Haute Lumière
Commerce · II · MMXXVI · daylight
For the Corporation
Volume II — Foundations: The Paradigm and the Science · Extension
Nine movements, one defence file.
THE PLATE
THE LETTER
You are not being asked whether this is true. You are being asked whether you will put your name to it.
That is a different question and it wants a different apparatus. A board does not adjudicate paradigms; it allocates capital and carries liability. An auditor does not care what an economy resembles; they care whether your useful-life estimate is supportable. A regulator does not read a chapter; they read a disclosure, and then they read the next one, and the thing they are looking for is a number that moved without an explanation.
So this extension reads the eleven chapters of Volume II with one question in front of it: which of these findings survives a hostile review, what is the strongest argument against each, and how do you hold a position that is genuinely uncertain without either overclaiming it or abandoning it?
The answers are unglamorous and they are countable. Of ten findings a corporation would actually want to use, three survive a hostile review outright, five survive inside a boundary that you must state before your critic does, and two should be conceded and left outside the room. The two you concede are not the weak ones you would guess. One of them is the sentence on the cover.
Chapter II.02 already did the generous half of this work: it listed five questions where the neoclassical instrument is simply the better tool, and called that list the boundary of the claim. This chapter does not repeat that list. It proposes the rule that generates it, tests the rule against II.02's five, tells you which one the rule misses and why, and then uses it to find three more rows nobody had listed and five where the instrument in this volume wins outright.
And then the part that most of this literature never reaches: what it is worth, in money, to say between eighteen and thirty percent to a board instead of thirty. It is worth 6.0 percent of the programme, it is computed in lib/verify/II_E2.py, and the mechanism is not virtue. It is that a stated band pre-agrees the response.
— The Editors
DISCOVERY
What is already working
Start with institutions that have already defended a contested position in front of exactly the audiences you face, and won — not by softening the claim, but by bringing it in the form the audience adjudicates in.
The Bank of England's housing model. In 2016 a central bank — the least romantic institution in the world — published an agent-based model of the UK housing market and used it to examine how loan-to-income limits would propagate. What made it survive was not that the method was fashionable. It was that the model answered a question the existing models could not answer at all: what happens at the distribution rather than at the average. A heterodox instrument entered a conservative institution by being the only tool that could answer a question the institution had already decided it needed answered. That is the entry route, and it is the only one with a record.
PUMA's Environmental Profit and Loss Account. In 2011 a listed company put a currency figure on its environmental impact across the whole value chain and published it, knowing it would be argued with. It was argued with, extensively, and it is still the clearest single demonstration of what happens to a company's self-portrait when the fence moves outward: €145 million of environmental cost for 2010, of which €8 million arose in PUMA's own operations. The defence was built into the publication — method disclosed, tier by tier, in euros.
Philips at Schiphol. In 2015 an airport bought light and a manufacturer kept the fittings. The contract is quoted everywhere as a sustainability case and it was signed as a balance-sheet case: capital expenditure moved off one party's books and title stayed with the other. The argument that got it through two finance functions was about title and working capital, not about materials, and the environmental outcome arrived as a consequence rather than as a premise.
SR 11-7, and the paid critic. In 2011 the Federal Reserve and the OCC required banks to validate every model independently of its developers and to document the effective challenge that validation provides — challenge that requires competence, incentive and influence together. A supervised standard now exists for paying somebody to attack your own numbers, with a defined output and an escalation path. A corporation adopting the register in Chapter II.11 is not inventing governance; it is pointing an existing, examined standard at a different object.
The IFRS 9 amendment of 2024. For several years lenders had a genuine drafting objection to sustainability-linked margin ratchets: a contingent cash-flow feature raised a question about the solely-payments-of-principal-and- interest test. The IASB's 2024 amendments settle it, effective 1 January 2026. The objection was real, it was technical, it was answered technically, and the answer is now citable in a term sheet. This is what a resolved contest looks like, and it is worth studying because most of the objections in this volume will be resolved the same way or not at all.
Five cases, one pattern: in each, a contested position was carried into a conservative venue in the form that venue adjudicates in — a distribution, a disclosed method, a title clause, a validation charter, an accounting standard. Not one of them won by being persuasive about a worldview.
THE ARITHMETIC
What survives review, what it costs to be honest, and where the honesty loses
Every figure below is computed in lib/verify/II_E2.py, which prints each input with its unit and its source before it prints a result.
First, the defence file. Ten findings a corporation would actually want to use, each with its strongest counter-argument, graded by what happens when somebody competent attacks it.
verdict findings share
------------------------------------------------------------------
SURVIVES a reviewer concedes it 3 30.0 %
BOUNDED it holds inside a stated boundary 5 50.0 %
CONCEDE do not take it into the room 2 20.0 %
------------------------------------------------------------------
total 10 100.0 %The three that survive. That aggregate behaviour is not derivable from the rationality of the parts — a theorem in the discipline's own journals, and the only counter available is that the representative agent is used anyway, which is a claim about convenience rather than about the theorem. That the consolidation boundary was repriced rather than abolished in 2003, by a factor of 3.33 — there is no counter, because that is what the standards say. And that a KPI under pressure degrades, and so does the thing it measures — Goodhart and Campbell, in their own words, with every auditor in the room already on your side.
The five that are bounded. These are the useful ones, and the boundary is the defence.
Size the buffer on expected shortfall, not Gaussian value at risk. The multiple, recomputed here rather than quoted: expected shortfall at the 99 percent point under a cubic tail is 4.04 sigma; Gaussian value at risk at the same point is 2.3263 sigma; so k is 1.7366, or 1.74×, and a Gaussian-sized buffer holds 42.4 percent less than the average loss in the tail it was built for. The counter is not about the arithmetic. It is about your sample, and it is the honest negative below.
Below a tail index of 1.0780, compartmentation beats full connection. Same estimation problem, plus the fact that diversification is the supervisory default and you will be arguing against the presumption.
A scope 1+2 fence captures between 5.5 and 8.1 percent of what a firm causes. The counter is sharp and correct: chain inventories overcount — at ten tiers by a factor of 5.50 — so the wider number is not yours to claim either. The defensible position is precise: report the wider boundary, never claim a chain reduction as your own achievement, and say both sentences in the same paragraph.
A ninety percent boiler is 6.20 percent second-law efficient and a heat pump is 4.03 times it. Carnot arithmetic, which nobody will dispute — bounded by the threshold Chapter II.06 states against itself: below roughly 2.0 percent of total cost, an exergy account is bookkeeping rather than management.
A five percent maintenance funding shortfall costs about a third of an asset life. The convexity is real; the decay rates behind the worked figure are modelled. Calibrate on your own asset history before it goes in a board paper, and say that you did.
The two to concede, and they will surprise you. The first is that useful work carries most of the Solow residual at an elasticity near 0.697 — a beautiful result that requires factor markets to misprice energy by an 8.3-times wedge against its cost share, which contradicts the first-order condition of production theory. It is a wonderful thing to believe and an indefensible thing to put in a capital paper. The second is the economy is a living system, which forbids no observation and therefore cannot support a business case. The edition classes it U in its own register; do the same and you have lost nothing, because none of the fifteen falsifiable claims depends on it.
Second, the cut — the rule that decides which vocabulary to use, and it is not about what you believe.
Chapter II.02 listed five questions where the neoclassical instrument remains the better one and called that list the boundary of the claim. It is the right list. What it lacks is a generating rule, and a list without one cannot be extended to the question you actually have on Thursday.
Here is the rule.
The neoclassical instrument wins wherever the decision will be adjudicated by a third party who must be able to reconstruct the counterfactual from the document alone.
Test it against II.02's five, because a rule that is not tested against the list it claims to explain is a slogan.
retrodicted allocating radio spectrum
losing bidders and a court must be able to check the rule
retrodicted matching doctors to hospitals, kidneys to patients
deferred acceptance is provably stable; a simulation is not
retrodicted tax incidence and comparative statics
a court or a tribunal needs a legible counterfactual
retrodicted index numbers and the cost of living
a statistical authority must publish a reproducible method
MISSED one competitive market, short horizon, stable tastes
nobody adjudicates this; the tool wins on parsimony instead
rows in II.02's table 5
retrodicted by the rule 4 80.0 %
missed 1 20.0 %Four of five, and the miss is the informative part. The fifth row is not about adjudication at all: extra structure there adds parameters without adding accuracy, and the neoclassical tool wins on parsimony. So the honest statement is that two criteria govern the boundary, not one — adjudication and parsimony — and this chapter has found the first.
Now use it forwards. The rule generates three rows II.02 does not list, and each is a real corporate decision: intra-group transfer pricing, where a tax authority adjudicates on the OECD guidelines; a regulated tariff or allowed return, where a regulator adjudicates on a published method; and a contractual damages claim, where a court must reconstruct the but-for world. In all three, reach for the orthodox instrument and do not apologise for it.
And it names five rows where the instrument in this volume wins outright, because the decision is yours and the counterfactual is internal: sizing a liquidity buffer, choosing a network position or a firebreak, setting a circulation target, funding a maintenance-of-order reserve, and electing a reporting boundary — where the election is yours and only the disclosure is adjudicated. Thirteen rows in all.
Read what that does to the question you thought you were being asked. Which economics you use is a venue decision, not a truth decision. The same corporation, in the same week, correctly uses partial equilibrium in a transfer pricing file and an expected-shortfall buffer in a treasury paper, and there is no inconsistency in it whatsoever — any more than there is in using a micrometer on a shaft and a tape on a room. A board that has been shown the rule stops asking you to declare an allegiance, which is the single most useful thing this extension can do for you.
Third, the honest negative — and it lands on the finding this volume is proudest of.
The network result says that below a tail index of 1.0780 a compartmented structure beats a connected one. Suppose your own estimate comes in at 1.30. Hill estimator asymptotics give a standard error of α/√n in the number of tail observations:
threshold tail index, alpha* 1.0780 (II.05)
your estimate, alpha-hat 1.3000
tail loss events in the sample 40
standard error 0.2055
95 % interval 0.897 to 1.703
does it clear the threshold? no — it straddles alpha*
tail events needed to clear it 132One hundred and thirty-two tail loss events before this decision is made by your data rather than by your preference. Most firms have forty. Most firms have fewer.
Do not read that as a defeat, because Chapter II.05 already told you what to do with it: if the estimate straddles the threshold you have learned that your structural choice is not determined by the data and must be made on other grounds — which is itself a finding worth having in writing. Put exactly that sentence in the board paper. It is a stronger paper for it, and the reason is the next block.
Fourth: what an honest band is worth, computed.
Take Chapter II.11's worked programme and its inputs, which are not ours to soften: a £4,000,000 commitment whose load-bearing assumption has a 15 percent chance of being wrong, where failure costs 60 percent of the programme.
programme at risk GBP 4,000,000
P(the load-bearing claim is wrong) 15.0 %
loss given wrong, point estimate 60.0 %
loss given wrong, band + pre-agreed response 20.0 %
expected loss, point estimate GBP 360,000
expected loss, honest band GBP 120,000
------------------------------------------------------------
difference GBP 240,000
as a share of the programme 6.0 %Six percent of the programme, for the cost of writing an interval.
And the mechanism matters more than the number, because it is what you say when somebody calls the band a hedge. The interval does not reduce the chance of being wrong — it is the same 15 percent in both columns. What it changes is the loss given wrong, and it changes it because a stated band forces a pre-agreed response: a trigger, a review date, a named person, a decision already taken about what happens if the low end turns out to be the truth. A point estimate has no such machinery, so being wrong arrives as a surprise, in a bad quarter, with nothing drafted.
That is the position to hold in front of a board, and it can be said in one sentence: I am not less certain than my competitor. I have priced my uncertainty and they have not.
Fifth, the second honest negative, and it is the one nobody writes down.
The honest band loses an unregulated tender, and the loss is computable.
rival's asserted saving, a point 30.0 %
the band your evidence supports 18.0 % to 30.0 %
your midpoint 24.0 %
credibility haircut the evaluator must apply
to the point estimate for the honest bid
to score level 20.0 %Unless the evaluator discounts the point estimate by at least a fifth, the honest bid loses on the number. In an unregulated beauty contest, scored by somebody with a spreadsheet and no liability, that discount is rarely applied. This is a real cost of the discipline recommended in this chapter and it is not recoverable by being more rigorous.
What it is recoverable by is venue, and there are three where a stated interval is priced rather than punished. An audited financial statement, where an unsupportable point estimate is a misstatement and your auditor is on the side of the band. A lender's covenant, where the interval becomes a trigger and the lender prefers a borrower who has drafted one — the boundary-lock clause of Chapter II.03 is exactly this move. A regulatory filing, where the calibrated-uncertainty conventions of the IPCC guidance note and the SR 11-7 validation opinion already exist as accepted formats. Choose the venue before you choose the vocabulary, which is the rule above, applied to yourself.
DREAM
What becomes ordinary
In the corporation that has absorbed this, the board pack carries an interval where the evidence supports one and a point where it does not, and nobody reads the interval as weakness.
The defence file sits behind the strategy paper as a standing annexe: each load-bearing finding, the strongest argument against it, and the boundary inside which it holds. It is written by the person proposing and reviewed by somebody who is not. Directors read it before the paper, the way an analyst reads the notes before the face of the accounts, and the discussion starts at the boundaries rather than at the conclusion. Meetings are shorter.
Nobody in the building is asked which economics they believe in. They are asked who adjudicates the decision, and the vocabulary follows from the answer. A transfer pricing file is orthodox because a tax authority will read it; a treasury buffer is sized on the tail because nobody outside adjudicates it; and the same analyst writes both in the same week without any sense of contradiction.
Estimation error is reported rather than smoothed. When a structural choice is not decided by the firm's own sample, the paper says so and names the number of observations that would decide it — and somebody starts collecting them, because a target that has a count attached gets a collection plan.
The concessions are published. A corporation that has struck two findings from its own business case is trusted on the twelve it kept, and the trust is measurable: it appears as a shorter due diligence, a tighter margin, a lower assurance fee. Being known as the firm that concedes accurately turns out to be a commercial asset, and the finance function is the first to notice.
And the paradigm has stopped being the subject. The building is full of people using instruments — a buffer multiple, a circulation fraction, an exergy column, a boundary register — and nobody needs to agree about what an economy is in order to use them. That is not a defeat for the idea. It is the idea having become ordinary, which is the only form in which an idea ever changes a balance sheet.
DESIGN
The defence file, built
One artifact, four columns, and a fortnight.
| Column | What goes in it |
|---|---|
| The finding | One sentence, stated so that two readers would test it identically |
| Verdict | Survives · bounded · concede |
| The strongest counter | Written by the person who would make it, not by you |
| The boundary | For every bounded row: the condition inside which it holds, and the number |
How it is built, and the order is the whole of it.
Days one to five: write the counters first. Before you write a single finding, list the arguments that will be made against the paper. If you cannot state the opposition's case in a form they would sign, you do not yet understand your own. This is the move Chapter II.02 makes when it honours the Walrasian programme before turning, and the turn only earns its force because the honouring was real.
Days six to ten: grade. Survives, bounded, concede. Be ruthless with survives — a finding survives only if the counter is about convenience or about interpretation, never if it is about the number.
Days eleven to fourteen: write the boundaries as numbers. Not in most circumstances — 2.0 percent of total cost, 132 tail events, ten tiers and a 5.50-times overcount. A boundary without a number is a caveat, and a caveat is the thing a hostile reader uses against you.
Rule one: you state the boundary before your critic does. A boundary you disclose is a demonstration of competence. The identical boundary disclosed by a reviewer is the end of the meeting. This is the single highest-return habit in the file and it costs nothing but sequence.
Rule two: a concession is struck from the business case, not softened in it. Chapter II.11's first rule, imported: an unfalsifiable claim may stay in the prose, may not appear in a term sheet, and may not be used to answer an objection. When somebody brings a number and receives a paragraph about interconnection, the file has already stopped working.
Rule three: the reviewer is not the proposer. Internal audit is the usual home, is usually already funded, and — under SR 11-7's shape — has a defined output and an escalation path. A validator embedded in the programme for three years has colleagues; rotate.
Sequencing the adoption, cheapest first, because each step pays for the next. The defence file itself is two weeks of one analyst and no capital. The interval discipline is a template change: every material estimate in a board paper carries a band and a pre-agreed response, and that is a paragraph in the paper template. The assurance instrument in the next movement is the only step that spends money, and by the time you reach it the first two have told you exactly which metric is worth assuring.
What to resist. The file is not a campaign to make every claim heterodox, and it is not a campaign to make every claim orthodox. Half its value is in the rows where it tells you to reach for the conventional tool and stop apologising — and a file that never does that has been written by somebody defending a position rather than choosing an instrument.
DESTINY
How it holds when the sponsor moves on
Three conditions, and the third is the one people skip.
It is in the standing pack. A defence file reviewed annually alongside the strategy paper survives a change of sponsor, which is the event that kills most of these. Reviewed by exception, it is reviewed once.
Somebody's appraisal does not depend on the verdict. This is the finding from model risk management and it was learned expensively: effective challenge needs competence, incentive and independence together, and removing any one reduces the function to ceremony.
It is used at least once a year to win something. An instrument that has never been used in an argument gets quietly dropped from the pack. Use it the first time a comparison with a competitor turns on a boundary — and it will, because most of them do.
Now the failure modes, named.
It fails when the verdicts inflate. Every finding you like drifts toward survives, and the drift does not feel like dishonesty; it feels like having thought about it more. The guard is the counter column: a survives row whose counter column is empty is a bounded row with the work not done.
It fails when a boundary is quietly loosened at revision — the sustainability- linked bond failure, where a step-up coupon is avoided by restating the KPI. Version the file; keep the previous boundary visible.
It fails when the file is built by a sustainability function rather than by the controller, because then it describes findings that govern no money, and a finding that governs no money is a diagram.
And it fails, most often, when the corporation wins an argument with it and then stops. The file's value is that it is current, and a current file requires somebody to go and check whether a boundary has moved — whether the tail estimate has thirty more observations this year, whether the IASB has settled the objection, whether the counter-argument has been published in a stronger form by somebody who read the first one.
DELIGHT
What it feels like
There is a particular pleasure in watching an objection arrive that you have already written down, in the words the objector would have used, on page two of your own annexe. The room changes temperature. The objection is still correct and it has stopped being an attack, because it has become part of the document.
And there is a quieter one, which belongs to the rule. The first time somebody asks you whether you believe the economy is a living system and you answer that depends on who is adjudicating the decision — and the question dissolves, because it was never a question about economics, it was a question about which tool to bring — you will feel a specific kind of lightness. You have been carrying an allegiance. You can put it down.
Best of all is the concession. Striking two findings from your own business case is a small, slightly frightening act that takes about a minute, and what comes back is out of all proportion: the twelve that remain are believed. Nothing else available to a corporation buys credibility that cheaply, and almost nobody spends it.
OPERATIONALIZE THIS
At the level of finance
The file is the discipline. The instrument that makes it worth money is limited assurance over the single contested metric, paired with a pre-agreed restatement protocol — and both halves already exist as standards, which is why this can be approved by a committee that already meets.
The structure. An ISAE 3000 limited assurance engagement over one metric — the one your defence file grades bounded and your business case depends on — performed by your existing auditor, with two additions the market does not currently standardise.
- The frozen boundary. The metric is measured on a stated organisational perimeter, named in the engagement by reference to the elected consolidation and emissions bases. Any change to that perimeter triggers restatement of the baseline on the new perimeter. This is Chapter II.03's boundary lock, carried into the assurance engagement rather than into the loan, so it protects the metric even where there is no facility.
- The restatement protocol. Agreed before the first measurement: who restates, on what trigger, by when, and how the prior period is presented alongside. An unagreed restatement protocol is a future dispute, exactly as an unagreed baseline is.
The counterparty. The existing auditor, at the existing engagement, where the marginal cost of adding a scope is far below the cost of a separate appointment. Take it to them in the planning meeting rather than after the year-end, because the conversation is about scope and materiality, which they have every year.
Balance-sheet treatment. The assurance fee is an operating cost. Where the assured metric supports a useful-life estimate or a provision — a regenerated asset life, a maintenance obligation, a restoration duty — the assurance opinion is the evidence supporting that estimate, and your auditors will want it long before you want to write it. Do not capitalise the assurance itself; its useful life is exactly as long as the uncertainty it resolves, which is not an asset life anybody will recognise.
The number that decides it. One inequality, on the first page:
limited assurance, ISAE 3000 EUR 120,000 /yr
P(challenge in a year) 20.0 %
cost of a forced restatement EUR 3,000,000
P(assurance prevents or contains it) 60.0 %
---------------------------------------------------------------
expected loss avoided EUR 360,000
ratio to cost 3.00 x
break-even challenge probability 6.67 %Above a 6.67 percent annual chance that the metric is challenged, the assurance is cheaper than the event it insures. Most corporations have never computed the right-hand side, and the act of computing it is worth more than the engagement — because the first thing it produces is a written estimate of what a restatement would actually cost you, which nobody in the building currently holds.
The first ninety days.
| Day | Action | Artifact |
|---|---|---|
| 1–15 | Write the counters to your own strategy paper, in their words | The counter list, unedited |
| 16–30 | Grade the findings; strike every concession from the business case | The defence file |
| 31–45 | Put a number on every boundary; compute your own tail estimate and its interval | The boundary schedule |
| 46–60 | Amend the board paper template: every material estimate carries a band and a pre-agreed response | The amended template |
| 61–75 | Scope the assurance with the auditor; agree the frozen perimeter and the restatement protocol | The signed perimeter statement |
| 76–90 | First assured measurement; first defence file to the audit committee | The committee minute |
Read alongside the other instruments in this edition, this one is the hinge: it is the only instrument in the catalogue whose purpose is to make the other instruments financeable, by turning a contested number into an assured one that a lender, an auditor and a regulator will each accept in the form they read.
APPRECIATIVE QUESTIONS
Twelve, for a room
Discovery — what is already working
- When has this company taken a genuinely contested position to a board or a regulator and been believed? What form did we put it in?
- Which of our numbers would survive a hostile reviewer today without any further work — and who here already knows which they are?
- Think of a time somebody named the weakness in their own paper before anyone asked. What happened in the room, and what did it cost them?
Dream — what becomes possible
- If every material estimate in our board pack arrived with a band and a pre-agreed response, which decision would change first?
- Imagine our reviewers were the best-paid analysts in the function. What would we find out in the first year that we do not know now?
- If nobody in this building were ever asked which economics they believed in, only who adjudicates the decision, what would we stop arguing about?
Design — what we build
- Which of our load-bearing findings are bounded rather than surviving, and is the boundary written as a number anywhere?
- What is the one metric where a challenge would be most expensive, and have we ever estimated what a restatement would actually cost?
- Who writes the counter-arguments to our own papers, and what would they need in order to write them properly?
Destiny — how it holds
- What would have to be true for the defence file to still be reviewed when everyone in this room has moved on?
- What is the first sign we would see that a boundary had been loosened rather than tightened, and who would notice it first?
- Ten years from now, which of our concessions would we most want to have made early — and what does that suggest we concede this quarter?
WORKS CITED
Acemoglu, D., Ozdaglar, A. and Tahbaz-Salehi, A. (2015). "Systemic Risk and Stability in Financial Networks." American Economic Review, 105(2), 564–608.
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.
Baptista, R., Farmer, J. D., Hinterschweiger, M., Low, K., Tang, D. and Uluc, A. (2016). "Macroprudential Policy in an Agent-Based Model of the UK Housing Market." Bank of England Staff Working Paper No. 619.
Board of Governors of the Federal Reserve System and Office of the Comptroller of the Currency (2011). Supervisory Guidance on Model Risk Management. SR Letter 11-7 / OCC Bulletin 2011-12.
Campbell, D. T. (1976). Assessing the Impact of Planned Social Change. Occasional Paper Series, Public Affairs Center, Dartmouth College.
CDP (2021). Transparency to Transformation: A Chain Reaction. CDP Global Supply Chain Report.
Financial Accounting Standards Board (2003). FIN 46(R): Consolidation of Variable Interest Entities.
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.
Goodhart, C. A. E. (1975). "Problems of Monetary Management: The UK Experience." Papers in Monetary Economics, Reserve Bank of Australia.
Hill, B. M. (1975). "A Simple General Approach to Inference About the Tail of a Distribution." Annals of Statistics, 3(5), 1163–1174.
International Accounting Standards Board (2011). IFRS 10: Consolidated Financial Statements.
International Accounting Standards Board (2024). Amendments to the Classification and Measurement of Financial Instruments (Amendments to IFRS 9 and IFRS 7).
International Auditing and Assurance Standards Board (2013). ISAE 3000 (Revised): Assurance Engagements Other than Audits or Reviews of Historical Financial Information.
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.
Milgrom, P. (2004). Putting Auction Theory to Work. Cambridge University Press.
Organisation for Economic Co-operation and Development (2022). OECD Transfer Pricing Guidelines for Multinational Enterprises and Tax Administrations.
PUMA SE (2011). PUMA's Environmental Profit and Loss Account for the Year Ended 31 December 2010.
Roth, A. E. and Peranson, E. (1999). "The Redesign of the Matching Market for American Physicians." American Economic Review, 89(4), 748–780.
Sonnenschein, H. (1972). "Market Excess Demand Functions." Econometrica, 40(3), 549–563.
World Resources Institute and World Business Council for Sustainable Development (2011). Corporate Value Chain (Scope 3) Accounting and Reporting Standard.
Note on figures. The defence file and its shares, the adjudication rule's retrodiction against Chapter II.02's five rows, the tail-index interval and the sample size that would clear it, the buffer multiple recomputed from the normal quantile, the value of an honest band and the credibility haircut it needs in a tender, and the assurance break-even are all computed in lib/verify/II_E2.py and reproduced by python3 lib/verify.py II.E2, which prints every input with its unit and its source before it prints a result. Figures carried in from the eleven chapters — the cubic-tail expected shortfall, the threshold tail index, the consolidation repricing, the scope multiples, the second-law efficiencies — are cited to the chapter that computed them and re-derived here only where a division makes them checkable.