This paper develops a formal, operational definition of intrinsic value and a four-source taxonomy of its causal origins. Intrinsic value is defined as the reservation price an agent or coherent group would pay for an object under a counterfactual that closes off speculative resale of the instrument itself while preserving all exercise of the instrument’s primary function. The four sources — Desirability, Utility, Necessity, and Enforceability (D.U.N.E.) — are characterized not as disjoint categories but as extreme points of a valuation simplex, distinguished by the type of dependence through which an object enters an agent’s utility function: phenomenal, instrumental, existential, and institutional.
A completeness result establishes that these four dependence types exhaust the ways in which an object can generate direct value under the no-resale constraint. A separate result, the generativity theorem for institutional objects, shows that for objects whose physical features are null — fiat currency, titles, licenses, bearer claims — enforceability is the necessary precondition for intrinsic value, and the full D.U.N.E. profile of such objects is bootstrapped from institutional recognition.
The framework is operationalized through standard willingness-to-pay elicitation under enforced no-resale, and an identification result establishes that the source coefficients are recoverable from discrete choice experiments with orthogonal attribute variation. The paper positions itself relative to four prior frameworks — Moore’s non-naturalism, Kant’s price–dignity distinction, Becker–Lancaster characteristics theory, and Arrow’s impossibility theorem — and treats each with the restrictions and concessions each demands.
Keywords: intrinsic value, no-resale counterfactual, D.U.N.E. taxonomy, institutional objects, generativity theorem, reservation price, willingness to pay, source identification, value theory
Intrinsic value is among the oldest and most persistently contested concepts in philosophy and economics. The term has been used to refer to a property of objects considered in themselves, a relation between objects and agents, a discounted stream of future monetary returns, a non-naturalistic ethical property irreducible to any descriptive fact, and a synonym for direct as opposed to speculative valuation. These usages are not merely different dialects of the same concept; they are different concepts sharing a name. The result is that when philosophers, economists, financial analysts, and institutional theorists invoke intrinsic value, they often talk past one another while believing themselves to be addressing a shared question.
This paper does not attempt to reconcile all prior usages. Such an attempt would either be so inclusive as to dissolve the concept entirely or so restrictive as to exclude most of the traditions that use the term. Instead, the paper develops one specific construct, defines it with precision, shows that it is operational and measurable, establishes its formal properties, and positions it relative to the most important rival frameworks while acknowledging openly where it departs from them.
The construct developed here is the following. The intrinsic value of an object to an agent (or to a coherent group of agents) is the maximum amount of money the agent would willingly sacrifice to obtain that object under a counterfactual in which the object cannot be resold for speculative gain, but can still be used for its intended purpose, redeemed for its stated claim, exercised as its constituted function, or consumed in the manner appropriate to its nature. This is not the value the object has in itself, in the Moorean sense; it is not the dignity of rational agency, in the Kantian sense; it is not the discounted cash flow of a security, in the finance sense; and it is not utility in the unqualified sense of neoclassical microeconomics. It is a specific, agent-relative, operational construct, and the decision to use the term “intrinsic value” for it reflects a judgment that this construct best captures the intuition that motivates most uses of the term in practice — that there is some value an object has to a person that is independent of any prospect of passing it on at a profit.
Given that definition, the paper makes three principal claims.
First, intrinsic value in this sense is well-defined, exists under standard behavioral assumptions, is unique under strict monotonicity, and can be elicited empirically through willingness-to-pay mechanisms that enforce the no-resale counterfactual. This claim is defended in Chapters 3, 5, 6, and 8.
Second, the direct value that an object contributes to an agent’s utility function arises from exactly four types of dependence: phenomenal (the object affects the agent’s inner states), instrumental (the object enables the agent to perform actions), existential (the object preserves the agent’s conditions of continuation), and institutional (the object’s relevant properties are constituted by collective recognition). These four dependence types correspond to four sources of intrinsic value — Desirability, Utility, Necessity, and Enforceability — which form an exhaustive basis rather than a disjoint partition. Objects do not sit in exactly one category; they admit convex combinations of source activations, and the extreme points of the resulting simplex correspond to objects for which a single source dominates. The completeness of this four-way classification is defended by a formal argument in Chapter 4 that does not depend on a category enumeration but on the structural features of how an object can enter an agent’s utility function at all.
Third, for a class of objects whose physical features are null — fiat currency, bearer claims, titles, licenses, and the full range of what may be called institutional objects — enforceability is the necessary precondition for intrinsic value, and the remaining sources are bootstrapped from institutional recognition. This is a non-trivial claim. It asserts that institutional objects do not have intrinsic value in spite of lacking physical features; they have intrinsic value because institutional recognition activates downstream desirability, utility, and necessity components that would otherwise be dormant. The claim is defended in Chapter 7 through a generativity theorem and is consistent with, but more general than, chartalist theories of money.
These three claims together define the scope of the paper. The paper is not an attempt to revolutionize economic theory; it is an attempt to specify, with sufficient precision to be usable, a concept that has been handled imprecisely for a long time, and to show that the specification has consequences that were not previously visible. The principal audience is anyone who has needed to invoke intrinsic value in a formal argument and has found existing definitions either metaphysically obscure, disciplinarily parochial, or operationally inert.
Chapter 2 of this paper engages directly with four frameworks that bear on the concept: Moore’s non-naturalism in Principia Ethica (1903), Kant’s distinction between Preis and Würde in the Groundwork (1785), the Becker–Lancaster characteristics theory of household production (Lancaster 1966; Becker 1965), and Arrow’s impossibility theorem on social aggregation (Arrow 1951, 1963). Each of these is treated not as a failed predecessor but as a commitment that presses on the framework developed here and that must be addressed on its own terms. Moore’s non-naturalism is conceded rather than refuted: the construct developed here is agent-relative and naturalistic, and Moore would reject the whole operationalization project for principled reasons. Kant’s distinction is accommodated rather than contested: the framework operates within the domain of things with price and makes no claim about dignity-bearers. Becker–Lancaster is complemented rather than replaced: the four sources developed here supply a causal taxonomy for the characteristics that Becker–Lancaster parameterizes. Arrow’s theorem is respected rather than circumvented: the group-level claims of this framework are conditional on domain restrictions under which Arrovian aggregation is known to be consistent. These positions are defended in detail in Chapter 2.
Chapter 2 treats the four priors. Chapter 3 states the no-resale definition of intrinsic value and clarifies the counterfactual that the definition invokes, with particular attention to distinguishing speculative resale (excluded) from exercise of primary function (preserved). Chapter 4 develops the D.U.N.E. taxonomy, presents the four sources as dependence types, and proves the completeness of the taxonomy by a structural argument. Chapter 5 states five axioms sufficient for the formal results and explains which standard rationality assumptions are taken as background. Chapter 6 proves the main theorems: existence, uniqueness, D.U.N.E. decomposition, basis non-reducibility, necessity threshold asymptotic, enforceability jump, and group aggregation under domain restriction. Chapter 7 develops the generativity theorem for institutional objects, which is the paper’s most specific and consequential result. Chapter 8 treats empirical measurement, identifies the core empirical obstacle (source correlation in naturally occurring data), and establishes an identification result for the source coefficients under orthogonal experimental variation. Chapter 9 concludes. Appendix A provides the experimental protocols and econometric specifications in full detail for replication purposes.
A formal definition of intrinsic value does not enter a vacuum. Four prior frameworks bear directly on any such definition, and each makes commitments that any new proposal must address explicitly. This chapter treats them in order: Moore’s non-naturalism, Kant’s distinction between price and dignity, the Becker–Lancaster characteristics theory, and Arrow’s impossibility theorem on social aggregation. The aim is not to claim that each of these frameworks failed where the present one succeeds. The aim is to identify what each prior framework is actually doing, show how the framework developed in this paper relates to it, and concede or accommodate rather than dismiss the objections each would raise.
G.E. Moore’s Principia Ethica (1903) is the twentieth-century locus classicus for the claim that intrinsic value is a non-natural, irreducible, simple property. For Moore, to ask whether something has intrinsic value is to ask whether the universe would be better or worse for its existence, considered in complete isolation from any relations it bears to anything else. The test for intrinsic value is the method of isolation: imagine the object existing alone in an otherwise empty universe, and ask whether its existence would be good. If the answer is yes, the object has intrinsic value; if the answer is indifferent or negative, it does not.
Two features of Moore’s conception matter for any formal framework. First, Moorean intrinsic value is ontologically objective. It is a property of the object, not a relation between the object and any particular valuer. Different observers may have different epistemic access to the property, but the property itself does not depend on who is observing. Second, Moorean intrinsic value is non-natural. Moore’s famous open question argument holds that any proposed naturalistic definition of the form “intrinsic value is X” — where X is some descriptive property such as pleasure, preference-satisfaction, evolutionary fitness, or willingness to pay — can always be meaningfully questioned (“but is X really good?”), and Moore takes this to show that intrinsic value cannot be reduced to any such descriptive property. The open question is always open; therefore the proposed reduction is always incomplete.
The framework developed in this paper departs from Moore on both points, and the departure should be stated openly rather than hidden behind a claim to have resolved Moore’s concerns. The construct defined here is agent-relative: it is a relation between an object and a coherent group, not a property of the object alone. The construct is also naturalistic: it is defined in terms of measurable willingness-to-pay under a specific counterfactual, and the measurement procedure produces a real number that can be compared, aggregated, and estimated statistically.
Moore’s response to this framework would be predictable and, on its own terms, correct. He would say that what is being measured is not intrinsic value but merely what agents are willing to sacrifice for things, which conflates evaluative reasoning with prudential reasoning. He would point out that an agent’s willingness to pay may be high for objects that lack intrinsic value (luxury goods that merely signal status) and low for objects that possess it (the continued existence of a beautiful landscape that the agent happens never to visit). He would conclude that the framework, whatever its merits as a theory of prudential valuation, has nothing to say about intrinsic value in his sense.
The honest response is to concede the point. Moorean intrinsic value and the intrinsic value defined here are not rivals; they are different concepts addressing different questions. Moore’s question is whether the world contains objects or states of affairs whose existence makes the world better considered sub specie aeternitatis. That is an ethical and metaphysical question, and the present framework is silent about it. The present framework’s question is how much an agent would sacrifice for an object under a counterfactual that isolates direct value from exchange value. That is an economic and decision-theoretic question, and the framework provides a precise answer. Both questions are worth asking. They are not the same question, and the same term cannot serve both without confusion.
The decision to use the term “intrinsic value” for the construct developed here is a terminological judgment, not a claim to have captured Moore’s concept. The term is used because it best names the intuition being formalized: the intuition that an object can be valuable to an agent for what it is, not merely for what it can be traded for. That intuition is present in ordinary language, in financial practice (where “intrinsic value” typically means direct rather than speculative valuation), and in institutional economics. None of these usages is Moorean. The framework developed here serves these usages and does not serve Moore’s.
Kant’s Groundwork of the Metaphysics of Morals (1785) draws a distinction that has a structural parallel to the one this paper develops, though the content differs. Kant distinguishes between Preis (price) and Würde (dignity). Things have price insofar as they can be substituted for other things of equivalent value; any object with a price has its place in a system of commensurable exchange, in which it is possible to say that one good is worth two of another. Rational agents have dignity insofar as they are ends in themselves, not means to any further end, and cannot be substituted for anything else. Dignity, for Kant, is not very high price. It is a categorically different kind of value that cannot be placed on the same scale as price at all.
Kant’s framework is silent about the valuation of things with price. Once an object has been sorted onto the price side of the Preis/Würde distinction, Kant offers no further analysis of how to think about its price. The Groundwork is a work of moral philosophy; it is concerned with what makes rational agency an unconditional end, not with what determines the relative prices of goods. The framework developed in this paper operates entirely within the price region of Kant’s distinction. It does not claim to apply to rational agents as ends in themselves, does not attempt to measure the dignity of persons, and makes no claim about whether there exist objects with value beyond commensurable exchange.
There is therefore no conflict between this framework and Kantian ethics. The two operate at different levels of a Kantian architecture. A committed Kantian can consistently adopt the framework developed here as an account of how to think about the price component of the Preis/Würde distinction, while retaining without revision all Kantian commitments about the categorical imperative, the inviolability of rational agency, and the unconditional value of persons. Nothing in the framework pushes against the dignity side of Kant’s distinction, and nothing in the framework requires the price side to encompass everything that has value.
One point of productive contact deserves note. Kant’s distinction is itself a source-taxonomic claim: it identifies a type of value (dignity) that cannot be reduced to price. In the framework developed here, the closest structural analog is the claim that enforceability is a source of intrinsic value that cannot be reduced to desirability, utility, or necessity. The two claims differ in content — Kant is distinguishing ethical value from economic value, while this paper is distinguishing four sources of economic value — but they share the structural commitment that certain categories of value cannot be reduced to others without loss. Both frameworks, in other words, are anti-reductionist about value. The anti-reductionism is defended on different grounds and for different purposes, but the commitment is the same.
Lancaster’s characteristics theory (1966) and Becker’s household production framework (1965) share a common strategic move: shift utility from goods themselves to their underlying attributes or to the outputs of household production processes. In Lancaster, a meal is valued not directly but through its characteristics — nutrition, taste, satiety, social context. In Becker, a meal is an input to the household production of health, pleasure, time allocation, and relational goods. Both frameworks have been empirically successful; they underlie hedonic regression, discrete choice modeling, and a large body of applied work on the valuation of non-market goods. The empirical methods developed in Chapter 8 of this paper depend on these frameworks and could not be constructed without them.
It is important to be clear about how the D.U.N.E. taxonomy relates to Becker–Lancaster, because superficially the two look like rivals. Both propose that utility is best understood as arising from something more granular than goods. The rivalry dissolves, however, when one attends to the level of abstraction each framework operates at. Becker–Lancaster is a modeling strategy. It says: parameterize preferences over underlying attributes rather than over goods, and doing so will give you a richer and more estimable model. It does not say what attributes to use, how to classify them, or why they enter the utility function. Those choices are left to the modeler.
The D.U.N.E. taxonomy operates at a higher level of abstraction. It is not a parameterization. It is a causal classification of what type of dependence a characteristic induces between object and agent. The question “what characteristics does this object have?” is a Becker–Lancaster question. The question “through what channel does each characteristic enter the agent’s utility function?” is a D.U.N.E. question. The two questions are complementary, and the answers combine: a Lancaster model with characteristics tagged by D.U.N.E. source is strictly more informative than a Lancaster model without such tagging, because it tells the researcher not only that a characteristic enters utility but why.
Concretely, a hedonic regression of housing prices on attributes might include square footage, number of bedrooms, school district quality, proximity to public transit, and legal zoning classification. Each of these is a characteristic in Lancaster’s sense. Under the D.U.N.E. taxonomy, square footage and bedroom count enter through utility (they enable the agent to perform domestic activities), school district quality enters through a mixture of utility and desirability (it provides an instrumental benefit and a status signal), proximity to public transit enters through utility (it enables commuting), and legal zoning classification enters through enforceability (it determines what can be built or conducted on the property). The D.U.N.E. classification does not replace the hedonic regression; it supplements it with a causal interpretation of each coefficient. That interpretation matters because it tells the researcher which coefficients should respond to shocks of which type: a zoning change will move the enforceability coefficient, a demographic shift will move the utility coefficient, a change in cultural norms will move the desirability coefficient, and none of these is substitutable for any other.
The relation between the two frameworks can therefore be summarized: Becker–Lancaster is silent on why characteristics matter, and D.U.N.E. supplies the why. Nothing in D.U.N.E. requires the researcher to abandon the Lancaster parameterization; the opposite is true. The empirical methods of Chapter 8 assume the Lancaster decomposition as their starting point and use D.U.N.E. only to classify the resulting attributes. A researcher committed to Becker–Lancaster loses nothing by adopting D.U.N.E., and a researcher committed to D.U.N.E. has every reason to retain Becker–Lancaster as the parameterization.
Arrow’s impossibility theorem (1951, 1963) states that no social welfare function can simultaneously satisfy universal domain, Pareto efficiency, independence of irrelevant alternatives, and non-dictatorship when there are at least three alternatives and at least two agents with unrestricted preferences. The theorem is one of the foundational results of social choice theory and places a sharp constraint on any framework that attempts to define group-level valuations as a function of individual preferences. Any framework that handles groups must either accept the constraint and work within the space of possibilities it leaves, or abandon one of the axioms that generates the impossibility. This framework does the former.
The original draft of this material contained an axiom asserting that every group admits a coherent aggregation rule mapping individual utilities into a group utility. That axiom is inconsistent with Arrow’s theorem on a universal domain and should not have been stated as a background assumption. The correct treatment, adopted in the present version, is to restrict the domain of the framework to groups for which aggregation is known to be consistent. The phrase “coherent group” throughout this paper refers to a group whose valuations can be aggregated in a manner that escapes Arrow’s impossibility through one of three standard routes.
The first route is domain restriction. If the group’s members have preferences that satisfy single-peakedness, value-restrictedness, or a similar structural condition, the aggregation problem becomes tractable because the conditions under which Arrow’s theorem bites are not met. Black’s median voter theorem (1948) and its extensions give sufficient conditions for well-defined group preferences in such cases. Coherent groups under domain restriction are therefore groups whose members happen to have preferences sufficiently aligned that social aggregation is consistent.
The second route is interpersonal utility comparability. Arrow’s theorem depends on the assumption that utilities cannot be meaningfully compared across agents. If one is willing to assume comparability — as Harsanyi (1955) did in his aggregation theorem, and as utilitarian welfare economics routinely does — then weighted utilitarian sums provide a well-defined social welfare function that satisfies Pareto, non-dictatorship, and a form of independence. Coherent groups under this route are groups for which the analyst accepts the comparability assumption on principled grounds.
The third route is institutional constitution. Many groups in the real world do not have preferences at all; they have procedures. A firm has a CEO whose decisions stand in for the firm’s preferences. A family may have a designated decision-maker for certain domains. A government has legislative and executive structures that produce binding collective decisions. In these cases, the group’s valuation of an object is not defined by aggregating individual valuations; it is defined by the output of the procedure that the group uses to produce its decisions. Coherent groups under this route are institutionally constituted entities whose decisions are produced by recognized procedures, and the “group’s intrinsic value” of an object is the reservation price the procedure would produce under the no-resale counterfactual.
The framework’s claims about group intrinsic value are conditional on one of these three restrictions being satisfied. When none of them is satisfied, the framework does not assign a well-defined group intrinsic value, and it does not claim to. This is a genuine restriction of the framework’s domain of application, and it is the honest way to respect Arrow’s theorem. The alternative — asserting an aggregation rule by axiom and hoping nobody notices — is not available to a paper that claims mathematical seriousness.
A last point should be made about the direction of the restriction. Group intrinsic value in this framework is well-defined for a large class of practically important groups: firms, households with coordinated decision-making, institutionally structured organizations, and any group whose members happen to have domain-restricted preferences over the object under consideration. It is not well-defined for heterogeneous publics with unrestricted preferences and no aggregation procedure. For such publics, the framework recommends disaggregation: estimate individual intrinsic values, report the distribution, and let the reader decide whether and how to aggregate. That recommendation is consistent with current best practice in welfare economics, and it is the most that a framework respecting Arrow can honestly offer.
This chapter presents the central definition of the paper. Before stating it formally, it is necessary to be precise about the counterfactual that the definition invokes, because the most common confusion about the no-resale constraint arises from conflating two things that the constraint treats differently: speculative resale of the instrument itself, and exercise of the instrument’s primary function. The first is excluded by the constraint; the second is preserved. A reader who collapses these two ends up with a definition under which fiat currency has zero intrinsic value, which is absurd on the framework’s own terms, because fiat currency is the clearest case in the world of an object whose value comes entirely from direct, non-speculative sources once the institutional backing is in place. The failure mode is avoidable, but only if the counterfactual is stated with care.
The counterfactual invoked by the no-resale constraint is the following. Fix an object x and an agent (or coherent group) g. Imagine a possible world in which g acquires x with the understanding that x cannot subsequently be resold, traded, transferred for gain, or otherwise disposed of for any consideration greater than zero. In this counterfactual world, g retains x permanently, or until the object is consumed, exercised, redeemed, or otherwise exhausted in the course of its ordinary use. The question the framework asks is: what is the maximum amount of money g would willingly pay to acquire x in this counterfactual world?
Two features of the counterfactual deserve emphasis. First, the counterfactual does not remove g’s ability to use x in the manner that x is intended to be used. A bond can still be held to maturity, even though it cannot be resold on a secondary market. A dollar bill can still be spent at a grocery store, even though it cannot be sold to a currency collector. A work of art can still be hung on a wall and contemplated, even though it cannot be auctioned. A pharmaceutical can still be consumed, even though it cannot be diverted to a black market. The no-resale constraint removes speculative and capital-gains-type motivations for acquisition while preserving every form of engagement with the object that constitutes its intended function.
Second, the counterfactual does not remove the ordinary monetary character of money used in the transaction to acquire x. The price p that g pays for x is paid in money, and that money retains its ordinary status in the counterfactual — it is the medium of exchange, it is enforceable legal tender, it is the unit of account. The no-resale constraint applies to x, the object being valued, not to the money used to value it. If the no-resale constraint were applied to money as well, the framework would become incoherent, because there would be no meaningful way to talk about what g would pay for x in a world where money itself has no value. The counterfactual should be read as local to x: hold everything else about g’s economic environment fixed, close off secondary-market resale of x, and ask how much of g’s ordinary monetary endowment g would surrender to obtain x.
A natural question is whether this counterfactual is well-defined in cases where the primary function of x is itself a form of exchange. Money is the canonical case. If the dollar bill is an object whose primary function is to be spent, and spending is a form of transfer, has the no-resale constraint not excluded the only thing that makes money valuable? The answer is no, and the reason is important. Spending money is not reselling it. Spending is the exercise of money’s primary function as a medium of exchange, and the medium-of-exchange function is fully preserved under the counterfactual. The speculative activity that the constraint excludes is not spending; it is the activity of acquiring money for the sake of a later sale of the same money at a profit, as a currency speculator might do. Under the counterfactual, g can still spend the money, pay taxes with it, use it as a unit of account, and receive it as payment; g simply cannot flip it as a speculative instrument. That restriction removes nothing that ordinary holders of money care about, which is why fiat currency retains full direct value under the constraint.
Let X denote the set of objects, services, and claims that can be valued. Let G denote the set of coherent groups (individuals are treated as singleton groups). For g in G, let mg denote the group’s monetary endowment, and let ∅ denote the status quo in which g does not acquire the object under consideration.
For each g and each x in X, define the no-resale utility function Ugnr(x, m) as the utility of holding x together with remaining monetary endowment m, evaluated under the counterfactual in which x cannot be resold for gain but can be exercised in its primary function. The superscript nr is used to distinguish this utility from ordinary utility functions that include speculative possibilities; readers familiar with the distinction between holding-period utility and full utility in finance may recognize the structure. The no-resale utility exists whenever g has coherent preferences over the bundle (x, m) that do not depend on any anticipated resale of x.
The intrinsic value of x to g, denoted IVg(x), is the supremum of the set of prices p ≥ 0 such that Ugnr(x, mg − p) ≥ Ugnr(∅, mg).
The definition says: the intrinsic value of x to g is the most money g would part with, from the no-resale counterfactual perspective, to obtain x rather than to remain in the status quo. Prices at which g is weakly better off acquiring x at price p belong to the set; the reservation price is the supremum of that set. Under the conditions stated in the axioms of Chapter 5, the supremum exists and is unique.
It is worth pausing to note that this definition is not new in isolation. The Becker–DeGroot–Marschak mechanism (Becker, DeGroot, and Marschak 1964) was designed precisely to elicit such reservation prices without strategic incentive to misreport, and the mechanism has been used since then in thousands of experimental studies. The Vickrey second-price auction (Vickrey 1961) achieves a similar objective under different incentive conditions. Contingent valuation methods in environmental economics (Mitchell and Carson 1989) use reservation price elicitation under explicit or implicit no-resale conditions. What is new here is not the mechanism but the conceptual move of treating the output of the mechanism as the definition of intrinsic value rather than as a measurement trick for estimating utility. The framework commits to the view that the reservation price under no-resale is not a proxy for intrinsic value; it is intrinsic value, in the sense that this paper is using the term.
The remaining work of this chapter is to enumerate, for avoidance of doubt, the specific types of motivation the no-resale constraint excludes and the specific types of motivation it preserves. This enumeration is important because the most serious objections to the framework turn on misreadings of which category a given motivation falls into.
The constraint excludes the following. First, it excludes speculative acquisition with the intent of resale at a higher price. An art buyer who acquires a painting only in the expectation of selling it for a gain in five years is motivated entirely by exclusions under the constraint. Second, it excludes acquisition as a store of value where the store is intended to be liquidated at a later date for consumption. A hoarder of physical gold whose entire motivation is to convert the gold back into money during a retirement period is excluded on the same grounds. Third, it excludes portfolio motivations, hedging, and diversification that rely on rebalancing through sale of the instrument. Fourth, it excludes any form of arbitrage or cross-market transfer. Fifth, it excludes anticipated resale to descendants, which is a form of transfer for consideration (the consideration being non-monetary affection but still a transfer in the economic sense).
The constraint preserves the following. First, it preserves consumption of the object in the manner appropriate to its kind. Food is eaten, pharmaceuticals are taken, books are read, paintings are contemplated. Second, it preserves the exercise of any primary function the object is constituted to perform. A bond’s coupon is collected and the principal is received at maturity; a stock’s dividends are received but the share is not sold; a license is used to perform the activity it authorizes; a patent is practiced but not sold or cross-licensed. Third, it preserves the medium-of-exchange function of money, including spending at retail, payment of taxes, settlement of debts, and receipt of income. Fourth, it preserves the unit-of-account and store-of-value functions of money insofar as those functions are instrumental to ordinary economic activity rather than speculative rebalancing. Fifth, it preserves the enforcement of institutional rights: a title holder can occupy the property, exclude trespassers, and make use of the land even though the title cannot be sold; a license holder can practice the authorized activity even though the license cannot be transferred for a fee.
The distinction between excluded and preserved motivations is clean in the canonical cases and admits edge cases in the peripheral ones. The framework does not claim to eliminate all edge cases; it claims that for the central applications of the concept — pricing of direct use value, valuation of institutional objects, assessment of goods with no active secondary market — the distinction is sufficiently clean to be operative. A corner case that is genuinely ambiguous under the framework is probably a corner case that is ambiguous in the underlying economic phenomenon as well, and a definition cannot be expected to resolve ambiguities that live in the phenomenon itself.
One implication of the distinction is worth stating explicitly because it addresses the most common misreading of the framework. Under the no-resale constraint, fiat currency retains its full value. The reason is that every ordinary use of money — spending at retail, settling debts, paying taxes, receiving wages — is an exercise of money’s primary function as a medium of exchange, not a form of speculative resale. A dollar bill under the counterfactual is still legal tender, is still accepted at the grocery store, is still required to be accepted in settlement of public and private debts. The only activity closed off by the counterfactual is the activity of treating dollars themselves as a speculative instrument to be bought and sold at a profit, which is not what ordinary holders of money do and not what motivates ordinary demand for money. The intrinsic value of fiat currency under the no-resale definition is therefore approximately equal to its ordinary value under full monetary functionality, which is what any reasonable definition of intrinsic value should yield for fiat currency.
The definition in Chapter 3 tells us what intrinsic value is but not where it comes from. An agent may be willing to pay a large amount for an object under the no-resale counterfactual, and the definition will register that willingness as a large intrinsic value, but the definition is silent about why. Answering the why is the work of this chapter. The claim of the chapter is that direct value under the no-resale constraint arises from exactly four types of dependence between the object and the agent, which correspond to four sources of intrinsic value: Desirability, Utility, Necessity, and Enforceability. These four sources are abbreviated D.U.N.E.
Before developing the taxonomy, it is important to be clear about what kind of claim the four-way classification makes. It is not a claim that objects can be sorted into four disjoint boxes. Most real objects have multiple sources active simultaneously, and for any given object the sources interact in ways that resist clean separation. Medicine is necessary and useful and often institutionally regulated. Wedding rings are desirable and institutionally recognized. Houses are useful and desirable and institutionally titled and, under certain conditions, necessary. The taxonomy is not a partition; it is a basis. The four sources are the extreme points of a space, and real objects occupy the interior of the simplex, with source intensities that can be any combination of the four corners.
The distinction between a partition and a basis matters because the standard objection to any four-way taxonomy of value — and the objection that would destroy this one if it were made to stand as a partition — is that the categories leak into each other. The aesthetic pleasure of looking at art is, in a loose sense, a functional capability of the art object: it does something, namely produce experiences in the viewer. The status-signaling value of a luxury watch is, in a loose sense, a utility: it performs the function of signaling status. A skeptic of the four-way classification can run this move on every boundary: desirability reduces to a kind of utility (utility of subjective experience), necessity reduces to utility with a steep marginal curve, enforceability reduces to a kind of utility (utility of legal recognition). If the taxonomy is a partition, every one of these reductions is fatal, because each shows that the supposed categories are not disjoint.
If the taxonomy is a basis, the reductions are not fatal. A basis is allowed to have elements that interact. What is required of a basis is not that its elements be disjoint but that they be linearly independent — that no element of the basis can be expressed as a function of the others — and that together they span the space. The D.U.N.E. taxonomy makes these two claims. First, each of the four sources is irreducible to the other three: one cannot express desirability as a function of utility, necessity, and enforceability without loss, and analogously for each other source. Second, the four together are sufficient: any direct value under the no-resale constraint can be expressed as some combination of the four sources, with no residue.
The skeptic’s observation that medicine is simultaneously necessary, useful, and institutionally regulated becomes a feature of the framework rather than a bug. Medicine is an object whose source vector has non-zero components on three of the four sources, and a complete account of medicine’s intrinsic value must track all three. The source vector of medicine might be something like (0.1, 0.4, 0.4, 0.1), with small components on desirability and enforceability and large components on utility and necessity. The source vector of a wedding ring might be (0.7, 0.05, 0.0, 0.25), dominated by desirability with a significant enforceability component (the ring functions as a socially recognized symbol of commitment, and the recognition is partly institutional). The source vector of a dollar bill is (0.2, 0.4, 0.2, 0.2) when the institutional structure is stable, with the important qualification — developed in Chapter 7 — that the three non-E components are bootstrapped from the E component for an institutional object with null physical features. Every object that has intrinsic value has a source vector, and the vector sits in the interior of the four-dimensional simplex unless a single source dominates so thoroughly that the object is effectively located at one of the extreme points.
The extreme points of the simplex are the clean cases that justify the basis claim. An object for which the source vector is (1, 0, 0, 0) — pure desirability, no utility, no necessity, no enforceability — is an object whose entire intrinsic value arises from its phenomenal impact on the agent and nothing else. A purely sentimental heirloom with no market, no function, no legal status, and no survival relevance is approximately such an object. An object for which the source vector is (0, 1, 0, 0) is a pure utility object: a tool that performs a function, with no aesthetic dimension, no necessity, and no legal status. A hand tool used in a trade is approximately such an object. The extreme points for necessity and enforceability are illustrated by basic survival goods in scarcity (food, water, medicine) and by purely institutional objects (titles, licenses, certificates) respectively.
The linguistic test for identifying the dominant source of an object is surprisingly reliable at the extreme points. When a person is asked why they want food in a survival situation, they do not say “because it is useful.” They say “because I need it.” When a person is asked why they want a painting they have inherited from a deceased parent, they do not say “because I need it” or “because it is useful.” They say “because I love it” or “because it reminds me of her.” When a person is asked why they want a particular tool, they say “because I can do such-and-such a job with it.” When a person is asked why they want a professional license, they say “because I am required to have it to practice.” These are not random choices of words. They are reports of which source dominates the person’s direct valuation, and ordinary language picks up the distinction with near-perfect reliability. A framework that cannot recover the linguistic distinction between need, love, use, and authorization is missing something that ordinary speakers have no difficulty tracking.
The basis claim needs more than linguistic intuition to rest on. What makes the four sources distinct is not that they are labeled with different words but that they correspond to structurally different types of dependence between the object and the agent. Each source is characterized by what the object does to or for the agent, and the four types of doing are ontologically distinct: they involve different features of the agent responding to different features of the object.
The source of direct value arising from dependence on the agent’s subjective phenomenal apparatus. An object contributes to an agent’s intrinsic value through desirability when it affects the agent’s inner states — qualia, experiences, emotions, aesthetic responses, symbolic resonances, sentimental associations — without requiring the agent to perform any instrumental action and without reference to the agent’s survival or institutional embedding.
Desirability is the source active when an agent values an object for the experience of having or perceiving it. Art, music, literature, beautiful landscapes, heirlooms with sentimental value, objects associated with loved ones or important memories, religious symbols, cultural artifacts, and decorative objects all have desirability as a principal source of their intrinsic value. The invariance that characterizes desirability is that a D-valued object retains its value under changes to the agent’s action set: if you would love the painting, you would love it whether or not you had any further opportunity to do anything about your love. D-value responds to what the agent is phenomenally, not what the agent can do or whether the agent continues to exist or whether the agent is institutionally embedded.
The source of direct value arising from dependence on the agent’s capacity for instrumental action. An object contributes to an agent’s intrinsic value through utility when it enables the agent to perform actions or achieve ends that the agent could not otherwise perform or achieve, or could perform or achieve only at greater cost.
Utility is the source active when an agent values an object for what the object lets the agent do. Tools, machines, vehicles, appliances, infrastructure, raw materials, information used to inform decisions, and capabilities of all kinds have utility as a principal source of their intrinsic value. The invariance that characterizes utility is that a U-valued object retains its value under changes to the agent’s phenomenal responses: a hammer drives nails whether or not the user finds the hammer beautiful or meaningful. U-value responds to what the agent can do, not what the agent experiences.
The source of direct value arising from dependence on the agent’s biological, physiological, or existential continuation. An object contributes to an agent’s intrinsic value through necessity when its presence is required for the agent (or the coherent group) to continue to exist, function, or maintain the conditions of existence.
Necessity is the source active when an agent values an object because the agent cannot survive, function, or persist without it. Food, water, air, shelter, medicine, safety goods, and under certain conditions energy, heat, and basic sanitation all have necessity as a principal source of their intrinsic value. The invariance that characterizes necessity is that an N-valued object retains its value under changes to both the agent’s action set and the agent’s phenomenal responses: food is necessary whether or not the agent enjoys it, whether or not the agent can do other things with it. N-value responds to whether the agent continues to exist, not to what the agent experiences or what the agent can do.
It is worth addressing here the reductionist move that would collapse necessity into utility with a hyperconvex marginal curve. The move says: necessity is not a separate source; it is utility with a function fN that happens to blow up near a threshold. If any fs can have a threshold, there is no taxonomic reason to privilege N. The response to this move is that threshold behavior is a mathematical signature of N-dominance, not the definition of N. The definition of N is the dependence type — the object is valued because the agent’s continuation depends on it. Threshold behavior is a consequence of this dependence type, because the conditions of existence are not a matter of degree: below a certain level of food or water or oxygen, the agent does not continue to exist, and above that level the agent does. The threshold is in the biology, not in the utility function. A gasoline utility curve can look hyperconvex near empty when a driver is running out of fuel on a long trip, but no one would say “I need gasoline the way I need water,” because the home-getting that gasoline enables is not an existential condition. The gasoline’s apparent necessity is inherited from something else — the need to be home — and the analysis of the gasoline’s intrinsic value should locate the N-component in the home-getting, not in the gasoline. Threshold behavior is neither necessary nor sufficient for N-dominance: one can have threshold behavior in U without N involvement, and one can have N-dominance without dramatic threshold behavior if the conditions of existence are satisfied at a level far above the threshold.
The source of direct value arising from dependence on the agent’s embedding in an institutional structure. An object contributes to an agent’s intrinsic value through enforceability when its relevant economic, legal, or social properties are constituted by institutional recognition — by rules, authorities, or collective practices that determine what the object is, what rights it confers, what obligations it creates, and what the agent is entitled to do with it.
Enforceability is the source active when an agent values an object because the object’s properties are constituted by institutional recognition. Legal tender, securities, titles, deeds, licenses, patents, contracts, enforceable claims, and under certain readings also citizenship and citizenship-dependent rights have enforceability as a principal source of their intrinsic value. The invariance that characterizes enforceability is that an E-valued object retains its value under changes to the agent’s phenomenal responses and action capacities but loses its value entirely under changes to the relevant institutional structure: a deed is valuable regardless of whether the holder likes the property or can physically improve it, but the deed ceases to have value entirely if the registrar’s office no longer recognizes it. E-value responds to the agent’s position in an institutional web, not to inner states, capacities, or existence.
The fourth source is the one most commonly overlooked in prior value theories, and it is the source whose addition most distinguishes the D.U.N.E. taxonomy from its predecessors. Classical economics and neoclassical utility theory have typically folded institutional value into utility by treating enforcement as a constraint on feasible actions rather than as a source of direct value. This folding obscures the distinctive dependence type of E: institutional objects do not have utility in the ordinary sense, because they do not enable an agent to perform a physical action that would otherwise be infeasible. They modify what counts as authorized or recognized action within a collective practice. The holder of a patent is not physically enabled by the patent to do something that would otherwise be impossible; the patent modifies the legal status of the holder’s activity relative to other potential practitioners. Collapsing E into U loses this distinctive structure, and the loss becomes acute when one tries to understand institutional objects whose physical features are null, which is the topic of Chapter 7.
The claim that these four sources are complete — that any direct value under the no-resale constraint arises from some combination of D, U, N, and E, with no residual source — needs a proof. The proof cannot be carried out by enumeration, because enumeration of categories is circular: any list of four categories can be defended by defining the fourth to include whatever the first three do not. What is needed instead is a structural argument operating on the space of possible dependence types.
The argument is the following. Consider an arbitrary agent g with a no-resale utility function Ugnr and an arbitrary object x that g values. By hypothesis, Ugnr(x, m) > Ugnr(∅, m) for some range of m, which is the condition under which x has positive intrinsic value to g. The question is: through what channel does x enter Ugnr? The answer must appeal to some feature of the agent that responds to some feature of the object. The channels available are given by the structural features of the agent that can respond to external objects under the no-resale counterfactual.
The structural features of an agent are these. An agent has phenomenal states (qualia, experiences, inner life). An agent has causal powers (capacities to act on the world). An agent has conditions of existence (biological and existential requirements for continued being). And an agent, if embedded in a society, has an institutional position (a set of rights, obligations, and recognitions conferred by collective practices). These four features are the features that can respond to external objects. There are no others that are structurally distinct.
The basis claim is that each of the four sources corresponds to dependence of Ugnr on x through exactly one of these four features. Desirability is the dependence of Ugnr on x mediated by x’s effect on g’s phenomenal states. Utility is the dependence mediated by x’s effect on g’s causal powers. Necessity is the dependence mediated by x’s effect on g’s conditions of existence. Enforceability is the dependence mediated by x’s effect on g’s institutional position.
Let g be an agent with a coherent no-resale utility function Ugnr, and let x be an object such that IVg(x) > 0. Then the dependence of Ugnr(x, m) on x can be expressed as a combination of four components, each corresponding to one of the four dependence types — phenomenal (D), instrumental (U), existential (N), or institutional (E) — and no residual component exists that is not reducible to some combination of these four.
Proof. The argument proceeds by exhaustion of dependence channels. Suppose there exists some fifth channel through which x enters Ugnr — call it the residual channel. This channel cannot be phenomenal, because the phenomenal channel is D. It cannot be instrumental, because the instrumental channel is U. It cannot be existential, because the existential channel is N. And it cannot be institutional, because the institutional channel is E. The residual channel must therefore correspond to some feature of the agent that is neither phenomenal, nor causal-active, nor existential, nor institutional. But these four features exhaust the structural categories of agency under which an external object can make a difference to the agent that the agent’s utility function can track. An object that makes no difference to any of these four features makes no difference to the agent at all, which contradicts the hypothesis that x has positive intrinsic value to g. Therefore no residual channel exists, and the four sources are complete. ∎
Two candidate residual sources deserve explicit discussion, because they are the ones that tend to be proposed when the completeness claim is challenged. The first is “relational value” — value arising from the agent’s relationships with other agents. Does not friendship, or community, or love constitute a fifth source of intrinsic value not captured by D, U, N, or E? The answer is that relational value is real but is not a fifth source; it decomposes into combinations of the four. The phenomenal pleasure of a friend’s company is D. The instrumental assistance a friend provides is U. The existential support a friend provides in times of crisis is N. The institutional recognition a formal relationship provides (marriage, partnership, shared legal status) is E. A relationship is a complex object whose source vector has components on all four, and the full analysis of a relationship’s intrinsic value tracks all four components separately. Nothing is lost by the D.U.N.E. analysis; what is sometimes called relational value is simply the integrated sum of the four components as they apply to relational objects.
The second candidate residual source is “informational value” — value arising from the possession of knowledge, data, or understanding. Is information a fifth source? Again, the answer is no. Information has D-value when it produces intellectual satisfaction or wonder or the phenomenal pleasure of understanding. Information has U-value when it enables the agent to make better decisions or perform actions that would be infeasible without it. Information has N-value when it is required for survival (medical information, hazard warnings, safety protocols). Information has E-value when its possession is institutionally constituted (a credential, a certification, a legally recognized qualification). Every case of informational value decomposes into some combination of the four, and no residual remains.
A more sophisticated challenge would propose “option value” — the value of flexibility or of preserving future possibilities — as a residual source. Option value in the strict sense of financial theory depends on future price variability and is excluded by the no-resale constraint. Option value in a broader sense, meaning the value of preserving future action possibilities regardless of future prices, is a form of utility: it is the instrumental value of future capacities. The D.U.N.E. framework handles option value by tracking the instrumental component separately from the phenomenal component and recognizing that an object may have utility value today because it enables action tomorrow. The framework does not need a fifth source to handle this case.
The completeness argument is structural rather than categorical, and that is the reason it works. Prior attempts at completeness for value taxonomies have failed because they proceeded by listing categories and daring the reader to propose a fifth, which is an argumentative form that cannot win. The completeness of D.U.N.E. rests instead on a claim about the structural features of agency: there are exactly four ways an external object can make a difference to an agent (through phenomenal states, causal powers, conditions of existence, and institutional position), and these four exhaust the space because they correspond to four disjoint structural features of the agent itself. The proof is no stronger than the claim that these four features exhaust the structural features of agency, but that claim is plausible independently, and is the right place to locate the argumentative burden.
The second condition for a basis, beyond completeness, is linear independence: no source is expressible as a function of the others. This condition is easier to defend than completeness, because it can be established by counterexample. For each pair of sources, one can exhibit an object whose value in one source is positive and whose value in the others is zero. Such objects serve as evidence that no reduction is possible.
A pure desirability object: a sentimental photograph of a deceased family member. The photograph has no functional utility (it performs no task), no necessity value (the agent does not require it for survival), and no enforceable status (possession confers no institutional rights). Its intrinsic value is entirely phenomenal: it induces emotional responses in the agent. Any framework that lacks D as a distinct source cannot assign positive value to this object.
A pure utility object: a commercial tool used in the agent’s trade, fungible and replaceable, bearing no sentimental meaning. The tool has no desirability value (the agent cares only that it works), no necessity value (the agent can survive without it), and no enforceable status (its possession confers no rights). Its intrinsic value is entirely instrumental: it enables the agent to perform a task. Any framework that lacks U as a distinct source cannot assign positive value to this object.
A pure necessity object: clean drinking water in a survival situation where there is no market for it, no aesthetic dimension, and no institutional claim on it. The water has no desirability value (the agent does not care how the water tastes or looks), no marginal utility beyond the survival threshold, and no enforceable status. Its intrinsic value is entirely existential: it keeps the agent alive. Any framework that lacks N as a distinct source cannot assign positive value above the threshold to this object.
A pure enforceability object: a notarized deed to an undeveloped plot of land the agent will never visit, in a jurisdiction whose institutions are stable. The deed has no desirability value (the agent has no emotional connection to the land), no utility value (the agent does not use the land), and no necessity value (the agent does not require the land to survive). Its intrinsic value is entirely institutional: the agent holds a recognized legal claim that can be exercised if desired, inherited by descendants, or simply counted as part of the agent’s institutional position. Any framework that lacks E as a distinct source cannot assign positive value to this object.
No source fs is expressible as a function of the other three sources alone. The four sources are linearly independent elements of the valuation basis.
Proof. For each source s ∈ {D, U, N, E}, there exists an object xs whose source vector is (0, …, zs, …, 0) with positive weight on source s and zero weight on every other source. The four examples just given (sentimental photograph, commercial tool, survival-threshold water, undeveloped-land deed) provide such objects. A framework that attempted to express source s as a function of the other three would assign zero intrinsic value to xs, since the other three are zero for xs. But xs has positive intrinsic value by construction. Therefore no such reduction exists, and the four sources are independent. ∎
Non-reducibility has a practical implication that will matter in Chapter 8. When an empirical researcher attempts to recover the coefficients of a D.U.N.E. model from data on willingness-to-pay, the researcher must ensure that the data contain variation in each source independently of the others. A dataset in which the four source intensities are perfectly correlated — for example, a dataset consisting entirely of medicines, which tend to have simultaneously high U, high N, and high E — does not permit separate identification of the coefficients. The non-reducibility result guarantees that the coefficients are in principle distinct and recoverable, but the recovery requires variation in the data, which the researcher must deliberately engineer or carefully locate in the natural world. This is an identification problem, and it is addressed formally in Chapter 8.
The definition and taxonomy developed in Chapters 3 and 4 require a small number of axioms to yield the theorems of Chapter 6. This chapter states five axioms and explains which other assumptions are treated as standard background rather than as load-bearing components of the framework.
The distinction between load-bearing axioms and background assumptions matters because axiomatic apparatus tends to inflate when it is not disciplined. The original version of this framework contained ten numbered axioms, most of which restated standard rationality conditions (completeness, transitivity, monotonicity, continuity) that are assumed throughout decision theory and economics without separate naming. Listing these as numbered axioms creates the impression that the framework depends on them in a special way, which is misleading; it depends on them in the ordinary way that any economic framework does. The present chapter separates the two categories.
The following conditions are assumed without separate numbering. They are the standard assumptions of decision theory under certainty and would be required by any framework in this space. A reader familiar with Debreu (1959) or Mas-Colell, Whinston, and Green (1995) will recognize them as the usual starting points.
Coherence of preferences. For each g ∈ G, preferences over bundles of the form (x, m) where x ∈ X and m ∈ ℝ+ are complete (every pair of bundles is comparable), transitive (if a ≻ b and b ≻ c then a ≻ c), and reflexive. These conditions are necessary for the existence of a utility representation and are background to everything that follows.
Continuity of preferences. For each g ∈ G and each x ∈ X, the function m ↦ Ugnr(x, m) is continuous in m on its domain. Continuity is necessary for the supremum in the intrinsic value definition to be approachable and for the existence results of Chapter 6 to go through.
Monotonicity in money. For each g ∈ G and each x ∈ X, Ugnr(x, m) is weakly increasing in m. More money is weakly preferred to less, holding the rest of the bundle fixed. This condition is required for the reservation price to be well-defined: without monotonicity in money, there may be no price p such that paying p makes the agent indifferent to the status quo.
Non-negativity of endowments and prices. mg ≥ 0 for all g, and prices p under consideration lie in the interval [0, mg]. This is a feasibility condition rather than a substantive assumption.
These four background conditions are not the substance of the framework. Any framework in decision theory or welfare economics would require them. They are listed here only so that a reader can identify what is standard machinery and what is specific to the D.U.N.E. framework.
The following five axioms are the substantive commitments of the framework. They are the ones that do work in the theorems of Chapter 6 and that distinguish D.U.N.E. from other decision-theoretic frameworks.
For all g ∈ G and all x ∈ X, the utility function Ugnr(x, m) is defined over the counterfactual in which x cannot be resold for gain but can be used in its primary function. The no-resale counterfactual is operative throughout the framework, and all utilities, reservation prices, and intrinsic values are computed under it.
This is the defining axiom of the framework. It is the axiom that makes intrinsic value distinct from ordinary market value or speculative value. The counterfactual it invokes was discussed at length in Chapter 3. The axiom does not merely state that no-resale is a constraint on the agent; it states that the entire framework operates within the counterfactual in which the constraint holds, so that questions about intrinsic value are questions about behavior under the counterfactual rather than behavior in actual markets.
For all g ∈ G, the no-resale utility function Ugnr(x, m) depends on x only through x’s source vector z(x) = (zD(x), zU(x), zN(x), zE(x)) and on the agent’s monetary endowment m. Formally, there exists a function Fg such that Ugnr(x, m) = Fg(zD(x), zU(x), zN(x), zE(x), m).
This axiom says that the only features of x that matter for the agent’s direct-value utility are the four source intensities. It is a strong assumption, but it is the assumption that makes the D.U.N.E. decomposition do work. Without it, the source vector z(x) would be descriptively useful but would not exhaust the determinants of utility; there could be features of x that affect utility without being captured by any of the four sources. Direct-value separability rules this out by definition: the four sources are, by axiom, the only features of x that matter for direct value.
The defense of this axiom is not that it is trivially true. It is that the completeness argument of Chapter 4 makes it defensible. If the four sources exhaust the structural features of agency through which an object can enter an agent’s utility function, then any feature of x that matters to the agent must matter through one of the four sources, and the axiom follows. The axiom is therefore a formal statement of the completeness result: what Chapter 4 argued on structural grounds, Axiom 2 states as a commitment of the formal framework.
The four sources D, U, N, and E are exhaustive: any direct-value motivation of g for x under the no-resale counterfactual can be expressed as a contribution to one of the four sources. No residual source exists.
This axiom is what Theorem 4.1 proved informally; Axiom 3 states the conclusion as a formal commitment. The axiom is needed explicitly because the completeness result in Chapter 4 rests on a structural argument that is persuasive but not airtight in the way a formal proof from more primitive assumptions would be. By stating completeness as an axiom, the framework makes its dependence on the four-way classification explicit, and any challenge to the framework can be directed to this axiom rather than to the derived theorems.
For each source s ∈ {D, U, N, E}, if zs(x′) ≥ zs(x) and zs′(x′) = zs′(x) for all s′ ≠ s, then Ugnr(x′, m) ≥ Ugnr(x, m). Increasing a source intensity while holding the other sources fixed weakly increases direct-value utility.
This is a monotonicity condition specific to the source decomposition. It says that each source makes a weakly positive contribution to utility, which is what it means for D, U, N, and E to be sources of value rather than merely features of objects. The axiom is needed for the theorems on cross-source comparative statics and for the identification results of Chapter 8.
For each g ∈ G, the enforceability component zE(x) of the source vector is non-zero if and only if x has recognized status in an institutional structure operative for g. The value of zE(x) depends on the strength, stability, and scope of the institutional recognition.
This is the axiom that grounds enforceability in actual institutional structures rather than in the agent’s private beliefs about institutional structures. An agent cannot generate E-value by privately believing that an object is institutionally recognized; the recognition must exist in the institutional structure that is operative for the agent. The axiom is necessary to avoid trivializing the E source, which would otherwise become a free parameter that agents could assign arbitrarily to any object. The axiom matters especially for Chapter 7, where the institutional constitution of objects plays a central role.
These five axioms are the load-bearing substance of the framework. The theorems of Chapter 6 can be derived from these five together with the background assumptions listed in Section 5.1. The original ten-axiom version of this framework repeated conditions that were either already implied by the background assumptions (monotonicity in money, continuity) or that were more appropriately stated as special conditions of particular theorems (non-satiation for necessity, which is a condition on fN rather than a general axiom about all objects). The present version separates background from substance and names only the commitments that do real work.
This chapter proves the main results of the framework. The theorems divide into three groups: foundational results on existence and uniqueness of intrinsic value, structural results on the D.U.N.E. decomposition, and behavioral results on how intrinsic value responds to changes in source intensities and institutional state. Throughout the chapter, the background assumptions of Section 5.1 and the five axioms of Section 5.2 are in force unless noted otherwise.
Under the background assumptions of coherence, continuity, and monotonicity in money, and under Axiom 1 (No-Resale), the intrinsic value IVg(x) exists for all g ∈ G and all x ∈ X.
Proof. Let S = { p ≥ 0 : Ugnr(x, mg − p) ≥ Ugnr(∅, mg) } denote the set of prices at which g weakly prefers acquiring x to the status quo. S is non-empty because p = 0 belongs to S whenever x is weakly preferred to ∅ at the full endowment mg, and p = 0 belongs to S trivially when x is indifferent to ∅. S is bounded above by mg, since any price p > mg is infeasible: the agent would have negative remaining monetary endowment, which lies outside the feasibility region. Monotonicity in money implies that as p increases from 0 to mg, Ugnr(x, mg − p) is weakly decreasing. Continuity of Ugnr in its second argument implies that the condition Ugnr(x, mg − p) ≥ Ugnr(∅, mg) defines an interval of prices. The supremum of this interval exists because S is non-empty and bounded, and continuity ensures the supremum is approachable in the sense required by the reservation-price definition. Therefore IVg(x) = sup S exists and is finite. ∎
Under the additional condition that Ugnr(x, m) is strictly decreasing in p over the relevant range (equivalently, strictly increasing in m), the intrinsic value IVg(x) is unique.
Proof. Strict monotonicity in m implies that the function p ↦ Ugnr(x, mg − p) is strictly decreasing in p. Consequently, the equation Ugnr(x, mg − p) = Ugnr(∅, mg) has at most one solution, and if such a solution exists it is the unique reservation price. Combined with existence from Theorem 6.1, the reservation price is unique. ∎
Under Axioms 2 and 3 (direct-value separability and source completeness), any intrinsic value IVg(x) can be expressed as a function of the source vector z(x) and the monetary endowment mg. There exists a group-specific function Ψg such that IVg(x) = Ψg(zD(x), zU(x), zN(x), zE(x), mg).
Proof. By Axiom 2, Ugnr(x, m) = Fg(zD(x), zU(x), zN(x), zE(x), m) for some function Fg. By Axiom 3, no features of x other than its source vector enter Ugnr. The intrinsic value IVg(x) is defined as the supremum of the set of prices at which Ugnr(x, mg − p) ≥ Ugnr(∅, mg), which by Axiom 2 equals the supremum of the set of prices at which Fg(zD(x), zU(x), zN(x), zE(x), mg − p) ≥ Fg(0, 0, 0, 0, mg), since the status quo ∅ corresponds to zero source activations. The supremum depends only on the source vector and on mg. Define Ψg as the function that maps (zD, zU, zN, zE, mg) to this supremum. Then IVg(x) = Ψg(zD(x), zU(x), zN(x), zE(x), mg). ∎
The practical consequence of Theorem 6.3 is that the intrinsic value of an object depends on the object only through its source vector. Two objects with identical source vectors have identical intrinsic values to the same agent, regardless of any other features they might have. This is a strong equivalence result, and it is the justification for empirical work that estimates intrinsic value by measuring source intensities and recovering the group-specific function Ψg.
No source function fs is expressible as a function of the other three source functions alone. The four sources are linearly independent elements of the valuation basis in the sense that the framework cannot be collapsed to a three-source framework without losing the ability to assign intrinsic value to objects whose source vector has positive mass on exactly one source.
Proof. For each source s ∈ {D, U, N, E}, Section 4.4 exhibits an object xs whose source vector has positive weight on s and zero weight on every other source. The photograph (pure D), commercial tool (pure U), survival-threshold water (pure N), and undeveloped-land deed (pure E) serve as witnesses. Under Theorem 6.3, the intrinsic value of each xs depends only on its source vector. If some source s were expressible as a function of the other three, then Ψg applied to the source vector of xs would yield the same value as Ψg applied to the source vector (0, 0, 0, 0), since the zero components on the other three sources would force the expressed source to also be zero. But the intrinsic value of xs is strictly positive by construction, whereas Ψg(0, 0, 0, 0, mg) is zero (since the agent is indifferent between the status quo and an object with null source activation). This contradiction rules out the expressibility, and the four sources are therefore non-reducible. ∎
Under Axiom 4 (source monotonicity), IVg is weakly increasing in each source intensity. That is, if zs(x′) ≥ zs(x) and zs′(x′) = zs′(x) for all s′ ≠ s, then IVg(x′) ≥ IVg(x).
Proof. Axiom 4 states that Ugnr is weakly increasing in each zs holding the others fixed. By Theorem 6.3, IVg(x) is the reservation price at which g is indifferent between x and ∅ under the no-resale utility. Increasing zs weakly increases Ugnr(x, m) for each m, which weakly increases the set of prices at which Ugnr(x, mg − p) ≥ Ugnr(∅, mg), which weakly increases the supremum of that set. Therefore IVg(x′) ≥ IVg(x). ∎
If x weakly dominates y in all four sources (zs(x) ≥ zs(y) for all s), then IVg(x) ≥ IVg(y) for any coherent group g.
Proof. Iterate Theorem 6.5 across all four sources. Start with y and increase zD to zD(x); the intrinsic value weakly increases. Then increase zU to zU(x); the intrinsic value weakly increases again. Repeat for zN and zE. The final object has source vector equal to z(x), and its intrinsic value is at least as large as IVg(y). ∎
The intuition behind the necessity threshold result is that necessity sources exhibit a qualitatively different behavior near the threshold of existential survival than the other three sources. The formal statement requires specifying the behavior of fN near the threshold.
Suppose that for a given agent g, the necessity component of the no-resale utility function has a survival threshold at zN = τ > 0, meaning that Ugnr(x, m) → −∞ as zN(x) → τ− for any fixed m ≥ 0. Suppose further that fN has polynomial blow-up in the sense that there exist constants c > 0 and α > 0 such that fN(zN) ~ c · (τ − zN)−α as zN → τ−. Then as zN(x) → τ−, the intrinsic value IVg(x) approaches mg at an asymptotic rate: mg − IVg(x) ~ C · (τ − zN(x))α for some constant C depending on g and on the other source intensities of x.
Proof. Near the threshold, the necessity component of the utility function dominates the other source components, because fN(zN) → ∞ as zN → τ−. The reservation price condition Ugnr(x, mg − p) = Ugnr(∅, mg) reduces, in the limit, to requiring the necessity component of Ugnr(x, mg − p) to match the necessity component of Ugnr(∅, mg). For x in the threshold regime, the necessity component of Ugnr(x, mg − p) is approximately c · (τ − zN(x))−α for any feasible mg − p. The status quo ∅ has zero necessity activation, so the necessity component of Ugnr(∅, mg) is the finite value fN(0). Matching the two requires the monetary term to absorb the imbalance, which forces mg − p → 0 at a rate controlled by α. Specifically, the monetary term in Fg has finite marginal utility, and the necessity component diverges at rate (τ − zN)−α; equating the two magnitudes yields mg − p ~ (τ − zN)α up to constants. Therefore p ~ mg and mg − IVg(x) ~ C · (τ − zN(x))α. ∎
The result gives a sharp approach rate rather than the mere statement that intrinsic value blows up near the threshold. The approach rate depends on the exponent α, which is a property of the agent’s necessity function rather than of the object alone. Agents with more sharply blowing-up necessity functions (smaller α) pay closer to their full endowment at equal distances from the threshold, while agents with more gradually blowing-up necessity functions (larger α) retain more of their endowment at the same distances. The result is empirically testable in survival-scarcity experiments and provides a handle for estimating α from data.
Let x be an object whose enforceability intensity zE(x) is a step function of institutional state: zE(x) = 0 when the institutional structure θ is in one regime and zE(x) = k > 0 when θ is in another regime. Then the intrinsic value IVg(x) is a discontinuous function of θ, jumping from a lower value (or zero, for objects whose other source components are zero) to a strictly higher value at the regime transition.
Proof. By Axiom 5, zE is non-zero only when institutional recognition is operative. Under the regime where institutional recognition is absent, zE(x) = 0 and the intrinsic value of x is determined by zD, zU, zN only. Under the regime where institutional recognition is operative, zE(x) = k > 0 and by Axiom 4, Ugnr weakly increases. Strict monotonicity of Fg in zE (which follows from Axiom 4 combined with the non-zero contribution of zE to direct value) implies that the increase in Ugnr is strict whenever k > 0. The reservation price correspondingly jumps, because the set of prices at which Ugnr(x, mg − p) ≥ Ugnr(∅, mg) expands discontinuously when zE(x) jumps from 0 to k. Therefore IVg(x) jumps discontinuously, and the jump is strictly positive. ∎
The enforceability jump result is the first hint of the generativity theorem that Chapter 7 will develop. It establishes that institutional state matters to intrinsic value in a way that is qualitatively different from how physical features of objects matter: physical features can be changed by continuous perturbations, but institutional state can change discontinuously, and the corresponding change in intrinsic value is equally discontinuous. This has implications for any analysis of monetary regimes, legal reform, and institutional transitions, and it distinguishes the framework from approaches that treat institutions as smoothly varying constraints rather than as discrete state variables.
For a coherent group g in the sense of Chapter 2 (domain-restricted, interpersonally comparable, or institutionally constituted), the group intrinsic value IVg(x) exists and satisfies mini IVi(x) ≤ IVg(x) ≤ maxi IVi(x), where the min and max are over individual members of g.
Proof. The proof is conditional on the route by which g escapes Arrow’s impossibility. Under domain restriction (single-peakedness or value-restriction), the median voter theorem or its extensions guarantee that group preferences are well-defined, and the aggregation function is monotone in individual preferences. A monotone aggregation of individual reservation prices must yield a group reservation price that lies between the minimum and maximum individual reservation prices. Under interpersonal comparability, weighted utilitarian aggregation produces a group utility function whose reservation price lies in the convex hull of individual reservation prices, which is bounded by the min and max. Under institutional constitution, the group’s reservation price is determined by the institutional procedure, and if the procedure aggregates individual inputs in any monotone way (which is standard for committee rules, voting procedures, and executive decision rules), the output lies between the min and max of the inputs. In all three routes, the bound holds. ∎
The group aggregation result is deliberately weak. It establishes that group intrinsic value exists and is bounded by individual values, which is enough for most applications, without attempting to say more than Arrow’s theorem permits. Stronger claims — that group intrinsic value equals a specific weighted average, or that it can be recovered from observed group choices in a parameterized form — require additional assumptions that the framework does not make. A researcher who wishes to make such stronger claims should explicitly state the additional assumptions and derive them within their chosen route around Arrow.
This chapter develops the paper’s most consequential result. The result concerns a class of objects that have long posed difficulties for theories of value: objects whose physical features are null or negligible, and whose value therefore cannot be explained by reference to any material property the object possesses in itself. Fiat currency is the canonical case. A paper bill has no nutritional content, no mechanical capability, no aesthetic property of any consequence, no surgical utility, and no direct physical impact on anyone’s conditions of existence. Yet a dollar bill is valuable to its holder, often very valuable, and a framework that claims to define intrinsic value must be able to account for this fact. The account given in this chapter is that the value of fiat currency, like the value of every other purely institutional object, is generated by enforceability, with the other three sources cascading downstream from the institutional recognition.
The generativity claim is stronger than the simple observation that fiat currency has E-value. Any framework that acknowledges enforceability as a source of value will say that fiat currency has E-value. The stronger claim of this chapter is that for institutional objects, E is the necessary precondition for the activation of any of the other sources. Without institutional recognition, fiat currency has no D-value, no U-value, and no N-value. With institutional recognition, all three of those sources can be activated downstream, and the full D.U.N.E. profile of fiat currency becomes active. E does not merely sit alongside the other sources as a fourth component; it generates them for institutional objects in a way that it does not for physical objects. This chapter develops that claim carefully, proves it as a theorem, and traces its implications for the theory of money.
The distinction between physical and institutional objects is not sharp, because most real objects have both physical and institutional components. A car has physical features (engine, chassis, wheels, fuel tank) and institutional features (registration, title, insurance, authorized inspection status). The framework handles mixed objects by assigning them mixed source vectors: the physical features contribute to D, U, and N through material properties, and the institutional features contribute to E through collective recognition. The two contributions combine in Ψg to yield a total intrinsic value. For mixed objects, the generativity claim of this chapter does not apply in its pure form; the non-E sources have independent grounding in physical features and do not need to be generated by institutional recognition.
The generativity claim applies to objects whose physical features are null, in the sense that removing institutional recognition would leave an object whose remaining intrinsic value is approximately zero. Fiat currency is the clearest case: a dollar bill without legal tender status is a piece of paper with no nutritional, functional, or material value to speak of. Titles and deeds are another clear case: a land deed without a recognizing registrar is an ink-on-paper document whose physical features contribute nothing to intrinsic value. Licenses, certificates, patents, bearer bonds, and digital monetary instruments fit the same pattern. For these objects, the physical substrate is a placeholder for institutional recognition; the intrinsic value lives entirely in the institutional relation the object constitutes, not in the material properties of the substrate.
An object x is institutional if its economically relevant features are constituted by institutional recognition — that is, if in the absence of institutional recognition, the source vector of x satisfies zD(x) ≈ 0, zU(x) ≈ 0, zN(x) ≈ 0, and zE(x) = 0. An object is physical if it has non-negligible source intensities on at least one of D, U, or N that are independent of institutional recognition. An object is mixed if it has both.
The definition is deliberately relative to institutional recognition. A dollar bill is an institutional object not because the paper it is printed on has no physical properties but because the paper’s physical properties do not contribute to the dollar’s intrinsic value. The same quantity of paper printed with a different design and carrying no legal tender status would not have the intrinsic value of a dollar bill, even though it is physically indistinguishable. The economic relevance of the paper is entirely mediated by the institutional recognition of the dollar bill’s status, which is what makes it institutional by the framework’s definition.
A useful diagnostic for identifying institutional objects is the substrate test. Ask: if one were to replace the physical substrate of the object with a different substrate that preserved the institutional recognition (electronic record instead of paper, cryptographic hash instead of ink), would the intrinsic value be preserved? For institutional objects, the answer is yes: dollar bills have been printed on cotton paper, issued as electronic reserves, and in principle could be issued as blockchain entries, with the intrinsic value preserved across substrates as long as institutional recognition is preserved. For physical objects, the answer is no: a car with a digital replacement for its engine block is not a car. The substrate test separates objects whose value is constituted by recognition from objects whose value is constituted by material properties.
Let x be an institutional object in the sense of Definition 7.1. Then: (i) IVg(x) > 0 only if zE(x) > 0. (ii) When zE(x) > 0, institutional recognition can activate non-zero values of zD, zU, and zN through the downstream consequences of recognition, yielding a full D.U.N.E. profile for x even though x has null physical features. (iii) The intrinsic value of x is generated by the cascade E → {D, U, N}, with enforceability as the necessary bootstrapping source and the remaining sources as derived consequences.
Proof. Claim (i). By Definition 7.1, an institutional object has zD(x) ≈ 0, zU(x) ≈ 0, and zN(x) ≈ 0 in the absence of institutional recognition. The only source that can contribute positively to IVg(x) in that case is zE. If zE(x) = 0 as well, then by Theorem 6.3 the source vector is (0, 0, 0, 0) and IVg(x) = Ψg(0, 0, 0, 0, mg), which equals zero because the agent is indifferent between the status quo and an object with null source vector. Therefore IVg(x) > 0 implies zE(x) > 0 for an institutional object. Claim (ii). When zE(x) > 0, institutional recognition creates authorized uses, obligations, and privileges that the holder can exercise. These authorized activities have consequences for the holder’s phenomenal states (status, social recognition, confidence: D-activation), instrumental action capacity (the ability to transact, to access institutionally-gated services, to discharge obligations: U-activation), and conditions of existence in a monetized or institutionally-structured society (the ability to pay for necessities, to access healthcare, to remain employed: N-activation). The downstream source activations are consequences of the institutional recognition; they are not independent. For fiat currency, for example, the U-value of being able to transact exists only because institutional recognition makes the currency acceptable in transactions; remove the recognition and the transactional capability vanishes. Claim (iii) follows from the combination of (i) and (ii): E is necessary (by (i)) for any positive intrinsic value, and the non-E sources are consequences of E (by (ii)) rather than independent contributors, so the cascade structure is E → {D, U, N}. ∎
The generativity theorem is the pivot of the chapter. It says that for institutional objects, the four-source decomposition still applies, but with a specific causal structure: one of the sources is necessary and generates the others, rather than four sources sitting in parallel as independent contributors. This structural asymmetry between E and the other three sources for institutional objects is not present for physical objects, where D, U, and N can be activated independently by material properties of the object. The asymmetry is not a weakness of the taxonomy; it is a feature of the subject matter. Institutional objects are objects whose value is constituted by collective recognition, and it should not be surprising that the source corresponding to collective recognition plays a privileged role in their valuation.
A useful way to see the asymmetry is to consider a thought experiment. Imagine that institutional recognition of fiat currency were suddenly removed — not just weakened, but removed entirely, by a collective withdrawal of acceptance. A dollar bill in this counterfactual world is still the same piece of paper, with the same printed design, the same physical weight, and the same historical provenance. None of its physical features has changed. Yet its intrinsic value to any holder has collapsed to approximately zero, because the downstream source activations that depended on institutional recognition (the ability to transact, to pay obligations, to access institutionally-gated goods) have all been removed. The physical features of the paper do not save the bill’s value because the physical features contribute nothing that was not already dependent on institutional recognition for its activation. This thought experiment is the clearest intuitive argument for the generativity theorem, and the formal proof given above is the rigorous version of the same argument.
A further implication is worth stating. The generativity theorem establishes that for institutional objects, the stability and durability of institutional recognition is the fundamental determinant of intrinsic value. Physical objects retain most of their value under institutional collapse because their D, U, and N sources continue to be activated by material properties. Institutional objects do not retain value under institutional collapse because all four sources are dependent on recognition that has been withdrawn. This gives a precise meaning to the intuition that fiat currency is “only as good as” the institution backing it, and it identifies the backing as the necessary bootstrapping source rather than as a parallel contributor.
For an institutional object x, the intrinsic value IVg(x) is a discontinuous function of the institutional regime θ that governs the recognition of x. Specifically, IVg(x) jumps at regime transitions that activate or deactivate the institutional recognition of x, even when all physical features of x remain constant across the transition.
Proof. This is an immediate corollary of the generativity theorem combined with the enforceability jump theorem from Chapter 6. Institutional regimes θ that activate recognition yield zE(x) > 0 and, via the cascade structure of Theorem 7.1, positive downstream activations of zD, zU, and zN. Regimes that deactivate recognition yield zE(x) = 0 and, via the same cascade, zero activations of the other sources for institutional objects. The transition from one regime to the other is discontinuous in the institutional state space, and by Theorem 6.8 the intrinsic value jumps at the transition. The magnitude of the jump is the full intrinsic value of the object under recognition, since the value under non-recognition is approximately zero for institutional objects. ∎
Regime dependence has consequences for every analytical task that requires comparing the intrinsic value of institutional objects across time or across institutional environments. The value of a dollar bill in the United States in 2026 cannot be compared directly to the value of a physically identical dollar bill in a hypothetical post-recognition environment, because the two live under different institutional regimes. Historical comparisons of the value of obsolete currencies are systematically misleading if they assume smooth functional dependence on physical features; the relevant variable is the institutional regime under which the currency was held, and the transition from recognition to non-recognition is a discontinuous event whose timing determines the trajectory of the currency’s value. Analysts who have worked on the dollarization of currencies in high-inflation economies, on the valuation of pre-euro European currencies after the euro’s introduction, or on the status of currencies in politically unstable regimes will recognize the regime-dependence phenomenon as the formal statement of intuitions they have operated with informally.
The generativity theorem is consistent with chartalist theories of money and provides a more general framework in which chartalism is a special case. Chartalism, in its various forms, holds that money is constituted by the authority that issues and accepts it — traditionally the state, though chartalism can be extended to any institution capable of enforcing acceptance. Knapp’s State Theory of Money (1905) and Innes’s essays on the credit theory of money (1913, 1914) are the historical roots of the view, and modern monetary theory developed by Wray, Mitchell, and others has brought chartalist ideas back into contemporary debate. The common thread across chartalist traditions is that money’s value depends on institutional acceptance, typically operationalized through the state’s willingness to accept the money in payment of taxes.
The framework developed here agrees with chartalism on the central claim that fiat currency’s value is constituted by institutional recognition, but it differs from chartalism in two ways. First, the framework is more general: it applies to institutional objects beyond currency (titles, deeds, licenses, patents, certificates) and to mixed objects (cars, houses, intellectual property) that have both physical and institutional components. Chartalism is typically formulated as a theory of money specifically, whereas the generativity theorem applies to any institutional object. Second, the framework explicitly treats the downstream source activations as cascade products rather than as independent parallel channels. Chartalism typically stops at the observation that institutional acceptance matters; the generativity theorem goes further, specifying that acceptance bootstraps downstream D, U, and N components through a specific causal structure.
The relation between the framework and chartalism is therefore one of subsumption rather than rivalry. A committed chartalist can adopt the framework without abandoning any substantive commitment; the framework provides a formal vocabulary and a more general structure in which chartalism’s core claims appear as instances. Conversely, a framework user who is interested specifically in monetary questions can recover chartalism by specializing the generativity theorem to objects whose institutional recognition is constituted by state authority and whose downstream sources are limited to transactional utility and tax-payment capacity.
One consequence of this relation is that the framework can be used to analyze monetary regimes that do not fit neatly into state-centric chartalism. Decentralized monetary instruments, algorithmic stablecoins, community currencies, and multi-issuer monetary systems are objects whose institutional recognition is distributed or non-state, and chartalist analyses of such objects have had to stretch the definition of “state” or “authority” to accommodate them. The generativity theorem does not depend on any particular source of institutional recognition; it depends only on the fact that recognition exists and is operative for the agent. This allows the framework to handle non-state monetary instruments symmetrically with state-issued currency, which is a generalization that chartalism typically achieves only with difficulty.
The relation also matters for physical objects that acquire institutional components. A house is primarily a physical object with U and D sources from material features, but it also has an E component from the title deed. The framework handles this mixed case by assigning the house a source vector with non-zero components on all four sources, with the E component generated institutionally and the D, U, N components generated partly physically and partly institutionally. A chartalist analysis of a house would have to contort itself to fit the mixed nature of the object; the D.U.N.E. framework handles it naturally by recognizing that physical and institutional sources coexist in the source vector of a single object without collapsing into each other.
The operational character of the framework depends on the ability to measure intrinsic value and to recover the source decomposition from data. This chapter develops the empirical methods that realize that operational character. The methods are of three kinds: elicitation mechanisms for individual reservation prices under no-resale (Section 8.1), the identification problem that prevents naive regression from recovering source coefficients (Section 8.2), and an identification result for discrete choice experiments with orthogonal variation (Section 8.3). Additional protocols and econometric templates are provided in Appendix A.
Four elicitation methods are suitable for measuring intrinsic value under the no-resale constraint. Each has a long history in experimental and environmental economics, and each yields reservation prices that, under the no-resale counterfactual, coincide with intrinsic value as defined in Chapter 3.
The BDM mechanism (Becker, DeGroot, and Marschak 1964) elicits a participant’s maximum willingness-to-pay for an object without giving the participant any incentive to misreport. The participant states a bid; a random price is drawn from a distribution with known support; the participant purchases the object if and only if the bid exceeds the drawn price, and pays the drawn price in that event. The mechanism is incentive-compatible in the sense that truth-telling (bidding one’s true maximum willingness-to-pay) weakly dominates any other strategy under standard preference assumptions. In the framework of this paper, a BDM mechanism with enforced no-resale (participants are instructed and contractually committed that they cannot resell the object) elicits the reservation price under the no-resale counterfactual, which by definition equals IVg(x) for the participant g and the object x.
The Vickrey auction (Vickrey 1961) is an alternative incentive-compatible mechanism that elicits true willingness-to-pay. Each participant submits a sealed bid for the object; the highest bidder wins and pays the second-highest bid. Truth-telling is weakly dominant for the same class of preference structures as BDM. When the auction is conducted with enforced no-resale, the winning bids measure IVg for the winners, and the full bid distribution provides a lower bound on IVg for the other participants (each losing bid is a reservation price at which the participant was unwilling to acquire the object, which places the bid at or below the participant’s true IV).
Contingent valuation (Mitchell and Carson 1989) uses direct survey questioning to elicit willingness-to-pay for non-market goods, goods with no active secondary market, or hypothetical goods that do not yet exist. The canonical question is of the form: “what is the maximum amount of money you would be willing to pay to obtain x, given that you cannot resell, trade, or exchange it at any future date?” Contingent valuation is particularly useful for goods that cannot be delivered in a laboratory setting or for which experimental manipulation is ethically or practically infeasible. The method has well-known limitations — hypothetical bias, yea-saying, protest responses — which require careful survey design to mitigate. For purposes of this framework, contingent valuation is a legitimate elicitation method when properly conducted, and its outputs correspond to IVg under the no-resale counterfactual because the no-resale constraint is embedded explicitly in the survey question.
Discrete choice experiments (DCEs) present participants with a series of choices between bundles of attributes at varying prices. Each bundle is a hypothetical object described by its attribute vector and a stated price. The participant chooses the bundle they prefer (or opts out, if that is permitted by the design). Estimation proceeds through conditional logit or mixed logit models, which recover the marginal utility of each attribute and the price coefficient, from which reservation prices and attribute-specific willingness-to-pay can be constructed. DCEs are the most flexible method of the four because they allow the researcher to engineer variation in the attributes of the bundles, which is essential for the identification strategy discussed in Section 8.3.
A naive use of any of these elicitation methods faces an identification problem that prevents straightforward recovery of the D.U.N.E. source coefficients. The problem arises from the correlation of sources in naturally occurring objects. Consider a dataset of willingness-to-pay values for medicines. Medicines tend to have simultaneously high values of zU (they perform a therapeutic function), zN (they are required for the patient’s continuation in cases of serious illness), and zE (they are institutionally regulated, prescribed, and dispensed through authorized channels). They may also have zD components in some cases (prestige associated with brand names, placebo effects mediated by perceived quality). A regression of willingness-to-pay on the four source intensities across a dataset of medicines will find that the four columns of the design matrix are nearly collinear, and the resulting estimates of the source coefficients will be imprecise or unidentified.
The same problem arises for other classes of naturally occurring objects. Houses have correlated U, D, and E components. Cars have correlated U, D, and E components. Education credentials have correlated U, D, and E components. In none of these cases does the naturally occurring dataset contain enough variation in individual sources to separate them. The researcher who wants to estimate the D.U.N.E. coefficients must therefore either find a setting in which the sources vary independently or engineer the variation experimentally.
The identification problem is not a flaw in the framework; it is the standard problem of any structural model with multiple correlated explanatory variables. Hedonic regression on housing attributes faces the same problem: square footage and number of bedrooms are highly correlated, and separating their coefficients requires either a large dataset with enough variation to break the correlation or an experimental design that orthogonalizes the attributes. The framework inherits the standard solutions to this standard problem, and the solutions come in three forms: orthogonal experimental variation, natural experiments, and functional form restrictions.
Orthogonal experimental variation is the cleanest solution to the identification problem because it eliminates source correlation by construction. In a discrete choice experiment with orthogonal attribute variation, the researcher designs the bundles so that each attribute varies independently of the others across the set of bundles. A fractional factorial design or an orthogonal array can produce such a design with a manageable number of bundles. Under orthogonal variation, the correlation between source intensities in the design matrix is zero (or small enough to be non-problematic), and the coefficient estimates are identified and well-behaved.
Let a discrete choice experiment elicit participant choices among bundles with attribute vectors (zD, zU, zN, zE) and prices p, where the attribute vectors are designed such that the sample correlation between any pair of attributes is zero. Suppose the participant’s choice probabilities follow a conditional logit model with utility linear in attributes and price: V(z, p) = αD zD + αU zU + αN zN + αE zE − γ p. Then the coefficient vector (αD, αU, αN, αE) is identified up to a normalizing scale constant (the price coefficient γ), and the attribute-specific willingness-to-pay values αs / γ are identified without normalization.
Proof. Under orthogonal variation, the design matrix of attribute levels has full column rank with orthogonal columns, and the conditional logit likelihood is strictly concave in the coefficient vector provided that the design includes sufficient choice pairs to distinguish all four attributes. Maximum likelihood estimation of the conditional logit model produces consistent estimates of the coefficient ratios αs / γ under standard regularity conditions (independent observations, correct specification of the choice probability structure, sufficient variation in prices to identify γ). The attribute-specific willingness-to-pay values are the ratios αs / γ, which are invariant to the normalization of the coefficient vector. The unnormalized coefficients are identified up to scale, which is the standard result for conditional logit models. ∎
The identification theorem establishes that the source coefficients are recoverable from experimental data when the researcher takes control of the design. It does not guarantee recovery from observational data, and the framework does not claim that it does. Researchers working with observational data must rely on the two other identification strategies — natural experiments and functional form restrictions — and must accept the stronger assumptions that those strategies require.
When experimental control is not available, natural experiments can sometimes identify individual sources by providing exogenous variation in a single source while holding the others approximately constant. Three classes of natural experiments are particularly useful for the framework.
Institutional shocks for E-identification. A change in the legal or institutional status of an object — legalization, licensing reform, registration requirement change, legal tender designation change — provides exogenous variation in zE without corresponding variation in zD, zU, or zN. The change in intrinsic value associated with the institutional shock identifies the E-contribution, provided that the other sources can be plausibly assumed to be held constant across the shock. Event studies of legalization events (for example, the legalization of cannabis in various jurisdictions, or the recognition of new currencies in monetary reforms) can be analyzed as natural experiments for E-identification.
Scarcity shocks for N-identification. A sudden change in the scarcity of a necessity good — a drought affecting water supply, a supply chain disruption affecting food availability, a shortage of critical medicine — provides exogenous variation in zN without corresponding variation in the other sources. The change in willingness-to-pay associated with the scarcity shock identifies the N-contribution. Studies of hurricane-affected areas, wartime rationing, and humanitarian crises provide natural experiments of this kind, though ethical considerations constrain the settings in which such studies can be conducted.
Aesthetic or cultural shifts for D-identification. A change in cultural attitudes or aesthetic preferences — a shift in art market valuations, a change in status symbols, a cultural rehabilitation or devaluation of a previously obscure or notorious figure — provides variation in zD without corresponding variation in the other sources. Art market data, celebrity endorsement studies, and cultural event studies can be analyzed as natural experiments for D-identification.
Identification of U is typically harder by natural experiment because utility changes tend to be accompanied by institutional or desirability changes. The most successful U-identification strategies rely on technological shocks — the introduction of a new tool that makes an existing tool obsolete, or the improvement of a capability that the tool enables — in settings where the institutional and desirability components are plausibly held constant. Hedonic regressions on product features in competitive markets provide another route, though they face the same correlation problems that afflict all observational work.
For researchers implementing the framework, three econometric specifications cover most applications. The first is a linear parametric specification in which intrinsic value is assumed to be a linear combination of source intensities:
The linear specification is straightforward to estimate by OLS and provides clean interpretation of the coefficients as marginal contributions of each source. Its limitation is that it cannot capture non-linearities, interactions, or threshold effects.
The second is a semi-parametric specification in which intrinsic value is modeled as a smooth unknown function of the source intensities: IV = f(zD, zU, zN, zE) + ε, with f estimated by splines, kernel smoothers, or generalized additive models. The semi-parametric specification accommodates non-linearities and interactions but requires larger sample sizes and does not yield clean coefficient interpretations.
The third is a structural specification in which the source functions fD, fU, fN, fE are specified parametrically with substantive functional forms: logarithmic for utility (following standard diminishing-marginal-value assumptions), power-law for desirability, hyperbolic or power-law-with-threshold for necessity, and linear or logistic for enforceability. The structural specification requires strong functional form assumptions but yields sharp tests of the framework’s predictions and can identify the threshold parameter α of Theorem 6.7 directly from data.
A researcher choosing among specifications should weigh the tradeoff between flexibility and identification. The linear specification is best for initial exploration and for large datasets with clear variation. The semi-parametric specification is best when the researcher has no strong prior on functional form and has enough data to estimate flexibly. The structural specification is best when the researcher wants to test specific predictions of the framework (threshold behavior, enforceability jumps, source-specific asymptotics) and is willing to commit to functional forms.
This paper has developed a formal, operational definition of intrinsic value as the reservation price an agent or coherent group would pay for an object under a counterfactual that closes off speculative resale while preserving the exercise of the object’s primary function. The definition rests on the careful distinction between resale and exercise, which was the source of most prior confusion about the no-resale constraint and which deserves the explicit treatment given in Chapter 3. Under the constraint, familiar elicitation methods — BDM, Vickrey auctions, contingent valuation, discrete choice experiments — recover intrinsic value empirically, and the framework inherits the operational maturity of experimental and environmental economics.
The paper has also developed a four-source taxonomy of the causal origins of intrinsic value. Desirability, Utility, Necessity, and Enforceability correspond to four types of dependence between object and agent: phenomenal, instrumental, existential, and institutional. The four sources are not a partition of objects into disjoint boxes but a basis for a valuation simplex whose extreme points are the clean cases and whose interior is the space of mixed objects. The completeness of the taxonomy is defended by a structural argument on the space of possible dependence types, and the non-reducibility of the sources is established by counterexample.
The most consequential result of the framework is the generativity theorem for institutional objects. For objects whose physical features are null, enforceability is the necessary precondition for any intrinsic value, and the remaining three sources are cascade products of institutional recognition. This result gives a precise meaning to the intuition that fiat currency, titles, deeds, and other purely institutional objects derive their value from institutional backing, and it extends the insight beyond monetary theory to any object whose properties are constituted by collective recognition. The result is compatible with chartalist theories of money and subsumes them as a special case.
The paper has been deliberately restrained in its scope. It does not attempt to resolve 2,500 years of philosophical disputes about value. It does not claim to render prior frameworks obsolete. It engages with Moore’s non-naturalism by conceding the conceptual difference rather than pretending to refute it, with Kant’s price–dignity distinction by respecting the categorical separation, with Becker–Lancaster characteristics theory by supplying a causal taxonomy that complements rather than replaces the parameterization, and with Arrow’s impossibility theorem by accepting its constraints and restricting the framework’s group-level claims to domain-restricted cases where Arrovian impossibility does not apply. This restraint is deliberate, and it reflects the judgment that a framework that engages prior work honestly is more durable than one that overreaches.
Three directions for future work are worth noting. First, the framework is static; a dynamic extension would treat intrinsic value as a function of time and would handle intertemporal tradeoffs, discounting, and the evolution of source intensities as the object ages. Second, the framework treats objects as fixed; an extension to stochastic source vectors would allow objects whose source intensities are uncertain, which is relevant to settings where institutional recognition is unstable or where necessity thresholds are themselves probabilistic. Third, the cross-cultural dimension of the framework deserves empirical attention: the source coefficients are group-specific, and systematic differences in source weightings across cultures, institutions, and historical periods would be both theoretically interesting and empirically tractable under the framework’s elicitation methods.
The framework developed here is not the last word on intrinsic value. It is a specification of one construct among many that have been called by that name, defended carefully enough to be usable in formal work and restrained enough to engage the priors of serious readers. Whether the framework earns its place in the literature will depend on how well it travels — whether it produces results that could not have been obtained with prior tools, whether it helps clarify debates that prior tools muddied, and whether it survives the empirical and theoretical tests that serious readers will bring to it. The paper invites those tests.
This appendix provides reference protocols and econometric templates for researchers implementing the framework empirically. The protocols are designed to enforce the no-resale constraint operationally and to support the identification strategies developed in Chapter 8.
Purpose. To elicit a participant’s reservation price for an object x under the no-resale counterfactual, yielding a direct measurement of IVg(x) for the participant g.
Instructions to participants. Participants are told that they will be asked to state the maximum amount of money they would pay to acquire the object, with the understanding that if they acquire the object they will not be permitted to resell it, trade it, give it away for any consideration, or otherwise transfer it for gain. They are informed that a random price will be drawn from a stated distribution, and that they will acquire the object at the drawn price if and only if their stated maximum is at least the drawn price. They are told, honestly, that truthful reporting of their maximum willingness-to-pay is the best strategy.
No-resale enforcement. The no-resale constraint is enforced through a combination of contractual commitment (participants sign an agreement not to transfer the object) and institutional features of the experiment (the object is marked, serialized, or otherwise identifiable as having been acquired through the experiment, making subsequent sale difficult). In field settings where contractual enforcement is weak, the constraint can be reinforced by selecting objects for which secondary markets are absent or illegal.
Data structure. Each observation records participant identifier, object identifier, stated maximum willingness-to-pay, random price draw, and acquisition outcome (purchased or not). The estimated intrinsic value is the stated maximum willingness-to-pay, averaged across participants if a group estimate is desired.
Purpose. To elicit reservation prices through a second-price sealed-bid auction with enforced no-resale for the winning participant.
Instructions to participants. Participants are told that they will submit a sealed bid for the object, that the highest bidder will win and will pay the second-highest bid, and that the winning bidder may not resell the object under any circumstances. They are informed that the weakly dominant strategy is to bid their true maximum willingness-to-pay.
Data structure. Each observation records participant identifier, auction identifier, and submitted bid. The estimated intrinsic value for each participant is the submitted bid. Group-level intrinsic value can be summarized by the median, mean, or maximum of the bid distribution, depending on the aggregation rule under which the group is coherent.
Purpose. To recover the source-specific coefficients (αD, αU, αN, αE) of the intrinsic value function from participant choices over hypothetical bundles with orthogonally varied attribute levels.
Design. The researcher constructs a fractional factorial design in which each of the four source attributes takes two or three levels (for example, low, medium, high) and the price takes several levels spanning the relevant range. The factorial design is selected so that the sample correlation between any pair of attribute levels is zero across the full set of choice cards. Each choice card presents two or more bundles, each characterized by its attribute vector and price, and the participant selects the preferred bundle or opts out.
Instructions to participants. Participants are told that the bundles are hypothetical but should be evaluated as if real, that they should choose the bundle they would prefer to acquire, and that the acquisition would be subject to a no-resale condition. Hypothetical bias is a known concern with DCEs, and the researcher should employ standard mitigation techniques (cheap talk scripts, consequential framing, commitment statements).
Estimation. Estimation proceeds through conditional logit or mixed logit models, recovering the coefficient vector (αD, αU, αN, αE) and the price coefficient γ. The attribute-specific willingness-to-pay is computed as αs / γ for each source s, and the implied intrinsic value of an object with source vector z is estimated as (αD zD + αU zU + αN zN + αE zE) / γ.
Purpose. To measure the discontinuous change in intrinsic value associated with a change in institutional recognition, providing an empirical test of Theorem 6.8.
Design. Two conditions are constructed that are identical except for the institutional status of the object. In the first condition, the object lacks institutional recognition (the license is not valid, the title is not registered, the currency is not legal tender). In the second condition, the institutional recognition is present. Participants are randomly assigned to one condition, and their reservation prices are elicited using BDM, Vickrey, or contingent valuation under the respective condition.
Estimation. The enforceability jump is the difference in mean reservation prices between the two conditions: ΔIV = IV(with recognition) − IV(without recognition). Standard errors are computed by bootstrap or by the asymptotic variance of the difference in means. The framework predicts that ΔIV > 0 for all non-trivial cases, and that ΔIV is approximately equal to the full intrinsic value under recognition for institutional objects with null physical features (by Theorem 7.1).
Purpose. To measure the asymptotic approach rate of intrinsic value to the full monetary endowment as scarcity approaches a survival threshold, providing an empirical test of Theorem 6.7.
Design. Participants are presented with scenarios that vary the availability of a necessity good (typically water or food) across a range of scarcity levels from abundant to critically scarce. For each scenario, participants elicit their willingness-to-pay for an additional unit of the good under no-resale. The scarcity levels are chosen to span a range that includes conditions well above the survival threshold and conditions approaching the threshold, without ethically crossing into actual scarcity.
Estimation. The estimated function IV(zN) is plotted against zN across scarcity levels. The framework predicts that as zN approaches the threshold τ from above, IV(zN) approaches mg (the participant’s stated monetary endowment) at a rate controlled by the parameter α of Theorem 6.7. The parameter α can be estimated by fitting the model mg − IV(zN) = C (τ − zN)α to the data, with τ either fixed by the researcher or estimated jointly.
For replication and cross-study comparison, data from empirical work with the framework should be stored with a consistent schema. The minimal schema records: participant identifier, object identifier, source intensity ratings (zD, zU, zN, zE) either assigned by the researcher or elicited from the participant, elicited willingness-to-pay, elicitation method used, and any experimental condition identifiers. An extended schema adds participant demographics, enforcement conditions of the no-resale constraint, and context variables relevant to the specific study.
Arrow, K. J. (1951). Social Choice and Individual Values. New York: John Wiley & Sons. (Second edition, 1963.)
Becker, G. S. (1965). A theory of the allocation of time. The Economic Journal, 75(299), 493–517.
Becker, G. M., DeGroot, M. H., and Marschak, J. (1964). Measuring utility by a single-response sequential method. Behavioral Science, 9(3), 226–232.
Black, D. (1948). On the rationale of group decision-making. Journal of Political Economy, 56(1), 23–34.
Debreu, G. (1959). Theory of Value: An Axiomatic Analysis of Economic Equilibrium. New Haven: Yale University Press.
Dekel, E., Lipman, B. L., and Rustichini, A. (2001). Representing preferences with a unique subjective state space. Econometrica, 69(4), 891–934.
Gul, F., and Pesendorfer, W. (2001). Temptation and self-control. Econometrica, 69(6), 1403–1435.
Harsanyi, J. C. (1955). Cardinal welfare, individualistic ethics, and interpersonal comparisons of utility. Journal of Political Economy, 63(4), 309–321.
Innes, A. M. (1913). What is money? The Banking Law Journal, 30(5), 377–408.
Innes, A. M. (1914). The credit theory of money. The Banking Law Journal, 31(2), 151–168.
Kahneman, D., and Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–292.
Kant, I. (1785). Groundwork of the Metaphysics of Morals. Trans. M. Gregor (1997), Cambridge University Press.
Knapp, G. F. (1905). Staatliche Theorie des Geldes. Leipzig: Duncker & Humblot. English translation: The State Theory of Money (1924), London: Macmillan.
Krantz, D. H., Luce, R. D., Suppes, P., and Tversky, A. (1971). Foundations of Measurement, Volume I: Additive and Polynomial Representations. New York: Academic Press.
Lancaster, K. J. (1966). A new approach to consumer theory. Journal of Political Economy, 74(2), 132–157.
Mas-Colell, A., Whinston, M. D., and Green, J. R. (1995). Microeconomic Theory. New York: Oxford University Press.
Mitchell, R. C., and Carson, R. T. (1989). Using Surveys to Value Public Goods: The Contingent Valuation Method. Washington: Resources for the Future.
Moore, G. E. (1903). Principia Ethica. Cambridge: Cambridge University Press.
Savage, L. J. (1954). The Foundations of Statistics. New York: John Wiley & Sons.
Sen, A. (1985). Commodities and Capabilities. Amsterdam: North-Holland.
Smith, A. (1776). An Inquiry into the Nature and Causes of the Wealth of Nations. London: W. Strahan and T. Cadell.
Thaler, R. H. (1980). Toward a positive theory of consumer choice. Journal of Economic Behavior and Organization, 1(1), 39–60.
Vickrey, W. (1961). Counterspeculation, auctions, and competitive sealed tenders. The Journal of Finance, 16(1), 8–37.
Wray, L. R. (2012). Modern Money Theory: A Primer on Macroeconomics for Sovereign Monetary Systems. Basingstoke: Palgrave Macmillan.