A governance research perspective on participation, power, and the problem no one wants to talk about....

There is a number that should trouble every ARB token holder: 787.

That is how many "Ghost Delegates" wallets holding real voting power but with zero governance participation were counted in Arbitrum DAO's February 2025 governance analytics report. Nearly eight hundred addresses sitting on delegated ARB, never showing up to vote. Not once.

This is not a fringe problem. This is the center of the governance crisis that Arbitrum is quietly trying to fix and largely failing to discuss openly.

Arbitrum governance analytics have begun categorizing delegates into a taxonomy that is more revealing than any voting chart:

Active Delegates — those who actually vote. In January 2025, there were 84. By February 2025, that number rose slightly to 91. A small improvement, but still a remarkably thin slice of a DAO managing a multi-billion dollar treasury.

Inactive Delegates — registered, hold delegated voting power, but have gone quiet. Around 123–125 as of early 2025. These are not newcomers. They were once active enough to attract delegation. Now they are dormant.

Ghost Delegates — the most alarming category. Wallets with meaningful voting power that have never participated. 787 as of February 2025. Down slightly from 791 in January but still a number so large it reframes the entire conversation about Arbitrum's governance health.

When you add inactive and ghost delegates together, you are looking at over 900 wallets holding real governance power that is functionally dead. The tokens delegated to them cannot vote. The communities that trusted them are unrepresented.

In October 2023, one of the most technically sophisticated risk management firms in DeFi Gauntlet was unable to vote on a $40 million ARB grants proposal. Not because they forgot. Not because they disagreed. Because their wallet address was incompatible with Arbitrum's chain.

Gauntlet had set up their governance wallet by reusing their Ethereum mainnet Safe address. It was incompatible with Arbitrum's governance contracts. The result: 12.6 million ARB in voting power more than many active top-tier delegates was rendered completely silent on one of the DAO's most consequential funding decisions.

An anonymous token holder had even added 8 million more ARB to Gauntlet's delegation just before the vote, unaware that Gauntlet was technically frozen out of governance.

This is not simply a user error story. It is a systems design story. It reveals how easily even well-resourced, well-intentioned delegates can become structurally silent and how the DAO has no real mechanism to detect or act on this until it is too late.

The governance participation picture in Arbitrum is one of concentrated activity among a very small group, with an enormous shadow of dormant power:

• On-chain participation dropped by 2.41 percentage points between January and February 2025

• On-chain voting power in active proposals fell from 210.7 million ARB to 198.6 million ARB in a single month

• Only 91 active delegates were participating in February 2025 in a DAO with hundreds of registered delegates and thousands of token holders

• The RAD (Rewarding Active Delegates) program, launched in January 2026 with $1.5 million in annual incentives, attracted only 38 applicants in its first month, of which only 24 completed compliance

That last number is worth sitting with. The DAO designed a $1.5 million incentive specifically to activate delegates and in its first month, fewer than 40 out of hundreds came forward.

The Rewarding Active Delegates program, approved in December 2025 with 55.8% support, is Arbitrum's most direct acknowledgment that silent delegates are a governance crisis.

The program's design is telling. To qualify, a delegate must:

• Hold a minimum of 200,000 ARB in voting power

• Vote on at least 75% of proposals each month

• Publicly post their voting rationale within five days of each vote

These are not high bars for someone genuinely engaged in governance. And yet, by March 2026, only 36 delegates were enrolled in the program. In February 2026, 37 were enrolled but only 33 met the base requirements for rewards.

This means that in a DAO that regularly sees 198–210 million ARB in voting power activated per proposal, fewer than 40 people are the ones doing the actual work. The rest hundreds of delegates, thousands of token holders are either passive, absent, or structurally blocked from participating.

The problem of silent delegates is made worse by what happens in the vacuum they leave behind.

In April 2025, a user called hitmonlee.eth spent just 5 ETH approximately $10,000 on LobbyFi to purchase 19.3 million ARB in voting power. That is more than the voting power held by established delegates like Wintermute or L2Beat. This single purchase helped secure a seat on Arbitrum's Oversight and Transparency Committee.

This is not an isolated incident. LobbyFi had already accumulated over 20.65 million ARB in delegated voting power by January 2025 making it one of the largest delegates in the entire DAO.

The connection to silent delegates is direct: LobbyFi's model works because of idle tokens. Token holders who would otherwise do nothing with their governance rights are incentivized to rent them out. The more ghost delegates there are, the more available voting power there is for purchase. Silence creates a market.

Behind the silence is a concentration problem that researchers are beginning to document more systematically.

A 2025 academic study analyzing governance data from Arbitrum, Aave, Uniswap, Compound, and ENS found that a large portion of voting power across major DAOs is concentrated within a small number of delegates. This concentration arises not only from intent but from structural platform bias Tally's default sorting by voting power, for instance, pushes the largest delegates to the top, making it more likely they attract further delegation, compounding their dominance.

Meanwhile, the majority of token holders abstain entirely. The result is a governance layer that looks democratic in theory but functions as an oligarchy in practice a small number of active delegates making decisions for millions of token holders who have no idea what is being decided.

Governance researchers and forum participants have identified several recurring patterns:

The Complexity Barrier. Arbitrum governance has grown significantly in scope and technical depth. Keeping up with treasury proposals, security council elections, LTIPP rounds, grant frameworks, and constitutional amendments requires sustained time investment. For delegates who originally raised their hands out of genuine interest but without institutional support, the workload becomes unmanageable.

The Incentive Gap. Until the RAD program launched in January 2026, there was no formal financial reward for consistent voting. Many delegates were active volunteers. Volunteers burn out.

The Technical Wall. The Gauntlet incident is the most dramatic example, but it is not unique. Wallet compatibility issues, Snapshot vs. Tally discrepancies, and unclear voting deadlines create technical friction that can silently sideline delegates.

The Rationale Reluctance. The RAD program requires public voting rationale within five days. For delegates who are uncertain about a proposal or who fear community backlash this public accountability requirement may actually discourage participation rather than encourage it.

Here is the question that goes largely unasked in Arbitrum governance discussions: if fewer than 40 delegates are doing the active governance work, and hundreds of millions of ARB in voting power is structurally dormant who is actually governing Arbitrum?

The answer is uncomfortable. A small, self-selected group of motivated participants, some of whom have clear financial stakes in the outcomes they vote on, are making decisions that affect a multi-billion dollar ecosystem. The silent majority of token holders have handed their power to representatives who have, in many cases, quietly stopped representing them.

This is not unique to Arbitrum. It is the structural condition of token-weighted governance across DeFi. But Arbitrum's scale its treasury, its transaction volume, its role in Ethereum's scaling roadmap makes the stakes higher than most.

The RAD program is a meaningful step. Financial incentives for consistent participation, public rationale requirements, and transparent tracking of delegate behavior are all positive governance design choices.

But 36–38 enrolled participants in a DAO of Arbitrum's size is not a solution. It is a floor. The question is whether the program can create a genuine culture shift or whether it will simply formalize the already-active delegates as a paid professional class while the silence around them deepens.

Some concerns already emerging from the forum: the 200,000 ARB minimum voting power threshold creates a barrier that favors established delegates over emerging ones. The KYC/KYB compliance requirement adds friction. And the payout structure, while transparent, still rewards voting power scale meaning larger delegates earn more, reinforcing the concentration dynamic.

From a governance research perspective, addressing the silent delegate problem requires more than incentive programs. It requires structural reform:

Delegation expiry mechanisms. Voting power delegated to addresses that have not participated in over six months should automatically be returned to the token holder or redirected to an active delegate of their choice. Perpetual silent delegation should not be the default.

Active notification systems. Token holders should receive clear, readable summaries of how their delegate voted or didn't on each proposal. Transparency tools exist; the DAO should fund and mandate their use.

Smaller delegate minimum thresholds. The current 200,000 ARB minimum for RAD eligibility concentrates incentives among large players. A tiered system acknowledging smaller but consistent delegates would broaden participation.

Governance literacy investment. Many ghost delegates are not bad actors. They are people who took on a role they did not fully understand, and received no ongoing support. Structured onboarding, ongoing education, and community check-ins could convert some fraction of the silent into the engaged.

Arbitrum has built one of the most sophisticated governance frameworks in DeFi. It has a written Constitution, a Security Council, elected oversight committees, incentive programs, and analytics dashboards. The infrastructure exists.

What it cannot engineer is presence. And right now, most of the power in Arbitrum governance belongs to addresses that are simply not there.

The silent delegates are not a conspiracy. They are a symptom of a governance model that delegated accountability without ensuring accountability followed. Until that changes, the 36 active participants in the RAD program will continue carrying a governance system designed for thousands.

That is not decentralization. That is a committee with extra steps.

Manoj Kumar Desai | DAO Governance Researcher from India | @MconnectDAO | New User of the Month — Aave, Optimism & Arbitrum (2026)