Who Decides

Volume VI — Governance and the Commons

THE PLATE

Four colleagues walking toward a long table set with dried grasses, the room bright with sun. — Plate VI.01The Table Before the Meeting.A decision rule is furniture. It decides who can be in the room, how long they will be there, and who is not in the room at all — and it does this before anybody has said a word.

THE LETTER

Every organisation you have ever been inside had an answer to the question at the top of this page, and almost none of them had written it down.

Who decides? In practice the answer is usually a sediment: a rule that was set once for a reason nobody remembers, amended twice under pressure, and now carried forward because changing it would require using it. Boards that vote by simple majority on a coffee supplier and on a pension buy-out. Cooperatives that require consensus on everything and therefore decide almost nothing. Treaty bodies where one member of twenty-seven can hold a directive for six months. Teams that call themselves flat and are run by whoever is most willing to keep talking.

None of that is a moral failure. It is an engineering failure, and engineering failures have the useful property that they can be measured and corrected.

This chapter treats a decision rule as what it is: a component with a specification, a cost, a failure mode and an operating range. We are going to compute what a rule costs, in two currencies that are always both being spent — the time and attention it takes to reach a decision, and the harm done to the people the decision goes against. Then we are going to look at five families of rule as they are actually practised, price each one, and say plainly which group pays in each case.

There is one result in here that will not go away once you have seen it. Every reduction in the cost of deciding raises the cost borne by somebody who did not want the decision, and the trade is exact. No rule escapes it. What a rule can do is place the cost deliberately, on people who can bear it, for stakes that justify it. What a rule cannot do is abolish it — and a governance system that claims to have abolished it has only stopped reporting where the bill went.

That is not a counsel of despair. It is the beginning of competence. A treasurer does not despair at the existence of interest rates; she prices them. This is the same move, applied to the one thing in your organisation that has never been priced at all.

— The Editors

DISCOVERY

What is already working

Start where the evidence is strongest, because the strongest evidence for governance design is not an argument — it is a set of institutions that have already run the experiment at scale and left the results in public.

A constitution that tells you its price. The United States Constitution can be amended when two-thirds of both chambers and three-fourths of the states agree. Since 1789 the Congressional Research Service counts roughly 11,848 measures proposed to amend it. Twenty-seven have been ratified — a rate of 0.228 per cent. Read that as a failure and you have misread it. Article V is working exactly as specified: it was designed so that a durable framework could not be rewritten by a transient majority, and it has held for more than two centuries. The price of that durability is written into the same arithmetic. Three-fourths of fifty states means thirty-eight must agree, which means thirteen states can block — and the thirteen least populous states hold 14,618,613 people, or 4.41 per cent of the country. The rule hands a veto over the constitutional text to four per cent of the people. That is not a defect; it is the purchase. The Framers bought stability and paid for it in minority veto power, and they were explicit about it.

A parliament that changed its rule and measured what happened. The United States Senate adopted its cloture rule in 1917 at two-thirds of senators present — with all present, 67 of 100, so 34 could block. In 1975 it was changed to three-fifths of senators duly chosen and sworn: 60 of 100, so 41 are needed to block. The blocking coalition grew by 7 senators, 20.6 per cent. Blocking was made harder on paper. The same reform removed the requirement to hold the floor, which made blocking almost free, and use of it rose. A rule has two parameters — the threshold and the effort — and changing only the threshold is how an institution reforms itself in the wrong direction. Forty-one senators come from at least 21 states, whose combined population is 37,174,921, or 11.22 per cent of the country.

A country that convened a random sample and then held a referendum. Ireland has done this twice and both times the result stood at the ballot box. The Convention on the Constitution (2012–2014) seated sixty-six randomly selected citizens alongside thirty-three parliamentarians and a chair; its recommendation on marriage became a referendum in May 2015 that passed with 62.07 per cent in favour. The Citizens' Assembly (2016–2018) seated 99 randomly selected citizens under Ms Justice Mary Laffoy and, over several weekends of evidence and deliberation, recommended repeal of the Eighth Amendment. The referendum of 25 May 2018 passed with 66.40 per cent on a turnout of 64.13 per cent — a margin of 16.40 points over a bare majority, on the question that had been considered unresolvable in Irish politics for thirty-five years. The assembly did not decide anything. It did something more useful: it made a decision available to a body that could.

A profession that abolished its managers and kept its accounts. Buurtzorg, the Dutch home-care organisation founded by Jos de Blok in 2006, runs nursing teams of at most 12 people who hire, schedule, budget and admit patients themselves. At around 10,000 nurses it carried a central staff of about 50 — a support ratio of 200 to 1, against roughly 833 self-managing teams. KPMG's 2015 review found Buurtzorg using approximately 40 per cent fewer care hours per client than comparable providers. The decision rule is consent within a team of twelve, and the reason it works is in the number twelve, which we will do the arithmetic on shortly.

And a discipline that counted the field. The OECD's Catching the Deliberative Wave (2020) identified 289 representative deliberative processes across member countries between 1986 and 2019 — 34 years, an average of 8.5 a year, heavily concentrated in the last decade. Sortition is no longer a thought experiment. It is a body of practice with a sample size.

Five institutions, five different rules, and not one of them arrived at its rule by accident or by principle alone. Each was bought, and each has a receipt.

THE ARITHMETIC

What works, what does not, and where the line sits

James Buchanan and Gordon Tullock published The Calculus of Consent in 1962 with a device that is still the most useful object in political economy and is almost never actually computed. Let us compute it.

The two curves.

Consider a body of n members that decides by a rule requiring k of them to agree. Two costs move in opposite directions as k rises.

External cost is what the rule does to people it decides against. A winning coalition of k imposes its choice on the n − k who are outside it; the expected share of the body outside the winning coalition is (n − k)/n, and the external cost is that share multiplied by the loss a member bears when a decision goes against them. It falls as k rises and reaches zero at unanimity, because under unanimity nothing happens to anyone without their agreement.

Decision cost is what it takes to assemble k agreements. It rises with k, and it rises faster than linearly, because each additional agreement must be won from a pool of remaining members that is getting smaller and more expensive.

Here is the micro-foundation, so the curve is a model and not a drawing. Suppose each member accepts a given proposal independently with probability p. A proposal carries in a round if at least k members accept, and the expected number of rounds is the reciprocal of that binomial tail probability:

    rounds(k)  =  1 / P(X >= k),   X ~ Binomial(n, p)

The optimal majority, computed. Take a real board: n = 9, p = 0.60, three hours of every member's time per round at a fully loaded £120 an hour, so one round costs 9 × 3 × 120 = £3,240. Twelve contested decisions a year. A member left outside the winning coalition bears £30,000.

   k    P(X>=k)   rounds    decision £/yr   external £/yr     total £/yr
   4     0.9006     1.11           43,169         200,000        243,169
   5     0.7334     1.36           53,011         160,000        213,011
   6     0.4826     2.07           80,562         120,000        200,562
   7     0.2318     4.31          167,740          80,000        247,740
   8     0.0705    14.18          551,146          40,000        591,146
   9     0.0101    99.23        3,858,025               0      3,858,025

The minimum is at k = 6 of 9 — 66.7 per cent — at £200,562 a year. A two-thirds rule is not a tradition in this table. It is the bottom of a sum.

Now notice the thing that makes this an instrument rather than a curiosity. Hold everything constant and move only the stake.

   loss to an excluded member      optimal rule
   £5,000                          4 of 9   (44.4 %)
   £10,000                         5 of 9   (55.6 %)
   £30,000                         6 of 9   (66.7 %)
   £100,000                        7 of 9   (77.8 %)
   £300,000                        8 of 9   (88.9 %)

Same nine people, same willingness to agree, same cost of a meeting. The right rule moves from below a majority to near-unanimity on the size of the stake alone. This is why a single decision rule for a whole organisation is always wrong. A board that votes 6-of-9 on the coffee supplier is paying £80,562 a year of decision cost to protect against a loss that does not exist; the same board voting 6-of-9 on a pension buy-out is under-protecting by two seats. The rule belongs to the decision, not to the body.

What unanimity actually costs. At k = 9 the tail is 0.6⁹ = 0.010078, which is 99.2 rounds — ninety-nine meetings for one decision, £3,858,025 a year for twelve of them. That is the honest price of a veto for everyone, and it is why no institution that has to act has ever run on real unanimity for long.

Consent rules, and what they really are. Sociocracy and its descendants solve this by narrowing what counts as an objection. Under consent, the question is not do you prefer this? but do you have an argued and paramount objection — a reason this will not work, or will harm the aim? A preference is not an objection. That single restriction changes the arithmetic completely. If the probability that a member holds an admissible objection is q = 0.10, the proposal carries with probability (1 − q)⁹ = 0.3874, which is 2.58 rounds. Consent is 38.4 times faster than unanimity.

And here is the part that consent's advocates rarely state. Find the k-of-9 rule with the same expected rounds and it is k = 6.23, or 69.2 per cent. Consent is not unanimity. It is a two-thirds supermajority wearing unanimity's clothes, and the difference is not free: the external cost at an equivalent 6.23-of-9 rule is £110,918 a year, against £0 at true unanimity. That £110,918 is billed, precisely, to the members whose objection is a preference they cannot phrase as a workability argument. The rule is a good one. It is good because of who it bills, not because it bills nobody.

Delegation, and the principal–agent arithmetic. Take an assembly of 1,200 deciding directly at a 60 per cent threshold — k = 720, expected 1.95 rounds, and each round costs 1,200 × 3 × £120 = £432,000. Twelve decisions a year is £10,114,446. Delegate to the board of nine at 6-of-9 and the same twelve decisions cost £80,562. The saving is £10,033,884 a year.

Against that stands agency loss. Suppose the board's choice diverges from what the assembly's majority would have chosen on 25 per cent of decisions. Set the saving against the expected loss and solve for the loss per divergent decision at which delegation stops paying:

    break-even loss  =  saving / (divergence x decisions)
                     =  10,033,884 / (0.25 x 12)
                     =  £3,344,628 per divergent decision

Delegate every decision whose loss-if-wrong is below £3.34m, and reserve every decision above it. That is not a principle about democracy. It is a schedule of reserved matters, derived, and it is what Operationalize This builds.

Delegation's real cost is not visible in that inequality, and this is the honest correction to it. In 1789 the House of Representatives seated 65 members for 3,929,214 people — 60,449 per seat. Today it seats 435 for 331,449,281 — 761,952 per seat. The principal's stake in each agent has thinned by 12.6 times, and the European Parliament's 720 members carry 623,898 each. The divergence rate is not a constant of nature. It rises as the ratio rises, and no reform of the rule touches it.

Sortition, and the arithmetic nobody does. A randomly selected assembly of 99 people has a 95 per cent margin of error of ± 9.8 points on a simple proportion. America in One Room (September 2019, 526 participants, reported by Fishkin, Siu, Diamond and Bradburn in 2021) sits at ± 4.3 points. A 1,200-member mini-public would reach ± 2.8. So a citizens' assembly is not a poll and must never be defended as one. Its claim is different and stronger: not this is what the public thinks, but this is what the public concludes when it has the evidence, the time and each other — a counterfactual no poll can produce at any sample size.

Now the honest negative, and it is the one this whole chapter is built on.

Look again at the two curves and notice that they are a single quantity split in two. Every step up in k buys external cost down and pays for it in decision cost; every step down does the reverse. The trade is exact and there is no rule outside it. Sortition does not escape it — it swaps external cost among groups by changing who is in the room, and buys its legitimacy with a sampling error of ten points. Delegation does not escape it — it converts decision cost into agency loss at a rate of £3.34m a divergent decision. Consent does not escape it — it converts a veto into a 69.2 per cent rule and bills £110,918 a year to whoever cannot argue in the required grammar.

So when a governance system claims to have abolished the trade-off, it has stopped reporting where the bill went. Here is a case, and here is the group.

Zappos adopted Holacracy across the company in 2014 and 2015 on the explicit claim that it removed managers, removed politics and gave everyone a voice — that is, that it lowered decision cost and external cost at once. In April 2015 the company offered every employee severance to leave rather than adopt it. 210 people took it — 14.0 per cent of a headcount of about 1,500 — and Zappos reported total turnover that year of 29 per cent. Zappos abandoned Holacracy in 2020.

The decision cost was not removed. It was moved off the org chart and onto every individual's week, where no line item reports it. At one hour per person per week of governance work across 48 weeks, that is 72,000 person-hours a year; at a loaded $40 an hour, $2,880,000 a year — a figure that had previously sat, visible and budgeted, in a management layer. The group that paid is the operational staff: the warehouse, customer-loyalty and merchandising employees whose work is not procedural, who had no appetite for constitutional drafting, and for whom an hour in a governance meeting bought nothing they had asked for. The people who gained were those already fluent in process. The system did not abolish the trade. It made the payers illegible, which is worse than making them poorer, because a cost nobody reports is a cost nobody can argue about.

DREAM

What becomes ordinary

In the organisation that has done this work, nobody argues about whether a decision needs a vote. They look it up.

There is a schedule. It has three or four columns and it lives where the delegated authority matrix used to live, and for each class of decision it names the rule, the body, and the number that put it there. Coffee suppliers and software licences sit at a bare majority of the operating team, because the loss if the decision goes wrong is four figures and the decision cost of anything heavier exceeds it. Hiring into a team sits at consent of that team, because the people who bear the consequence are exactly the people in the room. Anything that changes the ownership structure, the purpose, or a person's terms sits at a supermajority of the whole, because the loss-if-wrong crosses the threshold that the arithmetic set, and the threshold is written in the margin where anyone can check it.

The schedule is reviewed once a year, and the review is a computation rather than a debate. Somebody takes last year's decisions, sorts them by what it would have cost to get each one wrong, and asks whether any class has drifted across the line. Two or three usually have, in both directions, and they are moved. Nobody takes this personally, because the rule was never about trust.

When a decision goes against someone, they know the price they paid and they know it was named in advance. This turns out to matter more than winning. The most common source of institutional bitterness is not losing a vote; it is discovering afterwards that the rule was chosen, quietly, by people who knew which way it would fall. A rule that was set before anyone knew the question is a rule that can be lost to without injury.

And the meetings are shorter. Not because people talk less, but because the question what would it take to carry this? has a published answer, so the conversation is about the proposal rather than about the procedure. An organisation that has priced its decision rules spends its deliberation on the decision, which is what deliberation was for.

DESIGN

The structure that gets there

Four moves, in order. Each is a week's work and none of them requires anyone's permission but your own.

One — inventory the decisions, not the rules. List every recurring decision your unit made in the last twelve months. Do not start from the governance document; start from the calendar and the minutes, because the real rules are the ones that were used. For each, record who actually decided, how long it took from first raising to settled, and the honest answer to what would it have cost to get this wrong? That last column is the whole exercise. It is the loss term in the external-cost curve and nothing else in the model matters as much.

Two — price each class. Sort the list by loss-if-wrong and cut it into three or four bands. For each band, run the computation: with your n, your realistic p, the loaded cost of an hour of your people's time, and the band's loss, find the k that minimises the sum. The arithmetic is in lib/verify/VI_01.py and it takes four inputs. Most organisations discover two things at once: their highest band is under-protected, and their lowest band is carrying a rule that costs more than the decisions are worth.

Three — match the rule family to the shape of the cost, not to the culture.

Simple majority where the loss is small, the decisions are frequent, and the losers are the same people as the winners across a year. May's theorem (1952) is the reason this is the default: simple majority is the only rule that treats the proposal and the status quo alike. Every other rule is a deliberate thumb on one side of the scale, and should be able to say which.
Supermajority where the loss-if-wrong is large and largely irreversible, and where you are willing to say out loud whom the blocking coalition represents — thirteen states, forty-one senators, three of nine.
Consent where the affected group is small enough that (1 − q)ⁿ stays workable. At q = 0.10, a team of 12 expects 3.54 rounds; a body of fifty does not converge. Consent is a small-group instrument and Buurtzorg's twelve is not an accident.
Sortition where the decision is high-stakes, publicly contested, and distorted by the incentives of the people who would otherwise decide. Its output is a recommendation with a margin of error, and its power is that it makes a decision available to a body that could not otherwise reach one.
Delegation for everything below the break-even loss, with the reserved matters named explicitly above it.

Four — write the objection grammar down. If you adopt consent anywhere, the sentence that defines an admissible objection is the most load-bearing sentence in your governance document, because it is the sentence that sets q and therefore sets both curves. Write it, publish it, and keep a log of objections raised and how each was resolved. That log is how you measure q instead of assuming it, and a q you have measured turns this whole chapter from a model into an instrument.

DESTINY

How it holds when nobody is pushing

A decision rule is self-enforcing in a way that most governance is not, because it is used every time anybody decides anything. It does not need a champion. It needs three conditions, and it fails in three ways.

It holds when the schedule is consulted rather than remembered. A rule in somebody's head is a rule that changes with the weather. Put the schedule in the paper template so that every proposal states its own class and its own threshold in the first line, and the rule is enforced by the form rather than by a person.

It holds when the review is arithmetic. An annual re-computation is defensible in a way that an annual negotiation is not. The moment the review becomes a debate about how much trust people deserve, it will be won by whoever is most comfortable in that conversation, which is the exact failure the schedule existed to prevent.

It holds when the blocking coalition is named. Ostrom's third design principle — that most of those affected can participate in modifying the rules — is the condition under which people accept a rule that loses them a vote. A supermajority threshold that nobody can trace to a computation reads as a gerrymander, because usually it is one.

Now the failures. It fails when the loss-if-wrong column is filled in by the people who want a particular answer, because every threshold in the schedule is downstream of that column and it is the easiest number in the building to inflate. Have someone who will not be in the room fill it in. It fails when q drifts — when an objection grammar quietly widens until any preference qualifies, and the consent rule slides back toward true unanimity and ninety-nine rounds without anybody amending a word. The objection log is the only instrument that catches this, which is why it is worth the twenty minutes a month. And it fails when the rule is changed by the rule it governs, in the middle of a live question, by people who can see which way the vote will fall. Set the amendment threshold for the schedule higher than any threshold in it, and set it before you need it.

DELIGHT

What it feels like

There is a specific relief in losing a vote you expected to lose under a rule you helped set. It is not resignation. It is the feeling of having been dealt with fairly by a machine that was built before anyone knew the hand — the same pleasure as a well-cut game, or a contract that anticipated the thing that happened.

And there is a smaller, stranger pleasure on the other side of it: the meeting that ends early because the threshold was reached and everybody could see it. No one needs to keep speaking to establish that they cared. The proposal carried at six of nine, the two who objected have their objection in the log, and the room disperses into the afternoon with its attention intact.

Attention is the real currency here, and a priced decision rule is a way of spending it deliberately instead of continuously. What an organisation gets back is not efficiency, exactly. It is the ability to be serious about a few things — which is only possible if it has stopped being equally serious about all of them. That is what the schedule buys, and you feel it within a month.

OPERATIONALIZE THIS

At the level of finance

Decision rights are already a financial instrument. They are written into shareholders' agreements, joint-venture deeds and investment agreements as a schedule of reserved matters, and they change the balance sheet. Here is the instrument in the form a general counsel and a financial controller will both recognise.

The structure: a priced schedule of reserved matters.

The mechanics.

Three classes, named in the articles or the shareholders' agreement. Class A — ordinary operating decisions, board simple majority. Class B — decisions above the computed threshold, board majority plus the consent of the investor director or a stated percentage of a share class. Class C — constitutional matters, shareholder special resolution. In England and Wales, a special resolution is 75 per cent under section 283 of the Companies Act 2006, which gives you a statutory Class C threshold you do not have to invent.
The threshold that assigns a decision to a class is computed, not negotiated. Use the break-even: delegate below £3,344,628 of loss-if-wrong on the parameters above, reserve above it. Re-run it with your own n, p, hourly cost and divergence rate before you cite it; the method travels and the number does not.
The veto test, per reserved matter. A consent right is worth granting when the external cost it prevents exceeds the expected cost of the delay it creates. With a 20 per cent chance of a block, a 90-day delay and £4,000 a day at risk, the expected cost of the veto is £72,000 against £150,000 of external cost prevented — a ratio of 2.08 to 1, and the veto is granted. Run that line for every item on the schedule and delete the ones that fail it. A schedule of reserved matters that has never had an item deleted has never been priced.

The balance-sheet treatment, and this is the part most teams miss. Under IFRS 10, control turns on power over the relevant activities of an investee. A minority investor holding substantive participating rights over those activities can prevent the majority holder from consolidating; rights that are merely protective do not. US GAAP's consolidation guidance reaches the same place in substance. So the drafting of your reserved-matters schedule determines whether a subsidiary or joint venture appears on your balance sheet line by line or as a single equity-accounted investment. That is a governance decision with a direct, material effect on reported gearing and on covenant headroom. Bring your auditor into the drafting, not into the review — this is a conversation about the substance of rights, which is one they have every year.

The counterparty. Internal first. Write the schedule for one business unit, with the finance function as the counterparty for Class B consent, and run it for two quarters. You are building a track record of decisions made under a written rule, and that record is what makes the same structure negotiable with an external investor, a JV partner or a lender later. An investor who is handed a priced schedule rather than a boilerplate one is being told something true about the company, and it prices in.

The number that decides it. One inequality, on the front page:

      decision cost saved by delegating this class
   --------------------------------------------------  >  1
    divergence rate x decisions x loss if it is wrong

Above one, delegate and do not reserve. Below one, reserve it and accept the delay. If you cannot fill in the denominator, you do not have a governance problem — you have a measurement problem, and the first item on the schedule is to start recording loss-if-wrong on every decision your unit makes.

The first ninety days on a page.

Day	Action	Artifact
1–15	Inventory twelve months of real decisions from the minutes	The decision register, with loss-if-wrong
16–30	Band by loss; compute `k*` for each band	The priced schedule
31–45	Run the veto test on every proposed reserved matter	The veto ledger, items deleted
46–60	Draft Classes A, B, C into the agreement; auditor on consolidation	Draft schedule + IFRS 10 memo
61–75	Adopt for one business unit; begin the objection log	Objection log, week one
76–90	First review: measured `q`, decisions carried, hours spent	One page, one person

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

Think of a decision this organisation made in the last two years that everybody, including the people it went against, still thinks was decided well. What was the rule, and who was in the room?
Where do we already decide quickly and well without anybody checking a policy — and what is it about those decisions that makes that safe?
Which group here has the most reliable habit of raising an objection that turns out to be right? What have we been doing that makes that possible?

Dream — what becomes possible

If every proposal arrived with its own threshold printed in the first line, what would our meetings be about instead?
Imagine we could see, once a year, exactly what our decision rules cost us in hours and what they saved us in avoided harm. What would we do differently the following week?
If the people most affected by a decision were reliably the people who decided it, which three decisions would move, and who would they move to?

Design — what we build

What is the single decision in this unit with the largest loss-if-wrong, and what rule is it currently made under?
What sentence should define an admissible objection here, in our words — and who would we trust to apply it?
Which of our current approval steps would we delete tomorrow if we priced the delay it creates against the harm it prevents?

Destiny — how it holds

What would have to be true for this schedule to still be in use, and still accurate, when everyone in this room has moved on?
Who fills in the loss-if-wrong column, and what would make us confident that number was honest?
If our rule began quietly drifting back toward requiring everyone's agreement, what is the first thing we would see, and who would notice it first?

WORKS CITED

Arrow, K. J. (1951). Social Choice and Individual Values. Wiley.

Bernstein, E., Bunch, J., Canner, N. and Lee, M. (2016). "Beyond the Holacracy Hype." Harvard Business Review, July–August 2016.

Buchanan, J. M. and Tullock, G. (1962). The Calculus of Consent: Logical Foundations of Constitutional Democracy. University of Michigan Press.

Endenburg, G. (1998). Sociocracy: The Organization of Decision-Making. Eburon.

Farrell, D. M., Suiter, J. and Harris, C. (2019). "'Systematizing' Constitutional Deliberation: The 2016–18 Citizens' Assembly in Ireland." Irish Political Studies, 34(1), 113–123.

Fishkin, J. S. (2018). Democracy When the People Are Thinking: Revitalizing Our Politics Through Public Deliberation. Oxford University Press.

Fishkin, J. S., Siu, A., Diamond, L. and Bradburn, N. (2021). "Is Deliberation an Antidote to Extreme Partisan Polarization? Reflections on 'America in One Room'." American Political Science Review, 115(4), 1464–1481.

Financial Accounting Standards Board. Accounting Standards Codification Topic 810, Consolidation. FASB.

International Accounting Standards Board (2011). IFRS 10 Consolidated Financial Statements. IFRS Foundation.

Jensen, M. C. and Meckling, W. H. (1976). "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure." Journal of Financial Economics, 3(4), 305–360.

KPMG (2015). The Best of Both Worlds: Buurtzorg. KPMG International.

Laloux, F. (2014). Reinventing Organizations. Nelson Parker.

Landemore, H. (2020). Open Democracy: Reinventing Popular Rule for the Twenty-First Century. Princeton University Press.

May, K. O. (1952). "A Set of Independent Necessary and Sufficient Conditions for Simple Majority Decision." Econometrica, 20(4), 680–684.

OECD (2020). Innovative Citizen Participation and New Democratic Institutions: Catching the Deliberative Wave. OECD Publishing.

Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press.

Rae, D. W. (1969). "Decision-Rules and Individual Values in Constitutional Choice." American Political Science Review, 63(1), 40–56.

Robertson, B. J. (2015). Holacracy: The New Management System for a Rapidly Changing World. Henry Holt.

United Kingdom. Companies Act 2006, section 283.

United States Senate. Senate Cloture Rule (Rule XXII, adopted 1917, amended 1975). Committee on Rules and Administration.

Van Reybrouck, D. (2016). Against Elections: The Case for Democracy. Bodley Head.

Note on figures. Every figure above is computed in lib/verify/VI_01.py and reproducible with python3 lib/verify.py VI.01. The decision-cost model is a bargaining model with a stated micro-foundation, not a curve fitted to observed legislatures: the shape is the claim, and the level belongs to the parameters, which are printed with their units and labelled. Population figures are the 2020 United States Census apportionment counts and Eurostat's 1 January 2024 estimate; the proposed-amendment count is the Congressional Research Service tally through 2014.