The Asset the Law Gave Up On
Why money is the vanishing point of the good-faith-purchase problem, and what payments economics sees when it looks down the same line
Keywords: good faith purchase; payment finality; currency of money; identifiability; nemo dat; store-of-value; account-based payment; bona fide purchase; law and economics; transaction costs
Abstract. Property lawyers argue endlessly about the good-faith purchaser: a thief sells stolen goods to an innocent buyer, and the law must choose between two victims. Payments economists study why anyone accepts anything in settlement of a debt. This essay argues the two are one subject seen from opposite ends of a single axis — the identifiability of the asset. As identifiability falls, the owner’s incentive to search, the buyer’s incentive to verify, and the deterrent value of any title rule all fall with it. At the vanishing point, where the asset is perfectly fungible, the efficient legal rule is not a better allocation between the parties but the abolition of the contest: the honest recipient for value takes fresh title and the transaction is final. That asset has a name — money — and that rule has two names, the currency of money in law and payment finality in economics. Drawing on Crawford (2025), Fox (1996), Kahn and Roberds (2009), and Levmore (1987), the essay proposes a threshold, states it as a conjecture with untested premises, and traces where the deterrence of theft goes once title rules can no longer carry it.
I. Two literatures, one problem
There is a puzzle that property lawyers have argued about for as long as there has been property law, and there is a field of economics that most property lawyers have never read. This essay argues that they are the same subject, viewed from opposite ends of a single spectrum, and that once you see the spectrum whole, several things that look like anomalies in each literature turn out to be theorems of the other.
The lawyers’ puzzle is the good faith purchaser problem. A thief steals an asset from its owner and sells it to a buyer who pays full value and suspects nothing. The owner later finds the asset in the buyer’s hands and demands it back. The thief has vanished or is insolvent, so the dispute is between two innocent parties, and the law must pick one of them to bear the loss. Michael Crawford’s recent treatment in the Oxford Journal of Legal Studies (Crawford, 2025) is the sharpest statement of the problem in years, and its central provocation is that “why it is considered a riddle at all” is itself the question worth asking. Crawford’s answer, developed below, is that the law’s objective in these disputes should not be justice between owner and purchaser at all. It should be the suppression of theft. And for a large class of assets, he argues, no allocation of title between the two innocents can serve even that objective, because neither party’s behavior responds to the rule.
The economists’ field is payments economics, and its canonical introduction is Kahn and Roberds (2009), which opens by calling payment systems “the plumbing of the economy”: essential, pervasive, and ignored until something breaks. Payments economics asks why anyone accepts anything in settlement of a debt. Its answer is that payment arrangements exist to overcome two frictions: a time mismatch between when people produce and when they consume, and limited enforcement of promises about future behavior. Every payment system is a machine for solving those two frictions, and the two great architectures for doing so, which Kahn and Roberds call store-of-value systems and account-based systems, differ in exactly one fundamental respect: what must be verified for the payment to work. A store-of-value system verifies the object. An account-based system verifies the identity of the account holder.
Here is the connection, and it is not an analogy. The good faith purchaser problem is a problem about verification and identification of objects: can the owner identify her stolen asset in a stranger’s hands, and can the buyer verify his seller’s title before paying? Payments economics is a theory of what happens to exchange when verification and identification are costly, and of the institutional machinery that societies build when object-level verification fails. The property lawyers study the region of the asset space where identification partially works. The payments economists study the endpoint where it does not work at all, an endpoint that has a name. The name is money.
Money is the asset the law gave up on. Not by accident, not by oversight, and not as a grubby commercial exception grafted onto pure doctrine, but as the limiting solution of the very same optimization problem that generates the good faith purchase rules for paintings, cars, and cattle. When the identifiability of an asset falls toward zero, the owner’s incentive to search for it after a theft falls toward zero, the buyer’s incentive to investigate title falls toward zero, and the value of any title rule as a deterrent to theft falls toward zero. At that limit, the efficient legal rule is not a better-calibrated allocation between owner and purchaser. It is the abolition of the contest: the recipient in good faith and for value takes a fresh title good against the whole world, and the transaction is final. English law reached that answer for coin centuries before anyone could state the economics, and reached it for banknotes in a single famous case in 1758. The doctrinal name for the answer is the currency of money. The economic name for the same answer is payment finality. This essay traces the argument from one end of the spectrum to the other and shows what falls out along the way: why legal systems disagree so stubbornly about stolen goods but agree unanimously about stolen cash, why registers work for Ferraris and would be absurd for five-pound notes, why the deterrence of theft, when it can no longer live in title rules, migrates into the identity-verifying machinery of the payments system itself, and why the design of money is, at bottom, the choice of a point on the identifiability spectrum.
II. The riddle, restated as economics
Start where Crawford starts, with the structure of the dispute. Call the owner O, the thief T, and the good faith purchaser P. T steals from O and sells to P; O locates the asset and sues P in conversion. Who should win? The intuitive answer, that O should win because she is the owner, is circular: “owner” is just the label the law attaches to whoever holds the best claim, which is precisely what is in dispute. The corrective justice framework that organizes most of private law also fails here, and fails structurally. Corrective justice pairs a wrongdoer with a victim: the doer and sufferer of the same injustice. O and T form such a pair, and so do P and T. But O and P do not. They are linked only by their shared status as victims of a third party who is, in the standard phrase of the cases, not worth suing. The dispute the court must actually resolve is the one dispute in the triangle that corrective justice cannot reach.
Crawford’s move is to change the question. Instead of asking which allocation between O and P is just, ask what social loss the law should be minimizing. And here the analysis produces its first genuinely uncomfortable result: viewed narrowly, the theft itself might not be a loss at all. O’s loss is matched by T’s gain, so the involuntary transaction looks like a pure transfer, distributive rather than destructive. If P happens to value the asset more than O did, the sequence of theft and resale has even, in the narrowest accounting, increased total willingness-to-pay. So why does every legal system on earth prohibit it?
Because the accounting is wrong, and wrong in an instructive way. Theft is never a costless transfer. T burns real resources planning and executing the theft. O burns real resources on locks, guards, and precautions, which prompts T to invest in overcoming them, and the arms race between them continues until the returns to theft are competed away into pure waste. P bears verification costs precisely because theft exists as a background possibility, and where stolen goods pass through fences before reaching P, the chain of illicit intermediaries multiplies the transaction costs of moving one asset from an unwilling seller to a final buyer. Behind all of this sit the deadweight losses of distorted production: an economy in which theft is profitable is an economy in which people make things that are hard to steal rather than things that are valuable, and at the limit make nothing at all. The social cost of theft is not the misfortune of whichever innocent party loses the lawsuit, and it is not the gap between O’s and P’s subjective valuations. It is the entire negative-sum apparatus of theft as an enterprise. Figure 1 lays out the triangle and where the real costs sit.
Once the objective is restated this way, the law’s task in the O versus P dispute clarifies. The court is not distributing a loss; it is setting an incentive. The question is which rule, owner wins or purchaser wins, does more to reduce the returns to T. A rule favoring O preserves O’s incentive to search for stolen goods and forces P to worry about title, both of which lower the price T can get and raise T’s chance of being caught. A rule favoring P economizes on O’s socially wasteful recovery efforts, since a completed transfer to a good faith purchaser cannot be clawed back, but it also makes stolen goods easier to sell. There is a general belief among the law-and-economics writers, which Crawford shares with appropriate caution, that exceptions to the owner-wins rule increase the returns to theft: a purchaser-friendly rule means T needs only a thin cover story, O stops searching, the police lose their most motivated informant, and the effective price of stolen goods rises with their marketability. So far, so good for the traditional common law position, the rule that no one can give what he does not have, nemo dat quod non habet.
But the argument has a second act, and it is the second act that opens the door to money.
III. Why the title rules do not bite
The deep obstacle to using title rules as theft deterrence is a problem that Schwartz and Scott (2011) named the “double moral hazard” of good faith purchase. The owner takes optimal precautions against theft only if she expects to bear the loss when theft succeeds, which means only if she cannot recover from the purchaser. The purchaser investigates his seller’s title optimally only if he expects to bear the loss when the goods turn out to be stolen, which means only if the owner can recover from him. Both conditions cannot hold at once. Every categorical title rule, however refined, puts the full loss on one side and thereby destroys the other side’s incentive. The good faith qualification softens this only slightly, because the standard the purchaser must meet is honesty in fact, a test of subjective sincerity rather than of efficient investigation, and honesty is cheap. Schwartz and Scott’s own proposed escape, a negligence regime under which owners recover only if they took reasonable precautions, founders on the cost of specifying and litigating the standard of care for every category of chattel, and Crawford adds the sharper point that no jurisdiction has ever adopted it.
Crawford’s contribution is to show that for most assets the double moral hazard is not even the binding constraint, because something more brutal binds first: the owner will not search. A rational owner invests in searching for stolen goods up to the point where the marginal cost of search equals the marginal increase in the probability of recovery multiplied by the asset’s value to her. For goods that are cheap, or valuable but generic, or valuable and distinctive but easily altered, that calculation terminates almost immediately. The stolen laptop is one of ten million identical units. The stolen jewelry is melted by nightfall; the stolen car is in parts within a week; the recut diamond is a different stone. The owner knows the search is hopeless and rationally spends nothing on it, reporting the theft, if at all, only because her insurer requires it. And a purchaser who correctly anticipates that owners do not search has no reason to investigate title regardless of what the law says, because the probability that any owner ever appears to assert a claim is effectively zero. The parties behave as if the law favored the purchaser whatever the statute book says. In the language of incentives, the title rule does not bite. This is Crawford’s explanation for a pattern that Levmore (1987) documented and that later comparative work, on Crawford’s account, extended to a much larger sample (Crawford cites a survey putting the count at 247 jurisdictions; I read Levmore in full and take the larger figure from Crawford’s citation, not from the survey itself, which I have not read). Legal systems exhibit stubborn, convergence-resistant variety in their treatment of the good faith purchaser, from the strict owner-protection of the common law to the strong purchaser-protection of the Italian Civil Code, with many intermediate arrangements involving markets, time limits, and reimbursement conditions. Levmore’s functional thesis is that legal variety flourishes exactly where rules either do not matter behaviorally or present “the difficulty of discerning the best solution to a hard question”. The good faith purchaser problem, over most of the asset space, fits both of Levmore’s conditions at once: the rules barely move behavior, and to the extent they move it at all, reasonable lawmakers can disagree about the direction. On the reading advanced here, diversity of doctrine is what behavioral indifference looks like once it is written down across many legal systems.
Hold on to the variable doing the work in that argument, because it is the hinge of everything that follows. The reason the owner does not search is not that the asset is worthless; it is that the asset cannot be identified and proved hers at acceptable cost. The operative parameter is identifiability. Call it s. Everything in the good faith purchase problem, the owner’s search, the purchaser’s verification, the bite of the title rule, the deterrent effect on the thief, runs through s.
IV. Where the rules do bite: high identifiability and the register
Now run the machine in the other direction. Let the asset be valuable, unique, durable, and worthless if altered: a name painting, a documented classic car, a manuscript. For such assets the owner’s search calculus flips. The marginal return to search is high because the asset, wherever it surfaces, is recognizably hers and provably hers, and it cannot be anonymized without destroying the value that made it worth stealing. A mangled Renoir is a worthless piece of canvas, so the thief must preserve exactly the features that permit its recapture. Here the owner searches, the purchaser knows she searches, the purchaser therefore verifies, and the legal rule finally has behavior to act upon. It matters, at this end of the spectrum, whether the law says owner or purchaser.
And it is exactly this class of assets, Crawford observes, that is amenable to registration, which is where his affirmative proposal lives. A register is a technology for broadcasting ownership information at low cost, and its celebrated function is facilitative: it lets buyers verify title cheaply, unlocking trades that verification costs would otherwise block. But a register has a second function that matters more for theft: an obstructive function. A register does not merely certify the true owner’s information; it discredits the thief’s story. When the purchaser can check a database and see that the seller is not the owner, or that the item is recorded as stolen, the sale either dies or proceeds only at a deep discount that strips out the thief’s profit, and the purchaser, now knowingly dealing with a thief, becomes a source of information for the police. Registration attacks the returns to theft at the point of resale, which is the point where theft is monetized. Crawford’s doctrinal proposal follows: where a viable register exists, the law should resolve owner-purchaser disputes so as to maximize registration and consultation, which means the owner should prevail if and only if she registered the item or promptly registered the loss. An owner who could have registered and did not should lose to the good faith purchaser, on pain of which owners will register, registers will become comprehensive, and the obstructive machine will grind down the stolen-goods trade. The register need not even be public. Crawford records that the Art Loss Register, a private for-profit database of stolen art, charges owners to list items and buyers to search, and operates in exactly the high-value, high-identifiability asset class where the framework locates the register regime. I take that from Crawford’s reporting, which I read in full; I have not independently audited the register’s operations or verified that its coverage matches the framework’s prediction, and I do not claim its existence confirms the theory. It is consistent with the theory, and no more than that.
Registration has limits, and the limits are as instructive as the successes. A register is pointless for low-value goods, because the fixed cost of consulting it does not scale down with the price of a bottle of milk. It is pointless for goods that can be altered past recognition, because a register requires a stable alignment between the recorded description and the physical thing. It shifts crime rather than eliminating it, since a register that certifies titles invites identity fraud against the register itself, the classic Achilles heel of state-guaranteed title systems. And it redistributes regressively when funded from general taxation, because registrable assets are disproportionately the property of the rich. Every one of these limits is an identifiability condition or a cost condition, and together they carve the asset space into the three regions shown in Figure 2. At high s, the register regime: search pays, registration is feasible, and the law can condition title on it. In the broad interior, the contested zone: the double moral hazard is insoluble, search is marginal, the rules barely bite, and legal systems scatter across the solution space exactly as Levmore’s variety thesis predicts. And at the far left, as s falls toward zero, a third regime that neither Crawford nor the good faith purchase literature pursues to its conclusion, because at that end of the spectrum the asset stops being called goods and starts being called money.
V. The endpoint: how money escaped property law
The legal history of that endpoint has been reconstructed with great care by David Fox (1996), and it repays close attention, because the doctrinal record is, read correctly, a four-hundred-year natural experiment in exactly the mechanism this essay is describing.
English law gives money a property that no other chattel has, which the old books call currency: a recipient who takes money in good faith and for valuable consideration acquires a fresh legal title, good against the whole world, that does not derive from his transferor’s title at all. The thief who steals your coins acquires nothing, but the shopkeeper who innocently takes those coins from the thief in payment acquires everything, and your title is extinguished at that moment. Currency is the polar opposite of nemo dat. For every other kind of property, title survives non-consensual transfer; for money, title is reborn in each honest hand it passes through. The rule is not an exception grafted onto money for convenience. It is, Fox shows, the second of two distinct legal rationales that have successively performed the same function, and the succession between them is the revealing part.
The first rationale was evidentiary, and it was stated in the language of the sixteenth and seventeenth century courts as the maxim that money has no earmark. One coin of a given denomination was indistinguishable from every other. A plaintiff suing to recover specific coins had to prove that these very coins were his, and once coins left his possession and mingled with anyone else’s, that proof was impossible. The courts drew the consequence with complete candor: detinue, the action for the specific return of property, would not lie for money unless the money was sealed in a bag or box, because only the bag gave the coins an identity that could be proved at trial. Loose in another’s possession, coins effectively belonged to their possessor, not because any statute said so, but because no former owner could ever discharge the burden of identifying them. Property followed possession as a matter of evidence before it did so as a matter of doctrine. In the vocabulary of this essay: for coined money, s was zero as a physical fact, the owner’s probability of proof was zero at any level of search expenditure, and the law’s abandonment of recovery was not a policy choice but a recognition that the recovery machinery had nothing to grip.
Then the physical fact changed, and the law’s response is the crucial observation. From the 1640s, London goldsmiths began issuing paper instruments, promissory notes recording deposits, and these instruments were handwritten, dated, and named the original payee. Paper money, unlike coin, had earmarks. For the first time the owner of stolen money could identify it in a stranger’s hands: he could advertise the loss, notify the issuing bank, and stop payment, and the bank could recognize the note when presented. Technically, s had jumped from zero to something substantial. Under the old evidentiary rationale, banknotes should therefore have been governed by nemo dat like any other identifiable chattel, and the victim of theft should have recovered his note from whoever held it. Had the courts followed the logic of identifiability mechanically, paper money would have been legally traceable, every note a little lawsuit waiting to happen.
They did not follow it, and Miller v. Race, decided by Lord Mansfield in 1758, is the moment the refusal became doctrine. An innkeeper innocently gave change for a Bank of England note that, unknown to him, had been stolen from the mails; the owner had stopped payment; the Bank’s clerk refused to pay and detained the note; the innkeeper sued and won. Mansfield expressly rejected the no-earmark explanation as the basis of the result: “The true reason is on account of the currency of it” (Miller v. Race, 1758, as quoted in Fox, 1996). Money passes in currency; once it has passed to a good faith purchaser for value, the former owner’s title is simply gone, earmarks or no earmarks. Fox demonstrates that Mansfield was not inventing but ratifying: for sixty years the practices of bankers and merchants had treated notes as cash, an embryonic bona fide purchase rule had been operating in commercial usage, and the courts progressively absorbed it because the alternative was intolerable. The value of a banknote lay entirely in its unquestioned acceptability. If every recipient had to consider the possibility that the note in his hand could be reclaimed by some prior owner, notes would circulate at a discount reflecting title risk, transactions would slow while recipients investigated, and the note system, which existed precisely to economize on the costs of moving coin, would forfeit its reason for existing. Lindley LJ’s later formulation, quoted by Fox, compresses the whole economics into an epigram about land versus commerce: in dealings with land title is everything and can be leisurely investigated, while in commercial dealings possession is everything and there is no time to investigate title.
Read as legal history, Fox’s story is about the absorption of mercantile custom into common law. Read as economics, it is something stronger: it is a demonstration that the finality rule for money is not merely the passive consequence of low identifiability but an active legal commitment to treat money’s identifiability as zero even when it is technically positive. Banknotes had serial numbers and named payees; the law chose to make those earmarks legally inert in the hands of a good faith taker, because the exchange value of the instrument depended on their inertness. Identifiability, in other words, is not only a physical parameter of the asset. It is a design parameter of the legal regime, and for money the regime deliberately sets it to zero at the point of honest receipt. That observation, to which the essay returns in section IX, is what makes the analysis bite on the monetary design questions of the present day.
VI. Payments economics: what the finality rule is for
To see why the law was right to do what it did, step across to the economists’ side of the spectrum and ask the question payments economics asks: what problem does paying solve? Kahn and Roberds’ answer, built from a stripped-down model of agents arranged in a ring who periodically want each other’s goods, is that payment arrangements exist to overcome two frictions in combination. The first is a time mismatch: I want your good today, but the good you will want from the economy arrives later, so exchange cannot be a simultaneous swap. The second is limited enforcement: my promise to deliver later is only as good as the machinery for punishing me if I break it, and that machinery, courts, reputations, communal sanctions, is costly and incomplete. Where enforcement is perfect and cheap, pure credit suffices and nothing resembling money is needed: everyone delivers on schedule because defection triggers punishment, and the economy runs on promises. Payment instruments earn their existence exactly where promises fail.
When they fail, two architectures are available, and everything in the payments universe is one of them, or a hybrid. In an account-based system, the payer’s and payee’s positions are entries in a ledger, and payment is the adjustment of entries. The system works if and only if two informational technologies work: one that tracks an individual’s actions over time, and one that verifies that the individual standing in front of you is the individual in the ledger. Account systems live or die on identity. In a store-of-value system, payment is the physical transfer of an object, coins, notes, tokens, and the system works if the payee can verify one thing only: the genuineness of the object. He needs to know nothing about the payer. No history, no name, no creditworthiness, no ledger. The informational parsimony of the store-of-value architecture is its entire point: it is the payment system for strangers, for the anonymous, the itinerant, and the unrecorded, for every setting where the identity infrastructure that account systems require does not exist or costs too much. Figure 4 sets the two architectures side by side.
The literature Kahn and Roberds survey contains a result that gives this dichotomy its sharpest expression: in a class of models, money and perfect record-keeping are substitutes, in the sense that anything achievable with money is achievable with a complete public record of everyone’s transaction history, and vice versa. Money, on this view, is memory made portable: the balance in your pocket is a compressed, bearer-form summary of your net contributions to the economy, verifiable by inspection instead of by consulting a database. The formulation explains at a stroke why the store-of-value form dominates where record-keeping is expensive and why its territory shrinks as record-keeping gets cheap, a secular trend Kahn and Roberds document in the century-long migration from cash to accounts. But it also exposes exactly what the object-verification architecture cannot survive: any requirement to consult the object’s history. The moment the payee must ask where a coin has been, who has held it, and whether any prior holder’s title was defective, the coin has stopped being memory and become evidence, and the account-based system’s costs have been reimported into the store-of-value system without its benefits. A store-of-value instrument whose acceptability depends on its provenance is a self-contradiction.
Now the two literatures snap together. The legal rule of currency, Fox’s fresh title in every honest hand, is precisely the rule that forbids the coin’s history from mattering. Bona fide purchase for value is the doctrinal device that guarantees the informational parsimony on which the store-of-value architecture depends: it converts the question the payee would otherwise have to ask, is this payer’s title good, into a question he can answer by looking at the object and his own conscience. Payment finality is not a nice-to-have feature of money; it is the legal implementation of the only architecture that works when identity infrastructure is unavailable, and the good faith purchase rule for money is where private law delivers it. Conversely, the good faith purchase problem for goods is what the world looks like when finality is withheld: every buyer of a used chattel is, in a small way, a payee who has been told that the object’s history matters, and the verification costs, discounts, and litigation that follow are the price of that ruling. The property lawyers and the payments economists have been staring at the same trade-off, recovery versus circulation, history versus finality, from its two ends.
One further observation from the payments side completes the picture, because finality turns out to be load-bearing at the top of the monetary system as well as at the bottom. Kahn and Roberds document the wholesale migration, during the 1990s, of the world’s large-value interbank systems from net settlement to real-time gross settlement, an architecture in which every payment between banks is an immediate and irrevocable transfer of central bank funds. The motive was systemic risk: under netting, one bank’s failure to settle could cascade through the web of offsetting obligations, and regulators, after several near misses, chose to buy insulation at the price of vastly higher liquidity needs. What they bought, stated precisely, was finality: the guarantee that a payment received is a payment kept, proof against unwinding, no matter what later becomes of the payer. That is the same commodity, purchased at the same price of foreclosed recourse, that the currency rule supplies to the shopkeeper taking a note. From the innkeeper in Miller v. Race to the settlement systems moving trillions a day, the monetary system’s answer to the question of whether value received can be reclaimed is the same at every scale, and it is no. The system runs on the extinguishment of prior claims, and the deeper one descends into its plumbing, the more explicitly its engineers say so.
VII. One threshold for the whole spectrum
The claim that this is one problem rather than two similar problems can be made exact, and the exercise is worth doing in the open, because the formalism is small and every piece of it has already appeared in the prose. This section states the model that organizes my current research program on cash and payment finality; the two literatures reviewed above supply, respectively, its interior and its endpoint.
Let an asset have value V to its owner and identifiability s, where s indexes, on a scale from zero upward, the technical feasibility of identifying the specific asset and proving prior ownership of it: zero for a loose coin among coins, low for generic electronics, high for a documented painting. Let the owner’s search technology be summarized by R(s), the marginal increase in the probability of locating and provably recognizing the asset per unit of search expenditure, with R increasing in s and R(0) = 0: search productivity rises with identifiability and vanishes with it. Let θ be the probability that a located asset can be evidentially connected to the owner at trial, let δ be the proportional cost loading of the recovery process itself, litigation, enforcement, delay, and let c be the owner’s unit cost of search effort. A rational owner searches at all only if the expected value of recovery covers the cost of achieving it, and working through the owner’s first-order condition yields a threshold value
V*(s) = c / [ θ · R(s) · (1 + δ) ]
such that owners search, and the whole recovery apparatus engages, only for assets with V above V*(s). The threshold is monotone decreasing in s: the more identifiable the asset, the lower the value needed to make recovery rational. Figure 3 plots it. Above the curve is the recovery regime, where owners search, buyers verify, title rules bite, and Crawford’s register logic applies. Below the curve is the finality regime, where search is irrational at any legal rule, verification is worthless, and the efficient law writes off recovery between the parties entirely. The interior band around the curve, where V and V*(s) are comparable, is the contested zone where the double moral hazard is live and Levmore’s variety flourishes.
The unification is in the limit. As s falls to zero, R(s) falls to zero and V*(s) diverges: no finite value, however large, justifies search for a perfectly fungible asset. Money is not an exception to the good faith purchase framework; money is its s → 0 limit, the corner of the asset space where the threshold has gone to infinity and the recovery regime is empty. The good faith purchase literature, from the ancient codes through Levmore, Schwartz and Scott, and Crawford, is the theory of the interior, s strictly positive, where the question of who should win is hard because both answers have costs. The law of currency, from the no-earmark cases through Miller v. Race to modern banking law, is the theory of the endpoint, where the question of who should win has dissolved because one of the contestants, the searching owner, has rationally left the field. And Fox’s history supplies the experiment that separates the physical parameter from the legal one: when the physical s of money jumped upward with the arrival of earmarked paper, the law re-imposed the endpoint by doctrine, holding legal s at zero because the asset’s function demanded it. Fungibility in law is a commitment, not just a fact.
VIII. The premises, and what would test them
This is an essay, not a finished paper, and the distinction matters most here. The threshold conjecture rests on premises. None of them is established below; each is stated plainly as an untested premise, with the empirical test that would be required to earn it. Anyone who wants to turn the conjecture into a result has to do that work, and I have not done it here.
Premise 1: rational recovery behavior. Owners direct search and registration expenditure so as to maximize expected net recovery. This is not demonstrated. To test it one would need expenditure data — insurer recovery spending, private-investigator engagement, register subscriptions — matched to asset value and asset identifiability, and would have to show that expenditure tracks expected net recovery rather than, say, sentiment, insurance-contract requirements, or reporting mandates. I have not assembled that data and do not assert the premise holds.
Premise 2: search productivity vanishes with identifiability, R(0) = 0. The functional form R(s), increasing with R(0) = 0, is imposed, not derived. Fox’s cases show courts stating that specific recovery of loose coin was impossible for want of proof of identity, and allowing recovery when a sealed bag restored identity; that is historical support for the qualitative claim at the s = 0 corner, and nothing more. It does not establish the shape of R(s) anywhere in the interior, nor that recovery probability is continuous or monotone in identifiability across asset classes. Testing the shape would require recovery-rate data across a graded range of asset types with an independent measure of identifiability. I have not built that measure or that dataset.
Premise 3: the purchaser’s verification incentive is derivative of the owner’s search incentive. P investigates title only to the degree that an O might realistically appear. This is Schwartz and Scott’s complementarity between owner search and buyer investigation, extended by Crawford, and here it is adopted as a premise, not proven. Its test is the industrial organization of verification services: whether provenance research, title insurance, and vehicle-history checking exist only where owner search is observed, and whether their intensity scales with owner-search intensity. I have not conducted that test.
Premise 4: at s = 0 the least-cost deterrence margin is institutional rather than transactional. Because neither a fungible object nor an anonymous transaction can carry identity, the conjecture is that deterrence expenditure will locate at the account-based interfaces of the monetary system rather than at the point of exchange. This is the load-bearing claim for section IX, and it is a conjecture. Its test is a location study: where, measured in enforcement spending and regulatory obligation, does anti-theft and anti-laundering effort actually sit — at account opening and ledger-keeping institutions, or at hand-to-hand exchange? The claim is falsifiable and I have not falsified or confirmed it. I assert only that the conjecture makes the prediction, not that the prediction has been verified.
Two of these premises (2 at the s = 0 corner, and the qualitative direction of 4) have historical or architectural support in the four sources this essay actually read. The rest are open. A reader should treat sections VII through X as a research proposal with a formal skeleton, not as an established finding. That is the honest status of the argument.
IX. Where recovery goes when it leaves the courtroom
Nothing in the argument so far says that theft of money should be, or is, undeterred. It says that one particular deterrence instrument, the allocation of title between owner and good faith purchaser, has zero traction at s = 0. The deterrence problem does not disappear at the endpoint; it relocates. Understanding where it goes is the payoff of putting the two literatures together, and the guiding idea comes from a third literature entirely, the oldest one in institutional economics.
Coase (1937) asked why firms exist if the price mechanism allocates resources so well, and answered that “there is a cost of using the price mechanism”: discovering prices, negotiating contracts, and enforcing them transaction by transaction is expensive, and where those per-transaction costs exceed the costs of administrative coordination, activity moves inside an organization and the market is superseded. The boundary of the firm sits where the marginal cost of organizing internally equals the marginal cost of transacting externally. The structure of that argument, an activity migrating from a decentralized per-transaction mechanism to a centralized institutional one when per-transaction costs become prohibitive, is exactly the structure of what happens to the recovery of stolen value as s falls to zero.
In the recovery regime, deterrence is decentralized and per-transaction. Each owner searches for her own goods; each buyer verifies his own seller; each dispute is litigated on its own facts; the register, where it exists, is consulted purchase by purchase. The mechanism is the property-law analogue of the market: distributed, case-by-case, running on the initiative of the parties. At s = 0 every one of those per-transaction operations has infinite cost per unit of effect, for the reasons the model makes exact. So the function migrates, Coase-fashion, into institutions, and the institutions it migrates into are precisely the ones payments economics puts at the center of the monetary system: the keepers of the account-based layer. A bank verifies its customer’s identity once, at account opening, and thereafter every payment through the account inherits that identification at near-zero marginal cost. The ledger records history that the bearer instrument cannot carry. Reversal, freezing, monitoring, and reporting, all impossible against a coin in an unknown pocket, are routine operations against an entry in a known ledger. The economics that make it efficient for the law to abandon the recovery of specific coins between private parties are the same economics that make it efficient to concentrate the control of monetary crime at the institutional choke points where money’s anonymity ends: the deposit, the account, the wire, the threshold report. Deterrence of theft does not vanish from the world of money. It is vertically integrated into the payments system.
On the reading this essay proposes — and I stress it is a reading, an interpretive lens, not a demonstrated result — several features of the monetary landscape line up as though they were one design. Crawford notes that when private law’s instruments run out, raising the expected criminal sanction on the thief is the instrument that remains; for money, with title rules inert, the framework predicts the deterrence load falls disproportionately there, though I have not measured that load. The store-of-value layer carries no identity, no history, and delivers instant finality, which Kahn and Roberds identify as the economic function of that architecture. The account-based layer verifies identity as a condition of working at all, and the conjecture is that recording and monitoring obligations settle there because that is where the marginal cost of carrying them is lowest — a cost claim I have argued for but not quantified. The hybrid instruments Kahn and Roberds analyze — transferable debt, checks, notes tied to an identifiable issuer or endorser — occupy the middle of the identifiability spectrum, more traceable than coin and less anonymous, and their legal treatment, the holder-in-due-course rules descended from Miller v. Race through the Bills of Exchange Act, preserves circulation while retaining recourse against identified signatories. Whether the law of negotiable instruments as a whole maps onto the interior of the spectrum in the tidy way this lens suggests is a claim I find suggestive and have not tested; I offer it as interpretation, not proof.
The relocation argument also disciplines how one should think about proposals to re-engineer money’s identifiability, in either direction. Because s is partly a legal and technical design choice, Fox’s lesson, a society choosing the form of its money is choosing a point on Figure 3’s horizontal axis, and every point carries the full bundle of consequences the framework describes. Raise the effective identifiability of the circulating medium, by design features that attach history to value, and you buy back some per-transaction traceability at the price of exactly the verification burdens, discounts, and finality doubts that the currency rule was built to eliminate; carried far enough, the medium stops functioning as a store-of-value instrument at all and becomes an account system wearing a token costume. Lower identifiability at the account layer, and you strip the institutional choke points of the identification on which the entire relocated deterrence apparatus depends. The framework does not say which trade-off a society should make; it says the trade-off is a single dial, that the dial has been set before, deliberately, in 1758, and that the costs on each side of any setting are the ones this essay has been pricing all along.
X. What the conjecture predicts, and what would confirm or break it
A conjecture earns attention by making falsifiable predictions. Four follow from the threshold framework. I state each as a prediction and, honestly, as untested here: what would confirm it, and what would break it. I am not claiming the world has been checked.
Prediction 1 — regime monotonicity. As identifiability rises across asset classes, efficient legal treatment should shift from finality toward recovery, with registration-conditioned title at the top. The gross architecture of property law is consistent with this at a glance — land and ships and aircraft are registered, chattels are litigated, money has currency — but consistency at a glance is not a test. A real test requires an independent, operationalized measure of identifiability, a coding of legal treatment across many asset classes and jurisdictions, and a demonstration that treatment tracks identifiability after controlling for value, tradability, and durability. That confound-control has not been done here. The prediction would break if high-identifiability asset classes were commonly governed by finality rules, or low-identifiability classes by recovery rules, once value is held constant.
Prediction 2 — diversity concentrates in the interior. Doctrinal disagreement across jurisdictions should be large for mid-identifiability goods and vanish at both ends. Levmore’s comparative work, which I read in full, documents extensive variety in the treatment of stolen goods and is the source of the qualitative claim; it does not, by itself, establish that variety is a function of identifiability rather than of legal origin, colonial transplant, or doctrinal path-dependence. Distinguishing those requires a variance decomposition of legal treatment against identifiability with legal-family fixed effects. I have not run it. The prediction would break if endpoint treatments (money, land) varied across systems as much as mid-spectrum treatments do, or if interior variety were fully explained by legal-family membership with no residual role for identifiability.
Prediction 3 — enforcement location at s = 0. For monetary theft, deterrence effort should concentrate at identity-bearing institutional interfaces and be minimal at hand-to-hand exchange. This is the sharpest and most testable prediction, and it is untested here. Confirmation would be a location study of enforcement expenditure and regulatory obligation showing concentration at account opening, deposit, threshold reporting, and ledger-keeping institutions, and near-absence at the point of cash exchange. The prediction would break if substantial anti-theft or anti-laundering enforcement effort were found to operate at hand-to-hand cash exchange, or if the account-layer concentration turned out to be explained by something other than the marginal cost of identification. I assert the prediction, not its confirmation.
Prediction 4 — function dominates form at the boundary. Where an asset’s physical identifiability and its exchange function collide, the legal system should suppress the identifiability rather than surrender finality. Miller v. Race (1758) is a single instance consistent with this: an identifiable, stopped banknote whose owner had done everything the recovery regime asks, and a court that extinguished the owner’s title anyway, citing the needs of circulation. One case is an illustration, not a test. A test requires the assembly of boundary episodes — moments when a circulating medium acquired identifiability — and a demonstration that the legal response systematically favored finality over traceability across them, rather than one memorable decision selected because it fits. I have read Fox’s account of the one episode; I have not assembled the class. The prediction would break if, at such boundaries, legal systems generally chose traceability and accepted the loss of finality.
The status is therefore this: four falsifiable predictions, two of them (2 and 4) with a single documented instance apiece drawn from sources read in full, none of them subjected to the controlled test that would convert prediction into finding. The framework is offered as worth testing, not as tested.
XI. Conclusion: the riddle was a limit all along
Crawford ends his article with the observation that the real riddle is why the good faith purchaser problem is considered a riddle at all, and the answer this essay has assembled is that the problem looks riddling only when viewed at a single point of the identifiability spectrum, where the double moral hazard is live, search hovers at the margin of rationality, and every allocation of title is defensible and inadequate at once. Widen the lens and the riddle resolves into a threshold. Above V*(s), the law can and should conscript owners and buyers into the suppression of theft, and Crawford’s register-conditioned title rule is the right instrument, because registration is the one technology that attacks the thief’s revenue at the point of sale. Below V*(s), the conscription is impossible, the parties will ignore any rule the law writes, and the honest doctrinal response is the one Levmore’s comparative record shows lawmakers groping toward for four millennia: pick a rule, any rule, and resolve the dispute cheaply, because the choice moves nothing.
And at the limit, where identifiability is gone entirely, the law long ago settled on the answer of not asking the question. The currency of money, the fresh title in every honest hand, is what the good faith purchase doctrine becomes at s = 0, the corner where the owner’s search, the buyer’s verification, and the court’s inquiry have nothing to grip. Note the exact status of that claim: conditional on the imposed premise R(0) = 0, the threshold V*(s) diverges as s falls to zero, so the divergence is an analytic consequence of the premise, not an independent empirical finding — the premise is where the content sits, and the premise is untested for the interior. Payments economics supplies the reason the endpoint answer is not a defeat: Kahn and Roberds identify finality as the load-bearing property of the store-of-value architecture, the feature that lets strangers transact without ledgers, and bona fide purchase is the legal device that delivers it. Where the deterrence of theft goes once title rules go inert is, on the argument here, into the identity-bearing institutions of the account layer, by the Coasean logic that displaced functions migrate to the lowest-cost margin; that is a conjecture with a falsifiable location prediction (section X), not a documented fact. One problem, one spectrum, one threshold: the property lawyers work one end and the payments economists the other, and money is the name the far end already carries. What remains is to test it.
A note on method and sources
Six works underpin this essay. I read each in full before using it; none is cited from an abstract or a snippet. The four load-bearing sources are Crawford (2025) on the good-faith-purchaser problem, Fox (1996) on the legal history of money’s currency, Kahn and Roberds (2009) on the architecture of payment systems, and Levmore (1987) on comparative variety. Schwartz and Scott (2011) supplies the double-moral-hazard result, and Coase (1937) the migration-to-institutions logic in Section IX. The quotation from Miller v. Race (1758) is given as reported in Fox (1996); I have not consulted the original Burrow report and attribute the wording to Fox. Where I state a fact about the present-day world — enforcement locations, register operations, recovery rates — I mark it as untested conjecture, because I have not tested it.
References
Coase, R. H. (1937). The nature of the firm. Economica, 4(16), 386–405. https://onlinelibrary.wiley.com/doi/10.1111/j.1468-0335.1937.tb00002.x
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
Fox, D. (1996). Bona fide purchase and the currency of money. Cambridge Law Journal, 55(3), 547–565. https://www.cambridge.org/core/journals/cambridge-law-journal/article/abs/bona-fide-purchase-and-the-currency-of-money/86BB50EE4BCAC58064A13BA07C3F1969
Kahn, C. M., & Roberds, W. (2009). Why pay? An introduction to payments economics. Journal of Financial Intermediation, 18(1), 1–23. https://www.sciencedirect.com/science/article/abs/pii/S1042957308000533
Levmore, S. (1987). Variety and uniformity in the treatment of the good-faith purchaser. Journal of Legal Studies, 16(1), 43–65. https://www.journals.uchicago.edu/doi/abs/10.1086/467823
Schwartz, A., & Scott, R. E. (2011). Rethinking the laws of good faith purchase. Columbia Law Review, 111(6), 1332. https://scholarship.law.columbia.edu/faculty_scholarship/187/