Not Significant

Every figure below obeys one rule, declared here and never varied: the boundary of the ledger is drawn as a boundary. Quantities the accounting system records are solid. Quantities that are real but appear on no invoice are drawn open, at the identical size they would occupy if anyone had written them down.

Counted

A quantity that enters someone’s books. It has a payer, a payee, and standing.

Uncounted

A real quantity with no invoice. Drawn open at full size, never faded, never dropped. Fading it would concede the argument.

Priced register

Measurements of what markets and courts do respond to: second-party wrongs, penalties, valuations.

Measured harm

Damage that has been quantified by researchers, whether or not it was ever charged to anyone.

Where two quantities were measured in different units, they are not forced onto one axis. Several figures below are deliberately two panels rather than one chart, for that reason and stated each time.

There is a finding in this material I cannot put down, and it is not one of the large numbers.

When a firm is punished for a wrong done to someone who does not trade with it (a neighbour, a watershed, a person standing downwind) the share price moves by 0.24 percent. Not significant. Which is the technical way of saying: inside the noise. Inside the ordinary tremor of a Tuesday. The market has been informed that a company injured a stranger, and the market, listening with all its formidable attention, hears nothing.

Now hold that against its opposite. When a firm lies to the people who bought its stock, the reputational damage runs to more than seven and a half times everything the legal and regulatory system takes from it. Break the dollar down: for every dollar of value a company inflates by deception, it loses about thirty-six cents to penalties and two dollars seventy-one to reputation. The market is not indifferent to wrongdoing. It is exquisitely, expensively, almost vengefully sensitive to it.

It simply has an ear for one register and is deaf in the other.

This is not a scandal. It is a specification. A price is a tally of the people who showed up to bargain. It is the most efficient information-processing device our species has built, and its genius is precisely its boundary: it knows what someone was willing to pay, and what someone else was willing to refuse, and nothing whatsoever about anyone who was not in the room.

The person downwind was not in the room. She was never invited. Her lungs are not a counterparty.

Everything else in this essay follows from that one sentence.

Begin small, with the most elegant formula in microeconomics.

The Lerner index says that a firm’s markup over its own cost is the reciprocal of how easily you can walk away: L = (P − MC)/P = 1/|ε|. If your demand elasticity is four, if a ten percent price rise sends four in ten of you elsewhere, the firm can hold a quarter of the price as margin. Marginal cost of seventy-five dollars becomes a price of a hundred.

Read it the other way and it stops being algebra. Every price you have ever paid contains an estimate of how hard it would have been for you to leave. Brand, loyalty programmes, switching costs, the ecosystem that makes your photographs hostage to a login, these are not marketing. They are capital expenditure on your immobility. A firm that lowers your elasticity has, in the most literal accounting sense, bought something.

Which raises the hinge of the entire field: when a company spends money making you unwilling to leave, is that a cost of production, or is it the product?

Sit with the shape of that. The difference between an economy owned by monopolists and an economy that has barely changed is a decision about which column a number goes in.

Marginal cost is not a thing you can walk up to and touch. It is an inference from an accounting category, and no accounting category is theory-free. The most consequential fight in industrial organisation is a classification dispute.

Now push the boundary outward and watch it break.

Economists have measured the air-pollution damage of coal-fired electricity generation and found it runs between eight-tenths and five-and-a-half times the industry’s own value added. Not a rounding error against the industry. A multiple of it.

This is the externality, marginal social cost exceeding marginal private cost, but the diagram in the textbook does it no justice, because the diagram makes it look like an error. It is not an error. It is the ledger doing exactly what a ledger does: recording transactions between parties. The atmosphere did not sign anything.

We know how to fix this in principle. Put a price on the damage and let it enter the arithmetic. The United States did the work: in November 2023 its environmental agency published a social cost of carbon of about a hundred and ninety dollars a tonne for 2020 emissions at a two percent discount rate. A real number, with real uncertainty, produced by real scientists.

I want to be scrupulous here, because the researchers were: this is an instruction not to monetise. It is not an instruction to write zero. But notice what uncertainty has been permitted to do. It did not widen the estimate. It deleted it.

And in an arithmetic, a number you decline to write behaves in every operation exactly like a zero, silently, and with a modest expression.

That phrase, at a two percent discount rate, is the quietest moral statement in economics, and I think it is the most beautiful.

A discount rate is a claim about the weight of the future. Lower it and your great-grandchildren move closer; raise it and they recede into rounding. Nothing in the mathematics tells you where to set it. You choose, and then the mathematics obeys, and then the choice comes back to you wearing the costume of a calculation.

Corporations make this choice too, and here the research delivers its strangest news. The textbook chain is clean: central bank cuts the policy rate, the cost of capital falls, the present value of every project rises, marginal projects flip positive, investment blooms.

Central banks pull a lever attached to a rod attached to a hinge that has quietly rusted shut.

The economy has been under-built. Not for want of money. For want of patience, priced.

Then the cycle turns the knife. Research and development is procyclical, firms invent in booms and cut in slumps, and the evidence suggests it is too procyclical, that they cut invention exactly when invention is cheapest. Across four thousand seven hundred public companies and three recessions, seventeen percent did not survive and about nine percent flourished. Of the four postures a firm can take into a downturn, the one that paired selective discipline with continued investment had the best odds of breaking away: thirty-seven percent, against twenty-one for pure retrenchment.

The better bet was known. It was known, and mostly not taken. That is not a fact about spreadsheets. It is a fact about fear.

Fear has a shape inside a company, and it looks like a target.

At one American bank, leadership built what were internally called “fifty-fifty plans” sales goals designed so that only half the regions could meet them. Managers called subordinates several times a day to check the numbers. Between January 2011 and March 2016, roughly five thousand three hundred employees were fired for sales-practice violations. About a million and a half deposit accounts were opened without authorisation, and some five hundred and sixty-five thousand credit cards. The board learned the number of firings from a regulatory settlement.

Nobody in that story woke up wanting to commit fraud. They wanted to hit a number that had been engineered to be unhittable.

Nobody crosses a line. The line is moved, an inch at a time, by people acting reasonably within a system that rewards the inch.

In November 1999, a safety panel reviewing a new painkiller counted seventy-nine patients out of four thousand who had suffered serious heart problems or died, against forty-one among those taking an older drug.

The medicine was withdrawn in September 2004.

Return to section one. The reputational penalty for deceiving investors runs over seven and a half times the legal penalty. The reputational penalty for a wrong to a third party is 0.24 percent, not significant.

The families of three hundred and forty-six people are not a counterparty. That is not a metaphor. It is the finding.

The law knows this problem and has tried to legislate around it, and the attempt is worth studying closely, because it fails in an instructive way.

Delaware invented a corporate form, the public benefit corporation, whose directors must, by statute, balance shareholders’ financial interests against the interests of those materially affected by the company’s conduct and against a stated public benefit. A genuine reform. Then read the next subsection: a director “shall not… have any duty to any person on account of any interest of such person in the public benefit.”

The duty exists. There is no one who may enforce it.

That is not a loophole someone slipped past a legislature at midnight. That is the statute, doing what corporate law consistently does: allocating standing. Standing is the legal name for the accounting boundary. It is the list of people whose losses are permitted to become a case.

Everything else in the doctrine follows the same line. Spend on stakeholders all you like, so long as it plausibly serves shareholders. Adopt a policy that openly declines to maximise value and a Delaware court will strike it down: “The ‘Inc.’ after the company name has to mean at least that.” Sell the company and directors become, in the famous phrase, “auctioneers.”

R. Edward Freeman’s counter-move, fourteen years later, is beautiful in its simplicity: he redefined a stakeholder as “any group or individual who can affect or is affected by the achievement of the organization’s objectives.” He did not argue the boundary. He widened it by definition and dared the field to object.

And when the boundary was widened by proclamation rather than statute, we got to run the experiment. In August 2019, a hundred and eighty-one chief executives signed a statement redefining the purpose of a corporation to serve all stakeholders. Researchers wrote to the signatories and asked a single procedural question: did your board approve this?

Zoom out to the largest scale and the same question is waiting.

A tariff is levied on an importer. That is who signs the cheque. Measured against actual import prices, nearly ninety percent of the burden of the 2025 American tariffs fell on US firms and consumers. The statutory payer and the economic bearer are different people, and the gap between them is a boundary drawn by physics of pass-through, not by legislation.

Even how much tariff exists depends on where you draw the line: one estimate puts the effective rate at 11.8 percent on a statutory basis, another at 7.2 percent measured as duties actually collected over imports, a third at 6.6 percent as a calendar-year average. None is wrong. They are three different questions wearing the same name.

In February 2026 the Supreme Court held, six to three, that the emergency statute under which the largest tranche had been imposed “contains no reference to tariffs or duties.” Over a hundred and sixty-six billion dollars had already been collected from more than three hundred and thirty thousand importers. And on the nineteenth of August 2026, under a provision never before invoked by a president, a fifty percent tariff falls on roughly twenty billion dollars of Canadian dairy, motor vehicles and alcohol.

Now shrink it back down to the size of a person, because this is where I want to leave you.

An accounting boundary is not always an abstraction. Sometimes it is fifteen cents, and it is yours.

Here is the asymmetry that reorganised my thinking about all of this.

There exists a public database of corporate misconduct containing more than seven hundred thousand civil and criminal cases from more than four hundred and fifty agencies, with total penalties exceeding one trillion dollars, running back to the year 2000. One hundred and twenty-seven parent companies have each paid over a billion. One bank appears in two hundred and twenty-five separate cases of a million dollars or more.

We built that. Deliberately, over a decade, with funding and methodology and version control.

What a civilisation measures, it manages; what it declines to measure does not thereby cease to exist. It merely stops appearing in the minutes.

I have given you the machinery of harm. Now the other thing, which is true at the same time and which I find genuinely moving.

Every single one of the two hundred and ten new drugs approved by the American regulator between 2010 and 2016 was associated with published research funded by the National Institutes of Health. More than a hundred billion dollars of it. Over ninety percent was basic science on the biological target, not the pill, but the lock the pill was cut for.

The mRNA vaccines: at least thirty-one point nine billion dollars of American public investment, of which ninety-two percent took the form of advance purchase commitments, a government promising to buy a thing that did not yet exist, so that someone would build it.

And the countervailing forces are not dead. In the 2025 fiscal year, one thousand two hundred and ninety-seven new whistleblower suits were filed under the False Claims Act, a record, against a previous high of nine hundred and eighty. About twenty-seven thousand tips reached the securities regulator. Twenty-seven thousand people decided that what they had seen was not going to remain merely a thing they had seen.

In March 2026, a Los Angeles jury found for the first time that social media applications should be treated as defective products, engineered to exploit developing brains. The award was six million dollars, which against these companies is nothing, and which is also a door coming off its hinges. The plaintiff had started using one of the platforms at six years old.

And this year, global investment in clean energy runs to roughly two point two trillion dollars against about one point two trillion for fossil fuels, nearly two to one, with some eighty gigawatts of new nuclear capacity under construction across fifteen countries. Whatever you think of the pace, capital at that scale is not moved by sentiment. Something is being priced.

Here is why I think this material is beautiful and not merely grim.

Every boundary in it was drawn by a person.

Not by physics. Not by nature. Someone decided whether advertising is a cost of production or the product itself, and that decision determined whether we believe monopoly power tripled or barely moved. Someone chose a two percent discount rate, and that choice determined whether a tonne of carbon costs a hundred and ninety dollars or nothing you are obliged to write down. Someone set a scope threshold at a thousand employees and four hundred and fifty million euros of turnover, and thereby decided what an entire continent is required to see. Someone drafted a subsection removing standing from every person whose interest the statute had just recognised. Someone typed fifteen percent into a treaty.

None of these people thought of themselves as moral philosophers. They thought of themselves as doing technical work. That is exactly the point, and it is the reason I would tell anyone learning this trade that they have wandered into something far larger than they were promised.

Accounting is not bookkeeping. It is the technology by which a society decides what it is willing to notice. Recognition (putting a number in a column, giving a thing a name and a place in a statement) is the smallest possible act of moral inclusion. Derecognition is its opposite, and it is quieter, and it is performed constantly, by decent people, in good faith, on ordinary afternoons.

The market did not hear the harm because we did not put the harm where the market listens.

That is the whole finding. It is also, if you turn it over, the whole hope, because a boundary drawn by a person can be drawn again by a person. Not easily. Not without a fight over which column, and whose losses, and who has standing to complain. But drawn again.

Nothing in that number was inevitable. Not the boundary, not the silence, not the 0.24 percent.

Someone drew that line, in a room, on an unremarkable day, with a pen.

That is the bad news. It is also the only good news there is.

Every figure and quotation above comes from the audited dossier compiled 6 August 2026. No new research; no invented facts, names, quotations or events. Illustrative framing is mine; the specifics are not.

I

Karpoff, Lee & Martin (JFQA 2008) on misrepresentation penalties; Armour, Mayer & Polo on second- vs third-party wrongs; Karpoff, Lott & Wehrly (JLE 2005) on environmental violations.

II

Lerner index (standard theory); worked example my arithmetic. De Loecker, Eeckhout & Unger (QJE 2020) vs Traina (2018), Basu (JEP 2019), Benkard, Miller & Yurukoglu (NBER WP 34513).

III

Muller, Mendelsohn & Nordhaus (AER 2011); Trucost/TEEB (2013, 2009 data, a consultancy study, flagged); EPA social cost of greenhouse gases (Nov 2023); OMB M-25-27 (2025).

IV

Sharpe & Suarez (Federal Reserve / Management Science); Gormsen & Huber (AER 2025, via NBER Reporter); Barlevy (AER 2007); Gulati, Nohria & Wohlgezogen (HBR 2010), note the 9%-flourished and 37%/21% figures are two separate statistics, not one.

V

Wells Fargo independent directors’ report (2017), CFPB (2016), DOJ (2020); Graham, Harvey & Rajgopal (2005); Vaughan (1996, quoted 2016); Dyck, Morse & Zingales (RAS 2024).

VI

Vioxx timeline via NPR, Lancet estimates as reported there; Boeing via DOJ (2021), DOT OIG, and NPR (Nov 2025).

VII

DGCL §§365(a)–(b); eBay v. Newmark (Del. Ch. 2010); Revlon (Del. 1986); Friedman (NYT Magazine, 13 Sept 1970); Freeman (1984, p. 46); Bebchuk & Tallarita on the Business Roundtable statement.

VIII

New York Fed tariff incidence (Feb 2026); Yale Budget Lab / Penn Wharton / Tax Foundation effective-rate estimates; Learning Resources v. Trump (20 Feb 2026); Section 338 proclamation effective 19 Aug 2026; Canada–US tax convention Articles X(2) and XXI(2), Third Protocol 1995.

IX

Good Jobs First Violation Tracker; Tylenol via PBS; CVS press release (5 Feb 2014); Volvo Group heritage page (the company’s own account of its own conduct).

X

Cleary et al. (PNAS 2018, via two agreeing secondary sources, primary page inaccessible); BMJ (2023) on mRNA public investment; Jones & Summers (NBER 2020); Bloom, Schankerman & Van Reenen (Econometrica 2013); Scannell et al. (Nature Reviews Drug Discovery 2012); DOJ False Claims Act FY2025; SEC whistleblower FY2025; Los Angeles jury verdict (25 Mar 2026, on appeal); IEA World Energy Investment 2026.