For the Practitioner of Luminous Economics

Volume II — Foundations: The Paradigm and the Science · Extension

Nine movements, three grades.

THE PLATE

Hands holding an open book on a lap of white linen, the margin turned to the light from the side. — Plate II.E1The Margin, in Pencil.She is not deciding whether she agrees. She is deciding which sentences she would be willing to defend in a room where somebody has read the source.

THE LETTER

You have read eleven chapters of evidence and you are about to do something with them. Write a proposal. Sit on a panel. Advise a board, a fund, a council, a farm. Put a paragraph in a document that somebody else will be asked to sign.

This extension is about the part that happens next, and it is the part nobody teaches. Not what is true — what you may cite, at what strength, and with which source.

The distinction is not pedantry and it is not modesty. Volume II contains results of three quite different kinds, and they are printed in the same type, in the same voice, in the same nine movements. Some of them are theorems and official statistics that a hostile economist will concede in the first minute because they came out of their own journals. Some are real and disputed, and the dispute is live, and a practitioner who cites them without naming it will be corrected in public by somebody who has read the other side. And some are images — beautiful, useful for thinking, and not admissible as evidence anywhere, a fact this edition states about itself in Chapter II.11 and which almost nobody who quotes this literature has noticed.

Mixing the three is the single most expensive habit in this field, and this chapter prices the habit. The arithmetic is short and it is unkind: under the edition's own assumption about how sceptics read, three loose claims added to a bundle of seventeen sound ones remove 60.1 points of expected credibility. Adding evidence subtracted credibility. That is not a rhetorical flourish; it is a hypergeometric calculation and you can change its inputs in lib/verify/II_E1.py and watch it move.

What you get here is a ledger of thirty results, graded, with the grade's reason and the published origin attached; a rule for what each grade licenses; the arithmetic of what a misgrade costs; and an instrument that makes the discipline pay for itself in an advisory engagement. What you do not get is permission to stop thinking. The grades are ours, we wrote them, and Chapter II.11's third rule says plainly that the reader of a register is never its author. Regrade us.

— The Editors

DISCOVERY

What is already working

Begin with the good news, because it is substantial and it is not ours: several fields have already built exactly this instrument, published it, and found that it raised rather than lowered what they could claim.

GRADE, and the four levels. In 2008 the GRADE working group — Gordon Guyatt and some seventy collaborators — published in the BMJ a system for rating the quality of clinical evidence on four levels: high, moderate, low, very low, with the rating separated explicitly from the strength of the recommendation built on it. The move that made it work is the second half of that sentence. A strong recommendation may rest on low-quality evidence, and it must say so. GRADE is now used by the World Health Organization, by Cochrane, and by national guideline bodies across a dozen countries, and its adoption was driven by clinicians who found that a graded recommendation was easier to defend than an ungraded one, because the argument about the evidence had already happened on the page.

The IPCC's calibrated language. Since the guidance note of Mastrandrea and colleagues in 2010, every substantive statement in an Assessment Report carries two attached scales: a confidence level built from the type, amount, quality and consistency of evidence, and, where a probabilistic statement is warranted, a likelihood term with a numerical range behind it. Very likely means a stated probability band, not a mood. The result is a document that hostile readers attack at the level of individual findings rather than as a whole — which is exactly what a practitioner wants, because a finding can be defended and a whole cannot.

Ostrom's own principles, audited by other people. Michael Cox, Gwen Arnold and Sergio Villamayor-Tomás tested the eight design principles against 91 empirical studies in 2010 and found them broadly supported, with three of the eight needing to be split — taking the set from eight to eleven. Chapter II.11 tells this story to make a point about criticism. Take it here as a point about citation: the principles are more citable after the audit than before, and the practitioner who cites the eleven rather than the eight is visibly current.

The adversarial collaboration. Daniel Kahneman and Angus Deaton reported in 2010 that emotional wellbeing stopped rising with income around a threshold; Matthew Killingsworth found no satiation in 2021; and rather than trade rebuttals the two camps ran a joint reanalysis, published in 2023, which located the truth in both. Two teams, opposite findings, one procedure, a better answer, and nobody had to lose. For a practitioner this is the most useful case in the set, because it shows what to do when your grade-B claim meets its contest: propose the joint test rather than the rebuttal.

The Bermuda Principles. In February 1996 the public genome sequencing centres agreed to release assemblies above one kilobase within twenty-four hours and not to patent the sequence. What makes it a discovery case here is not the generosity. It is that a group of practitioners wrote down, in advance, what they would and would not claim — and the norm held for a decade without an enforcement mechanism, because everyone could see who had broken it.

Five cases, one pattern: in each, a community of practitioners published the strength of its own claims before anybody asked, and found that the published weakness was what made the strength usable. That is the whole of this extension, and the rest is arithmetic and a ledger.

THE ARITHMETIC

What you may claim, at what strength, and what a misgrade costs

Every figure below is computed in lib/verify/II_E1.py, which prints its inputs with their units and their sources before it prints a result.

First, the ledger. Thirty results, drawn from all eleven chapters of this volume, graded on three levels.

  grade                                 results     share
  ----------------------------------------------------------
  A  cite at full strength                 14       46.7 %
  B  cite with the contest named            9       30.0 %
  C  do not cite as evidence                 7      23.3 %
  ----------------------------------------------------------
  total                                    30      100.0 %

Grade A — cite at full strength, name the source, expect concession. These are theorems, identities, official statistics, or measurements published in journals whose editors are not friendly to this tradition. Among them: that aggregation of perfectly rational individuals preserves continuity, Walras's law and homogeneity and nothing else (Sonnenschein, Mantel and Debreu); that a competitive economy can have a price path which never converges (Scarf); that US firm sizes are Zipf with a Pareto exponent of 1.059, so the representative firm is a statistic the data does not possess (Axtell); that index returns carry a cubic tail; that cities scale superlinearly at about 1.15 while metabolism scales at 0.75, which means the economy is not an organism; that reserves are an economic and legal category rather than a geological one; that the competitive message space is minimal and therefore blind in k − 1 directions (Mount and Reiter); that a fully informative price cannot exist in equilibrium (Grossman and Stiglitz); that unpaid household production runs at 25.7 percent of US GDP and 63.1 percent of UK GDP; and that 147 holders accumulate forty percent of the control of the world's transnational corporations.

Grade B — cite with the contest in the same sentence. Nine results, each real, each disputed by somebody who has read it properly. The societal EROI floor near 11:1. The trust–growth magnitude, where two credible slopes exist — 0.80 and 0.67 points of growth per ten points of trust — and the coefficient does not survive extreme bounds analysis in the extended sample. The useful-work elasticity near 0.697, which requires factor markets to misprice energy by an 8.3-times wedge. Greif's reading of the Maghribi coalition, against Edwards and Ogilvie. Autopoiesis extended to social systems, which Maturana himself declined. Ulanowicz's window at 1/e. Agent-based validation. A routine heritability of 0.35, which is an assumption and not a measurement. And Dunbar's number, whose published 95 percent interval runs from 100.2 to 231.1 — the famous figure was never a number, it was an interval about 131 people wide.

The form for a grade-B citation is one sentence long and it is not a hedge:

Knack and Keefer find about 0.8 points of additional annual growth per ten points of generalised trust; Zak and Knack get 0.67 on a different specification; Beugelsdijk and colleagues show the coefficient does not survive extreme bounds analysis in the extended sample. The direction is established, the magnitude is a range, and I am using the low end.

A sceptic cannot do anything with that paragraph except agree with it.

Grade C — do not cite as evidence, at all, ever. Seven claims, and you have almost certainly said three of them out loud this year. That firms are autopoietic: they score three of six on a conjunctive key whose authors call this case allopoietic, in the same book. That systems self-organise to maximise entropy production: a conjecture, not a theorem. That thermodynamics forbids an economy of this size: the planet exports entropy 9,282 times faster than the human economy produces it, so the sink is not the binding constraint and a sceptic with a calculator finds that out in four minutes. Georgescu-Roegen's fourth law, false as stated since 1993. That a diverse, densely connected system is therefore resilient, which contradicts the only place that claim was ever made precisely. That humanity uses 1.71 Earths as an independent measure of ecological limits, when about sixty percent of the index is carbon restated as forest area. And the eight unfalsifiable sentences this edition names in its own register, beginning with the economy is a living system.

Second, the cut — and it is a fact about the ledger rather than an argument.

Take the fourteen grade-A results and ask who produced them.

  grade A results                                      14
  produced outside this tradition                      13     92.9 %
  produced inside it                                    1      7.1 %

Thirteen of the fourteen strongest things you can say came out of orthodox economics journals, general-science journals, or official statistical offices. The fourteenth is an identity anybody can rederive in a line. Not one of them is ours. Sonnenschein is in Econometrica. Scarf is in the International Economic Review. Axtell and Gabaix are in Science and Nature. Mount and Reiter are in the Journal of Economic Theory, Grossman and Stiglitz in the American Economic Review, the household figures in the Survey of Current Business and an ONS satellite account.

Read the consequence carefully, because it inverts the instinct. Your strongest material is the opposition's own published work, and your weakest material is your own vocabulary. The practitioner's move is therefore not to defend the paradigm and then deduce the findings from it. It is to lead with the findings — in the citation the room already respects — and to keep the vocabulary in the letter, where it belongs and where it is good.

Third, what a misgrade actually costs.

Model the reader. This edition states its own assumption about sceptics in three separate places: a reader who checks one number and finds it loose will not check the second one, and a sceptic who finds one claim loose discards both. Take that at face value. A bundle of twenty citations, some number of them grade C, and a reader who checks five of them drawn at random, discarding the whole bundle if any checked citation is loose.

  loose in 20     P(bundle survives 5 checks)     credibility lost
  ------------------------------------------------------------------
        0                    100.0 %                    0.0 %
        1                     75.0 %                   25.0 %
        2                     55.3 %                   44.7 %
        3                     39.9 %                   60.1 %
        4                     28.2 %                   71.8 %
        5                     19.4 %                   80.6 %

Now read the first and fourth rows together. Seventeen sound claims, standing alone, survive at 100.0 percent. The same seventeen, with three grade-C claims added for colour, survive at 39.9 percent. Adding three pieces of evidence removed 60.1 points of expected credibility from the seventeen that were already there.

That is the arithmetic of the thing everyone in this field does. The proposal has the SMD theorem in it, and the household satellite account, and the cubic tail — and then, in the second paragraph, because it felt like it tied the room together, the economy is a living system that self-organises to maximise entropy production. The reader checks five things. One of them is the second paragraph.

Fourth, the honest negative, and it is about our own model.

The discard assumption is an assumption. It is stated three times in this edition and measured nowhere in it. Run the kinder model instead — a reader who prices each citation independently and keeps the share that holds:

  discount model, 3 loose of 20                85.0 %
  discard model, same bundle                   39.9 %
  gap between the two models                   45.1 points

Forty-five points of the case for this discipline rest on which model of a reader is right, and this edition does not know. Under the discard model, grading your claim schedule is the highest-return hour of work available to you. Under the discount model it is worth roughly a fifth as much, and a practitioner who spends a week on it has overinvested. The honest position is to say which model you are assuming — and to notice that the two models make opposite recommendations about how much colour a proposal can carry.

There is a second negative and it cuts closer. The appreciative method that generates every one of the 924 room questions in this edition grades B by its own rule. Bushe and Kassam's meta-case analysis reviewed 20 published Appreciative Inquiry cases and found 7 reporting transformational outcomes — a rate of 35.0 percent. That is a real finding about a real method and it is not a dismissal; a third of interventions producing transformation is a good rate. But it is a case literature, not a controlled one, and a practitioner who presents the room questions as an evidence-based intervention has just put a grade-B claim in an A-grade voice. Say thirty-five percent of published cases, on a meta-case analysis, with the selection problem that implies — and then run them anyway, because the alternative on offer is an unmeasured deficit interview.

Fifth, what the volume has already published about itself, which you may quote at full strength because it is a count rather than a claim:

  load-bearing propositions audited in II.11        24
  of them falsifiable                               15        62.5 %
  of them unfalsifiable                              9        37.5 %

  systems sentences scored against II.04's licence   4
  of them licensed                                   2        50.0 %

A department that asks you whether this economics is science has, in those two blocks, a better answer than any assurance you could offer: sixty-two and a half percent of the load-bearing claims forbid an observation, the rest are struck from the business case, and the edition published the split itself.

DREAM

What becomes ordinary

In the practice that has absorbed this, a proposal arrives with its claim schedule attached and nobody remarks on it, the way nobody remarks on a set of accounts having notes.

The schedule is one page. Each row is a claim the recommendation depends on, its grade, its source, and — for every grade-A and grade-B row — the finding that would retire it. Reviewers read the schedule first and the argument second, the way an analyst reads the notes before the face of the accounts, and the reading is faster because the disagreements have addresses.

Practitioners have stopped defending the frame. A room that wants to argue about whether the economy is a living system is offered the register, told that the sentence is classed unfalsifiable and does not appear in the term sheet, and asked which of the fifteen falsifiable claims it would like to take. The conversation that follows is about thresholds, and it resolves, usually inside the hour.

Grade-B claims travel with their contest and are more persuasive for it. The practitioner who says the trust literature disagrees about the magnitude by a factor of about 1.2 and here is why I am using the low end is trusted on the next claim as well, and a room that has watched somebody give ground voluntarily extends credit it does not extend to certainty.

Regrading is routine and cheap. A result moves from B to A when a second team replicates it, and from A to C when somebody finds the coding error, and both movements are published with a date and a name. The ledger has a version number. Nobody experiences a downgrade as a defeat, because everybody has had one.

And the vocabulary has found its proper home. Regeneration, flow, wholeness, living system — these appear in the letter, in the plate, in the opening paragraph where they do what language does, and they are absent from the schedule, the arithmetic and the term sheet. Both halves of the document are stronger. The prose reads better without a burden of proof it was never built to carry, and the proof reads better without a metaphor standing next to it.

DESIGN

The claim schedule, built

The whole of this chapter becomes operational as one artifact, and it takes an afternoon.

The claim schedule: one row per load-bearing claim, five columns.

Column	What goes in it
The claim	One sentence, stated so that two readers would test it identically
Grade	A, B or C, with the reason in six words
Source	Author, year, venue. Not a secondary citation, not a summary, not this book
The contest	For every B: who disputes it, and what they say
Kill condition	For every A and B: the finding that would retire it

How it is built. Write the recommendation first, in your own voice, at full conviction. Then go back through it and underline every sentence a reasonable reader could ask you to support. You will find between eight and twenty-five. Those are the rows. Grade each one against the ledger and against its own source — read the source, not the summary, which is where four of this volume's seven grade-C claims were found. Then apply the three rules.

Rule one: a grade-C claim may stay in the prose and may not stay in the schedule. It is not a lie and it is not a failure. It is a figure of speech, and figures of speech belong in the letter. This is Chapter II.11's first rule, borrowed intact, and it is what lets you keep the writing you actually want to write.

Rule two: a grade-B claim is cited with its contest in the same sentence or not at all. Not in a footnote, not in an appendix, not in the oral answer if somebody asks. In the sentence. A contest disclosed by you is a strength; the same contest disclosed by a reviewer is the end of the meeting.

Rule three: the strongest citation in the bundle should be one your reader already respects. Lead with the Econometrica result, not with the regenerative one, even when the regenerative one is closer to your point. You are not conceding anything; you are choosing the order.

Governance. The schedule is graded by somebody who did not write the recommendation. This is the same separation Chapter II.03 puts on the boundary register and Chapter II.11 puts on the claim register, for the same reason — the author is the one person who cannot see the gap between what the source says and what the sentence says, because they read the sentence through the source. In a two-person practice this costs an hour a fortnight and it is traded.

Sequence. Grade the schedule before the proposal is finished, never after. A schedule graded after the recommendation is written becomes a search for support, and a search for support finds it. Graded midway, it changes what you recommend — which is the point, and the only reliable sign the discipline is working.

DESTINY

How it holds when nobody is watching

Three things keep a citation discipline alive past the enthusiasm that starts it, and they are all structural.

It is faster. A graded schedule shortens a review, because the reviewer stops hunting. Practitioners keep doing it for the same reason they keep a tidy model: the second engagement is cheaper than the first, and by the fourth the schedule is largely a copy-and-amend.

It is portable. The schedule outlives the engagement. A claim graded once, with its source read and its contest named, is graded for every proposal you write afterwards, and a practice that has been doing this for three years is carrying an asset its competitors would need three years to build.

Somebody who is not you reads it. One person grading their own claims is a conscience. Two is a procedure.

Now the failure modes, named so they can be seen coming.

It fails when the grades inflate. Every claim you like drifts toward A, and the drift does not feel like dishonesty — it feels like having thought about it more. The guard is the kill-condition column: a grade-A claim with no stated kill condition is a grade-C claim in a good suit, and the column will not let you write one.

It fails when the schedule becomes a compliance artifact — filled in once, copied forward, never re-derived from the sources. The trigger has to be the calendar and the engagement, never the conscience.

It fails when it is used as a weapon. A practitioner who grades a colleague's work in public to win an argument will get one honest schedule from that colleague and never another. Grade privately, argue publicly about thresholds.

And it fails, most commonly, when a practitioner answers a specific empirical objection with a general philosophical one. Somebody brings a number about cooperative failure rates; the answer comes back as a paragraph about interconnection. At that moment the schedule has already stopped working, whatever is written on it — and the person who notices first is usually the one who was about to be persuaded.

DELIGHT

What it feels like

There is a particular lightness in being able to say I don't know that one in a room where you are the expert. It sounds like a cost and it is not. The sentence buys you the whole rest of the meeting, because everything you do claim afterwards arrives with the weight of somebody who has demonstrated they will say the other thing.

And there is the pleasure of the ledger itself, which is a collector's pleasure. Thirty results, graded, with their origins written down — it is a small, well-kept set of tools, and a practitioner who has one finds themselves reaching for the right size without looking. The Scarf orbit for a room that believes in equilibrium. The household satellite account for a room that believes only in what is counted. The 92.9 percent figure for a room that thinks this is a fringe.

Best of all is the first time somebody uses one of your grade-A citations against you, correctly, in their own argument, for their own purposes. It means the citation has stopped being yours and become part of the furniture of the conversation, which is the only way an idea ever travels. You have not lost the point. You have won the vocabulary.

OPERATIONALIZE THIS

At the level of finance

An epistemology with no line in an engagement letter is a preference. Here is the instrument, in the form a client's procurement function will recognise immediately, because it is the shape of every performance-linked professional fee they have ever signed.

The structure: an advisory engagement with a graded claim schedule and a fee holdback released on independent review.

The mechanics.

The deliverable. The recommendation, plus a claim schedule in the five columns above, delivered as a numbered annexe to the engagement letter and referenced in it. The schedule is part of the deliverable, not a courtesy.
The holdback. Ten percent of the fee, retained by the client until their own reviewer — internal audit, a technical panel, an academic on retainer — has read the schedule. Released in full if no row is misgraded; released less the correction cost if a row is.
What counts as a misgrade. One definition, agreed in writing before the work starts: a claim graded A whose source does not support it as stated, or a claim graded B with no contest named. A claim that is wrong is not a misgrade — a correctly graded claim that later fails its kill condition is the system working, and the engagement letter should say so in one sentence.
Term. Review inside thirty days of delivery, holdback released on the earlier of the review or day sixty. A holdback with no long-stop is a discount.
Security. Nothing. This is a quality mechanism, not a credit one, and dressing it as credit will get it struck out by the client's counsel.

The balance-sheet treatment. The holdback is variable consideration under IFRS 15 and is constrained until the constraint is resolved — that is, you may not recognise it as revenue while the review is outstanding unless it is highly probable that a significant reversal will not occur. In practice a firm with two completed reviews and no misgrades can argue the constraint away; a firm with none cannot. Take it to your auditors at the first engagement rather than the fourth, because the conversation is about revenue recognition on variable consideration, which they have every year.

The counterparty. The client's internal audit function, not the sponsor. A review commissioned by the person who wants the recommendation to be right is a review with a known answer, and the whole instrument turns on the reviewer being someone whose standing does not move with the verdict — which is the same finding the model-risk literature reached expensively, and which Chapter II.11 operationalises as SR 11-7.

The number that decides it. One calculation, on the front page of the engagement letter:

  engagement fee                              GBP  120,000
  holdback rate                                       10.0 %
  holdback at risk                            GBP   12,000

  cost of grading the schedule
    3.0 days at GBP 900 fully loaded          GBP    2,700

  P(challenge), from the discard model
    at 3 loose claims in 20                             60.1 %
  expected holdback forfeited if ungraded     GBP    7,211
  ------------------------------------------------------------
  expected loss avoided / cost of grading              2.67 x
  break-even challenge probability                    22.5 %

Above a 22.5 percent chance that a reviewer finds one loose claim, grading the schedule is the cheapest professional indemnity available to you. Below it, grade the schedule anyway and charge for it — the client is buying a document they can defend after you have left the building, and that is worth more to them than the hours.

The first ninety days.

Day	Action	Artifact
1–15	Grade your last three proposals against the ledger, retrospectively	Three graded schedules, unpublished
16–30	Read the primary source for every claim you graded A	A corrected ledger, with the downgrades marked
31–45	Write your kill conditions; strike every grade C from the schedules	The kill list
46–60	Add the schedule and the holdback to the engagement letter template	The amended template
61–75	Find the second grader; agree the misgrade definition in writing	The signed definition
76–90	First engagement delivered with a schedule; first review commissioned	One reviewed schedule

Read alongside the other instruments in this edition, this one is the odd one out and deliberately so: it is the only instrument in the catalogue whose collateral is the practitioner's own accuracy, and the only one that pays a return in a currency — being believed next time — that never appears in a set of accounts.

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

Think of a claim you have made in the last year that a sceptic accepted immediately. What was it about the source that made that possible?
When has naming a weakness in your own evidence made a room more willing to act rather than less? What did you say, exactly?
Which of the results in this volume have you already checked at the source rather than taking from a summary — and what did you find there that the summary had lost?

Dream — what becomes possible

If every proposal you wrote carried a graded claim schedule, which conversation would you stop having entirely?
Imagine a room where the practitioner with the most downgrades on their ledger is the most trusted person present. What would you do differently on Monday?
If your grade-A citations were the only ones you were allowed to use, what would you have to go and find this quarter?

Design — what we build

What are the five claims your current largest piece of work actually rests on, and could two of us write them identically?
Which of your grade-B claims is closest to having its contest settled — and what would the joint test look like?
Who grades your schedule, and what would they need in order to tell you the uncomfortable answer?

Destiny — how it holds

What would have to be true for the ledger to still be current when you are no longer the person maintaining it?
What is the first sign you would see that your grades were inflating, and who is most likely to spot it before you do?
Ten years from now, which of these thirty results would you most want to have been the one you were careful about — and what does that tell you about what to check first?

WORKS CITED

Axtell, R. L. (2001). "Zipf Distribution of U.S. Firm Sizes." Science, 293(5536), 1818–1820.

Ayres, R. U., Ayres, L. W. and Warr, B. (2003). "Exergy, Power and Work in the US Economy, 1900–1998." Energy, 28(3), 219–273.

Beugelsdijk, S., de Groot, H. L. F. and van Schaik, A. B. T. M. (2004). "Trust and Economic Growth: A Robustness Analysis." Oxford Economic Papers, 56(1), 118–134.

Bianciardi, C., Tiezzi, E. and Ulgiati, S. (1993). "Complete Recycling of Matter in the Frameworks of Physics, Biology and Ecological Economics." Ecological Economics, 8(1), 1–5.

Bridgman, B., Dugan, A., Lal, M., Osborne, M. and Villones, S. (2012). "Accounting for Household Production in the National Accounts, 1965–2010." Survey of Current Business, 92(5), 23–36.

Bushe, G. R. and Kassam, A. F. (2005). "When Is Appreciative Inquiry Transformational? A Meta-Case Analysis." Journal of Applied Behavioral Science, 41(2), 161–181.

Cox, M., Arnold, G. and Villamayor-Tomás, S. (2010). "A Review of Design Principles for Community-based Natural Resource Management." Ecology and Society, 15(4), 38.

Dunbar, R. I. M. (1992). "Neocortex Size as a Constraint on Group Size in Primates." Journal of Human Evolution, 22(6), 469–493.

Edwards, J. and Ogilvie, S. (2012). "Contract Enforcement, Institutions, and Social Capital: The Maghribi Traders Reappraised." Economic History Review, 65(2), 421–444.

Gabaix, X., Gopikrishnan, P., Plerou, V. and Stanley, H. E. (2003). "A Theory of Power-Law Distributions in Financial Market Fluctuations." Nature, 423, 267–270.

Greif, A. (1993). "Contract Enforceability and Economic Institutions in Early Trade: The Maghribi Traders' Coalition." American Economic Review, 83(3), 525–548.

Grossman, S. J. and Stiglitz, J. E. (1980). "On the Impossibility of Informationally Efficient Markets." American Economic Review, 70(3), 393–408.

Guyatt, G. H., Oxman, A. D., Vist, G. E., Kunz, R., Falck-Ytter, Y., Alonso-Coello, P. and Schünemann, H. J. (2008). "GRADE: An Emerging Consensus on Rating Quality of Evidence and Strength of Recommendations." BMJ, 336(7650), 924–926.

Hall, C. A. S., Balogh, S. and Murphy, D. J. R. (2009). "What is the Minimum EROI that a Sustainable Society Must Have?" Energies, 2(1), 25–47.

Kahneman, D. and Deaton, A. (2010). "High Income Improves Evaluation of Life but Not Emotional Well-being." PNAS, 107(38), 16489–16493.

Killingsworth, M. A., Kahneman, D. and Mellers, B. (2023). "Income and Emotional Well-being: A Conflict Resolved." PNAS, 120(10), e2208661120.

Knack, S. and Keefer, P. (1997). "Does Social Capital Have an Economic Payoff? A Cross-Country Investigation." Quarterly Journal of Economics, 112(4), 1251–1288.

Kuznets, S. (1934). National Income, 1929–1932. Senate Document No. 124, 73rd Congress, 2nd Session. US Government Printing Office.

Lindenfors, P., Wartel, A. and Lind, J. (2021). "'Dunbar's Number' Deconstructed." Biology Letters, 17(5), 20210158.

Mantel, R. (1974). "On the Characterization of Aggregate Excess Demand." Journal of Economic Theory, 7(3), 348–353.

Martyushev, L. M. and Seleznev, V. D. (2006). "Maximum Entropy Production Principle in Physics, Chemistry and Biology." Physics Reports, 426(1), 1–45.

Mastrandrea, M. D., Field, C. B., Stocker, T. F. et al. (2010). Guidance Note for Lead Authors of the IPCC Fifth Assessment Report on Consistent Treatment of Uncertainties. Intergovernmental Panel on Climate Change.

Maturana, H. R. and Varela, F. J. (1980). Autopoiesis and Cognition: The Realization of the Living. D. Reidel.

May, R. M. (1972). "Will a Large Complex System be Stable?" Nature, 238, 413–414.

Mount, K. and Reiter, S. (1974). "The Informational Size of Message Spaces." Journal of Economic Theory, 8(2), 161–192.

Office for National Statistics (2018). Household Satellite Account, UK: 2015 and 2016.

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

Scarf, H. (1960). "Some Examples of Global Instability of the Competitive Equilibrium." International Economic Review, 1(3), 157–172.

Sonnenschein, H. (1972). "Market Excess Demand Functions." Econometrica, 40(3), 549–563.

Teubner, G. (1993). Law as an Autopoietic System. Blackwell.

Ulanowicz, R. E. (2009). A Third Window: Natural Life Beyond Newton and Darwin. Templeton Press.

U.S. Geological Survey (2024). Mineral Commodity Summaries 2024. Reserve and resource definitions in Appendix C.

Vitali, S., Glattfelder, J. B. and Battiston, S. (2011). "The Network of Global Corporate Control." PLoS ONE, 6(10), e25995.

Zak, P. J. and Knack, S. (2001). "Trust and Growth." The Economic Journal, 111(470), 295–321.

Note on figures. The ledger, the grade shares, the provenance count, the hypergeometric survival table under both reader models, the register and licence shares, and the holdback arithmetic are all computed in lib/verify/II_E1.py and reproduced by python3 lib/verify.py II.E1, which prints every input with its unit and its source before it prints a result. Results carried in from the eleven chapters are cited to the chapter that computed them and are not re-derived here. The discard model is this edition's own stated assumption about readers and is printed beside the weaker discount model precisely because it has never been measured.