Auditing governance concentration beyond token allocation: a live-governance study of 52 token protocols

Token-governance decentralization is not established by launch allocation, raw holder counts, or token-inequality metrics. It is an auditable current-control condition: who holds governance-relevant tokens after protocol-controlled addresses (PCAs) are removed, who retains insider positions, and how voting mechanisms transform holdings into rule-making power. Token allocation is a launch document.

Governance concentration is a live institutional state. We turn that standard into a method across a 52-protocol cross-section spanning DePIN, DeFi, infrastructure, and social tokens, computing Herfindahl-Hirschman Index (HHI) concentration after PCA exclusion. Under audit, launch-design and protocol-financial covariates (insider, team, and investor allocation; maturity; circulating float; valuation ratios) are each uninformative about steady-state concentration (insider allocation Pearson r = 0.09, p = 0.55, N = 50).

Current insider retention is the holder-side correlate that survives: protocols with more insider wallets among top holders are more concentrated (Spearman rho = 0.44, p = 0.005, N = 39, surviving a non-insider HHI tautology check at rho = 0.54). Voting mechanisms then separate rule-making power from holdings: delegation amplifies voting power above token holdings in thirteen of eighteen protocols with sufficient governance data, with five design-driven exceptions (ENS, GMX, HNT, JUP, LPT). A subsidy-to-concentration association appears only through a single outlier (Pearson r = 0.62 including Livepeer, r = 0.07 excluding it).

Applied across sectors, the audit also distinguishes DePIN from DeFi: DePIN governance is more concentrated than DeFi after correction, a directionally robust medium effect (Cohen’s d = 0.65, Mann-Whitney p = 0.028), reported as a descriptive sector contrast, not as the central claim. Holder-list concentration is meaningless until the unit of control is identified. The five-class PCA-exclusion typology is control attribution, not data cleaning, correcting systematic inflation in prior holder-list studies (median factor 2.3×, maximum approximately 18×); once control is attributed, Gini and HHI capture distinct properties of the same holder set (r = 0.52), and inequality metrics cannot substitute for direct concentration measurement.

All findings are descriptive associations from a single 2026 cross-section (holder snapshots collected March to May 2026), not causal claims; the audit standard and the current-control thesis are general, while the specific point estimates are bounded to that sample. The paper specifies five forward predictions with falsification thresholds and commits to Open Science Framework pre-registration before any panel or event-study extension. The practical implication is a changed audit default: a decentralization claim requires a live-governance audit of PCA-corrected holdings, insider retention, and voting power, not launch allocation or raw holder lists.

Concentration of this kind bears on the legitimacy of decentralized governance, not only its efficiency: where a small set of holders or delegates commands decisive voting weight, the broad participation in rule modification that the commons self-governance ideal presumes is nominal rather than operative.