Haute Lumière
Commerce · II.11 · MMXXVI · daylight
TEN CONCEPT BRIEFS · Chapter II.11 — The Case Against This Book
One page each. A reader who reads only these ten pages has the chapter.
BRIEF 1 — The Kill Condition
The idea. A claim is worth something in proportion to what it forbids. The kill condition is the observation that, if it appeared, would retire the claim — written down in advance, by the person making the claim.
Karl Popper's test is often taught as a rule about theories. It is more useful as a rule about sentences. Take any sentence in any business case and ask: what result would make this false? If nothing would, the sentence is not evidence. It may still be true, beautiful and worth saying. It cannot be used to win an argument.
Worked example. Two sentences about the same programme.
A — "Regenerative operations build the underlying stock over time." B — "A regenerative conversion starting 25 percent behind a conventional operation, with the conventional operation declining 3 percent a year and the conversion compounding 2 percent, crosses over in year six."
A forbids nothing; every outcome is consistent with it once "over time" is undefined. B forbids a great deal: a matched sample whose median crossover runs past year twelve kills it. B is the sentence that can lose, which is why B is the sentence that can persuade.
The discipline. A kill condition names three things — the study design, the sample, and the result. "Evidence to the contrary" is not a kill condition. "A matched-sample hazard study with the opposite sign and a confidence interval excluding one" is.
You already know this because you have sat in a meeting where somebody's position could not be dislodged by any fact, and you noticed that this made them less convincing rather than more.
BRIEF 2 — The Claim Register
The idea. A four-column table that converts a body of belief into an auditable inventory.
| Column | Contents |
|---|---|
| Claim | One sentence, written so two readers would test it identically |
| Class | F if it forbids an observation, U if it does not |
| Kill condition | For every F: the design, the sample, the result |
| Status | Standing · contested · retired, with a date and a name |
The audit of this edition. Twenty-four load-bearing propositions were drawn from the seven volumes and classified. Fifteen — 62.5 percent — are F. Nine — 37.5 percent — are U. All twenty-four print, with their kill conditions, from python3 lib/verify.py II.11.
The three rules.
- A U-claim may stay in the book, but may not be cited as evidence, appear in a term sheet, or be used to answer an objection.
- A kill condition may be tightened, never loosened. Amendments are versioned and the previous version stays visible. This is the rule that sustainability-linked bonds routinely fail, where a step-up coupon is avoided by restating the KPI rather than by meeting it.
- The reader of the register is not its author.
Why it matters. The denominator is the whole point. Twenty-four is not every sentence in the edition and it is not a random sample — it is a published list, so a reader who thinks a claim is misclassified has a specific line to name rather than a mood to express.
BRIEF 3 — The Degenerating Research Programme
The idea. Imre Lakatos's improvement on Popper. Whole programmes are rarely killed by one observation; they are judged by whether they are progressing or degenerating.
- A progressing programme predicts novel facts, and some of them turn out.
- A degenerating programme adds assumptions only to absorb results it did not predict, and forecasts nothing new.
Why this is the sharper test for us. Nobody is going to falsify "the economy is a living system" and nobody needs to. The live question is whether living-systems economics has told anyone something they did not already know and been right. The honest answer is sometimes: Ostrom's design principles predicted which commons endure, and they held up against 91 studies. The equally honest answer is that the paradigm's more metaphysical wing has a poor record of novel prediction and a good record of retrospective explanation.
The tell. When an unexpected result arrives and the response is a new distinction rather than a revised expectation, the programme is degenerating. "That case was not really regenerative" is the sentence to listen for.
You already know this because you can tell the difference between a colleague who updates and a colleague who reinterprets, and you trust the first one's forecasts.
BRIEF 4 — What a Price Knows
The idea. Hayek's argument in "The Use of Knowledge in Society" (1945) is not that markets are efficient. It is epistemic: the knowledge required to run an economy exists only as millions of fragments of local, tacit, transient circumstance, held by people who cannot articulate it and would not be believed if they did. A price is the only known device that aggregates it without collecting it.
The steelman against this book. Every "beyond GDP" dashboard, every regeneration-rate column, every impact metric is a centrally specified indicator. Somebody chose it, somebody defines it, somebody can be lobbied about it. It does not have the error-correcting property that makes a price useful — nobody loses money by being wrong about it, so nobody hurries to be right. A dashboard is a plan wearing the clothes of a measurement, and the socialist calculation debate is the recorded history of what happens next.
Where it holds. Fully, for goods that are excludable, rival, frequently traded and whose quality is observable at the point of sale. There the price beats any committee and the argument is over.
Where it does not. Where the effect is unpriced by construction — an externality, a stock nobody owns, a quality revealed years later. The Grossman–Stiglitz result adds the formal point that a price cannot fully reflect information that costs something to gather, because if it did nobody would pay to gather it.
The honest position. Prices are the best instrument available in their domain and the domain is narrower than the case for them usually admits. Both sentences.
BRIEF 5 — Capture
The idea. The public-choice objection to commons governance: a local institution is not automatically a democratic one. It is a smaller arena, and smaller arenas are cheaper to capture.
Mancur Olson's Logic of Collective Action (1965) supplies the mechanism. Concentrated interests organise; diffuse interests do not; so any governing body tends to be populated by those with most to gain per head. Bardhan and Mookherjee's work on decentralisation found exactly this in practice: devolving authority to local government improves outcomes where local accountability is strong and worsens them where it is not.
Applied here. "Let the community govern the resource" contains an unexamined noun. Communities have elders, creditors, landholders and incomers, and the rules that emerge usually favour whoever was in the room. Eduardo Araral's sceptical reading of the Ostrom programme (2014) presses precisely this point.
Why Ostrom is the answer and not the problem. Her design principles are a capture-resistance specification: clearly defined boundaries, congruence with local conditions, collective-choice arrangements that include those affected, graduated sanctions, accessible conflict resolution, minimal recognition of rights to organise, nested enterprises. Every one of those is a clause about who may not be excluded from the rule-making.
What remains conceded. Most of the durable cases involve between fifteen and a few thousand users, clear boundaries and long acquaintance. The scaling of these principles to an atmosphere shared by eight billion strangers is a hope that Ostrom herself described as a research agenda, not a finding.
BRIEF 6 — The Rarity Problem
The idea. Henry Hansmann's question in The Ownership of Enterprise (1996), and the hardest question in this book: if worker ownership is more productive, more resilient and more humane, why is there so little of it?
The number. In the United States there are on the order of 1,000 worker cooperatives employing roughly 10,000 people, against 160 million in work. That is 0.00625 percent — 6.25 workers in every hundred thousand.
Hansmann's answer, and it is not the one either side expects: the binding cost is not capital and not hostility. It is the cost of collective decision-making, which rises steeply with the heterogeneity of the owners' interests. Where employees are homogeneous — the plywood mills, the law firm, the taxi fleet — worker ownership appears and persists. Where they differ sharply in skill, tenure, risk appetite and time horizon, the governance cost swamps the benefit.
The testable consequence, which is why this brief matters more than it looks: Hansmann's theory predicts where cooperatives will be found, and it is broadly right about where they are. That is a novel prediction that came out — so the critique here is the progressing programme and the reply is the one under pressure.
What an honest paradigm does with that. Not deny it. Use it: design for homogeneity of interest where you can, and price the governance cost where you cannot. "Capital is hostile" is not an answer, because capital is hostile to many things that are nevertheless everywhere.
BRIEF 7 — Illyria, and the Horizon Problem
The idea. Two formal results that predict worker-managed firms will behave badly, and what the evidence did with them.
Ward's perverse supply response (1958). A firm maximising income per worker rather than total profit responds to a price rise by shrinking employment, because a smaller group divides the same surplus more richly. A whole economy of such firms would have a backward-bending supply curve. Domar (1966) and Vanek (1970) developed the model.
The horizon problem (Jensen and Meckling, 1979). A member whose claim on the firm ends when their employment does will under-invest in anything that pays back after they leave. Long-lived assets are systematically starved.
What happened when people looked. The perverse supply response is not generally observed. Craig and Pencavel found the Pacific Northwest plywood cooperatives adjusting pay where conventional mills adjusted headcount — which is the opposite behaviour to the model and, incidentally, the behaviour the cooperative case claims. Fakhfakh, Pérotin and Gago (2012) found French worker cooperatives at least as productive as comparable conventional firms. Burdín (2014) found Uruguayan worker-managed firms less likely to dissolve than conventional ones.
The honest reading of all four. The theory predicted a failure mode; the data did not find it; the theory's defenders reply that the surviving cooperatives are selected. That reply is testable and has not been settled, which is the correct place for an argument to be.
Why it matters. The horizon problem, unlike the Illyrian one, is real and is handled by an instrument: internal capital accounts that pay out on departure — Mondragon's central mechanism, and one of the few genuine institutional inventions in this field.
BRIEF 8 — Decoupling, Absolute and Relative, Territorial and Consumed
The idea. Four words that decide most arguments about growth, and they are routinely swapped mid-sentence.
- Relative decoupling — emissions per unit of output fall. Nearly universal.
- Absolute decoupling — emissions fall while output rises. Real, and demonstrated.
- Territorial — what is emitted inside the border.
- Consumption-based — what is emitted anywhere to serve domestic demand, including imports.
The arithmetic. UK territorial greenhouse gases fell from 811 to 417 MtCO₂e between 1990 and 2022: 48.6 percent, with output growing. Absolute decoupling, unambiguously. The consumption-based footprint fell over its comparable window by 15 to 20 percent. The territorial rate is therefore 2.43 to 3.24 times the consumption rate, midpoint 2.78×.
The rate question, which settles it. Global fossil CO₂ ran 37.4 GtCO₂ in 2023 against a remaining 275 GtCO₂ for an even chance at 1.5 °C — 7.35 years. Halving by 2035 needs 6.11 percent a year; with output at 3.0 percent, carbon intensity must fall 8.84 percent a year. It fell 1.45 percent a year from 1990 to 2019. Required over observed: 6.08×.
What this settles and what it does not. It does not refute green growth. It prices it: anyone holding that position now owes an account of how a rate sextuples. That is a far better argument than the one where each side asserts that decoupling is or is not possible, because it has a number in it and the number can be checked next decade.
BRIEF 9 — The Decomposition Test
The idea. Before trusting a composite index, take it apart and ask which component is doing the work. If one term carries the result, the index is that term wearing a costume.
Case one: the Ecological Footprint. Humanity's demand is reported at 1.71 Earths. The carbon component is about 60 percent of it — 1.026 Earths. Everything else — crops, grazing, timber, fish, built land — is 0.684 Earths. The overshoot above one planet is 0.71, and the carbon term alone exceeds the entire overshoot by 0.316 Earths.
So "we are using 1.7 Earths" means, arithmetically, "our carbon emissions restated as the forest area that would absorb them." Blomqvist and colleagues (2013) and Giampietro and Saltelli (2014) are right, and the index should not be cited as an independent measure of ecological limits.
And here is the part worth teaching. The finding survives the instrument's death, because the term carrying all of it — atmospheric carbon — is the one quantity with three independent physical checks: the Mauna Loa record, ocean pH, and ocean heat content. The composite fails the audit; the conclusion never needed the composite.
Case two: the ISEW threshold. Neumayer's charge is that the famous post-1970s divergence between welfare indices and GDP is produced by the cumulative depletion term, which rises monotonically by construction. Demonstrate the mechanism on a synthetic series: welfare 100 growing 1.5 percent a year, first-year depletion charge 1.0 index points.
| Depletion escalates at | Index turns down in |
|---|---|
| 3.0 %/yr | year 27 |
| 2.0 %/yr | year 81 |
| 1.5 %/yr | never |
| 1.0 %/yr | never |
The threshold is one assumption, read back. Run the decomposition before you cite the index, always, including when the index agrees with you.
BRIEF 10 — Effective Challenge
The idea. Criticism is a function with a budget, a reporting line and a supervised standard — not a virtue.
The Federal Reserve and OCC's Supervisory Guidance on Model Risk Management (SR 11-7, 2011) requires banks to hold a model inventory and to validate every model independently of its developers. Its central phrase is effective challenge, and the guidance is explicit that it requires three things together: competence, incentive and influence. Remove any one and validation becomes ceremony.
Applied to a paradigm. The model inventory is the claim register. The validation trigger is a change to a kill condition. The verdicts are three and never a score: standing · contested · retired. Scores get averaged, averages get reported, and reported averages stop being read.
The inequality that decides the budget.
P(claim wrong) x loss if wrong x P(validation detects it)
-------------------------------------------------------- > 1
validation budgetWorked. A £4,000,000 programme; a 15 percent chance the load-bearing assumption is wrong; 60 percent of the programme lost if it is — an expected loss of £360,000. At a 60 percent detection rate, £216,000 of expected loss avoided against a 3 percent validation budget of £120,000: a ratio of 1.80×. Break-even detection rate: 33.3 percent.
The soft term, named rather than buried. P(detect) is the only term the firm controls and the only one that can be gamed to zero. A validator without independence, competence or influence detects nothing at any budget — which is why a validation unit that has never overturned anything should be read as having found nothing, not as having confirmed everything.
All figures in these briefs are computed in lib/verify/II_11.py, printed with their inputs by python3 lib/verify.py II.11, and sourced in the chapter's Works Cited.