The rules are the experiment.

Resident reference needs are 20, 35 and 25 units. Equal envelopes divide max(supply − 20, 0) into thirds, without reallocating excess. Separate withdrawals proceed upstream to downstream; each place requests 120% of its reference need as a buffer. The river receives what remains, up to 20 units.

The coordinated rule reserves up to 20 units, then shares the remainder in proportion to resident needs, capped at each need. Unused water stays unallocated. The example excludes infrastructure losses, negotiation costs and uncertainty. A central institution with the same information could use the same formula; the question is who supplies, contests and acts on that information.

Scores in resident order: activity 5/40/20/0/0; outcomes 45/20/25/10/0; care 35/15/20/30/0. Dividend = surplus × equal-share percentage ÷ 100 ÷ 5. Reward = remaining pool × individual score ÷ total score. Unrounded payments sum to the realised surplus; displayed rounding can differ slightly.

The 120-unit essentials threshold is arbitrary. Revenue can fall, so a dividend is not itself a guaranteed basic income. The example excludes reserves, taxation, losses, household differences and the cost of evaluation. Asset control and membership need their own rules.

Revocable delegation returns min(requested withdrawal, initial delegated power). Locked delegation returns zero. All returned authority goes to its original holders. The model uses one representative and does not simulate chains, expertise, collusion or subsequent ballots.

Concentration is one observable feature, not a score of decision quality. Revocation can redistribute power without solving differences in time, knowledge or organisational capacity.

Capital voting counts 30 × resident turnout/100 supporting units and 70 opposing units. Individual voting counts 90 × turnout/100 supporting residents and 10 opposing investors. Support percentage = supporting weight ÷ total participating weight.

Investor opposition is an invented scenario, not a claim about investors generally. The example excludes identity fraud, persuasion, abstention among investors, strategic trading and minority safeguards. A majority result does not automatically fund or implement a repair.

Population shares are 36%, 30% and 34%. Open attendance produces invented counts of 3/2/15 without support and 6/5/15 with support. Selection invites a 7/6/7 panel; attendance is 4/2/7 without support and 7/6/7 with support. These scenarios illustrate possible barriers, not measured effects or a simulation of random selection.

Assume every panel recommends the same 60-unit repair, so the authority comparison isolates its remit. A panel with budget authority can authorise it only when the available budget reaches 60. An advisory panel can influence the decision through a response obligation, but cannot authorise it itself. Composition does not measure discussion quality or guarantee consensus.

Budget and a named operator must both be available before 100 units are committed and the 40-unit advance paid. The remaining 60 is held until the completion rule is satisfied. Under report-only checking, either closed-ticket scenario qualifies. An independent check qualifies only when the selected scenario says the pump is repaired.

When the pump remains broken and the challenge route exists, assume residents report the fault in time; final payment is paused. This simplified model assumes accurate independent checks and timely reviews, and does not recover advances automatically. Real procurement needs verification costs, incentives, liability and a remedy after payment. Choosing “repaired” describes a scenario; it does not cause work to occur without funding and an operator.

Reference needs are arbitrary teaching inputs: 40/20/25/20. Established claims allocate 50%/5%/20%/25% of supply. The essentials rule shares water proportionally among the first three claims until their combined 85 units are met, then assigns up to 20 units to productive users. The proportional rule meets the same fraction of every reference need, capped at 100%.

Shortfall = max(need − allocation, 0). Under essentials and proportional allocation, supply beyond total needs remains unallocated. Under established claims, excess allocated to one group is not automatically reassigned. The example excludes water quality, infrastructure, seasonality, jobs and long-term effects of different priorities.

Coverage = registered residents ÷ 2,400 thousand. Excluded population = 2.4 million − registered residents. Continuity is a separate input with a different denominator; it is not multiplied by coverage. Registration does not establish adequate water, and interval completion does not measure welfare.

The response choice changes an institutional obligation, not the calculated population or a predicted outcome. “Funded assessment” is an illustrative design assumption: no cost, capacity or success rate is simulated. Actual remediation must be funded, performed and verified separately.