewer Hands on the Wheel

A small crowd shows up, but an even smaller group decides. While hundreds of wallets might sign a proposal, the ultimate decision power is hyper-concentrated. Across the last 12 months, the median Nakamoto coefficient — the fewest wallets needed to command a strict majority of votes actually cast — is exactly two.

The Permanent Committee

Low headcount isn’t random churn. In the last year, 1,273 unique wallets voted at least once — but the top of the leaderboard barely missed a beat. Three wallets voted on every single one of 102 proposals. The next ten still showed up 98 times or more. Governance isn’t a crowd that occasionally stays home. It’s a standing committee with near-perfect attendance.

The Massive Illusion of Community

Let’s look at the people behind the numbers. Each dot represents roughly 600 unique wallets holding the power to govern — including token holders, stakers, and delegates. Together, they form an eligible community of nearly 60,000 wallets.

The Grey Majority

This is the reality of decentralised governance. The median turnout is just 1.3% of eligible wallets. The remaining 98.7% of the community stays completely grey and unactivated. Let’s zoom into a single real vote to see who actually occupies those few coloured dots.

A Real Vote

Consider the historic Aave Will Win Framework Temp Check, where Aave Labs requested up to ~$42.5M in stablecoins and 75k AAVE. By Aave standards, this was an exceptionally high-turnout vote — 302 unique wallets, roughly 2× the recent 12-month median of 126. Who were they?

Who Are They?

On paper, 302 wallets decided the future of a multi-billion-dollar protocol. Looking at this crowd, everyday individual wallets vastly outnumber the institutional players. But treating every wallet as equal is a structural illusion. Let’s resize them by their true voting power.

The Scale of Influence

On the Aave Will Win Temp Check, even among active voters, power is highly unequal. “FOR” won by individual wallet count, but a tiny handful of major institutional delegates wielded enough raw voting power to pass or sink the entire framework. A single group, the ACI, cast 25.9% of all voting power on the “AGAINST” side, nearly swinging a multi-million-dollar vote alone.

Anonymous Veto Power

The gap gets even sharper when rules are on the line. On a proposal requiring mandatory Conflict-of-Interest (COI) disclosures, a tiny group of just 31 wallets defeated a crowd of 138. Remarkably, three completely anonymous wallets held 42.5% of all voting power in the proposal — roughly four-fifths of the AGAINST coalition — using their anonymity to kill a rule about transparency.

What the Data Earned

Aave reaches roughly 60,000 wallets with governance power. In the last year, a median of 126 showed up per vote. Of those, two wallets typically held majority control.

This is not a broken system. It is a settled one. The permanent committee keeps the protocol running. The delegates absorb the complexity. The concentration is, in part, what makes fast, technically competent decisions possible at scale.

But a system that works is not the same as a system that is accountable. When three anonymous wallets can defeat 138 identified participants trying to pass a transparency rule, the gap is no longer about turnout. It is about whether the people with the most power have any obligation to the people they govern on behalf of. Right now, the answer is: only if they choose to.

Read Reflection →

Methods

Intervention

We pulled every closed off-chain vote on Aave DAO’s Snapshot space (aavedao.eth) [1] — 950 proposals in total — and measured three things: how many wallets were eligible to vote, how many actually did, and how concentrated the deciding power was among those who showed up.


For the most recent 12 months we went deeper, fetching every vote at wallet level to compute a per-proposal Nakamoto coefficient and track individual attendance across the year. Two proposals were selected as case studies — the Aave Will Win Framework Temp Check and the COI disclosure vote — and received full voter-level analysis including VP weight and actor classification.


A note on scope: these 950 are closed Snapshot votes, not forum discussion threads (Ayae Ide et al. [4] count ~2,700 forum threads for Aave), nor on-chain ProposalCreated events (~740 per that paper). Aave forum retrospectives report ~820 Snapshot proposals since 2020; our higher count reflects more recent closes captured by our fetch date.

Data sources

The pipeline fetches from three external sources and one hand-maintained seed:

Snapshot GraphQL API [1]
All closed proposals on aavedao.eth — title, created timestamp, snapshot block, unique voter count (votes field), total VP cast (scores_total), per-choice scores. For last-12-month proposals and case studies: per-voter address, VP, and choice. Display names for known addresses via the users query.
→ proposals.json, case_studies.json, recent_votes.json
Ethereum RPC [2]
Two reads against the Ethereum mainnet. First: AAVE.totalSupply() at each proposal’s snapshot block — used as an upper-bound proxy for eligible voting power. Second: Aave’s Gov V3 GovernancePowerStrategy.getFullVotingPower(address) [3] on a sample of holder addresses at recent snapshot blocks — used to estimate how many wallets had any voting power at all.
→ eligible_vp_by_block.json, wallet_metrics.json
Blockscout AAVE token holders [5]
Blockscout’s token-holder API for the AAVE contract provides a paginated list of non-contract holder addresses and reports a total holder count of ~205,000. We download ~10,000 holder addresses as a sample, then call getFullVotingPower on each to find what share have VP > 0. That rate is scaled to the full reported holder count to estimate total eligible wallets.
→ holder_addresses_sample.json, eligible_wallets_by_block.json
Seed labels (hand-curated)
A manually maintained list of known Aave governance actors — service providers (e.g. ACI with ~357 delegators), delegates, and the three case-study proposal IDs. These seed files are the ground truth for actor classification; everything else falls back to Snapshot display names or the VP-threshold heuristic below.
→ delegates.json (merged with Snapshot names)

Metrics

Wallets per vote
Snapshot’s votes field is the unique voter count per proposal — validated against paginated vote records on recent proposals. These differ because participation has declined over time; the story uses the recent figure as the more honest benchmark.
→ 767 all-time median · 126 last 12 months
Eligible wallets (~59,000, estimated)
We cannot check all ~205,000 AAVE holder addresses — that would require 205,000 RPC calls. Instead we survey-sample ~10,000 holders and call getFullVotingPower(address) [3] on each. This contract sums voting power across AAVE, stkAAVE, and aAAVE (own balance plus power delegated to that address). Addresses with power > 0 are “eligible.” We measured ~28.8% positive in sample, scaled: 205,000 × 28.8% ≈ 59,000. Wallets that delegated away return 0 and are excluded — power appears on the delegate’s wallet instead.
→ ~59,244 wallets · sample extrapolation
Wallet turnout (1.3%)
All-time median unique voters per proposal (767) divided by estimated eligible wallets (59,000). A ratio of two medians from different time windows — 767 is all-time, ~59k from recent blocks. Treat as order-of-magnitude, not precise.
→ 767 ÷ ~59,000 ≈ 1.3%
VP turnout (3.5%)
Voting power actually cast divided by AAVE totalSupply at the snapshot block. totalSupply is an upper bound — it does not subtract locked or unconfigured tokens — so true VP turnout is slightly higher. Frimpong et al. [6] report ~3.2% average on Aave on-chain votes (Dec 2020–Dec 2022), consistent with this.
→ ~3.5% median · totalSupply proxy
Nakamoto coefficient (median 2, last 12 months)
For each of 102 proposals, voters sorted by VP cast descending — count wallets needed to exceed 51% of total VP cast. We call this “per-vote Nakamoto.” Different from token-supply Nakamoto in concentration literature (Ayae Ide et al. [4]): ours asks how concentrated the outcome was among wallets that actually showed up.
→ median 2 wallets · >50% of VP cast
Repeat voters
For each address in any vote over the last 12 months, counted how many of 102 proposals they voted on. The 50%-threshold count and top-attendance list come from this per-address tally.
→ 72 wallets on 50%+ · 3 at 100% attendance
Actor classification
Priority order: (1) known actor from seed files; (2) Snapshot display name → delegate; (3) VP > 10,000 → unlabelled whale; (4) otherwise individual. Heuristic, not verified identity — same address may appear under different labels in different contexts.
→ service provider · delegate · whale · individual

Limitations

Snapshot off-chain layer only
Snapshot votes are signals, not binding execution. An ARFC or Temp Check that passes still requires a separate on-chain Aave Improvement Proposal. This story measures off-chain opinion, not on-chain outcome.
Eligible wallet estimate is extrapolation, not census
The ~59k figure compounds three assumptions: (a) Blockscout’s ~205k holder count is accurate; (b) our ~10k sample is representative; (c) VP-positive rate on recent Gov V3 blocks applies to all holders. Sample skews toward largest holders; getFullVotingPower fails on pre-V3 blocks; Blockscout covers AAVE token holders only — stkAAVE/aAAVE-only wallets may be absent.
Wallet turnout mixes time windows
1.3% divides all-time median voters (767, 2020–2025) by eligible-wallet estimate from recent Gov V3 blocks. The two inputs are not matched per proposal — best read as a rough floor.
VP turnout uses totalSupply as proxy
Exact eligible VP is the sum of each voter strategy output at snapshot block. We use AAVE totalSupply as stand-in — overstates eligible VP by including treasury, locked contracts, and addresses without governance access. True VP turnout is somewhat higher than 3.5%.
Aave Will Win is the Temp Check, not the binding ARFC
The vote analysed is the Snapshot Temp Check (~$42.5M stablecoins + 75k AAVE). The later binding ARFC revised the primary stable grant to ~$25M. Power dynamics are real for the Temp Check; outcomes may differ at ARFC stage.
COI 42.5% is share of total proposal VP
Three anonymous AGAINST wallets held 25.7%, 8.6%, and 8.2% of total proposal VP — summing to 42.5%. Within the AGAINST coalition alone, they held approximately 80% of that side’s voting power.

Resources

  1. Snapshot — off-chain governance voting platform. snapshot.org/#/aavedao.eth · GraphQL API: hub.snapshot.org/graphql
  2. Ethereum public RPC node used for totalSupply and getFullVotingPower reads. ethereum.publicnode.com
  3. Aave Governance V3 — GovernancePowerStrategy contract (BGD Labs). GovernancePowerStrategy.sol on GitHub · Deployed at 0xa198Fac58E02A5C5F8F7e877895d50cFa9ad1E04
  4. Messias J. & Ide A. et al. (2026). Fairness in Token Delegation: Mitigating Voting Power Concentration in DAOs. arXiv:2510.05830v2
  5. Blockscout — Ethereum block explorer, AAVE token holder API. eth.blockscout.com AAVE holders endpoint
  6. Frimpong K. et al. (2024). Understanding Blockchain Governance: Analyzing Decentralized Voting to Amend DeFi Protocols. arXiv:2407.10945
  7. Aave governance forum — proposal lifecycle documentation and ACI retrospectives. governance.aave.com
  8. BGD Labs — Aave Token V3 delegation properties. aave-token-v3/properties.md on GitHub