Token-weighted voting – often called one-token-one-vote – is the rule that a voter's influence is proportional to how many governance tokens they hold. It is the on-chain default: the great majority of DAOs count votes this way, and it is the baseline against which every other governance model is defined. Weighting is only half the machine: the rule that turns weighted ballots into a winner – plurality, approval, or one of the ranked methods – is a separate design choice layered on top of it. Its appeal is that it is simple, fully mechanical, and Sybil-resistant by construction – because influence is tied to a scarce, market-priced asset, splitting your holdings across a thousand wallets changes nothing, and buying more influence means buying more of the token. This page covers how it works on-chain, why it won, and the well-documented ways it fails.
The dominant implementation is the OpenZeppelin Governor pattern, standardized after Compound's 2020 Governor Bravo and now running under hundreds of DAOs. A governance token implements a votes extension (ERC-20 ERC20Votes) that records a checkpointed balance history, so the contract can ask "how many tokens did this address control at block N?". Voting power must be activated by delegation – holders delegate to themselves or to a representative – which is why raw token balances and active voting power often differ sharply.
A proposal then moves through a fixed lifecycle: it is submitted (subject to a proposal threshold), enters a voting period during which each address's weight is read from a snapshot block fixed at proposal creation, and – if support clears the threshold and quorum is met – is queued in a timelock before it can execute. The snapshot is load-bearing: because weight is counted as of a past block, tokens acquired after a proposal is filed carry no weight, which is the standard structural defense against flash-loan vote seizure.
Both numeric gates in that lifecycle – the proposal threshold on the way in and the quorum on the way out – are design choices rather than constants of the pattern. Which vote types count toward the quorum floor differs between Compound's Governor Bravo and OpenZeppelin's default counting module, and how the floor itself is denominated (an absolute token constant, a fraction of past total supply, delegated tokens under a clamp, or a function of the opposition) differs again across major deployments. Quorum and threshold design works through each and what it costs.
Token voting won for practical reasons, not because it is fair. It needs no identity layer – the hardest unsolved problem in on-chain governance – because it never has to know who anyone is, only how many tokens they moved. It is legible: anyone can compute the outcome from public balances. It is composable: a single ERC-20 doubles as the fundraising instrument, the liquidity-mining reward, and the ballot, and it drops straight into off-the-shelf tools like Tally and Snapshot. For a protocol that has to ship, "issue a token and wire it to a Governor" is the path of least resistance – and so it became the path almost everyone took. (Chainlink: governance tokens and DAO voting)
The defining objection is that one-token-one-vote is one-dollar-one-vote: whoever holds the most tokens decides, so a protocol that calls itself decentralized can be steered by a handful of founders, funds, and exchanges. The concentration is measurable – analysts track a DAO's Nakamoto coefficient (how few addresses it takes to reach a majority) – one of the core DAO health metrics – and for many large DAOs that number is uncomfortably small. Compounding it is voter apathy: because a small holder's vote almost never changes an outcome, turnout in token-weighted DAOs routinely runs in the single digits, which concentrates effective power even further into the few large holders who do show up. Forbes summarized a decade of findings under a blunt 2026 headline: DAOs keep centralizing.
The subtler failure is that token voting decouples influence from stake. A voter who borrows tokens to vote and returns them afterward has, as Vitalik Buterin argues, "zero net exposure" to the consequences of the decision – they capture any upside of a self-serving vote while bearing none of the downside to the token. This is not hypothetical: lending markets, liquid-staking wrappers, and outright bribe markets (the Curve Wars' vote-incentive platforms) all let voting power be rented. The result is a bundling vulnerability – voting rights that were supposed to be tied to long-term ownership become just another yield to be farmed, and a well-capitalized outsider can assemble a temporary majority without any lasting alignment. It is the mechanism behind several governance attacks.
Rather than abandon token voting, most DAOs bolt defenses onto it. Past-block snapshots neutralize flash-loaned weight. Timelocks and guardian vetoes buy time to react to a captured vote. Vote-escrow (ve) locks – as in Curve's veCRV – time-weight tokens so transient capital counts for less than committed capital. Staking gates – as on the dYdX Chain and other Cosmos-SDK DAOs – count only tokens actively staked to secure the chain, so idle balances carry no vote weight. Delegation lets apathetic holders hand their weight to engaged representatives, raising effective participation. And a growing number of designs combine stake with something money cannot cheaply buy – verified identity (for quadratic voting), elapsed time, or demonstrated participation – accepting a little more complexity to escape pure plutocracy. (a16z governance FAQ)
A caper deliberately departs from pure token weighting. Stake still matters – a member's holdings are one factor in their weight – but weight is the product of that stake and the member's earned participation, recorded as non-transferable tokens minted 1 per ballot cast and on nothing else. A balance whose holder has never cast a ballot – bought, transferred in, or borrowed for a block – therefore carries no weight, which blunts both the plutocracy problem (idle capital does not automatically rule) and the vote-buying problem (a borrowed position brings the lender's record, which is to say none of it, and no holder can rent out their own).
Combined with a ranked-choice super-majority, a market window that can veto a passed action, and an exit right that pays a treasury share sized by the same weight, tying each member's voice to their claim on the treasury, the design keeps token voting's Sybil-resistance while shedding its one-dollar-one-vote default. The mechanics are on the linked governance pages.