Nobody Asks Where Your Banknote Has Been
Cash as an institution of bounded bearer finality: how a lawsuit over a stolen banknote in 1758 built the most successful settlement technology in history, and why the law that made it also limits it
Keywords: money; bearer finality; institutions; transaction costs; property rights; good-faith purchase; payments economics; digital currency
Abstract. A shopkeeper takes a banknote from a stranger without asking where it came from; a large sum offered in physical cash to settle a property purchase is treated very differently. This essay explains the contrast by characterizing cash as an institution of bounded bearer finality. It separates four things usually run together under the word finality — clean title, discharge of the payer’s obligation, operational non-reversal, and the durability of the holder’s right against later recovery — and shows that title finality is categorical while institutional durability is probabilistic. Bearer cash economizes on transaction costs by conferring clean title on a good-faith taker for value, so ordinary low-value payments circulate without provenance inquiry; the legal mechanism is the common-law doctrine of currency. High value does not by itself destroy title — it bears on good faith and notice, and it triggers a regulatory and forfeiture overlay that makes the holder’s right fragile. The bound is set not by a decay of title but by the rule’s own conditions and that overlay. The argument yields observable, falsifiable implications, including for stablecoins, central bank digital currencies, and every claim that code can make a payment final.
1. Two questions nobody asks
A stranger hands a shopkeeper a twenty-pound note. The shopkeeper does not ask where the note has been, who held it last, or whether it was honestly come by. She could not answer those questions if she tried, and it does not matter: the payment is final on delivery. The buyer cannot reverse it, and the seller need not look behind it. If it later turns out the note was stolen from someone last week, the shopkeeper keeps it, and the victim’s remedy — if there is one — lies against the thief.
Now change one quantity. The same stranger offers to settle the purchase of a house in physical cash. Suddenly everything is different. The conveyancer will not touch it. The bank, if the money ever reaches one, files a report. Above some margin of value, opacity, and abnormality, the question the shopkeeper never asked — whose money is this, and how was it obtained? — becomes one that the institution requires be put, and the security the banknote enjoyed in the shop is qualified.
This contrast is worth an essay because almost everyone gets the explanation wrong, and the wrong explanation is doing real damage in current debates about digital money. The wrong explanation is that “finality” is a dial that turns down smoothly as the sum goes up — that a £20 note is very final, a £20,000 briefcase somewhat final, and £20 million hardly final at all, as though legal title to money evaporated with magnitude. That picture is false, and its falsity matters. What actually happens is stranger, older, and more precise: cash is an institution, in the strict sense that economists and philosophers give that word, and the institution has a very particular architecture. One part of it is categorical — it either applies or it does not, and where it applies it applies fully, at any amount. Another part is probabilistic — it weakens with circumstances, enforcement, and power. Confusing the two parts is the source of most of what is said loosely about cash, privacy, cryptocurrency, and “irreversible” payments.
The claim I want to defend runs in one sentence: cash is an institution of bounded bearer finality — a legal arrangement that manufactures a zone in which strangers settle with each other without producing any information about where the money came from, and that withdraws its protection, by specific and identifiable mechanisms, where the social cost of that ignorance becomes intolerable. Everything else in this essay is the unpacking of that sentence: what kind of thing the institution is, what economic problem it solves, what legal machinery implements it, where exactly the bound sits, and what all of this implies for the token systems and central bank digital currencies now being legislated into existence.
2. Not an object, not a technology
Start with the category error. Cash is usually treated as one of two things: a physical object (paper, polymer, metal) or a payment technology (a way of moving value that happens to be offline and anonymous). Both descriptions miss what does the work. A banknote is an object that carries a rule, and coin, paper, and — in principle — electronic bearer instruments are implementations of one and the same settlement arrangement. The arrangement, not the medium, is the thing.
What kind of thing is an arrangement like that? Institutional economics has spent decades sharpening the answer. Geoffrey Hodgson defines institutions as systems of established and embedded social rules that structure interaction, and gives money as a standard instance. John Searle goes deeper: institutions rest on constitutive rules of the form “X counts as Y in context C” — rules that do not merely regulate a pre-existing activity but create the very statuses the activity trades in. A piece of paper counts as money; crossing a line counts as scoring; uttering certain words counts as marrying. Money is Searle’s leading example, and for good reason. Frank Hindriks and Francesco Guala have shown how the rule-based and equilibrium-based views of institutions unify: an institution is a rule-in-equilibrium, at once a prescription and a self-enforcing pattern that nobody has a private incentive to deviate from. And John Commons, back in 1931, supplied the unit of analysis — the transaction, governed by working rules.
Cash fits this apparatus exactly, and not loosely. The legal doctrine I will describe in section 5 is a constitutive rule in Searle’s precise sense: delivery of a token, for value, to a good-faith taker counts as the creation of a fresh, clean title to the money. It is sustained as an equilibrium in which no party investigates provenance because none expects others to. And it is realized in the most elemental transaction there is: payment.
One more placement is worth making because it does real analytical work later. Claude Ménard and Gaetano Martino have recently argued that institutional analysis needs a meso level — between the macro level of constitutional and monetary order and the micro level of the individual deal sits a layer of transmitting devices that carry the general rules into particular transactions. Cash is exactly such a meso-institution. The macro layer defines legal tender, the unit of account, and the law of property in money; the micro layer is your purchase at the till; and the doctrine that makes the banknote pass clean between you is the transmitting device in between. Keep this three-layer picture in mind, because the essay’s central distinction — categorical title, probabilistic durability — is a statement about which layer fixes what. Title is fixed at the meso layer, by the currency rule. Durability is fixed where the meso-level overlay of anti-money-laundering and forfeiture machinery meets the macro level of enforcement capacity and sovereign power. Without the layered picture you cannot even say where each property lives.
The transaction-cost tradition supplies the institution’s function. Ronald Coase showed, in the two papers that founded the field, that institutions and the assignment of entitlements matter because transacting is costly — if bargaining were free, the firm in 1937 and the liability rule in 1960 would both be matters of indifference. Oliver Williamson developed the program into a theory of governance structures matched to transaction attributes and selected to economize on transaction costs; in his four-level scheme, the rules of property and contract sit above day-to-day exchange and condition it. Cash is functionally of exactly this kind. Verifying the title and history of money in every transaction would be prohibitive — imagine the corner shop conducting provenance due diligence on each note in the till — and the institution economizes on that cost by the most radical means available: making the verification legally unnecessary in the ordinary case. Not cheaper. Unnecessary.
Finally, cash is a settlement arrangement, and payments economics is its proper home. Charles Kahn and William Roberds, in their canonical introduction to the field, present payments as a distinct economic problem — the discharge of obligations and the transfer of value — whose essential function can be implemented through alternative arrangements, with settlement and its finality the central objects of study. Bearer cash is one settlement arrangement among others: account transfer, intermediated credit, netting. What distinguishes the arrangements from one another is how and on what terms each achieves finality, and comparing them is a comparison of transaction costs and of the kind of finality each delivers. That framing is exactly what most commentary lacks, and it is what the four-way distinction in section 6 will supply.
3. Money as memory, money as amnesia
If cash is a device for economizing on transaction costs, the obvious question is: which costs? Here the deep literature of monetary economics divides, and the division is the hinge of the whole argument.
One tradition runs through Armen Alchian’s 1977 paper “Why Money?” Goods are costly to identify — their quality, their condition, their authenticity — and institutions evolve to reduce the cost of that ignorance. Money is one such institution: the good that becomes money is the one most cheaply and uniformly identifiable, so that routing all exchange through it minimizes the verification cost that would otherwise be borne in every barter. Two features of Alchian’s account are decisive for us. First, he is explicit that money’s role does not rest on any bookkeeping or debt-recording function; recording could be done with anything, and it is precisely because goods are not costlessly identifiable that money is useful. Second, his problem is the verification of goods: he simply assumes the money itself is costlessly identifiable as money, and he is silent on its provenance. Karl Brunner and Allan Meltzer generalized this into a theory of money as an economizer on the information costs of exchange, and Nobuhiro Kiyotaki and Randall Wright later supplied the decentralized foundation, showing how a medium of exchange emerges among randomly matched traders.
Against this stands what is now the dominant foundation of formal monetary theory. Narayana Kocherlakota proved in 1998 that, in a broad class of environments, any allocation achievable with money is achievable with memory — a complete record of past actions — and concluded that money is a primitive, imperfect form of memory. The new monetarist program of Ricardo Lagos, Guillaume Rocheteau, and Randall Wright builds on this result. “Money is memory” has become one of the most quoted slogans in the field.
Both traditions are right about something, and the bearer case is where you can see exactly what. Kocherlakota’s theorem is an equivalence about feasible allocations; it does not say that a token carries a record. And notice what dimension of history money actually preserves: how much net value an agent has contributed to others, encoded in how much money they now hold. Bearer cash preserves exactly that quantity signal — your banknotes are the residue of what you have given up. What it erases is every other dimension: whose value this is, where it has been, and whether it was rightfully acquired. Bearer cash is, in a phrase, memory of quantity and amnesia of provenance.
Here is the crucial step, and it is the step that turns a nice slogan into a theorem-shaped claim: the amnesia is not incidental. It is constitutive. A bearer instrument that recorded whose value it was and how each holder came by it would, by that record, forfeit the three properties that let it settle at all. It would lose fungibility, since each unit would be individuated by a different history and no longer interchangeable with any other. It would re-import the very verification cost the institution exists to spare, since a recipient could be expected to read the record and would take subject to what it showed. And it would make clean title on delivery impossible, since title would again depend on the soundness of the entire prior chain rather than vesting afresh in the good-faith taker. A provenance-bearing banknote is not a better banknote; it is not a banknote.
Gary Gorton’s information economics states the same impossibility in another register, and it is worth having both formulations because they will both be needed when we come to digital money. Gorton and George Pennacchi showed that debt circulates as a payment medium because it is information-insensitive — designed so that no holder has anything to gain by investigating it, which is what lets strangers accept it at face value. Gorton and Guillermo Ordoñez showed that such private money works only while no agent has a reason to produce costly information about it; a financial crisis is precisely the switch from information-insensitivity to information-sensitivity, the moment everyone suddenly starts asking questions about paper they had been accepting blind. Now transpose: for collateralized debt the information at issue is the quality of the backing; for bearer cash, which has no backing, the information at issue is provenance and the holder’s entitlement. A provenance record is an information-production apparatus bolted to the instrument. To attach it to a bearer token is to make provenance information-sensitive by construction — to build the crisis into the design — destroying the no-questions-asked circulation that is the settlement property.
The Alchian–Kocherlakota dispute therefore resolves not by declaring a winner but by a division of mechanism. Memory is the true foundation of account money, which simply is a record — your bank balance is nothing but an entry in a ledger of who did what. Bearer cash discharges the monetary function by the opposite route: by forgetting provenance, where the forgetting is what makes settlement possible. Glen Whitman has recently made a general point about social technologies that fits this case exactly: number systems, he shows, are functionally heterogeneous — decimal persists for general calculation, binary for computation — and their coexistence is explained by differential fitness for different purposes, not by one being an imperfect approximation of the other. Monetary forms exhibit the same heterogeneity. Bearer cash and account money are not competing approximations to one ideal money; they are functionally distinct technologies — the record, fitted to relationships in which history is to be preserved, and the anti-record, fitted to arm’s-length settlement in which provenance is to be dispensed with. That they coexist rather than one displacing the other is exactly what the heterogeneity view predicts, and it is why a century of predictions that one form would extinguish the other keeps being wrong.
One asymmetry must be flagged now because the entire structure of the bound grows out of it. There are two verifications a recipient of cash could in principle perform, and the institution treats them in opposite ways. The first is that the token is genuine — not counterfeit. This is the verification Alchian assumed away, and the institution never waives it: a counterfeit note passes no value and no title, and the recipient who accepts one bears the loss, at any amount. The second is verification of provenance — that the genuine token was honestly come by and the transferor entitled to pass it. This, and only this, is what the institution economizes on. Genuineness is never waived; the economy on provenance cannot be unconditional, because the social cost of not asking rises with the stakes. Declining to investigate the provenance of a £20 note is cheap. Declining to investigate £10 million is not. Hold that asymmetry; it is the origin of everything in sections 7 and 8.
4. The register that would destroy its subject
There is a beautiful way to see how strange money’s legal position is, and it comes from a 2025 article in the Oxford Journal of Legal Studies by Michael Crawford, “The Riddle of the Good Faith Purchaser.” Crawford is writing about stolen goods, not money, and that is exactly why his analysis illuminates the monetary case: money is the limiting case his framework points at but does not reach.
Crawford’s setup is the oldest triangle in private law. An owner identifies an asset stolen from her in the possession of a good-faith purchaser who bought it, innocently, from the thief or down the chain. The purchaser refuses to give it back. Who should win? The instinctive answer — the owner, because she is the owner — begs the question, since “owner” just labels whoever has the best legal claim, which is the thing in dispute. And the honest answer of comparative law is that legal systems disagree spectacularly: even English law, which proclaims the general principle nemo dat quod non habet — no one gives what he does not have — admits a thicket of exceptions to it.
Crawford’s economic move is to ask what the loss in these cases actually is. The theft itself, he argues, is in the first instance a transfer, not a loss: the owner’s deprivation is matched by the thief’s gain, and if the purchaser happens to value the asset more than the owner did, the forced transaction has — on paper — increased total welfare. What makes theft unambiguously destructive is not the transfer but the enterprise: the resources the thief burns committing it, the resources owners burn guarding against it, the resources purchasers burn verifying titles, and the deadweight losses from distorted investment. Theft can be profitable to the thief; it is always socially wasteful. So the question the law of good faith purchase should answer is not which party “deserves” the asset but which liability rule best suppresses theft by minimizing the thief’s returns from selling stolen goods.
And here is his sobering conclusion: for most goods, no rule does. When the asset is of low value or hard to identify, the owner will rationally not search for it, the purchaser will rationally not investigate title, and the parties behave as if the law favors the purchaser whatever the statute book says. The allocation of liability has no behavioral effect at all. The exception — and Crawford builds his positive program on it — is the class of goods that are valuable, distinctive, and ruined by disguise: the Renoir that becomes worthless canvas if mangled, the classic car that becomes scrap if chopped. Such goods invite the owner’s search, and, crucially, they are amenable to registration. Registers, Crawford argues, are the most effective non-punitive tool the law has against theft: condition the owner’s victory over the good-faith purchaser on prior registration of the goods, and you give every future buyer a cheap way to check, every owner a cheap way to protect, and every thief a market that has been deliberately poisoned against him. Where no register is possible, Crawford is candid that there is little to choose between the parties, and he reaches for auction-theoretic devices to resolve what is otherwise a coin-flip.
Now hold Crawford’s framework up against money, and watch what happens. Money is valuable. Money is, in the aggregate, exactly the asset criminals most want. On Crawford’s logic you might expect the law to respond with the strongest possible register. Instead the law does the precise opposite, and has for centuries: it strips money of registrability altogether, refuses to let owners follow it into honest hands, and makes the good-faith purchaser’s victory not an unfortunate default but the constitutive rule. Why?
Because money is the one asset for which Crawford’s remedy would kill the patient. A register works by individuating: this Renoir, this chassis number, this parcel of land. Individuation is exactly what money cannot survive. Registered money is money whose every unit carries a history, which — as section 3 showed — means money that is no longer fungible, no longer cheap to accept, no longer capable of vesting clean title on delivery. The register that protects the Ferrari would, applied to the banknote, reintroduce into every corner-shop transaction the verification cost the whole institution exists to eliminate. Crawford’s analysis and the law of money are therefore not in tension; they are two halves of one theory of information. Where individuation is cheap and circulation is slow, the law builds registers and lets owners win. Where circulation is the entire point and individuation would destroy it, the law abolishes the owner’s pursuit and lets the taker win. Cash sits at the far pole of that spectrum — the asset the law has deliberately made unregisterable — and the doctrine that put it there has a name and a date.
5. 1758: the exception that makes the rule
Ordinary personal property passes, in English law, under nemo dat quod non habet: a transferee takes no better title than the transferor had, so a thief’s buyer acquires nothing and the true owner prevails. Money is the great exception, and it has been settled law since Miller v Race in 1758. A person who takes a banknote in good faith and for value acquires a good title to it even though it was stolen from a prior holder — because money passes as currency and, having “no earmark,” cannot be followed at law into the hands of such a taker. The court did not say the victim suffered no wrong; it said the wrong could not be visited on the innocent taker without destroying the note’s capacity to circulate as money at all. The exception was later generalized from notes to money as such, and the legal historian David Fox has given the doctrine its modern name: the currency of money — the attribute by which title to money is acquired afresh on transfer rather than derived from the transferor.
The modern law rationalizes the same structure through unjust enrichment, and the leading case repays attention because it shows exactly where the protection stops. In Lipkin Gorman v Karpnale (1991), a solicitor with a gambling problem stole client money from his firm’s account and lost it at a London casino. The firm sued the casino. The House of Lords confirmed two things at once. Stolen money can, in principle, be followed. But a bona fide purchaser who gives value takes free of the original owner’s claim — and the casino lost precisely because, as a matter of law, it had not given value: it had paid out on gaming contracts that were void under the legislation of the day, making it a volunteer, a recipient who gave nothing recognized in exchange. The casino had to make restitution, subject to a defense of change of position. Notice what did the work: not the size of the sums, which were large, but the absence of value. The doctrines interlock around a single pivot — the good-faith taker who gives value — and it is the presence or absence of that figure, not the magnitude of the payment, that determines whether the prior owner’s claim survives.
The conditions of the rule are worth stating exactly, because each is a boundary of the institution. Value is required: a donee — someone who gives nothing — takes subject to the prior owner’s claim, so the recipient of a gift of stolen money does not keep it. Good faith and absence of notice are required: a recipient who knows the money is tainted, or whose state of knowledge makes it unconscionable to retain it (the test the courts apply in the cognate doctrine of knowing receipt, from BCCI v Akindele), does not take clean. And tracing law reinforces rather than contradicts the structure: value can be followed through substitutions and into mixtures (Foskett v McKeown), and even through a bank account (Banque Belge pour l’Étranger v Hambrouck, 1921) — but every route of pursuit stops dead at the same figure, the good-faith purchaser for value without notice.
Two distinctions must be kept straight, because popular writing about money muddles both. First, currency is not legal tender. Legal tender concerns which forms of money a creditor must accept in discharge of a debt — in the United Kingdom a creature of statute (the Currency and Bank Notes Act 1954; the Coinage Act 1971) — and it is about the discharge of obligations, not the passing of title. Currency, the title-passing attribute, is what carries bearer finality, and the two merely travel together in ordinary cash. Second, currency is a constitutive rule, not a regulative one: it does not direct or restrict transfers that would happen anyway; it creates the clean title that the good-faith taker for value acquires. Delivery, for value, in good faith, counts as the making of a fresh title. That is Searle’s formula, implemented in the oldest commercial law we have.
One further contrast sharpens what is constructed about this. Institutional economists often find at the root of property a spontaneous possession convention — possessors defend, intruders defer — and experimental work by Marco Fabbri, Matteo Rizzolli, and Antonello Maruotti has shown that coordination on this “bourgeois” convention is strongest when possession is meritorious: people defer to possession they regard as honestly earned. Currency does not simply ride this convention. Toward the good-faith taker for value — himself a meritorious acquirer — the rule is consonant with it. What currency severs is the prior owner’s merit-based claim, which the spontaneous convention would protect and the doctrine deliberately extinguishes. The convention, left to itself, would let the victim pursue her stolen notes into every honest hand; the institution overrides that pursuit for the functional reason that money must circulate without inquiry, and a rule tracking the prior owner’s merit would defeat that. Bearer finality is not an evolved custom the law merely ratified. It is a constructed override of the custom, made for a reason.
6. One word, four finalities
Now the conceptual core. The word “finality” is doing at least four different jobs in ordinary and even expert usage, and the entire confusion about cash — and about crypto, and about CBDCs — comes from running them together.
Title finality is the acquisition by the transferee of a clean title, good against prior proprietary claimants. Obligation finality is the discharge of the payer’s underlying debt: after payment, the seller cannot come back for the price. Operational finality is non-reversal of the transfer within the system that effected it — for physical cash, the sheer irrevocability of handing the note over; for designated electronic systems, the statutory protection of completed transfers from being unwound (in the UK, the Financial Markets and Insolvency (Settlement Finality) Regulations 1999). Institutional finality — call it recoverability — is whether a court or the state will later restore the value to another claimant: through tracing, knowing receipt, or statutory forfeiture.
These come apart in every direction, and the combinations are not exotic — they are everyday legal reality. A transfer can be operationally irreversible yet institutionally recoverable: the ledger entry stands, but the court orders the recipient personally to repay. A payment can achieve obligation finality yet leave the recipient exposed to confiscation. A taker can hold clean title and still have committed a criminal reporting offense in accepting the money. Once you have the four categories, sentences like “cash is final” or “blockchain payments are final” stop being propositions and become questions: final in which of the four senses? Every serious claim in the rest of this essay is an answer to that question for a specific case.
The separation also travels. Wherever anyone calls a transfer “final” — in securities settlement, in correspondent banking, in a token protocol — the same four senses can be pried apart, so the payoff is not parochial to banknotes. But cash is where the anatomy is clearest, because cash is the instrument for which the four senses were institutionally engineered to coincide in the ordinary case: hand over the note, and title passes, the debt dies, the transfer cannot be undone, and no one is coming back for it. The engineering is the achievement. The next section is about where the engineering stops.
7. A pivot, not a slope
Here is the central correction this essay exists to make. Title finality is categorical. Where the conditions of the currency rule are met — value given, good faith, no notice — the taker acquires clean title, and this does not weaken smoothly as the sum rises. There is no doctrine anywhere in the law of money under which £9,999 passes title and £10,001 does not. Institutional finality is probabilistic. The durability of the holder’s right against later challenge depends on notice, abnormality, the regulatory overlay, enforcement capacity, and sovereign power — and it is this, not title, that erodes as the stakes and the strangeness rise.
So what does high value actually do, if it does not destroy title? Its work is real but indirect, and it runs in three precisely identifiable channels.
First, evidentially. A large, unusual, all-cash transfer is a circumstance from which a court may infer that the recipient knew, or was on notice, that something was wrong — and notice defeats good faith, which removes the doctrinal protection. But look at the mechanism: value operates through the notice condition, as evidence bearing on the recipient’s state of mind, not as a title-destroying fact in itself. A £2 million cash payment with an innocent explanation and a genuinely ignorant recipient passes title as cleanly as the £20 note in the shop.
Second, regulatorily. Anti-money-laundering and customer-due-diligence law — in the UK, the Money Laundering Regulations 2017 — imposes duties of inquiry above thresholds and in suspicious circumstances. These duties bind the recipient independently of title. A regulated dealer who accepts a large, unexplained cash payment may hold perfectly good title to the money and simultaneously have committed a reporting offense. Title law and regulatory law are answering different questions about the same event.
Third, by forfeiture. The proceeds-of-crime regime — in the UK, the Proceeds of Crime Act 2002 — permits the state to recover and confiscate criminal property, including through subsequent dealings, by a route that runs alongside private title rather than by denying it. The money-laundering offenses in sections 327 to 329 criminalize dealing with criminal property by reference to a person’s knowledge or suspicion, whether or not their civil title is good. Civil recovery under Part 5 acts against the property itself and can reach later holders. Confiscation after conviction acts against the offender’s assets. None of these is a rule about who holds clean title; each is part of institutional finality — whether the state will later extract the value from the holder.
And now the detail that clinches the whole argument. The forfeiture regime contains its own good-faith-purchaser bound: under section 308 of the 2002 Act, property ceases to be recoverable once it is obtained by a person who acquires it in good faith, for value, and without notice. Read that against Miller v Race. Parliament, building a twenty-first-century confiscation machine, reproduced — condition for condition — the protection the common law settled in 1758. The statutory overlay does not contradict the currency doctrine; it withdraws institutional finality along exactly the same fault lines — want of value, notice, complicity — on which title finality already turned. The same figure is protected at both layers, and it is the failure of the conditions, never the magnitude of the sum as such, that exposes the holder.
One more distinction, because it is the one professionals most often blur. The suspicion that triggers a reporting duty under anti-money-laundering law is not the notice that defeats a taker’s title. Suspicion for reporting is a deliberately low threshold — a possibility of criminal provenance, assessed against regulatory typologies — and it attaches to a regulated person’s conduct, generating a duty whose breach is an offense regardless of title. Notice that defeats title is a higher and different thing: actual knowledge, constructive notice, or a state of knowledge making retention unconscionable. A retail cashier faintly uneasy about a customer owes no report and certainly takes clean title. A regulated dealer in a large, opaque cash deal may owe a report and still take clean title, unless the circumstances rise to notice or bad faith. High value feeds both channels — as evidence bearing on notice, and as a fact engaging regulatory duty — without ever being, in itself, a condition of title. That is the precise sense in which bearer finality is bounded: not by a slope, but by a pivot — the conditions of the rule — surrounded by an overlay that makes the position of anyone near the pivot’s edge fragile.
8. The threshold moves, and it moves for a reason
Why does the institution bracket the provenance question below a threshold rather than pricing it? Why doesn’t the shopkeeper charge a small provenance-risk premium on every note, the way an insurer prices theft?
The primary reason is that she cannot. For the individual recipient, the provenance of a particular note from a particular stranger is not reducible to a transaction-level probability — there is no actuarial class over which the corner shop could price the chance of taint on this £20. Frank Knight’s old distinction is the right lens here, in the reading Chris Clarke has developed for financial governance: risk is where the distribution of outcomes is known; uncertainty is where no distribution can be formed. Aggregate actors — banks, insurers, regulators — do estimate laundering and counterfeit rates, so provenance is better described as costly to verify at the level of the single transaction than as strictly Knightian. But the institutional consequence is identical on either description: the individual recipient cannot economize by pricing, so she economizes by not asking, and the institution makes not-asking safe by guaranteeing her title if she takes honestly and for value. Above the threshold, the institution does not suddenly acquire the missing probability distribution; it manages the residual uncertainty provisionally, through the avowedly imperfect machinery of anti-money-laundering supervision.
This gives the bound an analytical location rather than leaving it a brute fact of regulation. The threshold at which provenance becomes information-sensitive is the value at which the expected social gain from inquiry — the crime it deters and the property it recovers — first exceeds the cost that inquiry imposes on ordinary exchange. That is a comparative-static claim with content: the threshold should fall as the social cost of opacity rises and as the cost of inquiry falls, and it should rise where inquiry is expensive and opacity comparatively harmless. The anti-money-laundering threshold is the institution setting this margin administratively. Its historical drift downward — reporting and due-diligence thresholds that once caught only exceptional transactions now catch merely large ones — is the predicted institutional response to a rising social cost of anonymous high-value transfer and a falling cost of inquiry, not a free-standing regulatory fashion. (This is an institutional interpretation of the pattern, not a claim that regulators consciously solve the social-cost calculation.)
The bound is therefore dynamic. The doctrine emerged from the law merchant and the circulation of coin, was absorbed into the common law in the eighteenth century, and was extended by analogy to negotiable instruments — bills, notes, checks — through mercantile usage recognized in Goodwin v Robarts (1875) and codified in the Bills of Exchange Act 1882. Note the direction of travel: negotiability did not found the monetary exception; it followed it, extending to certain instruments a protection modeled on the one money already enjoyed. And the threshold has moved ever since, each adjustment resetting the line as the social cost of bracketing provenance changes. An institution tracking a moving balance between the economy of verification and its abuse is exactly what the theory predicts an institution of bounded finality should look like from the outside.
9. Incomplete rules, imperfect rights, and the sovereign
Two features of the bound need a further layer of theory, because they determine whether “bounded finality” is a mechanism or just a relabeling of known law.
The first is how the bound gets set in the cases the rule never anticipated. The conditions — value, good faith, notice, the ordinary course — cannot specify in advance every situation to which they apply. What counts as value in a novel payment structure? What counts as notice when the red flags are algorithmic? Katharina Hingl and Milo Shera call this institutional incompleteness: rules expressed in words cannot coordinate expectations for every contingency, so unanticipated situations are settled ex post, with the rule’s purpose supplying the focal point. The boundary of bearer finality is incomplete in exactly this way, filled in case by case — as new instruments and new laundering schemes appear — by reference to the institution’s purpose: economize on verification, without surrendering the prior owner where the social cost of doing so becomes intolerable.
The second is whether the bound is real, which is an empirical question, not a doctrinal one. Douglas Allen has argued that economic property rights — a person’s ability, in expected terms, to actually exercise choices over a thing — are primal, and that legal rights, norms, and customs merely constrain them; crucially, economic rights are imperfect, carrying a probability that expresses the strength of the right. Possession of a banknote is an economic property right in exactly Allen’s sense, and the currency doctrine is the institution that converts it into fresh legal title for the good-faith taker. But Allen’s probability has to be set by something, and the candidate is enforcement capacity. Gustavo Magalhães de Oliveira and Bruno Varella Miranda’s study of the Brazilian Amazon bears on this by analogy: where rights over land and resources are imperfectly defined, strengthening the state’s monitoring and enforcement reduces violent appropriation, and where enforcement is weak the nominal right does little. Their data do not test bearer-cash finality; what they support is the general institutional proposition that formal rights become durable only where enforcement capacity makes them so. Applied here: the holder’s title and the prior owner’s recovery are both imperfect rights whose strength is a probability set, in part, by the capacity to trace, freeze, and recover.
The outer limit of the governance dimension is the case in which the adverse claimant is the sovereign itself. Mehrdad Vahabi’s study of Anfal in the Iranian constitutional order — the doctrine under which ownerless property belongs exclusively to the supreme jurisconsult — gives the pure instance of sovereign appropriation. Where the sovereign asserts an overriding entitlement, bearer finality is at its weakest, contested not by a prior private owner but by the state itself. Confiscation, demonetization, and exchange control are the familiar, milder instances of the same pole. Anyone who lived through a demonetization knows what it feels like when the macro layer reaches down and switches the meso-institution off: the note in your hand is unchanged as an object, and everything it counted as is gone.
10. What cash cannot do
Two recent contributions to the institutional literature locate the institution from the outside — by characterizing what cash is not — and both, read carefully, confirm the boundary rather than blurring it.
Ginny Seung Choi and Virgil Storr argue that the market is a discovery process not only for prices but for trust: traders learn, through repeated dealing, whom to trust and whom not to. Bearer cash makes that discovery unnecessary for the payment leg of an exchange — because settlement is final on delivery, the seller need not establish the buyer’s creditworthiness, history, or identity. But the right way to put this is that cash relocates trust rather than abolishing it. By the genuineness-provenance asymmetry of section 3, the recipient still bears counterfeit risk: trust shifts from the counterparty to the instrument and its issuer. And trust-discovery reappears, on schedule, exactly where the theory says the institution ends — above the threshold, where know-your-customer, knowing receipt, and forfeiture re-impose inquiry. The threshold is the boundary of the institution, visible from the trust side.
Anthony Gill and Michael Thomas open from the other direction with a lovely puzzle: if cash is so much more fungible than any gift, why do people give gifts at all? Their answer is that gifting is dynamically efficient because it is imperfect: the burnt-sacrifice character of a non-returnable, imperfectly matched gift signals trustworthiness and builds the relational networks markets need — something impersonal cash cannot do precisely because it carries no record of who gave it. The mirror-image is exact, and it has a sharp legal edge. Title finality is triggered by consideration, and the gift is the paradigm transfer without consideration. So the gift is precisely where title finality fails: the donee of stolen money takes no title, however pure her heart. A transfer can be socially efficient as a trust-signal and legally non-final as a conveyance of tainted money. Finality tracks the economy of verification, not social value — which is exactly what a verification-cost theory of the institution predicts, and what a “cash is just whatever people accept” theory cannot explain.
11. Code is not title
Now the payoff for the present. Bearer finality is not a property of paper. The institution can, in principle, be implemented electronically — the whole analysis of sections 2 through 9 was carried by a rule, not by cellulose. But most electronic monetary value is not bearer property, and the extension has to be disciplined, because this is where the four-finalities distinction earns its keep and where a decade of technological rhetoric has gone wrong.
Sort the instruments. Account-based bank money, and most stored-value and e-money products, are claims on an intermediary — reversible by chargeback, clawback, mistaken-payment recovery, or regulatory freezing. That is account money in electronic form, not bearer cash, whatever the app looks like. Stablecoins are bearer-like in their mode of transfer, but their finality depends on whether the legal order and the issuer’s governance recognize the transferee’s title against prior claims; where redemption rights and freezing powers sit with an issuer, institutional finality is the issuer’s to withdraw. Permissionless cryptocurrency tokens transfer by delivery and achieve genuinely strong operational finality — the ledger entry, once deep enough, does not come back. But their title and institutional finality remain matters for the recognizing legal order, not the protocol, as the migration of all serious provenance inquiry to exchanges and custodians demonstrates in practice: the questions the protocol refuses to ask get asked at the on-ramps and off-ramps instead. And a central bank digital currency can be placed anywhere on the spectrum — where it lands is a choice of the legal order, not of the technology.
Recent legislation has, in fact, decided the question — and it decided against the bearer end. The United States, in the GENIUS Act of 2025, gave payment stablecoins a statutory framework as redeemable claims on regulated issuers, with statutory redemption rights and insolvency priority for holders: a claim against an identified issuer, which is the anatomical opposite of a bearer title — and paired it with an executive policy against a retail central bank digital currency (Executive Order 14178). China moved the same direction from the other side of the design space: the People’s Bank of China’s e-CNY framework effective January 1, 2026, moved the instrument away from a cash-like bearer design toward digital deposit money, classifying balances held in commercial-bank wallets as deposit liabilities. The prepared digital euro and digital pound designs, with their holding caps and low-value-privacy tiers, are the bound written in at the drafting stage — an explicit legislative acknowledgment that unbounded electronic bearer finality is not on offer. Whatever one thinks of these choices, they confirm the analytical point: the bearer property — clean title, no provenance inquiry, finality conferred by legal recognition rather than by the ledger — is a specific institutional achievement that token technology does not, by itself, deliver.
The four finalities also discipline the recurring claim that electronic settlement can be made final by technical means. Operational finality is a real and valuable property, and for designated systems it is conferred by law, not by engineering alone. But operational finality is neither title finality nor institutional finality. A transfer can be operationally irreversible and confer no clean title — if the value was stolen and the recipient is not a good-faith purchaser for value. It can be operationally irreversible and institutionally recoverable — if forfeiture or tracing reaches it, with the court acting on the person rather than the ledger: the entry stands, and the holder is ordered to repay or to hold the value on constructive trust. The mistake in the “code is law” thesis, applied to money, is exactly this: reading operational irreversibility as title, or as immunity from recovery. Technical control is not legal title.
Marco Giraudo has given this error a name worth adopting: a legal bubble. A legal bubble inflates when market participants accumulate reliance on the assumption that their technological control of a resource will be ratified by courts as a protected property right — and it bursts when courts revise that ratification ex post to protect hierarchically superior rights. Treating technical control of a token as bearer finality is a legal bubble in precisely this sense: the incompleteness of the institution (section 9) gets settled, eventually, against the technologically controlling holder, and the “revision” that so outrages the code-is-law camp is simply the institution reasserting a bound the technology never displaced. Sinclair Davidson, Primavera De Filippi, and Jason Potts are right that distributed ledgers are institutional technologies — genuine alternatives for coordinating economic activity — and De Filippi and her co-authors are right that these “confidence machines” remain subject to governance in the end. The monetary case is the sharpest instance of both claims at once: because the finalities that matter — title and institutional — are conferred by a recognizing order, a technical system cannot abolish the recoverability that bounded finality entails. It can only relocate the threshold — move the place where provenance becomes information-sensitive and inquiry returns. The inquiry itself is conserved.
What carries across from banknote to token, then, is the lesson and not the rule. Whether a given token is property, and of what kind; whether nemo dat and a bona-fide-purchase defense apply to it; whether the transfer ran through an intermediary; what relief a claimant can seek — all of these are open, jurisdiction-specific questions that the banknote case settled for banknotes long ago and has not settled for tokens. A recipient of electronic value who took as a volunteer, with notice, or in a high-value abnormal transfer stands in an analogous bounded position — but only where the recognizing legal order supplies a clean-title rule comparable to currency. An electronic instrument occupies the bearer position only when law, not code, places it there. So far, the world’s legislatures have conspicuously declined to place any of them there.
12. The case against big notes is a calibration argument
The strongest case against cash runs through the very property this essay has been praising, and the theory should be able to absorb it. Kenneth Rogoff has argued for phasing out high-denomination paper currency on the ground that its anonymity and provenance-insensitivity — the exact features that make low-value settlement efficient — also make it the preferred instrument of tax evasion, corruption, and serious crime, so that the social costs of large notes exceed their benefits. And there is empirical bite behind the concern: Richard Wright, Erdal Tekin, Volkan Topalli and their co-authors found that when a US benefits program displaced cash with traceable electronic transfers, cash-dependent crime fell in the affected areas.
Notice what this argument is and is not. It is not a refutation of bearer finality; it is a statement about where the value-bound should fall. The objection identifies a social cost of bearer finality that rises with value and opacity — precisely the dimensions along which sections 7 and 8 located the bound — and the institutional response it recommends is the one the institution already embodies: grant the economy of verification where its benefits dominate, withdraw it where its costs dominate, through the consideration and notice conditions and the regulatory and forfeiture overlay. Whether the $100 bill or the €500 note should exist at all is the question of how high the bound is set — a calibration of the threshold, not a refutation of its existence. Kahn, McAndrews, and Roberds established that cash is privacy, in a precise economic sense; Rogoff establishes that privacy can shelter wrongdoing. Those are two faces of one institutional fact, and the bound is the line the institution draws between them — a line that has moved before and will move again.
This is also, candidly, where the thesis is most exposed, and therefore most testable. The theory predicts that durable legal orders contract the bound as the opacity and scale of anonymous transfer grow. A society that sustained very broad, high-value bearer finality without rising social cost — or a legal order that responded to rising laundering costs by widening the no-questions-asked zone and prospered — would be evidence against the account. The prediction is about the direction institutions move under identifiable pressure, and directions can be observed.
13. How this could be wrong
A framework that explains everything explains nothing, so it is worth stating plainly what observations would damage this one.
If a legal order destroyed title mechanically with value — a statute under which good-faith purchase simply fails above a stated sum, regardless of notice — the categorical/probabilistic distinction at the heart of the essay would be false for that order. If a legal order protected volunteers equally with purchasers — letting the donee of stolen money keep it on the same terms as a buyer — the consideration condition would not be doing the work claimed for it. If a proceeds-of-crime regime reached the good-faith purchaser for value without notice — if section 308 or its analogues did not exist, and forfeiture cut through the very figure the common law protects — then the claim that one logic runs through both layers, private and public, would be refuted; the alignment would be coincidence, not structure. If anti-money-laundering regimes were indifferent to value and suspicion — imposing identical inquiry on the £5 purchase and the £5 million one — the information-sensitivity account of the threshold would fail. And if electronic systems achieved title finality by engineering alone — if courts in fact treated deep ledger confirmation as extinguishing prior claims without any recognizing rule — the entire argument of section 11 would collapse.
English law, as it stands, does none of these things. Title turns on conditions, not magnitude; the volunteer is exposed and the purchaser protected; the forfeiture statute reproduces the common-law figure in section 308; the regulatory overlay is graduated by value and suspicion; and no court treats a ledger entry as self-executing title. That the doctrine has exactly this shape — when it could logically have had any of the others — is the evidence that “bounded bearer finality” is a mechanism the law implements, and not a vocabulary imposed on it after the fact. Other jurisdictions are implications to be tested, not assumed: many civil-law systems appear to reach a similar functional destination through good-faith acquisition of movables and special treatment of money, but whether the four-finalities structure travels to them is an open question — and naming the jurisdiction is what makes it a question rather than a smuggled generalization.
14. Coda: the note in your pocket
The banknote in your pocket is the strangest property you own. It is the only asset the law has deliberately made unpursuable — the one thing a victim of theft cannot follow into your honest hands, by a rule laid down when Britain still had colonies in America and preserved, condition for condition, in a forfeiture statute passed in 2002. It is a register-proof asset in Crawford’s sense: the one class of goods for which the law’s best anti-theft technology would destroy the thing it protected. It is memory of quantity and amnesia of provenance — an anti-record, engineered so that no rational recipient ever has a reason to ask it a question. And it is bounded: hedged by conditions that are categorical, wrapped in an overlay that is probabilistic, sitting under a sovereign who can switch the whole arrangement off.
None of this is visible on the note itself, which is the point. The most successful settlement institution in history works by making its own machinery unnecessary to know. But the machinery decides, right now, questions that are being answered in statutes and central bank design documents: whether a stablecoin is money or a claim, whether a CBDC can ever be cash, whether “irreversible” means anything a court must respect. Those questions cannot be answered by anyone who thinks finality is one thing that fades with the size of the payment. It is four things. One of them is absolute wherever its conditions hold. The rest is politics, enforcement, and power — which is to say, institutions all the way down.
Sources
Every source below was read in full before being relied on. Cases and statutes are cited as decided and enacted.
Alchian, A. A. (1977), “Why money?”, Journal of Money, Credit and Banking 9(1): 133–140.
Allen, D. W. (2023), “On the primacy of economic property rights”, Journal of Institutional Economics 19(2): 272–275.
Brunner, K. and Meltzer, A. H. (1971), “The uses of money: money in the theory of an exchange economy”, American Economic Review 61(5): 784–805.
Choi, G. S. and Storr, V. H. (2022), “The market as a process for the discovery of whom not to trust”, Journal of Institutional Economics 18(3): 467–482.
Clarke, C. (2021), “The legacy of Frank H. Knight for the politics of financial governance”, Journal of Institutional Economics 17(6): 973–987.
Coase, R. H. (1937), “The nature of the firm”, Economica 4(16): 386–405.
Coase, R. H. (1960), “The problem of social cost”, Journal of Law and Economics 3: 1–44.
Commons, J. R. (1931), “Institutional economics”, American Economic Review 21(4): 648–657.
Crawford, M. J. R. (2025), “The riddle of the good faith purchaser”, Oxford Journal of Legal Studies 45(1): 167–192. https://academic.oup.com/ojls/article/45/1/167/7900629
Davidson, S., De Filippi, P. and Potts, J. (2018), “Blockchains and the economic institutions of capitalism”, Journal of Institutional Economics 14(4): 639–658.
De Filippi, P., Mannan, M. and Reijers, W. (2020), “Blockchain as a confidence machine: the problem of trust and challenges of governance”, Technology in Society 62: 101284.
Fabbri, M., Rizzolli, M. and Maruotti, A. (2021), “Possession is nine-tenths of the law: possession, property, and coordination in a Hawk–Dove experiment”, Journal of Institutional Economics 17(2): 267–288.
Fox, D. (1996), “Bona fide purchase and the currency of money”, Cambridge Law Journal 55(3): 547–565.
Fox, D. (2008), Property Rights in Money, Oxford University Press.
Gill, A. and Thomas, M. D. (2023), “The dynamic efficiency of gifting”, Journal of Institutional Economics 19(1): 70–85.
Giraudo, M. (2022), “On legal bubbles: some thoughts on legal shockwaves at the core of the digital economy”, Journal of Institutional Economics 18(4): 587–604.
Gorton, G. and Ordoñez, G. (2014), “Collateral crises”, American Economic Review 104(2): 343–378.
Gorton, G. and Pennacchi, G. (1990), “Financial intermediaries and liquidity creation”, Journal of Finance 45(1): 49–71.
Hindriks, F. and Guala, F. (2015), “Institutions, rules, and equilibria: a unified theory”, Journal of Institutional Economics 11(3): 459–480.
Hingl, K. and Shera, M. (2025), “Ideas, institutions, and incompleteness”, Journal of Institutional Economics 21: e8.
Hodgson, G. M. (2006), “What are institutions?”, Journal of Economic Issues 40(1): 1–25.
Kahn, C. M., McAndrews, J. and Roberds, W. (2005), “Money is privacy”, International Economic Review 46(2): 377–399.
Kahn, C. M. and Roberds, W. (2009), “Why pay? An introduction to payments economics”, Journal of Financial Intermediation 18(1): 1–23.
Kiyotaki, N. and Wright, R. (1989), “On money as a medium of exchange”, Journal of Political Economy 97(4): 927–954.
Knight, F. H. (1921), Risk, Uncertainty, and Profit, Houghton Mifflin.
Kocherlakota, N. R. (1998), “Money is memory”, Journal of Economic Theory 81(2): 232–251.
Lagos, R., Rocheteau, G. and Wright, R. (2017), “Liquidity: a new monetarist perspective”, Journal of Economic Literature 55(2): 371–440.
Ménard, C. and Martino, G. (2025), “Unbundling institutions: the case for meso-institutions”, Journal of Institutional Economics 21: e31.
Oliveira, G. M. de and Miranda, B. V. (2024), “Environmental enforcement, property rights, and violence: evidence from the Brazilian Amazon”, Journal of Institutional Economics 20: e27.
Rogoff, K. S. (2015), “Costs and benefits to phasing out paper currency”, NBER Macroeconomics Annual 29: 445–487.
Searle, J. R. (2005), “What is an institution?”, Journal of Institutional Economics 1(1): 1–22.
Vahabi, M. (2025), “Recurrent exchange rate shocks and Anfal in Iran”, Journal of Institutional Economics 21: e36.
Whitman, G. (2025), “Countless ways to count: the functional heterogeneity of number systems”, Journal of Institutional Economics 21: e32.
Williamson, O. E. (1979), “Transaction-cost economics: the governance of contractual relations”, Journal of Law and Economics 22(2): 233–261.
Williamson, O. E. (2000), “The new institutional economics: taking stock, looking ahead”, Journal of Economic Literature 38(3): 595–613.
Wright, R., Tekin, E., Topalli, V., McClellan, C., Dickinson, T. and Rosenfeld, R. (2017), “Less cash, less crime: evidence from the electronic benefit transfer program”, Journal of Law and Economics 60(2): 361–383.
Cases: Miller v Race (1758) 1 Burr 452; 97 ER 398 (KB); Goodwin v Robarts (1875) LR 10 Ex 337, affd (1876) 1 App Cas 476 (HL); Banque Belge pour l’Étranger v Hambrouck [1921] 1 KB 321 (CA); Lipkin Gorman (a firm) v Karpnale Ltd [1991] 2 AC 548 (HL); BCCI (Overseas) Ltd v Akindele [2001] Ch 437 (CA); Foskett v McKeown [2001] 1 AC 102 (HL).
Legislation and policy: Bills of Exchange Act 1882; Currency and Bank Notes Act 1954; Coinage Act 1971; Financial Markets and Insolvency (Settlement Finality) Regulations 1999; Proceeds of Crime Act 2002 (ss. 327–329, Part 5, s. 308); Money Laundering, Terrorist Financing and Transfer of Funds (Information on the Payer) Regulations 2017; GENIUS Act of 2025 (Public Law 119-27); Executive Order 14178 (2025); People’s Bank of China e-CNY framework (effective 1 January 2026).