Marginal utility is the change in utility associated with consuming an additional unit of a good or with a sufficiently small increase in its quantity, holding other relevant circumstances constant. In microeconomics, it connects a representation of preferences to decisions about how scarce resources are allocated. The familiar proposition of diminishing marginal utility states that successive additions to consumption provide progressively smaller increments of utility. A first glass of water when thirsty illustrates the idea, but the proposition is conditional rather than a universal psychological law.
Marginal utility helped reshape economic theory during the nineteenth-century marginal revolution and remains central to models of consumer choice, uncertainty, saving, and welfare. Its interpretation nevertheless differs across these applications. In ordinary consumer theory, utility generally ranks alternatives rather than measuring satisfaction on an independently meaningful scale, so marginal-utility ratios are more fundamental than marginal-utility levels. In models of risky choice, the curvature of a utility function can carry additional behavioral meaning. These distinctions determine what can legitimately be inferred from diminishing marginal utility, market prices, and comparisons between people.
Definition and mathematical representation
Suppose a consumer chooses a bundle of goods represented by quantities , and preferences are represented by a utility function . When the function is differentiable, the marginal utility of good is its partial derivative:
The partial derivative holds the quantities of other goods constant. It therefore describes a local change, not the entire adjustment that might follow a change in income or prices. Whether the consumer has already eaten, how much leisure is available, and the quality of the good may also matter; a model must either include these circumstances explicitly or treat them as fixed.
For indivisible goods, marginal utility is naturally expressed as a finite difference. If counts identical units and other consumption is fixed, the utility of the next unit is
A hypothetical sequence of total utilities of and for zero through four units gives successive marginal utilities of and . Total utility continues to rise even though marginal utility falls. A fifth unit could add nothing, producing satiation at that margin, or reduce utility if consumption is compulsory and excessive.
Marginal utility must be distinguished from average utility, . Average utility depends on the utility assigned to the entire quantity and on the chosen zero of the utility scale. Marginal utility concerns only the increment. Neither is the same as a good’s market price, its physical productivity, or its moral importance. A lifesaving medicine can have great importance without its market price directly measuring the satisfaction it produces.
Units also matter. If an amount previously measured in kilograms is instead measured in grams, marginal utility per unit changes numerically. Changing the numerical scale used to represent utility changes it again. A number such as “five utility units per apple” therefore has no freestanding empirical meaning unless the interpretation and normalization of the utility function have been specified.
The derivative definition also presupposes sufficient smoothness. Perfect complements, quantity discounts, indivisible purchases, and threshold benefits can produce kinks or discontinuities. In these cases, choice can still be analyzed using comparisons between feasible bundles, one-sided changes, or mathematical optimization methods that do not require ordinary derivatives. Marginal utility is a convenient analytical device, not a prerequisite for representing every economically meaningful preference.
Diminishing marginal utility and its conditions
Diminishing marginal utility means that an additional unit contributes less utility when more of the same good is already consumed, with other conditions held constant. For a twice-differentiable utility function, this is represented by
The statement concerns the slope of utility with respect to one quantity. It does not mean that total utility declines: total utility rises whenever marginal utility remains positive. Nor does it imply that every additional unit is unwanted. Positive but diminishing marginal utility is compatible with consuming substantial amounts when their opportunity cost is sufficiently low.
Several mechanisms can make declining increments plausible. Immediate physiological needs may become less pressing as consumption proceeds. Units of a durable good may be allocated first to the most valuable uses and later to less valuable ones. A household might use its first available room for essential living space and additional rooms for increasingly optional purposes. Such examples explain why a diminishing-increments assumption can be useful, but they do not establish a universal measure of satisfaction.
The assumption can fail locally. Two components may be useful only when combined, and a collection may become especially valuable as completion approaches. An activity may require a minimum commitment before yielding substantial benefits. These cases often concern interactions among goods or discontinuous benefits rather than a simple sequence of identical, independently useful units. Network effects can likewise increase a service’s usefulness as other people adopt it, although that is a change in the surrounding environment rather than a clean test of own-quantity marginal utility.
Diminishing marginal utility is different from diminishing marginal rates of substitution. The latter describes how willingness to exchange one good for another changes along a fixed preference ranking and is associated with convex preferences. A utility function may represent convex preferences without having diminishing marginal utility in every coordinate. Conversely, negative own second derivatives alone do not establish all the curvature properties needed for a well-behaved multigood optimization problem.
It must also be separated from diminishing returns in production. Diminishing marginal product concerns the additional physical output generated by another unit of an input while other inputs remain fixed. Marginal utility concerns the valuation of consumption alternatives. Both use marginal reasoning, but one is a technological relationship and the other belongs to preference representation.
Finally, comparisons across time require care. Enjoyment from a tenth serving consumed immediately is not equivalent to enjoyment from a serving consumed the following week. Recovery of appetite, learning, habit, and changes in quality alter the conditions being held constant. The assumption becomes informative only after the relevant good, time period, and background circumstances have been defined.
Ordinal utility and the limits of numerical interpretation
In standard consumer theory, utility is ordinarily ordinal: it represents the ranking of alternatives. If , bundle is preferred to bundle . The numerical difference between the two utility values need not measure the intensity of that preference. A consumer assigned utilities of and is not thereby said to obtain twice as much satisfaction from the first bundle.
Any strictly increasing transformation of an ordinal utility function represents the same preferences. Let , where is increasing. Where differentiable with a positive derivative,
Marginal-utility levels therefore depend on the representation chosen. More importantly, even diminishing marginal utility need not survive a transformation:
The first term can alter the sign of the second derivative without changing any ranking of certain consumption bundles.
For example, on the domain , both and represent a preference for more rather than less of a single good. The first has diminishing marginal utility; the second has constant marginal utility. The transformation connects them. Ordinary choices over certain quantities cannot distinguish these representations merely by observing that more is preferred.
Ratios of marginal utilities are more robust. For two goods with nonzero relevant derivatives,
This ratio determines the local trade-off along an indifference curve. The positive magnitude is the marginal rate of substitution of good for good : the amount of a consumer is locally willing to surrender for an additional unit of , holding utility constant.
The distinction explains why contemporary consumer theory can retain marginal analysis without treating utility as a measurable psychological substance. Budget choices can reveal trade-offs, while the absolute marginal utility assigned to a particular good remains representation-dependent. The shape of indifference curves, not the arbitrary numerical spacing of utility labels, carries the relevant information.
Cardinal restrictions arise in some other settings, especially expected-utility theory under risk. There the admissible transformations are narrower, so curvature can acquire behavioral significance. Confusing that framework with ordinal choice under certainty leads to unwarranted claims about measurable satisfaction or universal diminishing marginal utility.
Consumer optimization and the equimarginal principle
The classic application of marginal utility is the allocation of a limited budget. A consumer facing positive prices and income solves
When preferences are locally nonsatiated, the budget is exhausted. If the optimum is interior and the utility function is differentiable, the first-order conditions give
Thus marginal utility per unit of expenditure is equal across goods consumed in positive quantities. This is the equimarginal principle.
The reasoning is an argument about feasible improvement. If one good provides more marginal utility per dollar than another, shifting a small amount of expenditure toward it raises utility while preserving the budget. At an interior optimum, no such improvement remains. For two goods, the condition is equivalently
The consumer’s willingness to exchange the goods matches the rate of exchange available in the market.
Consider , prices and , and income . The marginal-utility ratio is . Equating it to the price ratio gives ; combining this with yields and . Each good receives half the budget. The numerical marginal utilities would change under an increasing transformation, but the chosen bundle and the equality of the relevant ratios would not.
The multiplier is a shadow price of the budget. If the indirect utility function is , the envelope theorem gives, under suitable regularity conditions,
It is often called the marginal utility of income. Its utility units depend on the representation, but its role in relating small expenditure changes to utility changes is mathematically precise within that representation.
Equality conditions require qualifications. If a good is not consumed, its marginal utility per dollar can be below that of goods purchased. Perfect substitutes often produce such corner solutions. Perfect complements can produce optima at kinks, where the ordinary derivative ratio is unavailable. For indivisible goods, the relevant comparison concerns affordable discrete changes rather than infinitesimal reallocations.
First-order conditions also do not by themselves guarantee a maximum. Appropriate curvature conditions or direct comparison of feasible bundles are needed. When preferences are convex and the feasible set is well behaved, tangency conditions become especially useful. The principle is therefore a conditional result of mathematical optimization, not a claim that every observed purchase literally involves conscious calculations of satisfaction per dollar.
Demand, willingness to pay, and consumer surplus
Marginal utility enters the theory of demand through constrained choice, not through a simple identification of utility with price. A consumer’s demand specifies the quantity selected at given prices and income. The equilibrium condition links the marginal utility of a good to the utility value of the expenditure needed to obtain it. Both sides can change when a price changes.
Consequently, diminishing own marginal utility alone does not prove that ordinary demand slopes downward. A price change alters relative prices and purchasing power simultaneously. The Slutsky equation separates the resulting substitution effect from the income effect. For a good’s own price, a standard differentiable form is
evaluated at the original utility level. Here is compensated demand, which holds utility constant.
Under standard assumptions, the compensated own-price effect is nonpositive. For a normal good, higher real purchasing power increases demand, so the income effect reinforces the substitution effect when its price falls. For an inferior good, the effects oppose one another. A Giffen good is an exceptional case in which the income effect is strong enough to make uncompensated demand rise with price. This possibility illustrates why a law about a utility derivative cannot substitute for a full account of budget-constrained choice.
A particularly transparent relationship appears with quasilinear utility:
where is a numeraire good with price one. If both goods are consumed at an interior optimum and costs , then
The derivative can be read as marginal willingness to pay in units of the numeraire. With , it generates a downward-sloping inverse demand curve. The interpretation depends on the quasilinear structure and the availability of the numeraire margin.
This case also clarifies consumer surplus. With , the net benefit from purchasing units at price is
Graphically, this is the area under marginal willingness to pay and above the price. In more general preferences, willingness to pay varies with income, and the familiar area under ordinary demand is not an exact utility-independent welfare measure.
Exact monetary welfare comparisons can instead use compensating or equivalent variation, derived from the expenditure needed to reach specified utility levels. This distinction matters in cost-benefit analysis: a demand curve records willingness and ability to pay, not a direct interpersonal scale of need or satisfaction.
Historical development and the marginal revolution
The origins of marginal utility lie in attempts to explain value through the significance of additional units rather than the usefulness of an entire class of goods. The water–diamond comparison, associated especially with Adam Smith’s The Wealth of Nations of 1776, contrasted water’s great usefulness with its ordinarily low exchange value and diamonds’ high exchange value with their limited necessity. Marginal reasoning distinguishes the value of having water at all from the value of another unit when water is abundant.
Several writers anticipated aspects of the later theory. Daniel Bernoulli’s 1738 discussion of risky monetary prospects used a logarithmic relation between utility and wealth. Jules Dupuit’s 1844 work on public works connected willingness to pay, additional consumption, and benefits from infrastructure. Hermann Heinrich Gossen’s 1854 book developed propositions now associated with diminishing marginal utility and the equalization of marginal benefits across uses. His work initially attracted limited attention.
The “marginal revolution” conventionally centers on William Stanley Jevons, Carl Menger, and Léon Walras. Jevons published The Theory of Political Economy in 1871, describing the “final degree of utility.” Menger’s Principles of Economics, also published in 1871, explained valuation through the importance of the least important satisfaction dependent on an available quantity. Walras’s Elements of Pure Economics, issued in parts beginning in 1874, incorporated marginal valuations into a mathematical system of general equilibrium.
These authors did not present identical theories. Jevons used mathematical and pleasure-based language; Menger emphasized individual valuation and causal explanation without adopting the same mathematical method; Walras analyzed simultaneous market relations. The later label “marginal utility” brings together related ideas without eliminating these methodological differences. Friedrich von Wieser’s German term Grenznutzen helped establish the terminology during the 1880s.
Alfred Marshall’s Principles of Economics, first published in 1890, integrated marginal utility with production costs and supply conditions. In that framework, demand and supply jointly determine prices. Marginal utility was not a complete explanation of price independent of scarcity, technology, and the cost of bringing additional goods to market.
During the late nineteenth and early twentieth centuries, consumer theory increasingly separated preference ordering from measurable pleasure. Francis Ysidro Edgeworth and Vilfredo Pareto contributed to the development of indifference-curve analysis. John Hicks and R. G. D. Allen’s 1934 papers, “A Reconsideration of the Theory of Value,” helped formulate demand theory around ordinal preferences and marginal rates of substitution. Paul Samuelson’s 1938 revealed-preference approach further shifted attention toward restrictions on observable choices.
This evolution changed the status of marginal utility. It remained a useful derivative within utility representations, but claims about measurable satisfaction were no longer necessary foundations for ordinary demand theory.
Risk, expected utility, and diminishing marginal utility of wealth
Marginal utility has a distinct interpretation in expected-utility theory, where individuals choose among uncertain prospects. In a simple monetary model, a prospect with wealth outcomes and probabilities is evaluated as
Unlike ordinary ordinal utility over certain bundles, a von Neumann–Morgenstern utility function representing the same preferences over lotteries is unique up to a positive affine transformation, , with . This narrower transformation class preserves the sign of curvature.
For an increasing, concave utility function, the marginal utility of wealth is positive but diminishing. Concavity implies
The individual therefore weakly prefers a certain amount equal to a gamble’s expected wealth to the gamble itself. This is the connection between diminishing marginal utility of wealth and risk aversion within expected-utility theory. Concavity is not merely an arbitrary feature of a chosen ordinal representation in this setting.
The certainty equivalent satisfies . The difference between expected wealth and the certainty equivalent is the risk premium. For a small, zero-mean monetary risk around wealth , a second-order approximation gives
where is the risk’s variance. The measure , independently developed in work by John Pratt and Kenneth Arrow in the 1960s, is absolute risk aversion.
This framework helps explain demand for insurance. A loss reduces wealth precisely in circumstances where an additional unit of wealth may have greater marginal utility. Insurance transfers resources toward those states. A risk-averse person can therefore value coverage above its expected payout, although actual contracts also involve administrative costs, exclusions, moral hazard, and adverse selection.
Several distinctions remain essential. Diminishing marginal utility of one consumption good does not by itself establish aversion to every kind of risk. The relevant utility function must evaluate the outcomes of the particular gamble, and background risks, other goods, and state-dependent needs may affect preferences. Likewise, absolute and relative risk aversion are different concepts: relative risk aversion is , and it need not move with wealth in the same way as absolute risk aversion.
Expected utility is also a substantive behavioral model, not an unrestricted description of all uncertain choice. Prospect theory allows reference-dependent valuation and nonlinear probability weighting. In that framework, observed risk attitudes cannot be attributed solely to diminishing marginal utility of final wealth.
Consumption over time, saving, and asset valuation
Intertemporal models use marginal utility to explain how consumption is distributed across dates. A household may maximize discounted utility,
where is consumption and is a discount factor. Future consumption can be valuable both because of its amount and because of the circumstances in which it occurs. Saving transfers purchasing power from the present to the future, while borrowing makes the reverse transfer subject to repayment.
If an interior choice allows one unit of current consumption to be exchanged for units next period with certainty, the optimality condition is
This consumption Euler equation equates the utility cost of forgoing consumption now with the discounted marginal utility of the additional future consumption it finances. It is an intertemporal version of the equimarginal principle.
With diminishing marginal utility, a household generally has an incentive to avoid large avoidable differences in consumption across dates. Consumption smoothing does not necessarily mean identical consumption every period: the interest rate, time preference, household composition, expected needs, and uncertainty all matter. A household can rationally choose a rising or falling consumption path even when its period utility function is strictly concave.
A common specification is constant-relative-risk-aversion utility,
with the logarithmic case at . Marginal utility is . Under certainty, the Euler equation implies
In this time-additive expected-utility specification, governs both relative risk aversion and the inverse of the elasticity of intertemporal substitution. More general preference models separate these properties.
Under uncertainty, the corresponding condition is
This links marginal utility to asset valuation. A payoff delivered when consumption is low can be especially valuable because marginal utility is then high. An asset’s expected return alone therefore does not determine its value; the relationship between its payoff and economically adverse states also matters.
Borrowing constraints can prevent equality from holding. A household may have high current marginal utility but be unable to obtain funds against future income. Moreover, precautionary saving requires more than concavity alone: in standard models, a positive third derivative, , is a central condition supporting a precautionary response to future income risk. The additional curvature condition distinguishes aversion to risk from the incentive to accumulate a buffer against it.
Welfare, distribution, and the value of public resources
Marginal utility plays a prominent role in welfare economics, especially in arguments about distribution. The intuitive claim is that an additional unit of consumption may matter more to someone with little consumption than to someone with much. If individuals share an increasing, concave utility function and their utility levels are interpersonally comparable on the specified scale, transferring resources from a richer person to a poorer person can increase the sum of utilities.
Those assumptions are substantive. Suppose a social objective is
where are welfare weights. With a fixed total resource and no other constraints, an interior optimum satisfies
for every individual. Weighted social marginal utilities are equalized. Equal consumption follows only under additional symmetry assumptions, such as identical utility functions and equal welfare weights.
Ordinary ordinal consumer theory does not itself supply interpersonal comparisons. Each individual’s utility representation can be independently rescaled without changing that person’s choices. Adding such utility numbers therefore requires an extra ethical or measurement framework. Diminishing marginal utility within a person does not, by itself, establish that a particular transfer between two different people raises social welfare.
This distinction matters for progressive taxation. Models of optimal taxation often assign greater social value to resources accruing to lower-income households, but tax design also depends on behavioral responses, information constraints, and revenue needs. Differences in ability to pay, consumption needs, and access to public services cannot be reduced to a single observed income measure. Formal conclusions depend on the chosen social objective and on how taxes affect labor supply and other decisions.
For public goods, marginal valuations enter another allocation rule. When individuals consume the same quantity of a nonrival public good, an efficient interior provision level satisfies a Samuelson-type condition,
where is a private numeraire and is the marginal resource cost. Individual willingness to trade private consumption for the public good is added because the same extra unit benefits multiple people. This differs from private-good allocation, where one person’s consumption generally removes that unit from another’s use.
Marginal utility also helps clarify the limits of market valuation. Willingness to pay reflects the marginal utility of the good relative to the marginal utility of money, and is constrained by resources. A high monetary bid need not identify the person with the greatest need or the largest welfare gain. Evaluations involving income inequality, essential services, or environmental changes must distinguish observed purchasing power from explicitly chosen social welfare weights.
Evidence, measurement, and behavioral qualifications
Marginal utility is usually inferred within a model rather than directly observed. Purchases reveal selected bundles at particular prices and budgets; they do not reveal a numerical sensation attached to each unit. The revealed-preference approach asks whether observed choices are consistent with utility maximization. If a consumer chooses one bundle when another is affordable, the observation imposes a restriction on the underlying preference ranking.
Demand data can identify local trade-offs under suitable assumptions about prices, income, available goods, and preference stability. They generally cannot identify a unique ordinal utility scale. Empirical demand systems therefore impose structure and normalizations. An estimated marginal-utility coefficient acquires meaning through the model, while welfare calculations require attention to the income effects and substitution patterns that the specification permits.
Discrete-choice models make this dependence particularly visible. A model may represent the utility of alternative as , with an observed component and an unobserved disturbance. Price and quality coefficients can yield estimates of willingness to pay through ratios. Their absolute magnitude depends on the normalization of the disturbance scale, so comparing raw utility coefficients across populations or datasets can be misleading if scale differences are ignored.
Surveys of reported well-being provide another kind of evidence. Studies often examine how life evaluations or reported affect vary with income and consumption. Such measures concern particular questions and response scales, not an automatic observation of the utility function used in demand theory. Adaptation, comparison groups, household composition, and differences in reporting complicate the relationship between income and reported well-being. A diminishing association on a survey scale does not by itself identify an individual’s entire marginal-utility schedule.
Experiments can investigate satiation and trade-offs more directly by varying quantities, timing, and choice conditions. Yet even controlled evidence requires careful interpretation. A participant’s willingness to accept another food item can change because of fullness, boredom, expectations about later consumption, or the opportunity to exchange the item. These mechanisms need not imply the same underlying preference representation.
Behavioral economics further emphasizes dependence on context and reference points. Loss aversion, endowment effects, and mental accounting can affect valuations in ways that a stable utility function over final quantities may not capture without additional variables. Habit formation and present bias similarly alter the timing and interpretation of marginal trade-offs.
Neural responses to rewards may supply evidence about valuation processes, but a measured neural signal is not automatically an economic utility unit or a basis for interpersonal welfare comparison. Across observational, experimental, and biological approaches, the central empirical task is to specify which choices the utility model is intended to explain and which transformations leave its testable predictions unchanged.