Why Quadratic Funding's 'Wisdom of Crowds' is a Myth

QF's math assumes unique human voters, but identity is cheap to forge. The mechanism incentivizes coordinated collusion to maximize matching funds, not genuine preference revelation.\n- Gitcoin Grants data shows >60% of rounds are vulnerable to manipulation.\n- Matching pool becomes a prize for the best Sybil farmer, not the best project.

A single large contributor matching many small donations creates the illusion of broad consensus. The 'crowd' is often a capital-rich entity performing matching optimization, distorting the 'one person, one voice' ideal.\n- A $10K whale can dictate the outcome by strategically matching 100 fake $1 donations.\n- True small-donor signals are drowned out by capital efficiency games.

Optimism's Retroactive Public Goods Funding inverts the model: fund after proven impact, not before speculative promise. This shifts the attack surface from prediction to proof, though it introduces subjective curation challenges.\n- $100M+ distributed across three rounds to date.\n- Mitigates Sybil attacks but centralizes judgment in Citizens' Houses.

QF fails as a Hayekian information mechanism. Voters lack the incentive to research deeply for marginal matching gains, leading to low-information herding. Funding correlates with marketing spend, not underlying utility.\n- Data shows donation patterns cluster around already-popular projects.\n- The 'crowd' reveals popularity, not wisdom or future value.

QF's matching formula incentivizes the creation of fake identities, not genuine community support. The dominant strategy is to split capital across Sybil wallets, turning the mechanism into a capital efficiency contest. Projects like Gitcoin Grants spend millions on sophisticated Sybil detection, a cost that fundamentally undermines the model's efficiency.

A single large, coordinated donor can dictate which projects win by strategically distributing capital to trigger maximum matching. This creates de facto curation by capital, not by a diverse crowd. The result is funding centralization disguised as decentralization, where protocols like Optimism's RetroPGF see a handful of entities controlling outcome direction.

Optimism's evolution from QF to Retroactive Public Goods Funding (RetroPGF) admits the crowd's failure at forward-looking prediction. It funds proven impact, not popular promises. This shifts the 'oracle' from a manipulable crowd to a selected panel of domain experts, accepting that centralized judgment is more effective for certain value allocations.

Voter attention is the real scarce resource, not capital. QF assumes informed, distributed voters, but in practice, voter apathy and low information lead to following influencers or default options. The matching pool is thus allocated based on marketing and narrative, not a crowd's 'wisdom'—mirroring flaws in liquid democracy systems.

QF is efficient at the micro-level (allocating a fixed matching pool) but catastrophically inefficient at the macro-level. The administrative overhead, Sybil wars, and misallocated capital from gaming represent a massive deadweight loss. Compare this to a simple CLR-agnostic grant from a foundation, which may have lower 'legitimacy' but far higher capital efficiency.

The funding model must align with what can be measured. For prediction, use prediction markets. For proven utility, use RetroPGF. For permissionless inclusion, use lotteries or direct grants. QF tries to be a universal oracle and fails. New systems like Allo Protocol's strategy layers are experiments in separating funding mechanics from the flawed 'wisdom' assumption.